{ "generatedAt": "2026-09-15T00:00:00.000Z", "nodes": [ { "id": "ai-method-audit", "title": "AI method audit: each named route, its useful job and the harder limit", "shortTitle": "AI method audit", "type": "system", "status": "supported", "scope": [ "pdac", "pnet", "cholangiocarcinoma", "gallbladder", "hcc", "colorectal", "pan-cancer", "transfer-unproven" ], "summary": "The mission's thirty-two AI methods reduce to eight testable work packages; only image recognition and experiment selection are presently limited mainly by modelling.", "content": "Decision rule: AI earns a work package only when its input data exist, its target output changes a named human action, its current error can be measured and a prospective test can show benefit or useful failure. Every package must state the current human workflow, available training data, missing data, a model architecture no more complex than needed, an experimental validation loop, regulatory implications, expected economic value and whether AI is actually the limiting factor. The linked package pages use this template. Most routes fail before model choice: the missing item is a representative cohort, a trustworthy label, a treatment that can act on the output, trial-grade manufacture or a randomized comparison.\n\nThe mission names twenty applications and twelve model families: 20 + 12 = 32 method labels, not thirty-two separate investments. They collapse into eight work packages with explicit rejection rules. The useful unit is the clinical or experimental decision, not the fashionable model name. An architecture is compared with a simple fixed rule, and a model that cannot abstain on missing or unfamiliar inputs does not pass. Clinical models need locked versions, hospital-held-out tests, subgroup error bounds and prospective workflow trials. Models used only to choose the next laboratory experiment need blinded wet-lab repeat and a measured saving in time, wells or failed compounds. Neither route may claim patient benefit from technical accuracy alone.", "contentSections": [ "1. Existing-image decisions. Radiology and AI radiology map to opportunistic CT triage. Pathology and digital pathology map to chemotherapy selection from routine H&E. Patient stratification, recurrence prediction, treatment selection and multimodal clinical models map to the residual-disease state. Inputs are scans, slides or dated clinical measurements; outputs are an inspectable alert, a treatment preference or a calibrated recurrence state. Training data exist but are retrospective and enriched. Missing data include true false-positive cost, counterfactual treatment response, scanner and stain drift, low-shedding disease and external health systems. AI helps recognition and time-series fusion. Prospective utility, an effective action and safe workflow are harder limits.", "2. Molecular state and cause. Biomarker discovery, multi-omics, resistance prediction, causal inference, causal ML, mechanistic ML and genomic foundation models map to RAS-resistance routing. The input must be serial human plasma, tissue, drug exposure and outcome, not an undated omics table. The output is a fixed resistance class and confirmatory assay, not an invented treatment. Current error is unknown because uniform progression specimens and causal labels do not exist. Small interpretable multi-view or Bayesian state models must beat mutation-only rules. AI can join measurements; specimens, perturbation, pharmacology and available partner drugs are the limiting factors.", "3. Molecule and protein design. Protein modelling, protein language models, molecule generation and foundation models for biology map to the binder-design package. Drug repurposing uses the same gate but starts from approved-medicine exposure and safety. The output is a compound or binder for testing, never a clinical recommendation. AlphaFold 3 and BoltzGen support interaction or binder generation, while PDAC target validity, tissue delivery, healthy-organ selectivity, manufacture and exposure remain harder. Training leakage, target-dependent collapse and missing negative assay data make benchmark scores unreliable. AI can reduce design cycles only after the human target and assay are real.", "4. Experiment choice. Combination therapy discovery, active learning, Bayesian optimisation and reinforcement learning for experimental design map to the organoid active-learning package. Autonomous experimentation and robotic laboratories are execution layers around the same loop. Inputs are dose-response, image, genotype, cell state, time and prior failures; output is the next small batch of perturbations with uncertainty. The current human workflow uses broad fixed panels. The test is fewer than 25% of wells with the same useful hits and a result before cycle two. AI may be limiting for search efficiency. Culture success, metastatic fidelity, available drugs and the fact that a dish is not a patient remain the larger clinical limits.", "5. Virtual cells, not virtual patients. Virtual cells map to a closed perturbation loop that predicts measured cell-state changes and chooses the next live experiment. Digital twins are rejected for individual cure prediction because the required intervention data, whole-body state and counterfactual outcome do not exist. A transcript prediction must beat simple baselines in held-out patients and laboratories and then improve on the second experimental loop. The useful economic result is cheaper rejection of combinations. A simulated survival curve is not evidence of treatment effect.", "6. Trials. Clinical trial optimisation maps to eligibility extraction, site forecasting, data checks and missing-sample alerts. Adaptive trials use fixed, prospectively simulated rules; FDA final guidance requires the design, conduct and reporting to support reliable safety and effectiveness evidence. Synthetic controls may help planning or a single-arm safety comparison, but are rejected as the main proof of survival when treatment selection, supportive care and calendar time differ. Current error must be measured inside the target trials. AI can reduce coordinator time; the intervention, endpoint, randomisation and recruitment network remain the main limits.", "7. Literature and agents. Literature synthesis and agentic scientific systems can retrieve sources, compare claims and expose missing links. Their output is a candidate evidence record for expert checking, never evidence by itself. Training overlap, absent negative results, citation errors and shifting sources prevent a stable clinical error rate. Expected value is researcher time saved and broader adversarial search. The limiting factors are source access, complete reporting and expert review. No programme decision releases capital from an unverified generated summary.", "8. Regulation and economics. A model that detects, predicts recurrence or recommends treatment may become medical-device software; FDA's January 2025 full-life-cycle document is draft and non-binding, but it makes the evidence burden clear: intended use, data provenance, subgroup performance, version control, risk controls and monitoring must be planned. Models used to support drug decisions inherit the trial and product evidence burden. Economic value is counted as avoided scans, coordinator hours, wells, failed compounds or late relapse—not model accuracy. Each package has a cost gate and shuts down when it cannot beat a simple rule or cannot change a safe action.", "9. Gallbladder lesion models show the transport trap. GBCseeker reported external sensitivity 93.55% and specificity 77.78% in 98 people, 62 of whom had cancer; GAIA-MIL reported pooled external AUC 0.778 in 122 retrospectively assembled patients and 69.4% sensitivity below 10 mm. At 2% prevalence, unchanged GBCseeker performance would yield about 19 true positives and 218 false positives per 1,000, so PPV falls to 7.9%. At 0.2%, it would yield about 119 false positives per true positive. These models can be tested silently in consecutive lesion pathways, but they do not yet justify screening or preventive surgery. The limiting work is representative prospective sampling, fixed thresholds, country calibration, operation-harm measurement and an action trial.", "10. Endoscopic-ultrasound AI has crossed a reader test, not a care test. In a randomized crossover using selected image sets, novice solid-lesion sensitivity rose by 11.9 percentage points and accuracy by 7.7 points. Eight readers and 118 patient image sets cannot estimate what happens during live probe movement, whether extra findings are true, or whether management and outcome improve. The next unit is a prospective procedure with expert adjudication, false work-up, time, complications and action—not another frame benchmark." ], "sourceIds": [ "cao-2023-panda-ct-ai", "ygberg-2025-pancanai-validation", "hendifar-2026-ai-chemo-pathology", "botta-2024-tumor-informed-ctdna", "beutel-2021-pdac-organoids", "boyd-2025-closed-loop-single-cell", "abramson-2024-alphafold3", "stark-2025-boltzgen", "fda-2025-ai-device-lifecycle-draft", "fda-2023-externally-controlled-trials-draft", "fda-2019-adaptive-trial-guidance", "yang-2025-gbcseeker", "gupta-2026-gaia-mil", "clinicaltrials-nct06531408-pigbc", "tanaka-2026-ai-eus-reader" ], "links": [ { "target": "ai-programme", "relation": "audits" }, { "target": "technology-opportunistic-ct-ai-pancreas", "relation": "maps-to" }, { "target": "technology-ai-pathology-chemo-selection", "relation": "maps-to" }, { "target": "technology-ai-multimodal-mrd-state", "relation": "maps-to" }, { "target": "technology-ai-resistance-routing", "relation": "maps-to" }, { "target": "technology-ai-protein-design", "relation": "maps-to" }, { "target": "technology-ai-organoid-active-learning", "relation": "maps-to" }, { "target": "technology-ai-virtual-cell-loop", "relation": "maps-to" }, { "target": "technology-ai-trial-operations", "relation": "maps-to" }, { "target": "unknown-ctdna-actionability", "relation": "limited-by" }, { "target": "experiments", "relation": "tested-by" }, { "target": "risks", "relation": "bounded-by" }, { "target": "claim-gallbladder-diagnostic-models-existing-lesions-only", "relation": "audits" }, { "target": "trial-gallbladder-prevention-detection-action-map", "relation": "maps-to" } ], "evidence": { "humanProspective": "moderate — selected inputs only", "humanRetrospective": "strong — imaging, pathology and residual-disease inputs", "randomised": "none — no cited PDAC AI utility trial", "independentReplication": "moderate — uneven across methods", "sampleSize": "Twenty applications plus twelve model families, mapped to eight work packages with explicit rejection rules", "effect": "Method and investment audit; no pooled AI effect and no claim of mortality benefit", "limits": [ "Several frontier model sources are early or outside PDAC", "No cited PDAC AI route has randomized clinical utility evidence", "Economic values are programme estimates", "FDA AI lifecycle guidance cited here remains draft and non-binding", "Method labels overlap and do not represent independent interventions" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "The useful routes attach to observed human imaging, pathology, recurrence or resistance signals." }, "animalEvidence": { "rating": "not-applicable", "reason": "This method audit does not use animal performance as proof of AI utility." }, "inVitroEvidence": { "rating": "moderate", "reason": "Organoid, binder and closed-loop packages have laboratory evidence, mostly outside direct PDAC utility." }, "reproducibility": { "rating": "weak", "reason": "Some technical results repeat, but no complete PDAC decision package has randomized utility proof." }, "commercialConflicts": { "rating": "moderate", "reason": "Several models, assays and datasets are company-controlled or proprietary." }, "recency": { "rating": "current", "reason": "The map includes clinical, model and regulatory evidence checked in September 2026." } } }, "layout": { "x": 137, "y": 126 }, "updated": "2026-09-14" }, { "id": "ai-mission-work-packages", "title": "AI mission map: five useful work packages and five places AI is not the bottleneck", "shortTitle": "AI mission map", "type": "system", "status": "supported", "scope": [ "pdac", "pnet", "hepatobiliary", "colorectal", "pan-cancer", "transfer-unproven" ], "summary": "Fund AI where it closes a measured loop: CT triage, pathology choice, MRD time series, organoid experiment selection and resistance routing.", "content": "Ranked by expected decision value. 1. Opportunistic CT triage: high near-term reach because images already exist; requires prospective cross-hospital false-positive and lead-time measurement. 2. Pathology chemotherapy choice: near-universal H&E input and a real choice; requires a biomarker-stratified randomized trial. 3. Multimodal MRD state: could move recurrence earlier; assay shedding and an active treatment are harder than modelling. 4. Organoid active learning: can cut wells and time; culture success and fidelity dominate. 5. RAS resistance routing: can join serial plasma, tissue and drug data; the uniform specimens do not yet exist. Second tier: protein and binder design and closed perturbation loops can speed tool creation and reject mechanisms, but begin only after human targets and assays are valid. Trial automation can reduce delay but not replace controls. Low-value claims to reject now: a digital twin that predicts an individual's cure without intervention data; a general literature agent treated as evidence; a foundation model trained mostly outside PDAC and deployed without held-out institutions; a synthetic control used to claim survival; an autonomous lab optimizing a proxy that has not predicted human outcome. Economics: the £5m core can fund resistance routing plus organoid and data standards; the £25m observatory can lock CT, pathology and MRD models; the £100m programme should spend most capital on prospective intervention tests, not larger models.", "sourceIds": [ "cao-2023-panda-ct-ai", "hendifar-2026-ai-chemo-pathology", "botta-2024-tumor-informed-ctdna", "beutel-2021-pdac-organoids", "oreilly-2026-daraxonrasib", "stark-2025-boltzgen", "boyd-2025-closed-loop-single-cell" ], "links": [ { "target": "ai-programme", "relation": "part-of" }, { "target": "technology-opportunistic-ct-ai-pancreas", "relation": "ranks" }, { "target": "technology-ai-pathology-chemo-selection", "relation": "ranks" }, { "target": "technology-ai-multimodal-mrd-state", "relation": "ranks" }, { "target": "technology-ai-organoid-active-learning", "relation": "ranks" }, { "target": "technology-ai-resistance-routing", "relation": "ranks" }, { "target": "technology-ai-protein-design", "relation": "ranks" }, { "target": "technology-ai-virtual-cell-loop", "relation": "ranks" }, { "target": "technology-ai-trial-operations", "relation": "ranks" }, { "target": "ai-method-audit", "relation": "audited-by" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "strong", "randomised": "weak", "independentReplication": "moderate", "sampleSize": "Eight linked work packages or technology assessments", "effect": "Programme ranking; effect gates stated in each record", "limits": [ "Ranking contains judgment", "Most AI evidence is retrospective", "No cited PDAC AI tool has randomized utility proof", "Training data and missingness are major constraints", "Economics are planning estimates" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "Each ranked package begins with an observed human or experimental decision problem." }, "animalEvidence": { "rating": "not-applicable", "reason": "Animal evidence does not validate the ranking of AI work packages." }, "inVitroEvidence": { "rating": "moderate", "reason": "Some experiment-selection and molecule-design routes have laboratory evidence, with uneven PDAC relevance." }, "reproducibility": { "rating": "weak", "reason": "The component technical studies vary and none proves randomized PDAC utility." }, "commercialConflicts": { "rating": "moderate", "reason": "The ranking includes proprietary pathology, assay and model work alongside academic evidence." }, "recency": { "rating": "current", "reason": "The map includes current clinical and model evidence through September 2026." } } }, "layout": { "x": 139, "y": 124 }, "updated": "2026-09-14" }, { "id": "ai-programme", "title": "AI research programme", "shortTitle": "AI", "type": "technology", "status": "hypothesis", "scope": [ "pdac", "pnet", "cholangiocarcinoma", "gallbladder", "hcc", "colorectal", "pan-cancer" ], "summary": "The strongest AI routes use data already collected and control a real decision; each still needs a prospective experiment and an abstain rule.", "content": "AI matters when it closes a measured decision loop; it does not replace missing specimens, valid assays, effective drugs or randomized evidence. Priority 1, opportunistic CT triage: analyse abdominal scans already obtained for other reasons and flag subtle pancreatic change for expert review. Reach is high and no extra radiation is created, but performance must be measured prospectively across hospitals, scanners, time before diagnosis and benign pancreatic disease. Priority 2, computational pathology for chemotherapy choice: routine stained tissue slides are cheap and nearly universal. Valar's locked external cohort supports a test. PANCprAId now adds a stronger retrospective result: training in 231 resected patients followed by analysis of 313 archived slides from the randomized PRODIGE-24 trial produced treatment-interaction signals for disease-free and cancer-specific survival. The trial did not assign treatment using the model, so only a model-stratified randomized trial can show that switching treatment improves outcome. Priority 3, a multimodal residual-disease state: join tumour-informed DNA, fragment patterns, CA19-9, imaging, treatment exposure and immune state over time. The model must abstain when shedding or inputs are insufficient and can enter care only with a safe treatment action. Priority 4, organoid active learning: choose the next wells and combinations to reduce cost and return a result before cycle two; culture success and human fidelity are more limiting than model size. Priority 5, RAS-resistance routing: join serial plasma, progression tissue and perturbation data to assign a reproduced escape class. The required uniform longitudinal dataset does not yet exist, so the first task is data creation. Second tier: protein and binder design can accelerate tool molecules after a human target and assay are validated; closed perturbation models can select the next laboratory experiment; trial software can detect missing samples, protocol drift and slow sites. Explicitly reject five claims now: a digital twin that predicts individual cure without intervention data; a general literature model treated as evidence; a foundation model trained mainly outside PDAC and deployed without held-out hospitals; a synthetic control used to claim survival; and an autonomous laboratory optimizing a proxy that has never predicted human outcome. Spend logic: the £5m core funds data rules and resistance or organoid pilots; the £25m observatory locks CT, pathology and residual-disease models; the £100m programme spends most money on prospective intervention tests, not a larger foundation model. Every work package specifies input, output, intended user, error cost, missingness rule, external validation, prospective test, clinical action and shutdown condition.", "contentSections": [ "A 2026 crossover reader study supplies a bounded prospective AI result. Eight endosonographers read selected image sets from 118 patients with and without an AI overlay. Novice solid-lesion sensitivity rose from 76.8% to 88.7% and accuracy from 78.7% to 86.4%. This is a reader-assistance result, not a clinical utility trial: it did not test live procedures, management, false work-up, complications, stage or survival.", "PANCprAId is an unusually useful bridge between technical validation and an action test. It was trained on 231 retrospective cases, then applied to 313 archived slides from a randomized adjuvant-treatment trial. The reported interaction P values were 0.003 for disease-free survival and 0.001 for cancer-specific survival. This supports a forward test in which the model is locked before enrolment and treatment is assigned from its prediction. It does not yet support clinical use: the original trial did not randomize model-guided care, slide availability can select patients, and retrospective threshold choices can exaggerate an interaction." ], "sourceIds": [ "cao-2023-panda-ct-ai", "ygberg-2025-pancanai-validation", "hendifar-2026-ai-chemo-pathology", "beaufils-2026-pancpraid", "tanaka-2026-ai-eus-reader", "botta-2024-tumor-informed-ctdna", "kong-2022-pdac-organoid-drug-screen", "boyd-2025-closed-loop-single-cell", "stark-2025-boltzgen" ], "links": [ { "target": "detection", "relation": "supports" }, { "target": "treatment", "relation": "supports" }, { "target": "experiments", "relation": "supports" }, { "target": "people", "relation": "depends-on" }, { "target": "technology-opportunistic-ct-ai-pancreas", "relation": "contains" }, { "target": "technology-ai-pathology-chemo-selection", "relation": "contains" }, { "target": "company-valar-labs", "relation": "contains" }, { "target": "ai-mission-work-packages", "relation": "contains" }, { "target": "ai-method-audit", "relation": "contains" } ], "evidence": { "humanProspective": "moderate for reader assistance; weak for clinical utility", "humanRetrospective": "strong — includes a model applied to archived material from a randomized trial", "randomised": "weak — one reader-crossover study and one retrospective randomized-trial analysis; no randomized patient-utility result", "independentReplication": "moderate", "sampleSize": "Eight defined work packages grounded in imaging, pathology, residual-disease, organoid, molecule-design, laboratory and trial studies", "effect": "Programme ranking; no AI application has yet shown PDAC mortality benefit", "limits": [ "Most evidence is retrospective", "PANCprAId did not assign treatment prospectively", "Training populations, slides and scanners differ", "Missing treatment actions constrain MRD value", "Organoid fidelity and turnaround limit active learning", "Protein design does not validate targets", "Prospective utility trials are absent" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "The priority routes attach models to observed human imaging, pathology, recurrence or resistance signals." }, "animalEvidence": { "rating": "not-applicable", "reason": "Animal model performance is not used as proof that an AI decision improves care." }, "inVitroEvidence": { "rating": "moderate", "reason": "Organoid, binder and perturbation packages have laboratory support without clinical utility proof." }, "reproducibility": { "rating": "weak", "reason": "Technical results repeat unevenly and no cited PDAC AI route has randomized utility evidence." }, "commercialConflicts": { "rating": "moderate", "reason": "Several pathology, assay, model and dataset components are proprietary or company-controlled." }, "recency": { "rating": "current", "reason": "The ranking uses AI and clinical evidence checked through September 2026." } } }, "layout": { "x": 94, "y": 95 }, "updated": "2026-09-15" }, { "id": "biomarker-ca19-9-longitudinal-anchor", "title": "CA19-9 rises before diagnosis in some pancreatic cancers, but needs a risk gate and companion markers", "shortTitle": "CA19-9 trajectory", "type": "biomarker", "status": "supported", "scope": [ "pancreatic-all" ], "summary": "In blinded PLCO samples, CA19-9 reached 60% sensitivity at 99% specificity within six months of diagnosis and 50% for early-stage cases; adding LRG1 and TIMP1 recovered 13.2% of CA19-9-negative cases.", "content": "CA19-9 is not merely a late marker. In stored PLCO serum from 175 people later diagnosed with pancreatic cancer and 875 controls, levels began rising about two years before diagnosis. Within six months, sensitivity was 60% at 99% specificity for all cases and 50% for early-stage cases. Against chronic pancreatitis and benign cyst controls sampled at diagnosis, sensitivity at the same specificity fell to 46% and 30%. Adding LRG1 and TIMP1 recovered 13.2 percentage points among cases below the CA19-9 cutoff within one year. Base rate determines use. In 10,000 people with the observed 0.62% three-year risk after new-onset diabetes, a hypothetical test with 50% sensitivity and 99% specificity would produce about 31 true positives and 99 false positives: positive predictive value 31/(31+99)=23.8%, if performance transported perfectly. At general-population annual incidence near 13.9 per 100,000, the same test would yield about 7 true positives and 999 false positives: positive predictive value about 0.7%. The credible design is longitudinal change after a risk trigger, followed by focused imaging. Missing evidence: a locked threshold in real-time care, Lewis-antigen non-secretors handled explicitly, benign obstruction adjudicated, stage and actionability measured, and clinical utility shown.", "sourceIds": [ "fahrmann-2021-ca19-9-trajectory", "chari-2026-prospective-new-onset-diabetes", "seer-pancreas-statfacts" ], "links": [ { "target": "claim-new-onset-diabetes-enriches-risk", "relation": "depends-on" }, { "target": "claim-screening-base-rate-maths", "relation": "supports" }, { "target": "detection", "relation": "part-of" }, { "target": "experiments", "relation": "tests" } ], "evidence": { "humanProspective": "none", "humanRetrospective": "strong", "randomised": "none", "independentReplication": "moderate", "sampleSize": "175 future cases and 875 matched PLCO controls; independent diagnosis-time sets included 129 resectable cases and 275 controls", "effect": "Within six months: 60% sensitivity at 99% specificity overall and 50% for early-stage disease", "limits": [ "Stored-sample nested case-control design", "Not deployed screening", "Pancreatic histologies not fully separated", "Performance falls against pancreatic disease controls", "Some people cannot express CA19-9", "Obstruction can raise CA19-9", "Illustrative predictive-value maths assumes transport" ] }, "layout": { "x": 9, "y": 114 }, "updated": "2026-09-14" }, { "id": "bottlenecks", "title": "Ranked bottlenecks", "shortTitle": "Bottlenecks", "type": "bottleneck", "status": "supported", "scope": [ "pdac", "pnet", "cholangiocarcinoma", "gallbladder", "hcc", "colorectal" ], "summary": "The current evidence ranks elimination of occult systemic PDAC first, ahead of detection alone. The ranking remains provisional and shows what could change it.", "content": "Rank 1: safely eliminate occult and established systemic PDAC. Even among selected resected patients receiving mFOLFIRINOX, five-year disease-free survival was 26.1%, so 73.9% had recurrence or death. Durable common-RAS control is the leading current route: daraxonrasib raised median survival from 6.6 to 13.2 months after prior therapy, but median progression at 7.3 months shows resistance remains. Rank 2: detect biologically dangerous disease while it is still actionable, in a population enriched enough to avoid large screening harm. Only 15.4% of pancreatic cancers in SEER are localised at diagnosis. Moving every case to today's localised five-year relative survival would raise relative survival from 13.7% to 43.6%, a 29.9-point or 3.18-fold gain, but would still leave a 56.4-point deficit from 100% five-year relative survival. Rank 3: build a residual-disease measurement and action loop. Postoperative ctDNA predicts recurrence, but PDAC has no randomized proof that acting on it improves outcome. Rank 4: identify and remove the patient-specific immune block. Unselected checkpoint treatment has near-zero activity, while vaccines can produce T cells without yet proving recurrence reduction. Rank 5: prevent disease beyond the inherited minority. Known high-risk genes covered 5.5% of cases in one large study, and modifiable exposures cannot explain or prevent most PDAC. These ranks overlap: common-RAS durability is the strongest current route to systemic eradication, while better systemic treatment raises the value of detection. The ranking changes only when a contrary result changes patient reach or the lethal step—for example, lead-time-corrected evidence that most screen-found stage I PDAC is cured by current treatment, a large removable cause, a safe precursor rule, or a durable first-line systemic regimen. The research frontier page ranks 15 candidate routes underneath these five gaps; the programme pages fund the highest-information tests rather than every field.", "sourceIds": [ "conroy-2022-prodige24-five-year", "seer-pancreas-statfacts", "oreilly-2026-daraxonrasib", "botta-2024-tumor-informed-ctdna", "tie-2022-dynamic-colon", "oreilly-2019-durvalumab-tremelimumab", "hu-2018-germline-risk" ], "links": [ { "target": "unknowns", "relation": "depends-on" }, { "target": "experiments", "relation": "depends-on" }, { "target": "programme", "relation": "ranks" }, { "target": "ranking-provisional-pdac-bottlenecks", "relation": "ranks" }, { "target": "research-frontier-ranking", "relation": "contains" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "strong", "randomised": "strong", "independentReplication": "strong", "sampleSize": "Population registry, randomized systemic and adjuvant trials, residual-disease cohorts and surveillance evidence", "effect": "Five ranked missing capabilities; detection-only upper bound 43.6% five-year relative survival and resected five-year DFS 26.1%", "limits": [ "Priority ranking is judgement", "SEER pancreatic data mix histologies", "Detection calculation is an unrealistic upper bound", "Capabilities interact", "Costs and technical-success probabilities remain estimates" ] }, "layout": { "x": 50, "y": 65 }, "updated": "2026-09-15" }, { "id": "cholangiocarcinoma", "title": "Cholangiocarcinoma", "shortTitle": "Cholangiocarcinoma", "type": "cancer", "status": "supported", "scope": [ "cholangiocarcinoma" ], "summary": "Intrahepatic, perihilar and distal bile-duct cancers retain separate risk, molecular and treatment evidence where site changes the conclusion.", "content": "This branch follows chronic biliary injury and site-linked risk through invasive disease, molecular subgroups, resection, recurrence and advanced treatment. Results from intrahepatic FGFR2- or IDH1-altered disease are not silently applied to extrahepatic disease. PSC surveillance in Europe and Australia, liver-fluke prevention and screening in Thailand, and postoperative recurrence after resection remain separate evidence questions because their base rates, detectable states and interventions differ. The action map now follows each route from exposure or risk state through test, work-up, treatment, harm and survival.", "contentSections": [ "The prevention route is causal but incompletely measured. Chronic Opisthorchis viverrini and Clonorchis sinensis infection causes cholangiocarcinoma. A Thai One Health programme reported infection falling from 6.0% to 0.3%, but without a randomized comparator or cancer endpoint. The live eCHEC trial randomizes 480 people but ends at 12-month behaviour and stool infection. These studies can show exposure control; long-term linkage must show whether cancer incidence and death fall.", "The two early-action gaps remain distinct. In PSC, 23.8% of 512 people developed severe or progressive duct change under annual MRI/MRCP, with roughly nine non-malignant findings per malignancy and 13-month median survival after cancer. After resection, ctDNA repeatedly predicts recurrence and can lead imaging by months, but no biliary trial has shown that changing treatment from the result improves survival.", "PROOF 301 shows that precision treatment can fail at the network layer before the drug question is answered. It screened 1,127 people at 120 sites over 40 months for FGFR2-rearranged disease but randomized only 48 of about 300 planned, 4.3% of those screened, and stopped early. A molecularly narrow phase 3 therefore needs measured testing coverage, fast referral and enough connected centres before another comparison is launched." ], "sourceIds": [ "jusakul-2017-cholangiocarcinoma", "iarc-2012-liver-flukes", "charoensuk-2024-one-health-opisthorchis", "clinicaltrials-nct05321992-echec", "villard-2023-psc-surveillance", "tan-2024-psc-mrcp-surveillance", "khuntikeo-2020-thailand-cca-screening", "hu-2020-icc-resection-recurrence", "yu-2025-btc-ctdna-recurrence", "yoo-2025-extrahepatic-cca-ctdna", "abou-alfa-2026-proof301", "fda-2026-bizengri-cholangiocarcinoma" ], "links": [ { "target": "hepatobiliary", "relation": "part-of" }, { "target": "model-cholangiocarcinoma-causal-chain", "relation": "contains" }, { "target": "claim-cca-surveillance-depends-on-risk-population", "relation": "contains" }, { "target": "claim-icc-resection-often-leaves-recurrence-risk", "relation": "contains" }, { "target": "disease-system", "relation": "part-of" }, { "target": "detection", "relation": "part-of" }, { "target": "treatment", "relation": "part-of" }, { "target": "company-jazz-pharmaceuticals", "relation": "contains" }, { "target": "company-relay-therapeutics", "relation": "contains" }, { "target": "company-elevar-therapeutics", "relation": "contains" }, { "target": "claim-biliary-immunochemotherapy-tail", "relation": "contains" }, { "target": "company-astrazeneca", "relation": "contains" }, { "target": "trial-cca-action-map", "relation": "contains" }, { "target": "claim-cca-psc-surveillance-workload", "relation": "contains" }, { "target": "claim-fluke-control-prevents-exposure-not-yet-cancer", "relation": "contains" }, { "target": "claim-cca-ctdna-prognostic-not-action-guiding", "relation": "contains" }, { "target": "unknown-cca-postoperative-mrd-action", "relation": "contains" }, { "target": "hypothesis-cca-route-specific-action-platform", "relation": "contains" }, { "target": "trial-2026-randomised-result-stress-test", "relation": "contains" }, { "target": "regulatory-bizengri-nrg1-cholangiocarcinoma", "relation": "contains" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "strong", "randomised": "strong for selected advanced treatments; none for surveillance mortality", "independentReplication": "moderate", "sampleSize": "PSC surveillance cohorts, endemic-area screening, 1,221 curative-intent resections and molecularly selected trials", "effect": "Anatomical site and driver create different routes; surgery still leaves high recurrence and surveillance utility is unresolved", "limits": [ "Intrahepatic, perihilar and distal disease are often pooled", "PSC and liver-fluke results do not share a base rate", "No surveillance route has randomized mortality evidence", "Residual-disease action remains untested" ] }, "layout": { "x": 42, "y": 6 }, "updated": "2026-09-14" }, { "id": "claim-2026-blood-tests-improve-signal-not-screening-utility", "title": "New blood tests improve signal detection but still fail the screening-utility gate", "shortTitle": "2026 blood-test gate", "type": "claim", "status": "supported", "scope": [ "pdac", "pancreatic-all" ], "summary": "The strongest new prospective result found six of eight pancreatic cancers, but 40 false positives and no outcome test prevent it from becoming a screening programme.", "content": "Three 2026 studies improve the evidence without closing the decision. The Tianjin cell-free-DNA model is the most important because it was tested prospectively in 1,926 people with diabetes or obesity for 12 to 24 months. It found 6 of 8 PDAC cases, including 3 of 3 lesions labelled stage 0, but returned 40 false positives. That is 75% sensitivity, 97.9% specificity and 13.0% positive predictive value: for every true positive, 40 / 6 = 6.7 people without PDAC were sent into a positive-result path. Eight cancers are too few to estimate stage-specific sensitivity reliably, and the study did not test whether follow-up operations, cancer mortality or all-cause harm improved. A transport calculation shows why 97.9% specificity is not enough for ordinary-risk screening. At 13.9 PDAC cases per 100,000 people per year, holding 75% sensitivity and 97.9% specificity constant would produce about 10.4 true positives and 2,100 false positives, a positive predictive value near 0.49%. This is a workload illustration, not a forecast: performance can change when disease spectrum and population change. The PCDC blinded bakeoff is a better comparison of eight panels on the same 140 cancer cases and 140 controls; CA19-9 plus FUT2/3 genotype reached AUC 0.963, but cases spanned stages I to IV and controls were mainly healthy. The authors explicitly expect weaker performance against early asymptomatic disease and benign pancreatic conditions and have started a prospective biorepository. A separate 649-person ITIH3/CEACAM1/CA19-9 study reached AUC 0.917 in blinded validation, but at about 95% sensitivity specificity was 53.3%; nearly half the controls were positive at that operating point. Decision: these tests deserve locked, independent comparison inside an existing risk-enriched workflow. They do not justify general-population screening, a mortality claim or a new programme that duplicates EDI and NCT06271291. Promotion requires consecutive intended-use patients, complete negative follow-up, enough stage-I aggressive cancers for narrow intervals, a fixed work-up, counted procedures and complications, and randomised evidence that acting on the result improves patient outcomes.", "sourceIds": [ "wang-2026-cfdna-prospective-detection", "oberg-2026-pcdc-biomarker-bakeoff", "wang-2026-itih3-ceacam1-panel", "seer-pancreas-statfacts" ], "links": [ { "target": "detection", "relation": "extends" }, { "target": "detection-modality-ranking", "relation": "constrains" }, { "target": "claim-case-control-blood-tests-overstate-screening-readiness", "relation": "replicates" }, { "target": "trial-new-onset-diabetes-workup-map", "relation": "depends-on" }, { "target": "claim-screening-base-rate-maths", "relation": "replicates" } ], "evidence": { "humanProspective": "moderate — one 1,926-person cohort but only eight PDAC cases", "humanRetrospective": "strong — two blinded selected-case validation sets", "randomised": "none", "independentReplication": "weak — assays, features and intended populations differ", "sampleSize": "Prospective n=1,926 with 8 PDAC cases; PCDC n=280; ITIH3/CEACAM1 study n=649", "effect": "Prospective test: 6 true positives, 2 false negatives and 40 false positives; sensitivity 75%, specificity 97.9%, PPV 13.0%", "limits": [ "Only eight prospective PDAC cases", "No randomised clinical-utility or mortality result", "The two larger assay studies begin with selected known cases", "General-population calculation assumes performance transports unchanged", "Stage-0 category includes high-grade precursor lesions rather than only invasive cancer", "One prospective assay has company-affiliated authors" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "Tumour and host changes can alter cell-free DNA and blood proteins before imaging, but low-shedding lesions and non-cancer inflammation overlap." }, "animalEvidence": { "rating": "not-applicable", "reason": "Only intended-use human screening cohorts can establish false-positive burden, lead time and utility." }, "inVitroEvidence": { "rating": "moderate", "reason": "The assays detect measurable molecular features, but analytical separation does not establish patient benefit." }, "reproducibility": { "rating": "weak", "reason": "Each study used a different assay, population and operating point; no independent team has reproduced the prospective result." }, "commercialConflicts": { "rating": "moderate", "reason": "Several authors on the prospective cfDNA study were affiliated with the assay company; the PCDC comparison is the cleaner independent benchmark." }, "recency": { "rating": "current", "reason": "All three reports were published in 2026 and checked in September 2026." } } }, "layout": { "x": 126, "y": 64 }, "updated": "2026-09-15" }, { "id": "claim-3d-spatial-chemo-resistance-preprint", "title": "A 3D multi-omic preprint links chemotherapy persistence to tumour state and fibroblast crosstalk", "shortTitle": "3D resistance preprint", "type": "claim", "status": "open", "scope": [ "pdac", "transfer-unproven" ], "summary": "Deep-learning-guided mapping of treated human tissue found within-tumour response differences and syndecan-linked fibroblast signals; it has not predicted a patient's treatment response.", "content": "The 2026 preprint uses deep learning to classify three-dimensional morphology in human PDAC after chemotherapy, then applies spatial proteins and RNA to sensitive and persistent regions. It reports epithelial-to-mesenchymal and non-classical states in persistent cells and associates fibroblast–cancer syndecan signalling with resistance. The work is valuable because it measures treated human tissue in three dimensions rather than asking a model to invent resistance. It remains a discovery set: morphology labels, tissue availability and treatment selection can confound the association; the study is not peer reviewed; no locked rule predicted response in a new patient; and the linked patent creates a commercial conflict. The next test should lock a small feature set, apply it blind to pretreatment tissue from a separate neoadjuvant cohort and ask whether it predicts residual viable tumour better than clinical and two-dimensional pathology baselines.", "sourceIds": [ "forjaz-2026-3d-multiomic-preprint" ], "links": [ { "target": "technology-ai-resistance-routing", "relation": "could-inform" }, { "target": "datasets", "relation": "requires-validation-data-from" }, { "target": "claim-immune-exclusion-is-multicellular", "relation": "extends" }, { "target": "unknown-durable-ras-control", "relation": "may-address" } ], "evidence": { "humanProspective": "none", "humanRetrospective": "moderate", "randomised": "none", "independentReplication": "none", "sampleSize": "Treated human-tissue discovery cohorts; exact independent predictive cohort absent", "effect": "Associations with persistent tumour regions; no patient-outcome effect", "limits": [ "Preprint", "No prospective prediction", "Originating-team result", "Feature-selection risk", "Patent conflict", "No treatment decision test" ] }, "layout": { "x": 126, "y": 101 }, "updated": "2026-09-14" }, { "id": "claim-adc-pdac-preclinical", "title": "PDAC antibody–drug conjugates remain target-and-delivery experiments, not treatments", "shortTitle": "ADCs: preclinical", "type": "claim", "status": "open", "scope": [ "pdac", "pan-cancer", "transfer-unproven" ], "summary": "uPAR and Nectin-4 ADCs can kill PDAC models, but no cited programme has shown a patient-level therapeutic window or durable control.", "content": "Recent work shows that a uPAR-directed ADC can suppress PDAC models and alter immune state; separate Nectin-4 work remains cell-line and xenograft based. The attractive engineering claim is clear: use a surface marker to concentrate a potent payload in a poorly supplied tumour. The unsolved human properties are equally clear: uniform antigen across primary and metastases, internalisation, penetration through stroma, payload bystander range and acceptable expression in normal tissue. A mouse response can overstate both delivery and safety. The highest-information next step is not a broad efficacy trial. It is an antibody PET or labelled-microdose study with multi-lesion uptake, paired pathology and normal-organ dosimetry, followed by a small target-positive dose escalation. Stop if fewer than half of lesions meet the uptake threshold or healthy pancreas, bowel, liver or marrow exposure closes the window.", "sourceIds": [ "metrangolo-2025-upar-adc" ], "links": [ { "target": "treatment", "relation": "part-of" }, { "target": "claim-immune-exclusion-is-multicellular", "relation": "depends-on" }, { "target": "hypothesis-fapi-theranostic-pdac", "relation": "shares-selection-logic-with" }, { "target": "experiments", "relation": "requires-dosimetry-first" } ], "evidence": { "humanProspective": "none", "humanRetrospective": "none", "randomised": "none", "independentReplication": "weak", "sampleSize": "Laboratory and animal models only", "effect": "Model tumour suppression; no human effect", "limits": [ "Preclinical", "Antigen heterogeneity", "Normal-tissue expression", "Stromal penetration", "Payload toxicity", "No human dosimetry" ] }, "layout": { "x": 123, "y": 77 }, "updated": "2026-09-14" }, { "id": "claim-adjuvant-mfolfirinox-improves-but-recurrence-remains", "title": "Adjuvant mFOLFIRINOX improves survival after PDAC resection, but most patients still recur or die", "shortTitle": "Post-surgery limit", "type": "claim", "status": "supported", "scope": [ "pdac" ], "summary": "In PRODIGE 24, five-year disease-free survival rose from 19.0% to 26.1%, leaving 73.9% with recurrence or death by five years even in a selected resected population.", "content": "PRODIGE 24 randomised 493 people after complete or near-complete PDAC resection to modified FOLFIRINOX or gemcitabine. At mature follow-up, median disease-free survival was 21.4 versus 12.8 months, five-year disease-free survival was 26.1% versus 19.0%, and five-year overall survival was 43.2% versus 31.4%. The absolute gains were 7.1 percentage points for disease-free survival and 11.8 points for overall survival. This proves that systemic treatment after surgery matters. It also shows the remaining limit: 73.9% of the mFOLFIRINOX group had recurrence or died within five years. That figure is the complement of disease-free survival, not a direct recurrence rate. Trial eligibility selected patients who recovered sufficiently from surgery and were fit for combination chemotherapy, so population outcomes can be worse.", "sourceIds": [ "conroy-2022-prodige24-five-year" ], "links": [ { "target": "residual-disease", "relation": "supports" }, { "target": "treatment", "relation": "supports" }, { "target": "bottlenecks", "relation": "supports" }, { "target": "pdac", "relation": "part-of" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong", "independentReplication": "moderate", "sampleSize": "493 randomised patients after PDAC resection", "effect": "Five-year DFS 26.1% versus 19.0%; five-year OS 43.2% versus 31.4%; OS hazard ratio 0.68", "limits": [ "Selected resected and fit population", "Open-label treatment", "Disease-free survival combines recurrence and death", "Does not identify which occult disease state caused later recurrence" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Systemic treatment after resection addresses occult disease and improved both disease-free and overall survival." }, "animalEvidence": { "rating": "not-applicable", "reason": "The stated benefit and residual failure rate come directly from a randomized human trial." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "The claim does not depend on laboratory sensitivity assays." }, "reproducibility": { "rating": "strong", "reason": "Mature follow-up preserves the treatment effect and systemic adjuvant benefit is established across practice." }, "commercialConflicts": { "rating": "low", "reason": "The cited cooperative-group trial is not presented as a manufacturer-controlled result." }, "recency": { "rating": "recent", "reason": "The cited 2022 report supplies mature five-year outcomes." } } }, "layout": { "x": 71, "y": 63 }, "updated": "2026-09-14" }, { "id": "claim-africa-pdac-case-count-not-decision-data", "title": "A large African pancreatic-cancer case count still lacks the fields needed to choose an intervention", "shortTitle": "African PDAC data gap", "type": "claim", "status": "supported", "scope": [ "pdac", "pancreatic-all" ], "summary": "A 26,850-patient scoping review found more than 70% late-stage presentation, but staging was absent for nearly 40% and complete treatment paths appeared in fewer than half of the included studies.", "content": "Twenty African studies published from 1995 to 2023 contained 26,850 pancreatic-cancer patients. The number looks large, but it does not answer which failure to fund. More than 70% of reported cases were stage III-IV, staging was missing for nearly 40% of the full cohort, complete treatment paths were reported in fewer than half of studies, resection stayed below 15% in most cohorts, adjuvant treatment access varied, and survival was rarely reported. Nearly 40% of 26,850 is about 10,740 records without stage. The evidence was also concentrated in South Africa, Egypt and Morocco. This supports a late-presentation and delivery problem, not its cause. It cannot separate delayed symptom recognition, imaging, pathology, referral, operability, treatment supply, cost or follow-up. The next useful unit is therefore a consecutive case with a dated route from first presentation to outcome, not another pooled case count. A mission registry must record the population denominator, first presentation, imaging and pathology dates, stage, resectability decision, recommended and delivered treatment, reason for non-delivery, travel and cost, and twelve-month status.", "sourceIds": [ "gismalla-2026-africa-pancreatic-review", "asombang-2026-africa-hpb-registry" ], "links": [ { "target": "datasets", "relation": "constrains" }, { "target": "lab-africa-hpb-cancer-consortium", "relation": "defines-gate-for" }, { "target": "unknown-regional-route-loss", "relation": "leads-to" }, { "target": "trial-regional-route-completion-action-map", "relation": "informs" } ], "evidence": { "humanProspective": "none", "humanRetrospective": "moderate — synthesis of observational cohorts", "randomised": "none", "independentReplication": "moderate for late presentation, weak for any one route cause", "sampleSize": "20 studies, 26,850 patients; geographically concentrated and incompletely described", "effect": ">70% stage III-IV among reported cases; nearly 40% of the total cohort lacked stage; complete treatment paths in fewer than half of studies", "limits": [ "Scoping rather than effect review", "Study-level reporting gaps", "Geographic concentration", "Case definitions and care periods vary", "No population denominator", "Missing data may not be random", "No causal comparison" ], "dimensions": { "biologicalPlausibility": { "rating": "not-applicable", "reason": "This claim concerns whether records can support a care decision." }, "animalEvidence": { "rating": "not-applicable", "reason": "Animal evidence cannot fill missing patient routes." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Laboratory models do not identify where care was lost." }, "reproducibility": { "rating": "moderate", "reason": "Late presentation recurs across reports, but the missing fields prevent testing a common cause." }, "commercialConflicts": { "rating": "low", "reason": "The review declared no competing interests." }, "recency": { "rating": "current", "reason": "The synthesis was published in 2026 and searched through June 2025." } } }, "updated": "2026-09-14" }, { "id": "claim-age-alone-cannot-route-pdac-treatment", "title": "Age alone cannot route PDAC treatment; function, nutrition, cognition and disease speed matter", "shortTitle": "Age is not a treatment route", "type": "claim", "status": "supported", "scope": [ "pdac" ], "summary": "Trials in vulnerable older adults show that a birthday or a clinician's activity score cannot replace a structured inventory of what the person can do, tolerate and recover from.", "content": "An older person is not a treatment category. ECOG performance status is a clinician's 0-to-4 score of how active and independent a person appears; the main modern metastatic-PDAC trials largely admitted only scores 0 or 1. GIANT deliberately admitted vulnerable people aged 70-90 with ECOG 0-2. Its 176 participants had median age 77; 46% met the vulnerability rule through cognition, 36% through age at least 80 and 31.4% through other illnesses. Two reduced-intensity regimens produced median overall survival of 4.7 and 4.4 months, P=.72, and severe or life-threatening treatment toxicity in 45.6% and 58.7%, P=.10. The trial stopped at its futility boundary. A later analysis found that ability to perform complex daily tasks such as shopping, cooking and managing medicines, as well as nutrition, depression and quality of life, tracked survival. That association does not prove that fixing those factors extends life. GrantPax exposed the routing failure more directly: a structured ageing-and-function assessment classified 29/32 patients as vulnerable enough for the gentler route, but clinicians placed 15/32, 47%, on a route inconsistent with it. The study was not randomized and enrolled only 32 of 135 planned, so it cannot estimate which drug was better. Three mixed-cancer randomized trials provide the actionable part: teams that acted on nutrition, medicines, mobility, social support and symptoms reduced grade 3 or worse toxicity by 10.1, 20.0 and 15.2 percentage points. Grade 3 means severe enough to need substantial medical treatment and often admission. The first two trials did not improve survival; the newest missed its independence and quality-of-life primary outcomes. The valid conclusion is narrow: age alone must not deny or force treatment, a structured assessment can expose hidden risks, and assessment must trigger named actions. No completed study proves a PDAC survival-improving selection rule.", "sourceIds": [ "conroy-2011-folfirinox-metastatic", "wainberg-2023-napoli3", "dotan-2024-giant", "dotan-2025-giant-ga", "betge-2022-grantpax", "mohile-2021-gap70", "li-2021-gain", "ihorst-2026-geriatric-comanagement" ], "links": [ { "target": "treatment", "relation": "part-of" }, { "target": "unknown-older-pdac-treatment-route", "relation": "leaves-open" }, { "target": "failure-grantpax-geriatric-route", "relation": "includes" }, { "target": "trial-older-pdac-action-map", "relation": "tested-by" }, { "target": "hypothesis-older-pdac-fast-support-route", "relation": "motivates" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong for toxicity reduction; incomplete for PDAC selection", "independentReplication": "moderate", "sampleSize": "GIANT n=176; GrantPax n=32; GAP70+ n=718; GAIN n=605; 2026 co-management full analysis n=207", "effect": "GIANT OS 4.7 versus 4.4 months; randomized mixed-cancer severe-toxicity reductions 10.1 to 20.0 points", "limits": [ "GIANT meeting abstracts", "No proven PDAC survival selector", "Mixed-cancer support trials", "GrantPax under-recruited", "Different assessment and action bundles", "Age thresholds vary" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Organ reserve, nutrition, cognition, medicines and social support can change treatment exposure and harm without changing tumour biology." }, "animalEvidence": { "rating": "not-applicable", "reason": "Human function, treatment delivery and patient outcomes decide this route." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Cell systems cannot represent independence, cognition, treatment choice or hospital use." }, "reproducibility": { "rating": "moderate", "reason": "Three randomized mixed-cancer trials reproduce lower severe toxicity, while patient-valued primary outcomes and survival remain inconsistent or null." }, "commercialConflicts": { "rating": "low", "reason": "The decisive routing and care evidence comes mainly from public or academic studies; chemotherapy manufacturers do not define the claim." }, "recency": { "rating": "current", "reason": "Includes the 2025 GIANT analysis and a 2026 randomized co-management result." } } }, "updated": "2026-09-14" }, { "id": "claim-anticoagulation-prevents-pdac-vte-not-survival", "title": "Anticoagulant prophylaxis prevents many PDAC blood clots but has not extended cancer survival", "shortTitle": "Prevent clots, not tumour death", "type": "claim", "status": "supported", "scope": [ "pdac" ], "summary": "Randomized trials show a large and repeated fall in venous thrombosis, while the survival claim failed and agent, duration and bleeding selection remain patient decisions.", "content": "Three pancreatic-cancer randomized results separate complication control from tumour control. FRAGEM randomized 123 patients and cut VTE during 12 weeks of weight-adjusted dalteparin from 23% to 3.4%; all follow-up VTE fell from 28% to 12%. CONKO-004 randomized 312 and cut three-month symptomatic VTE from 15/152, 9.9%, to 2/160, 1.3%; overall symptomatic VTE fell from 15.1% to 6.4%. Its OS HR was 1.01 and PFS HR 1.06: no survival or cancer-control effect. In the 273-person pancreatic subgroup of CASSINI, rivaroxaban cut the primary event from 10.1% to 3.7% while patients were taking treatment, HR .35, but the day-180 result was 13.0% versus 9.6%, HR .70, P=.328; only 57% completed the double-blind period. A five-trial meta-analysis of 1,003 patients estimated RR .31 and an eight-point absolute fall in VTE, or 1 / 0.08 = 12.5 people treated per event prevented, close to its reported NNT 11.9. Major bleeding was 3.10% versus 2.84%, RR 1.08 with a wide 95% CI .47 to 2.52, so absence of a significant increase is not proof of zero bleeding harm. These data justify offering prophylaxis to suitable high-risk patients. They do not justify calling anticoagulation an anticancer treatment, using one drug for everyone or ignoring tumour invasion, thrombocytopenia, renal function, drug interactions and patient burden.", "sourceIds": [ "maraveyas-2012-fragem", "pelzer-2015-conko004", "vadhan-raj-2020-cassini-pdac", "frere-2020-pdac-thromboprophylaxis-meta", "ash-2023-primary-vte-prophylaxis-guide" ], "links": [ { "target": "treatment", "relation": "part-of" }, { "target": "failure-pdac-vte-prophylaxis-delivery-gap", "relation": "not-fully-delivered-by" }, { "target": "unknown-pdac-vte-prophylaxis-rule", "relation": "leads-to" }, { "target": "trial-pdac-vte-action-map", "relation": "informs" }, { "target": "hypothesis-pdac-vte-route-completion", "relation": "supports" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong", "independentReplication": "strong for VTE reduction; moderate for bleeding safety", "sampleSize": "Dedicated randomized n=123 and n=312; prespecified pancreatic subgroup n=273; five-trial pooled n=1,003", "effect": "Pooled VTE RR .31 and absolute difference -8 points; no OS effect in CONKO-004; major-bleeding RR 1.08 with wide limits", "limits": [ "Different agents and doses", "Older chemotherapy", "Different VTE definitions", "CASSINI treatment discontinuation", "Major bleeding is uncommon and confidence limits are wide", "No survival benefit" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Anticoagulants directly interrupt clot formation and repeated randomized trials reduce observed VTE." }, "animalEvidence": { "rating": "not-applicable", "reason": "The care decision rests on direct randomized patient outcomes." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Coagulation assays cannot establish the balance of clinical clot and bleeding events." }, "reproducibility": { "rating": "strong", "reason": "LMWH and rivaroxaban trials repeat a VTE reduction across separate studies." }, "commercialConflicts": { "rating": "moderate", "reason": "Several drug studies had commercial funding or authors, but effects repeat across agents and groups." }, "recency": { "rating": "mixed", "reason": "Dedicated LMWH trials are older; CASSINI, pooled evidence and current guidance remain relevant." } } }, "layout": { "x": 152, "y": 75 }, "updated": "2026-09-14" }, { "id": "claim-asp1929-pancreatic-human-test-not-started", "title": "ASP-1929 has a pancreatic human plan, not pancreatic human evidence", "shortTitle": "ASP-1929 pancreatic test pending", "type": "claim", "status": "supported", "scope": [ "pdac", "photoimmunotherapy", "egfr", "local-treatment" ], "summary": "The antibody-light product has conditional approval for one head-and-neck setting in Japan, while its first listed pancreatic study has not begun recruitment and has no control.", "content": "ASP-1929, also called cetuximab sarotalocan, joins cetuximab to the light-activated dye IR700. Cetuximab binds EGFR, the epidermal growth factor receptor, on a cell surface. Red 690 nm light then activates the attached dye and damages the membrane of illuminated bound cells. The antibody supplies a cell address; the light supplies a location switch. Japanese regulators gave conditional early approval for unresectable locally advanced or recurrent head and neck cancer in 2020, with all-patient surveillance and later trial data required. That approval does not cover pancreatic cancer. Pancreatic tissue studies support the address but not universal coverage: one found EGFR overexpression in 52/81 cancers, 64.2%. In a separate 745-person pancreatic trial, 90% of tested tumours expressed EGFR, yet ordinary cetuximab added to gemcitabine did not improve survival: 6.3 versus 5.9 months, hazard ratio 1.06. Photo-activated membrane injury is different from blocking EGFR signalling, so that failure does not disprove ASP-1929. It proves that EGFR staining alone is not a benefit argument. The first listed pancreatic ASP-1929 study plans 30 people and remained not yet recruiting at its July 2026 update. Zero pancreatic patients have public treatment outcomes.", "sourceIds": [ "clinicaltrials-nct07698613-asp1929-pancreas", "pmda-2020-akalux-review", "rakuten-medical-2026-asp1929", "kim-2015-egfr-expression-pdac", "philip-2010-swog-s0205-cetuximab", "maruoka-2018-interstitial-photoimmunotherapy" ], "links": [ { "target": "failure-asp1929-selection-light-and-endpoint-design", "relation": "bounded-by" }, { "target": "unknown-asp1929-pancreatic-target-coverage-and-benefit", "relation": "leaves-open" }, { "target": "hypothesis-asp1929-first-ten-measurement-gate", "relation": "motivates" }, { "target": "trial-pdac-asp1929-action-map", "relation": "mapped-by" }, { "target": "company-rakuten-medical", "relation": "involves" }, { "target": "treatment-modality-ranking", "relation": "updates" }, { "target": "treatment", "relation": "part-of" } ], "evidence": { "humanProspective": "none in pancreatic cancer", "humanRetrospective": "none in pancreatic cancer", "randomised": "none for photoimmunotherapy in pancreatic cancer", "independentReplication": "none", "sampleSize": "Pancreatic ASP-1929 outcomes n=0; planned single group n=30; supporting EGFR tissue n=81; cetuximab randomization n=745", "effect": "No pancreatic treatment estimate; head-and-neck regulatory experience cannot establish pancreatic benefit", "limits": [ "Trial has not recruited", "Cross-cancer approval", "EGFR expression varies", "Cetuximab blockade differs from light-triggered killing", "Interstitial-light evidence is in mice", "No pancreatic safety data", "Single-group plan" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "The antibody can bind EGFR and light can activate IR700, but pancreatic delivery and coverage are untested in people." }, "animalEvidence": { "rating": "moderate", "reason": "Inserted light improves deeper treatment in non-pancreatic mouse tumours; product-specific pancreatic evidence is limited." }, "inVitroEvidence": { "rating": "moderate", "reason": "Antibody-targeted light killing is established in model systems, but target density and tissue access vary." }, "reproducibility": { "rating": "none", "reason": "No treated pancreatic outcome is public." }, "commercialConflicts": { "rating": "high", "reason": "Rakuten Medical owns the platform and collaborates on the first pancreatic study." }, "recency": { "rating": "current", "reason": "The pancreatic registry was checked after its July 2026 update." } } }, "layout": { "x": 1006, "y": 930 }, "updated": "2026-09-15" }, { "id": "claim-bilcap-primary-endpoint-uncertainty", "title": "BILCAP supports adjuvant capecitabine, but its primary intention-to-treat survival test was not significant", "shortTitle": "BILCAP uncertainty", "type": "claim", "status": "contested", "scope": [ "cholangiocarcinoma", "gallbladder" ], "summary": "Median survival was 51.1 versus 36.4 months, but the primary adjusted hazard ratio was 0.81 with P=0.097; planned sensitivity and per-protocol analyses were positive.", "content": "BILCAP randomised 447 patients after macroscopically complete resection of cholangiocarcinoma or muscle-invasive gallbladder cancer. In the primary intention-to-treat analysis, median overall survival was 51.1 months with capecitabine and 36.4 months with observation, but the adjusted hazard ratio 0.81 crossed no effect and P=0.097. A planned sensitivity analysis adjusting for nodal status, grade and sex gave hazard ratio 0.71, P=0.010; the per-protocol analysis gave hazard ratio 0.75, P=0.028. The result is compatible with benefit and informed practice, but the strongest unbiased test missed its threshold. The mixed sites, adherence and wide uncertainty mean the atlas records this as contested rather than as either failure or settled cure. Recurrence-free survival remained measured in months, so adjuvant treatment did not remove the residual-disease problem.", "sourceIds": [ "primrose-2019-bilcap" ], "links": [ { "target": "cholangiocarcinoma", "relation": "part-of" }, { "target": "gallbladder", "relation": "part-of" }, { "target": "residual-disease", "relation": "supports" }, { "target": "failures", "relation": "supports" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong", "independentReplication": "moderate", "sampleSize": "447 randomised after resection", "effect": "ITT median OS 51.1 versus 36.4 months; HR 0.81, P=0.097; planned sensitivity HR 0.71, P=0.010", "limits": [ "Primary endpoint not statistically significant", "Open-label observation control", "Mixed biliary sites", "Per-protocol analysis loses full randomisation protection", "44% of treated patients had grade 3 toxicity" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "Cytotoxic treatment can remove residual proliferating disease, but benefit may differ by site and risk." }, "animalEvidence": { "rating": "not-applicable", "reason": "The decision rests on a human randomized trial." }, "inVitroEvidence": { "rating": "moderate", "reason": "Drug activity is established but cannot settle the postoperative survival effect." }, "reproducibility": { "rating": "weak", "reason": "The primary analysis missed its threshold and no independent randomized trial has yet supplied the same comparison." }, "commercialConflicts": { "rating": "moderate", "reason": "Roche supported the trial, although capecitabine is established and the programme was academically led." }, "recency": { "rating": "historical", "reason": "BILCAP remains a practice anchor while ACTICCA-1 results are still absent." } } }, "layout": { "x": 57, "y": 22 }, "updated": "2026-09-14" }, { "id": "claim-biliary-cancers-not-one-disease", "title": "Biliary tract cancers cannot share one molecular treatment map", "shortTitle": "Biliary differences", "type": "claim", "status": "supported", "scope": [ "cholangiocarcinoma", "gallbladder", "hepatobiliary" ], "summary": "Primary site and cause track distinct alteration patterns across intrahepatic, extrahepatic, gallbladder and liver-fluke-associated disease.", "content": "The Japanese 260-tumour study found different fusion and mutation patterns by primary site. The international cholangiocarcinoma study found four groups linked to liver-fluke exposure and distinct genomic or epigenomic alterations. A biomarker, trial or company labelled only ‘biliary tract cancer’ can therefore hide a small responsive subgroup or a negative result caused by mixing diseases. Every later treatment claim must retain the anatomical site, cause and biomarker.", "sourceIds": [ "nakamura-2015-biliary-genomics", "jusakul-2017-cholangiocarcinoma" ], "links": [ { "target": "hepatobiliary", "relation": "part-of" }, { "target": "disease-system", "relation": "supports" }, { "target": "treatment", "relation": "depends-on" }, { "target": "unknowns", "relation": "supports" } ], "evidence": { "humanProspective": "none", "humanRetrospective": "strong", "randomised": "none", "independentReplication": "moderate", "sampleSize": "260 biliary cancers in Nakamura; international cholangiocarcinoma cohort in Jusakul", "effect": "Distinct site- and cause-associated molecular groups", "limits": [ "Retrospective tissue cohorts", "Molecular actionability is not treatment benefit", "Regional causes and ancestry may alter frequencies" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Different duct sites and causes create different cell environments and selective pressures." }, "animalEvidence": { "rating": "moderate", "reason": "Cause-specific models support divergence but do not define human subgroup frequencies." }, "inVitroEvidence": { "rating": "strong", "reason": "Site-linked driver alterations have direct molecular and functional support." }, "reproducibility": { "rating": "moderate", "reason": "Independent international cohorts repeat site and cause differences, with changing exact frequencies." }, "commercialConflicts": { "rating": "low", "reason": "The classification evidence is primarily academic genomic work." }, "recency": { "rating": "historical", "reason": "The foundational cohorts remain valid and later treatment results reinforce subgroup separation." } } }, "layout": { "x": 67, "y": 18 }, "updated": "2026-09-14" }, { "id": "claim-biliary-drainage-is-route-not-cancer-treatment", "title": "Biliary drainage can protect a treatment route but is not itself a pancreatic-cancer treatment", "shortTitle": "Drainage protects the route", "type": "claim", "status": "supported", "scope": [ "pdac", "cholangiocarcinoma", "transfer-unproven" ], "summary": "Modern metal stents can make necessary drainage safer and more durable, while routine drainage still adds a procedure and has not improved cancer survival.", "content": "Two randomized eras answer different questions. In 2010, 202 patients with pancreatic-head cancer and bilirubin 40-250 micromol/L were assigned to surgery within one week or four to six weeks of mainly plastic-stent drainage. Serious complications were 37/96, 39%, versus 75/106, 74%; drainage caused complications in 47, 46%, including 27 cholangitis cases, and mortality did not improve. In 2026, 284 resectable pancreatic or periampullary cancers with bilirubin at least 5.8 mg/dL were assigned to metal-stent drainage or surgery within four weeks. Serious adverse events were 40/138, 29.0%, versus 36/136, 26.5%; the 2.5-point difference had an upper one-sided 95% limit of 11.7% and met a wide 15-point noninferiority margin. Pre-surgery serious events were still 15.2% versus 5.1%. The newer result makes metal-stent drainage a defensible option when cholangitis, symptoms, neoadjuvant therapy or unavoidable delay requires it. It does not show that routine drainage beats prompt surgery, and the expert-centre, sponsor-involved trial does not set a universal route. During neoadjuvant therapy, a small randomized trial cut dysfunction from 72.8% with plastic stents to 18.2% with covered metal stents. A broader review found less reintervention, treatment delay, obstruction and cholangitis with metal stents, but only two included studies were randomized and no R0-resection gain was shown. The patient benefit is therefore fewer infections, procedures and treatment interruptions. Do not count bilirubin fall, stent patency or successful insertion as cancer control.", "sourceIds": [ "van-der-gaag-2010-pbd-rct", "costamagna-2026-sems-early-surgery-rct", "tamura-2021-fcsems-plastic-rct", "lyu-2023-metal-plastic-nat-meta", "esge-2018-biliary-stenting-guideline" ], "links": [ { "target": "treatment", "relation": "part-of" }, { "target": "failure-routine-plastic-pbd", "relation": "learns-from" }, { "target": "failure-biliary-stent-treatment-interruption", "relation": "not-fully-delivered-by" }, { "target": "unknown-pdac-biliary-drainage-route", "relation": "leads-to" }, { "target": "trial-pdac-biliary-drainage-action-map", "relation": "informs" }, { "target": "hypothesis-pdac-biliary-route-completion", "relation": "supports" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "moderate", "randomised": "strong", "independentReplication": "moderate", "sampleSize": "Randomized n=202, n=284 and n=22; mixed-design neoadjuvant meta-analysis", "effect": "Old plastic route raised serious complications by 35 points; modern SEMS was within a 15-point noninferiority margin; no cancer-survival benefit", "limits": [ "Different eras and devices", "Mixed periampullary cancers", "Wide noninferiority margin", "Expert centres", "Sponsor involvement", "No survival benefit" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Removing obstruction can prevent infection and permit treatment, while the procedure itself creates infection and injury risks." }, "animalEvidence": { "rating": "not-applicable", "reason": "The decision rests on direct procedural outcomes in patients." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Laboratory tests cannot estimate procedural harm or treatment completion." }, "reproducibility": { "rating": "moderate", "reason": "Multiple trials support metal-stent durability, but route and population differ materially." }, "commercialConflicts": { "rating": "high", "reason": "The main modern trials used a named device and included manufacturer roles." }, "recency": { "rating": "current", "reason": "The evidence includes a 2026 randomized trial checked at publication." } } }, "updated": "2026-09-14" }, { "id": "claim-biliary-immunochemotherapy-tail", "title": "Durvalumab added a modest median benefit and a persistent three-year survival tail in mixed advanced biliary cancers", "shortTitle": "TOPAZ-1 tail", "type": "claim", "status": "supported", "scope": [ "cholangiocarcinoma", "gallbladder" ], "summary": "With 41.3 months' median follow-up, TOPAZ-1 reduced the death hazard by 26% and raised three-year survival from 6.9% to 14.6%, but pooled several biliary primary sites.", "content": "TOPAZ-1 randomised 685 untreated advanced biliary-tract cancer patients to durvalumab or placebo with gemcitabine and cisplatin. At the mature exploratory update, median overall survival was 12.9 versus 11.3 months, death hazard ratio 0.74 and estimated 36-month survival 14.6% versus 6.9%. The absolute median gain was 1.6 months; the three-year absolute gain was 7.7 points and the survival rate was about 2.12 times the control rate. That tail is clinically important, but 85.4% of the durvalumab group was not alive at three years. Because intrahepatic, extrahepatic and gallbladder cancers were pooled, site-specific certainty is lower than the overall result. The update is longer follow-up of the same trial, not independent replication. Unlike unselected PDAC checkpoint monotherapy, this tested checkpoint treatment with chemotherapy and in a different disease system; the result cannot identify which biological difference caused activity or transfer the effect to PDAC.", "sourceIds": [ "oh-2022-topaz1", "oh-2025-topaz1-three-year" ], "links": [ { "target": "cholangiocarcinoma", "relation": "part-of" }, { "target": "gallbladder", "relation": "part-of" }, { "target": "treatment", "relation": "supports" }, { "target": "transfer", "relation": "supports" }, { "target": "company-astrazeneca", "relation": "sponsored-by" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong", "independentReplication": "weak — mature follow-up of the same trial", "sampleSize": "685 randomised patients with advanced biliary-tract cancer", "effect": "Mature death HR 0.74; median OS 12.9 versus 11.3 months; estimated 36-month OS 14.6% versus 6.9%", "limits": [ "Mixed primary sites", "Does not isolate the immune component from chemotherapy interaction", "Most patients not alive at three years", "AstraZeneca-funded", "Does not transfer directly to PDAC" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "Chemotherapy can release antigens and durvalumab blocks PD-L1, but the active interaction is not isolated." }, "animalEvidence": { "rating": "not-applicable", "reason": "The clinical claim rests on randomized human survival rather than animal activity." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Laboratory evidence is not needed to establish the observed mixed-site survival effect." }, "reproducibility": { "rating": "weak", "reason": "The three-year analysis matures the same TOPAZ-1 population rather than independently repeating it." }, "commercialConflicts": { "rating": "high", "reason": "AstraZeneca funded the trial and company staff contributed to the evidence programme." }, "recency": { "rating": "recent", "reason": "The decisive three-year update was reported in 2025." } } }, "layout": { "x": 48, "y": 18 }, "updated": "2026-09-14" }, { "id": "claim-cabozantinib-pnet-pfs-not-cure", "title": "Cabozantinib strongly delayed pNET progression after prior treatment, with substantial toxicity", "shortTitle": "CABINET pNET", "type": "claim", "status": "supported", "scope": [ "pnet" ], "summary": "In the 95-patient pancreatic cohort, median progression-free survival was 13.8 versus 4.4 months, while severe adverse events affected roughly two thirds on treatment.", "content": "CABINET independently randomised pancreatic and extrapancreatic neuroendocrine cohorts after prior targeted treatment or radioligand therapy. In 95 pNET patients, cabozantinib increased median progression-free survival from 4.4 to 13.8 months; hazard ratio 0.23, and response was 19% versus 0%. Grade 3 or higher adverse events occurred in 62% to 65% of cabozantinib-treated patients across cohorts versus 23% to 27% with placebo. This is strong evidence of delayed progression in a heavily treated disease, not evidence that patients are cured or that life is extended by the same amount. Mature overall survival, sequencing against other active drugs and quality-adjusted time are needed to judge durable control.", "sourceIds": [ "chan-2025-cabinet" ], "links": [ { "target": "pnet", "relation": "part-of" }, { "target": "claim-pnet-targeted-control-differs-from-pdac", "relation": "extends" }, { "target": "treatment", "relation": "supports" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong", "independentReplication": "moderate", "sampleSize": "95 randomised patients in the pancreatic neuroendocrine cohort", "effect": "Median PFS 13.8 versus 4.4 months; HR 0.23; response 19% versus 0%", "limits": [ "Previously treated progressive disease", "PFS primary endpoint", "Overall survival not established in the primary report", "Severe adverse events common", "Small disease-specific cohort" ] }, "layout": { "x": 3, "y": 30 }, "updated": "2026-09-14" }, { "id": "claim-cachexia-is-parallel-lethal-system", "title": "PDAC cachexia is a parallel host failure that can shorten treatment and life even when the tumour is not controlled", "shortTitle": "Host wasting system", "type": "claim", "status": "supported", "scope": [ "pdac" ], "summary": "Rapid muscle loss doubled the observed death hazard; two randomized interventions changed weight or muscle, but neither completed the chain to longer survival.", "content": "Tumour burden is not the only system determining whether a patient can receive enough treatment or outlive PDAC. In 164 people with advanced pancreatic cancer, those in the quartile with the greatest skeletal-muscle loss over roughly two to four months had twice the death hazard of the quartile that gained muscle: HR 2.01, 95% CI 1.12 to 3.62. This is prognostic association, not proof that muscle loss itself caused death. Two randomized trials give the causal programme harder boundaries. Adding tocilizumab reduced early muscle loss, but the six-month survival endpoint failed, median survival, progression and response did not improve, and grade 3 or worse treatment-related events rose from 63.4% to 88.1%. Ponsegromab produced 1.22 to 2.81 kg placebo-adjusted weight gains at twelve weeks in a GDF-15-selected mixed-cancer trial and improved activity and symptoms at 400 mg; it did not test survival or chemotherapy delivery as primary outcomes. Cachexia therefore deserves its own treatment branch, but it cannot substitute for tumour control. The live 982-person PDAC ponsegromab trial should be used before another large study. Its result must be read through function, treatment dose delivered, quality of life, toxicity and survival, not weight alone.", "sourceIds": [ "babic-2019-muscle-loss-survival", "chen-2025-tocilizumab-cachexia", "groarke-2024-ponsegromab-cachexia", "clinicaltrials-nct06989437-ponsegromab-pdac" ], "links": [ { "target": "claim-il6-cachexia-signal", "relation": "constrained-by" }, { "target": "model-pdac-causal-chain", "relation": "supports" }, { "target": "treatment", "relation": "part-of" }, { "target": "risks", "relation": "contributes-to" }, { "target": "claim-ponsegromab-gdf15-cachexia-driver-not-survival", "relation": "supported-by" }, { "target": "unknown-cachexia-treatment-utility", "relation": "leads-to" }, { "target": "trial-pdac-cachexia-action-map", "relation": "tested-by" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "strong", "randomised": "strong", "independentReplication": "moderate", "sampleSize": "164-person longitudinal CT cohort and 147 treated in randomized phase 2 including safety run-in", "effect": "Greatest muscle-loss quartile death HR 2.01; tocilizumab improved muscle measures but missed OS6 and increased grade 3+ treatment events to 88.1%", "limits": [ "Muscle-loss survival link is observational", "Advanced disease only", "Interventions test different biological routes", "IL-6 primary survival endpoint was negative", "Ponsegromab primary endpoint was short-term weight", "Muscle or weight change may not mediate survival" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Longitudinal loss predicts death and randomized interventions can move host measures, but the survival route remains incomplete." }, "animalEvidence": { "rating": "moderate", "reason": "Models support inflammatory and appetite-control routes, while the decision depends on human function and survival." }, "inVitroEvidence": { "rating": "weak", "reason": "Cell systems cannot reproduce whole-body wasting, appetite, activity or chemotherapy completion." }, "reproducibility": { "rating": "moderate", "reason": "Different human studies repeat the existence of host wasting, but not a treatment that improves durable patient benefit." }, "commercialConflicts": { "rating": "high", "reason": "The strongest positive intervention evidence is owned and funded by its sponsor; the negative IL-6 result is less commercially promoted." }, "recency": { "rating": "current", "reason": "The evidence includes 2024 and 2025 randomized results and a trial record updated in August 2026." } } }, "layout": { "x": 106, "y": 60 }, "updated": "2026-09-14" }, { "id": "claim-case-control-blood-tests-overstate-screening-readiness", "title": "Case-control blood-test accuracy does not establish screening performance", "shortTitle": "Case-control trap", "type": "claim", "status": "supported", "scope": [ "pdac", "hepatobiliary", "colorectal", "pan-cancer" ], "summary": "CancerSEEK and early fragmentomics studies separated known cancers from controls, but did not test an asymptomatic population before diagnosis.", "content": "CancerSEEK reported a median sensitivity of 70% across eight nonmetastatic, clinically detected cancers and more than 99% specificity against healthy controls. A fragment-pattern classifier reported 57% to more than 99% sensitivity across seven cancers at 98% specificity and reached 91% when combined with mutation analysis. These were important technology demonstrations. They do not supply screening positive predictive value, lead-time distribution or mortality effect because cases already had clinically recognised cancer and controls did not represent the full range of inflammatory, benign and premalignant conditions encountered in practice. At pancreatic-cancer incidence near 13.9 per 100,000 per year, even 98% specificity would create roughly 2,000 false positives per 100,000 tests before accounting for repeat testing. Prospective intended-population studies are therefore a separate gate, not a routine extension of case-control accuracy.", "sourceIds": [ "cohen-2018-cancerseek", "cristiano-2019-fragmentomics", "seer-pancreas-statfacts" ], "links": [ { "target": "claim-screening-base-rate-maths", "relation": "supports" }, { "target": "detection", "relation": "part-of" }, { "target": "failures", "relation": "supports" } ], "evidence": { "humanProspective": "none", "humanRetrospective": "moderate", "randomised": "none", "independentReplication": "moderate", "sampleSize": "CancerSEEK: 1,005 cases and 812 controls; fragmentomics: 236 cases and 245 controls", "effect": "High case-control discrimination; population screening effect not measured", "limits": [ "Clinically detected cases", "Control spectrum narrower than clinical practice", "Cancer types pooled", "Stage-specific pancreatic sensitivity incompletely reported in abstracts", "Inventor and company interests" ] }, "layout": { "x": 22, "y": 114 }, "updated": "2026-09-14" }, { "id": "claim-cca-ctdna-prognostic-not-action-guiding", "title": "Postoperative biliary ctDNA predicts recurrence but does not yet select a treatment that improves survival", "shortTitle": "CCA ctDNA action gap", "type": "claim", "status": "supported", "scope": [ "cholangiocarcinoma", "gallbladder", "hepatobiliary" ], "summary": "Serial tumour-informed ctDNA found 15 of 16 recurrences early in one 56-person cohort and strongly separated risk in STAMP, but no study randomized an action from the result.", "content": "The signal is prognostic; the action is missing. In a retrospective multicentre cohort of 56 resected stage I-III biliary tract cancers, 16 recurred and ctDNA detected 15, 15/16 = 93.8%, an average 3.7 months before confirmed recurrence. Positivity in the postoperative residual-disease window was associated with relapse-free survival, HR 26, but the 95% confidence interval was 2.6-265 and median follow-up was only 12.8 months. In the STAMP extrahepatic cholangiocarcinoma biomarker cohort, 89 of 101 randomized patients supplied 254 prospective samples. ctDNA positivity was associated with worse disease-free survival before adjuvant treatment, HR 1.8, and more strongly at 12 weeks, HR 7.72, and 24 weeks, HR 5.24. Yet capecitabine and gemcitabine-cisplatin did not differ significantly in disease-free or overall survival, and ctDNA did not assign the treatment. A 2026 prospective feasibility study enrolled 18 but analysed 14 after one inoperable case and three panel-construction failures; only 12 had postoperative comparisons. Prediction, analytical delivery and treatment benefit are therefore three different gates. The next trial must randomize a safe treatment or imaging action among ctDNA-positive patients and keep ctDNA-negative de-escalation as a separate question.", "sourceIds": [ "yu-2025-btc-ctdna-recurrence", "yoo-2025-extrahepatic-cca-ctdna", "park-2026-btc-ctdna-feasibility", "clinicaltrials-nct05743959-btc-mrd" ], "links": [ { "target": "claim-icc-resection-often-leaves-recurrence-risk", "relation": "extends" }, { "target": "residual-disease", "relation": "supports" }, { "target": "failure-cca-mrd-prediction-without-action", "relation": "supports" }, { "target": "unknown-cca-postoperative-mrd-action", "relation": "exposes" }, { "target": "trial-cca-action-map", "relation": "informs" }, { "target": "cholangiocarcinoma", "relation": "part-of" } ], "evidence": { "humanProspective": "moderate — prospective nested and feasibility cohorts", "humanRetrospective": "moderate — multicentre real-world cohort", "randomised": "moderate — treatment comparison in STAMP, but not ctDNA-directed action", "independentReplication": "moderate for prognosis; none for utility", "sampleSize": "Retrospective n=56 with 16 recurrences; STAMP biomarker n=89 and 254 samples; prospective feasibility n=18 with 14 analysed", "effect": "15/16 recurrences detected with 3.7-month average lead; STAMP on-treatment ctDNA HRs 7.72 and 5.24 for disease-free survival", "limits": [ "Mixed biliary sites in two studies", "Short follow-up", "Wide confidence intervals", "Tumour-informed panels require tissue and construction time", "No ctDNA-directed treatment comparison", "Commercial assay involvement can affect publication and testing patterns" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Tumour-derived DNA after resection is a direct sign that malignant clones remain or have returned." }, "animalEvidence": { "rating": "not-applicable", "reason": "Human residual disease and treatment utility must be measured in patients." }, "inVitroEvidence": { "rating": "moderate", "reason": "Assay detection is technically supported, but analytical sensitivity differs by tumour shedding and panel construction." }, "reproducibility": { "rating": "moderate", "reason": "Several cohorts repeat the prognostic association, while no randomized action benefit exists." }, "commercialConflicts": { "rating": "high", "reason": "Tumour-informed assays are proprietary and some studies include assay-company authors or real-world testing selected through commercial use." }, "recency": { "rating": "current", "reason": "The record includes 2025 and 2026 cohorts and a live registry checked in September 2026." } } }, "updated": "2026-09-14" }, { "id": "claim-cca-psc-surveillance-workload", "title": "Annual PSC surveillance produces a large abnormal-imaging workload without durable curable detection", "shortTitle": "PSC surveillance workload", "type": "claim", "status": "supported", "scope": [ "cholangiocarcinoma", "hepatobiliary" ], "summary": "Five years of annual MRI/MRCP and CA19-9 in 512 people found severe or progressive duct change in 24%, but only about one in ten of those findings was malignant and median survival after cholangiocarcinoma was 13 months.", "content": "The prospective Swedish PSC cohort makes the workload visible. Among 512 unselected people followed for five years, 122 had severe or progressive duct change on MRI/MRCP and were referred for further investigation. That is 122/512 = 23.8% of the cohort. Malignancy was present in 10% of that imaging subgroup, so the reported threshold produced roughly nine non-malignant severe or progressive findings for each malignancy. Across the whole cohort, 11 cholangiocarcinomas and two high-grade dysplasias were diagnosed, 13/512 = 2.54% over five years, yet median survival after cholangiocarcinoma was 13 months. The imaging finding was strongly associated with later dysplasia or malignancy, but association did not turn into early curable detection. A retrospective Australian cohort of 298 people associated regular MRCP with lower all-cause mortality, HR 0.29, while survival after hepatobiliary cancer diagnosis remained unchanged. That result can reflect transplant access, referral and wider care rather than the surveillance test itself. The next PSC programme must therefore measure the entire chain: abnormal scan, ERCP or biopsy, benign procedure, transplant, interval cancer, resectability and cause-specific death. A new marker is useful only if it reduces this workload or moves action early enough to change those outcomes.", "sourceIds": [ "villard-2023-psc-surveillance", "tan-2024-psc-mrcp-surveillance" ], "links": [ { "target": "claim-cca-surveillance-depends-on-risk-population", "relation": "extends" }, { "target": "failure-psc-surveillance-no-curable-detection", "relation": "supports" }, { "target": "trial-cca-action-map", "relation": "informs" }, { "target": "unknown-cca-surveillance-utility", "relation": "constrains" }, { "target": "cholangiocarcinoma", "relation": "part-of" } ], "evidence": { "humanProspective": "strong — nationwide protocol-driven cohort", "humanRetrospective": "moderate — independent multicentre association", "randomised": "none", "independentReplication": "weak — direction differs by design and endpoint", "sampleSize": "Swedish prospective n=512 for five years; Australian retrospective n=298 and 2,117 person-years", "effect": "23.8% had severe or progressive duct change; malignancy in 10% of that subgroup; 11 cholangiocarcinomas; median survival 13 months", "limits": [ "The paper reports the imaging subgroup malignancy percentage rather than every diagnostic procedure", "No randomized surveillance comparison", "All-cause mortality can reflect wider care", "PSC severity and transplant access differ across centres", "Five years may miss later cancers" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "PSC creates duct inflammation and strictures that can both precede and mimic cholangiocarcinoma." }, "animalEvidence": { "rating": "not-applicable", "reason": "Animal models cannot estimate the human diagnostic workload or survival effect." }, "inVitroEvidence": { "rating": "weak", "reason": "Laboratory markers may improve classification but do not establish the care-path result." }, "reproducibility": { "rating": "weak", "reason": "Prospective and retrospective cohorts do not agree on the implied benefit and use different endpoints." }, "commercialConflicts": { "rating": "low", "reason": "The main evidence is academic, though imaging and endoscopy intensity still creates service incentives." }, "recency": { "rating": "current", "reason": "The prospective and retrospective reports were checked through September 2026." } } }, "updated": "2026-09-14" }, { "id": "claim-cca-surveillance-depends-on-risk-population", "title": "Cholangiocarcinoma surveillance cannot use one rule across PSC and endemic liver-fluke risk", "shortTitle": "CCA surveillance split", "type": "claim", "status": "controversial", "scope": [ "cholangiocarcinoma", "hepatobiliary" ], "summary": "Annual MRI/MRCP plus CA19-9 failed to create curable detection in prospective PSC surveillance, while endemic-area ultrasound shifted stage in a nonrandomized Thai programme.", "content": "The intended population changes both the base rate and the detectable precursor route. In Sweden, 512 unselected people with primary sclerosing cholangitis underwent annual MRI/MRCP, CA19-9 and clinical review for five years. Eleven cholangiocarcinomas were diagnosed, but median survival was 13 months and the programme did not find disease early enough for long-term survival. Australian retrospective data appear more favourable: among 298 people with PSC, regular MRCP was associated with lower all-cause mortality after weighting, but survival after hepatobiliary cancer diagnosis was unchanged. That pattern can reflect broader care intensity rather than cancer interception. In northeastern Thailand, an ultrasound programme found stage 0-II disease in 84.5% of 161 screen-detected histologically proven cancers versus 21.6% of 601 walk-in cancers. The comparison was not randomized, conditioned on diagnosed cancer and did not quantify population mortality or the full false-positive work-up. The evidence supports separate development paths: better molecular and duct-change classifiers in PSC, and controlled ultrasound-based delivery studies in very-high-incidence liver-fluke regions. It does not support general-population cholangiocarcinoma screening.", "sourceIds": [ "villard-2023-psc-surveillance", "tan-2024-psc-mrcp-surveillance", "khuntikeo-2020-thailand-cca-screening" ], "links": [ { "target": "model-cholangiocarcinoma-causal-chain", "relation": "supports" }, { "target": "detection-modality-ranking", "relation": "extends" }, { "target": "cholangiocarcinoma", "relation": "part-of" }, { "target": "unknowns", "relation": "exposes" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "moderate", "randomised": "none", "independentReplication": "weak — conflicting evidence", "sampleSize": "512 prospective PSC surveillance participants; 298 retrospective PSC participants; 762 histologically proven Thai cholangiocarcinomas", "effect": "Prospective PSC surveillance did not yield long survival; Thai screening cases were 84.5% stage 0-II versus 21.6% in walk-in cases", "limits": [ "No randomized cancer-mortality comparison", "Risk populations are biologically and epidemiologically different", "Retrospective care-intensity confounding", "Thai comparison conditions on diagnosed cancer", "False-positive procedures incompletely reported" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "PSC strictures and liver-fluke injury create different precursor and detection routes." }, "animalEvidence": { "rating": "moderate", "reason": "Models support liver-fluke carcinogenesis but not surveillance benefit." }, "inVitroEvidence": { "rating": "weak", "reason": "Laboratory evidence cannot establish a population action threshold." }, "reproducibility": { "rating": "weak", "reason": "PSC studies conflict by design and the Thai stage shift lacks a controlled mortality repeat." }, "commercialConflicts": { "rating": "low", "reason": "The main studies are health-system programmes, though more imaging and procedures can create service incentives." }, "recency": { "rating": "current", "reason": "PSC and endemic-area evidence was checked through September 2026." } } }, "layout": { "x": 45, "y": 23 }, "updated": "2026-09-14" }, { "id": "claim-cd40-signal-without-contribution-proof", "title": "CD40 activation has an uncontrolled mitazalimab signal but no proven added survival", "shortTitle": "CD40 contribution unresolved", "type": "claim", "status": "supported", "scope": [ "pdac", "pan-cancer", "transfer-unproven" ], "summary": "Mitazalimab plus mFOLFIRINOX produced 42.1% response and 14.9-month median survival in 57 people, but PRINCE and the missing control mean CD40 contribution is unknown.", "content": "Mechanism. CD40 is an activation switch on cells that show cancer fragments to T cells. An agonist presses that switch; it does not itself identify the cancer cell. OPTIMIZE-1 reported a confirmed response in 42.1% of 57 evaluable people, median response duration 12.6 months, progression-free survival 7.7 months and overall survival 14.9 months with mitazalimab plus mFOLFIRINOX. The first report shows the route denominator: 88 screened, 70 enrolled, 57 included in efficacy after at least two cycles at the chosen dose. The registry later grew to 94 but posts no final result. There was no mFOLFIRINOX-only group, so chemotherapy, patient selection, supportive care and the CD40 drug cannot be separated. Cross-check. In PRINCE, sotigalimab plus chemotherapy missed its one-year survival test against a historical number, as did the sotigalimab/nivolumab triplet. The study randomized three active combinations but had no chemotherapy-only group, mixed 12 non-randomized phase 1b patients into the 105-person efficacy set and was not powered to compare groups. Its immune measurements can generate a selector idea but cannot prove contribution. Decision. Mitazalimab is a real human signal worth preserving, not a treatment effect worth assuming. It advances only after final all-enrolled data and a concurrent fixed-backbone survival test.", "sourceIds": [ "van-laethem-2024-optimize1", "van-laethem-2025-optimize1-biomarkers", "clinicaltrials-nct04888312-optimize1", "padron-2022-prince", "clinicaltrials-nct03214250-prince", "alligator-2026-q2" ], "links": [ { "target": "treatment", "relation": "part-of" }, { "target": "failure-immune-reprogramming-without-patient-benefit", "relation": "bounded-by" }, { "target": "unknown-cd40-added-survival-pdac", "relation": "creates" }, { "target": "hypothesis-mitazalimab-controlled-survival-gate", "relation": "tested-by" }, { "target": "company-alligator-bioscience", "relation": "developed-by" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "not-applicable", "randomised": "weak", "independentReplication": "weak", "sampleSize": "OPTIMIZE-1 efficacy n=57 and safety n=70, registry n=94; PRINCE efficacy n=105 without chemotherapy-only control", "effect": "Mitazalimab ORR 42.1%, median PFS 7.7 months and OS 14.9 months; added effect unknown", "limits": [ "No concurrent mitazalimab control", "Efficacy excludes some enrolled people", "Final 94-person result absent", "Different CD40 drugs and chemotherapy backbones", "Marker associations were found inside treated groups" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The immune activation chain is measurable and human immune changes occurred." }, "animalEvidence": { "rating": "strong", "reason": "Mouse PDAC models supported CD40 combinations but do not isolate human benefit." }, "inVitroEvidence": { "rating": "strong", "reason": "Receptor activation and immune-cell effects can be measured directly." }, "reproducibility": { "rating": "weak", "reason": "The favourable product result is single-arm and another CD40 programme did not pass its historical survival gate." }, "commercialConflicts": { "rating": "high", "reason": "Alligator funded and owns the positive product record and needs an asset transaction." }, "recency": { "rating": "current", "reason": "Includes the current registry and August 2026 company position." } } }, "layout": { "x": 274, "y": 206 }, "updated": "2026-09-15" }, { "id": "claim-cd73-blockade-signal-without-concurrent-control", "title": "Quemliclustat reaches the human adenosine pathway; its survival signal still comes from an external control", "shortTitle": "CD73 signal, no concurrent control", "type": "claim", "status": "open", "scope": [ "pdac", "transfer-unproven" ], "summary": "ARC-8 measured tumour-pathway change and 15.7-month survival, but every randomized arm received quemliclustat and the drug-effect comparison used matched patients from other trials.", "content": "CD73 is an enzyme on cells and in fluid that converts AMP, a breakdown product released by stressed or damaged cells, into adenosine. Adenosine is a local stand-down signal: it limits inflammation in normal tissue but can also suppress cancer-killing immune cells around a tumour. Quemliclustat is a small molecule that blocks CD73. ARC-8 gives credible human mechanism evidence and weak efficacy evidence. It screened 158 and enrolled 116 people in dose expansion. Ninety were randomized, but the comparison was quemliclustat plus chemotherapy in 29 people versus the same package plus zimberelimab, a PD-1 immune-brake antibody, in 61. It therefore estimated the effect of adding zimberelimab, not the effect of quemliclustat. Across 122 people who received 100 mg quemliclustat, median overall survival was 15.7 months. A post-hoc matched group of 122 patients drawn from four other trials had survival of 9.8 months, HR 0.634, P=0.003. Yet objective response was 39% versus 41%, P=0.794, and progression-free survival was 6.3 versus 5.5 months, P=0.110. Matching balances recorded variables; it cannot balance unrecorded selection, care, calendar time or later treatment. In 37 paired biopsies, larger treatment-linked falls in the NR4A gene programme tracked T-cell activation and longer survival, HR 0.24, but this again defines groups after treatment. A different CD73 blocker, oleclumab, supplies the class warning: in 170 randomized patients its main response endpoint failed; oleclumab plus chemotherapy had OS HR 1.26, 95% CI 0.79-1.98, and adding durvalumab gave HR 0.75, 95% CI 0.50-1.13. Quemliclustat may differ by molecule and tumour penetration. Only PRISM-1's placebo-controlled survival result can estimate its added effect.", "sourceIds": [ "wainberg-2026-quemliclustat-arc8", "coveler-2024-oleclumab-pdac", "clinicaltrials-nct06608927-prism1" ], "links": [ { "target": "treatment", "relation": "constrains" }, { "target": "treatment-long-tail-audit", "relation": "expands" }, { "target": "failure-adenosine-control-and-marker-pdac", "relation": "explains" }, { "target": "unknown-quemliclustat-survival-contribution", "relation": "leaves-open" }, { "target": "trial-pdac-adenosine-action-map", "relation": "informs" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "moderate", "randomised": "weak", "independentReplication": "moderate", "sampleSize": "ARC-8 expansion n=116 with n=90 randomized between two quemliclustat regimens; matched external comparison n=122 versus n=122; oleclumab randomized n=170", "effect": "Quemliclustat matched-external OS HR 0.634; no significant response or PFS difference; oleclumab direct randomized OS intervals included no effect", "limits": [ "No concurrent randomized quemliclustat-free arm", "Post-hoc external control", "Post-treatment biomarker grouping", "ARC-8 was phase 1b and sponsor-authored", "Oleclumab is a different molecule", "PRISM-1 has no result" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Human tumours show the pathway and treatment-linked NR4A change." }, "animalEvidence": { "rating": "strong", "reason": "CD73 inhibition improves tumour immune activity in several models." }, "inVitroEvidence": { "rating": "strong", "reason": "Adenosine-driven NR4A expression and its inhibition were directly measured." }, "reproducibility": { "rating": "weak", "reason": "A second CD73 product did not establish overall-population benefit." }, "commercialConflicts": { "rating": "high", "reason": "Arcus employees authored ARC-8 and the product is licensed across commercial partners." }, "recency": { "rating": "current", "reason": "The decisive phase 1 report and phase 3 status are from 2026." } } }, "layout": { "x": 436, "y": 372 }, "updated": "2026-09-15" }, { "id": "claim-cldn18-bispecific-early", "title": "IBI389 produced early PDAC responses, but its mature paper withholds PDAC efficacy", "shortTitle": "CLDN18.2 bispecific", "type": "claim", "status": "open", "scope": [ "pdac", "pan-cancer", "transfer-unproven" ], "summary": "A 7/23 preliminary response result did not mature into public efficacy accounting for the 73 pancreatic patients in the full phase 1 paper.", "content": "Mechanism. IBI389 is a two-ended antibody: one end binds CLDN18.2 on a tumour cell and the other binds CD3 on a T cell, bringing the cell and killer together. Preliminary PDAC result. Sixty-four previously treated pancreatic patients received the drug. At 600 micrograms/kg, 23 people with CLDN18.2 staining of at least 2+ in at least 10% of tumour cells were response evaluable; seven responded, ORR 30.4% with a wide 95% CI of 13.2-52.9, and 16 had response or stable disease. Safety was heavy: 54.7% had grade 3 or worse treatment-related events and 51.6% had cytokine-release syndrome, an abrupt immune alarm that can cause fever, low blood pressure and breathing problems, though none was severe in that report. Mature-accounting problem. The later peer-reviewed paper includes 121 treated people and 73 pancreatic cancers, yet reports mature response, PFS and OS only for 27 gastric or gastro-oesophageal cancers. Across all 121, 99.2% had a treatment-related event, 21.5% had severe GGT elevation, a liver or bile-duct enzyme change, and 59.5% had cytokine-release syndrome, including one severe case. The current registry is stale and says unknown status after its planned completion. Decision. The 7/23 result is a real early signal, not proof of survival or even a mature pancreatic response rate. Request response, duration, progression, survival, dose, marker and adverse-event records for all 73 pancreatic patients. No new IBI389 cohort or capital is justified until every pancreatic denominator is public and a controlled test is designed.", "sourceIds": [ "hao-2024-ibi389-pdac", "xu-2025-ibi389-phase1", "clinicaltrials-nct05164458-ibi389" ], "links": [ { "target": "treatment", "relation": "part-of" }, { "target": "claim-immune-exclusion-is-multicellular", "relation": "must-overcome" }, { "target": "transfer-her2-selection", "relation": "warned-by" }, { "target": "unknowns", "relation": "part-of" }, { "target": "failure-antibody-target-and-payload-transfer-pdac", "relation": "bounded-by" }, { "target": "trial-cldn182-antibody-action-map", "relation": "mapped-by" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "not-applicable", "randomised": "none", "independentReplication": "none", "sampleSize": "Preliminary PDAC treated n=64 and selected efficacy n=23; mature paper total n=121 including 73 PDAC", "effect": "Preliminary PDAC ORR 7/23=30.4%; mature PDAC efficacy omitted; CRS 59.5% across the mature safety set", "limits": [ "Phase 1", "Wide response interval", "No control", "Mature PDAC efficacy not reported", "Common cytokine-release syndrome", "Registry status unknown", "Company-linked product" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "The molecule can physically recruit T cells to marker-bearing tumour cells, but entry and killing inside PDAC remain unmeasured." }, "animalEvidence": { "rating": "moderate", "reason": "Model support exists but is less useful than the incomplete human record." }, "inVitroEvidence": { "rating": "strong", "reason": "Dual binding and T-cell activation are measurable, but they do not establish patient benefit." }, "reproducibility": { "rating": "none", "reason": "No controlled or independent PDAC efficacy result exists, and the mature paper does not report the pancreatic outcome set." }, "commercialConflicts": { "rating": "high", "reason": "Innovent owns the product and controls the full trial record." }, "recency": { "rating": "current", "reason": "Uses the 2024 PDAC abstract, 2026 full paper and current stale-registry state." } } }, "layout": { "x": 120, "y": 74 }, "updated": "2026-09-15" }, { "id": "claim-cldn182-adc-selected-response", "title": "IBI343 has a marker-linked PDAC response signal, not yet a survival effect", "shortTitle": "IBI343 selected response", "type": "claim", "status": "mixed", "scope": [ "pdac", "pan-cancer", "transfer-unproven" ], "summary": "Ten of 44 high-marker tumours responded at 6 mg/kg versus none of 12 below the analysis cutoff, but the phase 1 had no control.", "content": "Engineering model. An antibody-drug conjugate is a guided carrier: the antibody is the address label, CLDN18.2 is the surface address, and exatecan is the DNA-damaging payload. IBI343 silences the antibody's immune-killing tail, then attaches exatecan so more payload should enter marker-bearing cells and less should circulate freely. Human signal. Entry required at least moderate CLDN18.2 staining in 40% of tumour cells. At 6 mg/kg, the reported analysis split patients at staining of at least 1+ in 60% of cells. Among 44 above that boundary, ten responded, ORR 22.7% (95% CI 11.5-37.8), disease control was 81.8%, median PFS 5.4 months and OS 8.5 months. Among 12 below it, nobody responded, disease control was 41.7%, PFS 1.4 months and OS 6.2 months. The response contrast is 10/44 minus 0/12 = 22.7 percentage points. Safety. Across 83 treated patients, 50.6% had grade 3 or worse treatment-emergent events, 7.2% stopped treatment because of an event, and no event caused death. Main severe problems were anaemia and low white-cell counts. Boundary. Free exatecan plus gemcitabine had already failed a 349-person randomized PDAC trial: OS 6.7 versus 6.2 months, P=.52, with more severe blood toxicity. IBI343 may change concentration and tumour delivery, so that failure does not invalidate the carrier. It does prove that payload identity is not enough. Conclusion. IBI343 is the strongest current antibody-delivered PDAC signal because response tracks a stated marker boundary. It remains Tier C until a randomized overall-survival test shows that the carrier adds useful life.", "sourceIds": [ "yu-2025-ibi343-pdac", "clinicaltrials-nct05458219-ibi343", "abou-alfa-2006-exatecan-pdac", "clinicaltrials-nct07066098-ibi343-phase3" ], "links": [ { "target": "treatment", "relation": "part-of" }, { "target": "treatment-modality-ranking", "relation": "ranked-by" }, { "target": "failure-antibody-target-and-payload-transfer-pdac", "relation": "bounded-by" }, { "target": "unknown-cldn182-adc-survival-effect", "relation": "leaves-open" }, { "target": "trial-cldn182-antibody-action-map", "relation": "mapped-by" }, { "target": "hypothesis-cldn182-adc-survival-test", "relation": "supports" }, { "target": "company-innovent-biologics", "relation": "developed-by" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "not-applicable", "randomised": "none", "independentReplication": "none", "sampleSize": "IBI343 PDAC treated n=83; 6 mg/kg high-marker n=44 and lower-marker n=12; free-payload randomized control n=349", "effect": "IBI343 confirmed ORR 22.7% versus 0%, PFS 5.4 versus 1.4 months and OS 8.5 versus 6.2 months across non-randomized marker groups", "limits": [ "Phase 1", "No concurrent control", "Marker groups differ in other ways", "Ad hoc group comparison", "Expression rules differ between entry and analysis", "Survival confounding by prior treatment", "Company-controlled data" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The carrier links a measurable surface address to an established DNA-damaging payload, and response differs across marker levels." }, "animalEvidence": { "rating": "moderate", "reason": "Models support targeted delivery, but the human response contrast now matters more." }, "inVitroEvidence": { "rating": "strong", "reason": "Binding, internalisation, payload release and DNA damage can be measured directly." }, "reproducibility": { "rating": "weak", "reason": "The signal comes from one uncontrolled phase 1 programme; a randomized PDAC result has not reported." }, "commercialConflicts": { "rating": "high", "reason": "Innovent controls the phase 1 and Greater China phase 3 programme; Takeda controls rights outside Greater China." }, "recency": { "rating": "current", "reason": "Uses the 2025 PDAC analysis and active phase 3 registry." } } }, "layout": { "x": 218, "y": 136 }, "updated": "2026-09-15" }, { "id": "claim-colon-ctdna-has-clinical-utility-proof", "title": "Stage II colon cancer has randomized proof for one ctDNA-guided treatment-omission rule", "shortTitle": "Colon ctDNA utility", "type": "claim", "status": "supported", "scope": [ "colorectal" ], "summary": "DYNAMIC cut adjuvant chemotherapy use from 28% to 15% while meeting its two-year recurrence-free-survival noninferiority test; that result does not validate other stages or actions.", "content": "DYNAMIC randomized 455 stage II colon-cancer patients to ctDNA-guided or clinicopathology-guided adjuvant treatment. Chemotherapy use fell from 28% to 15%, an absolute reduction of 13 percentage points, while two-year recurrence-free survival was noninferior at 93.5% versus 92.4%. This is clinical utility for one rule: a measured reduction in treatment burden with preserved outcome, not just risk association. The noninferiority margin was 8.5 percentage points and the low-risk stage II setting differs sharply from PDAC. CIRCULATE-Japan's 2,240-patient observational cohort strengthens prognosis: postoperative positivity had disease-free-survival hazard ratio 11.99, and sustained clearance during chemotherapy tracked better survival. DYNAMIC-III and ALTAIR subsequently failed different treatment rules, while FIND increased earlier curative-intent treatment without mature survival. Therefore this node must not be read as assay-wide utility. Transfer to PDAC requires a sufficiently sensitive assay and an action that improves a patient endpoint in the tested state.", "sourceIds": [ "tie-2022-dynamic-colon", "nakamura-2024-galaxy-ctdna", "tie-2025-dynamic3", "bando-2026-altair", "mo-2026-find-crc-surveillance" ], "links": [ { "target": "colorectal", "relation": "part-of" }, { "target": "unknown-ctdna-actionability", "relation": "may-transfer-to" }, { "target": "residual-disease", "relation": "supports" }, { "target": "transfer", "relation": "supports" }, { "target": "claim-crc-ctdna-utility-is-action-specific", "relation": "qualified-by" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "strong", "randomised": "strong", "independentReplication": "strong", "sampleSize": "455 randomized in DYNAMIC; 2,240 observed in GALAXY update; later action trials summarized in the linked utility record", "effect": "Chemotherapy 15% versus 28%; two-year RFS 93.5% versus 92.4% and noninferior; observational MRD DFS HR 11.99", "limits": [ "DYNAMIC tested one de-escalation rule, not positive-result escalation", "Stage II colon setting", "Noninferiority margin 8.5 percentage points", "GALAXY treatment analyses are observational", "Later randomized utility results are mixed", "Commercial assay interests" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Tumour-derived DNA after surgery tracks residual tumour burden and can support a treatment decision." }, "animalEvidence": { "rating": "not-applicable", "reason": "The utility claim is based on randomized human treatment allocation." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "The reported clinical utility does not depend on cell or organoid work." }, "reproducibility": { "rating": "strong", "reason": "Randomized utility and a large separate observational cohort support distinct parts of this narrow claim; later trials bound its transfer." }, "commercialConflicts": { "rating": "moderate", "reason": "Commercial assay interests are present and transfer to PDAC could create product incentives." }, "recency": { "rating": "current", "reason": "The 2022 result is interpreted beside action trials reported through 2026." } } }, "layout": { "x": 95, "y": 56 }, "updated": "2026-09-14" }, { "id": "claim-colorectal-screening-prevents-cancer-and-death", "title": "Colorectal screening can prevent cancer as well as detect it earlier", "shortTitle": "CRC prevention proof", "type": "claim", "status": "supported", "scope": [ "colorectal" ], "summary": "Repeated stool testing and one-time sigmoidoscopy reduced incidence and death; a one-colonoscopy invitation reduced incidence but has not shown a mortality benefit.", "content": "In the 46,551-person Minnesota trial, repeated faecal occult-blood testing followed by diagnostic colonoscopy reduced colorectal-cancer incidence by 20% with annual and 17% with biennial testing over 18 years. In the 170,432-person UK flexible-sigmoidoscopy trial, one invitation at ages 55 to 64 reduced incidence by 23% and colorectal mortality by 31% over median 11.2 years. The number invited and screened to prevent one colorectal death was 489. NordICC tests a one-colonoscopy invitation rather than these routes. At 13 years it reduced incidence from 1.80% to 1.46%, but colorectal-cancer mortality was 0.47% versus 0.41%, RR 0.88, 95% CI 0.68 to 1.08. Only 42.0% of invitees attended. Screening can therefore prevent cancer and death, but each invitation, test and follow-up route needs its own effect estimate. This success depends on properties PDAC lacks: a large accessible surface, a common removable adenoma precursor, and a tolerable confirmatory procedure. The transferable design is a risk-to-test-to-removal chain with a mortality endpoint; the colon tools do not transfer directly to microscopic PanIN or deep pancreatic tissue.", "sourceIds": [ "mandel-2000-fobt-incidence", "atkin-2010-flexible-sigmoidoscopy", "bretthauer-2022-nordicc-10y", "kaminski-2026-nordicc-13y" ], "links": [ { "target": "colorectal", "relation": "part-of" }, { "target": "detection", "relation": "supports" }, { "target": "transfer", "relation": "supports" }, { "target": "claim-general-population-screening-2019", "relation": "contradicts" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong", "independentReplication": "strong", "sampleSize": "46,551 in Minnesota trial; 170,432 in UK sigmoidoscopy trial", "effect": "Incidence reduced 17% to 23%; CRC mortality reduced 31% in UK trial; number screened per death prevented 489", "limits": [ "Older stool test methods", "Sigmoidoscopy mainly reaches distal colon", "Willingness-based entry in UK trial", "Only 42.0% accepted NordICC colonoscopy", "NordICC mortality confidence interval includes no benefit", "Long follow-up needed", "Pancreatic precursor access is fundamentally different" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Removing adenomas and serrated precursors can interrupt progression before invasive cancer." }, "animalEvidence": { "rating": "not-applicable", "reason": "The claim is based on randomized population outcomes." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Laboratory evidence cannot establish invitation, attendance or mortality effects." }, "reproducibility": { "rating": "strong", "reason": "Several randomized routes reduce incidence, though mortality and anatomical reach differ by route." }, "commercialConflicts": { "rating": "low", "reason": "The main trials were publicly or charitably supported; procedure-company relationships remain disclosed in NordICC." }, "recency": { "rating": "current", "reason": "The record includes the 2026 NordICC 13-year update." } } }, "layout": { "x": 92, "y": 13 }, "updated": "2026-09-14" }, { "id": "claim-commercial-capital-ledger-boundaries", "title": "Commercial cancer capital must remain split by financing, deal cash, contingent value, acquisition price and R&D", "shortTitle": "Commercial capital ledger", "type": "claim", "status": "supported", "scope": [ "pdac", "colorectal", "hepatobiliary", "pan-cancer" ], "summary": "Primary filings expose large but incompatible commercial amounts; they show scale and incentives, not a disease-specific global investment total.", "content": "Observed recent venture financing: PAQ reported $77m for its Series B and extension, Valar reported a $22m Series A, and Mainz reported a $6m placement. Those disclosed events total $105m, but they cover different dates, technologies and intended uses and omit earlier or undisclosed rounds. PAQ's first $39m named six investors; the later extension did not disclose allocations. Historical origin remains separate: Third Rock launched Revolution Medicines with $45m in 2015, about 11 years before PDAC approval, but the original platform differed and the money cannot be assigned to daraxonrasib. Cosomil shows another boundary: NEDO lists a ¥499m public grant, while two private rounds name investors but disclose no amounts. Yen public support cannot be added to dollar venture financing. Observed licensing cash: Jacobio reported $100m upfront from AstraZeneca for JAB-23E73. A further $1.915bn is contingent on development, regulatory and commercial milestones and is not deployed capital. Observed acquisition price: Illumina agreed to pay $7.1bn for GRAIL. That price bought ownership expectations; it is not R&D expenditure and later impairment is not research cash.\n\nObserved company R&D uses yet another boundary. Natera reported $320.678m, $404.138m and $624.110m of company-wide R&D in 2023, 2024 and 2025, totalling $1.348926bn. The filing does not allocate that total among women's health, organ health, oncology, Signatera, colorectal cancer or pancreatic cancer. Revolution Medicines reported $738.9m of company-wide R&D in the first half of 2026, including $227.8m identified as third-party daraxonrasib cost. GRAIL reported $95.45m of R&D in the same half-year. These figures show that clinical platforms require hundreds of millions, but they cannot be added into a ten-year disease total without double counting financing later spent as R&D.\n\nThe atlas therefore reports separate columns for venture financing, public startup grants, non-contingent deal cash, contingent headline value, acquisition consideration and reported R&D expense. It adds only amounts inside a single column with the same time, currency and company boundary. Capital efficiency is then compared with the strongest output produced—randomized survival, a validated decision rule, a decisive negative result or reusable trial infrastructure—not with market value or press-release milestones.", "sourceIds": [ "paq-2025-series-b-investors", "paq-2026-seriesb-pt0511", "valar-2024-series-a", "mainz-2026-private-placement", "third-rock-2015-revolution-medicines-launch", "nedo-2024-cosomil-profile", "cosomil-2026-series-a1", "jacobio-2026-annual-results", "ftc-2024-illumina-grail", "natera-2025-10k", "revolution-medicines-2026-q2-10q", "grail-2026-q2-10q" ], "links": [ { "target": "investment", "relation": "part-of" }, { "target": "claim-current-capital-frontier-2026", "relation": "extends" }, { "target": "company-paq-therapeutics", "relation": "contains" }, { "target": "company-valar-labs", "relation": "contains" }, { "target": "company-mainz-biomed", "relation": "contains" }, { "target": "company-jacobio", "relation": "contains" }, { "target": "company-grail", "relation": "contains" }, { "target": "company-natera", "relation": "contains" }, { "target": "company-revolution-medicines", "relation": "contains" }, { "target": "investor-third-rock-revolution-medicines", "relation": "contains" }, { "target": "investor-paq-series-b-syndicate", "relation": "contains" }, { "target": "investor-valar-series-a-syndicate", "relation": "contains" }, { "target": "investor-cosomil-public-private-capital", "relation": "contains" } ], "evidence": { "humanProspective": "not-applicable", "humanRetrospective": "not-applicable", "randomised": "not-applicable", "independentReplication": "moderate", "sampleSize": "Three recent disclosed financing events, one historical founding round, one public startup grant, two undisclosed private rounds, one licensing agreement, one acquisition record and three company R&D disclosures", "effect": "$105m recent disclosed financing events; separate historical $45m launch round; separate ¥499m public grant; $100m non-contingent licensing cash; $1.915bn excluded contingent value; $7.1bn acquisition price kept separate", "limits": [ "Selected events, not a market census", "Private rounds and deal terms are often undisclosed", "Historical founding capital cannot be traced to the later asset", "Company R&D is not reported by disease", "Financing may later be spent as R&D", "Acquisition price embeds expectations and control value", "No cross-currency global sum" ] }, "layout": { "x": 101, "y": 106 }, "updated": "2026-09-14" }, { "id": "claim-cordis-title-visible-target-cancer-capital", "title": "CORDIS exposes EUR 312.720m of title-visible target-cancer commitments since 2015", "shortTitle": "EU title-visible floor", "type": "claim", "status": "supported", "scope": [ "pancreatic-all", "pdac", "pnet", "cholangiocarcinoma", "gallbladder", "hcc", "colorectal", "pan-cancer" ], "summary": "A strict query of official Horizon 2020 and Horizon Europe project titles finds 159 unique projects starting from 2015, with maximum EU commitments totalling EUR 312,719,948.49.", "content": "This is a lower-bound project ledger, not a complete estimate of European research spending. The 6 August 2026 CORDIS bulk files contain public project records and maximum EU contribution. A fixed title rule searched pancreatic cancer and pancreatic neuroendocrine disease, cholangiocarcinoma, biliary-tract and gallbladder cancer, hepatocellular carcinoma and liver cancer, colorectal cancer, and colorectal liver metastasis. Restricting start dates to 1 January 2015 through 14 September 2026 produced 94 Horizon 2020 matches worth EUR 125,968,482.13 and 65 Horizon Europe matches worth EUR 186,751,466.36. The non-overlapping programme files therefore give 159 projects and EUR 312,719,948.49. Disease-term counts overlap when one title names two cancers, so only unique projects enter the total. Sixty-two titles named pancreatic cancer, 67 colorectal cancer, 12 hepatocellular carcinoma, 12 liver cancer, five cholangiocarcinoma, two colorectal liver metastasis, one gallbladder cancer and one biliary-tract cancer. The query misses target biology, shared facilities and broad cancer projects whose titles omit those disease words. It also assigns the whole commitment to the ledger when a title names pancreatic plus another cancer. Maximum EU contribution is a commitment, not annual or realised expenditure. The number can therefore show the order of magnitude and the visible programme mix, but it cannot be added to US or UK totals or treated as PDAC-only spend.", "sourceIds": [ "cordis-h2020-target-cancer-title-query-2026", "cordis-horizon-target-cancer-title-query-2026" ], "links": [ { "target": "investment", "relation": "part-of" }, { "target": "claim-observed-ten-year-research-capital", "relation": "extends" }, { "target": "claim-ukri-title-visible-pancreatic-capital", "relation": "complements" }, { "target": "funding", "relation": "supports" } ], "evidence": { "humanProspective": "not-applicable", "humanRetrospective": "not-applicable", "randomised": "not-applicable", "independentReplication": "moderate", "sampleSize": "159 unique matched projects: 94 Horizon 2020 and 65 Horizon Europe", "effect": "EUR 312,719,948.49 in maximum EU commitments for strict title-visible target-cancer projects starting from 2015", "limits": [ "Title rule omits relevant target-led and shared-infrastructure grants", "Full values of multi-cancer projects are included", "Maximum EU contribution is commitment rather than expenditure", "The source snapshot changes as CORDIS is updated", "Disease-term category counts overlap", "Not PDAC-only and not a global total" ] }, "layout": { "x": 95, "y": 107 }, "updated": "2026-09-14" }, { "id": "claim-crc-classification-produces-different-treatment-routes", "title": "Colorectal molecular subgroups require different treatments, not one precision-oncology label", "shortTitle": "CRC treatment routes", "type": "claim", "status": "supported", "scope": [ "colorectal", "pan-cancer", "transfer-unproven" ], "summary": "MSI-high disease produced durable first-line checkpoint benefit, BRAF V600E required dual pathway blockade and HER2-positive disease showed a smaller proof-of-concept response.", "content": "Colorectal cancer demonstrates what useful biological classification looks like: a measured state selects a treatment with a different mechanism and a materially different outcome. In 307 MSI-high or mismatch-repair-deficient metastatic cancers, first-line pembrolizumab produced median overall survival of 77.5 months versus 36.7 months and median response duration of 75.4 versus 10.6 months, despite 62% effective checkpoint crossover from chemotherapy. In 665 previously treated BRAF V600E cancers, encorafenib plus cetuximab improved median survival from 5.9 to 9.3 months; adding a third pathway drug did not improve median survival. HERACLES screened 914 KRAS-wild-type cases, found 48 HER2-positive tumours and treated 27, with a 30% response but no randomized comparator. These are three different evidence levels and clinical reaches. The transfer to PDAC is the process—find a reproducible dependency, lock the assay and test the treatment interaction—not the drugs themselves. Most PDAC is mismatch-repair proficient, and a rare biomarker result cannot support an unselected therapy claim.", "sourceIds": [ "andre-2025-keynote177-five-year", "tabernero-2021-beacon-crc", "sartore-bianchi-2016-heracles" ], "links": [ { "target": "model-colorectal-causal-chain", "relation": "supports" }, { "target": "colorectal", "relation": "part-of" }, { "target": "transferability-matrix", "relation": "supports" }, { "target": "failure-checkpoint-blockade-unselected-pdac", "relation": "contrasts-with" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong", "independentReplication": "moderate", "sampleSize": "307 KEYNOTE-177, 665 BEACON and 27 treated HERACLES patients after 914 screened", "effect": "MSI-high OS 77.5 versus 36.7 months; BRAF V600E OS 9.3 versus 5.9 months; HER2 response 30% without control", "limits": [ "Each result applies to a small molecular subgroup", "KEYNOTE-177 had 62% effective crossover", "All three drug programmes had commercial sponsorship", "HERACLES was small and nonrandomized", "No direct drug transfer to PDAC" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Mismatch repair, BRAF and HER2 define different tumour dependencies and immune states." }, "animalEvidence": { "rating": "moderate", "reason": "Models support the dependencies but do not establish the clinical effect sizes." }, "inVitroEvidence": { "rating": "strong", "reason": "Target and pathway studies support each selected route." }, "reproducibility": { "rating": "moderate", "reason": "The major randomized routes are strong, while HER2 evidence here is small and nonrandomized." }, "commercialConflicts": { "rating": "high", "reason": "All three drug programmes had substantial company sponsorship or investigator relationships." }, "recency": { "rating": "recent", "reason": "The record includes five-year KEYNOTE-177 follow-up reported in 2025." } } }, "layout": { "x": 68, "y": 28 }, "updated": "2026-09-14" }, { "id": "claim-crc-colonoscopy-invitation-benefit-limited-by-participation", "title": "One colonoscopy invitation prevented some colorectal cancers but has not shown fewer deaths", "shortTitle": "Colonoscopy invitation effect", "type": "claim", "status": "supported", "scope": [ "colorectal" ], "summary": "At 13 years, NordICC reduced colorectal-cancer incidence from 1.80% to 1.46%, while mortality was 0.47% versus 0.41% with a confidence interval that included no benefit.", "content": "NordICC randomized 84,583 people aged 55 to 64 to one colonoscopy invitation or no screening. Only 11,843 of 28,220 invited people underwent colonoscopy: 42.0%. At 13 years, colorectal cancer occurred in 375 of 28,217 invited people, 1.46%, and 912 of 56,366 controls, 1.80%; RR 0.81, 95% CI 0.71 to 0.90. The absolute incidence difference was 1.80% - 1.46% = 0.34 percentage points, or about 1 / 0.0034 = 294 invitations per cancer prevented over 13 years. Colorectal-cancer death was 0.41% versus 0.47%; RR 0.88, 95% CI 0.68 to 1.08, so the randomized comparison did not show fewer deaths. The 0.06-point observed difference would equal about 1,667 invitations per death, but that number is not a proven benefit because the confidence interval crosses 1. The large per-protocol estimates are vulnerable to differences between people who accepted and declined screening. The mission implication is direct: a technically effective prevention procedure still needs uptake, complete diagnostic follow-up and long mortality observation. Do not estimate population effect from the people who complied alone.", "sourceIds": [ "bretthauer-2022-nordicc-10y", "kaminski-2026-nordicc-13y" ], "links": [ { "target": "colorectal", "relation": "part-of" }, { "target": "claim-colorectal-screening-prevents-cancer-and-death", "relation": "qualifies" }, { "target": "prevention-ranking", "relation": "informs" }, { "target": "trial-crc-action-map", "relation": "informs" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong", "independentReplication": "moderate across endoscopy methods; one invitation trial for this exact rule", "sampleSize": "84,583 randomized; 28,217 invited and 56,366 controls in the 13-year analysis", "effect": "CRC incidence RR 0.81, 95% CI 0.71-0.90; CRC mortality RR 0.88, 95% CI 0.68-1.08; 42.0% attended", "limits": [ "Invitation effect is diluted by 58% nonattendance", "One colonoscopy at ages 55-64", "Mortality lower than assumed in trial design", "Per-protocol estimates are not randomized", "Results differ by proximal versus distal site and by sex", "Published erratum must be read with the report" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Colonoscopy can find and remove mucosal precursors before invasion." }, "animalEvidence": { "rating": "not-applicable", "reason": "Population prevention benefit is established or rejected in people." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Laboratory results cannot measure uptake, complications, incidence or death." }, "reproducibility": { "rating": "moderate", "reason": "Other endoscopy trials support prevention, but NordICC is the main randomized test of a one-colonoscopy invitation." }, "commercialConflicts": { "rating": "low", "reason": "Public and cancer-charity funders supported the trial; some investigators disclosed endoscopy-company relationships." }, "recency": { "rating": "current", "reason": "The 13-year randomized update was published in 2026." } } }, "updated": "2026-09-14" }, { "id": "claim-crc-ctdna-utility-is-action-specific", "title": "Colorectal ctDNA utility belongs to the action, not to the assay alone", "shortTitle": "ctDNA action-specific utility", "type": "claim", "status": "supported", "scope": [ "colorectal", "transfer-unproven" ], "summary": "Randomized trials support one stage II de-escalation rule and an earlier-imaging route, but reject stage III chemotherapy rules and trifluridine/tipiracil at molecular relapse.", "content": "Four randomized comparisons now prevent a blanket statement that colorectal ctDNA either works or fails. DYNAMIC stage II reduced chemotherapy from 28% to 15% while meeting its two-year recurrence-free-survival noninferiority test: a low-risk de-escalation rule worked. DYNAMIC-III reduced oxaliplatin and hospital admission in blood-negative stage III disease but failed noninferiority, while more chemotherapy in blood-positive disease did not improve recurrence-free survival. ALTAIR treated blood-positive, scan-negative molecular relapse with trifluridine/tipiracil and missed its disease-free-survival endpoint while causing much more severe blood toxicity. FIND used a positive methylation ctDNA result to trigger immediate CT: among 584 eligible patients, recurrence rates were similar, but 48.1% versus 23.6% of recurrent patients received curative-intent treatment, RR 2.03, P=0.008. Liver or lung-confined recurrence resection was 42.3% versus 18.2%. FIND shows an action gain at median 23.3 months, not yet a survival gain. Therefore the unit of evidence is assay + population + threshold + action + comparator + patient endpoint. Prognostic separation can coexist with failed treatment. Earlier surgery can rise without yet proving longer life.", "sourceIds": [ "tie-2022-dynamic-colon", "tie-2025-dynamic3", "bando-2026-altair", "mo-2026-find-crc-surveillance" ], "links": [ { "target": "colorectal", "relation": "part-of" }, { "target": "claim-colon-ctdna-has-clinical-utility-proof", "relation": "qualifies" }, { "target": "failure-dynamic3-ctdna-guided-chemo-no-rfs-gain", "relation": "depends-on" }, { "target": "failure-altair-mrd-treatment-no-dfs-gain", "relation": "depends-on" }, { "target": "unknown-crc-ctdna-positive-action", "relation": "leads-to" }, { "target": "transfer-colorectal-mrd-action", "relation": "qualifies" }, { "target": "trial-crc-action-map", "relation": "informs" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong", "independentReplication": "strong for prognosis; mixed and action-dependent for utility", "sampleSize": "DYNAMIC n=455; DYNAMIC-III n=968 evaluable; ALTAIR n=243; FIND n=584 eligible", "effect": "One de-escalation success, two treatment-rule failures and one earlier-curative-treatment signal without mature survival", "limits": [ "Different stages and assays", "Different action rules", "FIND survival is immature", "DYNAMIC stage II used a permissive noninferiority margin", "Commercial assay interests in several studies", "No pooled effect is valid" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Residual tumour DNA can precede imaging, while treatment sensitivity depends on the residual clone and chosen action." }, "animalEvidence": { "rating": "not-applicable", "reason": "Clinical utility is a human decision effect." }, "inVitroEvidence": { "rating": "weak", "reason": "Analytical performance cannot establish benefit from a care rule." }, "reproducibility": { "rating": "strong", "reason": "Four randomized studies test distinct action rules and repeatedly separate prognosis from utility." }, "commercialConflicts": { "rating": "high", "reason": "Commercial assay involvement is material in the platform and ALTAIR records." }, "recency": { "rating": "current", "reason": "The record includes published 2025 and 2026 randomized results." } } }, "updated": "2026-09-14" }, { "id": "claim-crc-liver-ablation-selected-survival", "title": "Aggressive local treatment extended survival in selected unresectable colorectal liver metastases", "shortTitle": "CLOCC local survival", "type": "claim", "status": "supported", "scope": [ "colorectal", "transfer-unproven" ], "summary": "CLOCC randomized 119 people with fewer than ten liver metastases and no extrahepatic disease; ablation with or without resection plus systemic therapy improved long-term overall survival.", "content": "CLOCC tested whether complete treatment of visible liver disease could matter when colorectal metastases were unresectable but still confined to one organ. The 119 patients had fewer than ten liver metastases and no extrahepatic disease. At median follow-up of 9.7 years, systemic treatment plus radiofrequency ablation with or without resection reduced the death hazard versus systemic treatment alone, HR 0.58, 95% CI 0.38 to 0.88. Eight-year overall survival was 35.9% versus 8.9%; median survival differed less, 45.6 versus 40.5 months, because the gain sat in the long-survival tail. This supports an organ-control route in a selected subgroup and complements the larger TransMet effect. It does not show that ablation cures unselected metastatic colorectal cancer or that the result transfers to PDAC. The study began as phase 3 but closed as a small phase 2, used older systemic treatment, and was not built around a modern molecular or ctDNA selector. The next gain is more likely to come from identifying truly organ-confined biology than ablating increasingly numerous lesions.", "sourceIds": [ "ruers-2017-clocc-crc-liver-ablation", "adam-2024-transmet" ], "links": [ { "target": "colorectal-liver-metastasis", "relation": "part-of" }, { "target": "claim-transmet-selected-liver-only-crc", "relation": "replicates-principle" }, { "target": "claim-extend-oligometastatic-local-therapy", "relation": "may-transfer-to" }, { "target": "failure-crc-radioembolization-pfs-without-os", "relation": "contrasts-with" }, { "target": "model-colorectal-causal-chain", "relation": "supports" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong", "independentReplication": "moderate", "sampleSize": "119 randomized patients; fewer than ten liver metastases and no extrahepatic disease", "effect": "OS HR 0.58; 8-year OS 35.9% versus 8.9%; median OS 45.6 versus 40.5 months", "limits": [ "Small phase 2", "Older chemotherapy", "Highly selected organ-confined population", "Local treatment mixed ablation and resection", "Not powered originally for a conventional comparative OS endpoint" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Complete treatment of visible disease can create a survival tail if occult systemic disease is sufficiently limited." }, "animalEvidence": { "rating": "not-applicable", "reason": "The clinical claim rests on randomized human survival." }, "inVitroEvidence": { "rating": "weak", "reason": "Cell response cannot identify organ confinement." }, "reproducibility": { "rating": "moderate", "reason": "TransMet supports the organ-confined principle using a different local method and a narrower selector." }, "commercialConflicts": { "rating": "low", "reason": "The claim is not built around a proprietary systemic product, although local devices and procedures have incentives." }, "recency": { "rating": "mixed", "reason": "The long-term CLOCC report is from 2017 and is interpreted with the 2024 TransMet result." } } }, "layout": { "x": 130, "y": 48 }, "updated": "2026-09-14" }, { "id": "claim-cryoablation-pain-signal-not-survival-proof", "title": "Pancreatic cryoablation has a useful pain signal but no reliable survival proof", "shortTitle": "Cryo pain signal, no survival proof", "type": "claim", "status": "supported", "scope": [ "pdac", "unresectable", "pain", "local-treatment" ], "summary": "Freezing a pancreatic tumour may reduce pain and opioid need in selected people; uncontrolled imaging, immune and survival changes do not establish cancer control.", "content": "Cryoablation places needles into or beside a tumour and cycles extreme cold and thaw. Ice crystals, loss of blood supply and membrane damage kill tissue; the expanding ice ball is visible on computed tomography. In a 2026 retrospective series of 11 people with painful unresectable pancreatic ductal adenocarcinoma, mean pain on a 0-to-10 scale fell from 6.72 to 3.45 at one month, a 3.27-point change. All 11 had documented opioid reduction and 5/11 stopped opioids. This is a strong signal worth testing. It is not a treatment effect estimate: there was no comparator, opioid dose was recorded only as categories, ongoing cancer care varied and later pain scores excluded people who died. Nine of 11 had metastatic disease, so local imaging cannot represent total cancer control. A separate open-surgery series of 24 reported median survival of 16.8 versus 11.4 months in 24 selected chemotherapy controls, P=0.021, but treatment was not assigned. An older 118-person bypass comparison found smaller tumours and lower CA19-9, a blood marker shed by some pancreatic cancers, without a significant prognosis difference. Cryoablation is a plausible palliative treatment. A cure or survival claim is not supported.", "sourceIds": [ "pusceddu-2026-percutaneous-cryoablation", "kang-2025-laparotomic-cryoablation", "song-2014-cryo-bypass-comparison", "clinicaltrials-nct07306286-cryo-immunotherapy" ], "links": [ { "target": "failure-cryoablation-selection-attrition-and-route-harm", "relation": "bounded-by" }, { "target": "unknown-cryoablation-randomized-pain-and-net-benefit", "relation": "leaves-open" }, { "target": "hypothesis-cryoablation-pragmatic-pain-trial", "relation": "motivates" }, { "target": "trial-pdac-cryoablation-action-map", "relation": "mapped-by" }, { "target": "claim-hifu-pain-signal-without-survival-proof", "relation": "compares-with" }, { "target": "treatment-modality-ranking", "relation": "updates" }, { "target": "treatment", "relation": "part-of" } ], "evidence": { "humanProspective": "weak", "humanRetrospective": "moderate", "randomised": "none", "independentReplication": "weak", "sampleSize": "Percutaneous pain series n=11; open cryo n=24 plus 24 controls; older bypass comparison total n=118", "effect": "Mean pain change -3.27/10 at one month in an uncontrolled selected series; survival unproved", "limits": [ "No randomized comparison", "Selected anatomy and fitness", "Qualitative opioid measurement", "Survivor-only later pain", "Mixed systemic treatment", "Open and percutaneous routes differ", "Imaging and markers are not survival" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Freeze-thaw injury kills tissue and can reduce local pressure and nerve irritation." }, "animalEvidence": { "rating": "moderate", "reason": "Cryogenic injury and ice-ball geometry are established, but patient palliation is the key question." }, "inVitroEvidence": { "rating": "strong", "reason": "Ice formation, membrane injury and cell death are direct measurable effects." }, "reproducibility": { "rating": "weak", "reason": "Pain reduction repeats in small series without a controlled estimate." }, "commercialConflicts": { "rating": "moderate", "reason": "Commercial needles are used, but the two recent reports declare no author conflict." }, "recency": { "rating": "current", "reason": "The record includes May 2026 pain data and a recruiting 2026 combination study." } } }, "layout": { "x": 844, "y": 768 }, "updated": "2026-09-15" }, { "id": "claim-current-capital-frontier-2026", "title": "The 2026 capital frontier is broad RAS control and risk-enriched detection, but money is still ahead of proof", "shortTitle": "2026 capital frontier", "type": "claim", "status": "supported", "scope": [ "pdac", "colorectal", "pan-cancer", "transfer-unproven" ], "summary": "At least $193m of disclosed upfront or venture capital is attached to new RAS and detection programmes, while a further $1.915bn Jacobio deal value is contingent.", "content": "Disclosed recent examples are not a market total. PAQ reported $77m in Series B financing for KRAS degraders. Jacobio reported $100m upfront from AstraZeneca for JAB-23E73 and up to $1.915bn more in contingent value; only the upfront belongs in observed capital. Mainz raised $6m mainly for runway during a pancreatic-detection pivot. Adenocyte's lead investor reported a $10m first close for ultrasound-assisted pancreatic-fluid sampling. Cosomil raised an undisclosed Series A1 for screening development. Disclosed non-contingent amounts therefore sum to at least $193m = $77m + $100m + $6m + $10m, excluding the undisclosed Cosomil round and company internal R&D. The pattern says sophisticated capital expects broad RAS depth, degradation, resistance and risk-enriched detection to be fundable in the next two to four years. It does not say those approaches work. Crowded: KRAS G12D and broad-RAS molecules. Becoming crowded: degraders and first-line combinations. Underexplored: prospective resistance routing, MRD action and workflow-level detection harm. Scientifically attractive but commercially hard: dangerous-precursor classification and long prevention trials. Commercially attractive but biologically weak: broad screening sales before utility. Newly possible: randomized common-RAS residual-disease trials and national real-workflow biomarker pilots. The best investment is the measurement and controlled-decision layer shared across competing drugs, not another undifferentiated molecule.", "sourceIds": [ "paq-2026-seriesb-pt0511", "jacobio-2026-annual-results", "mainz-2026-private-placement", "cosomil-2026-series-a1", "igan-2026-adenocyte-financing" ], "links": [ { "target": "investment", "relation": "part-of" }, { "target": "company-paq-therapeutics", "relation": "contains" }, { "target": "company-jacobio", "relation": "contains" }, { "target": "company-mainz-biomed", "relation": "contains" }, { "target": "company-cosomil", "relation": "contains" }, { "target": "company-adenocyte", "relation": "contains" }, { "target": "recommended-strategy", "relation": "supports" } ], "evidence": { "humanProspective": "not-applicable", "humanRetrospective": "not-applicable", "randomised": "not-applicable", "independentReplication": "moderate", "sampleSize": "Five financing or licensing events", "effect": "At least $193m disclosed non-contingent capital; $1.915bn additional contingent deal value excluded", "limits": [ "Not a market-wide total", "Company announcements", "Internal R&D excluded", "Cosomil amount undisclosed", "Deal terms and timing may differ", "Capital does not establish science" ] }, "layout": { "x": 96, "y": 105 }, "updated": "2026-09-14" }, { "id": "claim-cyst-classifier-current-disease-not-future-risk", "title": "A cyst test that detects current advanced disease is not yet a forecast of future progression", "shortTitle": "Diagnosis is not prognosis", "type": "claim", "status": "supported-with-limits", "scope": [ "pdac" ], "summary": "PancreaSeq GC detects current HGD or PDAC well in a highly enriched sampled cohort, but no cited study shows that it predicts later progression in routine low-risk surveillance.", "content": "PancreaSeq GC is a real advance in the decision it studied. In a blinded prospective validation of 241 people with diagnostic follow-up, including 186 mucinous cysts and 97 cases of advanced neoplasia, it detected current HGD or PDAC with 86.6% sensitivity and 97.9% specificity. The cohort prevalence was 97/241 = 40.2% overall and 97/186 = 52.2% among mucinous cysts, far above routine low-risk surveillance. A transport calculation shows why that matters. If the same 86.6% sensitivity and 97.9% specificity held at a 1.90% event prevalence, 100,000 people would produce about 1,645 true positives and 2,060 false positives, PPV 44.4% and 1.25 false positives per true positive. That is a model, not an observed result, and performance may fall when disease spectrum, sampling and verification change. More importantly, the study classified disease already present at sampling; it did not follow low-risk marker-positive and marker-negative cysts to test future HGD or invasion. GNAS and KRAS remain useful lineage markers, not progression clocks. Required next test: freeze the classifier, enroll consecutive surveillance patients before EUS selection, retain all negative follow-up, separate current advanced disease from later progression, and report complications and surgery yield. Commercial conflicts matter because investigators disclosed assay intellectual property, royalties and related consulting; they do not erase the result, but they raise the bar for independent validation.", "sourceIds": [ "singhi-2026-pancreaseq-gc", "wu-2011-gnas-ipmn", "ohtsuka-2024-bd-ipmn-surveillance", "mirzaian-2026-low-risk-cysts" ], "links": [ { "target": "claim-ipmn-mutations-identify-lineage-not-progression", "relation": "extends" }, { "target": "unknown-dangerous-precursor-selection", "relation": "constrains" }, { "target": "hypothesis-dynamic-cyst-interception-rule", "relation": "constrains" }, { "target": "detection", "relation": "supports" } ], "evidence": { "humanProspective": "moderate — blinded multi-institution diagnostic validation", "humanRetrospective": "moderate — reference pathology and diagnostic follow-up", "randomised": "not-applicable", "independentReplication": "weak for the exact classifier and future-risk use", "sampleSize": "n=241; 186 mucinous cysts; 97 advanced-neoplasia cases", "effect": "Current advanced-neoplasia sensitivity 86.6%, specificity 97.9%, AUC 0.923; no observed future-progression performance", "limits": [ "Advanced-neoplasia prevalence 40.2%", "EUS-FNA selected population", "Diagnostic rather than future-progression endpoint", "Spectrum and verification bias can affect transport", "Assay intellectual-property and royalty disclosures" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Late tumour-suppressor and expression changes can identify advanced neoplasia, but their timing and persistence before progression are not established." }, "animalEvidence": { "rating": "weak", "reason": "Model systems support molecular progression but do not validate the human clinical forecast." }, "inVitroEvidence": { "rating": "moderate", "reason": "The panel measures DNA and RNA linked to cyst type and advanced disease, but laboratory separation is not clinical utility." }, "reproducibility": { "rating": "weak", "reason": "The exact classifier needs independent, consecutive low-risk validation with complete negative follow-up." }, "commercialConflicts": { "rating": "high", "reason": "Authors disclosed intellectual property, royalties and consulting tied to cyst testing, requiring independent analysis and replication." }, "recency": { "rating": "current", "reason": "The validation was published online in December 2025 and in print in 2026." } } }, "layout": { "x": 50, "y": 39 }, "updated": "2026-09-14" }, { "id": "claim-daraxonrasib-common-ras-survival-breakthrough", "title": "Multi-RAS inhibition doubled median survival against chemotherapy in previously treated metastatic PDAC, but did not produce durable control", "shortTitle": "Common-RAS survival gain", "type": "claim", "status": "supported", "scope": [ "pdac" ], "summary": "In a 500-patient randomised phase 3 trial, daraxonrasib produced median overall survival of 13.2 months versus 6.6 months in RAS-G12 PDAC, with hazard ratio 0.40.", "content": "Daraxonrasib changes the PDAC treatment map because it inhibits the active form of several mutant and wild-type RAS proteins instead of one rare allele. In 500 patients with previously treated metastatic PDAC, 91.8% had a RAS G12 mutation. Median overall survival in that group was 13.2 months with daraxonrasib and 6.6 months with investigator-choice chemotherapy; hazard ratio 0.40. Median progression-free survival was 7.3 versus 3.5 months; hazard ratio 0.45. Grade 3 or worse adverse events occurred in 61.8% versus 69.6%, and treatment-related discontinuation in 1.2% versus 11.2%. This is strong proof that common-RAS suppression can change survival in PDAC. It is not cure or durable control: median progression still occurred within 7.3 months and median survival remained 13.2 months. The decisive next tests are first-line combination, treatment after resection or molecular residual disease, resistance sampling at progression, and whether continuous RAS pressure can be combined safely with immune or cell-death mechanisms. The trial was open-label, used several possible chemotherapy comparators, and was funded by the drug's manufacturer; those limits do not erase the large randomised survival difference but matter for toxicity, quality-of-life and subgroup interpretation.", "sourceIds": [ "oreilly-2026-daraxonrasib" ], "links": [ { "target": "treatment", "relation": "supports" }, { "target": "claim-kras-g12c-inhibition-proves-targetability-not-general-control", "relation": "succeeds-where" }, { "target": "ranking-provisional-pdac-bottlenecks", "relation": "supports" }, { "target": "model-pdac-causal-chain", "relation": "part-of" }, { "target": "experiments", "relation": "depends-on" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong", "independentReplication": "none", "sampleSize": "500 randomised patients; 248 daraxonrasib and 252 investigator-choice chemotherapy; 91.8% RAS G12", "effect": "RAS-G12 median OS 13.2 versus 6.6 months, HR 0.40; median PFS 7.3 versus 3.5 months, HR 0.45", "limits": [ "Previously treated metastatic population", "Open-label", "Investigator-choice comparator included four regimens", "Manufacturer-funded", "Company employees were authors", "No independent phase 3 replication", "Median control remained measured in months", "First-line and low-burden benefit not established" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "A large randomized survival effect confirms that common-RAS dependence is clinically actionable in this setting." }, "animalEvidence": { "rating": "not-applicable", "reason": "The claim is established by a human phase 3 trial and does not rely on animal efficacy." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "The survival claim does not rely on cell-line or organoid results." }, "reproducibility": { "rating": "weak", "reason": "The phase 3 result is large but has not yet been independently repeated." }, "commercialConflicts": { "rating": "high", "reason": "The manufacturer funded the trial and company employees were authors." }, "recency": { "rating": "current", "reason": "The decisive phase 3 report and approval are from 2026." } } }, "layout": { "x": 91, "y": 61 }, "updated": "2026-09-14" }, { "id": "claim-daraxonrasib-kras-amplification-resistance", "title": "Mutant KRAS amplification is the first recurrent human escape class after daraxonrasib, but not yet a validated treatment selector", "shortTitle": "KRAS amplification escape", "type": "claim", "status": "supported", "scope": [ "pdac" ], "summary": "Paired plasma from 44 treated patients found new RAS-pathway alterations in 59% at progression, led by mutant KRAS amplification in 36%, crossing the mission's 25% frequency gate.", "content": "The first paired human resistance series narrows the durable-RAS problem. Targeted sequencing of pretreatment and end-of-treatment plasma from 44 phase 1/2 daraxonrasib patients found treatment-emergent RAS-pathway changes in 26/44 = 59%. Mutant KRAS amplification was the largest repeated class at 16/44 = 36%; MAPK-pathway changes occurred in 11/44 = 25%, and RTK and PI3K-pathway changes in 4/44 = 9% each. Eleven patients acquired more than one alteration, and no secondary KRAS mutations were observed. This clears the mission's first frequency gate because 36% exceeds the required 25%. It does not clear the action gate. The series was sponsor-led, used circulating DNA rather than systematic multi-lesion tissue, and did not randomize patients to a matched combination. Models suggest several possible responses—deeper RAS blockade, G12D-selective plus broad RAS inhibition, RTK or DNA-damage combinations—but multiple coexisting changes mean amplification is not automatically the sole driver. A separate G12R case linked EGFR and wild-type RAS escape to about five months of control on a trametinib-based sixth-line combination, but n=1 cannot establish treatment effect. Independent phase 3 plasma and tissue should now test whether mutant KRAS amplification repeats near 36%, appears before clinical progression and predicts benefit from a fixed combination.", "sourceIds": [ "aronchik-2026-daraxonrasib-resistance", "dorbin-2026-allele-specific-daraxonrasib-resistance", "lin-2026-vertical-kras-inhibition", "oreilly-2026-daraxonrasib" ], "links": [ { "target": "claim-daraxonrasib-common-ras-survival-breakthrough", "relation": "explains-limit-of" }, { "target": "claim-kras-mutant-dosage-may-shape-resistance", "relation": "extends" }, { "target": "unknown-durable-ras-control", "relation": "narrows" }, { "target": "hypothesis-prospective-ras-resistance-routing", "relation": "passes-frequency-gate-for" }, { "target": "trial-daraxonrasib-expansion-programme", "relation": "must-reproduce-in" }, { "target": "company-revolution-medicines", "relation": "reported-by" } ], "evidence": { "humanProspective": "moderate — paired samples from a prospective phase 1/2 trial", "humanRetrospective": "not-applicable", "randomised": "none — resistance substudy and combinations were not randomized", "independentReplication": "none", "sampleSize": "Paired ctDNA n=44; KRAS-amplification class n=16; separate allele-routed treatment case n=1", "effect": "Emergent RAS-pathway alterations 26/44, 59%; mutant KRAS amplification 16/44, 36%; MAPK 11/44, 25%; RTK and PI3K 4/44 each, 9%", "limits": [ "Phase 1/2 subset", "Sponsor-led analysis", "Circulating DNA may miss low-shedding and spatially distinct clones", "Eleven patients had more than one new alteration", "Model combinations have no randomized human proof", "Single routed-treatment case cannot establish efficacy" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Increasing mutant KRAS dosage is a direct route to restore pathway output under a RAS inhibitor." }, "animalEvidence": { "rating": "strong", "reason": "Human and mouse PDAC models reproduced amplification and pathway reactivation and supported several combination concepts." }, "inVitroEvidence": { "rating": "strong", "reason": "Engineered and acquired-resistance systems reproduced pathway escape, although they do not select one safe clinical combination." }, "reproducibility": { "rating": "weak", "reason": "The human frequency comes from one sponsor-led cohort; the related single-patient routing result is not a repeat." }, "commercialConflicts": { "rating": "high", "reason": "Most laboratory authors worked for Revolution Medicines, held company interests and in several cases were named patent inventors." }, "recency": { "rating": "current", "reason": "The paired human resistance study and related mechanistic reports were published in 2026." } } }, "layout": { "x": 99, "y": 71 }, "updated": "2026-09-14" }, { "id": "claim-dendritic-cell-vaccine-uncontrolled", "title": "Dendritic-cell vaccination remains an uncontrolled immune-response programme", "shortTitle": "Dendritic vaccine gap", "type": "claim", "status": "open", "scope": [ "pdac" ], "summary": "A 38-patient resected-PDAC study estimated two-year recurrence-free survival at 64%, but selection and standard therapy prevent attribution.", "content": "A single-arm study gave autologous dendritic-cell immunotherapy after PDAC resection and standard treatment. Thirty-eight patients enrolled, 28 completed five vaccinations, immune responses were detected and estimated two-year recurrence-free survival was 64%. The trial met its historical threshold, but without randomization the contribution of surgery, chemotherapy, patient selection and vaccine cannot be separated. Earlier Japanese retrospective and small prospective series similarly connected immune skin-test responses with survival, which can arise because healthier immune systems both respond and live longer. The remaining step is a randomized, manufacturable adjuvant trial with a fixed antigen, timely-release rate, complete intention-to-treat accounting and recurrence or survival endpoint. Personalized neoantigen RNA and off-the-shelf KRAS vaccines currently have clearer product definitions, but they too lack efficacy proof.", "sourceIds": [ "lau-2024-dendritic-resected-pdac" ], "links": [ { "target": "claim-neoantigen-vaccine-frontier", "relation": "competes-with" }, { "target": "failure-algenpantucel-l-vaccine-phase3", "relation": "warned-by" }, { "target": "treatment-long-tail-audit", "relation": "ranked-by" }, { "target": "treatment", "relation": "part-of" } ], "evidence": { "humanProspective": "weak", "humanRetrospective": "weak", "randomised": "none", "independentReplication": "weak", "sampleSize": "38; 28 completed five vaccinations", "effect": "Estimated two-year RFS 64%; no controlled treatment effect", "limits": [ "Single arm", "Historical threshold", "Selected resected patients", "Complex personalized manufacture", "Immune response is not efficacy" ] }, "layout": { "x": 137, "y": 95 }, "updated": "2026-09-14" }, { "id": "claim-dpyd-testing-prevents-avoidable-fluoropyrimidine-harm", "title": "DPYD-guided fluoropyrimidine dosing prevents some severe harm but does not make chemotherapy safe", "shortTitle": "DPYD prevents some 5-FU harm", "type": "claim", "status": "supported", "scope": [ "pdac", "cholangiocarcinoma", "colorectal", "treatment-safety", "pharmacogenetics" ], "summary": "Testing the gene that controls much of 5-FU clearance can prevent dangerous overexposure in some patients, while a negative panel leaves most toxicity unexplained.", "content": "System model. DPYD is the gene for DPD, the main enzyme that breaks down 5-fluorouracil, or 5-FU. Reduced DPD is like a garbage collector running too slowly: the same input dose produces a larger and longer drug exposure. Severe diarrhoea, mouth injury, infection from low white cells, nerve damage and death can follow. This matters directly to FOLFIRINOX, NALIRIFOX and other pancreatic or gastrointestinal regimens containing 5-FU. Evidence. A prospective 17-hospital study found 85 four-variant carriers among 1,103 evaluable patients. Starting-dose reduction lowered severe-toxicity risk against historical full-dose carrier controls for three variants, although one variant comparison did not improve. PACIFIC-PGx later reported severe toxicity of 7% in DPYD carriers given guided dosing versus 39% in historical controls: 32 fewer events per 100 carriers, but only 19 carriers were found among 462 tested and the study was not randomized. Policy. UK and European regulators already recommend testing every patient before systemic fluoropyrimidines, withholding them in complete deficiency and reducing the initial dose in partial deficiency. NHS England reports testing about 38,000 patients. Boundary. This is an established safety action, not a cancer cure and not a licence to underdose everyone. Most tested patients do not carry the small standard panel, and severe toxicity still occurs after a negative test. The useful mission question is whether every eligible pancreatic patient receives a timely result, the dose action is correct, later dose adjustment preserves treatment, and residual failures reveal missing variants or non-genetic causes.", "sourceIds": [ "ema-2020-dpd-testing", "mhra-2020-dpd-testing", "nhs-england-dpyd-implementation", "henricks-2018-dpyd-guided-dosing", "glewis-2024-pacific-pgx", "chan-2024-dpyd-non-european-review" ], "links": [ { "target": "treatment", "relation": "part-of" }, { "target": "claim-modern-metastatic-chemotherapy-remains-short-control", "relation": "bounded-by" }, { "target": "failure-pharmacogenetic-panel-and-route-coverage", "relation": "bounded-by" }, { "target": "trial-pdac-pharmacogenetic-safety-action-map", "relation": "mapped-by" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "strong", "randomised": "weak — dose guidance is compared mainly with historical controls", "independentReplication": "moderate — multiple systems reproduce feasibility and lower carrier toxicity", "sampleSize": "Prospective n=1,181 enrolled and 1,103 evaluable; PACIFIC-PGx n=462 tested with 19 DPYD intermediate metabolizers", "effect": "PACIFIC-PGx severe toxicity 7% versus 39% historical in carriers; established regulator-backed safety action without pancreatic-specific survival proof", "limits": [ "Historical controls", "Few variant carriers", "Mixed cancers", "Four-variant coverage", "Residual toxicity after a negative result", "Reduced-dose efficacy and escalation not fully resolved" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "DPD directly clears 5-FU, so reduced enzyme activity raises drug exposure." }, "animalEvidence": { "rating": "not-applicable", "reason": "The clinical dosing action is supported by human pharmacology and toxicity." }, "inVitroEvidence": { "rating": "moderate", "reason": "Variant function supports the mechanism but cannot measure the full patient route." }, "reproducibility": { "rating": "moderate", "reason": "Prospective programmes repeat feasibility and lower carrier risk, mainly against historical controls." }, "commercialConflicts": { "rating": "low", "reason": "The decisive policy and multicentre evidence are not tied to one proprietary assay." }, "recency": { "rating": "current", "reason": "The established rule was checked against current NHS and regulator pages in September 2026." } } }, "layout": { "x": 1880, "y": 1480 }, "updated": "2026-09-15" }, { "id": "claim-early-panin-driver-mutations-not-progression-markers", "title": "Driver mutations appear in the earliest PanIN lesions and therefore do not alone mark lethal progression", "shortTitle": "Early PanIN mutations", "type": "claim", "status": "supported", "scope": [ "pdac" ], "summary": "More than 99% of studied low-grade PanIN-1 lesions carried KRAS, CDKN2A, GNAS or BRAF mutation, making mutation presence an initiation signal rather than a sufficient danger signal.", "content": "Microdissection and sequencing found a mutation in KRAS, CDKN2A, GNAS or BRAF in more than 99% of the earliest low-grade PanIN-1 lesions studied. This places genetic change very early in the common microscopic precursor route. It also creates a detection trap: finding a mutant fragment does not tell whether the lesion will remain indolent, progress slowly or has already acquired invasive and metastatic ability. The missing marker is change over time or a later biological state—clonal expansion, genomic disruption, cell-state shift, tissue invasion, immune escape or organ response—that predicts dangerous progression with enough specificity to justify intervention.", "sourceIds": [ "kanda-2012-early-panin-mutations", "wu-2011-gnas-ipmn" ], "links": [ { "target": "claim-ipmn-mutations-identify-lineage-not-progression", "relation": "supports" }, { "target": "detection", "relation": "supports" }, { "target": "unknowns", "relation": "supports" }, { "target": "pdac", "relation": "part-of" } ], "evidence": { "humanProspective": "none", "humanRetrospective": "moderate", "randomised": "not-applicable", "independentReplication": "moderate", "sampleSize": "Microdissected early low-grade PanIN specimen series", "effect": "More than 99% contained at least one of four named driver mutations", "limits": [ "Specimen study", "Does not measure population prevalence", "No longitudinal progression outcome", "Assay on tissue is not equivalent to a blood test", "Mutation abundance and combinations may add information not tested here" ] }, "layout": { "x": 11, "y": 45 }, "updated": "2026-09-14" }, { "id": "claim-electrochemotherapy-local-effect-without-patient-benefit", "title": "Electrochemotherapy can alter pancreatic tumours, but added patient benefit is unproved", "shortTitle": "ECT local effect, no benefit proof", "type": "claim", "status": "supported", "scope": [ "pdac", "locally-advanced", "resectable", "local-treatment" ], "summary": "Small single-arm studies show that bleomycin plus reversible electric pulses can be delivered to pancreas or a surgical margin; none can estimate added survival, recurrence or quality-adjusted life.", "content": "Electrochemotherapy, ECT, gives bleomycin and then sends short high-voltage pulses through electrodes around the target. The pulses temporarily open cell membranes so far more drug enters; unlike irreversible electroporation, the electric field is intended to deliver a drug rather than kill mainly by permanent membrane failure. A prospective study selected 25 people with locally advanced pancreatic cancer who had stable or responding disease after chemotherapy and then treated all with open ECT. At one month, 19/25 were called partial responses and 5/25 stable. At six months, 11/25 were partial responses and 3/25 stable, while 11/25 had no scan because they had died. Median survival from ECT was 11.5 months. Thirteen surgeon-selected fixed-electrode cases had median survival of 6 months and 12 variable-geometry cases had 12 months, but the difference was not significant, P=0.18, and was not randomly assigned. PanECT applied ECT to the back cut surface after pancreatic-head surgery. Seven preliminary cases showed delivery was possible; its completed 10-person registry set reports four post-procedure complications, one serious pancreatic fistula followed by death, and seven deaths over follow-up. There is no control group and disease-free and overall-survival tables are not posted. These studies establish technical delivery and give a direct safety warning. They do not show that ECT adds useful life, prevents recurrence or beats the same operation or chemotherapy without ECT.", "sourceIds": [ "izzo-2021-pancreatic-ect-phase12", "cebron-2023-panect-preliminary", "clinicaltrials-nct04281290-panect-results", "igea-2026-cliniporator-vitae" ], "links": [ { "target": "failure-electrochemotherapy-endpoint-and-trial-drift", "relation": "bounded-by" }, { "target": "unknown-electrochemotherapy-added-benefit-and-delivery", "relation": "leaves-open" }, { "target": "hypothesis-electrochemotherapy-randomized-data-gate", "relation": "motivates" }, { "target": "trial-pdac-electrochemotherapy-action-map", "relation": "mapped-by" }, { "target": "company-igea", "relation": "involves" }, { "target": "trial-pdac-irreversible-electroporation-action-map", "relation": "compares-with" }, { "target": "treatment-modality-ranking", "relation": "updates" }, { "target": "treatment", "relation": "part-of" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "not-applicable", "randomised": "none", "independentReplication": "weak", "sampleSize": "Locally advanced single arm n=25; surgical-margin PanECT n=10", "effect": "Technical delivery and imaging change observed; added patient benefit unknown", "limits": [ "No concurrent control", "Selection after chemotherapy response", "Eleven of 25 lacked six-month imaging because of death", "Electrode geometry was surgeon-selected", "Small samples", "Open operations add harm", "PanECT survival outcomes are not posted" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Measured electric pulses temporarily open membranes and can increase intracellular bleomycin." }, "animalEvidence": { "rating": "moderate", "reason": "Tumour drug delivery is supported outside these human outcome studies but does not settle patient benefit." }, "inVitroEvidence": { "rating": "strong", "reason": "Electroporation-driven drug uptake is a direct and repeatable physical mechanism." }, "reproducibility": { "rating": "weak", "reason": "Small human series reproduce feasibility, not added survival or recurrence benefit." }, "commercialConflicts": { "rating": "high", "reason": "The same device maker supplied technology, employed authors and helped fund the randomized programme." }, "recency": { "rating": "current", "reason": "The record includes PanECT results posted in March 2026 and the transitioned phase 2b trial." } } }, "layout": { "x": 736, "y": 660 }, "updated": "2026-09-15" }, { "id": "claim-elemental-diet-survival-signal-needs-replication", "title": "A small nutrition trial reported a large survival signal that needs independent replication", "shortTitle": "Nutrition survival signal", "type": "claim", "status": "uncertain", "scope": [ "pdac", "supportive-care", "treatment" ], "summary": "Elemental diet during chemotherapy produced median survival of 21.4 versus 12.8 months in one 60-person centre; the effect is decision-relevant but too fragile to treat as established care.", "content": "A Japanese single-centre trial randomized 60 people with advanced pancreatic cancer to usual diet or at least one packet per day of an elemental diet, a nutritionally complete formula made of readily absorbed components, during chemotherapy. Three people withdrew from the intervention. At two months, body-mass index fell by 0.3 rather than 0.6 kg/m2 and grip strength was better preserved. Median progression-free survival was 9.5 versus 6.4 months, p=0.068, while median overall survival was 21.4 versus 12.8 months, hazard ratio 0.55, 95% CI 0.31-0.98, p=0.038. The survival result is large enough to matter and small enough to be unstable. It comes from one open centre, 60 assignments, several secondary measures and an overall-survival confidence interval just below no effect. It could reflect nutrition, better chemotherapy delivery, chance, unmeasured imbalance or differences after progression. The correct action is a cheap independent multicentre replication that records every assigned patient, formula adherence, actual chemotherapy dose, pancreatic enzymes, weight, muscle, function, symptoms, progression, later treatment and survival. No specialist formula purchase or survival claim follows from this one study.", "contentSections": [ "A separate 32-person ketogenic-diet screening trial was feasible but did not confirm benefit: PFS HR 0.53, 95% CI 0.21-1.37, and OS HR 0.58, 95% CI 0.25-1.37. Nutritional ketosis occurred on a median 39.4% of days. Its wide intervals and one-sided screening tests keep the result exploratory. It supports testing host support; it does not reproduce the elemental formula or its survival estimate." ], "sourceIds": [ "uemura-2026-elemental-diet-pdac", "jameson-2026-ketogenic-diet-pdac", "chen-2025-tocilizumab-cachexia", "groarke-2024-ponsegromab-cachexia" ], "links": [ { "target": "treatment", "relation": "part-of" }, { "target": "claim-cachexia-is-parallel-lethal-system", "relation": "supports" }, { "target": "unknown-cachexia-treatment-utility", "relation": "informs" }, { "target": "trial-pdac-cachexia-action-map", "relation": "informs" }, { "target": "hypothesis-older-pdac-fast-support-route", "relation": "informs" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "not-applicable", "randomised": "moderate — randomized but one small open centre", "independentReplication": "weak", "sampleSize": "n=60 randomized; three intervention withdrawals", "effect": "Median OS 21.4 vs 12.8 months, HR 0.55, 95% CI 0.31-0.98; PFS p=0.068", "limits": [ "Single centre", "Open intervention", "Small sample", "Three intervention withdrawals", "Several secondary measures", "Survival confidence interval close to no effect", "Chemotherapy exposure and later treatment could mediate or confound the route", "Does not establish a formula-specific standard" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "Nutrition and preserved function could improve treatment delivery, but the causal component is not isolated." }, "animalEvidence": { "rating": "not-applicable", "reason": "The decision rests on a human care-route intervention." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "A formula cannot be judged from tumour-cell systems." }, "reproducibility": { "rating": "weak", "reason": "No independent matching randomized survival result was found." }, "commercialConflicts": { "rating": "low", "reason": "No product-sponsor role was reported in the primary record checked." }, "recency": { "rating": "current", "reason": "Published and checked in 2026." } } }, "updated": "2026-09-15" }, { "id": "claim-eli002-mrd-frontier", "title": "ELI-002 produced persistent KRAS immunity, but its randomized phase 2 trial missed the main endpoint", "shortTitle": "ELI-002 MRD signal", "type": "claim", "status": "weakened", "scope": [ "pdac", "colorectal" ], "summary": "Phase 1 showed strong CD4/CD8 responses, but randomized AMPLIFY-7P did not meet its prespecified intention-to-treat disease-free-survival endpoint.", "content": "AMPLIFY-201 tested an amphiphile vaccine targeting KRAS G12D and G12R in 25 people with molecular residual disease after local treatment: 20 had pancreatic cancer and five colorectal cancer. At final follow-up, 71% of evaluable patients induced both CD4 and CD8 responses and 67% showed antigen spreading. Stronger post-treatment responders had much longer relapse-free and overall survival, but everyone received vaccine and the groups were defined after treatment. AMPLIFY-7P then supplied a randomized test in 144 resected mutant-KRAS PDAC patients. It missed its prespecified intention-to-treat disease-free-survival endpoint. The sponsor highlighted a post-hoc R0-resection subgroup and immune-response associations, but neither replaces the failed main comparison. The combined evidence establishes that ELI-002 can induce immunity; it does not yet establish that the vaccine prevents recurrence. The remaining path is a prospectively defined R0 trial with balanced risk factors, adequate duration and complete effect reporting, plus a separate randomized test of any checkpoint or RAS-inhibitor combination.", "sourceIds": [ "pant-2024-eli002", "wainberg-2025-eli002-final", "elicio-2026-amplify7p-results" ], "links": [ { "target": "residual-disease", "relation": "targets" }, { "target": "failure-eli002-phase2-missed-dfs", "relation": "weakened-by" }, { "target": "transfer", "relation": "supports" }, { "target": "experiments", "relation": "tests" }, { "target": "treatment", "relation": "part-of" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong", "independentReplication": "none", "sampleSize": "Phase 1 n=25; randomized phase 2 n=144", "effect": "Phase 1 produced strong immune responses; phase 2 missed prespecified ITT DFS. Post-hoc R0 subgroup HR 0.65, P=0.048", "limits": [ "Phase 2 main estimate not publicly reported", "Full peer-reviewed phase 2 report unavailable", "Post-hoc subgroup", "Immune-response groups were defined after treatment", "Response association is not a treatment-effect estimate", "Company-sponsored programme", "Company employees were authors", "Overall survival immature" ] }, "layout": { "x": 86, "y": 91 }, "updated": "2026-09-14" }, { "id": "claim-elraglusib-randomized-survival-signal", "title": "Elraglusib has a randomized survival signal that needs clean confirmation", "shortTitle": "Elraglusib survival signal", "type": "claim", "status": "mixed", "scope": [ "pdac" ], "summary": "Adding elraglusib to first-line chemotherapy improved survival in a randomized phase 2 trial, including the all-randomized analysis, but did not improve tumour-control measures and has not been independently confirmed.", "content": "System model. GSK-3β is an enzyme that helps several cell-control circuits decide whether a cell grows, survives stress, changes state or interacts with immunity. Elraglusib blocks that enzyme. This is a wide control-node intervention, so direct survival evidence matters more than the many proposed downstream mechanisms. Randomized result. NCT03678883 compared weekly elraglusib plus gemcitabine/nab-paclitaxel with the same chemotherapy alone in untreated metastatic PDAC. In the primary treated population, 155 versus 78 people, median overall survival was 10.1 versus 7.2 months: a 2.9-month gain, HR 0.62, 95% CI 0.46-0.84, P=0.010. HR 0.62 means the observed death rate at any point was 38% lower, not that 38% were cured. The stricter all-randomized analysis, 171 versus 98, still favoured treatment: 8.9 versus 7.2 months, HR 0.68, 95% CI 0.51-0.89, P=0.022. One-year survival was 42.1% versus 22.3%. Signal boundary. Tumour-control outputs did not confirm the survival story: PFS was 5.6 versus 5.1 months, HR 0.90, P=0.624; response was 28.4% versus 21.8%, P=0.281; disease control was 40.0% versus 33.3%, P=0.323; and response duration was 5.5 versus 4.0 months, HR 0.81, P=0.984. The study was open label, selected one of two drug schedules adaptively, changed its primary analysis after starting and made receipt of treatment the primary counting rule. A separate 49-person FOLFIRINOX study was explicitly non-comparator; its elraglusib groups reached PFS 6.0-6.5 months and the authors said outcomes did not exceed historical standards. Decision. This is stronger than an uncontrolled response and deserves a high Tier B rank. It is not survival proof. Recover every randomized record and the dated analysis history, then run an independently analysed blinded phase 3 with all-randomized survival primary.", "sourceIds": [ "mahalingam-2026-elraglusib-randomized", "clinicaltrials-nct03678883-elraglusib", "pathak-2026-elraglusib-folfirinox", "clinicaltrials-nct05077800-elraglusib-folfirinox" ], "links": [ { "target": "treatment", "relation": "part-of" }, { "target": "failure-elraglusib-analysis-and-survival-discordance", "relation": "bounded-by" }, { "target": "unknown-elraglusib-survival-replication", "relation": "creates" }, { "target": "trial-elraglusib-action-map", "relation": "mapped-by" }, { "target": "hypothesis-elraglusib-survival-confirmation-gate", "relation": "supports" }, { "target": "company-actuate-therapeutics", "relation": "developed-by" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "moderate", "independentReplication": "weak", "sampleSize": "Randomized trial n=286 total; relevant weekly and control ITT n=269; primary treated set n=233; separate non-comparator study n=49", "effect": "ITT OS 8.9 versus 7.2 months, HR 0.68, 95% CI 0.51-0.89; PFS HR 0.90, 95% CI 0.68-1.20", "limits": [ "Open label", "Adaptive schedule selection", "Primary analysis changed after trial start", "Unequal pre-treatment exclusion", "No PFS or response confirmation", "No independent randomized replication", "Sponsor funded" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "GSK-3β controls several relevant cell and immune circuits, but the exact route to longer survival is unresolved." }, "animalEvidence": { "rating": "moderate", "reason": "Models support activity, while the randomized human result now sets the decision boundary." }, "inVitroEvidence": { "rating": "moderate", "reason": "Target inhibition and several downstream effects can be measured, but none explains the survival-only pattern yet." }, "reproducibility": { "rating": "weak", "reason": "The survival signal exists in one randomized phase 2; the small FOLFIRINOX study did not establish a matching treatment effect." }, "commercialConflicts": { "rating": "high", "reason": "Actuate funded the trial, employs authors and controls the product and participant data." }, "recency": { "rating": "current", "reason": "Uses the May 2026 full paper, June 2026 abstract and September 2026 registry states." } } }, "layout": { "x": 198, "y": 116 }, "updated": "2026-09-15" }, { "id": "claim-extend-oligometastatic-local-therapy", "title": "EXTEND gives a real oligometastatic PDAC signal, but not yet cure or overall survival", "shortTitle": "EXTEND local signal", "type": "claim", "status": "open", "scope": [ "pdac" ], "summary": "Metastasis-directed therapy increased median PFS from 2.5 to 10.3 months in 40 eligible patients, HR 0.43, and now needs larger external confirmation.", "content": "EXTEND randomized patients with five or fewer metastases to systemic treatment with or without comprehensive metastasis-directed treatment. In the PDAC cohort, 40 were eligible for the primary per-protocol analysis. Median PFS was 10.3 versus 2.5 months, HR 0.43, 95% CI 0.20–0.94; no grade 3 or worse event was attributed to local treatment. This is stronger than a surgical case series because it has a concurrent randomized control. It remains fragile: the cohort was tiny, the primary analysis was per protocol, treatment types varied, crossover and scan timing can influence PFS, and OS is not established. It also sits beside larger local-control trials that failed survival. The next study should lock oligometastatic biology, require at least six months of systemic control, stratify ctDNA and RAS state, and use OS plus treatment-free time. Until then local ablation is a selected hypothesis, not a general cure route.", "sourceIds": [ "ludmir-2024-extend", "timmer-2024-crossfire", "fietkau-2025-conko007" ], "links": [ { "target": "treatment", "relation": "part-of" }, { "target": "failure-local-intensification-pdac-survival-not-shown", "relation": "contradicts" }, { "target": "hypothesis-selected-local-consolidation", "relation": "supports" }, { "target": "claim-pdac-death-can-be-local-or-metastatic", "relation": "depends-on" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "moderate", "randomised": "moderate", "independentReplication": "none", "sampleSize": "40 eligible PDAC patients", "effect": "PFS 10.3 vs 2.5 months, HR 0.43, 95% CI 0.20-0.94", "limits": [ "Very small", "Per-protocol primary analysis", "OS not established", "Heterogeneous local treatment", "External replication absent", "Selection rule needs locking" ] }, "layout": { "x": 132, "y": 76 }, "updated": "2026-09-14" }, { "id": "claim-fap-radioligand-feasible-pdac-efficacy-unproven", "title": "FAP-targeted radiation reaches selected patients; pancreatic-cancer benefit is unproved", "shortTitle": "FAP treatment gate", "type": "claim", "status": "mixed", "scope": [ "pdac", "cholangiocarcinoma", "pan-cancer", "transfer-unproven" ], "summary": "Human studies establish delivery, organ-dose measurement and a workable phase 2 dose, but they do not yet show tumour shrinkage or longer useful life in PDAC.", "content": "Plain definition. Fibroblast activation protein, or FAP, is a surface marker found on many cancer-associated fibroblasts: support cells that help build the tissue around a tumour. A radioligand is an address label joined to a radioactive payload. The scan asks whether the label visits the disease; treatment asks whether enough radiation stays there to kill disease without damaging marrow, kidney, bowel or healing tissue. Human feasibility is established. In 2022, 11 compassionate-use patients included five pancreatic cancers; after one cycle, two breast cancers were stable and the other nine patients progressed. In 2026, LuMIERE scanned 35 people, treated 27 and included nine pancreatic cancers. Two dose-limiting toxicities occurred; 11 of 27, 40.7%, had an all-cause grade 3 or worse event. Of 25 people with response scans, one appendiceal cancer shrank and ten were stable, including four pancreatic cancers. No pancreatic partial response was reported. Measurement boundary. Gigabecquerel, GBq, is radioactive decays per second administered. Gray, Gy, is energy absorbed per kilogram of tissue. They are not interchangeable. A PET standardized uptake value, SUV, is a brightness snapshot; absorbed dose also depends on retention over time and isotope physics. LuMIERE selected people when every target lesion had SUVmax at least 1.5 times blood-pool SUVmean, but the paper did not report tumour-lesion dosimetry and could not robustly relate baseline uptake to response. Normal-organ estimates project six 9.25-GBq cycles to mean 19.0 Gy kidney and 2.3 Gy red marrow, with wide person-to-person spread; two people crossed the 23-Gy kidney limit and stopped. A separate uncontrolled 3BP-3940 series reported 34 complete or partial responses among 51 evaluable people, but began with 88 people across 21 cancers, used three isotopes, allowed other treatment and gave no PDAC-specific response denominator. Its single pancreatic dosimetry example delivered 2.2 Gy to tumour and 0.265 Gy to kidney, a ratio of 2.2 / 0.265 = 8.3; one example proves measurement is possible, not benefit. Decision: call the platform human-feasible and PDAC-efficacy unproved. Do not convert scan brightness, stable disease or mixed-cancer response into a survival claim.", "sourceIds": [ "baum-2022-fap2286-firsthuman", "mcconathy-2026-lumiere-phase1", "clinicaltrials-nct04939610-lumiere", "baum-2026-3bp3940", "clinicaltrials-nct05432193-frontier", "clinicaltrials-nct07229768-ctr-fapi", "ozdemir-2014-fibroblast-depletion", "rhim-2014-stroma-restrains" ], "links": [ { "target": "treatment", "relation": "part-of" }, { "target": "technology-fapi-pet-staging-not-screening", "relation": "distinguishes-treatment-from" }, { "target": "failure-fapi-scan-dose-benefit-collapse", "relation": "constrained-by" }, { "target": "unknown-pdac-fap-radioligand-treatment-effect", "relation": "leaves-open" }, { "target": "trial-pdac-fap-radioligand-action-map", "relation": "mapped-by" }, { "target": "failure-bulk-stroma-removal", "relation": "constrained-by" } ], "evidence": { "humanProspective": "moderate — multicentre dose escalation with nine pancreatic cancers", "humanRetrospective": "moderate — two uncontrolled mixed-cancer experiences", "randomised": "none", "independentReplication": "moderate for delivery, none for PDAC benefit", "sampleSize": "LuMIERE 35 scanned, 27 treated, 9 pancreatic; compassionate-use n=11 with 5 pancreatic; 3BP-3940 n=88 with response reported for 51", "effect": "No reported pancreatic partial response in LuMIERE phase 1; four pancreatic cancers had stable disease", "limits": [ "No PDAC control group", "Small pancreatic denominator", "LuMIERE did not report tumour-lesion dose", "Scan uptake was not linked robustly to response", "Stable disease can reflect natural history or prior treatment", "3BP-3940 mixes cancers, isotopes and other treatments", "Sponsors and investigators have development interests" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "FAP-rich tumour support tissue offers a physical address, but target cells are not always the malignant cells and FAP also marks repair." }, "animalEvidence": { "rating": "moderate", "reason": "Pancreatic models support radiation delivery, while past broad fibroblast removal warns that the support tissue has mixed functions." }, "inVitroEvidence": { "rating": "moderate", "reason": "Binding and isotope studies support the carrier, not whole-patient benefit." }, "reproducibility": { "rating": "moderate", "reason": "Multiple agents reach FAP-positive lesions in people, but no PDAC response or survival result has repeated." }, "commercialConflicts": { "rating": "high", "reason": "LuMIERE was funded and analysed with Novartis involvement, and uncontrolled agent reports create strong development incentives." }, "recency": { "rating": "current", "reason": "Includes phase 1 publication and registry states checked through 15 September 2026." } } }, "layout": { "x": 61, "y": 133 }, "updated": "2026-09-15" }, { "id": "claim-ferroptosis-pdac-selectivity-not-class-effect", "title": "Ferroptosis is a cell-state vulnerability, not a safe pancreatic drug class", "shortTitle": "Ferroptosis needs cell selectivity", "type": "claim", "status": "supported", "scope": [ "pdac", "ferroptosis", "metabolism", "immune-microenvironment", "drug-delivery" ], "summary": "Iron-driven membrane damage can kill PDAC cells, but the effect reverses with cell type and context; tumour-selective delivery is the core requirement.", "content": "Ferroptosis is cell death caused when iron-driven oxidation destroys lipid membranes faster than the cell can repair them. Two repair inputs are SLC7A11/xCT, which imports cystine used to make the antioxidant glutathione, and GPX4, which clears lipid peroxides. Removing xCT killed PDAC cells and delayed xenografts, but other cysteine routes prevented tumour elimination in living animals. Fentomycin-1 killed iron-rich, CD44-high primary PDAC cells, including drug-tolerant states, in culture but has no pancreatic animal or human treatment result. A tumour-selective GPX4 degrader improved immune attack in pancreatic mouse models because it spared measured immune cells. The opposite experiment exposes the danger: high iron or pancreas-wide GPX4 loss accelerated KRAS-driven pancreatic tumour formation in mice by releasing oxidized DNA, activating STING and recruiting macrophages. These results do not cancel each other. Together they define the engineering problem: deliver enough iron chemistry or antioxidant failure to vulnerable tumour cells while sparing immune, pancreatic and other normal cells, and measure whether the resulting inflammation helps or harms. A drug does not inherit a class effect from the word ferroptosis.", "sourceIds": [ "daher-2019-xct-pdac-ferroptosis", "dai-2020-ferroptotic-damage-pdac", "li-2023-tumour-selective-gpx4-degrader", "caneque-2025-fentomycin-lysosomal-iron" ], "links": [ { "target": "treatment", "relation": "part-of" }, { "target": "failure-cnsi-fe-resensitization-and-mechanism-claim", "relation": "bounded-by" }, { "target": "unknown-ferroptosis-cell-type-coverage-and-human-benefit", "relation": "leaves-open" }, { "target": "hypothesis-cnsi-fe-current-study-mechanism-gate", "relation": "motivates" }, { "target": "trial-pdac-ferroptosis-action-map", "relation": "mapped-by" }, { "target": "claim-pdac-metabolic-adaptation-is-redundant", "relation": "supports" }, { "target": "failure-immune-reprogramming-without-patient-benefit", "relation": "supports" } ], "evidence": { "humanProspective": "none for a PDAC ferroptosis treatment effect", "humanRetrospective": "none", "randomised": "none", "independentReplication": "moderate — model vulnerabilities repeat, direction depends on cell context", "sampleSize": "Multiple cell and mouse systems; human CNSI-Fe exposure includes two pancreatic cases", "effect": "Tumour-cell ferroptosis can suppress models; non-selective pancreatic ferroptotic damage can accelerate tumour formation", "limits": [ "Mostly model evidence", "Different induction methods", "Tumour initiation differs from established disease", "No clinical GPX4 inhibitor", "Cell-selective degrader untested in humans", "Human biomarkers absent" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Iron, lipid oxidation and antioxidant repair form a measured causal chain." }, "animalEvidence": { "rating": "strong", "reason": "Genetic and drug studies show both tumour suppression and context-dependent promotion." }, "inVitroEvidence": { "rating": "strong", "reason": "PDAC cells show xCT, GPX4 and lysosomal-iron vulnerabilities." }, "reproducibility": { "rating": "moderate", "reason": "The mechanism repeats, but useful direction is conditional on cell and tissue context." }, "commercialConflicts": { "rating": "moderate", "reason": "Academic tool compounds and a company iron product make different claims that must not be pooled." }, "recency": { "rating": "current", "reason": "The route includes 2025 chemical tools and 2026 first-human reporting." } } }, "layout": { "x": 1168, "y": 1092 }, "updated": "2026-09-15" }, { "id": "claim-fgfr2-targeting-valid-but-not-durable", "title": "FGFR2 is a valid target in rearranged intrahepatic cholangiocarcinoma, but current control is not durable", "shortTitle": "FGFR2 precision limit", "type": "claim", "status": "supported", "scope": [ "cholangiocarcinoma" ], "summary": "Single-arm inhibitors produced 35.5% to 42% response; in 2026 first-line pemigatinib improved progression-free survival but not overall survival versus chemotherapy.", "content": "FIGHT-202 found 38 responses among 107 previously treated FGFR2 fusion- or rearrangement-positive cholangiocarcinomas, 35.5%. FOENIX-CCA2 found 43 responses among 103 FGFR2-rearranged intrahepatic cancers, 42%, with median response duration 9.7 months and progression-free survival 9.0 months. The 2026 FIGHT-302 trial moved the question into first-line treatment: among 167 randomised patients before early closure, pemigatinib increased median progression-free survival from 6.8 to 8.3 months and response from 15% to 47%, while median overall survival was 24.4 versus 25.0 months. This validates oncogene targeting and routine molecular testing for this subgroup. It also shows the remaining problem: resistance and later therapy erase or obscure survival gain. The evidence applies mainly to FGFR2-rearranged intrahepatic disease, not all cholangiocarcinoma or gallbladder cancer.", "sourceIds": [ "abou-alfa-2020-pemigatinib", "goyal-2023-futibatinib", "bekaii-saab-2026-fight302" ], "links": [ { "target": "cholangiocarcinoma", "relation": "part-of" }, { "target": "treatment", "relation": "supports" }, { "target": "claim-biliary-cancers-not-one-disease", "relation": "supports" }, { "target": "bottlenecks", "relation": "supports" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "moderate", "independentReplication": "strong", "sampleSize": "107 in FIGHT-202 target cohort; 103 in FOENIX-CCA2; 167 randomised in FIGHT-302", "effect": "ORR 35.5% and 42% in later-line studies; first-line PFS 8.3 versus 6.8 months, OS 24.4 versus 25.0 months", "limits": [ "Rare molecular subgroup", "Mainly intrahepatic disease", "Two key studies were single-arm", "Phase 3 closed early", "Crossover and later therapy affect OS", "Company-funded programmes" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "FGFR2 fusions create a direct oncogenic dependency in a defined intrahepatic subgroup." }, "animalEvidence": { "rating": "strong", "reason": "Models support FGFR dependence and acquired kinase-domain resistance." }, "inVitroEvidence": { "rating": "strong", "reason": "Cell and molecular studies support target engagement and resistance mutations." }, "reproducibility": { "rating": "strong", "reason": "Multiple inhibitors and trials reproduce response, although overall-survival benefit remains absent." }, "commercialConflicts": { "rating": "high", "reason": "The pivotal programmes are company-funded and compare proprietary inhibitors." }, "recency": { "rating": "current", "reason": "The record includes the 2026 randomized first-line result." } } }, "layout": { "x": 36, "y": 11 }, "updated": "2026-09-14" }, { "id": "claim-fibrosis-nerve-mechanosensing", "title": "Fibrosis can activate tumour-supporting Schwann cells through mechanical force", "shortTitle": "Stroma–nerve mechanics", "type": "claim", "status": "open", "scope": [ "pdac", "transfer-unproven" ], "summary": "Two human-tissue series and model experiments show that Schwann cells can support nerve invasion through cell-state and force responses, but no patient treatment has tested either route.", "content": "Perineural invasion is not passive tumour growth along an empty space. A Keio series of 168 resected PDACs found focal epithelial differentiation at nerve invasion sites in 124 tumours, 74%; Schwann-cell co-culture induced the same direction of change, but culture cannot prove causality in a patient. A 2026 cross-disciplinary study then connected stiffer human PDAC stroma with activated Schwann cells. Mouse live imaging and force-applied systems linked nuclear compression and phospholipase A2 to c-Jun activation; fibrosis without cancer cells was sufficient to activate Schwann cells in models. Together these results make a tumour–nerve–matrix loop plausible and provide independent human observations, while the causal steps remain model-derived. This creates a narrower route than removing all stroma: interrupt a measured Schwann-cell state or force signal. The next experiment is not an unselected drug trial. Confirm the state in independent untreated PDAC and chronic-pancreatitis tissue, perturb it in intact human tumour–nerve–stroma cultures, and measure invasion, pain, nerve function and tumour growth without destroying protective matrix.", "sourceIds": [ "fujii-nishimura-2018-schwann-met", "stupakov-2026-schwann-mechanics" ], "links": [ { "target": "claim-immune-exclusion-is-multicellular", "relation": "extends" }, { "target": "failure-bulk-stroma-removal", "relation": "suggests-narrower-route-than" }, { "target": "research-frontier-ranking", "relation": "candidate-for" }, { "target": "treatment", "relation": "part-of" }, { "target": "model-pdac-causal-chain", "relation": "supports" } ], "evidence": { "humanProspective": "none", "humanRetrospective": "moderate", "randomised": "none", "independentReplication": "moderate", "sampleSize": "Keio series of 168 resected tumours plus a separate human-tissue, mouse and cell-mechanics programme", "effect": "Mechanistic state change; no clinical effect", "limits": [ "Causal evidence remains model-heavy", "No human intervention", "Causal chain incomplete in native human tissue", "Possible wound-healing and nerve toxicity", "Different studies identify different Schwann-cell routes" ] }, "layout": { "x": 122, "y": 104 }, "updated": "2026-09-14" }, { "id": "claim-fluke-control-prevents-exposure-not-yet-cancer", "title": "Liver-fluke control removes a proven cause but has not yet quantified prevented cholangiocarcinoma", "shortTitle": "Fluke control outcome gap", "type": "claim", "status": "supported", "scope": [ "cholangiocarcinoma", "hepatobiliary", "prevention" ], "summary": "Opisthorchis viverrini is a proven human cause of cholangiocarcinoma, and a Thai One Health programme reported a 95% relative fall from 6.0% infection to 0.3%, but no controlled cancer-incidence result.", "content": "The causal link is stronger than the programme-effect estimate. IARC concludes that chronic Opisthorchis viverrini and Clonorchis sinensis infection causes cholangiocarcinoma. That makes avoidance of raw infected fish, sanitation, diagnosis and treatment true primary prevention rather than screening. A three-year rural Thai programme reported 218 infections among 3,600 participants at baseline, 6.0%, followed by no human reinfections in year two and 0.3% prevalence in year three after education, anthelminthic treatment, household and animal surveillance and environmental measures. The reported prevalence change is (6.0 - 0.3) / 6.0 = 95% relative, but it is not a randomized effect: the intervention was bundled, the time-specific denominators are not equivalent in the abstract, and no cancer endpoint was measured. The live randomized eCHEC study plans 480 participants and measures self-reported behaviour plus stool infection at 12 months. It can test exposure control, but not the number of cholangiocarcinomas or deaths prevented over the following decade. A defensible prevention programme must therefore preserve the near-term infection endpoint and link every cluster to long-term cancer incidence, stage, surgery and death. It must not relabel behaviour change as cancer prevention achieved.", "sourceIds": [ "iarc-2012-liver-flukes", "charoensuk-2024-one-health-opisthorchis", "clinicaltrials-nct05321992-echec" ], "links": [ { "target": "cholangiocarcinoma", "relation": "part-of" }, { "target": "prevention-risk-interception-audit", "relation": "extends" }, { "target": "failure-echec-cancer-outcome-mismatch", "relation": "supports" }, { "target": "trial-cca-action-map", "relation": "informs" }, { "target": "unknown-cca-surveillance-utility", "relation": "distinct-from" } ], "evidence": { "humanProspective": "moderate — longitudinal infection programme", "humanRetrospective": "strong for causal exposure classification", "randomised": "weak — planned for behaviour and infection, not cancer", "independentReplication": "strong for causation; weak for the bundled programme effect", "sampleSize": "One Health report n=3,600; eCHEC planned randomized n=480", "effect": "Reported infection prevalence fell from 6.0% to 0.3%, a 95% relative change; cancer effect unmeasured", "limits": [ "Before-and-after programme", "Bundled components", "No cancer incidence or mortality endpoint", "Cancer latency is much longer than follow-up", "Reinfection and food practice vary by community" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Removing a Group 1 carcinogenic chronic infection removes a direct cause of bile-duct injury." }, "animalEvidence": { "rating": "moderate", "reason": "Experimental infection supports carcinogenic mechanisms but does not quantify human programme benefit." }, "inVitroEvidence": { "rating": "moderate", "reason": "Inflammatory and mutagenic routes support causation, not delivery effectiveness." }, "reproducibility": { "rating": "weak", "reason": "Multiple programmes improve behaviour or infection measures, but a controlled cancer-incidence effect has not been reproduced." }, "commercialConflicts": { "rating": "low", "reason": "The interventions are mainly public-health measures rather than proprietary products." }, "recency": { "rating": "current", "reason": "The active trial and 2024 longitudinal evidence were checked in September 2026." } } }, "updated": "2026-09-14" }, { "id": "claim-gallbladder-diagnostic-models-existing-lesions-only", "title": "Gallbladder AI can classify selected lesions but has not shown population early-detection utility", "shortTitle": "Lesion diagnosis, not screening", "type": "claim", "status": "supported", "scope": [ "gallbladder" ], "summary": "GBCseeker and GAIA-MIL have useful external diagnostic results, but both start after suspicious pathology is visible and their predictive values depend on highly enriched cohorts.", "content": "GBCseeker combined targeted plasma cfDNA, CT radiomics, CA19-9 and sex in 301 Chinese patients with gallbladder-occupying lesions. Its independent 98-person set contained 62 cancers and 36 benign lesions. Sensitivity was 93.55%, specificity 77.78%, AUC 0.93 and accuracy 87.76%. Those are strong lesion-diagnosis results, but the reported PPV of 87.88% is partly a consequence of 63.3% cancer prevalence. If the same sensitivity and specificity were transported without loss to 1,000 people at 2% prevalence, it would produce about 19 true positives and 218 false positives: PPV 7.9% and 11.6 false positives per true positive. At the roughly 0.2% incidental-cancer rate assumed by the UK P-iGBC protocol, PPV would be about 0.84% and false positives about 119 per true positive. These are transport calculations, not observed model performance in either population.\n\nGAIA-MIL used ultrasound and reported pooled external sensitivity 78.2%, specificity 73.4% and AUC 0.778 in 122 retrospectively assembled patients from three Indian centres. Centre AUC ranged from 0.699 to 0.950 and external sensitivity was 69.4% for lesions below 10 mm, the size range most important to early action. At 2% prevalence, unchanged performance would model to about 16 true positives, four missed cancers and 261 false positives per 1,000. Both models may improve work-up after a visible lesion; neither has been tested as general screening, future-risk prediction or a randomized decision aid. Their next test is silent prospective use in every consecutive eligible patient, fixed thresholds, local calibration, stage-specific errors and comparison against clinician plus standard imaging—not another enriched AUC.", "sourceIds": [ "yang-2025-gbcseeker", "gupta-2026-gaia-mil", "clinicaltrials-nct06531408-pigbc" ], "links": [ { "target": "detection", "relation": "part-of" }, { "target": "ai-method-audit", "relation": "supports" }, { "target": "unknown-gallbladder-prevention-selector", "relation": "supports" }, { "target": "hypothesis-gallbladder-two-stage-interception-rule", "relation": "could-test" } ], "evidence": { "humanProspective": "moderate for selected-lesion diagnosis", "humanRetrospective": "moderate for external ultrasound validation", "randomised": "none", "independentReplication": "weak — external centres but no independent build or action test", "sampleSize": "GBCseeker n=301 with external n=98; GAIA-MIL total 1,151 patients with pooled external n=122", "effect": "GBCseeker external sensitivity 93.55%, specificity 77.78%, AUC 0.93; GAIA-MIL pooled external sensitivity 78.2%, specificity 73.4%, AUC 0.778", "limits": [ "Both begin after visible or suspected gallbladder pathology", "External test sets are small and cancer-enriched", "GBCseeker requires blood sequencing and CT", "GAIA-MIL external cohorts were retrospective", "GAIA-MIL sensitivity fell for sub-10-mm lesions", "No randomized management or mortality result", "Transport calculations assume unchanged accuracy" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "Existing cancer can alter cfDNA, CA19-9 and imaging, but early or future disease may not produce the same signal." }, "animalEvidence": { "rating": "not-applicable", "reason": "Diagnostic performance must be established in the intended human pathway." }, "inVitroEvidence": { "rating": "moderate", "reason": "Mutation assays and image features are measurable, but technical detection does not establish decision benefit." }, "reproducibility": { "rating": "weak", "reason": "Each model has external-site data, yet neither has an independent prospective action test and centre performance varies." }, "commercialConflicts": { "rating": "moderate", "reason": "Broader testing could benefit assay and software owners; fixed code, independent analysis and data rights are needed." }, "recency": { "rating": "current", "reason": "The multimodal and ultrasound reports were published in 2025 and 2026." } } }, "layout": { "x": 46, "y": 75 }, "updated": "2026-09-14" }, { "id": "claim-gallbladder-endemic-late-presentation", "title": "A North Indian prospective cohort shows that gallbladder-cancer failure begins before treatment selection", "shortTitle": "Gallbladder late presentation", "type": "claim", "status": "supported", "scope": [ "gallbladder", "hepatobiliary" ], "summary": "At a high-volume Varanasi centre, 80.2% of 1,500 consecutive new patients had metastatic disease; only 92 received definitive surgery.", "content": "A prospective Tata Memorial Centre cohort in Varanasi recorded 1,500 newly diagnosed gallbladder-cancer patients over nine months. At presentation, 1,203, or 80.2%, already had metastatic disease. Among 75 assigned neoadjuvant chemotherapy for borderline resectable disease, 61.3% progressed and only 12.7% reached curative surgery. Of 116 explored surgically, 92 received definitive surgery and selected one-year survival was 88.5%. Among 768 planned for palliative chemotherapy, only 294 received at least one cycle and median survival was five months. These groups were not randomized, so surgery did not create all of the survival difference; selection and stage did. The stronger conclusion is operational: in this high-incidence region, most patients never enter a curative pathway and many do not receive planned systemic treatment. Prevention and early detection therefore have unusually high value, but a screening programme still needs a risk denominator, a safe work-up and mortality evidence. This cohort adds Indian evidence that mixed biliary trials cannot supply.", "sourceIds": [ "vineet-2026-north-india-gallbladder-cohort", "gonzalez-2025-cholecystectomy-chile", "lorenzo-bermejo-2025-gallbladder-specimens" ], "links": [ { "target": "gallbladder", "relation": "part-of" }, { "target": "claim-gallbladder-prevention-selection-bottleneck", "relation": "raises-value-of" }, { "target": "model-gallbladder-causal-chain", "relation": "supports" }, { "target": "detection", "relation": "supports" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "none", "independentReplication": "moderate", "sampleSize": "1,500 prospectively recorded new patients at one endemic-region centre", "effect": "80.2% metastatic at presentation; 92 definitive operations; palliative-treatment median OS five months", "limits": [ "Single centre", "Treatment groups selected by stage and fitness", "Short recruitment window", "No population denominator", "One-year surgical follow-up", "Cannot estimate screening benefit" ] }, "layout": { "x": 137, "y": 35 }, "updated": "2026-09-14" }, { "id": "claim-gallbladder-incidental-residual-disease", "title": "Incidental gallbladder cancer is a pathology-and-restaging emergency, not automatic proof that re-resection caused longer survival", "shortTitle": "Incidental GBC residual disease", "type": "claim", "status": "supported", "scope": [ "gallbladder" ], "summary": "A Dutch national cohort found residual tumour in 35% of re-resected incidental cancers, usually lymph nodes; re-resection correlated with longer survival but was not randomized.", "content": "The Netherlands Cancer Registry identified 463 people whose gallbladder cancer was discovered after cholecystectomy; only 110, 24%, underwent re-resection, after a median 66 days. Residual disease was found in 35% and most often in lymph nodes, 23%. R0 resection was achieved in 92% of those re-resected. Median overall survival was 52.6 months with re-resection versus 13.7 months without it. That fourfold difference is not a causal treatment effect because patients selected for another operation were likely fitter, earlier-stage, better staged and able to survive until surgery.\n\nWithin the re-resected group, residual liver disease carried HR 5.54 and nodal disease HR 2.35 for death; median survival was not reached without residual disease and 23.1 months with it. Those associations make the first cholecystectomy specimen and complete staging decisive information. The mission needs rapid expert pathology, tumour-depth and margin reporting, nodal and liver staging, and referral for a prospective re-resection pathway. It also needs to record patients screened out or never referred so that the benefit of re-resection is not exaggerated by selection.", "sourceIds": [ "de-savornin-lohman-2020-incidental-gallbladder", "singh-2026-gallbladder-liver-resection-rct" ], "links": [ { "target": "gallbladder", "relation": "part-of" }, { "target": "model-gallbladder-causal-chain", "relation": "supports" }, { "target": "claim-gallbladder-liver-resection-extent", "relation": "depends-on" }, { "target": "treatment", "relation": "supports" } ], "evidence": { "humanProspective": "none", "humanRetrospective": "strong", "randomised": "none for re-resection versus no re-resection", "independentReplication": "moderate", "sampleSize": "463 incidental gallbladder cancers; 110 underwent re-resection", "effect": "Residual disease in 35%; median OS 52.6 versus 13.7 months with versus without re-resection; residual liver disease HR 5.54 and nodal disease HR 2.35 within re-resected patients", "limits": [ "Registry cohort", "Strong selection and referral bias", "Immortal-time bias is possible", "Only 24% underwent re-resection", "Pathology review was limited to re-resected patients", "No randomized estimate of re-resection benefit" ] }, "layout": { "x": 63, "y": 41 }, "updated": "2026-09-14" }, { "id": "claim-gallbladder-liver-resection-extent", "title": "More liver removal did not improve gallbladder-cancer outcomes in the first randomized comparison", "shortTitle": "Gallbladder surgery extent", "type": "claim", "status": "supported", "scope": [ "gallbladder", "hepatobiliary" ], "summary": "A 163-patient phase 3 trial found no disease-free or overall-survival advantage for segment 4b/5 resection over a wedge during radical cholecystectomy.", "content": "A single-centre Delhi phase 3 trial randomized patients intraoperatively, after excluding metastatic or unresectable gallbladder adenocarcinoma, to non-anatomical liver wedge resection or removal of segments 4b and 5. The final analysis included 163 patients. Segment 4b/5 surgery took 318 versus 287 minutes and blood loss was 265 versus 223 mL. R0 resection, morbidity and mortality did not differ. At median follow-up of 27 months, mean disease-free survival was 41.8 versus 44.7 months, HR 0.80, 95% CI 0.47 to 1.40, P=0.50; mean overall survival was 45.3 versus 50.7 months, HR 0.60, 95% CI 0.36 to 1.14, P=0.12.\n\nThis trial rejects the simple belief that a larger standard liver resection necessarily improves control. It does not prove that the two operations are equivalent: the confidence intervals allow material benefit or harm, follow-up is short for cure, the trial was open-label and single-centre, and only patients considered resectable at operation entered the comparison. The usable result is a surgical-quality boundary. Secure a negative margin and adequate nodal staging, but do not substitute greater tissue removal for better staging, pathology, systemic control or a trial-powered subgroup rule.", "sourceIds": [ "singh-2026-gallbladder-liver-resection-rct" ], "links": [ { "target": "gallbladder", "relation": "part-of" }, { "target": "model-gallbladder-causal-chain", "relation": "supports" }, { "target": "treatment", "relation": "supports" }, { "target": "claim-gallbladder-incidental-residual-disease", "relation": "extends" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong", "independentReplication": "weak", "sampleSize": "163 patients in final analysis; 83 segment 4b/5 and 80 wedge", "effect": "DFS HR 0.80, 95% CI 0.47 to 1.40; OS HR 0.60, 95% CI 0.36 to 1.14; no difference in R0 resection, morbidity or mortality", "limits": [ "Single-centre Indian trial", "Open-label", "Median follow-up 27 months", "Final-analysis rather than clearly stated intention-to-treat denominator in the abstract", "Registration identifier not reported in PubMed", "Wide confidence intervals", "Only intraoperatively resectable disease" ] }, "layout": { "x": 67, "y": 47 }, "updated": "2026-09-14" }, { "id": "claim-gallbladder-polyp-guidelines-still-overtreat", "title": "Current gallbladder-polyp rules still produce many low-yield scans and operations", "shortTitle": "Polyp-rule operation burden", "type": "claim", "status": "supported", "scope": [ "gallbladder" ], "summary": "In a prospective 26-centre Dutch cohort, the 2022 rule changed only three surgery indications and found one high-grade lesion among 68 guideline-indicated operations.", "content": "POLYP prospectively followed 302 Dutch patients already selected for gallbladder-polyp surveillance or surgery. Of 231 people with repeat imaging, the median was three scans over 23.2 months. Eighty-eight people had cholecystectomy; 68 met the 2022 European polyp indication, only three fewer than under the 2017 rule. Pathology among those 68 found six adenomas, four with dysplasia, one with high-grade dysplasia and no cancer. That is 11.3 operations per adenoma, 17 per dysplastic adenoma and 68 per high-grade lesion in this selected cohort. Forty-one of 68 had cholesterolosis or another non-neoplastic wall finding and 21 had no polyp or wall abnormality. The 2022 revision therefore changed age eligibility but did not solve measurement error or identify the dangerous minority. This does not show that the single high-grade lesion would have become lethal cancer or that surgery was unnecessary; it shows that size, growth and age alone have low action yield. The earlier 622,227-person Kaiser cohort supplies the opposing population scale: initial size at least 10 mm carried more cancer risk, but overall cancer occurrence was almost identical with and without a recorded polyp. The next rule must be tested against high-grade dysplasia, invasive cancer, surgical harm and interval cancer, not the number of operations that follow a guideline.", "sourceIds": [ "van-dijk-2025-polyp-study", "szpakowski-2020-gallbladder-polyps" ], "links": [ { "target": "claim-gallbladder-polyp-selector", "relation": "extends" }, { "target": "failure-polyp-surveillance-low-yield", "relation": "supports" }, { "target": "unknown-gallbladder-prevention-selector", "relation": "supports" }, { "target": "hypothesis-gallbladder-two-stage-interception-rule", "relation": "constrains" } ], "evidence": { "humanProspective": "moderate — multicentre cohort but enrolled at different points in care", "humanRetrospective": "strong for the separate population cohort", "randomised": "none", "independentReplication": "moderate for low risk; weak for a replacement rule", "sampleSize": "POLYP n=302, repeat imaging n=231, cholecystectomy n=88 and 2022-guideline operations n=68; Kaiser ultrasound cohort n=622,227", "effect": "One high-grade dysplastic adenoma and no cancer among 68 guideline-indicated operations; 41 non-neoplastic abnormalities and 21 with no polyp or wall abnormality", "limits": [ "Surgical-service recruitment enriched the operation rate", "Many follow-up paths were unfinished", "Only one high-grade lesion makes the yield imprecise", "Dutch results may not transport to endemic regions", "No mortality or cancer-prevention endpoint" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Adenoma and dysplasia can be precursor states, but most reported polyps are not neoplastic." }, "animalEvidence": { "rating": "not-applicable", "reason": "The claim concerns human clinical selection and operation yield." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Laboratory activity cannot determine whether a measured polyp warrants surgery." }, "reproducibility": { "rating": "moderate", "reason": "Large population and prospective clinical cohorts both show low absolute cancer yield, with different selection mechanisms." }, "commercialConflicts": { "rating": "low", "reason": "The checked report declares no conflicts and evaluates guidelines rather than a product." }, "recency": { "rating": "current", "reason": "The prospective results were published in 2025 and assessed the 2022 rule." } } }, "layout": { "x": 41, "y": 71 }, "updated": "2026-09-14" }, { "id": "claim-gallbladder-polyp-selector", "title": "Most ultrasound-detected gallbladder polyps are not a useful cancer-prevention target", "shortTitle": "Gallbladder polyp risk", "type": "claim", "status": "supported", "scope": [ "gallbladder" ], "summary": "In a 622,227-person cohort, recorded gallbladder polyps were common and overall cancer risk was almost identical with and without them; initial size carried more information.", "content": "A Kaiser Permanente cohort followed 622,227 adults who had abdominal ultrasound over 20 years. Gallbladder polyps were recorded in 35,856 people. Nineteen later received a gallbladder-cancer diagnosis: 0.053%, almost identical to 316 of 586,357 people without a recorded polyp, 0.054%. The unadjusted cancer incidence rose from 1.3 per 100,000 person-years for an initial polyp smaller than 6 mm to 128.2 per 100,000 person-years for an initial polyp of at least 10 mm. Size matters, but the absolute event count remained small.\n\nAmong 5,721 people with a polyp who underwent cholecystectomy, 14 had cancer in the specimen, about one cancer per 409 operations. That ratio is not a number needed to treat because indications varied, cancers may already have been present, and removing the gallbladder prevents future disease rather than merely detecting it. Polyp growth was also common: reaching 10 mm during follow-up did not identify a cancer among 507 people over 1,549 person-years. Imaging-report extraction, selective repeat scanning and censoring at surgery create further bias. The current prevention rule therefore needs a second biological or imaging feature beyond a small polyp or small size change, followed by prospective measurement of cancers prevented, cancers missed and surgical harm.", "sourceIds": [ "szpakowski-2020-gallbladder-polyps", "lorenzo-bermejo-2025-gallbladder-specimens" ], "links": [ { "target": "gallbladder", "relation": "part-of" }, { "target": "model-gallbladder-causal-chain", "relation": "supports" }, { "target": "claim-gallbladder-prevention-selection-bottleneck", "relation": "supports" }, { "target": "detection", "relation": "depends-on" } ], "evidence": { "humanProspective": "none", "humanRetrospective": "strong", "randomised": "none", "independentReplication": "moderate", "sampleSize": "622,227 adults undergoing ultrasound; 35,856 had a quantitative gallbladder polyp and 19 developed gallbladder cancer", "effect": "Cancer proportion 0.053% with a polyp versus 0.054% without; incidence 1.3 per 100,000 person-years below 6 mm and 128.2 at 10 mm or larger", "limits": [ "Retrospective integrated-health-system cohort", "Only 19 cancers after a recorded polyp", "Ultrasound reports were extracted by text processing", "Repeat imaging was selective", "Follow-up ended at cholecystectomy", "Large masses may have been labelled polyps", "No direct estimate of deaths or cancers prevented by surgery" ] }, "layout": { "x": 38, "y": 66 }, "updated": "2026-09-14" }, { "id": "claim-gallbladder-prevention-must-beat-surgery-harm", "title": "A gallbladder-prevention rule must beat the local harm of surgery, not a universal polyp cutoff", "shortTitle": "Prevention net-benefit gate", "type": "claim", "status": "supported", "scope": [ "gallbladder" ], "summary": "Global cholecystectomy harm is measurable and varies by setting, so the intervention threshold must be country-, centre- and patient-specific.", "content": "AMBROSE prospectively recorded 21,706 consecutive cholecystectomies for benign disease in 57 countries: 8.0% had a 30-day complication, 0.4% died, 1.3% had a bile leak and 0.2% had a severe bile-duct injury. Per 1,000 operations that is about 80 complications, four deaths, 13 leaks and two severe duct injuries. These totals mix elective, emergency and delayed surgery and therefore cannot be assigned to an asymptomatic prevention operation. They establish the size and variation of the harm that a programme must measure locally.\n\nThe mortality-only break-even equation is p × b > m, where p is the probability that the removed gallbladder contains or will develop lethal preventable disease, b is the fraction of those deaths surgery prevents and m is operative mortality. Using the mixed AMBROSE mortality only as a visible upper-level reference, m=0.004 means p must exceed 0.4% even if b=1; if surgery prevents half of target deaths, p must exceed 0.8%. A safety target of at least five prevented cancer deaths per operative death would require p above 2.0% when b=1. Those are illustrations, not a clinical threshold, because elective mortality, competing death, symptom relief, nonfatal harm and country incidence change the result. The rule should therefore publish expected cancer deaths prevented, operation deaths, severe duct injuries, other complications, quality-adjusted life and cost for each country and age group. A biomarker that raises pathology yield but fails that comparison has not solved prevention.", "sourceIds": [ "wong-2025-ambrose-cholecystectomy", "gonzalez-2025-cholecystectomy-chile", "van-dijk-2025-polyp-study" ], "links": [ { "target": "claim-gallbladder-prevention-selection-bottleneck", "relation": "constrains" }, { "target": "unknown-gallbladder-prevention-selector", "relation": "constrains" }, { "target": "hypothesis-gallbladder-two-stage-interception-rule", "relation": "constrains" }, { "target": "risks", "relation": "supports" } ], "evidence": { "humanProspective": "strong for short-term surgical harm", "humanRetrospective": "moderate for prevention association", "randomised": "none for preventive cholecystectomy", "independentReplication": "strong for surgery having nonzero harm; weak for net prevention benefit", "sampleSize": "AMBROSE n=21,706 operations in 57 countries", "effect": "Thirty-day complications 8.0%, mortality 0.4%, bile leak 1.3% and severe bile-duct injury 0.2% in a mixed benign-disease surgical cohort", "limits": [ "Mixed elective, emergency and delayed operations", "No asymptomatic prevention subgroup effect supplied here", "Thirty-day follow-up misses long-term harm", "No estimate of individual cancer deaths prevented", "Worked thresholds omit symptom benefit and quality weights" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Removing the organ prevents gallbladder cancer but also creates immediate surgical risk." }, "animalEvidence": { "rating": "not-applicable", "reason": "Human operative outcomes determine this gate." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Laboratory evidence cannot set an acceptable surgical-risk threshold." }, "reproducibility": { "rating": "strong", "reason": "Large international and national cohorts consistently show low but nonzero mortality and bile-duct injury." }, "commercialConflicts": { "rating": "low", "reason": "The core result comes from a global collaborative outcomes cohort rather than a prevention product." }, "recency": { "rating": "current", "reason": "The international cohort reported contemporary 2022 practice in 2024-2025." } } }, "layout": { "x": 52, "y": 79 }, "updated": "2026-09-14" }, { "id": "claim-gallbladder-prevention-selection-bottleneck", "title": "Gallbladder cancer has a physically removable precursor organ, but gallstones are too common to define the surgical target alone", "shortTitle": "Gallbladder prevention gap", "type": "claim", "status": "open", "scope": [ "gallbladder" ], "summary": "Chile's programme increased cholecystectomy and ecological mortality fell, but individual benefit is unproved; 10,561 surgical specimens show risk markers that now need prospective validation.", "content": "Gallbladder cancer is unusual because removing the gallbladder before invasion could prevent the disease. Chile has guaranteed cholecystectomy for selected gallstone patients aged 35–49 since 2006. An interrupted time-series study found the operation rate rose by 45 per 100,000 people per year. Each 1% increase in people living without a gallbladder for ten years was associated with 0.73% lower gallbladder-cancer mortality, but only in women, southern Chile and people over 60, and several other digestive cancers fell too. That pattern is compatible with benefit but also with ecological confounding. In 10,561 gallstone patients operated in Argentina, Bolivia, Chile and Peru, cancer or dysplasia risk tracked with clinical suspicion, open surgery, female sex, stones over 3 cm, age, smoking and other factors; effects varied by site and differed between dysplasia and cancer.\n\nPolyps are not a simple alternative selector. In a 622,227-person ultrasound cohort, overall gallbladder-cancer occurrence was 0.053% among people with a recorded polyp and 0.054% without one. Initial polyps at least 10 mm carried much higher incidence than polyps below 6 mm, but growth to 10 mm did not identify a cancer among 507 people observed for 1,549 person-years. Among 5,721 patients with a polyp who underwent surgery, 14 cancers were found, about one per 409 operations, though that is not a causal number needed to treat. The engineering bottleneck is a prospectively validated rule combining geography, age, stone or polyp features, biological markers and imaging that selects enough future cancers while keeping operations, bile-duct injuries and system cost acceptable. The test must be evaluated on cancer incidence and mortality, not retrospective pathology enrichment alone.", "sourceIds": [ "gonzalez-2025-cholecystectomy-chile", "lorenzo-bermejo-2025-gallbladder-specimens", "szpakowski-2020-gallbladder-polyps" ], "links": [ { "target": "model-gallbladder-causal-chain", "relation": "supports" }, { "target": "detection", "relation": "depends-on" }, { "target": "programme", "relation": "depends-on" }, { "target": "experiments", "relation": "tests" }, { "target": "claim-gallbladder-polyp-selector", "relation": "depends-on" } ], "evidence": { "humanProspective": "none", "humanRetrospective": "strong", "randomised": "none", "independentReplication": "moderate", "sampleSize": "National aggregate data, 10,561 operated gallstone patients across four South American countries, and 622,227 adults undergoing ultrasound in a US health system", "effect": "Programme added 45 cholecystectomies per 100,000 people per year; ecological association 0.73% lower GBC mortality per 1% more people without a gallbladder for 10 years; cancer proportion 0.053% with versus 0.054% without a recorded polyp", "limits": [ "Ecological analysis", "No individual causal estimate", "Operated-cohort selection", "Polyp cohort was retrospective with only 19 cancers after a recorded polyp", "Risk model not prospectively validated", "Regional heterogeneity", "Surgical harms and full costs not quantified here" ] }, "layout": { "x": 42, "y": 72 }, "updated": "2026-09-14" }, { "id": "claim-general-population-screening-2019", "title": "The 2019 USPSTF assessment found no net benefit for general-population pancreatic screening", "shortTitle": "No general screening", "type": "claim", "status": "supported", "scope": [ "pancreatic-all" ], "summary": "The recommendation was grade D for asymptomatic adults not known to have high inherited or familial risk.", "content": "The dated claim matters: in 2019 the USPSTF found low incidence, uncertain test accuracy, no validated early biomarker and at least moderate harms from false positives and treatment. It explicitly excluded people at high inherited or familial risk. New tests must be judged against that net-benefit problem; a retrospective area-under-the-curve result is not enough. This record will be revised if a later evidence review supersedes the recommendation.", "sourceIds": [ "uspstf-2019-pancreatic-screening" ], "links": [ { "target": "detection", "relation": "supports" }, { "target": "failures", "relation": "supports" }, { "target": "unknowns", "relation": "supports" } ], "evidence": { "humanProspective": "none", "humanRetrospective": "moderate", "randomised": "none", "independentReplication": "not-applicable", "sampleSize": "Evidence review rather than one cohort", "effect": "Grade D recommendation in the stated population", "limits": [ "Assessment published in 2019", "Does not apply to high inherited or familial risk" ] }, "layout": { "x": 19, "y": 108 }, "updated": "2026-09-14" }, { "id": "claim-germline-risk-subset", "title": "Known high-risk germline mutations explain an important minority of pancreatic cancer", "shortTitle": "Inherited-risk subset", "type": "claim", "status": "supported", "scope": [ "pancreatic-all" ], "summary": "Six established cancer-predisposition genes carried risk in a 3,030-case study, but mutations in those genes occurred in only 5.5% of cases.", "content": "A Mayo Clinic case-control sequencing study compared 3,030 people with pancreatic cancer against large public exome reference groups. Deleterious variants in CDKN2A, TP53, MLH1, BRCA2, ATM and BRCA1 were associated with increased risk. Mutations in these six genes were found in 5.5% of cases overall, including 5.2% without a pancreatic-cancer family history. This supports germline testing as a way to find a high-risk minority and can affect treatment for some DNA-repair defects. It cannot be the sole population-risk gate: 94.5% of cases did not carry a mutation in these six genes. The case group was 95.6% non-Hispanic White and used external reference controls, so exact prevalence and odds ratios need population-specific validation.", "sourceIds": [ "hu-2018-germline-risk" ], "links": [ { "target": "detection", "relation": "supports" }, { "target": "pdac", "relation": "part-of" }, { "target": "claim-high-risk-surveillance", "relation": "supports" } ], "evidence": { "humanProspective": "none", "humanRetrospective": "strong", "randomised": "not-applicable", "independentReplication": "strong", "sampleSize": "3,030 pancreatic cancer cases; 123,136 gnomAD and 53,105 ExAC reference individuals", "effect": "Six associated genes present in 5.5% of cases; gene-level odds ratios 2.58 to 12.33", "limits": [ "Predominantly White registry population", "External reference controls", "Pancreatic cancer histology not limited to PDAC in the abstract", "Association does not measure screening benefit" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The implicated genes have established roles in DNA repair, cell-cycle control or inherited cancer syndromes." }, "animalEvidence": { "rating": "moderate", "reason": "Models support gene function but are not needed to establish carrier prevalence in human cases." }, "inVitroEvidence": { "rating": "moderate", "reason": "Functional work supports damaging variants, while this record relies on human sequencing association." }, "reproducibility": { "rating": "strong", "reason": "The major genes and the existence of an inherited minority repeat across sequencing studies and clinical guidelines." }, "commercialConflicts": { "rating": "low", "reason": "The cited case-control study is not a test-company performance report, though testing markets can influence wider claims." }, "recency": { "rating": "recent", "reason": "The large cited sequencing study is from 2018 and the prevalence conclusion remains clinically relevant." } } }, "layout": { "x": 9, "y": 30 }, "updated": "2026-09-14" }, { "id": "claim-hcc-alcohol-interception-window", "title": "Alcohol-related HCC may have an interception window before liver decompensation", "shortTitle": "Alcohol HCC window", "type": "claim", "status": "contested", "scope": [ "hcc", "hepatobiliary" ], "summary": "In a prospective cirrhosis cohort, abstinence was associated with lower HCC incidence before prior decompensation but not after it; the nonrandomized result may contain behavioural and clinical confounding.", "content": "A Spanish surveillance cohort prospectively followed 727 people with alcohol-related cirrhosis for a median 54 months; 354 remained abstinent and 104 developed HCC. Among 247 people without previous liver decompensation, abstinence was independently associated with lower HCC incidence, HR 0.35, 95% CI 0.13 to 0.94. Among 480 people with previous decompensation, the association was absent, HR 0.95, 95% CI 0.59 to 1.52.\n\nThe pattern is consistent with an intervention window: removing the causal exposure may reduce new malignant initiation while the liver remains compensated, but may be too late to erase an advanced damaged field. It is not a randomized comparison. People able to maintain abstinence can differ in care, nutrition, comorbidity and disease severity, and the compensated estimate is based on a subgroup with a wide confidence interval. The result supports early alcohol-use treatment and liver-disease detection because those interventions already have broader health value; it does not justify promising a 65% causal reduction in HCC. Surveillance must continue after abstinence because risk remains.", "sourceIds": [ "rodriguez-2021-alcohol-abstinence-hcc" ], "links": [ { "target": "hcc", "relation": "part-of" }, { "target": "model-hcc-causal-chain", "relation": "supports" }, { "target": "claim-hcc-prevention-is-proven-for-viral-cause", "relation": "contrasts-with" }, { "target": "prevention-ranking", "relation": "supports" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "not-applicable", "randomised": "none", "independentReplication": "weak", "sampleSize": "727 people with alcohol-related cirrhosis; 247 without and 480 with previous decompensation", "effect": "Adjusted HR 0.35 before prior decompensation; HR 0.95 after prior decompensation", "limits": [ "Observational abstinence exposure", "Residual and behavioural confounding", "Subgroup interaction needs independent testing", "Wide confidence interval in compensated disease", "Does not show that established cirrhosis becomes low risk", "No evidence here for PDAC prevention" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Stopping alcohol can reduce continuing injury before decompensation fixes a high-risk liver state." }, "animalEvidence": { "rating": "moderate", "reason": "Alcohol-related injury models support the direction but cannot estimate the human cancer effect of abstinence." }, "inVitroEvidence": { "rating": "weak", "reason": "Cell studies cannot reproduce the long course from abstinence through cirrhosis to HCC." }, "reproducibility": { "rating": "weak", "reason": "The stage-specific HCC association comes from one prospective observational cohort." }, "commercialConflicts": { "rating": "low", "reason": "Abstinence is not a proprietary intervention." }, "recency": { "rating": "recent", "reason": "The cohort was published in 2021 and remains directly relevant to the interception question." } } }, "layout": { "x": 68, "y": 2 }, "updated": "2026-09-14" }, { "id": "claim-hcc-cares310-survival", "title": "A second immune–vascular combination extended HCC survival, but liver reserve and toxicity still bound the result", "shortTitle": "CARES-310 HCC gain", "type": "claim", "status": "supported", "scope": [ "hcc" ], "summary": "CARES-310 raised median survival from 15.2 to 23.8 months against sorafenib in first-line unresectable HCC; 25% had treatment-related serious adverse events.", "content": "CARES-310 randomized 543 untreated unresectable or metastatic HCC patients across 95 sites to camrelizumab plus rivoceranib or sorafenib. At final analysis, median overall survival was 23.8 versus 15.2 months, hazard ratio 0.64, and median progression-free survival was 5.6 versus 3.7 months, hazard ratio 0.54. The result independently supports the class-level idea that joining immune release to vascular control can improve HCC survival. It does not establish cure: median progression remained under six months. Treatment-related serious adverse events occurred in 25% versus 7%; hypertension, liver-enzyme rises and hand-foot syndrome were common. Eighty-three percent of participants were Asian, so global transport should be measured rather than assumed. Hengrui and Elevar funded the open-label trial. The result strengthens tumour control but does not answer whether active preservation of liver reserve adds benefit, because trial eligibility and subgroup results do not randomize liver care.", "sourceIds": [ "qin-2025-cares310-final", "finn-2020-imbrave150", "storandt-2024-atezo-bev-liver-function" ], "links": [ { "target": "hcc", "relation": "part-of" }, { "target": "claim-hcc-systemic-combination-improves-not-cures", "relation": "replicates" }, { "target": "claim-hcc-tumour-and-liver-are-competing-failure-axes", "relation": "supports" }, { "target": "company-elevar-therapeutics", "relation": "funded-by" }, { "target": "company-hengrui", "relation": "funded-by" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong", "independentReplication": "moderate", "sampleSize": "543 randomized first-line unresectable HCC patients at 95 sites in 13 countries and regions", "effect": "Median OS 23.8 versus 15.2 months, HR 0.64; median PFS 5.6 versus 3.7 months, HR 0.54", "limits": [ "Open-label", "Eighty-three percent Asian population", "Company-funded", "Comparator was sorafenib", "Treatment-related serious adverse events 25% versus 7%", "Liver-reserve intervention not tested" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "A second randomized immune–vascular regimen supports a reproducible HCC treatment class." }, "animalEvidence": { "rating": "not-applicable", "reason": "The survival claim rests on a large human phase 3 trial." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "The stated benefit does not depend on laboratory response." }, "reproducibility": { "rating": "moderate", "reason": "Another immune–vascular combination improved HCC survival, but this exact regimen has one phase 3 result." }, "commercialConflicts": { "rating": "high", "reason": "Hengrui and Elevar funded the study and hold programme rights." }, "recency": { "rating": "current", "reason": "The final survival analysis was published in December 2025." } } }, "updated": "2026-09-14" }, { "id": "claim-hcc-early-hbv-treatment-signal-immature", "title": "Earlier HBV treatment may prevent HCC, but the randomized signal is still immature", "shortTitle": "Early HBV treatment signal", "type": "claim", "status": "contested", "scope": [ "hcc", "hepatobiliary" ], "summary": "ATTENTION recorded two HCCs with tenofovir alafenamide and seven under observation, but only 11 composite events occurred and the interim result did not cross its stopping boundary.", "content": "ATTENTION tests whether antiviral treatment should start before noncirrhotic chronic HBV reaches the fibrosis or ALT thresholds used by many guidelines. Among 734 randomized adults with moderate or high viraemia, the composite of HCC, decompensation, transplantation or death occurred in 2/369 treated people = 0.54% and 9/365 observed people = 2.47% after median follow-up of 17.7 months. HCC alone occurred in two versus seven people, 0.54% versus 1.92%. The descriptive absolute HCC difference is 1.38 percentage points, equivalent to about 73 people treated per HCC avoided over the observed period if the small interim difference were stable. It is not a reliable treatment estimate: the composite HR was 0.21 with a 97.5% CI from 0.04 to 1.20, only 11 primary events occurred, and the prespecified early-stop boundary was not crossed. The South Korean government and Gilead funded the trial. The result strengthens the prevention route enough to await the planned analysis; it does not yet justify booking a population cancer effect or replacing final trial follow-up.", "sourceIds": [ "lim-2025-attention-hbv", "chang-1997-hbv-vaccine-hcc", "liaw-2004-lamivudine-hbv" ], "links": [ { "target": "claim-hcc-prevention-is-proven-for-viral-cause", "relation": "extends" }, { "target": "model-hcc-causal-chain", "relation": "supports" }, { "target": "trial-hcc-action-map", "relation": "informs" }, { "target": "prevention-ranking", "relation": "supports" } ], "evidence": { "humanProspective": "strong design, immature follow-up", "humanRetrospective": "not-applicable", "randomised": "strong design, sparse events", "independentReplication": "moderate for later-stage HBV treatment; weak for this early-treatment population", "sampleSize": "734 randomized; 11 composite events; 9 HCCs", "effect": "Composite HR 0.21, 97.5% CI 0.04-1.20; HCC 2 versus 7; early-stop boundary not crossed", "limits": [ "Interim analysis", "Only nine HCCs", "Confidence interval includes no effect", "South Korean and Taiwanese population", "Open observation comparator", "Gilead co-funding", "The descriptive number needed to treat is unstable" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Reducing HBV replication earlier can reduce continued liver injury and malignant selection." }, "animalEvidence": { "rating": "moderate", "reason": "Viral carcinogenesis models support cause control, but the treatment threshold is a human clinical question." }, "inVitroEvidence": { "rating": "strong", "reason": "Tenofovir alafenamide suppresses HBV replication, the intended upstream cause." }, "reproducibility": { "rating": "weak", "reason": "The early-treatment population has one sparse interim randomized result." }, "commercialConflicts": { "rating": "moderate", "reason": "The trial had government funding and Gilead co-funding; the drug maker benefits from broader treatment thresholds." }, "recency": { "rating": "current", "reason": "The interim analysis was published in 2025 and remains awaiting planned follow-up." } } }, "updated": "2026-09-14" }, { "id": "claim-hcc-local-control-long-survival-not-eradication", "title": "HCC local treatment can extend survival in selected states, while recurrence remains common", "shortTitle": "HCC local control", "type": "claim", "status": "supported", "scope": [ "hcc", "hepatobiliary" ], "summary": "SURF found surgery and radiofrequency ablation comparable for small operable HCC; TORCH found that ablation added after arterial chemoembolisation improved survival in selected liver-confined unresectable HCC. Neither removes the damaged liver field or recurrence risk.", "content": "SURF randomized 302 patients at 49 Japanese institutions when both liver resection and radiofrequency ablation were considered feasible for small HCC. Five-year overall survival was 74.6% with surgery and 70.4% with ablation, HR 0.96 after adjustment; five-year recurrence-free survival was 42.9% and 42.7%, HR 0.90. Serious adverse events occurred in 3.3% after surgery and none after ablation. A separate prospective nonrandomized cohort of 753 did not show a survival difference after adjustment. This establishes that a less invasive local treatment can achieve similar long survival in a carefully selected group, not that either eradicates the disease system.\n\nTORCH tested a different state: 241 people with liver-confined, unresectable BCLC stage B HCC at two Chinese tertiary centres. Adding selective radiofrequency ablation after arterial chemoembolisation increased median progression-free survival from 7.3 to 17.7 months, HR 0.47, and median overall survival from 35.1 to 88.6 months, HR 0.50. Grade 3 or 4 treatment-related events occurred in 23.2% versus 18.3%. This is a large randomized survival result for a selected local-treatment sequence. It does not transfer to extrahepatic disease, poor liver reserve, inaccessible tumour geometry or PDAC.\n\nMore than half of the small-HCC SURF population recurred by five years, consistent with both residual metastatic clones and new primaries arising in a damaged liver field. Treatment choice should therefore include stage, tumour geometry, liver reserve, repeatability and future transplant options. Recurrence work needs clonal classification rather than assuming every new lesion is the old cancer returning.", "sourceIds": [ "kawaguchi-2025-surf-hcc", "lyu-2026-torch-hcc", "zhang-2026-planet-hcc-recurrence", "yopp-2026-imbrave050-update" ], "links": [ { "target": "hcc", "relation": "part-of" }, { "target": "claim-hcc-recurrence-distinct-clonal-routes", "relation": "leads-to" }, { "target": "failure-sorafenib-adjuvant-hcc", "relation": "constrained-by" }, { "target": "failure-imbrave050-interim-benefit-disappeared", "relation": "constrained-by" }, { "target": "model-hcc-causal-chain", "relation": "supports" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong", "independentReplication": "moderate", "sampleSize": "SURF: 302 randomized and 753 in a parallel prospective cohort; TORCH: 241 randomized", "effect": "SURF five-year OS 74.6% versus 70.4%, adjusted HR 0.96; TORCH median OS 88.6 versus 35.1 months, HR 0.50", "limits": [ "The two trials address different HCC states", "SURF selected small HCC eligible for either procedure", "TORCH selected liver-confined BCLC stage B HCC at two Chinese centres", "TORCH was open-label", "Repeated treatment after recurrence affects OS", "Neither compares transplantation", "Neither transfers to pancreatic cancer" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Resection and ablation remove visible small tumours but leave the injured liver and possible occult clones." }, "animalEvidence": { "rating": "not-applicable", "reason": "The survival and recurrence comparison rests on randomized human treatment." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Laboratory response cannot establish long-term local-treatment outcomes." }, "reproducibility": { "rating": "moderate", "reason": "Two randomized trials support local treatment in different selected HCC states, but neither has a matching independent replication." }, "commercialConflicts": { "rating": "low", "reason": "The comparison does not depend on a proprietary drug." }, "recency": { "rating": "current", "reason": "The long-term SURF report was published in 2025; TORCH and recurrence evidence were published in 2026." } } }, "layout": { "x": 126, "y": 38 }, "updated": "2026-09-15" }, { "id": "claim-hcc-molecular-groups", "title": "HCC has molecular groups, but classification has not solved treatment selection", "shortTitle": "HCC groups", "type": "claim", "status": "supported", "scope": [ "hcc" ], "summary": "TCGA integrated five data types and found three HCC groups, including one associated with poorer prognosis in three cohorts.", "content": "The study establishes molecular diversity and links one group to adverse prognosis. It does not show that assigning a patient to a group selects a treatment that improves survival. HCC also develops in a diseased organ whose reserve and underlying cause constrain therapy; tumour classification must therefore be joined to liver function and cause rather than used alone.", "sourceIds": [ "tcga-2017-hcc" ], "links": [ { "target": "hepatobiliary", "relation": "part-of" }, { "target": "disease-system", "relation": "supports" }, { "target": "treatment", "relation": "depends-on" }, { "target": "unknowns", "relation": "supports" } ], "evidence": { "humanProspective": "none", "humanRetrospective": "moderate", "randomised": "none", "independentReplication": "moderate", "sampleSize": "363 patients for mutation and copy number; 196 for five-platform analysis", "effect": "Three integrated groups; one adverse group reproduced prognostically across three cohorts", "limits": [ "Retrospective tissue cohort", "Prognosis is not predictive treatment utility", "Underlying liver disease is not captured by tumour class alone" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Distinct genomic and expression states are expected across HCC causes and evolutionary routes." }, "animalEvidence": { "rating": "moderate", "reason": "Models reproduce parts of HCC diversity but not the linked human cause and liver state." }, "inVitroEvidence": { "rating": "moderate", "reason": "Tumour systems can reproduce pathway differences but do not establish treatment selection." }, "reproducibility": { "rating": "moderate", "reason": "One adverse group was prognostic in three cohorts, but no group has broad prospective treatment utility." }, "commercialConflicts": { "rating": "low", "reason": "The TCGA classification was a public consortium analysis." }, "recency": { "rating": "historical", "reason": "The integrated classification was published in 2017 and must be read against newer treatment states." } } }, "layout": { "x": 87, "y": 13 }, "updated": "2026-09-14" }, { "id": "claim-hcc-prevention-is-proven-for-viral-cause", "title": "HCC prevention is possible when an upstream infectious cause can be removed or suppressed", "shortTitle": "HCC prevention proof", "type": "claim", "status": "supported", "scope": [ "hcc" ], "summary": "Taiwan's hepatitis B vaccination programme cut childhood HCC incidence, and antiviral treatment reduced HCC and liver-disease progression in advanced chronic HBV.", "content": "After Taiwan introduced universal infant hepatitis B vaccination in 1984, annual HCC incidence among children aged 6 to 14 fell from 0.70 to 0.36 per 100,000 across successive periods; among children aged 6 to 9 it fell from 0.52 in earlier birth cohorts to 0.13 after the programme. In a separate randomised trial of 651 people with advanced chronic hepatitis B, lamivudine reduced the composite liver-disease endpoint from 17.7% to 7.8% and HCC from 7.4% to 3.9% during shortened follow-up. This is proof that at least one cancer in scope is partly preventable by attacking an upstream cause. It does not prevent HCC driven by hepatitis C, alcohol, metabolic liver disease or persistent cirrhosis, and modern antiviral regimens have different resistance profiles.", "sourceIds": [ "chang-1997-hbv-vaccine-hcc", "liaw-2004-lamivudine-hbv" ], "links": [ { "target": "hcc", "relation": "part-of" }, { "target": "detection", "relation": "supports" }, { "target": "transfer", "relation": "supports" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "strong", "randomised": "strong", "independentReplication": "strong", "sampleSize": "Nationwide Taiwan childhood registry; 651 randomised chronic-HBV patients", "effect": "Child HCC incidence fell roughly 49%; lamivudine HCC 3.9% versus 7.4% and composite progression 7.8% versus 17.7%", "limits": [ "Vaccination evidence is a natural experiment", "Childhood HCC is rare", "Antiviral trial was 98% Asian and stopped early", "Older lamivudine had 49% resistance", "Does not apply to non-viral HCC causes" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Vaccination prevents HBV infection and antivirals reduce replication-driven injury upstream of malignant change." }, "animalEvidence": { "rating": "strong", "reason": "Viral liver-cancer models support persistent infection as a cause, though the stated effects are human." }, "inVitroEvidence": { "rating": "strong", "reason": "Vaccine immunity and antiviral suppression act on the intended viral cause." }, "reproducibility": { "rating": "strong", "reason": "A nationwide vaccination change and a randomized antiviral trial support the same upstream direction." }, "commercialConflicts": { "rating": "moderate", "reason": "Population vaccination evidence is public; antiviral expansion creates drug-company interests." }, "recency": { "rating": "mixed", "reason": "The decisive records are older but newer ATTENTION evidence tests an earlier treatment boundary." } } }, "layout": { "x": 74, "y": 6 }, "updated": "2026-09-14" }, { "id": "claim-hcc-recurrence-distinct-clonal-routes", "title": "HCC recurrence contains distinct clonal routes that should not share one adjuvant treatment", "shortTitle": "HCC recurrence routes", "type": "claim", "status": "supported", "scope": [ "hcc", "hepatobiliary" ], "summary": "PLANet found polyclonal and monoclonal recurrence patterns in paired multiregion tumours, linking early intrahepatic recurrence and distant metastasis to different routes.", "content": "The Asia-Pacific PLANet cohort performed multiregion whole-genome sequencing on 408 samples and RNA sequencing on 406 samples from 106 primary HCCs. Twenty-four patients had paired primary and recurrent tumours with deeper multiregion analysis. More than half of intrahepatic recurrences were polyclonal, associated with earlier recurrence, greater state variation and a regulatory-T-cell-rich immune environment. Most distant metastases appeared monoclonal and traced to a dominant aggressive primary subclone. This is a stronger recurrence map than time-to-event alone, but it is still observational and the paired recurrence group is small. It supports a treatment design rule: classify whether recurrence is seeded from the primary, polyclonal or a new field cancer, then match systemic, immune, local and cause-control strategies. It does not yet validate the reported treatment-sensitivity associations. A prospective adjuvant trial should preserve multiregion primary tissue and require recurrence biopsy so that failure can be assigned to the route actually treated.", "sourceIds": [ "zhang-2026-planet-hcc-recurrence", "yopp-2026-imbrave050-update", "bruix-2015-storm-sorafenib" ], "links": [ { "target": "hcc", "relation": "part-of" }, { "target": "claim-hcc-local-control-long-survival-not-eradication", "relation": "explains" }, { "target": "failure-sorafenib-adjuvant-hcc", "relation": "possible-cause-of" }, { "target": "failure-imbrave050-interim-benefit-disappeared", "relation": "possible-cause-of" }, { "target": "model-hcc-causal-chain", "relation": "supports" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "strong", "randomised": "none", "independentReplication": "moderate", "sampleSize": "106 primary tumours; 24 patients with paired recurrent tumours and multiregion sampling", "effect": "More than half of intrahepatic recurrences polyclonal; most distant metastases monoclonal in this cohort", "limits": [ "Only 24 paired recurrences", "Observational treatment-sensitivity associations", "Surgical cohort selection", "Geographic and cause mix may not transport", "No validated clinical classifier" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Recurrence can arise from surviving tumour clones or a new primary tumour in the injured liver." }, "animalEvidence": { "rating": "moderate", "reason": "Models can trace clonal spread but do not reproduce years of human field injury." }, "inVitroEvidence": { "rating": "moderate", "reason": "Functional systems can test clone-specific sensitivity after the routes are identified." }, "reproducibility": { "rating": "weak", "reason": "Only 24 paired recurrences carried the main multiregion route analysis." }, "commercialConflicts": { "rating": "low", "reason": "The route claim comes from academic genomic research rather than a product trial." }, "recency": { "rating": "current", "reason": "The PLANet analysis was published in 2026." } } }, "layout": { "x": 132, "y": 42 }, "updated": "2026-09-14" }, { "id": "claim-hcc-repeat-treatment-sequence", "title": "Recurrent HCC needs a repeatable treatment sequence, not a single best procedure", "shortTitle": "Repeat HCC treatment", "type": "claim", "status": "supported", "scope": [ "hcc", "hepatobiliary" ], "summary": "In 240 selected patients with early recurrent HCC, repeat surgery did not significantly improve survival over ablation; it reduced local repeat recurrence but increased complications.", "content": "A Chinese open-label trial randomized 240 people with early recurrent HCC after an initial complete resection to repeat hepatectomy or percutaneous radiofrequency ablation. All had Child-Pugh A liver function, no macrovascular or distant disease, and tumour geometry technically treatable by either approach. Five-year overall survival was 43.6% after repeat surgery and 38.5% after ablation, P=0.17. Five-year repeat recurrence-free survival was 36.2% and 30.2%, P=0.09. Neither primary comparison showed a significant advantage.\n\nThe trade-off was real. Ablation produced more local repeat recurrence, 37.8% versus 21.7%, while surgery produced more complications, 22.4% versus 7.3%. That is roughly 16 fewer local repeat recurrences but 15 more complications per 100 treated, before allowing for uncertainty and different severity. Apparent survival advantages for surgery in tumours larger than 3 cm or AFP above 200 ng/mL came from subgroup analyses and should be tested rather than treated as rules. Because another recurrence remained common after both choices, the mission is to optimize a sequence across surveillance, tumour geometry, liver reserve, previous treatment, clonal origin and future treatment options. A procedure that maximizes control today can still be inferior if it removes the liver reserve needed for the next recurrence.", "sourceIds": [ "xia-2020-repeat-hcc-treatment", "zhang-2026-planet-hcc-recurrence" ], "links": [ { "target": "hcc", "relation": "part-of" }, { "target": "model-hcc-causal-chain", "relation": "supports" }, { "target": "claim-hcc-local-control-long-survival-not-eradication", "relation": "extends" }, { "target": "claim-hcc-recurrence-distinct-clonal-routes", "relation": "depends-on" }, { "target": "claim-hcc-transplant-selection", "relation": "depends-on" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong", "independentReplication": "weak", "sampleSize": "240 randomized; 217 completed the trial", "effect": "Five-year OS 43.6% versus 38.5%, P=0.17; local repeat recurrence 21.7% versus 37.8%; complications 22.4% versus 7.3% for surgery versus ablation", "limits": [ "Single Chinese centre system", "Ninety percent male", "Age limited to 20 to 60", "Only Child-Pugh A liver function", "Only small disease technically treatable by either route", "Open-label", "Subgroup findings were exploratory", "Post-publication correction to Figure 3" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Repeated intrahepatic recurrence makes preservation of liver tissue and future treatment options central." }, "animalEvidence": { "rating": "not-applicable", "reason": "The procedure trade-off is established through a human randomized comparison." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Cell response cannot estimate procedure complications or later treatment access." }, "reproducibility": { "rating": "weak", "reason": "The randomized comparison came from one selected Chinese centre system." }, "commercialConflicts": { "rating": "low", "reason": "Neither procedure requires a proprietary systemic drug." }, "recency": { "rating": "recent", "reason": "The randomized report was published in 2020 and is interpreted with 2026 recurrence-route evidence." } } }, "layout": { "x": 112, "y": 31 }, "updated": "2026-09-14" }, { "id": "claim-hcc-stride-five-year-tail", "title": "STRIDE doubled five-year survival against sorafenib in selected unresectable HCC, but four in five still died", "shortTitle": "HIMALAYA five-year tail", "type": "claim", "status": "supported", "scope": [ "hcc" ], "summary": "In HIMALAYA, five-year survival was 19.6% with tremelimumab plus durvalumab versus 9.4% with sorafenib; the death hazard ratio was 0.76.", "content": "HIMALAYA randomised 393 people to the STRIDE tremelimumab–durvalumab regimen and 389 to sorafenib in the comparison used here. At about five years' median follow-up, the death hazard ratio was 0.76 and 60-month overall survival was 19.6% versus 9.4%. The absolute five-year gain was 10.2 percentage points and the survival rate was about 2.09 times the sorafenib rate. This is durable human survival evidence, not just response. It is still not reliable control for most patients: 80.4% in the STRIDE group was not alive at five years. The five-year report is an exploratory update of the same randomized trial and not independent replication. Eligibility and liver reserve restrict how far the result travels into routine HCC; no inference should be transferred directly to PDAC, where unselected checkpoint therapy has almost no activity. The decision lesson is narrower: long survival tails are possible when immune treatment fits the disease and host setting, so HCC programmes should measure both tumour control and liver failure rather than treating all death as one process.", "sourceIds": [ "rimassa-2025-himalaya-five-year" ], "links": [ { "target": "hcc", "relation": "part-of" }, { "target": "treatment", "relation": "supports" }, { "target": "company-astrazeneca", "relation": "sponsored-by" }, { "target": "claim-hcc-systemic-combination-improves-not-cures", "relation": "extends" }, { "target": "failure-checkpoint-blockade-unselected-pdac", "relation": "contrasts-with" }, { "target": "hypothesis-hcc-liver-reserve-co-control", "relation": "supports" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong", "independentReplication": "weak — five-year update of the same trial", "sampleSize": "STRIDE n=393; sorafenib n=389", "effect": "Death HR 0.76; 60-month OS 19.6% versus 9.4%; absolute gain 10.2 percentage points", "limits": [ "Exploratory long-term update", "Same trial rather than independent replication", "Selected unresectable-HCC population", "Liver reserve affects transport to routine care", "AstraZeneca staff and financial ties", "Does not transfer directly to PDAC" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "Dual checkpoint treatment can deepen immune activation, but the responder biology is not fully resolved." }, "animalEvidence": { "rating": "not-applicable", "reason": "The survival claim is established by randomized human follow-up rather than animal evidence." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Laboratory activity is not needed to establish the observed human survival tail." }, "reproducibility": { "rating": "weak", "reason": "Five-year follow-up strengthens durability inside HIMALAYA but is not an independent repeat." }, "commercialConflicts": { "rating": "high", "reason": "AstraZeneca staff contributed and many investigators disclosed company relationships." }, "recency": { "rating": "recent", "reason": "The five-year analysis was reported in 2025." } } }, "layout": { "x": 83, "y": 17 }, "updated": "2026-09-14" }, { "id": "claim-hcc-surveillance-mortality-signal", "title": "Six-month HCC surveillance reduced HCC mortality in one large Chinese randomised trial", "shortTitle": "HCC surveillance signal", "type": "claim", "status": "supported", "scope": [ "hcc" ], "summary": "AFP plus ultrasound every six months reduced HCC mortality from 131.5 to 83.2 per 100,000 in 18,816 high-risk people despite 58.2% attendance.", "content": "A Shanghai trial randomised 18,816 people aged 35 to 59 with hepatitis B infection or chronic hepatitis to six-month alpha-fetoprotein plus ultrasound invitations or usual care. Screening attendance was only 58.2%. HCC mortality was 83.2 versus 131.5 per 100,000, a mortality rate ratio of 0.63, and resection was performed in 46.5% versus 7.5% of detected cancers. The mortality endpoint makes this evidence stronger than a stage-shift study. Important limits remain: one older urban Chinese trial, incomplete attendance, changing imaging and treatment, and risk dominated by viral hepatitis. The result supports targeted organ surveillance where disease prevalence is high; it does not validate average-risk screening or direct transfer to PDAC.", "sourceIds": [ "zhang-2004-hcc-screening" ], "links": [ { "target": "hcc", "relation": "part-of" }, { "target": "detection", "relation": "supports" }, { "target": "claim-high-risk-surveillance", "relation": "may-transfer-to" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong", "independentReplication": "weak", "sampleSize": "18,816 randomised high-risk participants", "effect": "HCC mortality 83.2 versus 131.5 per 100,000; mortality rate ratio 0.63, 95% CI 0.41–0.98", "limits": [ "Single older trial", "58.2% attendance", "Urban Shanghai high-risk population", "Screening and treatment have changed", "Does not separate ultrasound from AFP contribution" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Repeated surveillance can reduce HCC death only if detection advances access to curative treatment." }, "animalEvidence": { "rating": "not-applicable", "reason": "Screening attendance, work-up and mortality require human evidence." }, "inVitroEvidence": { "rating": "weak", "reason": "AFP biology cannot establish the programme-level mortality result." }, "reproducibility": { "rating": "weak", "reason": "The mortality signal rests on one older randomized programme trial." }, "commercialConflicts": { "rating": "low", "reason": "The trial tested an organised programme rather than a proprietary new test." }, "recency": { "rating": "historical", "reason": "The report dates from 2004 and predates current imaging and treatment." } } }, "layout": { "x": 70, "y": 14 }, "updated": "2026-09-14" }, { "id": "claim-hcc-surveillance-test-improvement-not-mortality", "title": "A better HCC surveillance test does not yet mean fewer deaths", "shortTitle": "HCC test gain is not mortality gain", "type": "claim", "status": "supported", "scope": [ "hcc", "hepatobiliary" ], "summary": "Randomized surveillance trials have changed stage, false referrals or lesion counts without yet showing that a newer test reduces HCC mortality.", "content": "The evidence separates the value of surveillance from the value of a particular test. The older Shanghai trial found an HCC-mortality rate ratio of 0.63 for a six-month ultrasound-and-AFP programme versus no organised screening, but it cannot tell whether a new scan or marker is better. In the 414-person MIRACLE-HCC trial, noncontrast MRI produced more BCLC stage 0 diagnoses than ultrasound, 8% versus 2%, and fewer false-positive referrals, 0.7% versus 3.1%. Yet BCLC 0/A detection rates were 12% versus 7% and not statistically different, and the paper did not report a mortality comparison. In 1,208 high-risk adults, adding AFP, AFP-L3 and DCP to ultrasound did not improve early-stage detection: HR 0.81, 95% CI 0.47 to 1.40. Scanning every three rather than six months likewise produced more lesions 10 mm or smaller but no significant gain in HCC 30 mm or smaller. These trials rule out an easy substitution claim. The useful endpoint chain is test completion, false work-up, late-stage HCC, curative treatment, liver decompensation, HCC death and all-cause death. TRACER tests late-stage disease and harms; PREMIUM is powered for HCC mortality. Until those results mature, stage and referral gains should be reported as intermediate outcomes, not lives saved.", "sourceIds": [ "zhang-2004-hcc-screening", "rhee-2025-miracle-hcc", "hirode-2026-hcc-biomarker-rct", "trinchet-2011-hcc-surveillance-interval", "singal-2024-tracer-protocol", "clinicaltrials-nct06084234-tracer", "ioannou-2026-premium-protocol", "clinicaltrials-nct05486572-premium" ], "links": [ { "target": "claim-hcc-surveillance-mortality-signal", "relation": "constrains" }, { "target": "failure-hcc-three-month-ultrasound-no-small-cancer-gain", "relation": "supported-by" }, { "target": "failure-hcc-biomarker-addition-no-early-stage-gain", "relation": "supported-by" }, { "target": "trial-hcc-action-map", "relation": "informs" }, { "target": "unknown-hcc-surveillance-modality-utility", "relation": "leads-to" }, { "target": "detection", "relation": "part-of" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong", "independentReplication": "moderate across different test changes; no repeat mortality result", "sampleSize": "Shanghai n=18,816; MIRACLE-HCC n=414; biomarker RCT n=1,208; interval RCT n=1,278", "effect": "MRI false referrals 0.7% versus 3.1%; added biomarkers early-stage HR 0.81; three-month scanning found more tiny lesions without more small HCC", "limits": [ "Only the older Shanghai programme reported an HCC-mortality comparison", "The newer trials changed different parts of the surveillance route", "MIRACLE-HCC was single-centre and restricted to Child-Pugh A cirrhosis", "The biomarker trial was underpowered for comparisons beyond ultrasound plus AFP", "Treatment access can change the value of the same detection test", "TRACER and PREMIUM have no results yet" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Earlier or clearer detection can help only when it changes treatment before tumour or liver failure closes the route." }, "animalEvidence": { "rating": "not-applicable", "reason": "Animal models cannot estimate screening attendance, work-up harm or cancer mortality." }, "inVitroEvidence": { "rating": "weak", "reason": "Assay performance can support marker selection but cannot prove benefit across the care route." }, "reproducibility": { "rating": "moderate", "reason": "Several randomized trials show that test-level gains can fail to improve the next clinical endpoint, but mortality evidence is sparse." }, "commercialConflicts": { "rating": "moderate", "reason": "The biomarker paper includes Fujifilm-affiliated authors and TRACER investigators disclose assay-company relationships; PREMIUM is publicly sponsored." }, "recency": { "rating": "current", "reason": "Completed trials and live registries were checked through September 2026." } } }, "updated": "2026-09-14" }, { "id": "claim-hcc-systemic-combination-improves-not-cures", "title": "Atezolizumab plus bevacizumab changed advanced HCC survival but did not make it durably controllable for most", "shortTitle": "HCC systemic gain", "type": "claim", "status": "supported", "scope": [ "hcc" ], "summary": "IMbrave150 reduced the death hazard by 42% versus sorafenib and raised one-year survival to 67.2%, while median progression-free survival remained 6.8 months.", "content": "IMbrave150 randomised 501 untreated patients with unresectable HCC to atezolizumab plus bevacizumab or sorafenib. At the first analysis, the death hazard ratio was 0.58 and one-year overall survival was 67.2% versus 54.6%. Median progression-free survival was 6.8 versus 4.3 months. This is a large advance compared with many PDAC trials and shows that changing both immune inhibition and tumour vasculature can matter in HCC. It is not cure: most disease still progressed within a year, eligibility excluded some patients with poor liver function or bleeding risk, and severe adverse events affected more than half of both groups. The contrast with PDAC checkpoint failure points to different immune, vascular and liver contexts rather than a generic immunotherapy effect.", "sourceIds": [ "finn-2020-imbrave150" ], "links": [ { "target": "hcc", "relation": "part-of" }, { "target": "treatment", "relation": "supports" }, { "target": "failure-checkpoint-blockade-unselected-pdac", "relation": "contradicts" }, { "target": "transfer", "relation": "supports" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong", "independentReplication": "strong", "sampleSize": "501 randomised patients with unresectable HCC", "effect": "Death HR 0.58; 12-month OS 67.2% versus 54.6%; median PFS 6.8 versus 4.3 months", "limits": [ "Open-label", "Selected liver function and bleeding risk", "First-analysis follow-up", "Roche/Genentech-funded", "HCC biology does not directly transfer to PDAC" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Vascular and immune treatment can act together in the immune-active, highly vascular HCC setting." }, "animalEvidence": { "rating": "moderate", "reason": "Models support immune and vascular interaction but do not predict the survival size in cirrhotic patients." }, "inVitroEvidence": { "rating": "strong", "reason": "The targets and direct drug activities are established, though clinical benefit depends on the whole tumour and liver." }, "reproducibility": { "rating": "strong", "reason": "CARES-310 and HIMALAYA support survival benefit from other immune combinations, although each regimen differs." }, "commercialConflicts": { "rating": "high", "reason": "Roche and Genentech funded IMbrave150 and hold the treatment interests." }, "recency": { "rating": "recent", "reason": "The pivotal 2020 result now has newer class-level survival comparisons." } } }, "layout": { "x": 79, "y": 15 }, "updated": "2026-09-14" }, { "id": "claim-hcc-transplant-selection", "title": "HCC transplantation can remove tumour and the injured liver, but the durable result depends on narrow selection", "shortTitle": "HCC transplant selection", "type": "claim", "status": "supported", "scope": [ "hcc", "hepatobiliary" ], "summary": "The original Milan series reported 92% four-year recurrence-free survival in 35 patients whose explants met strict tumour limits, compared with 59% in 13 patients beyond them.", "content": "Liver transplantation differs from resection or ablation because it removes the visible HCC and the chronically injured liver in which another primary tumour could arise. The 1996 Milan series prospectively followed 48 people with cirrhosis and small unresectable HCC. Eligibility used one tumour no larger than 5 cm, or no more than three tumours each no larger than 3 cm, without known spread. Among the 35 patients whose explanted livers met those limits, actuarial four-year overall survival was 85% and recurrence-free survival was 92%. Among 13 whose pathology exceeded the limits, the corresponding estimates were 50% and 59%.\n\nThis is evidence that durable elimination is possible in a selected HCC state, not proof that transplantation will cure an unselected population. The study was a 48-person, single-centre, nonrandomized series; median follow-up was only 26 months, staging before 1996 transplantation was less sensitive than current imaging, and 33 patients received treatment before transplant. Donor supply, waiting-list progression, operative mortality, lifelong immunosuppression and competing liver-disease mortality constrain scale. The operational problem is therefore a selection-and-allocation problem: identify disease still confined enough for transplant before it progresses, while showing that the expected survival gain justifies a scarce organ.", "sourceIds": [ "mazzaferro-1996-hcc-transplant" ], "links": [ { "target": "hcc", "relation": "part-of" }, { "target": "model-hcc-causal-chain", "relation": "supports" }, { "target": "claim-hcc-local-control-long-survival-not-eradication", "relation": "extends" }, { "target": "claim-hcc-recurrence-distinct-clonal-routes", "relation": "depends-on" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "not-applicable", "randomised": "none", "independentReplication": "moderate", "sampleSize": "48 transplanted patients; 35 met the prespecified pathological limits and 13 exceeded them", "effect": "Four-year OS 85% and recurrence-free survival 92% within limits, versus 50% and 59% beyond limits", "limits": [ "Single-centre nonrandomized series", "Small sample", "Median follow-up 26 months", "Selection was confirmed on explant pathology rather than entirely before treatment", "Most patients had viral cirrhosis", "Donor scarcity and waiting-list selection limit population reach" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Transplant can remove both the visible tumour and the chronically injured liver that can generate new HCC." }, "animalEvidence": { "rating": "not-applicable", "reason": "Human graft allocation, selection and recurrence determine the claim." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Laboratory systems cannot estimate transplant survival or donor constraints." }, "reproducibility": { "rating": "moderate", "reason": "The original small series shaped a widely repeated selection result, but it was not randomized." }, "commercialConflicts": { "rating": "low", "reason": "The finding concerns clinical selection and scarce organs rather than a proprietary therapy." }, "recency": { "rating": "historical", "reason": "The source is the 1996 Milan series and should be treated as historical selection evidence." } } }, "layout": { "x": 106, "y": 24 }, "updated": "2026-09-14" }, { "id": "claim-hcc-tumour-and-liver-are-competing-failure-axes", "title": "HCC control has two failure axes: malignant burden and liver reserve", "shortTitle": "HCC dual failure axes", "type": "claim", "status": "supported", "scope": [ "hcc", "hepatobiliary" ], "summary": "Early and advanced HCC cohorts show that liver dysfunction is not background noise: it independently changes cause of death and the survival observed under cancer treatment.", "content": "A four-health-system cohort followed 1,336 people with BCLC 0/A HCC. Of 598 deaths, 220 were assigned to HCC, 114 to non-HCC liver disease and 179 to other causes. Thus non-HCC liver deaths were about 52 for every 100 HCC deaths, before counting unassigned deaths. Child-Pugh B carried a 2.64-fold liver-death hazard and ALBI grade 2 a 2.47-fold hazard versus grade 1; MASLD cause carried a 1.99-fold hazard. These are adjusted observational associations, not intervention effects.\n\nThe same split persists in advanced treatment. IMbrave150 established atezolizumab plus bevacizumab in selected patients with preserved liver function. A later 322-person multi-institutional cohort found median survival of 21.6 months for Child-Pugh A, 9.1 months for B7 and 4.7 months for B8-C12. Within Child-Pugh A, ALBI grade 1 versus 2 separated median survival at 34.9 versus 14.2 months. The comparison mixes tumour burden, treatment selection and organ reserve and cannot estimate treatment benefit by class. It does show why an HCC programme must model two changing states.\n\nThe practical rule is to record tumour state and liver state at every decision. A treatment earns credit only if it preserves or improves both total survival and usable liver reserve. Trials should report death from tumour, hepatic decompensation and other causes; time to first decompensation; ALBI and Child-Pugh change; dose intensity; and the next treatment actually received. A tumour response that consumes the reserve needed for later treatment is not durable control.", "sourceIds": [ "seif-el-dahan-2026-early-hcc-causes-death", "storandt-2024-atezo-bev-liver-function", "finn-2020-imbrave150" ], "links": [ { "target": "hcc", "relation": "part-of" }, { "target": "model-hcc-causal-chain", "relation": "supports" }, { "target": "claim-hcc-systemic-combination-improves-not-cures", "relation": "constrains" }, { "target": "claim-hcc-repeat-treatment-sequence", "relation": "constrains" }, { "target": "hypothesis-hcc-liver-reserve-co-control", "relation": "motivates" } ], "evidence": { "humanProspective": "weak", "humanRetrospective": "strong", "randomised": "strong for systemic tumour treatment; none for integrated liver-preservation strategy", "independentReplication": "moderate", "sampleSize": "Early HCC n=1,336 with 598 deaths; advanced treatment cohort n=322; IMbrave150 n=501", "effect": "Early HCC: 114 liver-related versus 220 HCC-related deaths; advanced cohort median OS 21.6 months Child-Pugh A, 9.1 B7 and 4.7 B8-C12", "limits": [ "Cause-of-death assignment is retrospective", "Liver-function groups are not randomized", "Confounding by tumour burden and treatment selection", "High-volume US centres", "Integrated intervention not tested" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "HCC grows in an organ whose failure can independently cause death and remove treatment options." }, "animalEvidence": { "rating": "moderate", "reason": "Models reproduce tumour growth and liver injury but not the full human treatment sequence." }, "inVitroEvidence": { "rating": "weak", "reason": "Cell systems omit portal hypertension, hepatic reserve and competing liver death." }, "reproducibility": { "rating": "moderate", "reason": "Independent early and advanced cohorts show the same liver-reserve constraint, but not its modifiability." }, "commercialConflicts": { "rating": "moderate", "reason": "The prognostic cohorts are mainly academic; one linked treatment trial is company-funded." }, "recency": { "rating": "current", "reason": "The cause-of-death analysis was published in 2026 and the treatment cohort is recent." } } }, "layout": { "x": 81, "y": 17 }, "updated": "2026-09-14" }, { "id": "claim-hcq-combinations-fail-pdac-not-autophagy", "title": "Hydroxychloroquine combinations failed in PDAC; that does not test every autophagy target", "shortTitle": "Autophagy drug boundary", "type": "claim", "status": "supported", "scope": [ "pdac", "cholangiocarcinoma", "pan-cancer", "transfer-unproven" ], "summary": "Three hydroxychloroquine combinations produced no useful pancreatic-cancer response signal, while the only direct ULK1/2 programme ended without final results.", "content": "Plain definition. Autophagy is a cell's controlled recycling process: damaged or surplus parts are packed, delivered to a lysosome, broken down and reused. The lysosome is the acidic recycling compartment. Pancreatic cancer models suggested that blocking ERK or MEK, parts of the RAS growth-control circuit, makes tumour cells depend more heavily on this recycling route. Hydroxychloroquine, or HCQ, disrupts the lysosome but is a broad and incomplete way to block the whole process. Direct contribution test. NCT04386057 randomized 20 previously treated metastatic PDAC patients to LY3214996, an ERK inhibitor, plus HCQ and 19 to the ERK inhibitor alone. Four-month disease control was 1/20 = 5.0% versus 1/19 = 5.3%. Median progression-free survival was 1.3 versus 1.9 months and median overall survival 2.4 versus 4.6 months. These small survival estimates are not a powered harm comparison, but they rule out the hoped-for 20% disease-control signal. The combination also used 200 mg of LY3214996 rather than 400 mg alone, so added HCQ consumed treatment capacity. Independent direction. NCT05221320 gave ulixertinib, another ERK inhibitor, plus HCQ to 18 pancreatic-cancer patients: 0 responded, median progression-free survival was 1.9 months among 16 and 10/18 had a serious adverse event. MEKiAUTO treated 14 PDAC patients with cobimetinib plus HCQ, adding atezolizumab in four. Four of ten dose-limiting-toxicity-evaluable patients had a dose-limiting toxicity, 7/14 had a grade 3 or worse treatment-related event, best response was stable disease, and the planned phase 2 cohorts never opened. Boundary. These results reject HCQ plus these ERK or MEK drugs in these advanced patients. They do not prove that tumour-cell recycling is dispensable. The randomized study could not show consistent autophagy inhibition in blood cells, had no reliable measure of blockade in tumour, and its eleven tested patient-derived organoids showed no clear synergy. A direct ULK1/2 inhibitor, inlexisertib, showed blood-cell target inhibition in an early sponsor report, but its 91-person trial later ended for a stated business decision and posted no final results. The correct state is: failed agents, unresolved target class.", "sourceIds": [ "surana-2025-erk-hcq-pdac", "clinicaltrials-nct05221320-ulixertinib-hcq", "manji-2023-mekiauto", "clinicaltrials-nct04214418-mekiauto", "clinicaltrials-nct04892017-inlexisertib", "deciphera-2022-dcc3116-phase1", "karasic-2019-hydroxychloroquine-pdac" ], "links": [ { "target": "treatment", "relation": "part-of" }, { "target": "failure-metabolic-repurposing-pdac", "relation": "supported-by" }, { "target": "claim-pdac-metabolic-adaptation-is-redundant", "relation": "bounded-by" }, { "target": "unknown-selective-autophagy-pdac-contribution", "relation": "leaves-open" }, { "target": "trial-pdac-autophagy-action-map", "relation": "mapped-by" }, { "target": "hypothesis-tumour-verified-selective-autophagy-blockade", "relation": "gates" } ], "evidence": { "humanProspective": "strong — three prospective PDAC programmes and one direct randomized contribution test", "humanRetrospective": "none needed", "randomised": "moderate — 39 treated patients randomized between ERK inhibitor plus HCQ and ERK inhibitor alone", "independentReplication": "moderate for failure of HCQ combinations; none for a selective tumour-confirmed inhibitor", "sampleSize": "Randomized n=39 after safety run-in n=13; ulixertinib-HCQ pancreatic n=18; MEKiAUTO treated n=14; inlexisertib whole trial n=91 without posted final results", "effect": "Four-month disease control 5.0% with ERK plus HCQ versus 5.3% with ERK alone; 0/18 responses with ulixertinib plus HCQ", "limits": [ "Small late-line cohorts", "Combination arm reduced the ERK-inhibitor dose", "HCQ is not a selective autophagy target", "Tumour target engagement was not established", "The inlexisertib final dataset is unavailable", "A business termination is not a biological result" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "RAS-driven PDAC can use autophagy in models, but human tumour dependence and degree of blockade were not established." }, "animalEvidence": { "rating": "moderate", "reason": "Model combinations motivated the trials but did not predict their clinical results." }, "inVitroEvidence": { "rating": "weak", "reason": "Earlier models reported synergy; eleven trial-derived organoids did not reproduce clear synergy." }, "reproducibility": { "rating": "moderate", "reason": "Lack of useful HCQ-combination activity repeats across separate ERK and MEK programmes." }, "commercialConflicts": { "rating": "moderate", "reason": "Several drugs and studies had sponsor interests; the strongest negative randomized result was investigator-led and openly reported." }, "recency": { "rating": "current", "reason": "Includes the April 2026 registry termination and September 2025 randomized publication." } } }, "layout": { "x": 69, "y": 139 }, "updated": "2026-09-15" }, { "id": "claim-heat-regional-hyperthermia-primary-endpoint-negative", "title": "HEAT did not establish better disease-free survival after pancreatic cancer surgery", "shortTitle": "HEAT primary result negative", "type": "claim", "status": "supported", "scope": [ "pdac", "resected", "regional-hyperthermia" ], "summary": "A 117-person randomized trial missed its main endpoint, and its multi-part treatment change cannot identify the effect of heating itself.", "content": "Regional hyperthermia means controlled heating of a deep body region, usually to about 40-43 degrees C. It aims to make chemotherapy or radiation more damaging to cancer cells. It is not thermal ablation, which tries to destroy tissue with much higher focal heat. HEAT randomized 117 eligible people after pancreatic ductal adenocarcinoma surgery: 57 to gemcitabine, cisplatin and regional hyperthermia and 60 to a control treatment. Median disease-free survival, the locked main endpoint measuring time until recurrence or death, was 12.7 versus 11.2 months, P=0.394. That is a failed primary test. Median overall survival was 33.2 versus 25.2 months, P=0.099, and therefore also did not cross the usual 0.05 statistical boundary. Survival after recurrence was 15.3 versus 9.8 months, P=0.031, but that later secondary comparison cannot replace the failed primary result. Grade 3 or worse adverse events occurred in 61.5% versus 63.6%; two treatment-related deaths occurred in the control arm. Only 117 of 336 planned people were randomized: 117 / 336 = 34.8%, leaving 65.2% of the planned sample absent. The result is useful evidence against adopting the package. It is weak evidence about heat itself because the experimental package also added cisplatin and changed the gemcitabine schedule, while the registered comparator included capecitabine but the paper says only eight control participants received it.", "sourceIds": [ "issels-2023-heat-rct", "clinicaltrials-nct01077427-heat", "eudract-2008-004802-14-heat" ], "links": [ { "target": "failure-heat-component-and-registry-drift", "relation": "bounded-by" }, { "target": "unknown-regional-hyperthermia-added-benefit-modern-care", "relation": "leaves-open" }, { "target": "hypothesis-heat-existing-randomization-recovery-gate", "relation": "motivates" }, { "target": "trial-pdac-regional-hyperthermia-action-map", "relation": "mapped-by" }, { "target": "treatment-modality-ranking", "relation": "updates" }, { "target": "treatment", "relation": "part-of" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "not-applicable", "randomised": "moderate", "independentReplication": "none", "sampleSize": "117 eligible randomized of 336 planned; 107 started assigned treatment", "effect": "Disease-free survival 12.7 versus 11.2 months, P=0.394; overall survival 33.2 versus 25.2 months, P=0.099", "limits": [ "Stopped far below planned enrollment", "Open-label", "Multiple treatment components changed together", "Registry and publication disagree on control treatment delivery", "Main endpoint negative", "Post-recurrence survival was a secondary analysis", "Older chemotherapy setting", "No independent randomized reproduction" ] }, "layout": { "x": 616, "y": 540 }, "updated": "2026-09-15" }, { "id": "claim-hifu-pain-signal-without-survival-proof", "title": "Focused ultrasound has a repeated pain signal, but survival benefit is not established", "shortTitle": "HIFU pain, no survival proof", "type": "claim", "status": "supported", "scope": [ "pancreatic-all", "pdac", "locally-advanced", "metastatic", "local-treatment" ], "summary": "Uncontrolled human studies repeatedly report pain relief after thermal focused ultrasound; the survival claim rests on biased or internally inconsistent comparisons and remains open in live randomized trials.", "content": "High-intensity focused ultrasound, HIFU, concentrates sound energy into a small volume and heats it enough to damage tumour and nearby pain-carrying nerves without inserting a probe. This makes two distinct claims possible: symptom relief and tumour control. A prospective uncontrolled German study treated 50 people with advanced pancreatic cancer and reported early pain relief in 84%; median survival was 16.2 months from diagnosis but only 8.3 months from HIFU. Without a concurrent control, that survival clock combines prior survival, patient selection, chemotherapy and HIFU. A Chinese two-centre record of 120 people after gemcitabine failure reported pain remission of 57% with HIFU plus S-1 versus 20% with S-1, a crude 37-point difference and 1 / 0.37 = 2.7 people treated per additional pain response if causal. It also reported survival of 10.3 versus 6.6 months. But its abstract says randomly assigned while its methods call the work retrospective, say records were collected after treatment and provide no sequence, concealment or registry. That contradiction blocks a randomized-effect claim. The only published prospective randomized HIFU-PC15 document is a protocol, not a result. It targets 40 people across stage III and IV disease, first and second lines and several chemotherapy choices; safety and local response are primary, while pain, quality of life and survival are secondary. Thermal HIFU therefore has a credible symptom signal and direct tissue-destruction mechanism. It does not yet have trustworthy evidence that it extends life.", "sourceIds": [ "li-2015-hifu-s1", "marinova-2018-hifu-prospective", "marinova-2024-hifu-pc-protocol", "drks-00012367-hifu-pc", "clinicaltrials-nct05601323-suizenji", "sonire-2026-nedo-award" ], "links": [ { "target": "failure-hifu-randomization-and-trial-drift", "relation": "bounded-by" }, { "target": "unknown-hifu-pain-survival-and-system-transfer", "relation": "leaves-open" }, { "target": "hypothesis-hifu-existing-randomized-gate", "relation": "motivates" }, { "target": "trial-pdac-hifu-action-map", "relation": "mapped-by" }, { "target": "company-sonire-therapeutics", "relation": "involves" }, { "target": "failure-local-control-without-survival-lap07", "relation": "learns-from" }, { "target": "treatment-modality-ranking", "relation": "updates" }, { "target": "treatment", "relation": "part-of" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "moderate", "randomised": "weak", "independentReplication": "moderate — pain signal only", "sampleSize": "Prospective single arm n=50; retrospective comparison n=120; HIFU-PC15 target n=40; Suizenji randomized target n=90", "effect": "Pain response repeatedly reported; added overall survival remains unproved", "limits": [ "No published prospective randomized result", "The 120-person paper contradicts itself about random assignment", "Uncontrolled studies start survival clocks at different times", "Disease stage and chemotherapy vary", "Pain measures and analgesic use are not consistently controlled", "Thermal HIFU devices and dose plans differ", "Pancreatic histology is not always separated" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Focused acoustic energy can heat tissue and interrupt pain-carrying structures directly." }, "animalEvidence": { "rating": "not-applicable", "reason": "The bounded claim concerns symptom and survival effects measured in people." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Cell destruction cannot establish pain benefit or longer life." }, "reproducibility": { "rating": "moderate", "reason": "Pain relief repeats across cohorts, but controlled survival does not." }, "commercialConflicts": { "rating": "moderate", "reason": "Device firms fund or sponsor parts of the evidence and current trials." }, "recency": { "rating": "current", "reason": "The map includes trial states checked through September 2026." } } }, "layout": { "x": 706, "y": 630 }, "updated": "2026-09-15" }, { "id": "claim-high-risk-surveillance", "title": "Expert surveillance can shift PDAC diagnosis earlier in selected high-risk people", "shortTitle": "High-risk surveillance", "type": "claim", "status": "supported", "scope": [ "pdac" ], "summary": "MRI and endoscopic-ultrasound surveillance cohorts show more stage I diagnoses and longer observed survival in people at high inherited or familial risk.", "content": "In the 2024 comparative cohort, 10 of 26 high-risk people with PDAC had stage I disease versus 155 of 1,504 matched SEER controls; five-year overall survival was 50% versus 9%. The result is clinically important but not a randomised screening trial. The screened PDAC group was only 26 people, all were White, selection can remain despite matching, and surveillance occurred in expert centres. CAPS5 separately reported a stage shift in a prospective multicentre surveillance cohort. The evidence supports selective high-risk surveillance, not population screening.", "sourceIds": [ "blackford-2024-high-risk-surveillance", "dbouk-2022-caps5", "chen-2023-fdr-cancer-risk" ], "links": [ { "target": "detection", "relation": "supports" }, { "target": "pdac", "relation": "part-of" }, { "target": "bottlenecks", "relation": "supports" }, { "target": "claim-universal-germline-testing-needs-action", "relation": "depends-on" }, { "target": "hypothesis-pdac-germline-family-route-completion", "relation": "receives-carriers-from" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "moderate", "randomised": "none", "independentReplication": "moderate", "sampleSize": "26 PDAC cases in 2024 comparison; CAPS cohorts larger but few cancers", "effect": "Stage I: 38.5% versus 10.3%; five-year overall survival: 50% versus 9%", "limits": [ "Nonrandomised", "Small number of cancers", "Expert centres", "Limited population diversity", "Possible residual lead-time and selection bias" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "Inherited-risk enrichment raises prevalence and imaging can identify resectable lesions, but not every detected lesion is lethal." }, "animalEvidence": { "rating": "not-applicable", "reason": "The surveillance claim is measured in human cohorts." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "No laboratory model is needed to support the observed stage shift." }, "reproducibility": { "rating": "moderate", "reason": "More than one expert-centre cohort reports stage shift, but cancer counts remain small and nonrandomized." }, "commercialConflicts": { "rating": "low", "reason": "The cited programme evidence is academic surveillance rather than a product-validation study." }, "recency": { "rating": "recent", "reason": "The comparison was reported in 2024 and is supported by the 2022 CAPS5 cohort." } } }, "layout": { "x": 9, "y": 94 }, "updated": "2026-09-14" }, { "id": "claim-hipec-local-control-without-survival-proof", "title": "HIPEC can reduce local recurrence, but has not extended pancreatic survival", "shortTitle": "HIPEC local control, no survival proof", "type": "claim", "status": "supported", "scope": [ "pdac", "hipec", "resectable", "peritoneal-metastases", "local-treatment" ], "summary": "A small randomized resection study changed where cancer returned without changing survival; metastatic reports combine extreme selection, major surgery and several local treatments.", "content": "HIPEC circulates heated liquid chemotherapy through the open abdomen during surgery. Heat may increase tissue penetration and drug effect on microscopic surface disease. In 42 people randomly assigned after margin-clear pancreatic resection, gemcitabine HIPEC reduced local recurrence from 11/21, 52%, to 2/21, 10%, a 42-point difference. Median overall survival was 18.0 months with surgery alone and 17.1 with HIPEC, p=0.899. Median disease-free survival was 10 versus 14 months, p=0.888. One person in each arm died in hospital. The local endpoint moved; useful life did not. The peritoneal-metastasis evidence answers a different question and is uncontrolled. An 18-person Mayo pilot required peritoneal-only disease and response to at least six months of chemotherapy, then combined complete cytoreduction with pancreatic resection, irreversible electroporation or intraoperative radiation plus HIPEC. Median survival was 26 months, while 8/18 had severe complications, 7/18 were readmitted and 1/18 died within 30 days. A later ten-person series reported 24.2 months without a control. These reports prove that a highly selected operation package is possible. They do not prove HIPEC contribution or cure.", "sourceIds": [ "padilla-valverde-2024-adjuvant-hipec-rct", "grotz-2022-crs-hipec-pdac-pilot", "yan-2024-crs-hipec-pdac-retrospective", "clinicaltrials-nct04858009-hipec-pdac" ], "links": [ { "target": "treatment", "relation": "part-of" }, { "target": "failure-hipec-local-endpoint-and-package-selection", "relation": "bounded-by" }, { "target": "unknown-hipec-added-value-peritoneal-pdac", "relation": "leaves-open" }, { "target": "hypothesis-hipec-existing-trial-measurement-gate", "relation": "motivates" }, { "target": "trial-pdac-hipec-action-map", "relation": "mapped-by" }, { "target": "failure-local-control-without-survival-lap07", "relation": "supports" }, { "target": "treatment-modality-ranking", "relation": "updates" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "weak", "randomised": "moderate — one small adjuvant study", "independentReplication": "none for survival benefit", "sampleSize": "Adjuvant randomized n=42; metastatic pilot n=18; retrospective metastatic n=10", "effect": "Adjuvant local recurrence 10% versus 52%, but median survival 17.1 versus 18.0 months; uncontrolled metastatic median survival 24.2 to 26 months", "limits": [ "Small single-centre randomization", "No adjuvant survival difference", "Imbalanced adjuvant chemotherapy", "Extreme metastatic selection", "Mixed local procedures", "No metastatic control", "Severe operative harm", "Short follow-up" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Heated local chemotherapy can kill exposed residual cells on abdominal surfaces." }, "animalEvidence": { "rating": "not-applicable", "reason": "Human randomized local-control data exist." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "The decision is whether local killing repays systemic failure and operative harm." }, "reproducibility": { "rating": "weak", "reason": "Feasibility repeats; survival benefit does not." }, "commercialConflicts": { "rating": "low", "reason": "The decisive reports are investigator-led; surgical enthusiasm remains a non-financial bias risk." }, "recency": { "rating": "current", "reason": "The route includes the 2024 randomized report and a registry updated in September 2026." } } }, "layout": { "x": 1108, "y": 1032 }, "updated": "2026-09-15" }, { "id": "claim-histotripsy-pancreas-visibility-limited", "title": "Pancreatic histotripsy works only when the target and bubble cloud stay visible", "shortTitle": "Histotripsy visibility constraint", "type": "claim", "status": "supported", "scope": [ "pdac", "local-treatment", "focused-ultrasound", "medical-device" ], "summary": "Mechanical ultrasound can break pancreatic tissue without an incision, but pig work shows that bowel gas can hide the target and turn a non-invasive treatment into severe off-target injury.", "content": "Histotripsy sends very short, high-pressure ultrasound pulses into a small focus. The negative-pressure part creates a cloud of bubbles; their rapid expansion and collapse mechanically fragments tissue with little intended heat. This differs from thermal HIFU, which cooks tissue. In nine young pigs with healthy pancreases, the planned 1.5-centimetre treatment zone was placed in the pancreas in six. The bubble cloud was visible during treatment in five. When bowel gas blocked the ultrasound view, three treatments hit bowel or nearby tissue. One pig developed septic inflammation across the abdomen and was killed after two days; another developed a gastrointestinal blockage and was killed around three weeks; the third bowel lesion had no observed clinical consequence. Two pigs had possible pancreatitis, one with enzyme elevation and duct widening. A portal-vein clot and two skin burns also occurred. The mechanism is physically credible and the device can spare some vessels, but its control loop has a hard condition: if target or bubble cloud visibility is lost, energy delivery must not start or must stop immediately. There is no human pancreatic outcome yet.", "sourceIds": [ "gannon-2026-pancreas-histotripsy-safety", "clinicaltrials-nct06282809-gannon", "fda-den220087-edison" ], "links": [ { "target": "failure-histotripsy-acoustic-window-and-translation", "relation": "bounded-by" }, { "target": "unknown-histotripsy-human-coverage-and-benefit", "relation": "leaves-open" }, { "target": "hypothesis-gannon-independent-safety-gate", "relation": "motivates" }, { "target": "trial-pdac-histotripsy-action-map", "relation": "mapped-by" }, { "target": "company-histosonics", "relation": "involves" }, { "target": "claim-hifu-pain-signal-without-survival-proof", "relation": "compares-with" }, { "target": "treatment-modality-ranking", "relation": "updates" }, { "target": "treatment", "relation": "part-of" } ], "evidence": { "humanProspective": "none — pancreatic trial recruiting", "humanRetrospective": "none", "randomised": "none", "independentReplication": "weak", "sampleSize": "Nine healthy pigs; one 50-person human feasibility study recruiting", "effect": "Six of nine pancreas targets achieved; three off-target bowel hits under poor visibility; no human benefit result", "limits": [ "Healthy young pigs, not human tumours", "Small 1.5 cm target", "Pancreas depth differed from humans", "Safety controls were bypassed for shallow targets", "Industry supplied system and funding", "Human trial is single arm", "Only 30-day safety is primary" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Cavitation-driven mechanical tissue fragmentation is directly observed." }, "animalEvidence": { "rating": "moderate", "reason": "Pancreatic targeting and failure modes were measured in nine living pigs." }, "inVitroEvidence": { "rating": "strong", "reason": "Bubble-cloud location and tissue breakup are measurable physical outputs." }, "reproducibility": { "rating": "weak", "reason": "The pancreas evidence comes from a small linked research and company network." }, "commercialConflicts": { "rating": "high", "reason": "HistoSonics supplied and funded the work and employs or advises several authors." }, "recency": { "rating": "current", "reason": "The 2026 animal report and July 2026 human registry update are current." } } }, "layout": { "x": 808, "y": 732 }, "updated": "2026-09-15" }, { "id": "claim-hrs4642-g12d-chemo-signal", "title": "Chinese KRAS-G12D inhibition plus chemotherapy produced a strong response signal with heavy toxicity", "shortTitle": "HRS-4642 early signal", "type": "claim", "status": "open", "scope": [ "pdac" ], "summary": "HRS-4642 plus gemcitabine and nab-paclitaxel produced a 63.3% response rate in 30 untreated patients; 90.3% had grade 3 or worse treatment-related events.", "content": "A Chinese multicentre phase 1b/2 trial tested an intravenous liposomal KRAS-G12D inhibitor, HRS-4642, with gemcitabine and nab-paclitaxel. Thirty treatment-naive phase 2 patients produced a confirmed response rate of 63.3%, 95% CI 43.9–80.1, at 12.3 months median follow-up. Grade 3 or worse treatment-related adverse events occurred in 90.3%, mainly blood toxicities attributed to chemotherapy; none caused treatment discontinuation or death in this small series. The result matters because it is a non-Western, mutation-selective alternative to common-RAS treatment and directly tests liposomal delivery. It cannot show that HRS-4642 adds benefit because every patient received all three drugs. The double-blind phase 3 comparison against placebo plus the same chemotherapy is the decisive experiment. The likely Hengrui compound and liposome patent families make compound supply, formulation know-how, combination rights and trial-data access practical constraints. Those ownership records do not strengthen the efficacy claim, and exact HRS-4642 claim coverage still requires counsel.", "sourceIds": [ "cui-2026-hrs4642-pdac", "nct07232875-hrs4642-phase3", "hengrui-2025-hk-prospectus", "patent-wo2022268051-hengrui-kras-g12d", "patent-wo2024022507-hengrui-liposome" ], "links": [ { "target": "claim-setidegrasib-g12d-degradation", "relation": "competes-with" }, { "target": "claim-daraxonrasib-common-ras-survival-breakthrough", "relation": "competes-with" }, { "target": "company-hengrui", "relation": "developed-by" }, { "target": "patent-landscape", "relation": "constrained-by" }, { "target": "person-liwei-wang", "relation": "led-by" }, { "target": "treatment-modality-ranking", "relation": "updates" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "not-applicable", "randomised": "none", "independentReplication": "none", "sampleSize": "31 treated; 30 treatment-naive phase 2 patients", "effect": "Confirmed ORR 63.3%, 95% CI 43.9–80.1; grade 3+ treatment-related events 90.3%", "limits": [ "Single arm", "Very small", "Company employees among authors", "Chemotherapy contribution cannot be separated", "No mature PFS or OS comparison", "G12D subgroup only", "Patent records affect access but do not prove efficacy", "Exact HRS-4642 patent claim coverage remains unconfirmed" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "KRAS-G12D is a direct tumour driver and selective inhibition is a clear intervention, but the uncontrolled combination cannot assign the response to HRS-4642." }, "animalEvidence": { "rating": "moderate", "reason": "Non-human development evidence supports the route to testing, but the clinical claim stands on a small human cohort rather than independent animal reproduction." }, "inVitroEvidence": { "rating": "strong", "reason": "The linked compound and formulation records report KRAS-G12D and ERK-pathway activity, but laboratory activity cannot establish the observed human contribution." }, "reproducibility": { "rating": "none", "reason": "No independent cohort or randomized result has reproduced the treatment effect." }, "commercialConflicts": { "rating": "high", "reason": "The sponsor controls the compound, formulation, trial and much of the public evidence, with company employees among authors." }, "recency": { "rating": "current", "reason": "The phase 1b/2 result and phase 3 record are current in 2026." } } }, "layout": { "x": 128, "y": 82 }, "updated": "2026-09-14" }, { "id": "claim-human-pdac-dtc-prognostic-not-dormancy", "title": "Human disseminated-cell assays predict risk but do not prove PDAC dormancy", "shortTitle": "DTC risk is not dormancy", "type": "claim", "status": "supported", "scope": [ "pdac" ], "summary": "Bone-marrow tumour-cell signals associate with poor survival, but no study shows that the measured cells are dormant, seed recurrence or respond to a useful action.", "content": "Human studies show dissemination, not a complete dormancy route. At primary surgery, cytokeratin-positive bone-marrow cells were found in 24/175, 13.7%, pancreatic-cancer patients. In multivariable analysis, positivity associated with progression-free survival HR 2.057 and overall survival HR 2.283. In a separate prospective study of 48 unresectable PDAC patients, an mRNA panel found disseminated-cell signals before treatment in 15/48, 31%, and after two months in 8/25, 32%. Pretreatment positivity associated with PFS and OS HR 2.0, while the during-treatment result was not prognostic. These assays do not show whether a positive cell is viable, quiescent, growing, killed by treatment or ancestral to the later recurrence. Bone is also not a common clinical PDAC metastasis site. Marker choice matters: the second study's restricted keratin panel produced HRs as high as 22 from a small cohort, which is a warning about selection and unstable estimates. Human primary-tumour data support route differences: among 744 resected patients, 333 recurred, and selected primary tumours that later recurred first in liver or lung carried different expression programmes. Yet only 21 primary specimens entered the single-nucleus comparison, and the work did not sample residual cells. The valid conclusion is that occult dissemination and pre-existing recurrence states are plausible in people. Dormancy, timing and action remain unproved.", "sourceIds": [ "effenberger-2012-pdac-bone-marrow-dtc", "nordgard-2022-unresectable-dtc", "chalabi-hajkarim-2025-organotropism", "tanaka-2012-microscopic-local-recurrence" ], "links": [ { "target": "residual-disease", "relation": "part-of" }, { "target": "claim-pdac-dormancy-model-proven-human-uncertain", "relation": "constrains" }, { "target": "unknown-dormant-cell-actionability", "relation": "leads-to" }, { "target": "failure-pdac-dormancy-sample-gap", "relation": "exposes" }, { "target": "trial-pdac-residual-cell-state-action-map", "relation": "informs" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "moderate", "randomised": "none", "independentReplication": "moderate for prognosis; none for dormancy", "sampleSize": "Surgical marrow cohort n=175; unresectable prospective cohort n=48; recurrence cohort n=744 with 21 sequenced primaries", "effect": "Bone-marrow positivity associated with OS HR 2.283 and HR 2.0 in separate cohorts", "limits": [ "One-time marrow sampling", "Different assays and thresholds", "No viability test", "No lineage link to recurrence", "No dormant-state definition", "No action trial" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "Occult cells can persist and predict relapse, but the measured marrow signal may not be the lethal clone." }, "animalEvidence": { "rating": "strong", "reason": "Several PDAC models directly track quiescent or residual cells and later growth." }, "inVitroEvidence": { "rating": "moderate", "reason": "Human-derived models can sustain quiescent states, but culture changes the state being measured." }, "reproducibility": { "rating": "moderate", "reason": "Prognostic direction repeats across human assays; dormancy and treatment utility do not." }, "commercialConflicts": { "rating": "low", "reason": "The central marrow studies are academic biomarker reports without a named product route." }, "recency": { "rating": "mixed", "reason": "Human DTC studies are older, while organ-linked cellular-state work is from 2025." } } }, "updated": "2026-09-14" }, { "id": "claim-hybridtherm-randomized-no-patient-benefit", "title": "HybridTherm ablated pancreatic tissue but did not improve survival or resection", "shortTitle": "HybridTherm: no patient benefit", "type": "claim", "status": "supported", "scope": [ "pdac", "locally-advanced", "borderline-resectable", "local-treatment" ], "summary": "The only randomized HybridTherm comparison found a small uncertain six-month disease-control difference, no survival gain and fewer resections with the device.", "content": "HybridTherm was a 14-gauge probe placed through an endoscope under ultrasound. Radiofrequency energy heated the tumour while internal carbon-dioxide cooling controlled the probe; the registry's word cryothermal does not mean that the tumour was frozen. Forty people were assigned to HybridTherm plus chemotherapy or chemotherapy alone. Three assigned to HybridTherm were removed after new scans found metastases, leaving 17 versus 20 in the paper's all-assigned analysis. At six months, 7/17 versus 6/20 were alive without measured progression: 41.2% versus 30%, an 11.2 percentage-point difference, P=0.48. Median time without progression was 11.1 versus 9.4 months, but the interval was wide and the comparison was not significant. Median overall survival was 13 versus 17 months, hazard ratio 1.18, 95% confidence interval 0.6 to 2.4, P=0.79. Resection occurred in 2/17 versus 4/20. The probe produced local tissue change and 11/37 procedures had an adverse event, though none was severe. This is evidence of feasible local heating. It is not evidence of longer life, more surgery or net patient benefit.", "sourceIds": [ "testoni-2021-hybridtherm-rct", "clinicaltrials-nct02336672-hybridtherm" ], "links": [ { "target": "failure-hybridtherm-product-accrual-and-custody", "relation": "bounded-by" }, { "target": "unknown-eus-rfa-added-benefit-after-hybridtherm", "relation": "leaves-open" }, { "target": "hypothesis-eus-rfa-existing-study-gate", "relation": "motivates" }, { "target": "trial-pdac-eus-rfa-action-map", "relation": "mapped-by" }, { "target": "company-erbe-elektromedizin", "relation": "involves" }, { "target": "claim-pelican-rfa-no-survival-and-more-harm", "relation": "supports" }, { "target": "treatment-modality-ranking", "relation": "updates" }, { "target": "treatment", "relation": "part-of" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "moderate", "independentReplication": "weak", "sampleSize": "40 randomized; 17 intervention and 20 control retained in the reported all-assigned set", "effect": "Six-month progression-free difference +11.2 points, p=0.48; median survival 13 versus 17 months, HR 1.18, p=0.79", "limits": [ "Stopped below the planned phase 2 size", "Three intervention assignments excluded after randomization", "Open-label single centre", "Different chemotherapy regimens allowed", "Wide confidence intervals", "Manufacturer controlled allocation and monitoring", "Registry and paper enrollment counts conflict" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Radiofrequency heating produced visible ablation in the intended pancreatic target." }, "animalEvidence": { "rating": "moderate", "reason": "Thermal ablation has model support, but the human randomized result is more decision-relevant." }, "inVitroEvidence": { "rating": "strong", "reason": "Radiofrequency heating and tissue death are direct physical effects." }, "reproducibility": { "rating": "weak", "reason": "No independent randomized replication with the same probe exists." }, "commercialConflicts": { "rating": "high", "reason": "The manufacturer held the allocation list, monitored the data and employed two authors." }, "recency": { "rating": "current", "reason": "The completed record and active successor study were checked in September 2026." } } }, "layout": { "x": 772, "y": 696 }, "updated": "2026-09-15" }, { "id": "claim-hypoxia-control-without-survival-pdac", "title": "A hypoxia-activated drug delayed PDAC progression but did not pass its survival test", "shortTitle": "Hypoxia control, no survival proof", "type": "claim", "status": "supported", "scope": [ "pdac", "pan-cancer", "transfer-unproven" ], "summary": "Evofosfamide improved PFS from 3.7 to 5.5 months in 693 people, while OS HR 0.84 missed the planned phase 3 threshold.", "content": "Mechanism. Hypoxia means low oxygen. Evofosfamide is a masked cell-killing drug intended to open and release its payload in low-oxygen tissue. That is a conditional circuit: tumour oxygen is the input, chemical activation is the switch, and local poison release is the output. MAESTRO randomized 693 people to evofosfamide or placebo with gemcitabine. Median progression-free survival improved from 3.7 to 5.5 months, HR 0.77, P=.004, and confirmed response rose from 9% to 15%, P=.009. Median overall survival was 8.7 versus 7.6 months, HR 0.84, 95% CI 0.71-1.01, P=.059; the main endpoint failed. Blood-count harm was more common. No prospective tumour-hypoxia measurement selected entry or tested whether the circuit was switched on in each patient. Decision. The drug showed real tumour-control activity but not proven longer life. Do not call the route dead and do not rerun an unselected evofosfamide cohort. Any reuse needs a fixed, independently validated human hypoxia assay, adequate payload activation in tumour, preserved normal tissue and a survival control. PFS alone cannot release clinical or mission adoption.", "sourceIds": [ "clinicaltrials-nct01746979-maestro", "van-cutsem-2016-maestro" ], "links": [ { "target": "treatment", "relation": "part-of" }, { "target": "failure-vascular-starvation-pdac", "relation": "contrasts-with" }, { "target": "trial-pdac-vascular-hypoxia-action-map", "relation": "bounded-by" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong", "independentReplication": "weak", "sampleSize": "693 randomized, 346 evofosfamide and 347 placebo", "effect": "PFS HR 0.77; OS HR 0.84, 95% CI 0.71-1.01, main endpoint failed", "limits": [ "Conference report rather than full paper", "Old gemcitabine backbone", "No hypoxia-selected entry", "PFS and OS point in the same direction but only PFS passed", "Drug development stopped" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "PDAC can be hypoxic and the drug uses oxygen-dependent chemical activation." }, "animalEvidence": { "rating": "strong", "reason": "Models supported hypoxia-dependent activation before phase 3." }, "inVitroEvidence": { "rating": "strong", "reason": "Activation and cell killing under controlled oxygen can be measured." }, "reproducibility": { "rating": "weak", "reason": "The phase 3 control effect is real, but useful-life benefit did not pass and no selected replication exists." }, "commercialConflicts": { "rating": "moderate", "reason": "Product sponsors ran the study; the registry and failed primary endpoint are public." }, "recency": { "rating": "historical", "reason": "The result remains the largest direct PDAC hypoxia-drug test." } } }, "layout": { "x": 322, "y": 270 }, "updated": "2026-09-15" }, { "id": "claim-icc-resection-often-leaves-recurrence-risk", "title": "Curative-intent intrahepatic cholangiocarcinoma resection is followed by recurrence in most patients", "shortTitle": "ICC recurrence after surgery", "type": "claim", "status": "supported", "scope": [ "cholangiocarcinoma", "hepatobiliary" ], "summary": "In an international 1,221-patient surgical cohort, 59.4% recurred and median overall survival was 36.8 months.", "content": "Surgery remains the only established route to eliminate localized intrahepatic cholangiocarcinoma, but an operation labelled curative-intent is not equivalent to eradication. Across 1,221 patients from an international multi-institutional database, 725, or 59.4%, developed recurrence and median overall survival was 36.8 months. People who remained recurrence-free for several years had progressively better conditional survival, which is useful for follow-up counselling but does not identify an intervention. The result makes the missing step explicit: postoperative disease measurement must distinguish residual metastatic clones from a genuinely cleared state, and any adjuvant treatment must prove that acting on that state changes recurrence or survival. The cohort cannot establish the value of surgery against no surgery because resectability and fitness determined entry.", "sourceIds": [ "hu-2020-icc-resection-recurrence", "primrose-2019-bilcap" ], "links": [ { "target": "model-cholangiocarcinoma-causal-chain", "relation": "supports" }, { "target": "residual-disease", "relation": "extends" }, { "target": "cholangiocarcinoma", "relation": "part-of" }, { "target": "unknown-ctdna-actionability", "relation": "needs-analogue-of" } ], "evidence": { "humanProspective": "none", "humanRetrospective": "strong", "randomised": "none", "independentReplication": "moderate", "sampleSize": "1,221 curative-intent intrahepatic cholangiocarcinoma resections", "effect": "59.4% recurrence; median overall survival 36.8 months", "limits": [ "Retrospective surgical selection", "Multi-era treatment and surveillance", "Intrahepatic disease only", "Conditional survival is not a treatment effect", "No molecular residual-disease action test" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "High recurrence after complete-intent resection is consistent with occult local or metastatic disease." }, "animalEvidence": { "rating": "weak", "reason": "Models support spread but are not needed to establish the human recurrence burden." }, "inVitroEvidence": { "rating": "weak", "reason": "Laboratory systems do not estimate postoperative recurrence or treatment benefit." }, "reproducibility": { "rating": "moderate", "reason": "High recurrence is common across surgical series, though this exact estimate comes from one database." }, "commercialConflicts": { "rating": "low", "reason": "The cohort does not depend on a proprietary treatment or assay." }, "recency": { "rating": "historical", "reason": "The large cohort remains useful for burden, while current adjuvant and ctDNA evidence is tracked separately." } } }, "layout": { "x": 46, "y": 24 }, "updated": "2026-09-14" }, { "id": "claim-il6-cachexia-signal", "title": "IL-6 blockade reduced muscle loss but missed the PDAC survival endpoint and increased severe toxicity", "shortTitle": "Cachexia signal", "type": "claim", "status": "contested", "scope": [ "pdac", "transfer-unproven" ], "summary": "Tocilizumab did not improve six-month or median survival, PFS or response, but reduced muscle loss; grade 3+ treatment toxicity rose to 88.1%.", "content": "A randomized phase 2 added tocilizumab to gemcitabine/nab-paclitaxel for patients selected by inflammatory score. The six-month OS endpoint was missed: 68.6% versus 62.0%, P=.409; median OS, PFS and response did not differ. An 18-month survival comparison and several muscle measures favoured tocilizumab, including less early skeletal-muscle loss. Grade 3 or worse treatment-related events rose from 63.4% to 88.1%. This may be a supportive-care signal, not tumour control. Cachexia can shorten treatment and life, so it deserves a separate programme, but late survival and multiple exploratory measures are vulnerable to chance after the primary endpoint failed. Next test: a cachexia-specific trial powered for function, treatment delivery, quality of life and survival, with multiplicity control and a lower-toxicity dose or narrower IL-6-selected group.", "sourceIds": [ "chen-2025-tocilizumab-cachexia" ], "links": [ { "target": "treatment", "relation": "part-of" }, { "target": "unknowns", "relation": "part-of" }, { "target": "claim-statistical-significance-can-hide-trivial-benefit", "relation": "warned-by" }, { "target": "experiments", "relation": "requires-confirmation" }, { "target": "failure-tocilizumab-cachexia-muscle-without-survival", "relation": "bounded-by" }, { "target": "unknown-cachexia-treatment-utility", "relation": "leads-to" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong", "independentReplication": "none", "sampleSize": "147 treated including 6-person safety cohort", "effect": "Primary OS6 missed; muscle loss improved; grade 3+ treatment events 88.1% vs 63.4%", "limits": [ "Primary endpoint negative", "Exploratory late survival", "Multiple comparisons", "Increased toxicity", "Supportive-care versus tumour effect unresolved", "No replication" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "IL-6 can contribute to inflammatory wasting, but the selected human intervention did not improve the main survival outcome." }, "animalEvidence": { "rating": "moderate", "reason": "Model evidence supports IL-6 involvement but cannot outweigh human toxicity and the failed endpoint." }, "inVitroEvidence": { "rating": "moderate", "reason": "Pathway activity does not establish useful whole-patient benefit." }, "reproducibility": { "rating": "none", "reason": "No independent randomized repeat of this exact PDAC strategy was identified." }, "commercialConflicts": { "rating": "moderate", "reason": "A commercial medicine was tested, but the publication reports the negative primary result and toxicity." }, "recency": { "rating": "current", "reason": "The randomized result was published in 2025." } } }, "layout": { "x": 135, "y": 79 }, "updated": "2026-09-14" }, { "id": "claim-immune-exclusion-is-multicellular", "title": "PDAC immune exclusion is produced by interacting tumour, fibroblast and myeloid programmes", "shortTitle": "Multicell immune block", "type": "claim", "status": "open", "scope": [ "pdac" ], "summary": "Mouse and paired human trial evidence show that antigen-specific T cells can enter PDAC while suppressive myeloid and regulatory programmes still prevent useful tumour response.", "content": "In native mouse PDAC, FAP-positive fibroblast-derived CXCL12 coated tumour cells and was linked to T-cell exclusion; blocking CXCR4 allowed T-cell entry and exposed anti-PD-L1 activity. Another mouse study found that tumour-derived GM-CSF recruited suppressive Gr-1-positive CD11b-positive myeloid cells and that loss of tumour GM-CSF blocked tumour development through CD8 T cells. Two later human trials show why entry is not enough. Plerixafor plus cemiplimab increased CD8 T-cell infiltration in 21 reported patients but produced no objective responses; granulocytes and macrophages also increased, including an immunosuppressive macrophage state. CRS-207 plus nivolumab and ipilimumab, with or without GVAX, generated mesothelin- and mutant-KRAS-specific T-cell clones inside tumours but produced only two partial responses among 57 treated patients; myeloid and regulatory T-cell signatures tracked poor response. A separate selected signal remains possible: POLAR combined pembrolizumab and olaparib after platinum in core homologous-recombination-deficient disease and reported a 44% three-year survival estimate, but missed both prespecified response and six-month progression thresholds and lacked a randomized comparator. Together, the evidence makes a single-checkpoint or single-trafficking account untenable. The clinical model needs at least four measured layers: antigen-specific function, T-cell location, suppressive myeloid and regulatory state, and tumour-cell susceptibility. A programme advances only when changing those layers produces tumour killing and durable patient benefit, not infiltration alone.", "sourceIds": [ "feig-2013-cxcl12-immune-exclusion", "bayne-2012-gmcsf-myeloid", "ho-2025-cxcr4-pd1-pdac", "bever-2026-vaccine-checkpoint-pdac", "park-2026-polar-pdac" ], "links": [ { "target": "treatment", "relation": "supports" }, { "target": "failure-bulk-stroma-removal", "relation": "succeeds-where" }, { "target": "failure-immune-recruitment-without-response", "relation": "supported-by" }, { "target": "unknowns", "relation": "supports" }, { "target": "pdac", "relation": "part-of" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "weak", "randomised": "moderate", "independentReplication": "moderate", "sampleSize": "Two founding mouse programmes; human CXCR4 plus PD-1 n=21; vaccine plus dual-checkpoint n=57; selected POLAR maintenance n=63", "effect": "Human immune recruitment occurred without useful response in two programmes; selected HRD maintenance showed a possible survival tail but missed its primary thresholds", "limits": [ "Small human trials", "Different immune interventions and patient settings", "No standard-care efficacy comparator in the recruitment studies", "Selected POLAR result is single arm", "Mouse mechanisms may not match every human state", "No demonstrated broad survival benefit" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Paired human samples show that T-cell entry can coexist with macrophage and regulatory-cell states linked to poor response." }, "animalEvidence": { "rating": "strong", "reason": "Independent mouse systems identify stromal exclusion and tumour-driven myeloid suppression, with intervention experiments supporting causal roles." }, "inVitroEvidence": { "rating": "moderate", "reason": "Antigen-specific T-cell expansion and functional assays support target recognition but did not predict clinical response." }, "reproducibility": { "rating": "moderate", "reason": "Different human programmes repeat the pattern of immune engagement without broad tumour control, but the exact blocking state varies." }, "commercialConflicts": { "rating": "moderate", "reason": "Company-supplied agents and extensive investigator relationships are present, while negative results and limits are reported." }, "recency": { "rating": "current", "reason": "The claim integrates human reports published in 2025 and 2026." } } }, "layout": { "x": 62, "y": 72 }, "updated": "2026-09-14" }, { "id": "claim-ipmn-has-two-cancer-routes", "title": "IPMN surveillance must watch both the cyst and the rest of the pancreas", "shortTitle": "Two IPMN cancer routes", "type": "claim", "status": "supported", "scope": [ "pdac" ], "summary": "Cancer can arise by progression within a known IPMN or as a separate PDAC elsewhere, and the two routes do not share the same observed predictors.", "content": "The largest direct prospective evidence separates two routes. In 2,104 non-operated Japanese patients examined every six months for 5.17 years, cumulative high-grade dysplasia or invasive carcinoma within the branch-duct IPMN was 1.90%, while concomitant PDAC distinct from the IPMN was 2.11%. Cyst size, main-duct size and mural nodules predicted the first route; male sex and older age predicted the separate-PDAC route, while cyst size did not. The categories were defined as distinct, so together they imply roughly 4% pancreatic high-grade or malignant disease over the period, but the paper reports route-specific cumulative incidences rather than a formally tested summed endpoint. A separate 6,064-person low-risk-cyst cohort found 38 HGD or cancers: 26 at the cyst site and 12 elsewhere. The practical result is strict: a cyst-only model cannot cover the whole risk. The action path needs one score for change inside the lesion and another for cancer arising elsewhere in the gland, with whole-pancreas imaging and complete interval-cancer follow-up.", "sourceIds": [ "ohtsuka-2024-bd-ipmn-surveillance", "mirzaian-2026-low-risk-cysts", "assawasirisin-2025-long-term-ipmn" ], "links": [ { "target": "unknown-dangerous-precursor-selection", "relation": "constrains" }, { "target": "detection", "relation": "supports" }, { "target": "claim-ipmn-mutations-identify-lineage-not-progression", "relation": "extends" }, { "target": "trial-pancreatic-precursor-action-map", "relation": "constrains" } ], "evidence": { "humanProspective": "strong — 2,104-person fixed six-month surveillance cohort", "humanRetrospective": "moderate — independent 6,064-person health-system cohort", "randomised": "none", "independentReplication": "moderate", "sampleSize": "Japan n=2,104; Massachusetts low-risk cohort n=6,064 with 38 endpoints", "effect": "Japan: 1.90% HGD or invasive carcinoma within IPMN and 2.11% separate PDAC over 5.17 years; Massachusetts: 26/38 at the cyst and 12/38 elsewhere", "limits": [ "No randomized surveillance comparison", "Different lesion definitions and follow-up rules", "Small numbers of cancer endpoints despite large cohorts", "Route-specific cumulative incidences should not be treated as a formally analysed combined endpoint" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "IPMN can progress clonally while the cancer-prone pancreas can also generate a separate ductal cancer." }, "animalEvidence": { "rating": "not-applicable", "reason": "The decision concerns observed human cancer routes during surveillance." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Cell systems cannot measure where cancers arise during long-term clinical follow-up." }, "reproducibility": { "rating": "moderate", "reason": "Two independent human cohorts found substantial numbers of cancers outside the indexed cyst, though event totals remain small." }, "commercialConflicts": { "rating": "low", "reason": "The decisive evidence comes from academic surveillance cohorts rather than a product study." }, "recency": { "rating": "current", "reason": "The main prospective report appeared in 2024 and the independent cohort in 2026." } } }, "layout": { "x": 49, "y": 38 }, "updated": "2026-09-14" }, { "id": "claim-ipmn-mutations-identify-lineage-not-progression", "title": "GNAS and KRAS mutations identify IPMN lineage but do not by themselves predict malignant progression", "shortTitle": "IPMN marker limit", "type": "claim", "status": "supported", "scope": [ "pdac" ], "summary": "GNAS or KRAS was found in 96% of studied IPMNs, but a common lineage marker cannot alone decide which cyst needs surgery.", "content": "Sequencing of IPMN cyst fluid found GNAS codon-201 mutations in 66% of 132 lesions and either GNAS or KRAS mutations in 96%. In seven of eight evaluable IPMN-associated invasive cancers, the invasive lesion retained the GNAS mutation found in the cyst. This supports IPMN as a molecularly traceable PDAC precursor route. It does not make either mutation a progression test: the study did not prospectively distinguish indolent cysts from lesions destined to invade. Because the markers are common in IPMN and most detected pancreatic cysts will not become lethal cancer, the operational need is a longitudinal marker of dangerous change, not just lineage identity.", "sourceIds": [ "wu-2011-gnas-ipmn" ], "links": [ { "target": "detection", "relation": "supports" }, { "target": "unknowns", "relation": "supports" }, { "target": "pdac", "relation": "part-of" } ], "evidence": { "humanProspective": "none", "humanRetrospective": "moderate", "randomised": "not-applicable", "independentReplication": "strong", "sampleSize": "132 IPMNs for prevalence; 8 linked invasive cancers", "effect": "GNAS in 66%; GNAS or KRAS in 96%; shared GNAS in 7/8 linked invasive lesions", "limits": [ "Retrospective specimen study", "Only eight cyst-cancer pairs", "No prospective progression endpoint", "Presence does not distinguish dangerous from indolent IPMN" ] }, "layout": { "x": 15, "y": 39 }, "updated": "2026-09-14" }, { "id": "claim-ipmn-surgery-timing-tradeoff", "title": "The IPMN surgery rule trades avoidable operations against waiting until invasion", "shortTitle": "IPMN timing trade-off", "type": "claim", "status": "supported-with-limits", "scope": [ "pdac" ], "summary": "Earlier surgery exposes many people without invasive disease to major harm; later surgery improves pathology yield but can miss the HGD-only prevention window.", "content": "Neither side of the trade-off is small. In a prospective German registry of 1,137 operations for mixed pancreatic cystic lesions, 137 patients, 12.0%, had histological malignancy; 323, 28.4%, had severe morbidity and 29, 2.6%, died within 30 days. Scaled to 1,000 comparable operations, those observed rates equal about 120 malignancies, 284 severe morbidities and 26 early deaths; these counts can overlap and the registry's malignancy count omits high-grade lesions whose removal may prevent cancer, so it cannot label the remaining operations unnecessary. In a more selected prospective-surveillance cohort without baseline worrisome features, 26 of 267 eventually underwent resection; five had adenocarcinoma and one carcinoma in situ, a 6/26 = 23.1% advanced-pathology yield under that paper's categories. Selection improved in a 2026 Verona series: after 2017, post-surveillance resections contained 28% HGD and 45% invasive cancer while low-grade dysplasia fell to 17%. Yet 45% already invasive is the other failure: waiting for high-risk stigmata can improve yield after the prevention window has narrowed. Worrisome features alone did not clearly separate HGD from invasion. The action rule therefore cannot optimize pathology yield alone. It must minimize two losses together: serious surgery harm in people without HGD or cancer, and invasive cancers occurring before a planned HGD interception. The threshold must be adjusted for age, operative fitness, lesion site and procedure, and compared prospectively with current care.", "sourceIds": [ "henn-2023-cyst-surgery-registry", "lattimore-2023-bd-ipmn-surveillance", "pea-2026-post-surveillance-resection", "wu-2026-cyst-surveillance-practice" ], "links": [ { "target": "unknown-dangerous-precursor-selection", "relation": "supports" }, { "target": "hypothesis-dynamic-cyst-interception-rule", "relation": "constrains" }, { "target": "failures", "relation": "supports" }, { "target": "programme-100m-randomized-platform", "relation": "constrains" } ], "evidence": { "humanProspective": "moderate — prospective surgical registry and surveillance cohort", "humanRetrospective": "strong — 3,304-person trajectory analysis", "randomised": "none", "independentReplication": "moderate", "sampleSize": "German operations n=1,137; low-risk BD-IPMN n=267; Verona presumed IPMN n=3,304", "effect": "German mixed-cyst surgery: 12.0% malignancy, 28.4% severe morbidity, 2.6% 30-day mortality; recent post-surveillance IPMN resections: 28% HGD and 45% invasive cancer", "limits": [ "Mixed cyst types in the German registry", "Malignancy count excludes some preventive value", "Surgery complications and pathology groups overlap", "Verona study is retrospective across changing eras", "No randomized surgery-threshold trial" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "HGD precedes invasion, but radiographic features overlap and pancreatic surgery has intrinsic organ-loss risk." }, "animalEvidence": { "rating": "not-applicable", "reason": "The balance depends on human operative harm and clinical timing." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Laboratory work cannot estimate the net harm of competing clinical actions." }, "reproducibility": { "rating": "moderate", "reason": "Multiple cohorts show both low overall malignant transformation and substantial invasive disease among late-selected resections." }, "commercialConflicts": { "rating": "low", "reason": "The main evidence is academic registry and cohort work without a central commercial assay claim." }, "recency": { "rating": "current", "reason": "The synthesis includes a 2026 post-surveillance pathology study and 2026 practice data." } } }, "layout": { "x": 51, "y": 40 }, "updated": "2026-09-14" }, { "id": "claim-ire-ablation-without-proven-added-survival", "title": "Irreversible electroporation destroys local tissue; added survival remains unproved", "shortTitle": "IRE mechanism boundary", "type": "claim", "status": "contested", "scope": [ "pdac", "local-control", "device-evidence" ], "summary": "High-voltage pulses can destroy pancreatic tumour tissue without relying on heat, but physical ablation, scan control and longer life are three different claims.", "content": "Irreversible electroporation, or IRE, places needle electrodes around a tumour and sends short high-voltage pulses through it. The electric field makes permanent holes in cell membranes and kills cells without heat being the main mechanism. In engineering terms, this establishes that the actuator can alter tissue. It does not establish that the treated volume contains every cancer cell, that distant microscopic disease is controlled, or that patients live longer. The procedure needs general anaesthesia, full muscle paralysis and cardiac timing because the same pulses can cause violent contraction or dangerous heart rhythm. CROSSFIRE is the strongest completed randomized check: after FOLFIRINOX chemotherapy, 68 people were assigned to stereotactic ablative radiotherapy or IRE. Median survival was 16.1 versus 12.5 months, IRE-versus-radiotherapy HR 1.39, 95% CI 0.84-2.30, P=0.21. A hazard ratio above 1 points toward faster death with IRE, but the interval includes both harm and benefit; the trial stopped for futility. One treatment-related death occurred in each arm. Because there was no chemotherapy-only arm, this rejects superiority to the tested radiotherapy package, not the added value of IRE over systemic treatment alone. NICE therefore still limits pancreatic IRE to research. The open question is not whether voltage can kill cells. It is whether adding this procedure after controlled chemotherapy produces more useful life than the same chemotherapy without it.", "sourceIds": [ "nice-2017-ire-pancreatic", "timmer-2024-crossfire", "clinicaltrials-nct03899636-direct-rct", "rai-2022-lappie-protocol", "angiodynamics-2026-10k" ], "links": [ { "target": "treatment", "relation": "qualifies" }, { "target": "treatment-long-tail-audit", "relation": "expands" }, { "target": "failure-local-intensification-pdac-survival-not-shown", "relation": "supports" }, { "target": "failure-direct-selection-and-randomized-reporting-gap", "relation": "bounded-by" }, { "target": "unknown-ire-added-survival-after-modern-chemotherapy", "relation": "leaves-open" }, { "target": "trial-pdac-irreversible-electroporation-action-map", "relation": "mapped-by" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "moderate", "randomised": "strong", "independentReplication": "weak", "sampleSize": "CROSSFIRE randomized n=68; DIRECT randomized target n=528 estimated, actual enrollment unknown", "effect": "Local tissue ablation is plausible; randomized added survival over chemotherapy remains unknown", "limits": [ "CROSSFIRE lacked a chemotherapy-only arm", "CROSSFIRE stopped early", "The phase 3 DIRECT result is missing", "Device delivery does not prove complete tumour coverage", "Local control does not measure distant disease" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Short high-voltage pulses can permanently disrupt cell membranes and create a local ablation zone." }, "animalEvidence": { "rating": "strong", "reason": "Tissue ablation is well established in laboratory and animal systems; the patient-benefit question is separate." }, "inVitroEvidence": { "rating": "strong", "reason": "Electric-field cell killing is repeatable in model systems." }, "reproducibility": { "rating": "weak", "reason": "No completed randomized trial has shown added survival over the same modern chemotherapy." }, "commercialConflicts": { "rating": "high", "reason": "AngioDynamics owns NanoKnife and funded CROSSFIRE and DIRECT evidence." }, "recency": { "rating": "current", "reason": "The boundary combines current trial records, a 2024 randomized report and the 2026 company filing." } } }, "layout": { "x": 556, "y": 480 }, "updated": "2026-09-15" }, { "id": "claim-ivospemin-signal-without-controlled-benefit", "title": "Ivospemin has an uncontrolled response signal and unresolved sight-threatening harm", "shortTitle": "Ivospemin signal boundary", "type": "claim", "status": "contested", "scope": [ "pdac", "transfer-unproven" ], "summary": "Twelve of 28 evaluable phase 1 patients responded, but there was no control arm, six serious retinal toxicities were reported and the pivotal result is missing.", "content": "Polyamines are small charged molecules that cells use to build DNA, RNA and proteins. Cancer cells often increase this supply. Ivospemin, also called SBP-101, resembles the natural polyamine spermine and disturbs this growth system. The human evidence does not yet show added benefit. Its phase 1 trial enrolled 50 people in one group and was designed to choose a dose, not compare survival. At the recommended schedule, 12 of 28 evaluable patients had a partial response: 12 / 28 = 42.9%. Another 11 had stable disease at eight weeks. Median survival later reported by the sponsor was 14.6 months, but comparison with an older chemotherapy trial cannot isolate drug effect because patient mix, scans, later care and calendar time differ. The harm signal is material. The abstract attributed six serious liver and six serious retinal toxicities to ivospemin: each count is 6 / 50 = 12%, although the same people may appear in both counts. A later independent case describes a patient with healthy retinas and 20/20 vision before treatment whose sight fell to 20/5000 and 20/4000 by six months, with bilateral loss of retinal pigment epithelium, the support layer under light-sensing cells, and no available treatment. One case cannot estimate frequency or fully separate drug, chemotherapy and cancer. It does show that excluding people with prior eye disease cannot be assumed to remove the risk. Only the blinded ASPIRE comparison can estimate benefit and arm-specific harm, but its result and actual randomized denominator are not public.", "sourceIds": [ "singhal-2021-sbp101-phase1", "clinicaltrials-nct03412799-sbp101", "carla-2024-ivospemin-retinal-atrophy", "clinicaltrials-nct05254171-aspire" ], "links": [ { "target": "treatment-modality-ranking", "relation": "qualifies" }, { "target": "treatment", "relation": "qualifies" }, { "target": "failure-aspire-trial-operational-collapse", "relation": "limited-by" }, { "target": "unknown-aspire-assigned-denominator-and-results", "relation": "creates" }, { "target": "trial-pdac-polyamine-action-map", "relation": "informs" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "weak", "randomised": "weak", "independentReplication": "weak", "sampleSize": "One-arm phase 1 n=50; efficacy denominator n=28; one independent severe-eye case; ASPIRE result absent", "effect": "Phase 1 response 12/28 = 42.9%; six serious retinal and six serious hepatic toxicities among 50 treated", "limits": [ "No phase 1 control", "Efficacy used an evaluable subset", "Survival benchmark was historical", "The eye paper is one case", "ASPIRE arm-level outcomes are unavailable" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "Polyamine demand and drug-linked pathway effects are plausible, but the exact causal action remains unsettled." }, "animalEvidence": { "rating": "moderate", "reason": "PDAC models support activity but cannot estimate human benefit or retinal harm." }, "inVitroEvidence": { "rating": "strong", "reason": "Ivospemin changes polyamine pools and growth in cell systems." }, "reproducibility": { "rating": "weak", "reason": "No controlled patient benefit or repeated independent response result is public." }, "commercialConflicts": { "rating": "high", "reason": "The efficacy signal comes from sponsor-led development; the severe-eye case is independent." }, "recency": { "rating": "current", "reason": "The harm report is from 2024 and the pivotal registry remains unresolved in 2026." } } }, "layout": { "x": 484, "y": 408 }, "updated": "2026-09-15" }, { "id": "claim-k912-first-human-local-signal-no-component-proof", "title": "K-912 plus focused ultrasound produced local damage, not proof of sonodynamic synergy", "shortTitle": "K-912 local signal, synergy unproved", "type": "claim", "status": "supported", "scope": [ "pdac", "cholangiocarcinoma", "sonodynamic-therapy", "focused-ultrasound" ], "summary": "A 12-person first-human study found short-term scan changes with little early severe harm, but both active components were always given together and patient survival remained short.", "content": "Sonodynamic therapy is meant to use focused sound to activate or release a sensitizing drug inside a tumour. K-912 packages epirubicin, a DNA-damaging chemotherapy, inside polymer particles. The MS-2 device delivers high-intensity focused ultrasound, or HIFU, which can itself heat and kill tissue. In the first-human study, 11 people had stage IV pancreatic cancer and one had cholangiocarcinoma; all 12 had cancer spread across the peritoneum, the lining of the abdomen. Everyone received K-912 one day before HIFU. At one month, scans classified 4/12 as complete and 5/12 as partial tumour coagulation, while actual tumour-size reduction occurred in 5/12 and disease control in 8/12. Two of six people with pain improved. No grade 3 or worse event appeared through 30 days; one mild skin burn and one moderate biliary blood-test change were linked to ultrasound. The individual table gives median survival of 140 days, 4.6 months, among the 11 pancreatic participants. One person labelled complete coagulation already had progressive disease at one month and died after 74 days. Another labelled complete died after 37 days. The study proves that the package can be delivered and can alter a local scan. With no K-912-only, HIFU-only or standard-care group, it cannot prove drug activation, synergy, durable control or added life.", "sourceIds": [ "muragaki-2026-k912-sonodynamic" ], "links": [ { "target": "failure-k912-component-and-local-endpoint", "relation": "bounded-by" }, { "target": "unknown-k912-ultrasound-drug-interaction-and-benefit", "relation": "leaves-open" }, { "target": "hypothesis-k912-component-identification-gate", "relation": "motivates" }, { "target": "trial-pdac-k912-sonodynamic-action-map", "relation": "mapped-by" }, { "target": "company-sonire-therapeutics", "relation": "involves" }, { "target": "treatment-modality-ranking", "relation": "updates" }, { "target": "treatment", "relation": "part-of" } ], "evidence": { "humanProspective": "weak", "humanRetrospective": "none", "randomised": "none", "independentReplication": "none", "sampleSize": "n=12 total; n=11 pancreatic and n=1 cholangiocarcinoma", "effect": "Nine of 12 scan-defined coagulation responses, five size reductions, two of six pain improvements, zero grade 3-plus events through 30 days; pancreatic median survival 140 days", "limits": [ "Single centre", "No component controls", "Four sequential dose groups of three", "Mixed cancers", "One-month efficacy window", "Thirty-day main safety window", "All had peritoneal spread", "Company-supported medical writing" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "Both focused heat and epirubicin can injure cancer, but an extra sound-drug interaction is unproved." }, "animalEvidence": { "rating": "moderate", "reason": "Earlier animal work supports combined treatment but does not identify its human component effect." }, "inVitroEvidence": { "rating": "moderate", "reason": "Drug release and acoustic effects are testable, but were not isolated in the human report." }, "reproducibility": { "rating": "none", "reason": "No independent human cohort has repeated the package." }, "commercialConflicts": { "rating": "moderate", "reason": "Government funded the study, while Sonire supported medical writing and the drug came from Kowa." }, "recency": { "rating": "current", "reason": "The first-human report appeared in 2026." } } }, "layout": { "x": 1042, "y": 966 }, "updated": "2026-09-15" }, { "id": "claim-kinisoquin-access-stack", "title": "A Kinisoquin decision needs the exact formula, licence chain, comparator and event data", "shortTitle": "Kinisoquin access stack", "type": "claim", "status": "supported", "scope": [ "pdac", "pan-cancer", "transfer-unproven" ], "summary": "Two public patent families do not establish the product or trial rights; an independent decision also needs manufacturing, formulation, trial and comparator access.", "content": "The access question is not solved by the word isoquercetin or a claimed 2045 runway. US10391096B2 covers a composition containing isoquercetin, vitamins B3 and C and folic acid for venous thrombosis, including cancer-induced thrombosis, and names Quercis in the ownership history. WO2020112828A1, later including US11872241B2, covers isoquercetin or quercetin routes for thrombotic events and biomarkers in cancer and begins with Beth Israel Deaconess. The phase 3 registry describes Kinisoquin capsules with vitamins B3 and C but does not publish the full formulation, manufacturing controls, licence terms or exact claim mapping. Independent evaluation therefore needs at least seven layers: exact active and excipient specification; trial-grade manufacture and stability; licence chain between inventors, Beth Israel, Quercegen and Quercis; live national claims and freedom to operate; protocol, placebo and dose-selection rules; patient-level clot, bleed, discontinuation, PFS and OS data; and a fair comparison with generic LMWH or direct factor-Xa prophylaxis. The 2021 announcement of up to $150m was a conditional share-subscription facility after a public listing, not proof that $150m was received, and is excluded from the capital case. No licence, acquisition or programme money should move until these layers are documented by contract and counsel.", "sourceIds": [ "patent-us10391096-kinisoquin", "patent-wo2020112828-isoquercetin-cancer", "clinicaltrials-nct06861088-catiq-p3", "quercis-2026-kinisoquin-profile", "zwicker-2019-isoquercetin-catiq" ], "links": [ { "target": "patent-landscape", "relation": "extends" }, { "target": "company-quercis-pharma", "relation": "depends-on" }, { "target": "claim-kinisoquin-phase2-marker-not-clinical-proof", "relation": "constrains" }, { "target": "programme-500m-integrated-mission", "relation": "constrains" }, { "target": "risks", "relation": "supports" } ], "evidence": { "humanProspective": "not-applicable", "humanRetrospective": "not-applicable", "randomised": "not-applicable", "independentReplication": "moderate for public records; none for contract access", "sampleSize": "Two patent families, one phase 3 registry, one company profile and one phase 2 report", "effect": "Seven access layers identified; no efficacy or valuation effect", "limits": [ "Not a legal opinion", "No licence inspected", "No manufacturing dossier", "Patent family search incomplete", "Company ownership is private", "Exact phase 3 contract unavailable" ], "dimensions": { "biologicalPlausibility": { "rating": "not-applicable", "reason": "This record maps ownership, product and data access rather than biological truth." }, "animalEvidence": { "rating": "not-applicable", "reason": "Animal evidence cannot establish rights or supply." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Assay activity cannot establish the product, licence or data stack." }, "reproducibility": { "rating": "moderate", "reason": "Public patents and registries show the layers, while the contracts have not been tested." }, "commercialConflicts": { "rating": "high", "reason": "The sponsor controls the product, trial, key data and much of the public narrative." }, "recency": { "rating": "current", "reason": "The public ownership and trial records were checked in September 2026." } } }, "layout": { "x": 166, "y": 90 }, "updated": "2026-09-14" }, { "id": "claim-kinisoquin-phase2-marker-not-clinical-proof", "title": "Kinisoquin's 57-person phase 2 proves marker activity, not prevention of pancreatic-cancer clots", "shortTitle": "Kinisoquin marker result", "type": "claim", "status": "supported", "scope": [ "pdac", "colorectal", "pan-cancer", "transfer-unproven" ], "summary": "Sequential open-label dose cohorts changed D-dimer and coagulation markers with zero primary events, but no control group permits a clinical-effect estimate.", "content": "CATIQ enrolled 57 people with several advanced cancers into sequential open-label cohorts: 28 received 500 mg and 29 received 1,000 mg isoquercetin daily for 56 days. The primary efficacy endpoint was D-dimer, not clinical thrombosis. At 1,000 mg, median D-dimer fell 21.9%, P=.0002; plasma PDI inhibition increased 73.3%, platelet-dependent thrombin generation fell 57.2% and soluble P-selectin fell 57.9%. No primary VTE or major hemorrhage occurred in either cohort. Those zero events are encouraging but have no untreated counterfactual, short follow-up and no PDAC-only estimate. Quercegen supplied product and funded the study alongside NHLBI. Quercis now describes the result as zero VTE and no major bleeding and sponsors NCT06861088, a randomized 480-person PDAC phase 3. That trial—not the biomarkers or zero-event marketing line—can estimate prevention. It must report all randomized patients, dose selection, event adjudication, discontinuation, major and relevant non-major bleeding, comparison with available anticoagulant care, quality of life and survival. A marker change without fewer clinical events fails.", "sourceIds": [ "zwicker-2019-isoquercetin-catiq", "clinicaltrials-nct06861088-catiq-p3", "quercis-2026-kinisoquin-profile" ], "links": [ { "target": "claim-anticoagulation-prevents-pdac-vte-not-survival", "relation": "competes-with" }, { "target": "trial-pdac-vte-action-map", "relation": "tested-by" }, { "target": "company-quercis-pharma", "relation": "developed-by" }, { "target": "claim-kinisoquin-access-stack", "relation": "constrained-by" }, { "target": "unknown-pdac-vte-prophylaxis-rule", "relation": "leaves-open" } ], "evidence": { "humanProspective": "moderate for markers; weak for clinical events", "humanRetrospective": "not-applicable", "randomised": "none for completed evidence", "independentReplication": "none", "sampleSize": "Open-label sequential cohorts n=28 and n=29 across several cancers; phase 3 estimated n=480", "effect": "At 1,000 mg D-dimer -21.9%, PDI inhibition +73.3%, thrombin generation -57.2% and P-selectin -57.9%; no controlled clinical effect", "limits": [ "No control group", "Mixed cancers", "Biomarker primary endpoint", "Only 56 days", "Zero events do not estimate relative effect", "Company-funded and company-supplied product" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "Several coagulation measures changed in the expected direction, but their link to fewer clinical clots is unproved." }, "animalEvidence": { "rating": "moderate", "reason": "PDI and P-selectin routes have model support, which does not establish patient benefit." }, "inVitroEvidence": { "rating": "strong", "reason": "Target and coagulation assays support pharmacological activity at the tested dose." }, "reproducibility": { "rating": "none", "reason": "The completed clinical marker result has no independent controlled reproduction." }, "commercialConflicts": { "rating": "high", "reason": "Commercial product supply and funding feed directly into the current sponsor's phase 3 claim." }, "recency": { "rating": "current", "reason": "The completed result is from 2019 and the decisive phase 3 registry is current in 2026." } } }, "layout": { "x": 162, "y": 86 }, "updated": "2026-09-14" }, { "id": "claim-kras-egfr-stat3-mouse-regression", "title": "RAS, EGFR/HER2 and STAT3 blockade can prevent resistance in mouse PDAC, but the drug combination is not ready for humans", "shortTitle": "Triple-node RAS escape", "type": "claim", "status": "supported", "scope": [ "pdac", "pan-cancer", "transfer-unproven" ], "summary": "Daraxonrasib, afatinib and a STAT3 degrader produced durable regressions across several PDAC mouse systems; dose, drug quality, conflicts and lack of human evidence are decisive limits.", "content": "The biological signal is unusually strong for a mouse combination. Removing RAF1, EGFR and STAT3 genetically caused complete regression in 21 orthotopic tumours with no detected return by 300 days. Pharmacological RAS, EGFR/HER2 and STAT3 blockade used daraxonrasib, afatinib and SD36: in one orthotopic survival experiment, 10 triple-treated mice remained tumour-free for more than 200 days, compared with eight daraxonrasib-treated and nine vehicle mice that reached endpoints. In 12 genetically engineered mice, every tumour regressed and six completely regressed; none relapsed during 80–200 days of treatment, although several animals were euthanised for tumour-unrelated problems. Six patient-derived xenograft models also regressed, using three mice per model, without an untreated comparator shown in that experiment. The result supports a testable resistance model: simultaneous upstream, downstream and parallel escape suppression may be required around RAS inhibition. It does not supply a clinical regimen. The paper says its mouse daraxonrasib dose was about five times the dose used in trials, flags the afatinib dose as a human tolerability problem, and says SD36 lacks suitable drug properties for clinical testing. Small mouse groups, incomplete adaptive immunity in xenografts and no independent repeat further weaken translation. Publication history matters: the original paper was retracted for undisclosed financial interests in Vega Oncotargets, then the work was republished after direct review with expanded declarations. The mission should fund independent reproduction using clinically achievable exposure, human organoid–immune–fibroblast systems and direct normal-tissue toxicity before considering a phase 1 combination. Passing gate: tumour regression across at least five genetically distinct human models at unbound exposures attainable in patients, with two independent laboratories and no severe normal-organ injury. Failing that gate retires the exact triple combination while leaving the multi-node resistance idea open.", "sourceIds": [ "liaki-2026-triple-combination-republished", "pnas-2026-liaki-retraction", "cnio-experimental-oncology-2026" ], "links": [ { "target": "research-frontier-ranking", "relation": "extends" }, { "target": "unknown-durable-ras-control", "relation": "tests" }, { "target": "claim-daraxonrasib-common-ras-survival-breakthrough", "relation": "extends" }, { "target": "failure-liaki-undisclosed-conflict-retraction", "relation": "depends-on" }, { "target": "treatment-modality-ranking", "relation": "supports" }, { "target": "lab-cnio-experimental-oncology", "relation": "depends-on" } ], "evidence": { "humanProspective": "none", "humanRetrospective": "none", "randomised": "none", "independentReplication": "none", "sampleSize": "Orthotopic survival cohorts n=9 vehicle, n=8 daraxonrasib and n=10 triple therapy; genetically engineered triple-therapy n=12; six xenograft models with n=3 each", "effect": "Complete durable regression beyond 200 days in the small orthotopic triple-therapy cohort; regression in all 12 genetically engineered tumours, six complete", "limits": [ "Mouse and xenograft evidence only", "Small groups", "Xenograft experiment lacks a displayed concurrent control", "Daraxonrasib dose about five times the trial dose", "Afatinib exposure may not be tolerable in patients", "SD36 is not a clinical drug", "No independent repeat", "Declared founder and company interests after an earlier retraction" ] }, "layout": { "x": 72, "y": 76 }, "updated": "2026-09-14" }, { "id": "claim-kras-g12c-inhibition-proves-targetability-not-general-control", "title": "Direct KRAS G12C inhibition proves targetability in PDAC but reaches only a small subgroup and responses are usually short", "shortTitle": "KRAS proof and limit", "type": "claim", "status": "supported", "scope": [ "pdac", "hepatobiliary" ], "summary": "Sotorasib produced responses in 8 of 38 PDAC patients and adagrasib in 7 of 21; KRAS G12C occurs in only about 1% to 2% of pancreatic cancers.", "content": "Two single-arm studies show that direct mutant-KRAS inhibition can shrink pancreatic tumours. Sotorasib produced eight confirmed responses among 38 previously treated patients, 21%, with median progression-free survival 4.0 months and overall survival 6.9 months. Adagrasib produced responses in seven of 21 pancreatic and five of 12 biliary-tract patients in a mixed solid-tumour cohort; median response duration across tumour types was 5.3 months. This overturns the old claim that KRAS is wholly undruggable. It does not solve PDAC because G12C is only about 1% to 2% of pancreatic cancers, neither study was randomised, and resistance emerged within months for most. The high-value next test is whether inhibitors for common G12D, G12V or multi-RAS states can produce deeper responses and whether rational combinations prevent resistance without intolerable toxicity.", "sourceIds": [ "strickler-2022-sotorasib-pdac", "bekaii-saab-2023-adagrasib-solid-tumors" ], "links": [ { "target": "treatment", "relation": "supports" }, { "target": "unknowns", "relation": "supports" }, { "target": "claim-biliary-cancers-not-one-disease", "relation": "may-transfer-to" }, { "target": "pdac", "relation": "part-of" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "not-applicable", "randomised": "none", "independentReplication": "moderate", "sampleSize": "38 PDAC patients with sotorasib; 21 pancreatic and 12 biliary patients with adagrasib", "effect": "Objective response 21% with sotorasib and 33.3% in the adagrasib pancreatic subgroup; response duration measured in months", "limits": [ "Rare G12C subgroup", "Single-arm studies", "Previously treated selected patients", "Small cancer-specific cohorts", "Industry-funded", "Response is not durable control" ] }, "layout": { "x": 88, "y": 80 }, "updated": "2026-09-14" }, { "id": "claim-kras-g12v-tcr-early-signal", "title": "KRAS-G12V TCR cells produced one short complete response among five pancreatic patients", "shortTitle": "KRAS-G12V TCR signal", "type": "claim", "status": "mixed", "scope": [ "pdac", "transfer-unproven" ], "summary": "One of five HLA-A*11:01-positive patients had a complete response lasting 5.5 months; repeat infusions produced no responses and sometimes triggered rapid immune rejection.", "content": "Engineering model. A TCR-engineered cell reads a short protein fragment displayed by an HLA molecule: the KRAS mutation is the message and HLA is the display socket. This can target an intracellular cancer driver that CAR-T cannot see, but both mutation and socket must match, tumour cells must keep displaying the fragment and the manufactured cells must persist. Human result. Five men with recurrent KRAS-G12V pancreatic cancer and HLA-A*11:01 received autologous engineered T cells after lymphocyte-depleting treatment. One had complete disappearance of four liver metastases for 5.5 months, two had short stable disease without shrinkage, median PFS was 14.3 weeks and OS 12.6 months. The observed response rate is 1/5=20%, with an exact 95% interval that is necessarily very wide. Repeat-dose failure. Three patients received another infusion after progression. None responded; two showed very fast rejection associated with antibodies against the engineered receptor. Severe events were mainly low blood counts from conditioning. Boundary. Together with the earlier split KRAS-G12D report—one 72% partial response and one nonresponse—this proves that mutant KRAS can be seen by engineered cells in some human PDACs. It does not show a reproducible rate, durable control, broad population reach or a safe repeat-dose system.", "sourceIds": [ "xu-2026-kras-g12v-tcr", "clinicaltrials-nct04146298-kras-g12v-tcr", "leidner-2022-kras-tcr" ], "links": [ { "target": "treatment", "relation": "part-of" }, { "target": "claim-kras-tcr-one-response-not-platform-proof", "relation": "extends" }, { "target": "failure-cell-therapy-route-and-denominator-pdac", "relation": "bounded-by" }, { "target": "unknown-engineered-cell-reproducible-benefit-pdac", "relation": "leaves-open" }, { "target": "trial-pdac-engineered-cell-action-map", "relation": "mapped-by" }, { "target": "hypothesis-kras-tcr-human-persistence-gate", "relation": "supports" } ], "evidence": { "humanProspective": "weak", "humanRetrospective": "not-applicable", "randomised": "none", "independentReplication": "weak", "sampleSize": "KRAS G12V n=5; earlier KRAS G12D case series n=2", "effect": "One complete response lasting 5.5 months among five G12V patients; no response to three repeat infusions", "limits": [ "Non-prespecified interim analysis", "Single group", "Five men", "One mutation-HLA pair", "Custom autologous manufacture", "Conditioning toxicity", "No durable response beyond 5.5 months" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The receptor recognizes a driver-mutation fragment and has produced human tumour regression in two KRAS-HLA settings." }, "animalEvidence": { "rating": "strong", "reason": "Matched organoid and xenograft work supported natural presentation before treatment." }, "inVitroEvidence": { "rating": "strong", "reason": "Mutation specificity, HLA restriction and cell killing were directly tested." }, "reproducibility": { "rating": "weak", "reason": "Two mutation-HLA routes each contain a responder, but most reported patients did not respond and no controlled cohort exists." }, "commercialConflicts": { "rating": "moderate", "reason": "The academic team developed the receptor and controls the ongoing study." }, "recency": { "rating": "current", "reason": "Uses the July 2026 report and current registry." } } }, "layout": { "x": 232, "y": 150 }, "updated": "2026-09-15" }, { "id": "claim-kras-interception-vaccine-immune-response-not-prevention", "title": "A KRAS vaccine induced immune responses in high-risk people; it has not yet shown cancer prevention", "shortTitle": "KRAS prevention gap", "type": "claim", "status": "open", "scope": [ "pdac" ], "summary": "Eighteen of 20 high-risk participants developed KRAS-specific T-cell responses, but 16.5 months of single-arm follow-up cannot establish prevention.", "content": "The first reported cancer-interception trial of a pooled mutant-KRAS long-peptide vaccine enrolled 20 people at inherited or familial risk who also had a pancreatic radiographic abnormality. Adverse events were grade 1 or 2, and 18 of 20 developed a KRAS-specific T-cell response; some responses persisted for two years. No participant had developed PDAC at the reported median follow-up of 16.5 months. Without a control group, an expected short-term cancer rate for the exact lesion mix, or enough follow-up, zero cancers is not evidence that the vaccine prevented cancer. The result establishes early safety and immune activity. Prevention requires a larger controlled study with a defined cancer-incidence endpoint and long follow-up.", "sourceIds": [ "haldar-2026-kras-interception" ], "links": [ { "target": "detection", "relation": "depends-on" }, { "target": "experiments", "relation": "tests" }, { "target": "unknowns", "relation": "supports" }, { "target": "pdac", "relation": "part-of" } ], "evidence": { "humanProspective": "weak", "humanRetrospective": "not-applicable", "randomised": "none", "independentReplication": "none", "sampleSize": "20 high-risk participants with pancreatic radiographic abnormalities", "effect": "KRAS-specific T-cell response in 18 of 20; no PDAC at median 16.5 months", "limits": [ "Single arm", "Very small sample", "Short follow-up for a prevention endpoint", "Radiographic lesions were heterogeneous", "Immune response is not cancer prevention" ] }, "layout": { "x": 34, "y": 115 }, "updated": "2026-09-14" }, { "id": "claim-kras-mutant-dosage-may-shape-resistance", "title": "KRAS allele dosage marks a more aggressive PDAC state and may matter for resistance, but is not yet a predictive biomarker", "shortTitle": "KRAS dosage", "type": "claim", "status": "supported", "scope": [ "pdac" ], "summary": "In 2,336 patients, mutant-KRAS dosage gains occurred in 20% of KRAS-mutated diploid tumours and were associated with advanced disease and worse outcomes.", "content": "A clinicogenomic study spanning 2,336 patients found widespread allelic imbalance at KRAS. Mutant-allele dosage gain occurred in one in five KRAS-mutated diploid tumours and tracked with advanced disease and prognosis across stages. This gives a concrete resistance question for broad RAS inhibitors: does higher mutant dosage reduce depth or duration of response, and does dosage rise at progression? The published association cannot answer that because it predates routine broad-RAS treatment and is observational. The useful experiment is paired tumour and plasma sequencing before daraxonrasib, during response and at progression, with allele-specific copy number and drug exposure measured together.", "sourceIds": [ "varghese-2025-kras-dosage", "oreilly-2026-daraxonrasib" ], "links": [ { "target": "claim-daraxonrasib-common-ras-survival-breakthrough", "relation": "depends-on" }, { "target": "model-pdac-causal-chain", "relation": "part-of" }, { "target": "experiments", "relation": "tests" }, { "target": "unknowns", "relation": "supports" } ], "evidence": { "humanProspective": "none", "humanRetrospective": "strong", "randomised": "none", "independentReplication": "none", "sampleSize": "2,336 patients across stages", "effect": "KRAS mutant dosage gains in 20% of KRAS-mutated diploid tumours; associated with advanced disease and prognosis", "limits": [ "Observational", "Treatment-predictive value not tested", "Broad-RAS resistance inference is a hypothesis", "Single institutional sequencing programme", "Survival association may retain confounding" ] }, "layout": { "x": 94, "y": 67 }, "updated": "2026-09-14" }, { "id": "claim-kras-tcr-one-response-not-platform-proof", "title": "KRAS-G12D TCR cells can regress PDAC in one patient, but a second patient with the same receptors progressed", "shortTitle": "KRAS TCR split result", "type": "claim", "status": "contested", "scope": [ "pdac", "transfer-unproven" ], "summary": "One HLA-C*08:02-positive patient had a 72% partial response lasting at least six months; another treated with the same TCRs progressed and died within six months.", "content": "A patient with metastatic KRAS-G12D PDAC received 16.2 billion autologous T cells engineered with two HLA-C*08:02-restricted receptors. Visceral metastases regressed by 72% and the response was ongoing at six months; engineered cells still formed more than 2% of circulating T cells. That proves a PDAC can be attacked through a shared driver neoantigen. It does not prove a reproducible therapy. The same report describes another PDAC patient treated with the same receptors: lung lesions initially regressed, liver lesions remained stable, then disease progressed and the patient died six months after treatment despite blood persistence. No common antigen-loss or presentation-loss mechanism was found in the progressing lesion. Reach is also narrow: the tumour needs KRAS G12D, the patient needs the matching HLA allele, and a custom autologous product plus lymphodepletion and high-dose interleukin-2 is complex. The decisive programme would build a library across common KRAS alleles and HLA types, measure tumour entry and state, and prospectively explain responder versus nonresponder before scaling manufacturing.", "sourceIds": [ "leidner-2022-kras-tcr" ], "links": [ { "target": "treatment", "relation": "part-of" }, { "target": "claim-immune-exclusion-is-multicellular", "relation": "depends-on" }, { "target": "claim-daraxonrasib-common-ras-survival-breakthrough", "relation": "may-transfer-to" }, { "target": "experiments", "relation": "tests" }, { "target": "unknowns", "relation": "supports" } ], "evidence": { "humanProspective": "weak", "humanRetrospective": "not-applicable", "randomised": "none", "independentReplication": "none", "sampleSize": "Two individually treated PDAC patients discussed; one major response and one progression", "effect": "Responder had 72% RECIST regression ongoing at six months; second patient progressed and died at six months", "limits": [ "Case reports", "HLA-C*08:02 restriction", "KRAS G12D restriction", "Custom autologous manufacture", "High-dose IL-2 and lymphodepletion", "No reproducible response rate", "Durability beyond report unknown" ] }, "layout": { "x": 85, "y": 97 }, "updated": "2026-09-14" }, { "id": "claim-kras-wildtype-pdac-needs-driver-route", "title": "KRAS-wild-type pancreatic cancer is a routing signal, not one disease or one treatment", "shortTitle": "KRAS-wild-type driver route", "type": "claim", "status": "supported", "scope": [ "pdac", "precision-treatment", "gene-fusion", "care-delivery" ], "summary": "About one in ten PDAC samples lacks a KRAS mutation, enriching several rare drivers that need different DNA, RNA and treatment paths.", "content": "System model. KRAS-wild-type means a tumour test did not find the usual activating KRAS mutation. It is a branch condition, not a biological subtype with one drug. In 2,483 profiled PDAC samples, 266, or 10.7%, were KRAS wild type. That group contained BRAF, FGFR2, ALK, RET and NRG1 fusions, NTRK and ERBB2 changes, DNA-repair changes and more MSI-high tumours. A separate 795-case cohort found 73 KRAS-wild-type tumours and alternative MAPK growth-circuit drivers in 32/73, 43.8%. Fusion checks matter: a 100-person cohort found targetable fusions in 4/13 KRAS-wild-type and 0/87 KRAS-mutant tumours, with external rates of 18/97 and 2/10. Human treatment evidence is real but sparse. One NTRK-fusion pancreatic tumour responded then became resistant to larotrectinib. One RDX-MET fusion response lasted at least 12 months. A 14-person KRAS-wild-type series treated five people by matched alteration and reported three lasting 11 to more than 18 months. NRG1 has the only larger pancreatic drug cohort in this route, with 12/30 responses in the FDA review. Boundary. These cases prove that finding the alternative driver can matter. They do not prove that profiling itself extends survival, that every detected change is causal, or that tumour-wide pooled response transfers to PDAC. Each result must lead to its own drug, evidence and access decision.", "sourceIds": [ "philip-2022-kras-wildtype-pdac", "fusco-2021-kras-wildtype-fusions", "singh-2023-kras-wildtype-drivers", "mehdi-2024-kras-wildtype-case-series", "oreilly-2019-ntrk-pdac-response", "fda-2018-larotrectinib-approval", "fda-2024-bizengri-multidisciplinary-review" ], "links": [ { "target": "treatment", "relation": "part-of" }, { "target": "claim-nrg1-approved-response-not-durable-control", "relation": "contains" }, { "target": "failure-kras-wildtype-matching-route-and-selection", "relation": "bounded-by" }, { "target": "trial-pdac-kras-wildtype-action-map", "relation": "mapped-by" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "strong", "randomised": "weak — rare-driver PDAC treatment is not tested as one randomized route", "independentReplication": "moderate — several cohorts reproduce enrichment, but treatment evidence is mostly cases", "sampleSize": "Molecular cohorts n=2,483, n=795 and n=100; treatment evidence includes cases, a 14-person series and NRG1 n=30", "effect": "KRAS-wild-type frequency 10.7%; 43.8% alternative MAPK driver in one cohort; drug sensitivity established for selected rare fusions", "limits": [ "Retrospective profiling", "Referral selection", "Different assays", "Tiny treatment groups", "Tumour-wide approvals", "No route-level survival comparison" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Mutually alternative growth drivers can replace KRAS signalling in a measured tumour." }, "animalEvidence": { "rating": "moderate", "reason": "Driver-specific models support some routes but cannot establish patient benefit." }, "inVitroEvidence": { "rating": "strong", "reason": "Organoid and molecular studies show direct pathway dependence for selected changes." }, "reproducibility": { "rating": "moderate", "reason": "Driver enrichment repeats across cohorts, while each treatment result remains sparse." }, "commercialConflicts": { "rating": "moderate", "reason": "Profiling programmes and tumour-wide drugs have commercial owners and mixed sponsor involvement." }, "recency": { "rating": "current", "reason": "The route includes current English panel coverage and the 2024 case series." } } }, "layout": { "x": 2100, "y": 1480 }, "updated": "2026-09-15" }, { "id": "claim-lanreotide-pnet-control-not-survival", "title": "Lanreotide can hold selected low-grade enteropancreatic NETs stable without proving longer survival", "shortTitle": "Lanreotide control", "type": "claim", "status": "supported", "scope": [ "pnet" ], "summary": "CLARINET halved progression or death risk in pooled somatostatin-receptor-positive NETs, but most patients were already stable and no overall-survival gain was shown.", "content": "CLARINET randomized 204 people with grade 1 or 2, non-functioning, somatostatin-receptor-positive enteropancreatic neuroendocrine tumours to lanreotide or placebo. Median progression-free survival was not reached versus 18.0 months, HR 0.47, and 24-month progression-free survival was 65.1% versus 33.0%. This establishes prolonged disease control for the selected pooled population. It does not establish cure or a pancreatic-specific survival effect. Ninety-six percent had shown no progression during the three to six months before enrolment, pancreatic and intestinal sites were pooled, subgroup intervals were wide, and overall survival and quality of life did not differ in the primary report. Somatostatin-receptor expression is therefore both a treatment handle and a boundary: it supports lower-toxicity chronic control in indolent pNET, but receptor loss, higher grade or fast progression require another path.", "sourceIds": [ "caplin-2014-clarinet-lanreotide" ], "links": [ { "target": "pnet", "relation": "part-of" }, { "target": "claim-pnet-targeted-control-differs-from-pdac", "relation": "extends" }, { "target": "claim-prrt-pnet-randomized-control", "relation": "precedes" }, { "target": "model-pnet-causal-chain", "relation": "supports" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong", "independentReplication": "moderate", "sampleSize": "204 randomized across pancreatic, midgut, hindgut and unknown-primary NETs", "effect": "PFS HR 0.47; 24-month PFS 65.1% versus 33.0%; no OS or quality-of-life difference", "limits": [ "Primary sites pooled", "Mostly stable disease before enrolment", "Low-grade receptor-positive non-functioning disease only", "No overall-survival benefit", "Industry-funded" ] }, "layout": { "x": 111, "y": 30 }, "updated": "2026-09-14" }, { "id": "claim-lapc-response-warrants-surgical-reassessment", "title": "Initially inoperable local PDAC should be reconsidered after treatment, not presumed operable", "shortTitle": "Reconsider, do not presume", "type": "claim", "status": "supported", "scope": [ "pdac" ], "summary": "A second expert review after systemic treatment can find a technically removable state that the original scan cannot show reliably, but it does not prove the operation will extend life.", "content": "Locally advanced PDAC means that scans show no distant spread but the tumour has extensive contact with major blood vessels, making immediate complete removal unsafe or impossible. The useful rule is dynamic rather than permanent. After at least four months of multiagent systemic treatment, a person without new distant disease, with adequate physical reserve and a favourable disease trajectory should receive a fresh review by an expert pancreas team. Persistent vessel contact on CT is not an automatic veto: in NEOLAP, formal anatomical downstaging occurred in only 7.7%, although tumour shrinkage was common and complete-margin resection still occurred. CA19-9, a protein shed into blood by many but not all pancreatic cancers, adds information when interpreted after relief of jaundice and as a continuous change, not as a magic cutoff. PREOPANC-4 shows what the operational ceiling looks like: 155 of 180 selected explorations produced resection after non-progressive disease, but 44% had a major complication, 0.6% died in hospital or within 30 days and 3% died within 90 days. Three specialist centres performed 74% of the resections, and only 37% met the narrower NCCN locally advanced definition. This supports mandatory reconsideration and specialist concentration. It does not support automatic surgery, because the programme had no continued-treatment comparison and did not isolate a survival effect.", "sourceIds": [ "kunzmann-2021-neolap", "guggenberger-2023-neolap-ct", "hartlapp-2022-neolap-ca199", "stoop-2026-preopanc4-surgery", "verkolf-2026-taps-resection-prediction", "nice-2018-ng85-pancreatic" ], "links": [ { "target": "treatment", "relation": "clarifies" }, { "target": "disease-system", "relation": "depends-on" }, { "target": "trial-lapc-conversion-surgery-action-map", "relation": "supported-by" }, { "target": "unknown-lapc-surgery-causal-benefit", "relation": "bounded-by" }, { "target": "failure-conversion-surgery-selection-bias", "relation": "bounded-by" }, { "target": "claim-staging-laparoscopy-avoids-incision-not-proven-survival", "relation": "requires" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "strong", "randomised": "moderate for treatment route; absent for the modern surgery decision", "independentReplication": "moderate", "sampleSize": "PREOPANC-4 explored n=180; NEOLAP randomized n=130; CT subset n=83; TAPS baseline model n=1,835", "effect": "PREOPANC-4 resection after exploration 86%; major morbidity 44%; 90-day mortality 3%", "limits": [ "Reassessment criteria differ between centres", "CA19-9 is absent or misleading in some people", "CT analysis was small", "PREOPANC-4 used broader Dutch anatomy criteria", "No modern randomized surgery-versus-continued-treatment effect", "Expert-centre results may not transport" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Time on systemic treatment reveals rapid spread and treatment resistance that a baseline anatomy label cannot." }, "animalEvidence": { "rating": "not-applicable", "reason": "Human anatomy, treatment response, operative harm and survival decide the route." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Cell response cannot determine technical removal or whole-route benefit." }, "reproducibility": { "rating": "moderate", "reason": "Several expert cohorts find resectable responders, but the rule and denominator vary." }, "commercialConflicts": { "rating": "moderate", "reason": "The operation is not a proprietary product, but centre and operator incentives can raise procedure use." }, "recency": { "rating": "current", "reason": "Includes nationwide and TAPS results published in 2026." } } }, "layout": { "x": 100, "y": 156 }, "updated": "2026-09-15" }, { "id": "claim-latam-cca-referral-cohort-shows-route-loss", "title": "The Latin American CCA registry shows treatment loss after referral but cannot measure who never arrived", "shortTitle": "LATAM CCA route loss", "type": "claim", "status": "supported", "scope": [ "cholangiocarcinoma", "hepatobiliary" ], "summary": "Five expert centres recorded 309 eligible patients; among 276 with treatment data, 29.0% had surgery, 39.9% chemotherapy and 31.2% supportive care alone, but there was no population denominator.", "content": "The LATAM-CCA Registry joined five referral centres in Argentina, Brazil, Chile, Ecuador and Peru. All five reported access to standard surgery, chemotherapy and radiotherapy. Among 309 eligible patients, 47% had local, 19% locally advanced and 34% metastatic disease. Among 276 with treatment data, 80/276 = 29.0% underwent resection, 110/276 = 39.9% received chemotherapy and 31.2% received supportive care alone. Median survival measured after treatment start was 33.7 months after resection, 8.3 months after chemotherapy and 2.3 months with supportive care. These are selected groups, not randomized effects. Patients classified as African or Hispanic had shorter survival from diagnosis than those classified as Caucasian, while survival after the same treatment did not differ clearly across the small groups. Ethnicity was self-identified or taken from notes and may stand in for geography, income, referral delay, stage and many unmeasured factors; it is not evidence for a biological cause. The main blind spot is before referral: the registry excluded cases lacking mandatory data, did not count people never diagnosed or referred, and was not population based. It can reveal loss after arrival at an expert centre. A linked population denominator and dated referral route are needed to find upstream loss.", "sourceIds": [ "da-fonseca-2024-latam-cca" ], "links": [ { "target": "lab-latam-cca-registry", "relation": "observed-at" }, { "target": "datasets", "relation": "constrains" }, { "target": "unknown-regional-route-loss", "relation": "leads-to" }, { "target": "trial-regional-route-completion-action-map", "relation": "informs" } ], "evidence": { "humanProspective": "none", "humanRetrospective": "moderate — five-centre observational registry", "randomised": "none", "independentReplication": "none for this exact Latin American CCA route", "sampleSize": "309 eligible patients from five expert referral centres; 276 in treatment analysis", "effect": "29.0% resection, 39.9% chemotherapy and 31.2% supportive care; no population denominator or before-referral count", "limits": [ "Referral-centre selection", "Retrospective investigator-reported data", "Excluded missing mandatory records", "No external central review", "Self-reported or note-derived ethnicity", "Small treatment subgroups", "No causal treatment comparison", "Commercial registry funding and author relationships" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Stage and performance at referral determine whether surgery or chemotherapy remains possible." }, "animalEvidence": { "rating": "not-applicable", "reason": "The claim concerns a human care route." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Laboratory work cannot recover people absent from a referral registry." }, "reproducibility": { "rating": "weak", "reason": "This is the first reported multicentre Latin American CCA cohort and lacks population linkage." }, "commercialConflicts": { "rating": "moderate", "reason": "Incyte funded the registry and a senior author disclosed grants, fees and consulting relationships, although funders were reported to have no study role." }, "recency": { "rating": "current", "reason": "The cohort was published in late 2024 and remains the current regional registry report." } } }, "updated": "2026-09-14" }, { "id": "claim-live-funding-routes-2026", "title": "Live 2026 funding routes cover slices of the mission, not the whole programme", "shortTitle": "Live funding routes", "type": "claim", "status": "supported", "scope": [ "pdac", "pnet", "cholangiocarcinoma", "gallbladder", "hcc", "colorectal", "pan-cancer" ], "summary": "Current UK, EU, US and charitable routes range from £250,000 translation awards to a £25.4m diagnostic call and rolling ARPA-H concepts, each with different eligibility and scope.", "content": "A fundable mission must be split by evidence stage and by the restrictions of each source. As of 14 September 2026, Innovate UK and the Office for Life Sciences have a £25.4m call-level fund for collaborative clinical validation of early cancer diagnostics, closing 28 October; it requires a UK-led consortium with an SME. Wellcome Discovery Awards accept investigator-led discovery programmes, average £3.5m over seven years and have a 22 September deadline, but do not replace a pivotal commercial trial. Cancer Research UK's Therapeutic Catalyst offers £250,000 or £350,000 for 12-24 months of antibody or small-molecule translation with its drug-discovery team, closing 25 September. The 2026 Horizon Cancer Mission call closes 15 September and covers fixed topics such as microbiome-based early prediction and pragmatic immunotherapy trials; a broad PDAC proposal outside those topics is ineligible. ARPA-H's four mission-office routes accept rolling concept summaries for distinct high-risk work, while its existing ADAPT programme already covers evolving-tumour measurement and treatment recommendation, so a proposal must identify a non-duplicative bottleneck. NCI mechanisms, SPOREs and cooperative groups remain the main US academic and trial routes. No listed route alone fits £100m-£500m. That scale needs a lead philanthropic or public anchor, then health-system, drug, assay and trial-network commitments tied to one protocol, data rights and stop rules.", "sourceIds": [ "ukri-accelerated-cancer-diagnosis-2026", "wellcome-2026-discovery-awards", "cruk-therapeutic-catalyst-2026", "horizon-cancer-mission-calls-2026", "arpa-h-open-funding-2026", "arpa-h-adapt-2026", "nci-2026-spore-funding" ], "links": [ { "target": "funding", "relation": "part-of" }, { "target": "funding-route-stage-fit", "relation": "extends" }, { "target": "programme-5m-decision-core", "relation": "may-fund-part-of" }, { "target": "programme-25m-prospective-observatory", "relation": "may-fund-part-of" }, { "target": "programme-100m-randomized-platform", "relation": "cannot-fund-alone" } ], "evidence": { "humanProspective": "not-applicable", "humanRetrospective": "not-applicable", "randomised": "not-applicable", "independentReplication": "moderate", "sampleSize": "Five current scheme pages plus NCI route guidance", "effect": "Live route and stage-fit map as of 14 September 2026", "limits": [ "Deadlines and calls change", "Call-level funds are not project award sizes", "Eligibility requires full scheme review", "ARPA-H rolling routes do not imply programme interest", "Large syndicate terms require negotiation" ] }, "layout": { "x": 82, "y": 118 }, "updated": "2026-09-14" }, { "id": "claim-liver-niche-before-metastasis", "title": "PDAC can alter the liver before detectable metastasis in models; the human importance is uncertain", "shortTitle": "Liver-niche signal", "type": "claim", "status": "open", "scope": [ "pdac", "hepatobiliary" ], "summary": "Tumour exosomes carrying MIF primed a fibrotic, macrophage-rich liver niche in mice; a small human association has not established a treatment target.", "content": "Mouse experiments found that PDAC-derived exosomes were taken up by liver Kupffer cells, inducing TGF-beta, fibronectin production by hepatic stellate cells and macrophage recruitment. Blocking exosomal macrophage migration inhibitory factor reduced niche formation and metastasis in the model. A patient sample comparison reported higher exosomal MIF in stage I tumours that later produced liver metastases than in tumours that did not progress. This makes host-organ conditioning a plausible step between local disease and liver metastasis. The human effect size, timing, causal share and treatment window remain uncertain. The decisive work would prospectively sample patients before recurrence, replicate the marker independently, and test whether the pathway can be changed without harming liver immunity or repair.", "sourceIds": [ "costa-silva-2015-liver-niche" ], "links": [ { "target": "residual-disease", "relation": "supports" }, { "target": "hepatobiliary", "relation": "may-transfer-to" }, { "target": "unknowns", "relation": "supports" }, { "target": "pdac", "relation": "part-of" } ], "evidence": { "humanProspective": "none", "humanRetrospective": "weak", "randomised": "none", "independentReplication": "weak", "sampleSize": "Mouse intervention studies plus a small human exosome comparison", "effect": "Exosomal MIF blockade reduced mouse liver niche formation and metastasis; human samples showed an association", "limits": [ "Causal treatment result is from mice", "Small human comparison", "No prospective clinical validation", "Exosome measurement methods can vary", "Host-pathway blockade may cause normal-tissue harm" ] }, "layout": { "x": 89, "y": 53 }, "updated": "2026-09-14" }, { "id": "claim-localised-outcomes-differ-by-cancer", "title": "A stage shift has very different value in different cancers", "shortTitle": "Stage shift differs", "type": "claim", "status": "supported", "scope": [ "pancreatic-all", "colorectal", "hepatobiliary" ], "summary": "Five-year relative survival for localised disease is 91.3% in colorectal cancer, 43.6% in pancreatic cancer and 37.4% in the combined liver and intrahepatic-bile-duct category.", "content": "These registry groups are not biologically equivalent, and the liver category itself combines different cancers. The figures are useful precisely because they prevent a false shared-cancer strategy. For colorectal cancer, preventing or finding disease while local can already approach the mission outcome at population level. For pancreatic and liver or intrahepatic-bile-duct cancer, earlier diagnosis is necessary but current treatment of localised disease remains insufficient. None of these associations proves the result of a new screening programme.", "sourceIds": [ "seer-pancreas-statfacts", "seer-colorectal-statfacts", "seer-liver-intrahepatic-statfacts" ], "links": [ { "target": "mission", "relation": "supports" }, { "target": "detection", "relation": "supports" }, { "target": "treatment", "relation": "supports" }, { "target": "transfer", "relation": "supports" } ], "evidence": { "humanProspective": "none", "humanRetrospective": "strong", "randomised": "not-applicable", "independentReplication": "moderate", "sampleSize": "Population registries, SEER 21", "effect": "Localised five-year relative survival: colorectal 91.3%; pancreatic 43.6%; liver and intrahepatic bile duct 37.4%", "limits": [ "Registry association", "Different tumour group definitions", "Liver category combines HCC and intrahepatic cholangiocarcinoma", "Five-year survival is not proof of eradication" ] }, "layout": { "x": 62, "y": 35 }, "updated": "2026-09-14" }, { "id": "claim-mced-early-stage-sensitivity-limit", "title": "High-specificity methylation testing still misses many stage I cancers", "shortTitle": "Stage I signal limit", "type": "claim", "status": "supported", "scope": [ "pan-cancer", "pdac", "hepatobiliary", "colorectal" ], "summary": "In CCGA validation, sensitivity was 18% across all stage I cancers and 39% in a 12-cancer high-mortality set at 99.3% specificity.", "content": "The CCGA targeted-methylation validation set achieved 99.3% specificity. Sensitivity rose sharply with stage. Across all cancer types it was 18% at stage I, 43% at stage II, 81% at stage III and 93% at stage IV. For a named set of 12 high-mortality cancers that included pancreas, liver or bile duct and colorectum, stage I sensitivity was 39%. Those are pooled results, not pancreatic-specific estimates. The pattern exposes the central physics and biology problem: the cancers for which earlier detection matters most release the least signal. Better classification of abundant late-stage signal does not solve stage I detection. Required progress is higher signal capture at fixed specificity, repeated longitudinal measurement, or prior risk enrichment that raises prevalence before the blood test is used.", "sourceIds": [ "liu-2020-ccga-methylation" ], "links": [ { "target": "detection", "relation": "supports" }, { "target": "claim-screening-base-rate-maths", "relation": "supports" }, { "target": "bottlenecks", "relation": "supports" } ], "evidence": { "humanProspective": "none", "humanRetrospective": "moderate", "randomised": "none", "independentReplication": "moderate", "sampleSize": "6,689 participants: 2,482 cancer and 4,207 non-cancer; training and validation split", "effect": "Validation specificity 99.3%; all-cancer stage I sensitivity 18%; 12-cancer-set stage I sensitivity 39%", "limits": [ "Case-control design", "Pooled cancer estimates", "No pancreatic-specific stage I estimate in the abstract", "No mortality or net-benefit endpoint", "GRAIL-funded study" ] }, "layout": { "x": 29, "y": 119 }, "updated": "2026-09-14" }, { "id": "claim-mced-workflow-feasible-utility-unproved", "title": "Multicancer blood-test follow-up is feasible, but mortality benefit remains unproved", "shortTitle": "MCED utility gap", "type": "unknown", "status": "open", "scope": [ "pan-cancer", "pdac", "hepatobiliary", "colorectal" ], "summary": "Prospective studies show that MCED workups can operate; the 142,250-person NHS randomized trial missed its main late-stage endpoint and has not shown fewer deaths.", "content": "DETECT-A tested 10,006 women and used PET-CT to confirm and locate positive blood tests. It found 26 cancers through blood testing; false-positive results sent 1% to PET-CT and 0.22% to a futile invasive procedure. PATHFINDER returned methylation-test results to 6,621 analysable adults. Ninety-two had a cancer signal; 35 were diagnosed with cancer and 57 were not, a 38% positive predictive value. Median time to resolution was 79 days and 93% of false-positive participants had imaging. NHS-Galleri then randomized about 142,250 adults to annual Galleri plus usual screening or usual screening alone. It did not reduce the prespecified combined stage III/IV endpoint, IRR 1.03, but reported a secondary 14% reduction in stage IV disease. These studies prove that the workflow can operate at scale. They do not yet prove fewer cancer deaths or benefit for pancreatic cancer specifically. Longer randomized mortality follow-up, cancer-specific effects, harms, overdiagnosis and cost remain the gate.", "sourceIds": [ "lennon-2020-detect-a", "schrag-2023-pathfinder", "sasieni-2026-nhs-galleri" ], "links": [ { "target": "detection", "relation": "part-of" }, { "target": "unknowns", "relation": "part-of" }, { "target": "failure-nhs-galleri-primary-stage-shift", "relation": "constrained-by" }, { "target": "claim-case-control-blood-tests-overstate-screening-readiness", "relation": "extends" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "not-applicable", "randomised": "strong", "independentReplication": "moderate", "sampleSize": "DETECT-A 10,006; PATHFINDER 6,621 analysable; NHS-Galleri approximately 142,250 randomized", "effect": "Workflow feasible; NHS-Galleri main stage III/IV IRR 1.03, secondary stage IV IRR 0.86; mortality unknown", "limits": [ "No mortality result", "Main NHS-Galleri endpoint missed", "Cancer types pooled", "Few or unreported pancreatic-specific events", "Follow-up work-up varied", "Company funding and inventor interests", "Overdiagnosis and net benefit unresolved" ] }, "layout": { "x": 37, "y": 113 }, "updated": "2026-09-14" }, { "id": "claim-mcrc-early-seeding", "title": "Colorectal metastases can be seeded before the primary tumour is clinically detectable", "shortTitle": "Early CRC seeding", "type": "claim", "status": "supported", "scope": [ "colorectal" ], "summary": "A multi-region sequencing and modelling study inferred early metastatic seeding in 17 of 21 evaluable metastatic colorectal cancers.", "content": "The result warns against equating removal of a small primary tumour with elimination of systemic risk. It also found low genomic divergence between primary and metastasis and early acquisition of canonical drivers, which may preserve targetability across sites. The timing is inferred from 23 metastatic patients and a spatial model, not observed prospectively. It should guide residual-disease research, not be treated as a universal percentage.", "sourceIds": [ "hu-2019-crc-early-seeding" ], "links": [ { "target": "colorectal", "relation": "part-of" }, { "target": "disease-system", "relation": "supports" }, { "target": "residual-disease", "relation": "supports" }, { "target": "transfer", "relation": "may-transfer-to" } ], "evidence": { "humanProspective": "none", "humanRetrospective": "moderate", "randomised": "not-applicable", "independentReplication": "weak", "sampleSize": "118 biopsies from 23 patients; 21 evaluable for timing", "effect": "Early seeding inferred in 17 of 21 evaluable patients", "limits": [ "Small selected metastatic cohort", "Model-derived timing", "Liver and brain metastases", "Not a population prevalence estimate" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Dissemination before diagnosis is compatible with tumour growth and phylogenetic branching." }, "animalEvidence": { "rating": "moderate", "reason": "Models support early spread but cannot give the human frequency." }, "inVitroEvidence": { "rating": "weak", "reason": "Cell systems do not reconstruct clinical seeding time." }, "reproducibility": { "rating": "weak", "reason": "The exact 17-of-21 estimate comes from one small selected study." }, "commercialConflicts": { "rating": "low", "reason": "The main concern is model and sampling bias, not a product incentive." }, "recency": { "rating": "historical", "reason": "The study was published in 2019 and remains a bounded timing result." } } }, "layout": { "x": 93, "y": 42 }, "updated": "2026-09-14" }, { "id": "claim-men1-surveillance-finds-lesions-not-danger", "title": "MEN1 surveillance finds pNET early but still cannot identify the dangerous lesion", "shortTitle": "MEN1 surveillance limit", "type": "claim", "status": "supported", "scope": [ "pnet" ], "summary": "Inherited risk justifies imaging from adolescence, but tumour count, size and a circulating RNA score do not yet form a proven metastasis-prevention rule.", "content": "DutchMEN data from 350 people show why inherited pNET risk is actionable but not solved. Five developed a clinically relevant non-functioning pNET before age 18, defined as at least 20 mm or rapid growth, and two later developed lymph-node metastases. The modelled ages at which 1%, 2.5% and 5% had developed a clinically relevant tumour were 9.5, 13.5 and 17.8 years, supporting a discussion of starting pancreatic imaging around age 13-14. Detection is not the same as knowing which lesion to treat. In a DutchMEN treatment cohort covering more than 90% of the national MEN1 population, 53 of 152 patients had surgery and 99 watchful waiting. Adjusted surgery was not associated with significantly less liver metastasis or death: HR 0.73, 95% CI 0.25-2.11. The under-2-cm estimate was HR 2.04 with CI 0.31-13.59, too imprecise to claim benefit or harm. Above 3 cm, five of six observed patients reached liver metastasis or death, so the data do not support indefinite observation of large disease. A prospective 42-person non-randomized comparison found much less radiographic progression with lanreotide than observation over median 73 months, but did not show fewer operations, metastases or deaths. The current action is therefore surveillance tied to size, rate, duct effects, hormone syndrome and shared review. The missing action is a validated marker of future spread or functional harm that improves the timing of surgery or treatment.", "sourceIds": [ "klein-haneveld-2021-men1-young-screening", "nell-2018-men1-surgery-watchful-waiting", "faggiano-2020-men1-lanreotide", "van-vliembergen-2026-men1-mnet-assay" ], "links": [ { "target": "pnet", "relation": "supports" }, { "target": "claim-germline-risk-subset", "relation": "contrasts-with" }, { "target": "failure-men1-mnet-assay-no-progression-prediction", "relation": "limited-by" }, { "target": "unknown-small-pnet-surgery-selector", "relation": "exposes" }, { "target": "trial-small-pnet-action-map", "relation": "supports" } ], "evidence": { "humanProspective": "moderate — prospective imaging, drug and biomarker components", "humanRetrospective": "strong — national population cohort", "randomised": "none", "independentReplication": "moderate — lesion detection and size risk repeat, action benefit does not", "sampleSize": "Young-screening cohort n=350; management cohort n=152; lanreotide comparison n=42; mNET assay n=110 eligible", "effect": "Imaging detects early lesions; no cited selector or intervention has shown lower pNET mortality in MEN1", "limits": [ "Rare childhood events", "Size-based clinical relevance includes no pathology", "Confounded surgery choice", "Small non-randomized drug comparison", "No schedule randomization", "No quality-adjusted lifetime outcome" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Germline MEN1 creates early multifocal pNET risk, but individual lesions can follow very different courses." }, "animalEvidence": { "rating": "weak", "reason": "MEN1 models support tumour formation but cannot set human imaging or surgery timing." }, "inVitroEvidence": { "rating": "weak", "reason": "Molecular assays have not predicted the clinical course in the prospective human cohort." }, "reproducibility": { "rating": "moderate", "reason": "Early and multifocal pNET detection repeats, while mortality-reducing action selection does not." }, "commercialConflicts": { "rating": "moderate", "reason": "Long-term imaging, drug treatment and proprietary blood assays can each benefit from broader use." }, "recency": { "rating": "current", "reason": "The conclusion includes a 2026 negative prospective biomarker result." } } }, "updated": "2026-09-14" }, { "id": "claim-metastases-can-be-polyclonal", "title": "A PDAC metastasis can be founded by more than one tumour clone in mice", "shortTitle": "Polyclonal seeding", "type": "claim", "status": "open", "scope": [ "pdac" ], "summary": "Lineage tracing found frequent multi-clone metastatic seeding in a mouse model, raising the treatment bar if this is common in patients.", "content": "Multicolour lineage tracing in a native mouse PDAC model found that a substantial fraction of metastases were seeded by more than one tumour clone and that later clonal expansion differed by organ. If common in human PDAC, this means a metastasis can import complementary resistance traits at founding rather than acquire them one at a time. It would weaken any treatment plan based on one sampled clone. The observation is strong within the model but its human frequency is unresolved. Multi-region primary and metastatic sequencing, ideally with phylogenetic and spatial methods that preserve low-frequency clones, is needed before using it to design patient treatment.", "sourceIds": [ "maddipati-2015-polyclonal-metastasis" ], "links": [ { "target": "residual-disease", "relation": "supports" }, { "target": "claim-pdac-progression-timing-uncertain", "relation": "supports" }, { "target": "unknowns", "relation": "supports" }, { "target": "pdac", "relation": "part-of" } ], "evidence": { "humanProspective": "none", "humanRetrospective": "none", "randomised": "not-applicable", "independentReplication": "weak", "sampleSize": "Genetically engineered mouse PDAC lineage-tracing study", "effect": "A substantial fraction of model metastases had more than one colour lineage", "limits": [ "Mouse-model result", "Colour lineage is not full genomic resolution", "Human frequency unknown", "Organ-specific outgrowth complicates sampling" ] }, "layout": { "x": 96, "y": 36 }, "updated": "2026-09-14" }, { "id": "claim-metastatic-folfirinox-not-durable-control", "title": "FOLFIRINOX improved metastatic PDAC survival without producing durable control for most patients", "shortTitle": "Metastatic ceiling", "type": "claim", "status": "supported", "scope": [ "pdac" ], "summary": "A major chemotherapy advance extended median overall survival from 6.8 to 11.1 months in fit patients, a large relative gain but far short of durable control.", "content": "The French PRODIGE 4/ACCORD 11 trial randomised 342 patients with metastatic pancreatic cancer and ECOG performance status 0 or 1. FOLFIRINOX increased median overall survival from 6.8 to 11.1 months, a 4.3-month absolute gain and a 43% relative reduction in the hazard of death. Median progression-free survival was 6.4 versus 3.3 months, and response was 31.6% versus 9.4%. Toxicity increased. This was a decisive treatment improvement, but median survival below one year in the fitter trial population means the regimen did not make metastatic PDAC durably controllable.", "sourceIds": [ "conroy-2011-folfirinox-metastatic" ], "links": [ { "target": "treatment", "relation": "supports" }, { "target": "bottlenecks", "relation": "supports" }, { "target": "pdac", "relation": "part-of" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong", "independentReplication": "strong", "sampleSize": "342 randomised patients with metastatic pancreatic cancer and ECOG 0 or 1", "effect": "Median OS 11.1 versus 6.8 months; death hazard ratio 0.57; response 31.6% versus 9.4%", "limits": [ "Patients had good performance status", "Regimen caused more adverse events", "The source used the broader label pancreatic cancer although the trial population is treated as pancreatic adenocarcinoma in practice" ] }, "layout": { "x": 83, "y": 68 }, "updated": "2026-09-14" }, { "id": "claim-modern-metastatic-chemotherapy-remains-short-control", "title": "Modern first-line chemotherapy still controls metastatic PDAC for months, not years, for most patients", "shortTitle": "Modern chemo ceiling", "type": "claim", "status": "supported", "scope": [ "pdac" ], "summary": "NAPOLI 3 improved median survival by 1.9 months over gemcitabine plus nab-paclitaxel, while median survival remained 11.1 months and severe adverse events affected 87%.", "content": "NAPOLI 3 randomised 770 fit, untreated patients with metastatic PDAC across 18 countries. NALIRIFOX increased median overall survival from 9.2 to 11.1 months; hazard ratio 0.83, P=0.036. The absolute median gain was 1.9 months. Grade 3 or higher treatment-emergent adverse events occurred in 87% and 86% of the groups, and treatment-related deaths in 2% of each. This is a valid treatment gain, but it leaves the mission almost untouched: the median patient still died within one year and treatment burden was high. Compared with the 2011 FOLFIRINOX trial's 11.1-month median, regimen engineering has not moved median survival into durable-control territory, although the trials cannot be directly compared across eras and populations.", "sourceIds": [ "wainberg-2023-napoli3", "conroy-2011-folfirinox-metastatic" ], "links": [ { "target": "claim-metastatic-folfirinox-not-durable-control", "relation": "extends" }, { "target": "treatment", "relation": "part-of" }, { "target": "bottlenecks", "relation": "supports" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong", "independentReplication": "strong", "sampleSize": "770 randomised patients in NAPOLI 3; 342 in the earlier FOLFIRINOX trial", "effect": "NALIRIFOX median OS 11.1 versus 9.2 months; hazard ratio 0.83; severe adverse events 87% versus 86%", "limits": [ "ECOG 0 or 1 population", "Open-label trial", "Cross-trial comparison to 2011 is descriptive", "Median survival does not describe rare long survivors", "Ipsen-funded" ] }, "layout": { "x": 77, "y": 76 }, "updated": "2026-09-14" }, { "id": "claim-modifiable-risk-reduces-but-cannot-eliminate-pdac", "title": "Smoking and excess weight are real prevention targets, but their removal would prevent only a minority of pancreatic cancers", "shortTitle": "Partial prevention ceiling", "type": "claim", "status": "supported", "scope": [ "pancreatic-all" ], "summary": "Cohorts consistently support smoking and excess weight as modifiable risk, but modelled removal covers a minority of cases and intervention studies have not proved a pancreatic-specific prevention effect.", "content": "Smoking cessation and weight control are useful public-health actions, not a complete pancreatic-cancer strategy. In ten Japanese population cohorts with 1,779 cancers, current smoking was associated with hazard ratios of 1.59 in men and 1.81 in women. Male risk returned near never-smoker levels after five years in that analysis; female attenuation was not seen. A later Multiethnic Cohort analysis followed 182,011 people and observed 1,831 cancers: 50 pack-years was associated with 91% excess risk and each year since quitting with 9% lower excess risk. This is a strong dose-and-time pattern, but quitting was not randomized. A UK Biobank model with 728 incident cancers estimated that removing smoking and obesity together could prevent about 16% of cases, while abdominal obesity alone had a 22% modelled fraction. Those fractions cannot be added because exposures overlap, and they assume associations are causal and completely removable. Bariatric surgery and GLP-1 receptor agonist cohorts report lower pancreatic-cancer incidence, but rare outcomes, treatment selection, short exposure windows and reverse causation prevent a prescription for cancer prevention. Diabetes is especially hard to count because recent diabetes can be caused by occult cancer. The safe conclusion is that cessation, healthy weight and metabolic care can reduce risk and improve health beyond cancer, while most pancreatic cancers remain and no named intervention has pancreatic-specific randomized prevention proof.", "sourceIds": [ "ando-2019-smoking-japan", "bogumil-2024-smoking-cessation-multiethnic", "zhang-2022-uk-biobank-risk-fractions", "schauer-2019-bariatric-pancreatic-risk", "wang-2025-glp1-pancreatic-risk" ], "links": [ { "target": "model-pdac-causal-chain", "relation": "supports" }, { "target": "programme", "relation": "part-of" }, { "target": "claim-germline-risk-subset", "relation": "extends" }, { "target": "ranking-provisional-pdac-bottlenecks", "relation": "supports" }, { "target": "claim-pdac-primary-prevention-action-gap", "relation": "depends-on" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "none", "independentReplication": "strong", "sampleSize": "Japanese pooled cohorts n=354,154 with 1,779 cancers; Multiethnic Cohort n=182,011 with 1,831 cancers; UK Biobank 728 cancers; intervention-adjacent cohorts n=88,625 and n=1,636,056", "effect": "Current-smoking HR 1.59 in men and 1.81 in women; 9% lower excess risk per year quit; UK model attributed roughly 16% to smoking plus obesity", "limits": [ "Observational exposure and treatment evidence", "All-pancreatic histologies may be included", "Population-attributable fractions require causal and removal assumptions", "Fractions overlap and cannot be added", "UK Biobank is not population representative", "No randomized pancreatic-cancer prevention trial" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Tobacco exposure has a dose-and-time pattern and obesity or metabolic injury can affect inflammation and growth signals." }, "animalEvidence": { "rating": "moderate", "reason": "Models support carcinogen and metabolic mechanisms but cannot establish the population effect or removable fraction." }, "inVitroEvidence": { "rating": "weak", "reason": "Cell work cannot reproduce decades of exposure, quitting or competing disease." }, "reproducibility": { "rating": "strong", "reason": "Smoking and excess-weight associations repeat across large cohorts, although intervention effects remain unproved." }, "commercialConflicts": { "rating": "moderate", "reason": "Risk cohorts are mainly noncommercial, while surgery and diabetes-drug prevention claims create commercial incentives." }, "recency": { "rating": "mixed", "reason": "The claim combines long-running exposure evidence with 2024 and 2025 intervention-adjacent analyses." } } }, "layout": { "x": 22, "y": 39 }, "updated": "2026-09-14" }, { "id": "claim-mrd-platform-access-is-more-than-patent-license", "title": "An MRD programme needs assay, data, supply and regulatory access, not only a patent licence", "shortTitle": "MRD access stack", "type": "claim", "status": "supported", "scope": [ "pdac", "colorectal", "pan-cancer", "transfer-unproven" ], "summary": "Natera's filing describes more than 650 patents, centralized Signatera processing, one validated sequencing platform and pending companion-diagnostic submissions; each can block an independent action trial.", "content": "The ownership question is not solved by naming WO2019200228A1. Natera's 2025 annual filing says it held more than 650 issued or pending US and foreign patents across its businesses. It describes Signatera as a personalized residual-disease and recurrence test, explicitly not a test that matches a patient to a particular therapy. It is offered clinically as a laboratory-developed test in Natera's own certified laboratories. Austin and San Carlos process Signatera and two other products that form the large majority of company revenue, and the filing says there is otherwise no backup or redundant site for tests that run at only one laboratory. Natera also says its molecular tests are validated only on Illumina sequencing, that Illumina is the sole supplier of sequencers and related reagents under an agreement running to August 2033, and that primary Signatera analysis uses a sole-sourced blood tube. Signatera companion-diagnostic premarket applications remain under review; breakthrough designation does not establish approval. Therefore an independent MRD action trial needs at least seven negotiated items: assay specification, sample processing, result turnaround, raw and derived data, algorithm version control, supply continuity, and permission to publish negative results. If the assay controls randomization, it also needs a failure plan and a frozen rule that survives a vendor update. The patent family matters, but it is only one layer of operational control. This is not a claim that Natera will deny access or that every patent is valid. It is a contract and replication requirement inferred from the company's stated operating model.", "sourceIds": [ "natera-2025-10k", "patent-wo2019200228-natera-mrd", "botta-2024-tumor-informed-ctdna", "bando-2026-altair", "clinicaltrials-nct05174169-circulate-us" ], "links": [ { "target": "patent-landscape", "relation": "extends" }, { "target": "company-natera", "relation": "depends-on" }, { "target": "trial-crc-action-map", "relation": "constrains" }, { "target": "trial-pdac-ctdna-action-map", "relation": "constrains" }, { "target": "programme-500m-integrated-mission", "relation": "constrains" }, { "target": "risks", "relation": "supports" } ], "evidence": { "humanProspective": "not-applicable", "humanRetrospective": "not-applicable", "randomised": "not-applicable", "independentReplication": "moderate for patent family and trial use; weak for contract availability", "sampleSize": "One audited company filing, one priority patent family, one PDAC cohort and two colorectal action programmes", "effect": "Seven access layers identified; no efficacy effect", "limits": [ "Not a legal opinion", "More than 650 patents cover all Natera businesses, not only Signatera", "The filing states business risks rather than a mission contract position", "No licence terms were obtained", "Supply arrangements can change", "Pending regulatory submissions are not approvals or denials", "The cited patent family is not an exhaustive claim map" ], "dimensions": { "biologicalPlausibility": { "rating": "not-applicable", "reason": "The claim concerns control of a testing operation rather than tumour biology." }, "animalEvidence": { "rating": "not-applicable", "reason": "Animal evidence cannot establish assay, data, supply or legal access." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Analytical performance does not answer ownership, continuity or publication rights." }, "reproducibility": { "rating": "moderate", "reason": "The filing and public patent record establish the layers, but no independent contract test has been performed." }, "commercialConflicts": { "rating": "high", "reason": "The assay company controls key laboratory, algorithm, data and supply relationships and benefits from wider use." }, "recency": { "rating": "current", "reason": "The operating facts come from the 2025 annual report filed in February 2026 and a trial record checked in September 2026." } } }, "updated": "2026-09-14" }, { "id": "claim-mtap-prmt5-selected-human-signal", "title": "MTAP deletion creates a selected PRMT5 route with human target inhibition and early responses", "shortTitle": "MTAP-PRMT5 signal", "type": "claim", "status": "mixed", "scope": [ "pdac", "cholangiocarcinoma", "pan-cancer" ], "summary": "About 28% of PDAC loses MTAP; two PRMT5 inhibitors show human activity, but pancreatic survival and the added value of the RAS combination remain unproved.", "content": "System model. MTAP is an enzyme in a recycling path. The MTAP gene sits beside the CDKN2A tumour-suppressor gene, so cancers often delete both. MTAP loss lets the metabolite MTA accumulate. MTA partly slows PRMT5, an enzyme the cell still needs to control RNA processing and protein marks. An MTA-cooperative PRMT5 inhibitor binds best in this already stressed state and can push the cancer below its viability threshold while leaving MTAP-intact cells more headroom. Synthetic lethality means either fault alone is survivable but the pair is lethal. Reach. In 8,834 Japanese pancreatic cancers, MTAP deletion occurred in 27.3% overall and 28.0% of PDAC. A separate Foundation Medicine series found 28.9% in pancreatic tissue. That is a large selected segment, not a rare mutation. Selection must be exact: 33.7% of tumours with CDKN2A loss did not have MTAP loss, liquid biopsy agreement depended on tumour fraction and partial deletions have uncertain sensitivity. Human mechanism. In the first AMG 193 study, paired biopsies at doses at least 480 mg showed complete tumour PRMT5 inhibition. At active tolerable doses, nine of 42 people across several cancer types responded, including pancreatic and biliary cancers; the paper does not give a pancreatic denominator. Vopimetostat monotherapy later produced a 15% response rate and 71% disease control in 39 pancreatic patients at active doses with at least six months since first dose; the second-line subset response was 25% and median PFS 7.2 months. Combination signal. Vopimetostat plus daraxonrasib produced 11 responses among 12 evaluable PDAC patients, nine confirmed. Vopimetostat plus zoldonrasib produced 14 among 27, ten confirmed. The combined 25 of 39 is 64%, but this arithmetic joins two drugs, several doses and short follow-up. There was no concurrent RAS-drug-alone arm, so it cannot identify how much vopimetostat adds. Decision. Promote the route above broad epigenetic treatment and model-only combinations, but below controlled survival evidence. Require exact MTAP loss, public all-treated follow-up and a randomized contribution test before clinical adoption or mission funding.", "sourceIds": [ "matsumoto-2026-mtap-pancreatic", "pavlick-2026-mtap-genomics", "rodon-2024-amg193-prmt5", "clinicaltrials-nct05732831-vopimetostat", "tango-2025-vopimetostat-monotherapy", "clinicaltrials-nct06922591-vopimetostat-ras", "tango-2026-vopimetostat-ras-data" ], "links": [ { "target": "treatment", "relation": "part-of" }, { "target": "failure-broad-epigenetic-treatment-pdac", "relation": "differs-from" }, { "target": "failure-vopimetostat-ras-contribution-unknown", "relation": "bounded-by" }, { "target": "unknown-mtap-prmt5-ras-contribution", "relation": "creates" }, { "target": "trial-mtap-prmt5-action-map", "relation": "mapped-by" }, { "target": "hypothesis-mtap-prmt5-ras-deeper-control", "relation": "supports" }, { "target": "company-tango-therapeutics", "relation": "developed-by" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "strong", "randomised": "none", "independentReplication": "moderate", "sampleSize": "Prevalence n=8,834 and 29,136 pancreatic tissue samples; AMG 193 active-dose n=42 pan-cancer; vopimetostat monotherapy PDAC n=39; combination evaluable PDAC n=39", "effect": "Vopimetostat monotherapy ORR 15%; two RAS combinations give 25 responses among 39 evaluable PDAC patients, without a control", "limits": [ "No randomized treatment result", "Short interim follow-up", "Company-controlled combination report", "Different RAS drugs and doses", "No pancreatic denominator for AMG 193" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "A defined deletion creates a measurable metabolite and enzyme dependency." }, "animalEvidence": { "rating": "moderate", "reason": "Models support selection and combinations but human data now set the gate." }, "inVitroEvidence": { "rating": "strong", "reason": "MTAP loss, MTA cooperation and PRMT5 inhibition are directly measurable." }, "reproducibility": { "rating": "moderate", "reason": "Two PRMT5 molecules show pan-tumour responses, while pancreatic contribution is not independently controlled." }, "commercialConflicts": { "rating": "high", "reason": "The largest pancreatic response estimates come from Tango disclosures." }, "recency": { "rating": "current", "reason": "Includes June 2026 combination data and current registries." } } }, "layout": { "x": 184, "y": 105 }, "updated": "2026-09-15" }, { "id": "claim-multiantigen-t-cells-feasible-not-effective-yet", "title": "Multiantigen T cells are feasible and can persist, but selected uncontrolled cohorts do not establish benefit", "shortTitle": "Multiantigen T cells", "type": "claim", "status": "open", "scope": [ "pdac" ], "summary": "Of 56 patients whose cells were collected, 37 received product; disease control was 84.6% in chemotherapy responders and 25% in refractory disease, and 2 of 9 resected patients remained disease-free at 66 months.", "content": "This autologous product expands nonengineered T cells against five tumour-associated antigens: PRAME, SSX2, MAGEA4, Survivin and NY-ESO-1. Thirty-seven of 56 patients whose cells were collected were infused. Only one treatment-related serious event occurred, and product-derived clones persisted for as long as 12 months. Disease control was 84.6% in 13 patients already responding to first-line chemotherapy and 25% in 12 refractory patients. Among nine resected evaluable patients, two remained disease-free at 66 months. These groups lack controls and differ strongly before cell therapy; disease control in prior responders cannot be assigned to the product. The procurement-to-infusion fraction was 37/56=66%, so one third did not reach treatment. The study proves manufacture, safety, persistence and biological activity. A next trial needs randomisation within a single disease state, an intention-to-treat endpoint from collection, antigen-expression tracking and a manufacturing time compatible with rapidly progressing PDAC.", "sourceIds": [ "musher-2026-multiantigen-t-cells" ], "links": [ { "target": "failure-mesothelin-car-t-pdac", "relation": "succeeds-where" }, { "target": "claim-kras-tcr-one-response-not-platform-proof", "relation": "extends" }, { "target": "treatment", "relation": "part-of" }, { "target": "experiments", "relation": "tests" } ], "evidence": { "humanProspective": "weak", "humanRetrospective": "not-applicable", "randomised": "none", "independentReplication": "none", "sampleSize": "56 procured, 37 infused across three cohorts; 9 evaluable after resection", "effect": "Disease control 84.6% in prior responders and 25% in refractory disease; 2 of 9 resected patients disease-free at 66 months", "limits": [ "No control group", "Safety and feasibility primary", "Cohorts selected by prior response", "One third of procured patients were not infused", "Small resected cohort", "Exploratory efficacy", "Extensive commercial conflicts" ] }, "layout": { "x": 82, "y": 103 }, "updated": "2026-09-14" }, { "id": "claim-mwa-pain-signal-no-added-benefit", "title": "Pancreatic microwave ablation has a pain signal but no reliable added-benefit estimate", "shortTitle": "MWA pain signal, benefit unknown", "type": "claim", "status": "supported", "scope": [ "pdac", "unresectable", "pain", "local-treatment" ], "summary": "Microwave heating can destroy part of a pancreatic tumour and may reduce pain; current survival comparisons are dominated by selection, missing imaging and bundled treatment.", "content": "Microwave ablation inserts an antenna and uses an oscillating electric field to make water molecules move, producing heat and local tissue death. In a four-hospital retrospective series of 72 people, the antenna was placed successfully in every procedure but only 30/74 tumours, 40.5%, were completely ablated. Seven of 72, 9.8%, had a grade IIIa complication requiring an intervention without general anaesthesia. Among 33 people with pain, 32 reported relief and the mean 0-to-10 score fell from 6.3 to 2.0. That is a large uncontrolled palliative signal. Median follow-up was only 4.6 months and median survival was 5.6 months. A newer 150-person record reported median survival 14.6 months after artery-delivered chemotherapy plus microwave versus 9.0 months after artery treatment alone, but microwave patients started with tumours 31% smaller and metastases 20 percentage points less often. Local progression among imaged people was almost identical, 10.6% versus 9.6%, and 20 microwave patients lacked imaging. A small prospective immune combination caused one treatment-path death. Microwave ablation heats tissue; whether it adds useful life or safer pain relief remains unknown.", "sourceIds": [ "zhou-2024-pancreatic-mwa", "vogl-2026-tace-mwa", "topal-2025-mwa-dual-ici", "clinicaltrials-nct06231160-mwa-cohort" ], "links": [ { "target": "failure-mwa-selection-bundling-and-missing-imaging", "relation": "bounded-by" }, { "target": "unknown-mwa-safe-coverage-and-patient-benefit", "relation": "leaves-open" }, { "target": "hypothesis-mwa-existing-data-gate", "relation": "motivates" }, { "target": "trial-pdac-mwa-action-map", "relation": "mapped-by" }, { "target": "claim-cryoablation-pain-signal-not-survival-proof", "relation": "compares-with" }, { "target": "claim-pelican-rfa-no-survival-and-more-harm", "relation": "compares-with" }, { "target": "treatment-modality-ranking", "relation": "updates" }, { "target": "treatment", "relation": "part-of" } ], "evidence": { "humanProspective": "weak", "humanRetrospective": "moderate", "randomised": "none", "independentReplication": "weak", "sampleSize": "Retrospective n=72 and n=150; prospective package n=12", "effect": "Pain improved in 32/33 uncontrolled cases; complete ablation 40.5%; survival difference confounded", "limits": [ "No randomized device comparison", "Major baseline imbalance", "Twenty of 67 microwave cases lacked imaging", "Microwave bundled with artery therapy or immunotherapy", "Short follow-up", "Only 40.5% complete ablation", "One fatal treatment-path event" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Microwave energy produces direct local heating and tissue death." }, "animalEvidence": { "rating": "moderate", "reason": "Thermal field behavior is established, but human net benefit is the missing quantity." }, "inVitroEvidence": { "rating": "strong", "reason": "Temperature, energy and cell death can be measured directly." }, "reproducibility": { "rating": "weak", "reason": "Feasibility repeats, while complete coverage and patient benefit do not." }, "commercialConflicts": { "rating": "moderate", "reason": "Multiple commercial systems are used; supplier-neutral data remain necessary." }, "recency": { "rating": "current", "reason": "The record includes February 2026 clinical data and current registry status." } } }, "layout": { "x": 874, "y": 798 }, "updated": "2026-09-15" }, { "id": "claim-neoadjuvant-therapy-improves-selection-not-cure", "title": "Treatment before PDAC surgery improves some route outcomes, but it is not a universal survival rule", "shortTitle": "Treatment before surgery", "type": "claim", "status": "supported", "scope": [ "pdac" ], "summary": "Across nine randomized resectable-PDAC trials, treatment first improves event-free survival and reduces noncurative exploration, while overall survival remains inconclusive and individual trials conflict.", "content": "The strongest bounded conclusion is about the route, not a universal sequence. PREOPANC randomized 246 eligible resectable or borderline patients. Gemcitabine-based chemoradiotherapy before surgery improved long-term OS, HR .73, and five-year survival from 6.5% to 20.5%, but changed timing, radiation and chemotherapy together. NORPACT-1 then randomized 140 resectable pancreatic-head cases and moved in the opposite direction: treatment-first median OS was 25.1 versus 38.5 months, HR 1.52; only 61/77 assigned patients received neoadjuvant FOLFIRINOX and severe events were 58% versus 40%. CISPD-1, a 324-person single-centre Chinese phase 3 trial, again favoured treatment first: event-free survival 15.3 versus 10.9 months, HR .71, and median OS 35.4 versus 27.2 months, HR .73 with the confidence interval reaching 1.00; severe events were 47.6% versus 30.7%. Its sequential two-regimen route and 50% body-or-tail tumour mix limit transfer. The 2026 nine-trial analysis, 604 treatment-first and 527 surgery-first patients, found event-free survival HR .77, fewer noncurative explorations, RR .90, and more node-negative specimens, RR 1.73. Overall survival remained inconclusive, HR .85, 95% CI .68-1.05. PREOPANC-2 found no survival difference between all-neoadjuvant FOLFIRINOX and gemcitabine chemoradiotherapy plus postoperative gemcitabine, HR .88, while A021501 argues against assuming radiation is the active component. Therefore treatment before surgery can improve event-free control and avoid futile operations, but the survival effect depends on patient state, regimen and reliable delivery. NCT04340141 A021806 has completed accrual of 358 and directly compares the same twelve mFOLFIRINOX cycles split before and after surgery against all twelve after upfront surgery. Its overall-survival result is the next gate.", "sourceIds": [ "versteijne-2022-preopanc-longterm", "preopanc2-2025", "katz-2022-a021501", "labori-2024-norpact1", "bai-2025-cispd1", "tan-2026-neoadjuvant-ipd-meta", "ghaneh-2023-espac5", "clinicaltrials-nct04340141-a021806" ], "links": [ { "target": "treatment", "relation": "part-of" }, { "target": "residual-disease", "relation": "depends-on" }, { "target": "failure-local-intensification-pdac-survival-not-shown", "relation": "constrains" }, { "target": "claim-adjuvant-mfolfirinox-improves-but-recurrence-remains", "relation": "complements" }, { "target": "ranking-provisional-pdac-bottlenecks", "relation": "supports" }, { "target": "failure-norpact1-neoadjuvant-delivery", "relation": "contradicted-by" }, { "target": "unknown-pdac-perioperative-treatment-route", "relation": "leads-to" }, { "target": "trial-pdac-perioperative-sequencing-action-map", "relation": "mapped-by" }, { "target": "lab-alliance-a021806-network", "relation": "tested-by" }, { "target": "lab-zhejiang-pancreatic-centre", "relation": "supported-by" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong but conflicting", "independentReplication": "moderate for event-free survival; none for a universal OS rule", "sampleSize": "Nine randomized trials n=1,131; PREOPANC n=246; NORPACT-1 n=140; CISPD-1 n=324; A021806 actual n=358", "effect": "Pooled EFS HR .77; pooled OS HR .85, 95% CI .68-1.05", "limits": [ "Regimens and timing differ", "Resectable and borderline states differ", "PREOPANC bundled radiation and timing", "NORPACT-1 had delivery failure", "CISPD-1 was single-centre", "A021806 survival is pending" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Early systemic treatment can address occult disease and avoid futile surgery, while added toxicity and delay can remove benefit." }, "animalEvidence": { "rating": "not-applicable", "reason": "Human randomized route evidence decides this claim." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Laboratory systems cannot estimate operation loss, treatment delivery or survival." }, "reproducibility": { "rating": "moderate", "reason": "Pooled event-free benefit repeats, but overall-survival trials point in different directions." }, "commercialConflicts": { "rating": "low", "reason": "The main evidence comes from academic trials using established chemotherapy." }, "recency": { "rating": "current", "reason": "Includes a 2025 phase 3 result, a 2026 pooled analysis and September 2026 registry status." } } }, "layout": { "x": 10, "y": 111 }, "updated": "2026-09-14" }, { "id": "claim-neoantigen-vaccine-frontier", "title": "Personalised RNA neoantigen vaccination can produce long-lived T cells after PDAC surgery; efficacy remains unproved", "shortTitle": "RNA vaccine signal", "type": "claim", "status": "open", "scope": [ "pdac" ], "summary": "Eight immune responders in a 16-patient phase 1 cohort had longer observed recurrence-free survival than eight non-responders, but the comparison was not randomised.", "content": "A single-centre phase 1 programme combined an individualised RNA neoantigen vaccine, atezolizumab and mFOLFIRINOX after PDAC surgery. At median 3.2-year follow-up, vaccine-expanded T-cell clones persisted in immune responders. Recurrence-free survival was not reached in eight responders and was 13.4 months in eight non-responders, P=0.007. This is evidence of immune engagement and durable T-cell memory. It is not an estimate of treatment effect: response status was observed after treatment, there was no untreated vaccine control, the cohort was tiny, and a person's ability to mount a response may track other favourable biology. A randomised phase 2 study, NCT05968326, is the test that can move this from biological signal toward clinical efficacy.", "sourceIds": [ "rojass-2025-neoantigen-followup" ], "links": [ { "target": "residual-disease", "relation": "targets" }, { "target": "experiments", "relation": "tests" }, { "target": "treatment", "relation": "part-of" }, { "target": "pdac", "relation": "part-of" } ], "evidence": { "humanProspective": "weak", "humanRetrospective": "not-applicable", "randomised": "none", "independentReplication": "none", "sampleSize": "16 vaccinated and evaluable patients; 8 immune responders and 8 non-responders", "effect": "RFS not reached in responders versus 13.4 months in non-responders; P=0.007", "limits": [ "Phase 1", "Single centre", "No untreated vaccine control", "Response-defined groups can be confounded", "Combination treatment prevents attribution to one component", "Company involvement" ] }, "layout": { "x": 73, "y": 91 }, "updated": "2026-09-14" }, { "id": "claim-new-onset-diabetes-enriches-risk", "title": "New-onset diabetes can enrich pancreatic-cancer risk enough to justify a one-time work-up study", "shortTitle": "Diabetes risk gate", "type": "claim", "status": "open", "scope": [ "pancreatic-all" ], "summary": "Prospective U.S. surveillance found 0.62% three-year incidence after new-onset diabetes, but completed imaging studies contain only one pancreatic cancer each and EDI has no posted result.", "content": "Prospective active surveillance of 18,838 U.S. adults aged at least 50 with glycaemically defined new-onset diabetes found 82 pancreatic cancers and a race-adjusted three-year incidence of 0.62%. Incidence varied from 0.84% in non-Hispanic White participants to 0.22% in Asian or Pacific Islander participants. Diabetes preceded cancer diagnosis by eight months on average. A 402,663-person Korean cohort found a stepwise risk increase with initial treatment intensity and additional risk when treatment escalated within six months; that supports a longitudinal signal but not screening utility. END-PAC uses age, weight change and glucose change. In its locked retrospective validation cohort, a score of at least 3 found seven of nine cancers, 78% sensitivity, at 85% specificity; cancer prevalence in the high-score group was 3.6%, roughly one per 28. This is not the same threshold as EDI's score above 0. Completed work-up evidence remains small. A 93-person pancreas-CT pilot found one PDAC and it was stage IV; 56% had extrapancreatic incidental findings. PANDOME enrolled 109 and found one stage-I PDAC in its 12-person deteriorating-diabetes subgroup, after four pancreatic biopsies. The 8,869-participant EDI study is active but not recruiting, uses non-randomized site pathways, has no posted results and is expected to complete in 2028. The current bottleneck is not finding a relative-risk signal. It is producing a transportable absolute-risk rule, then proving that imaging or a blood-led work-up detects aggressive treatable disease with acceptable incidental procedures and reduces death.", "sourceIds": [ "chari-2026-prospective-new-onset-diabetes", "lee-2026-diabetes-progression-korea", "sharma-2018-endpac", "wu-2022-nod-imaging-pilot", "frank-2025-pandome", "edi-nct04662879" ], "links": [ { "target": "detection", "relation": "supports" }, { "target": "claim-screening-base-rate-maths", "relation": "succeeds-where" }, { "target": "experiments", "relation": "tests" }, { "target": "trial-new-onset-diabetes-workup-map", "relation": "tested-by" } ], "evidence": { "humanProspective": "weak", "humanRetrospective": "moderate", "randomised": "none", "independentReplication": "weak", "sampleSize": "Prospective GNOD cohort n=18,838 with 82 cancers; Korean cohort n=402,663; END-PAC validation n=1,096 with 9 cancers; CT pilot n=93; PANDOME n=109; EDI n=8,869 with results pending", "effect": "Prospective three-year incidence 0.62%; END-PAC at least 3 found 7/9 cancers at 85% specificity; completed imaging studies found one pancreatic cancer each", "limits": [ "Prospective cohort reports pancreatic cancer rather than PDAC only", "Marked racial and geographic differences", "Risk does not prove screening benefit", "END-PAC has very few validation cancers", "Other illnesses can create high scores", "Pilot imaging evidence is extremely small", "EDI is non-randomized by site and unfinished" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Occult pancreatic cancer can alter glucose and weight before symptoms, and timing is concentrated near diagnosis." }, "animalEvidence": { "rating": "not-applicable", "reason": "The risk-gate and care-pathway questions require intended-population human evidence." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Laboratory systems cannot establish absolute risk, racial transport or work-up benefit." }, "reproducibility": { "rating": "moderate", "reason": "Risk enrichment repeats across large cohorts, but thresholds, absolute incidence and imaging yield do not yet transport reliably." }, "commercialConflicts": { "rating": "low", "reason": "The main risk and imaging studies are academic or nonprofit-led rather than product-performance reports." }, "recency": { "rating": "current", "reason": "The risk estimate, PANDOME report and EDI registry status were current through September 2026." } } }, "layout": { "x": 12, "y": 105 }, "updated": "2026-09-14" }, { "id": "claim-nis793-target-engagement-without-benefit", "title": "NIS793 changed the intended TGF-beta tissue system and made patient outcomes worse", "shortTitle": "TGF-beta target hit, no benefit", "type": "claim", "status": "supported", "scope": [ "pdac", "pan-cancer", "transfer-unproven" ], "summary": "A 511-person blinded phase 3 rejected unselected NIS793: median survival was 9.2 versus 11.2 months and serious adverse events rose by 14.6 percentage points.", "content": "Plain meaning. TGF-beta is a signal that can restrain early abnormal cells but later help established cancers build scar-like tissue and suppress immune attack. NIS793 was designed to block that signal. The drug did what it was built to do: paired tissue and blood measurements showed lower TGF-beta activity, fewer activated-fibroblast signals and less collagen-related activity. The patient result went the other way. In daNIS-2, 245 people were assigned NIS793 plus gemcitabine/nab-paclitaxel and 245 placebo plus the same chemotherapy. Median overall survival was 9.2 versus 11.2 months, a loss of 2.0 months. Median progression-free survival was 4.6 versus 5.4 months, and response was 21.6% versus 25.3%. Serious adverse events affected 144/239=60.3% versus 110/241=45.6%, an absolute rise of 14.6 points. Fatal adverse events were 12/239=5.0% versus 8/241=3.3%. The earlier randomized phase 2 gave the same direction: NIS793 plus chemotherapy had OS HR 1.32, 95% CI 0.84-2.07, despite target engagement. Decision. Broad NIS793 treatment in unselected PDAC is rejected. A tissue change is an internal system measurement, not evidence that a person lives longer or better. The distinct high-angiogenin ALK5-inhibitor claim remains a testable marker question, not permission to revive broad TGF-beta blockade.", "sourceIds": [ "clinicaltrials-nct04935359-danis2", "pelletier-2026-nis793-biomarkers", "clinicaltrials-nct04390763-nis793-phase2", "clinicaltrials-nct05546411-nis793-neoadjuvant", "melisi-2018-galunisertib", "pietrobono-2026-angiogenin-alk5" ], "links": [ { "target": "treatment", "relation": "part-of" }, { "target": "failure-immune-reprogramming-without-patient-benefit", "relation": "supports" }, { "target": "hypothesis-angiogenin-tgfb-existing-data-test", "relation": "bounds" }, { "target": "trial-immune-reprogramming-action-map", "relation": "changes" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "moderate", "randomised": "strong", "independentReplication": "strong", "sampleSize": "Phase 3 n=511, randomized efficacy groups n=245 versus n=245; phase 2 n=164", "effect": "Phase 3 OS 9.2 versus 11.2 months; PFS 4.6 versus 5.4 months; serious adverse events +14.6 percentage points", "limits": [ "Registry does not post an OS hazard ratio", "Chemotherapy backbone was gemcitabine/nab-paclitaxel", "NIS793 does not test every possible TGF-beta intervention", "High-angiogenin ALK5 subgroup was not reported" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Direct tissue measurements confirm pathway blockade while the pathway's opposing functions explain possible harm." }, "animalEvidence": { "rating": "strong", "reason": "Models motivated treatment but failed to predict patient benefit." }, "inVitroEvidence": { "rating": "strong", "reason": "Drug binding and downstream tissue changes were measurable." }, "reproducibility": { "rating": "strong", "reason": "Randomized phase 2 and phase 3 both failed to show benefit." }, "commercialConflicts": { "rating": "moderate", "reason": "Novartis sponsored the trials, while the phase 3 registry exposes the negative result." }, "recency": { "rating": "current", "reason": "Phase 3 results and the tissue analysis were posted in 2026." } } }, "layout": { "x": 268, "y": 198 }, "updated": "2026-09-15" }, { "id": "claim-nk-cell-therapy-early", "title": "Allogeneic NK cells are feasible in PDAC, but a 19-patient response set cannot show added benefit", "shortTitle": "NK cells: early", "type": "claim", "status": "open", "scope": [ "pdac", "transfer-unproven" ], "summary": "A Chinese phase 1b/2 reported tolerable dosing and 31.6% response with gemcitabine/S-1, but there was no chemotherapy-only control.", "content": "A Beijing phase 1b/2 treated 24 people with allogeneic NK cells plus gemcitabine and S-1; 19 were efficacy-evaluable. ORR was 31.6%, disease control 73.7%, median PFS 6.6 months and OS 10.8 months, with no graft-versus-host disease or dose-limiting toxicity. Serial single-cell blood work found NK and T-cell states associated with response. That establishes feasibility and a measurement platform. It does not estimate the NK-cell contribution because the study was small, single-arm and used an active chemotherapy backbone. Blood expansion also does not prove entry into every tumour lesion. The next trial needs chemotherapy control, labelled or genetically traceable NK cells, tumour biopsies, product potency and an early futility rule. Until then this ranks below common-RAS drugs and randomized treatment paths.", "sourceIds": [ "tan-2025-nk-gem-s1" ], "links": [ { "target": "treatment", "relation": "part-of" }, { "target": "claim-immune-exclusion-is-multicellular", "relation": "must-overcome" }, { "target": "transfer-car-t-lineage-access", "relation": "connects-to" }, { "target": "experiments", "relation": "requires-randomized-test" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "not-applicable", "randomised": "none", "independentReplication": "none", "sampleSize": "24 treated; 19 efficacy-evaluable", "effect": "ORR 31.6%, median PFS 6.6 months, OS 10.8 months", "limits": [ "Single-arm", "Small efficacy set", "Active chemotherapy backbone", "One-country programme", "Tumour entry not established", "No product-contribution estimate" ] }, "layout": { "x": 117, "y": 76 }, "updated": "2026-09-14" }, { "id": "claim-nrg1-approved-response-not-durable-control", "title": "NRG1 targeting produces pancreatic responses, but durable control is not established", "shortTitle": "NRG1 response, not durable control", "type": "claim", "status": "supported", "scope": [ "pdac", "nrg1", "precision-oncology", "rare-driver", "treatment" ], "summary": "Zenocutuzumab is the first approved NRG1-fusion PDAC treatment; 40% responded in a 30-person approval set, while 80% had stopped treatment by the cutoff.", "content": "NRG1 fusion joins two genes so that neuregulin growth signalling is forced through HER3 and HER2. Zenocutuzumab is a two-target antibody that blocks this signalling pair. FDA accelerated approval used 30 previously treated advanced pancreatic patients: 12 responded, 40%, 95% CI 23% to 59%. Eight of the 12 responses lasted at least six months and two lasted at least twelve months. This is far above the 7.7% response cited for later-line liposomal irinotecan plus fluorouracil, an absolute difference of 32.3 percentage points, but it is not a randomized comparison. At cutoff, 24 of 30, 80%, had stopped zenocutuzumab and 20 had scan or clinical progression. A later paper reported 15 responses among 36 pancreatic patients, 42%, but still without a control. The correct conclusion is a real target-linked response signal and an indicated treatment route for a rare subgroup, not cure, proven survival gain or durable control for most.", "sourceIds": [ "fda-2024-bizengri-approval", "fda-2024-bizengri-multidisciplinary-review", "schram-2025-zenocutuzumab-nrg1" ], "links": [ { "target": "claim-pdac-molecular-subtypes", "relation": "specializes" }, { "target": "failure-nrg1-testing-and-confirmation-route", "relation": "limited-by" }, { "target": "unknown-nrg1-identification-and-survival-benefit", "relation": "leaves" }, { "target": "trial-pdac-nrg1-action-map", "relation": "mapped-by" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "weak", "randomised": "none", "independentReplication": "none for comparative benefit", "sampleSize": "FDA pancreatic efficacy set n=30; later pancreatic report n=36", "effect": "ORR 40%, 95% CI 23% to 59%; 67% of responders at least six months; 80% of all treated had stopped by cutoff", "limits": [ "Single arm", "Rare selected subgroup", "Response endpoint", "No comparative overall survival", "Sponsor-funded trial", "Approval set and publication set differ", "Accelerated approval still unverified" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The fusion, receptor pair and blocking antibody form a direct target chain." }, "animalEvidence": { "rating": "moderate", "reason": "Model evidence supports target action but is no longer the decision limit." }, "inVitroEvidence": { "rating": "strong", "reason": "Target blocking is measured, while clinical durability remains the question." }, "reproducibility": { "rating": "moderate", "reason": "Responses occur across partners and tumour types, but all decisive pancreatic data come from one study." }, "commercialConflicts": { "rating": "high", "reason": "The pivotal study was funded by the drug owner and underpins an accelerated approval." }, "recency": { "rating": "current", "reason": "Evidence includes the current September 2026 FDA status." } } }, "layout": { "x": 1210, "y": 1134 }, "updated": "2026-09-15" }, { "id": "claim-observed-ten-year-research-capital", "title": "Ten-year research capital can be measured in bounded ledgers, not one global total", "shortTitle": "Bounded capital ledger", "type": "claim", "status": "supported", "scope": [ "pancreatic-all", "pdac", "pnet", "hcc", "colorectal", "hepatobiliary" ], "summary": "NCI tagged $1.9407bn to pancreatic cancer in 2015–2024 and DoD appropriated another $66m in FY2020–FY2024; charity and company figures use different boundaries and remain separate.", "content": "Observed US public series: summing the ten NCI pancreatic-cancer research-area values for fiscal years 2015–2024 gives $1.9407bn: 125.3 + 152.6 + 178.3 + 182.1 + 187.0 + 178.5 + 218.1 + 226.8 + 246.0 + 246.0 million. The same method gives $2.3194bn for colorectal cancer and $973.3m for liver cancer. The Department of Defense Pancreatic Cancer Research Program adds a distinct appropriation ledger: $66m across FY2020–FY2024 and 102 awards, with a separate $20m FY2026 appropriation. For the common 2015–2024 window, the two observed federal ledgers total $2.0067bn = $1.9407bn + $66m. A separate UKRI title-visible query finds £24.106m in displayed commitments for explicitly named pancreatic-cancer projects starting from 2015. It is not added across currencies or accounting bases. These still do not cover all public funding: the NCI figures are tagged categories rather than a project export, other NIH institutes can fund relevant work, earlier DoD awards sat in a broader cancer programme, UKRI's title filter misses broad and shared grants, and appropriations or commitments are not identical to final research expenditure. NCI also states that categories overlap, so a RAS trial can appear under both pancreatic cancer and another research heading. Charity anchors use different accounting: PanCAN reported $249m cumulative research investment by June 2025, FY2025 Research and Scientific Affairs expense of $13.823m and a narrower direct Research line of $6.494m; Lustgarten reported more than $301m of cumulative grants since 1998 and $13.85m of new 2025 commitments; Pancreatic Cancer UK reported £17.5m cumulative investment, £2.552m in new 2024/25 grants and £1.3m of research support. These figures cannot be added: they mix countries, lifetimes, commitments, expenses and possible co-funding. Missing categories remain material: other non-US public bodies, university subsidy, in-kind clinical care, undisclosed company spend, venture rounds and the disease share of broad RAS, immune and detection platforms. Therefore $2.0067bn and £24.106m are bounded observations from different ledgers, not a global total or PDAC-only estimate.", "sourceIds": [ "nci-2022-budget-factbook", "nci-2023-budget-factbook", "nci-2026-research-funding", "cdmrp-2026-pcarp-funding", "ukri-gtr-pancreatic-title-query-2026", "pancan-2025-audited-financials", "pancan-2025-grants", "lustgarten-2025-grants", "pcuk-2025-annual-report" ], "links": [ { "target": "investment", "relation": "part-of" }, { "target": "funding", "relation": "depends-on" }, { "target": "programme", "relation": "constrains" }, { "target": "claim-ukri-title-visible-pancreatic-capital", "relation": "contains" }, { "target": "claim-statistical-significance-can-hide-trivial-benefit", "relation": "uses-same-discipline" } ], "evidence": { "humanProspective": "not-applicable", "humanRetrospective": "not-applicable", "randomised": "not-applicable", "independentReplication": "moderate", "sampleSize": "Ten NCI fiscal years, five DoD programme years and four charity disclosures", "effect": "$2.0067bn observed US federal total across the NCI category series and DoD PCARP, FY2015–FY2024; not a global or PDAC-only estimate", "limits": [ "NCI categories overlap", "2020–2024 NCI values are estimates in the current table", "DoD values are appropriations, not final project expenditure", "Pancreatic cancer includes non-PDAC histologies", "Charity figures mix commitments, expenses and lifetime totals", "No full global commercial ledger" ] }, "layout": { "x": 91, "y": 104 }, "updated": "2026-09-14" }, { "id": "claim-oncosil-local-control-signal-awaits-randomized-result", "title": "Modern phosphorus-32 implantation has a local-control signal, but the randomized answer is still hidden", "shortTitle": "OncoSil signal, result pending", "type": "claim", "status": "supported", "scope": [ "pdac", "locally-advanced", "brachytherapy", "local-treatment" ], "summary": "Silicon phosphorus-32 microparticles can keep beta radiation inside a pancreatic tumour and have passed multicentre delivery; an 88-person same-chemotherapy comparison has completed its main follow-up but reports no result.", "content": "OncoSil places radioactive phosphorus-32 microparticles into a pancreatic tumour through an endoscope-guided needle. The particles emit short-range beta radiation and are dosed to deliver about 100 gray, a unit of absorbed energy, to the tumour. A ten-centre pilot enrolled 50 people; 42 received implantation. Radiation remained localized on early imaging in 40/42. Sixteen of 42 had a device- or procedure-related event and three, 7.1%, had at least one grade 3 or worse event. Counting all 50 who started chemotherapy, 41, 82%, had no local progression at 16 weeks; 10, 20%, later underwent resection; median survival was 15.2 months. There was no control. A two-hospital analysis estimates 189 more days alive within 30 months after adjustment, but it first excluded people who died before a three-month landmark and still cannot remove unknown selection. The decisive TRIPP-FFX study has now randomized 88 people to the same FOLFIRINOX chemotherapy with or without OncoSil. Main follow-up completed in January 2026, but no result is posted. The route has crossed delivery, not benefit.", "sourceIds": [ "ross-2022-panco-p32", "lim-2026-p32-landmark", "clinicaltrials-nct05466799-tripp-ffx", "rosemurgy-2008-p32-randomized" ], "links": [ { "target": "failure-p32-liquefaction-survival-and-reporting-gap", "relation": "bounded-by" }, { "target": "unknown-oncosil-randomized-survival-and-delivery", "relation": "leaves-open" }, { "target": "hypothesis-tripp-ffx-independent-analysis-gate", "relation": "motivates" }, { "target": "trial-pdac-p32-action-map", "relation": "mapped-by" }, { "target": "company-oncosil-medical", "relation": "involves" }, { "target": "treatment-modality-ranking", "relation": "updates" }, { "target": "treatment", "relation": "part-of" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "moderate", "randomised": "moderate — modern trial complete at primary endpoint but unreported", "independentReplication": "weak", "sampleSize": "Modern pilot n=50; adjusted comparison n=86; randomized modern n=88; older randomized n=30", "effect": "Pilot local control 82% and resection 20%; causal survival and quality-of-life result pending", "limits": [ "Pilot single arm", "Eight of 50 were not implanted", "Company funded pilot", "Adjusted comparison excludes early deaths", "Randomized main endpoint is local control", "Overall survival is secondary", "Old and modern formulations differ" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Imaging shows localized radioactive particles and beta radiation deposits energy in nearby tissue." }, "animalEvidence": { "rating": "moderate", "reason": "Dosimetry is supported, but the human randomized record is now more useful." }, "inVitroEvidence": { "rating": "strong", "reason": "Radioactive decay and absorbed dose are measurable physical quantities." }, "reproducibility": { "rating": "moderate", "reason": "Multicentre implantation repeats, but comparative patient benefit has not been reported." }, "commercialConflicts": { "rating": "high", "reason": "The manufacturer funded and staffed the pilot and sponsors the randomized study." }, "recency": { "rating": "current", "reason": "The randomized registry was checked after its January 2026 primary completion." } } }, "layout": { "x": 904, "y": 828 }, "updated": "2026-09-15" }, { "id": "claim-onomichi-community-pathway-stage-shift", "title": "A Japanese community-to-hospital pathway shifted pancreatic diagnosis earlier, but did not prove fewer deaths", "shortTitle": "Onomichi stage shift", "type": "claim", "status": "supported", "scope": [ "pdac", "pancreatic-all", "transfer-unproven" ], "summary": "Risk review in primary care, tiered referral, EUS and selective pancreatic-juice testing raised stage 0/I diagnoses by 8.2 percentage points across reported eras.", "content": "The Onomichi Method is important because it tests delivery rather than one assay. From 2007 to 2023, 29,642 people received risk-based primary-care assessment and 925 pancreatic cancers were diagnosed. Stage 0/I cases rose from 13.9% in 2007-2015 to 22.1% in 2016-2017: an absolute gain of 8.2 percentage points and a relative increase of 59%. Observed five-year survival rose from 14.9% to 20.7%, a 5.8-point gain. Survival did not improve significantly within early or later stage groups, which fits diagnosis moving earlier more than treatment improving. The design cannot prove mortality benefit: it has no concurrent control and is exposed to lead time, referral changes, diagnostic intensity and era effects. The next test is a matched or stepped-wedge health-system comparison that counts all referrals, false-positive procedures, interval cancers, stage distribution, treatment received and cause-specific mortality.", "sourceIds": [ "hanada-2026-onomichi-method", "ja-onomichi-hanada-profile-2026" ], "links": [ { "target": "detection-integrated-workflows", "relation": "supports" }, { "target": "person-keiji-hanada", "relation": "led-by" }, { "target": "lab-onomichi-pancreas-biliary-centre", "relation": "delivered-by" }, { "target": "trial-enzeavour-japan-screening", "relation": "informs" }, { "target": "hypothesis-one-time-risk-triggered-workup", "relation": "supports" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "strong", "randomised": "none", "independentReplication": "none", "sampleSize": "29,642 assessed; 925 pancreatic cancers", "effect": "Stage 0/I 13.9% to 22.1%; observed five-year survival 14.9% to 20.7%", "limits": [ "Observational era comparison", "No concurrent control", "Lead-time and referral bias", "No cause-specific mortality effect", "Complex pathway prevents attribution to one component" ] }, "layout": { "x": 138, "y": 72 }, "updated": "2026-09-14" }, { "id": "claim-pancreatic-laser-heat-feasibility-not-benefit", "title": "Pancreatic laser heating can create a small dead zone, but patient benefit is unproved", "shortTitle": "Laser heat feasible, benefit unproved", "type": "claim", "status": "supported", "scope": [ "pdac", "locally-advanced", "laser-ablation", "local-treatment" ], "summary": "A nine-person endoscopic study proves that a thin fibre can deliver heat inside selected pancreatic tumours; later surgical work found important leaks and no controlled benefit result survives.", "content": "Laser ablation here means heat, unlike photodynamic therapy's light-triggered chemical reaction. A fibre delivers near-infrared light; tissue absorbs the energy, temperature rises and proteins coagulate. In nine people with pancreatic cancer that had not responded to chemoradiation, endoscopic ultrasound guided a 300 micrometre fibre into the tumour. All procedures completed without a reported adverse event. One application produced 0.4 to 6.4 cubic centimetres of dead tissue across the tested settings, while tumours were 21 to 45 mm across. This proves access and local heating, not full coverage or patient benefit. A separate immunostimulating laser protocol required open surgery. Of 15 assigned people, 13 were treated; three treated people developed a late pancreatic fistula, meaning digestive fluid leaked from the pancreatic duct system. The authors judged safety, feasibility and device handling unsatisfactory. A planned randomized study stopped with eight people and no posted result. Another endoscopic study stopped at five because the device was unavailable. No human comparison establishes better symptoms, quality of life, progression or survival. The route has crossed a local engineering threshold only.", "sourceIds": [ "dimatteo-2018-eus-laser-ablation", "paiella-2021-imilt-pancreas", "clinicaltrials-nct03187587-imilt-randomized", "clinicaltrials-nct03784417-eus-laser" ], "links": [ { "target": "failure-laser-fistula-accrual-and-device-continuity", "relation": "bounded-by" }, { "target": "unknown-laser-ablation-delivery-and-net-benefit", "relation": "leaves-open" }, { "target": "hypothesis-laser-existing-study-recovery-gate", "relation": "motivates" }, { "target": "trial-pdac-laser-ablation-action-map", "relation": "mapped-by" }, { "target": "company-clinical-laserthermia-systems", "relation": "involves" }, { "target": "treatment-modality-ranking", "relation": "updates" }, { "target": "treatment", "relation": "part-of" } ], "evidence": { "humanProspective": "weak", "humanRetrospective": "none", "randomised": "weak — eight-person terminated study without results", "independentReplication": "weak", "sampleSize": "Endoscopic feasibility n=9; surgical studies n=15 assigned and 13 treated; terminated randomized n=8; terminated endoscopic n=5", "effect": "Local dead volume 0.4-6.4 cubic centimetres in n=9; three fistulas among 13 surgical treatments; no comparative benefit", "limits": [ "Tiny cohorts", "No published randomized result", "Different access routes", "Different laser control methods", "No whole-tumour coverage measure", "No causal survival or quality-of-life estimate", "Device availability failure" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Absorbed laser energy predictably heats and coagulates tissue." }, "animalEvidence": { "rating": "moderate", "reason": "Animal pancreas studies refine diffuser geometry, but the human route is the decision target." }, "inVitroEvidence": { "rating": "strong", "reason": "Light absorption, heat diffusion and thermal injury are measurable." }, "reproducibility": { "rating": "weak", "reason": "Human programmes remain small, use different systems and repeatedly fail to complete." }, "commercialConflicts": { "rating": "high", "reason": "The central immunostimulating system's manufacturer sponsored or collaborated on the programme." }, "recency": { "rating": "current", "reason": "The registry and company pipeline were checked in September 2026." } } }, "layout": { "x": 970, "y": 894 }, "updated": "2026-09-15" }, { "id": "claim-pancreatic-stage-distribution", "title": "Most pancreatic cancers are not localised when diagnosed", "shortTitle": "Late diagnosis", "type": "claim", "status": "supported", "scope": [ "pancreatic-all" ], "summary": "In SEER 2016–2022, 15.4% were localised, 28% regional, 51% distant and 5% unstaged at diagnosis.", "content": "This is a pancreatic-cancer registry result, not a PDAC-only estimate. Stage is strongly associated with outcome: five-year relative survival was 43.6% for localised, 17.0% for regional and 3.4% for distant disease. Association is not proof that moving every diagnosis earlier would reproduce the localised survival rate, because tumour biology, length bias and treatment eligibility differ.", "sourceIds": [ "seer-pancreas-statfacts" ], "links": [ { "target": "detection", "relation": "supports" }, { "target": "bottlenecks", "relation": "supports" }, { "target": "pdac", "relation": "part-of" } ], "evidence": { "humanProspective": "none", "humanRetrospective": "strong", "randomised": "not-applicable", "independentReplication": "moderate", "sampleSize": "Population registry, SEER 21", "effect": "15.4% localised; 51% distant; 43.6% versus 3.4% five-year relative survival", "limits": [ "Mixed pancreatic cancer histologies", "Registry association", "US population" ] }, "layout": { "x": 18, "y": 67 }, "updated": "2026-09-14" }, { "id": "claim-pancreatitis-high-relative-low-absolute-risk", "title": "Chronic pancreatitis creates high relative PDAC risk but modest absolute risk for most patients", "shortTitle": "Pancreatitis risk", "type": "claim", "status": "supported", "scope": [ "pancreatic-all", "pdac" ], "summary": "Long-term pancreatitis risk is real, but a 2026 Danish cohort showed that much of the high early rate disappears after excluding the first two years, when occult cancer can mimic pancreatitis.", "content": "In 2,015 clinical-centre patients with chronic pancreatitis, pancreatic-cancer incidence was 26.3 times the country-, age- and sex-adjusted expectation. Excluding the first five years reduced the ratio to 14.4, which helps limit reverse causation. Absolute cumulative risk was 1.8% at ten years and 4.0% at twenty years. A 2026 Danish national cohort of 14,565 incident chronic-pancreatitis cases makes the early problem clearer: five-year pancreatic-cancer incidence was 3.4%, but only 0.5% after excluding the first two years. A new diagnosis can therefore be a warning that cancer is already present, not only a cause of future cancer. Korean data still found persistent excess risk in recurrent acute pancreatitis and chronic pancreatitis with acute episodes after longer follow-up. In the rare hereditary form, eight cancers occurred in 8,531 person-years and estimated cumulative risk approached 40% by age 70. The action should match the time state: exclude occult cancer promptly after a new or unexplained pancreatitis diagnosis; reserve long-term specialist surveillance for groups with sustained high absolute risk; do not infer that prophylactic surgery or a pancreatitis medicine prevents cancer. A second marker of active dangerous change is still needed to avoid repeated invasive work-up.", "sourceIds": [ "lowenfels-1993-chronic-pancreatitis", "lowenfels-1997-hereditary-pancreatitis", "kim-2023-pancreatitis-risk-korea", "davidsen-2026-chronic-pancreatitis-mortality" ], "links": [ { "target": "detection", "relation": "supports" }, { "target": "claim-high-risk-surveillance", "relation": "supports" }, { "target": "claim-screening-base-rate-maths", "relation": "supports" }, { "target": "pdac", "relation": "part-of" } ], "evidence": { "humanProspective": "none", "humanRetrospective": "moderate", "randomised": "not-applicable", "independentReplication": "strong", "sampleSize": "Historical chronic-pancreatitis cohort n=2,015; Danish incident cohort n=14,565; Korean pancreatitis cohorts n=451,496; hereditary cohort n=246 with eight cancers", "effect": "Historical 20-year risk 4.0%; Danish five-year incidence 3.4% before and 0.5% after a two-year lag; hereditary risk near 40% by age 70", "limits": [ "Historical and registry cohorts", "Clinical and code-based case definitions", "Large early reverse-causation effect", "Changing diagnosis and treatment", "Rare hereditary cohort and only eight cancers", "Histology not always explicitly PDAC" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Sustained pancreatic inflammation can promote cancer, while occult cancer can also produce pancreatitis." }, "animalEvidence": { "rating": "moderate", "reason": "Inflammation accelerates pancreatic carcinogenesis in models but cannot separate human cause from early diagnostic confusion." }, "inVitroEvidence": { "rating": "weak", "reason": "Cell systems do not estimate the absolute human risk or the large effect of excluding early follow-up." }, "reproducibility": { "rating": "strong", "reason": "Several countries show excess risk and newer data reproduce the importance of time since diagnosis." }, "commercialConflicts": { "rating": "low", "reason": "The decisive estimates come from academic cohorts and public registries rather than product studies." }, "recency": { "rating": "mixed", "reason": "Rare inherited-risk estimates are historical, while national cohort evidence was updated in 2023 and 2026." } } }, "layout": { "x": 6, "y": 36 }, "updated": "2026-09-14" }, { "id": "claim-panec-needs-separate-classification", "title": "High-grade pancreatic neuroendocrine carcinoma needs separate classification and treatment", "shortTitle": "Separate PanNEC", "type": "claim", "status": "supported", "scope": [ "pnet" ], "summary": "Morphology alone poorly separates high-grade pancreatic neuroendocrine disease, while RB loss and KRAS mutation mark a platinum-responsive PanNEC group in small retrospective data.", "content": "High proliferation is not enough to treat every pancreatic neuroendocrine neoplasm as one disease. In a 70-sample multi-institutional study, three expert cytopathologists agreed on well versus poorly differentiated status in only 38% overall. A final classification needed RB, DAXX and ATRX staining, mutation results, prior histology and consensus. In a separate 49-patient Japanese PanNEC-G3 cohort, RB loss and/or KRAS mutation was associated with a 70.0% response to first-line platinum compared with 33.3% without either feature. The confidence interval was wide and the apparent survival difference was not significant. This evidence is enough to prevent the atlas from merging well-differentiated G3 pNET with poorly differentiated PanNEC. It is not enough to make RB/KRAS status a validated treatment-selection rule. A prospective classifier must combine differentiation, Ki-67, RB/TP53/KRAS and receptor expression, then test therapy by that declared state.", "sourceIds": [ "sigel-2018-panenen-differentiation", "tanaka-2020-panec-g3" ], "links": [ { "target": "model-pnet-causal-chain", "relation": "supports" }, { "target": "pnet", "relation": "part-of" }, { "target": "claim-pnet-distinct-from-pdac", "relation": "extends" }, { "target": "unknowns", "relation": "exposes" } ], "evidence": { "humanProspective": "none", "humanRetrospective": "moderate", "randomised": "none", "independentReplication": "weak", "sampleSize": "70 evaluable diagnostic samples and 49 PanNEC-G3 treatment records", "effect": "Morphology agreement 38%; platinum response 70.0% versus 33.3% with RB loss and/or KRAS mutation, OR 9.22, 95% CI 1.26-67.3", "limits": [ "Small retrospective cohorts", "No prospective treatment interaction test", "Wide confidence interval", "Overall-survival difference was not significant", "Diagnosis used expert consensus and added molecular tests" ] }, "layout": { "x": 42, "y": 20 }, "updated": "2026-09-14" }, { "id": "claim-panova3-ttfields-small-os-benefit", "title": "PANOVA-3 proves a small, narrow survival gain from tumour-treating fields", "shortTitle": "TTFields survival signal", "type": "claim", "status": "supported", "scope": [ "pdac" ], "summary": "Adding an abdominal electric-field device to gemcitabine and nab-paclitaxel prolonged median survival by 2.0 months, but did not establish better progression-free survival and imposed a large daily burden.", "content": "Tumour-treating fields are low-intensity alternating electric fields delivered through four insulated skin arrays. The device is intended to disrupt cell division. PANOVA-3 randomized 571 people with newly diagnosed, unresectable, locally advanced pancreatic adenocarcinoma to gemcitabine plus nab-paclitaxel with or without the device. Median overall survival was 16.2 versus 14.2 months, hazard ratio 0.82, 95% confidence interval 0.68-0.99, P=0.039. The one-year survival estimates were 68.1% versus 60.2%, an absolute difference of 7.9 percentage points. If that difference repeats, about 1 / 0.079 = 13 people need device treatment for one additional person to be alive at one year. The trial did not establish better progression-free survival: 10.6 versus 9.3 months, hazard ratio 0.85, 95% CI 0.68-1.05, P=0.137. Local progression-free survival and response also did not establish a significant gain. Surgery was not increased: 20/285, 7.0%, versus 29/286, 10.1%. The physical burden matters. Median device use was 11.2 hours each day for 27.6 weeks, below the trial's requested 18 hours. Device-related skin toxicity affected 76.3%; 7.7% had grade 3 or worse skin toxicity and 8.4% stopped the device because of a device-related adverse event. Serious adverse events occurred in 53.6% versus 48.0%. FDA approved Optune Pax on 11 February 2026 only for adults with locally advanced pancreatic cancer together with this chemotherapy pair. This is strong evidence for that narrow use and against the blanket claim that physical treatment cannot extend life. It is not evidence of cure, independent reproduction, benefit with another drug backbone or benefit in metastatic disease. The survival result sits near its statistical boundary and conflicts with unchanged scan-based progression, so both replication and a better causal account remain important.", "sourceIds": [ "macarulla-2025-panova3", "clinicaltrials-nct03377491-panova3", "fda-2026-optune-pax-ssed" ], "links": [ { "target": "failure-local-intensification-pdac-survival-not-shown", "relation": "contradicts" }, { "target": "claim-pdac-death-can-be-local-or-metastatic", "relation": "supports" }, { "target": "regulatory-fda-optune-pax-2026", "relation": "supports" }, { "target": "unknown-ttfields-survival-replication-and-access", "relation": "leaves-open" }, { "target": "trial-pdac-ttfields-action-map", "relation": "mapped-by" }, { "target": "company-novocure", "relation": "commercialized-by" }, { "target": "treatment-modality-ranking", "relation": "updates" }, { "target": "treatment", "relation": "part-of" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong", "independentReplication": "none", "sampleSize": "571 randomized", "effect": "Median OS +2.0 months; HR 0.82, 95% CI 0.68-0.99; one-year survival +7.9 percentage points; P=0.039", "limits": [ "Open-label", "Sponsor-run device trial", "No independent randomized reproduction", "PFS was not significant", "Local PFS and response did not establish benefit", "No cure or durable-tail evidence", "Median use 11.2 hours per day", "High skin-event burden", "Serious adverse events were 5.6 percentage points higher" ] }, "layout": { "x": 44, "y": 117 }, "updated": "2026-09-15" }, { "id": "claim-particle-therapy-local-control-without-survival-proof", "title": "Particle radiation can control the pancreatic target but has not proved longer survival", "shortTitle": "Particle local control, no survival proof", "type": "claim", "status": "supported", "scope": [ "pdac", "locally-advanced", "proton", "carbon-ion" ], "summary": "Proton and carbon-ion series report tolerable treatment and strong local control, but none provides a randomized survival estimate against current photon treatment or chemotherapy.", "content": "Protons and carbon ions deposit much of their energy near the end of their path, which can reduce dose beyond a tumour. Carbon ions also deposit energy densely and may cause harder-to-repair DNA damage. This is a physical advantage in a treatment plan, not automatic patient benefit. In a 76-person carbon-ion and gemcitabine dose-escalation study, 72 were treated, two-year freedom from local progression was 83% and two-year overall survival was 35%; the selected higher-dose stage III group was 48%. A separate retrospective 72-person Japanese series reported two-year overall survival of 46%, but freedom from distant spread was only 28%. A 2026 Taiwanese series of 84 selected people after induction chemotherapy reported one-year local control of 94.7% while freedom from distant spread was 44.5% and progression-free survival was 42.9%. This 50.2-point local-versus-distant gap shows why beam precision cannot solve systemic disease. Proton evidence is smaller. A 15-person phase 2 trial produced 47% one-year survival, missing its 75% target against a 50% historical benchmark. A prospective registry included 19 people across seven centres, reported no grade 3 or worse treatment-related event and median survival of 14.6 months, but had no photon control. These records support delivery and local-control capability. They do not establish that particles extend overall or quality-adjusted survival over well-delivered modern photon radiation, or that either is better than systemic treatment alone.", "sourceIds": [ "shinoto-2016-carbon-ion-gemcitabine", "kawashiro-2018-jcros-carbon-ion", "lai-2026-carbon-ion-risk-adapted", "rapp-2022-proton-phase2", "eckstein-2023-proton-registry" ], "links": [ { "target": "failure-particle-therapy-comparator-and-accrual", "relation": "bounded-by" }, { "target": "unknown-particle-therapy-added-patient-benefit", "relation": "leaves-open" }, { "target": "hypothesis-particle-therapy-existing-data-gate", "relation": "motivates" }, { "target": "trial-pdac-particle-therapy-action-map", "relation": "mapped-by" }, { "target": "failure-local-intensification-pdac-survival-not-shown", "relation": "supports" }, { "target": "treatment-modality-ranking", "relation": "updates" }, { "target": "treatment", "relation": "part-of" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "moderate", "randomised": "none", "independentReplication": "moderate", "sampleSize": "Carbon: 72 treated in dose escalation, 72 retrospective multicentre and 84 retrospective Taiwan; proton: 15 phase 2 and 19 prospective registry", "effect": "Strong uncontrolled local control; no randomized overall-survival or quality-adjusted-survival effect", "limits": [ "No randomized particle-versus-photon result", "No randomized particle-versus-systemic-treatment result", "Single-arm and retrospective designs", "Different chemotherapy and radiation schedules", "Selected non-metastatic or oligometastatic populations", "Small proton cohorts", "Distant spread remains common", "Dose-model differences hinder comparison" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Particle depth-dose and ionization properties can improve target-to-normal-tissue exposure." }, "animalEvidence": { "rating": "moderate", "reason": "Radiation physics and model systems support local tumour damage, not human survival." }, "inVitroEvidence": { "rating": "strong", "reason": "Dense carbon-ion energy deposition can produce complex DNA damage in cells." }, "reproducibility": { "rating": "moderate", "reason": "Several centres report local control, but no randomized patient-benefit result exists." }, "commercialConflicts": { "rating": "moderate", "reason": "The reports come from particle centres whose clinical programmes depend on the technology, even when trials are academic." }, "recency": { "rating": "current", "reason": "The evidence includes a 2026 carbon-ion cohort and registries checked in September 2026." } } }, "layout": { "x": 646, "y": 570 }, "updated": "2026-09-15" }, { "id": "claim-pdac-death-can-be-local-or-metastatic", "title": "PDAC has both locally destructive and widely metastatic lethal courses", "shortTitle": "Two lethal courses", "type": "claim", "status": "supported", "scope": [ "pdac", "pancreatic-all" ], "summary": "A 76-person rapid-autopsy series found 30% locally destructive and 70% widely metastatic end-stage patterns; a Japanese resection-autopsy series judged metastasis the direct cause in most deaths.", "content": "In 76 rapid autopsies, 30% of patients died with locally destructive pancreatic cancer and 70% with widespread metastases. In a separate Japanese series after apparently curative resection, local recurrence was present in 18 of 24 autopsies, but it was judged the direct cause in only four of 20 cancer deaths; distant metastasis was present in 18 of 24. The studies are small and selected, but together reject both extremes: not every patient dies of distant tumour burden, and local recurrence on imaging or autopsy is not necessarily the lethal event. This supports systemic control as the largest population bottleneck while preserving a substantial local-control subgroup. A useful clinical system must predict the failure route before treatment, rather than use one local-versus-systemic rule for everyone.", "sourceIds": [ "iacobuzio-donahue-2009-autopsy-patterns", "hishinuma-2006-recurrence-autopsy" ], "links": [ { "target": "ranking-provisional-pdac-bottlenecks", "relation": "supports" }, { "target": "failure-local-control-without-survival-lap07", "relation": "extends" }, { "target": "residual-disease", "relation": "part-of" }, { "target": "pdac", "relation": "part-of" } ], "evidence": { "humanProspective": "none", "humanRetrospective": "moderate", "randomised": "not-applicable", "independentReplication": "moderate", "sampleSize": "76 rapid autopsies; 27 autopsies after resection", "effect": "30% locally destructive versus 70% widespread metastatic pattern; local recurrence present in 75% but direct cause in 4/20 cancer deaths in second series", "limits": [ "Autopsy-selection bias", "Small cohorts", "Different treatment eras", "Judging direct cause of death is subjective", "One series used pancreatic-cancer rather than PDAC-only wording" ] }, "layout": { "x": 84, "y": 57 }, "updated": "2026-09-14" }, { "id": "claim-pdac-delivery-is-variable-not-single-barrier", "title": "PDAC drug delivery is poor in some tumours, measurable in people and not explained by one stromal barrier", "shortTitle": "Variable drug delivery", "type": "claim", "status": "supported", "scope": [ "pdac" ], "summary": "Human CT and intraoperative gemcitabine measurements show large transport variation; mouse barrier removal improved delivery, yet phase III hyaluronan depletion did not extend survival.", "content": "The phrase 'PDAC has poor drug delivery' hides patient variation and an important failed test. A human study estimated transport from 176 pancreatic-protocol CT scans and infused gemcitabine during resection in 12 patients. Tumour transport and drug incorporation varied widely, and greater stroma tracked lower incorporation. In autochthonous mouse PDAC, hyaluronan-rich matrix generated pressure, collapsed vessels and limited perfusion; enzymatic depletion reopened vessels and improved gemcitabine response. The direct human test then failed: in 492 patients with hyaluronan-high metastatic PDAC, PEGPH20 raised objective response from 36% to 47% but left progression-free survival at 7.1 months and overall survival at 11.2 versus 11.5 months, death HR 1.00. Therefore delivery matters, but removing one bulk matrix component is not a sufficient cancer-control strategy. A useful engineering route must measure exposure in each tumour, preserve restraining stroma, prove that target concentrations reach malignant cells, and combine this with a drug capable of eradicating the exposed cells.", "sourceIds": [ "koay-2014-human-pdac-drug-transport", "provenzano-2012-stromal-delivery-barrier", "hingorani-2020-halo301" ], "links": [ { "target": "failure-pegph20-response-without-survival", "relation": "constrained-by" }, { "target": "failure-bulk-stroma-removal", "relation": "constrained-by" }, { "target": "claim-immune-exclusion-is-multicellular", "relation": "extends" }, { "target": "model-pdac-causal-chain", "relation": "supports" }, { "target": "treatment", "relation": "part-of" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "moderate", "randomised": "strong", "independentReplication": "strong", "sampleSize": "176 human CT scans, 12 intraoperative drug measurements, and 492 patients in phase 3", "effect": "Variable gemcitabine incorporation; PEGPH20 ORR 47% versus 36% but OS HR 1.00", "limits": [ "Direct drug measurement cohort was only 12", "Gemcitabine transport may not represent antibodies or nanoparticles", "Metastatic phase 3 may not represent small residual tumours", "CT transport estimates require prospective routing validation", "Matrix has both tumour-supporting and restraining roles" ] }, "layout": { "x": 72, "y": 37 }, "updated": "2026-09-14" }, { "id": "claim-pdac-dormancy-model-proven-human-uncertain", "title": "Dormant PDAC cells can cause relapse in models, but their human states and locations remain unmeasured", "shortTitle": "Dormancy: model proof", "type": "claim", "status": "contested", "scope": [ "pdac", "transfer-unproven" ], "summary": "Two PDAC model systems show residual cells surviving regression or reawakening after chemotherapy; human evidence is correlational and cannot distinguish dormancy from undetected slow growth.", "content": "Dormancy is a plausible route from apparently complete treatment to recurrence, but it is not yet a clinically mapped PDAC state. In a reversible mouse PDAC model, turning off c-Myc caused visible primary, metastatic and transplanted tumours to regress while residual neoplastic cells remained; restoring c-Myc caused rapid recurrence. A September 2026 study used a separate liver dormancy-tracking system and reported that chemotherapy-induced hepatocyte senescence drove TGF-beta, neutrophil extracellular traps and CCDC25 to restart disseminated-cell growth. Blocking parts of that chain suppressed metastatic reactivation in models without reducing chemotherapy activity. Human samples linked chemotherapy with hepatocyte senescence and linked circulating NET-DNA with poor outcome, but did not prove that a dormant human cell was awakened through this route. Clinical recurrence after a quiet interval is also compatible with continuous growth below imaging detection. The mission therefore cannot treat a dormancy marker or anti-NET drug as ready. It needs paired primary tumour, serial blood, normal-liver tissue where safely available, recurrence tissue and rapid-autopsy material, with an agreed human dormant-cell definition before intervention testing.", "sourceIds": [ "lin-2013-reversible-pdac-dormancy", "shi-2026-chemo-dormant-dtc", "conroy-2022-prodige24-five-year", "effenberger-2012-pdac-bone-marrow-dtc", "nordgard-2022-unresectable-dtc", "chalabi-hajkarim-2025-organotropism", "pei-2025-spatial-metastatic-pdac", "tanaka-2012-microscopic-local-recurrence", "mcclellan-2026-quiescent-pdac-car-t" ], "links": [ { "target": "residual-disease", "relation": "part-of" }, { "target": "unknown-dormant-cell-actionability", "relation": "leaves-open" }, { "target": "claim-liver-niche-before-metastasis", "relation": "extends" }, { "target": "claim-adjuvant-mfolfirinox-improves-but-recurrence-remains", "relation": "possible-cause-of" }, { "target": "model-pdac-causal-chain", "relation": "supports" }, { "target": "claim-human-pdac-dtc-prognostic-not-dormancy", "relation": "bounded-by" }, { "target": "failure-pdac-dormancy-sample-gap", "relation": "bounded-by" }, { "target": "trial-pdac-residual-cell-state-action-map", "relation": "bounded-by" } ], "evidence": { "humanProspective": "weak", "humanRetrospective": "weak", "randomised": "none", "independentReplication": "moderate", "sampleSize": "Two distinct PDAC model programmes plus human correlative samples; human cohort sizes were not exposed in the accessible 2026 record", "effect": "Residual cells caused recurrence after oncogene restoration in one model; blocking the proposed reawakening route suppressed liver metastasis in another", "limits": [ "Causal evidence is from engineered models", "No validated human dormant-cell marker", "No human anti-dormancy intervention", "Clinical recurrence does not itself prove dormancy", "The new 2026 result has no independent replication" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Distinct PDAC models causally show residual-cell survival, later growth and treatment-specific quiescent states." }, "animalEvidence": { "rating": "strong", "reason": "Independent mouse systems support persistence, liver reawakening and EREG-linked immune escape." }, "inVitroEvidence": { "rating": "moderate", "reason": "Cell and organoid systems support candidate states but remove organ, immune and treatment-history context." }, "reproducibility": { "rating": "weak", "reason": "Model routes differ, and no human residual state has been reproduced." }, "commercialConflicts": { "rating": "moderate", "reason": "Assay and later target access require declared ownership, open methods and independent analysis." }, "recency": { "rating": "current", "reason": "The evidence set includes 2025 human spatial studies and 2026 treatment-specific models." } } }, "layout": { "x": 84, "y": 67 }, "updated": "2026-09-14", "contentSections": [ "Human evidence narrows rather than closes the gap. Bone-marrow cytokeratin or RNA signals predict poor outcome in separate cohorts of 175 and 48 people, but neither measured viability, low cycling or ancestry to recurrence. Primary-tumour single-nucleus work linked organ-specific states to later recurrence in 21 selected cases, while rapid autopsy mapped 49 end-stage sections from 13 people. These samples bracket the disease-free interval; none observes a human residual cell during it." ] }, { "id": "claim-pdac-exercise-function-not-survival", "title": "Exercise can change activity and some quality measures but has not improved pancreatic survival or major complications", "shortTitle": "Exercise changes function, not survival", "type": "claim", "status": "supported", "scope": [ "pdac", "supportive-care", "surgery", "exercise", "quality-of-life" ], "summary": "Several randomized trials show that movement can be increased safely, while their main patient-outcome tests are negative or incomplete.", "content": "System model. Physical training can raise strength, walking capacity and reserve. That can matter if it prevents a complication, keeps chemotherapy on schedule, preserves independence or adds good days. A step count or muscle measurement is an internal state, not the patient return. Advanced disease. APACaP randomized 313 people to standard care with or without a 16-week home programme. Its primary week-16 analysis retained 172 and missed all three endpoints: global health difference -0.98, physical function -2.08 and fatigue 4.16, all P at least 0.17. Secondary longitudinal analyses among 259 people suggested improvement across several quality measures, so the signal is plausible but sensitive to who remained alive and answered. Neoadjuvant care. PancFit randomized 151; both groups improved six-minute walking distance, weekly aerobic activity was similar and quality of life and clinical outcomes did not differ. The intervention did increase strength sessions. Surgery. A 152-person phone-and-wearable trial increased steps 11% among 83 completers but severe complications were 22.9% versus 20.5%, P=0.807. A 135-person early-mobilization trial improved activity but missed its complication index. A 180-person mixed hepatobiliary and pancreatic trial found 31 versus 27 metres walk improvement against active prehabilitation, P=0.633. Boundary. Exercise is reasonable supportive care when safe and wanted. The evidence does not show cure, longer survival or fewer major pancreatic complications, and it does not justify delaying cancer treatment for a training target.", "sourceIds": [ "neuzillet-2023-apacap-rct", "ngohuang-2023-pancfit-rct", "ausania-2019-pancreatic-prehab-rct", "steindorf-2019-pancreatic-exercise-rct", "zarate-rodriguez-2023-pancreatectomy-steps-rct", "mizuno-2024-hpb-prehab-rct", "li-2024-pancreatic-early-mobilization-rct" ], "links": [ { "target": "treatment", "relation": "part-of" }, { "target": "claim-pdac-palliative-care-improves-qol-not-survival", "relation": "coordinates-with" }, { "target": "failure-pdac-exercise-endpoint-and-attrition", "relation": "bounded-by" }, { "target": "trial-pdac-exercise-action-map", "relation": "mapped-by" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "moderate", "randomised": "strong — several randomized studies exist", "independentReplication": "moderate — activity changes repeat but patient benefit does not", "sampleSize": "Randomized studies n=313, n=151, n=152, n=135, n=180, n=65 and n=40", "effect": "Activity and selected secondary quality measures improve; main quality, function or complication tests were negative or incomplete", "limits": [ "High attrition", "Different disease stages", "Mixed interventions", "Active controls", "Small trials", "No survival-powered study" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Training directly changes strength and aerobic capacity that can support treatment." }, "animalEvidence": { "rating": "not-applicable", "reason": "Human function and useful life are the decisive outputs." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "A laboratory model cannot measure participation or patient-valued benefit." }, "reproducibility": { "rating": "moderate", "reason": "Training changes some function repeatedly, but main clinical results are not positive." }, "commercialConflicts": { "rating": "low", "reason": "The interventions are low-cost behaviours without a required proprietary product." }, "recency": { "rating": "current", "reason": "The evidence includes randomized surgical studies through 2024 and a 2026 live protocol." } } }, "layout": { "x": 2320, "y": 1480 }, "updated": "2026-09-15" }, { "id": "claim-pdac-metabolic-adaptation-is-redundant", "title": "PDAC can draw nutrients through several routes in a poorly supplied tumour", "shortTitle": "Metabolic redundancy", "type": "claim", "status": "open", "scope": [ "pdac" ], "summary": "Human tissue and model studies support protein scavenging and stromal alanine supply, warning that one metabolic blockade may be bypassed.", "content": "Human PDAC tissue was depleted of glucose and several amino acids and showed active macropinocytosis, while cell and mouse work showed that extracellular protein can supply amino acids. Separate coculture and animal work showed pancreatic stellate cells releasing alanine through autophagy, which tumour cells used as carbon. These routes explain how a KRAS-driven tumour can grow in a poorly vascularised environment. They also expose a treatment problem: if the tumour can switch between free nutrients, engulfed protein and stromal exchange, blocking one route may select another. This is a biological rationale, not clinical proof of redundancy. A useful experiment needs isotope tracing in patients or fresh tissue, parallel pathway measurement, and a combination that is safe for normal pancreas, liver and immune cells.", "sourceIds": [ "kamphorst-2015-protein-scavenging", "sousa-2016-stellate-alanine" ], "links": [ { "target": "treatment", "relation": "supports" }, { "target": "unknowns", "relation": "supports" }, { "target": "pdac", "relation": "part-of" } ], "evidence": { "humanProspective": "none", "humanRetrospective": "weak", "randomised": "none", "independentReplication": "moderate", "sampleSize": "Human tissue metabolomics plus cell, coculture and animal models", "effect": "Active macropinocytosis in human specimens; alanine and protein supported model growth under nutrient limitation", "limits": [ "No patient treatment test", "Human flux contribution is not quantified", "Parallel routes may be context-specific", "Normal-tissue toxicity could narrow the treatment window" ] }, "layout": { "x": 77, "y": 44 }, "updated": "2026-09-14" }, { "id": "claim-pdac-microbiome-mechanism-without-human-benefit", "title": "Pancreatic microbes alter treatment in models, but no human microbiome treatment has proved patient benefit", "shortTitle": "Microbiome mechanism, no human benefit", "type": "claim", "status": "supported", "scope": [ "pdac", "transfer-unproven" ], "summary": "The mechanism is credible; the clinical treatment claim is not yet earned.", "content": "Microbiome means the bacteria, fungi and other microbes living in or around the body. Human pancreatic tumours are not sterile: bacteria were detected in 86 of 113 PDAC samples, and a bacterial enzyme converted gemcitabine into inactive dFdU in laboratory systems. Antibiotics reversed that resistance in mice. Separate mouse work linked bacteria to immune suppression, fungi to complement signalling, and stool from long-term survivors to slower tumour growth. The human parts remain associations. The long-survivor comparison began with 21 long-term and 22 short-term survivors; their microbes were sampled after survival had happened, so treatment, surgery, diet and survival itself can change the signal. The fungal drug experiments were in mice. The only published randomized human intervention found here assigned 40 people to 1-kestose, a fermentable sugar intended to change gut bacteria, or no supplement. Thirty-eight completed. Response was 14/20 versus 11/18 and disease control 19/20 versus 15/18; neither between-arm difference was significant. CA19-9, albumin and inflammation markers moved within one arm, but that is not the randomized question, which is whether the change differs between arms. Engineering translation: the system has plausible state variables and a controllable input, but no validated input-to-survival transfer function in people. Clinical decision: do not prescribe antibiotics, antifungals, probiotics or stool transfer to control PDAC outside a trial.", "sourceIds": [ "geller-2017-bacterial-gemcitabine", "pushalkar-2018-pdac-microbiome", "riquelme-2019-pdac-microbiome-survival", "aykut-2019-pdac-mycobiome", "nakaoka-2024-kestose-pdac", "jrct-kestose-pdac" ], "links": [ { "target": "treatment", "relation": "constrains" }, { "target": "treatment-long-tail-audit", "relation": "expands" }, { "target": "failure-microbiome-proxy-without-patient-benefit", "relation": "explains" }, { "target": "unknown-microbiome-treatment-contribution-pdac", "relation": "leaves-open" }, { "target": "trial-pdac-microbiome-action-map", "relation": "informs" } ], "evidence": { "humanProspective": "weak", "humanRetrospective": "moderate", "randomised": "weak", "independentReplication": "moderate", "sampleSize": "Mechanism n=113 human tumours; survivor discovery n=43 plus validation; randomized kestose n=40, 38 complete", "effect": "Credible human presence and model causality; no significant randomized tumour-response result", "limits": [ "Most causal intervention evidence is in mice", "Low-biomass tumour samples are contamination-sensitive", "Survivor sampling is after the outcome", "The human randomized trial was small, open and not powered for survival" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Bacterial gemcitabine metabolism and immune effects name measurable physical routes." }, "animalEvidence": { "rating": "strong", "reason": "Antibiotic, fungal and stool-transfer interventions repeatedly changed mouse tumours." }, "inVitroEvidence": { "rating": "strong", "reason": "Bacterial cytidine deaminase directly inactivated gemcitabine in controlled systems." }, "reproducibility": { "rating": "moderate", "reason": "Different groups support microbial involvement, but no human treatment effect has repeated." }, "commercialConflicts": { "rating": "moderate", "reason": "Supplement, live-bacterium and stool products create product and sponsor interests." }, "recency": { "rating": "current", "reason": "Foundational 2017-2019 mechanisms are checked against the 2024 human randomized pilot and 2026 trial map." } } }, "layout": { "x": 354, "y": 312 }, "updated": "2026-09-15" }, { "id": "claim-pdac-molecular-subtypes", "title": "PDAC contains reproducible tumour-cell expression states, but purity can create false subtype structure", "shortTitle": "PDAC subtypes", "type": "claim", "status": "supported", "scope": [ "pdac" ], "summary": "TCGA supported two tumour-specific expression groups, basal-like or squamous and classical or progenitor, after accounting for low tumour purity.", "content": "The finding supports biological classification beyond anatomy, but a subtype becomes clinically useful only if it is stable enough, measurable from routine material and predicts a treatment decision that improves outcomes. Bulk samples mix tumour, stromal and immune cells; subtype labels that track sample composition rather than tumour state can mislead both biology and trials.", "sourceIds": [ "tcga-2017-pdac" ], "links": [ { "target": "pdac", "relation": "part-of" }, { "target": "disease-system", "relation": "supports" }, { "target": "treatment", "relation": "depends-on" }, { "target": "unknowns", "relation": "supports" } ], "evidence": { "humanProspective": "none", "humanRetrospective": "moderate", "randomised": "none", "independentReplication": "moderate", "sampleSize": "150 clinically annotated pancreatic cancers in TCGA analysis", "effect": "Two tumour-specific expression groups after purity adjustment", "limits": [ "Retrospective tissue collection", "Bulk-sample composition", "Classification is not proof of treatment benefit" ] }, "layout": { "x": 30, "y": 36 }, "updated": "2026-09-14" }, { "id": "claim-pdac-nerve-ablation-pain-benefit-inconsistent", "title": "Pancreatic nerve ablation can reduce pain, but routine early use is not established", "shortTitle": "Nerve-ablation pain effect varies", "type": "claim", "status": "supported", "scope": [ "pdac", "pain", "palliative-care", "celiac-plexus", "neurolysis" ], "summary": "Randomized trials range from a three-month pain benefit to no added effect over modern opioids, while more aggressive nerve destruction has produced survival warnings.", "content": "The celiac plexus is a network of pain-carrying nerves behind the upper abdomen. Neurolysis deliberately damages it, usually with alcohol delivered through an endoscope or needle. In a 96-person randomized study, doing this early improved three-month pain change but did not improve quality of life or survival. In a later modern-opioid study, 24 intervention and 22 control participants had four-week pain of 1.3 versus 2.3 on a 0-to-10 scale, p=0.10, with no difference at any time, no opioid saving and no quality-of-life gain. More destruction was not better. Direct ganglion plus plexus neurolysis produced similar 12-week pain response to plexus treatment alone, 46.2% versus 40.4%, but median survival was 5.59 versus 10.46 months, HR 1.49. Alcohol destruction of the upstream splanchnic nerves improved pain by at most 0.7/10 and lowered opioid use, yet survival favoured saline, HR 1.56, with no quality-of-life gain. Those survival results are safety signals, not a proven biological effect, but they require active monitoring. Heat ablation looked better than alcohol in only 26 people over four weeks. Nerve treatment can be useful for selected refractory pain; the evidence does not support routine early destruction for everyone.", "sourceIds": [ "wyse-2011-early-eus-cpn-rct", "kanno-2020-eus-cpn-modern-opioid-rct", "levy-2019-celiac-ganglia-versus-plexus-rct", "dong-2021-splanchnic-neurolysis-rct", "bang-2019-celiac-rfa-versus-cpn-rct", "clinicaltrials-nct06160323-upfront-cgn-cpn" ], "links": [ { "target": "treatment", "relation": "part-of" }, { "target": "failure-pdac-pain-death-and-evaluable-denominators", "relation": "bounded-by" }, { "target": "unknown-pdac-nerve-ablation-net-benefit", "relation": "leaves-open" }, { "target": "hypothesis-pdac-nerve-ablation-existing-trial-safety-gate", "relation": "motivates" }, { "target": "trial-pdac-celiac-pain-action-map", "relation": "mapped-by" }, { "target": "unknown-cryoablation-randomized-pain-and-net-benefit", "relation": "compares-with" }, { "target": "treatment-modality-ranking", "relation": "updates" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "weak", "randomised": "strong", "independentReplication": "moderate — conflicting direction and technique", "sampleSize": "Key randomized studies n=96, n=46 analysed, n=110, n=96 and n=26", "effect": "Possible short pain relief; inconsistent opioid and quality-of-life benefit; two survival-harm signals with more aggressive targets", "limits": [ "Different nerve targets", "Different opioid eras", "Small studies", "Short pain follow-up", "Death competes with pain assessment", "Technique variation", "Survival often secondary", "No shared patient-level synthesis" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Destroying afferent sympathetic pathways can reduce visceral pain transmission." }, "animalEvidence": { "rating": "not-applicable", "reason": "Several randomized human pain tests exist." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Pain and survival cannot be represented in isolated cells." }, "reproducibility": { "rating": "moderate", "reason": "Pain relief appears in some trials but disappears against modern drug care and varies by target." }, "commercialConflicts": { "rating": "low", "reason": "Most interventions use common alcohol, needles or standard endoscopy; procedure preference remains relevant." }, "recency": { "rating": "current", "reason": "A 2026 randomized trial is actively testing early use." } } }, "layout": { "x": 1138, "y": 1062 }, "updated": "2026-09-15" }, { "id": "claim-pdac-organoid-correlation-not-routing-utility", "title": "Pancreatic organoids can correlate with drug response, but no complete treatment-routing benefit is established", "shortTitle": "Organoid correlation is not utility", "type": "claim", "status": "supported", "scope": [ "pdac", "transfer-unproven" ], "summary": "A miniature tumour grown from a biopsy can reject or retain drug options in the laboratory; the unresolved claim is whether using that answer makes the patient live longer or better.", "content": "A patient-derived organoid is a three-dimensional culture grown from a person's tumour cells. Think of it as an executable sample of one cancer clone, not a full digital twin: it can expose cells to drugs, but it omits circulation, drug metabolism, much of the immune system, nerves, stroma and competition among metastases. Conditional response correlation is real. The 2021 German feasibility study produced profiles for 28 of 44 confirmed pancreatic cancers, 63.6%, and reported strong first-line prediction among the subset, but median turnaround was 53 days and accuracy fell to 40% after prior treatment. HOPE was closer to an unselected service denominator: only 12 of 76 enrolled people produced a drug-testable model, 15.8%, after a median 96 days; correlation used 12 models. ORGANOPREDICT generated 54 models from 87 pretreated patients, 62%, in a mean 6.8 weeks; only 34 had complete evaluable follow-up. Its reported 83.3% sensitivity and 92.9% specificity therefore describe the conditional subset, not every person entering the route, and treatment was not assigned by randomization. PASS-01 supplies the current timing warning. Organoids were generated in 50%, only 75 of 140 treated people reached second-line treatment, and the median time on second line was 2.1 months. The assay is biologically credible and useful for drug development now. Clinical routing remains unproved until a randomized policy beats standard choice from the original enrolled denominator.", "sourceIds": [ "beutel-2021-pdac-organoids", "grossman-2022-hope-organoids", "boileve-2024-organopredict", "knox-2025-pass01", "tiriac-2018-organoid-profiling" ], "links": [ { "target": "treatment", "relation": "limits" }, { "target": "technology-ai-organoid-active-learning", "relation": "constrains" }, { "target": "hypothesis-organoid-guided-early-switch", "relation": "constrains" }, { "target": "unknown-pdac-organoid-guided-treatment-utility", "relation": "creates" }, { "target": "trial-pdac-organoid-action-map", "relation": "summarized-by" } ], "evidence": { "humanProspective": "moderate — several prospective feasibility and correlation cohorts", "humanRetrospective": "moderate", "randomised": "weak — treatment was randomized in PASS-01 but organoid guidance was not", "independentReplication": "moderate — response correlation and route attrition repeat under different methods", "sampleSize": "Beutel n=54 attempted and 28 profiles; HOPE n=76 and 12 drug-testable patients; ORGANOPREDICT n=87, 54 models and 34 evaluable; PASS-01 n=160 randomized and 50% organoid generation", "effect": "Conditional assay sensitivity and specificity are promising; randomized clinical utility remains absent", "limits": [ "Different culture methods and denominators", "Small conditional comparison sets", "Treatment was not assigned by the organoid result", "Missing host and metastatic context", "Failed cultures can select easier tumours", "Drug access and patient decline sit downstream of the assay" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Living tumour cells can retain drug-response differences that static sequence alone does not measure." }, "animalEvidence": { "rating": "moderate", "reason": "Matched model work supports response fidelity but cannot estimate patient-level routing benefit." }, "inVitroEvidence": { "rating": "strong", "reason": "Multiple pancreatic organoid programmes reproduce heterogeneous drug response." }, "reproducibility": { "rating": "moderate", "reason": "Correlation repeats, while yield, timing and methods vary enough to block a fixed clinical rule." }, "commercialConflicts": { "rating": "moderate", "reason": "Laboratories and screening platforms benefit from adoption; independent protocol and data access are required." }, "recency": { "rating": "current", "reason": "Includes the 2025 PASS-01 result and trial states checked in September 2026." } } }, "layout": { "x": 101, "y": 155 }, "updated": "2026-09-15" }, { "id": "claim-pdac-palliative-care-improves-qol-not-survival", "title": "Systematic palliative care can improve pancreatic quality of life without proving longer survival", "shortTitle": "Palliative quality, not survival", "type": "claim", "status": "supported", "scope": [ "pdac", "palliative-care", "quality-of-life", "supportive-care", "treatment-route" ], "summary": "One pancreatic randomized trial improved quality-of-life measures and reduced late chemotherapy, while survival was unchanged and another symptom trial missed both main endpoints.", "content": "Palliative care is active management of pain, nausea, appetite, fatigue, mood, family load and treatment decisions alongside cancer treatment; it is not a signal that cancer treatment has stopped. In 207 people with advanced pancreatic cancer, systematic visits improved the 12-week FACT-Hep Trial Outcome Index change by 3.83 points, 95% CI 0.10 to 7.57, and reduced chemotherapy in the last 30 days of life from 27.8% to 18.7%, an absolute 9.1-point reduction. It did not change overall survival. A separate 288-person pancreatic-and-biliary trial missed both main endpoints for 50% pain and depression reduction, although some secondary pain measures improved. A 2026 home programme reported 8.0 versus 4.9 months median survival and fewer hospital days against an earlier control cohort; the time-period comparison cannot show that palliative care caused survival. The stable conclusion is better quality and decision support in some systems, not tumour control or a survival treatment.", "sourceIds": [ "maltoni-2016-systematic-palliative-qol", "maltoni-2016-systematic-palliative-end-of-life", "woo-2019-pancreatobiliary-palliative-rct", "dufva-2026-home-palliative-pdac", "sanders-2024-asco-palliative-guideline" ], "links": [ { "target": "claim-cachexia-is-parallel-lethal-system", "relation": "complements" }, { "target": "claim-pdac-nerve-ablation-pain-benefit-inconsistent", "relation": "offers-system-alternative-to" }, { "target": "failure-pdac-palliative-timing-and-analysis", "relation": "limited-by" }, { "target": "unknown-pdac-palliative-route-net-benefit", "relation": "leaves" }, { "target": "trial-pdac-palliative-care-action-map", "relation": "mapped-by" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "moderate", "randomised": "moderate", "independentReplication": "moderate — mixed results", "sampleSize": "Pancreatic randomized n=207; pancreatobiliary randomized n=288; historical-control pancreatic n=125", "effect": "QoL change difference 3.83 points; late chemotherapy 18.7% versus 27.8%; no randomized survival gain", "limits": [ "Open-label care", "Different intervention bundles", "Missing 12-week reports after death or decline", "Some secondary endpoints", "Historical survival control", "No common route denominator" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Direct symptom treatment and decision support can change lived burden without changing tumour biology." }, "animalEvidence": { "rating": "not-applicable", "reason": "This is human care delivery." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "No laboratory model can measure patient quality or end-of-life decisions." }, "reproducibility": { "rating": "moderate", "reason": "Several studies support parts of the route, while primary endpoints and packages differ." }, "commercialConflicts": { "rating": "low", "reason": "The intervention is a multidisciplinary care service rather than a proprietary product." }, "recency": { "rating": "current", "reason": "The synthesis includes current trials and a 2026 home-care cohort." } } }, "layout": { "x": 1246, "y": 1170 }, "updated": "2026-09-15" }, { "id": "claim-pdac-primary-prevention-action-gap", "title": "Smoking cessation and metabolic care are justified, but no pancreatic-specific intervention has proved prevention", "shortTitle": "Primary-prevention action gap", "type": "claim", "status": "open", "scope": [ "pancreatic-all" ], "summary": "Cessation, bariatric-surgery and GLP-1 studies point in a favourable direction, but every pancreatic-cancer estimate is observational and none supports prescribing an intervention solely to prevent this cancer.", "content": "The action question is stricter than the risk-factor question. In 182,011 Multiethnic Cohort participants with 1,831 pancreatic cancers, 50 pack-years was associated with 91% excess risk and each year since quitting with 9% lower excess risk. That strengthens the case for cessation, but it models observed quitting rather than randomizing a cessation programme. A five-site cohort compared 22,198 bariatric-surgery patients with 66,427 matched nonsurgical patients and reported pancreatic-cancer HR 0.46, 95% CI 0.22–0.97, after mean follow-up of 3.5 years. The interval is wide, the endpoint rare and treatment selection remains. A target-trial imitation in 1,636,056 people with type 2 diabetes, including 167,091 prescribed GLP-1 receptor agonists, reported pancreatic-cancer HRs from 0.42 to 0.82 against six other medicine classes over five years. Matching cannot remove unrecorded treatment choice, adherence, surveillance or reverse-causation differences. None of these studies justifies bariatric surgery or a GLP-1 drug solely to prevent pancreatic cancer. Smoking cessation, healthy weight and diabetes care remain warranted for their established total-health benefits. Scale explains the proof problem: at 13.9 cancers per 100,000 person-years, even a true 50% relative reduction in an average-risk population prevents about 6.95 cancers per 100,000 person-years, or one cancer per 14,388 person-years of exposure. A modest bias could erase that apparent effect. The next useful work is not a stand-alone pancreatic prevention trial in average-risk adults. Reproduce medicine and surgery estimates in independent data with active comparators, exposure lags, competing-risk analysis and negative-control outcomes; then obtain long-term cancer follow-up from randomized metabolic or cessation trials. Stop a pancreatic-specific programme if the association disappears after lagging exposure, appears equally for negative-control outcomes, fails across data systems or produces an absolute effect too small to justify added treatment harm and cost.", "sourceIds": [ "bogumil-2024-smoking-cessation-multiethnic", "schauer-2019-bariatric-pancreatic-risk", "wang-2025-glp1-pancreatic-risk", "seer-pancreas-statfacts" ], "links": [ { "target": "prevention-ranking", "relation": "supports" }, { "target": "claim-modifiable-risk-reduces-but-cannot-eliminate-pdac", "relation": "extends" }, { "target": "ranking-provisional-pdac-bottlenecks", "relation": "supports" }, { "target": "experiments", "relation": "constrains" } ], "evidence": { "humanProspective": "moderate — smoking cessation association", "humanRetrospective": "moderate — surgery and medicine comparisons", "randomised": "none for pancreatic-cancer prevention", "independentReplication": "moderate for risk factors; weak for named interventions", "sampleSize": "Smoking cohort n=182,011 with 1,831 cancers; surgery cohort n=88,625; GLP-1 cohort n=1,636,056", "effect": "Observed pancreatic-cancer estimates: 9% lower excess smoking risk per year quit; surgery HR 0.46; GLP-1 comparator HRs 0.42-0.82", "limits": [ "No randomized pancreatic-cancer prevention result", "Pancreatic cancer was a rare secondary outcome", "Treatment selection and unmeasured confounding", "Reverse causation and exposure-lag uncertainty", "Histology not restricted to PDAC", "Relative effects do not establish worthwhile absolute benefit" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "Tobacco carcinogens, obesity and abnormal metabolism have plausible links to pancreatic carcinogenesis, but several pathways and reverse causation remain." }, "animalEvidence": { "rating": "weak", "reason": "Model evidence can support mechanisms but cannot establish the observed human prevention effects or their size." }, "inVitroEvidence": { "rating": "weak", "reason": "Cell systems cannot reproduce years of exposure, competing risks or treatment selection in these cohorts." }, "reproducibility": { "rating": "weak", "reason": "Risk-factor direction repeats, while the named surgery and drug effects lack independent causal confirmation for pancreatic cancer." }, "commercialConflicts": { "rating": "moderate", "reason": "The cited reports do not establish a commercial indication, but surgery and diabetes-drug findings can create strong market incentives and need independent repeat." }, "recency": { "rating": "mixed", "reason": "The evidence spans a 2019 surgery cohort, a 2024 smoking analysis and a 2025 medicine study." } } }, "layout": { "x": 114, "y": 92 }, "updated": "2026-09-14" }, { "id": "claim-pdac-progression-timing-uncertain", "title": "The usable early-detection window in PDAC is not one settled number", "shortTitle": "Timing disputed", "type": "claim", "status": "contested", "scope": [ "pdac" ], "summary": "Human evolutionary models suggest a long average interval, while mouse lineage tracing and punctuated-genome evidence show routes to earlier dissemination or rapid change.", "content": "Yachida and colleagues modelled a mean 11.8 years from initiating alteration to the parental non-metastatic founder and another 6.8 years to the metastatic clone. That supports a possible detection window, but the estimates came from selected rapid-autopsy tumours and model assumptions. Rhim and colleagues observed dissemination before overt tumour formation in a mouse model; that finding does not prove that early disseminated cells form lethal human metastases. Notta and colleagues found evidence for punctuated genomic change in many human PDAC genomes, challenging a uniformly gradual sequence. The operational conclusion is narrower: precursor time probably exists for many patients, but the duration and point of irreversible systemic risk must be measured in longitudinal human cohorts.", "sourceIds": [ "yachida-2010-metastatic-timing", "rhim-2012-early-dissemination", "notta-2016-punctuated-evolution", "effenberger-2012-pdac-bone-marrow-dtc", "chalabi-hajkarim-2025-organotropism" ], "links": [ { "target": "disease-system", "relation": "supports" }, { "target": "detection", "relation": "supports" }, { "target": "residual-disease", "relation": "supports" }, { "target": "unknowns", "relation": "supports" }, { "target": "claim-human-pdac-dtc-prognostic-not-dormancy", "relation": "constrained-by" }, { "target": "trial-pdac-residual-cell-state-action-map", "relation": "constrained-by" } ], "evidence": { "humanProspective": "none", "humanRetrospective": "moderate", "randomised": "not-applicable", "independentReplication": "weak", "sampleSize": "Small rapid-autopsy timing cohort; more than 100 genomes in punctuated-evolution analysis; mouse lineage-tracing cohort", "effect": "Competing estimates and mechanisms; no validated patient-level window", "limits": [ "Autopsy selection", "Evolutionary-clock assumptions", "Mouse-to-human transfer", "Genome change is not identical to metastatic competence" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Human recurrence and evolution establish heterogeneous routes rather than one universal clock." }, "animalEvidence": { "rating": "moderate", "reason": "Mouse lineage tracing supplies an early-spread route but not human timing." }, "inVitroEvidence": { "rating": "weak", "reason": "Cell systems cannot estimate years to invasion or metastatic competence." }, "reproducibility": { "rating": "weak", "reason": "Different human and mouse methods estimate different biological events." }, "commercialConflicts": { "rating": "low", "reason": "Assay and later target access require declared ownership, open methods and independent analysis." }, "recency": { "rating": "mixed", "reason": "The evidence set includes 2025 human spatial studies and 2026 treatment-specific models." } } }, "layout": { "x": 6, "y": 44 }, "updated": "2026-09-14", "contentSections": [ "Two newer human observations do not turn this into one timetable. Bone-marrow tumour-cell signals can be present at diagnosis and predict poor outcome, but do not date dissemination or prove later ancestry. Primary-tumour states associated with liver or lung recurrence were found in a 744-person clinical cohort, with single-nucleus comparison in only 21 cases; association with destination is not a clock." ] }, { "id": "claim-pdac-resistance-uses-reversible-cell-states", "title": "PDAC treatment resistance can arise through reversible cell-state change, not only new mutation", "shortTitle": "State-based resistance", "type": "claim", "status": "supported", "scope": [ "pdac", "transfer-unproven" ], "summary": "Human tumours contain mixed classical and basal-like states, while lineage-traced models show KRAS and immune treatment selecting different relapse states.", "content": "PDAC classification cannot be a one-time mutation list. Purified human primary and metastatic tumours contain mixtures along classical and basal-like expression states, partly associated with copy-number change. In mouse models and patient-derived xenografts, KRAS inhibition rapidly enriched classical cells; lineage tracing showed those cells seeding relapse after drug withdrawal, and state-directed ablation deepened response in the model. In another mouse PDAC system, relapse after immune treatment involved ZEB1 and SNAIL-driven transition, epigenetic silencing of Irf6 and reduced sensitivity to T-cell killing. These results do not establish one universal resistant state: the state selected depended on the treatment. They establish a stronger design rule. A useful classifier must be repeated during treatment, measure state mixture rather than force one label, and be tied to a combination that removes the observed escape route at human drug exposure. The clinical gap is paired tissue before response and at progression in current broad-RAS trials; without it, the model evidence cannot choose a patient combination.", "sourceIds": [ "chan-seng-yue-2020-evolution-subtypes", "singhal-2024-classical-kras-resistance", "kim-2024-irf6-immune-plasticity" ], "links": [ { "target": "claim-pdac-subtypes-are-mixed-evolving-states", "relation": "extends" }, { "target": "unknown-durable-ras-control", "relation": "possible-cause-of" }, { "target": "hypothesis-prospective-ras-resistance-routing", "relation": "supports" }, { "target": "model-pdac-causal-chain", "relation": "supports" }, { "target": "treatment", "relation": "part-of" } ], "evidence": { "humanProspective": "none", "humanRetrospective": "strong", "randomised": "none", "independentReplication": "moderate", "sampleSize": "Human genomic and transcriptomic tumour cohort plus independent mouse and patient-derived model programmes", "effect": "Treatment-specific state enrichment and relapse causality in models; no human combination effect", "limits": [ "Treatment causality is model-derived", "Different treatments selected different states", "Patient-derived xenografts lack an intact human immune system", "No validated serial clinical classifier", "No state-directed human trial" ] }, "layout": { "x": 48, "y": 43 }, "updated": "2026-09-14" }, { "id": "claim-pdac-stemness-is-dynamic", "title": "The fixed pancreatic cancer stem-cell hierarchy is weaker than a dynamic, environment-driven model", "shortTitle": "Dynamic stemness", "type": "claim", "status": "contested", "scope": [ "pdac", "transfer-unproven" ], "summary": "Marker-free tracing and serial human xenografts show broad or changing tumour-initiating capacity, so eliminating one marker-positive minority may not prevent relapse.", "content": "The clinically important question is not whether a marker identifies cells that can form a tumour after transplantation. It is whether a stable minority exclusively maintains an established human PDAC and can be eliminated without other cells replacing it. Marker-free clonal labelling in established PDAC tissue models found that all tumour cells retained clonogenic potential in place and that proximity to activated fibroblasts shaped which cells expanded. A separate serial human PDAC xenograft study found successive bursts from different tumour-initiating clones with little overlap between passages. KRAS-inhibitor lineage tracing further showed a treatment-enriched classical state acting as a relapse reservoir. Together these data fit dynamic stem-like capacity controlled by environment and treatment better than a permanently fixed rare hierarchy. They do not prove every patient cell is equally dangerous, and xenograft passage can create selection. The programme consequence is to target the transition or niche that creates persistence, and to test marker-defined cell depletion for replacement by previously negative cells, rather than assuming one surface marker defines the curative target.", "sourceIds": [ "lodestijn-2021-environment-clonogenicity", "ball-2017-transient-tic-clones", "singhal-2024-classical-kras-resistance" ], "links": [ { "target": "claim-pdac-resistance-uses-reversible-cell-states", "relation": "supports" }, { "target": "claim-immune-exclusion-is-multicellular", "relation": "depends-on" }, { "target": "claim-pdac-dormancy-model-proven-human-uncertain", "relation": "could-overlap" }, { "target": "model-pdac-causal-chain", "relation": "supports" } ], "evidence": { "humanProspective": "none", "humanRetrospective": "none", "randomised": "none", "independentReplication": "moderate", "sampleSize": "Independent lineage and clonal-tracking programmes in tissue, mouse and human-xenograft systems", "effect": "Broad or changing tumour-initiating capacity in the tested systems", "limits": [ "No direct lineage tracing in a person", "Xenograft passage changes selection", "Clonogenic capacity is not identical to metastatic lethality", "Marker-independent tracing still depends on model assumptions", "Does not exclude clinically important rare states" ] }, "layout": { "x": 60, "y": 48 }, "updated": "2026-09-14" }, { "id": "claim-pdac-subtypes-are-mixed-evolving-states", "title": "PDAC transcriptional subtypes can coexist and change during tumour evolution", "shortTitle": "Subtype plasticity", "type": "claim", "status": "supported", "scope": [ "pdac" ], "summary": "Purified tumour genomes and transcriptomes showed a continuum of classical and basal-like states driven partly by copy-number change and mixed cell populations.", "content": "Whole genomes and transcriptomes from purified primary and metastatic human PDAC epithelium showed that expression subtypes are not always fixed boxes. Tumours contained mixtures of cell populations along a transcriptional continuum, confirmed with single-cell analysis. Copy-number changes affecting mutant KRAS, GATA6 and other genes tracked the states, and reconstructed histories linked some changes to genome doubling and ongoing instability. This sharpens the classification problem: a label from one biopsy can miss a mixed or changing tumour. A clinically useful classifier must report mixture, sampling site and time, and it must predict a treatment decision that improves outcome. Taxonomy alone is not precision treatment.", "sourceIds": [ "chan-seng-yue-2020-evolution-subtypes", "tcga-2017-pdac" ], "links": [ { "target": "claim-pdac-molecular-subtypes", "relation": "extends" }, { "target": "treatment", "relation": "depends-on" }, { "target": "residual-disease", "relation": "supports" }, { "target": "pdac", "relation": "part-of" } ], "evidence": { "humanProspective": "none", "humanRetrospective": "strong", "randomised": "not-applicable", "independentReplication": "moderate", "sampleSize": "Whole-genome and transcriptome cohort of purified primary and metastatic tumours with single-cell confirmation", "effect": "Mixed classical and basal-like populations linked to copy-number and evolutionary events", "limits": [ "Observational tumour sampling", "Purification and sampling can alter measured proportions", "Does not prove a subtype-directed treatment benefit", "One biopsy may not capture spatial mixture" ] }, "layout": { "x": 40, "y": 30 }, "updated": "2026-09-14" }, { "id": "claim-pdac-surgery-is-a-complete-treatment-route", "title": "PDAC surgery succeeds only when the whole curative route is completed", "shortTitle": "Surgery is a whole route", "type": "claim", "status": "supported", "scope": [ "pdac", "pancreatic-all" ], "summary": "The operation is one dangerous step between expert selection and systemic treatment; current national data do not show that every planned step was delivered.", "content": "England already concentrates pancreatic surgery in specialist networks, yet the whole route is not reliably visible or completed. The 2026 national audit reports 23 English HPB specialist centres. Among people having Whipple surgery, 96.4% were alive at 90 days, so 3.6% died within 90 days. Only 62% received chemotherapy or chemoradiotherapy within 14 weeks, with centre IQR 54-68%; that means 38% had no recorded timely postoperative treatment. The audit counts receipt, not completion of every planned cycle, and it does not tell whether missing treatment followed progression, operative harm, frailty, patient choice, delayed referral or a broken handoff. Older PDAC cohorts show the same chain. In a 203-person surgery-first cohort, only 85, 41.9%, completed chemotherapy; in a Dutch 1,071-person resection cohort, major complications tracked with shorter disease-free interval and survival, with much of the association passing through adjuvant treatment. These are not randomized causal estimates. The valid unit of success is the complete patient route: expert review, justified resection, safe operation, rescue from complications, recovery, and delivery of the planned systemic treatment. Resection count, margin status, hospital volume or length of stay alone cannot stand in for that endpoint.", "sourceIds": [ "npaca-2026-state-of-nation", "nhs-england-2024-hpb-pancreatic-service", "nice-2018-ng85-pancreatic", "labori-2016-adjuvant-completion", "henry-2023-complications-adjuvant", "kollbeck-2025-complications-chemo", "coupland-2016-england-resection-volume" ], "links": [ { "target": "treatment", "relation": "part-of" }, { "target": "claim-neoadjuvant-therapy-improves-selection-not-cure", "relation": "complements" }, { "target": "claim-adjuvant-mfolfirinox-improves-but-recurrence-remains", "relation": "depends-on" }, { "target": "system-england-pancreatic-surgery-network", "relation": "delivered-through" }, { "target": "failure-pdac-postoperative-treatment-loss", "relation": "exposes" }, { "target": "unknown-pdac-surgical-route-completion", "relation": "leads-to" }, { "target": "trial-pdac-surgical-route-action-map", "relation": "informs" }, { "target": "hypothesis-pdac-surgical-route-completion", "relation": "supports" } ], "evidence": { "humanProspective": "moderate for national audit capture; none for intervention", "humanRetrospective": "strong", "randomised": "strong for systemic treatment efficacy; none for whole-route delivery", "independentReplication": "moderate", "sampleSize": "Current English audit n=17,672 diagnoses; 23 specialist centres; historical route cohorts n=203, n=920 and n=1,071", "effect": "96.4% 90-day survival after Whipple; 62% received postoperative chemotherapy or chemoradiotherapy within 14 weeks, centre IQR 54-68%", "limits": [ "National audit includes exocrine pancreatic cancers rather than PDAC only", "Treatment receipt is not full-course completion", "Reasons for route loss are not fully recorded", "Historical cohorts used older treatment", "Observational complication and survival associations" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "PDAC has high occult systemic risk, so local removal without systemic treatment cannot be assumed to complete curative-intent care." }, "animalEvidence": { "rating": "not-applicable", "reason": "Human route delivery and surgical outcomes decide this claim." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Laboratory models cannot measure referral, recovery or treatment delivery." }, "reproducibility": { "rating": "moderate", "reason": "National and multi-centre cohorts repeatedly show postoperative treatment loss, but definitions and treatment eras differ." }, "commercialConflicts": { "rating": "low", "reason": "The main route evidence comes from public audit and academic observational cohorts, not a product comparison." }, "recency": { "rating": "current", "reason": "The record uses the September 2026 national audit and current NHS service requirements." } } }, "updated": "2026-09-14" }, { "id": "claim-pdac-vaccine-immunity-without-proven-benefit", "title": "PDAC vaccines can teach T cells what to attack; none has yet proved that patients live longer", "shortTitle": "Vaccine immunity, no proven benefit", "type": "claim", "status": "supported", "scope": [ "pdac", "transfer-unproven" ], "summary": "Four randomized treatment programmes covering 1,812 participants failed their main survival or disease-free-survival test despite repeated immune-response signals.", "content": "A therapeutic cancer vaccine is software for the immune system: it presents a tumour feature and attempts to compile that feature into T cells able to find and kill cancer cells. PDAC has repeatedly passed the compile test and failed the patient-outcome test. TeloVac randomized 1,062 people: median overall survival was 7.9 months with chemotherapy, 6.9 with chemotherapy followed by the GV1001 telomerase-peptide vaccine, HR 1.19, P=0.05, and 8.4 with concurrent vaccine, HR 1.05, P=0.64. A 90-person GVAX/CRS-207 trial then reported survival of 6.1 versus 3.9 months, HR 0.59, alongside stronger mesothelin-specific T-cell responses. ECLIPSE did not reproduce this against chemotherapy: among its prespecified 213-person later-line cohort, survival was 3.7 months with the combined vaccines versus 4.6 with chemotherapy, HR 1.17, 95% CI 0.84-1.64. Algenpantucel-L added no survival in 303 randomized patients. ELI-002 generated KRAS-specific T cells but missed the prespecified disease-free-survival endpoint in 144 randomized patients. The arithmetic is 1,062 + 303 + 303 + 144 = 1,812 randomized participants across four failed later-stage programmes. Newer products solve some engineering problems, not the outcome problem. A personalized RNA vaccine produced durable T-cell clones in eight of 16 vaccinated patients, but responders were defined after treatment and had no randomized untreated counterpart. A dendritic-cell product reported two-year recurrence-free survival in 38 selected patients without a control. A preventive KRAS vaccine generated T-cell responses in 18 of 20 high-risk people, but the current study's primary outcomes are safety and immunity, not cancer incidence. The valid statement is narrow: several platforms induce target-specific immunity. No platform has yet shown a randomized survival or recurrence benefit in PDAC.", "sourceIds": [ "middleton-2014-telovac", "le-2015-gvax-crs207", "le-2019-eclipse", "hewitt-2022-algenpantucel-l", "pant-2024-eli002", "wainberg-2025-eli002-final", "elicio-2026-amplify7p-results", "rojass-2025-neoantigen-followup", "lau-2024-dendritic-resected-pdac", "haldar-2026-kras-interception", "clinicaltrials-nct05013216-kras-prevention" ], "links": [ { "target": "treatment", "relation": "constrains" }, { "target": "treatment-long-tail-audit", "relation": "expands" }, { "target": "failure-pdac-vaccine-escalation-reversals", "relation": "explains" }, { "target": "unknown-pdac-vaccine-regimen-contribution", "relation": "leaves-open" }, { "target": "trial-pdac-vaccine-action-map", "relation": "informs" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong", "independentReplication": "moderate", "sampleSize": "Four failed randomized programmes n=1,812 in total; newer uncontrolled cohorts n=16, n=20 and n=38", "effect": "No randomized PDAC vaccine programme has yet established survival or recurrence benefit", "limits": [ "Products, targets and disease settings differ", "A class-level failure does not prove every future vaccine will fail", "AMPLIFY-7P full peer-reviewed results are not public", "IMCODE003 has no result", "Immune-response groups are not randomized groups" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Several products generate tumour-targeted T cells in people." }, "animalEvidence": { "rating": "strong", "reason": "Multiple platforms have extensive model support, which has not predicted patient benefit reliably." }, "inVitroEvidence": { "rating": "strong", "reason": "Antigen recognition and T-cell function are directly measurable." }, "reproducibility": { "rating": "weak", "reason": "Immune activation repeats; randomized clinical benefit does not." }, "commercialConflicts": { "rating": "high", "reason": "Every current product is sponsor-controlled and several key results are company reports." }, "recency": { "rating": "current", "reason": "The record includes trial states and results through September 2026." } } }, "layout": { "x": 396, "y": 342 }, "updated": "2026-09-15" }, { "id": "claim-pdt-local-ablation-without-benefit-proof", "title": "Photodynamic therapy can kill a small illuminated region, but has not shown patient benefit", "shortTitle": "PDT local kill, benefit unproved", "type": "claim", "status": "supported", "scope": [ "pdac", "locally-advanced", "photodynamic-therapy", "local-treatment" ], "summary": "Three small uncontrolled human studies show that a light-activated drug can create a dead zone inside a pancreatic tumour; none establishes longer or better life.", "content": "Photodynamic therapy gives a light-sensitive drug, places an optical fibre into the tumour and switches on light of a matching wavelength. Where drug, light and oxygen overlap, short-lived reactive oxygen molecules damage cells and small blood vessels. This is a local chemical reaction, not whole-body treatment. Across three uncontrolled studies, 35 people received verteporfin or porfimer treatment. A 15-person verteporfin dose study found consistent 12 mm dead zones at 40 joules but large variation in volume; median survival after treatment was 8.8 months under mixed cancer care. A 12-person porfimer study produced more image-defined dead tissue in 6/12, followed by median progression-free survival of 2.6 months and survival of 11.5 months; every participant then received chemotherapy, so attribution is impossible. An eight-person verteporfin study produced a dead zone in 5/8 and none in 3/8. The mean dead-zone diameter was 15.7 mm inside tumours averaging 33.3 mm. A crude sphere model gives (15.7/33.3)^3 = 0.105, about 10% of tumour volume. The shape is not spherical, but the calculation exposes the scale mismatch. These studies verify local tissue injury. They do not show complete tumour coverage, symptom relief, delayed spread or longer survival.", "sourceIds": [ "huggett-2014-verteporfin-pdt", "dewitt-2019-eus-porfimer-pdt", "chandrasekhara-2021-eus-verteporfin-pdt", "clinicaltrials-nct03033225-vertpac02" ], "links": [ { "target": "failure-pdt-geometry-selection-and-bundling", "relation": "bounded-by" }, { "target": "unknown-pdt-spatial-coverage-and-patient-benefit", "relation": "leaves-open" }, { "target": "hypothesis-pdt-measured-spatial-gate", "relation": "motivates" }, { "target": "trial-pdac-pdt-action-map", "relation": "mapped-by" }, { "target": "treatment-modality-ranking", "relation": "updates" }, { "target": "treatment", "relation": "part-of" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "none", "randomised": "none", "independentReplication": "weak", "sampleSize": "Three uncontrolled studies, total n=35; individual n=8, 12 and 15", "effect": "Image-defined dead tissue in 5/8 and 6/12; variable dose response in n=15; no comparative patient benefit", "limits": [ "No randomized comparison", "Small single-centre cohorts", "Different drugs and routes", "Mixed prior and later treatment", "Image change is not patient benefit", "Partial tumour coverage", "Selected anatomy" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Light activation in the presence of drug and oxygen has a measured local chemical effect." }, "animalEvidence": { "rating": "moderate", "reason": "Animal work established tissue injury and informed dose, but cannot answer human benefit." }, "inVitroEvidence": { "rating": "strong", "reason": "Drug activation and reactive oxygen injury are reproducible in controlled systems." }, "reproducibility": { "rating": "moderate", "reason": "Three human studies produced local dead tissue, with substantial person-to-person variation." }, "commercialConflicts": { "rating": "moderate", "reason": "One registry lists a drug-company collaborator; the central evidence is mainly academic and publicly funded." }, "recency": { "rating": "current", "reason": "The record includes the 2026 state of the active combination study." } } }, "layout": { "x": 940, "y": 864 }, "updated": "2026-09-15" }, { "id": "claim-pelican-rfa-no-survival-and-more-harm", "title": "PELICAN shows pancreatic radiofrequency ablation adds harm without survival benefit", "shortTitle": "RFA no benefit, more harm", "type": "claim", "status": "supported", "scope": [ "pdac", "locally-advanced", "radiofrequency-ablation" ], "summary": "In a 188-person randomized analysis, open radiofrequency ablation did not improve survival, increased severe serious events by 16 percentage points and worsened quality of life.", "content": "Radiofrequency ablation inserts electrodes into a tumour and uses alternating current to create enough heat to destroy tissue by coagulation and protein damage. This is high-temperature destruction, not the mild sensitizing heat in regional hyperthermia. PELICAN randomized 190 people with nonprogressive locally advanced pancreatic cancer after about two months of chemotherapy; two control participants withdrew consent, leaving 95 assigned to ablation plus chemotherapy and 93 analysed controls. Median overall survival from randomization was 12.1 versus 11.6 months, hazard ratio 1.07, 95% confidence interval 0.80-1.45, P=0.64. Median progression-free survival was 5.8 versus 6.9 months, HR 1.12, 95% CI 0.82-1.52, P=0.47. Grade 3 or worse serious adverse events occurred in 26/95, 27%, versus 10/93, 11%, P=0.004. The absolute excess was 16 percentage points. If repeated, 1 / 0.16 = 6.25, so about six people would need assignment to the ablation route for one additional severe serious event. Global health scores fell by 14.6, 12.0 and 18.4 points at one, three and 18 months in the ablation group, beyond the study's ten-point clinically important threshold; control changes stayed below it. There was one RFA-related death. This is direct evidence against open pancreatic RFA after short induction chemotherapy. The confidence interval does not prove zero effect in every narrower state, but it excludes the planned large HR 0.65 benefit and gives no reason to expose another unselected cohort.", "sourceIds": [ "seelen-2026-pelican", "clinicaltrials-nct03690323-pelican" ], "links": [ { "target": "failure-rfa-selection-and-route-attrition", "relation": "explains" }, { "target": "unknown-rfa-human-mechanism-after-negative-trial", "relation": "leaves-open" }, { "target": "hypothesis-pelican-specimen-salvage-gate", "relation": "motivates" }, { "target": "trial-pdac-rfa-action-map", "relation": "mapped-by" }, { "target": "failure-local-intensification-pdac-survival-not-shown", "relation": "supports" }, { "target": "treatment-modality-ranking", "relation": "updates" }, { "target": "treatment", "relation": "part-of" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong", "independentReplication": "moderate", "sampleSize": "190 randomized; 188 analysed after two control consent withdrawals; 95 ablation and 93 control", "effect": "OS HR 1.07, 95% CI 0.80-1.45; severe serious events +16 points; clinically important quality-of-life loss", "limits": [ "Open-label", "Two control withdrawals excluded", "Only about two months of induction chemotherapy", "Only 70 of 95 assigned to RFA received it", "Surgery and ablation route cannot be blinded", "Does not test every biological subgroup", "Registry is stale and lacks posted results" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The procedure destroys local tissue, but local destruction did not create net patient benefit." }, "animalEvidence": { "rating": "moderate", "reason": "Model immune and ablation effects cannot override the randomized human result." }, "inVitroEvidence": { "rating": "strong", "reason": "Heat destroys cells but that is not the disputed clinical question." }, "reproducibility": { "rating": "moderate", "reason": "The large modern trial aligns with an older randomized negative result described by the authors." }, "commercialConflicts": { "rating": "low", "reason": "The trial was academic and publicly funded, although Olympus was listed as a registry collaborator." }, "recency": { "rating": "current", "reason": "The randomized result was published in March 2026." } } }, "layout": { "x": 712, "y": 636 }, "updated": "2026-09-15" }, { "id": "claim-pert-survival-observational-randomized-unproved", "title": "Pancreatic enzymes treat maldigestion, but their reported PDAC survival advantage remains unproved", "shortTitle": "PERT survival gap", "type": "claim", "status": "contested", "scope": [ "pdac", "pancreatic-all" ], "summary": "Large treated cohorts report longer survival, while two small randomized trials failed weight endpoints and did not establish survival; treatment selection can explain much of the gap.", "content": "Pancreatic duct obstruction, gland loss and surgery can cause exocrine insufficiency, so replacing digestive enzymes has a direct reason: improve digestion and symptoms. The stronger claim—that PERT extends pancreatic-cancer survival—is not settled. A 501-person advanced-PDAC cohort found median survival of 17.1 months with PERT versus 12.5 months without it, adjusted HR 0.73. A 1,247-person case-control study found adjusted one-year mortality OR 0.49 overall, but the chemotherapy-only subgroup was not significant: OR 0.75, 95% CI 0.44 to 1.24. These are associations. Patients who receive enzymes may have better specialist care, longer survival to prescription, recognized maldigestion, better adherence or different disease. The randomized checks are smaller and short. In 67 unresectable patients, PERT did not improve eight-week weight, nutrition, quality of life or survival; median OS was 5.84 versus 8.13 months, P=.744. In 88 chemotherapy-treated patients, BMI did not differ and median OS was 19.0 versus 12.0 months, P=.070. This does not prove no survival effect; it proves that the current randomized evidence is too small and inconsistent to support one. PERT can still be correct clinical care for confirmed or likely insufficiency. The mission must separate relief of maldigestion from an unproved cancer-survival claim.", "sourceIds": [ "woo-2016-pert-rct", "saito-2018-pert-rct", "picozzi-2025-pert-advanced-pdac", "dugic-2026-pert-mortality" ], "links": [ { "target": "claim-cachexia-is-parallel-lethal-system", "relation": "extends" }, { "target": "failure-pert-short-randomized-trials", "relation": "bounded-by" }, { "target": "unknown-pert-pdac-utility", "relation": "leads-to" }, { "target": "trial-pert-pdac-action-map", "relation": "informs" }, { "target": "treatment", "relation": "part-of" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "strong", "randomised": "moderate — two small trials", "independentReplication": "moderate for association; none for survival causality", "sampleSize": "Randomized n=67 and n=88; retrospective n=501 and n=1,247", "effect": "Observational survival HR .73 and one-year mortality OR .49; randomized BMI and weight endpoints negative and survival not significant", "limits": [ "Confounding by care and patient selection", "Possible survivor-time bias around prescription", "Short randomized endpoints", "Small trials", "Different PERT doses and enzyme-deficiency definitions", "Some commercial author ties" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Replacing missing digestive enzymes can correct maldigestion, but the route from absorption to cancer survival contains several unproved links." }, "animalEvidence": { "rating": "not-applicable", "reason": "Human maldigestion and clinical outcomes decide this supportive-care question." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "A laboratory digestion result cannot establish treatment delivery or survival." }, "reproducibility": { "rating": "moderate", "reason": "Several cohorts repeat an association, while randomized trials have not reproduced a survival effect." }, "commercialConflicts": { "rating": "moderate", "reason": "One large cohort included AbbVie employees and several authors in another study disclosed enzyme-company relationships." }, "recency": { "rating": "current", "reason": "The evidence includes 2025 and 2026 observational analyses beside older randomized trials." } } }, "layout": { "x": 149, "y": 73 }, "updated": "2026-09-14" }, { "id": "claim-pipac-local-delivery-signal-without-benefit-proof", "title": "PIPAC reaches peritoneal disease, but no pancreatic study proves added patient benefit", "shortTitle": "PIPAC signal, benefit unproved", "type": "claim", "status": "supported", "scope": [ "pdac", "peritoneal-metastases", "pipac", "drug-delivery" ], "summary": "Repeated abdominal aerosol chemotherapy has produced tissue responses with a modest procedure-harm rate, while all pancreatic evidence remains uncontrolled and heavily selected.", "content": "PIPAC uses keyhole surgery to fill the abdominal cavity with a pressurized chemotherapy mist for about 30 minutes, then removes the aerosol through a closed system. Its engineering aim is surface coverage: expose cancer deposits spread across the peritoneum, the abdominal lining, while limiting blood concentrations. The largest pancreatic report retrospectively combined six centres: 156 people received 350 procedures, 2.24 each on average. Grade 3-4 adverse events followed 10/350 procedures, 2.9%, and six people, 3.8%, died within 30 days from disease progression. Median survival was nine months from first PIPAC. Only 55/156, 35.2%, completed at least three procedures. Progression or declining condition stopped 87/156, 55.8%. Older five- to twenty-person series reported tissue regression, but repeated biopsies were available mainly in people who survived and stayed fit for another operation. No randomized pancreatic comparison has held systemic chemotherapy constant. PIPAC therefore has a credible local-delivery and tissue-effect signal, not proof of longer or better life.", "sourceIds": [ "digiorgio-2026-pipac-pdac-registry", "graversen-2017-pipac-pdac", "clinicaltrials-nct05371223-nab-pipac", "digiorgio-2024-nab-pipac-protocol" ], "links": [ { "target": "treatment", "relation": "part-of" }, { "target": "failure-pipac-repeat-procedure-selection-and-bundling", "relation": "bounded-by" }, { "target": "unknown-pipac-added-survival-quality-and-systemic-control", "relation": "leaves-open" }, { "target": "hypothesis-pipac-existing-data-and-control-gate", "relation": "motivates" }, { "target": "trial-pdac-pipac-action-map", "relation": "mapped-by" }, { "target": "treatment-modality-ranking", "relation": "updates" } ], "evidence": { "humanProspective": "weak", "humanRetrospective": "moderate", "randomised": "none in pancreatic cancer", "independentReplication": "weak — uncontrolled tissue-response series across centres", "sampleSize": "Largest pancreatic registry n=156 and 350 procedures; older series n=5 to n=20", "effect": "Local tissue-response signal and 9-month median survival from first PIPAC; no controlled patient-benefit estimate", "limits": [ "Retrospective treatment selection", "No standard-care control", "Repeat-procedure survivor selection", "Changing drugs", "Concurrent systemic treatment", "Local response endpoints", "Peritoneal-only reach", "Centre differences" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "A pressurized intraperitoneal aerosol can directly expose abdominal-surface deposits to chemotherapy." }, "animalEvidence": { "rating": "not-applicable", "reason": "Human delivery and tissue-response data are already available." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "The decision turns on net human benefit, not cell killing." }, "reproducibility": { "rating": "weak", "reason": "Several small series and a multicentre registry repeat feasibility, but none controls treatment choice." }, "commercialConflicts": { "rating": "moderate", "reason": "Specialist centres and device workflows can favour expansion; the central gap is still study design." }, "recency": { "rating": "current", "reason": "The evidence includes an August 2026 six-centre registry report and the current trial record." } } }, "layout": { "x": 1078, "y": 1002 }, "updated": "2026-09-15" }, { "id": "claim-pnet-distinct-from-pdac", "title": "Pancreatic neuroendocrine tumours require a separate disease model from PDAC", "shortTitle": "pNET ≠ PDAC", "type": "claim", "status": "supported", "scope": [ "pnet", "pdac" ], "summary": "Whole-genome pNET data show different drivers, inherited contributions and telomere biology from the dominant PDAC model.", "content": "Scarpa and colleagues sequenced 102 primary pancreatic neuroendocrine tumours and described mutational processes, inherited DNA-repair variants and recurrent pathways unlike the canonical PDAC landscape. Combining the two because both arise in the pancreas would corrupt detection assumptions, treatment rankings and survival estimates. Transfer between them must name the shared mechanism, not the organ alone.", "sourceIds": [ "scarpa-2017-pnet", "tcga-2017-pdac" ], "links": [ { "target": "pnet", "relation": "part-of" }, { "target": "pdac", "relation": "contradicts" }, { "target": "disease-system", "relation": "supports" }, { "target": "transfer", "relation": "depends-on" } ], "evidence": { "humanProspective": "none", "humanRetrospective": "strong", "randomised": "not-applicable", "independentReplication": "moderate", "sampleSize": "102 primary pNET whole genomes; 150 clinically annotated pancreatic cancers in TCGA PDAC analysis", "effect": "Materially different recurrent genomic programmes", "limits": [ "Retrospective tissue cohorts", "Within-disease diversity remains", "Genomic distinction does not alone determine clinical management" ] }, "layout": { "x": 5, "y": 20 }, "updated": "2026-09-14" }, { "id": "claim-pnet-seqtor-no-sequence-winner", "title": "SEQTOR did not establish one superior order for everolimus and chemotherapy in advanced pNET", "shortTitle": "pNET sequence unresolved", "type": "claim", "status": "supported", "scope": [ "pnet" ], "summary": "A 141-person randomized crossover trial found no significant PFS or survival difference between two treatment orders, while chemotherapy produced more shrinkage.", "content": "SEQTOR randomized 141 people with advanced well-differentiated pNET to everolimus then streptozotocin plus 5-FU, or the reverse order. First-line 12-month PFS was 71.4% versus 61.8%, with an odds ratio of 0.65 and 95% CI 0.32-1.32; median first-line PFS was 19.4 versus 22.7 months. Median overall survival was 61.7 versus 50.6 months, HR 1.43, 95% CI 0.86-2.37. These intervals do not establish a winner or equivalence. Chemotherapy produced more first-line response, 30.3% versus 11.6%, which matters when rapid shrinkage is needed. The trial also exposes a design problem: slow accrual and longer survival led investigators to replace the planned 35-month two-line PFS endpoint with first-line 12-month PFS. A useful next trial should route by current grade, growth speed, receptor distribution, symptoms and need for shrinkage, then compare complete treatment strategies using quality-adjusted time without progression or major toxicity rather than another isolated drug PFS.", "sourceIds": [ "capdevila-2025-seqtor-pnet" ], "links": [ { "target": "pnet", "relation": "part-of" }, { "target": "model-pnet-causal-chain", "relation": "supports" }, { "target": "claim-pnet-state-changes-at-progression", "relation": "depends-on" }, { "target": "unknown-pnet-reclassification-actionability", "relation": "supports" }, { "target": "trials", "relation": "part-of" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong", "independentReplication": "none for this exact sequence", "sampleSize": "141 randomized patients: 72 everolimus first, 69 STZ/5-FU first", "effect": "No significant PFS or OS sequence difference; first-line response 11.6% versus 30.3%", "limits": [ "Open-label", "Under-recruited", "Primary endpoint changed", "Wide survival confidence interval", "No receptor-directed radiation sequence arm", "No locked biological routing rule" ] }, "layout": { "x": 24, "y": 17 }, "updated": "2026-09-14" }, { "id": "claim-pnet-state-changes-at-progression", "title": "Progressive pNET can change grade and receptor state, so the original biopsy can become obsolete", "shortTitle": "pNET changes at progression", "type": "claim", "status": "supported", "scope": [ "pnet" ], "summary": "In a small prospective cohort, repeat biopsy and dual PET found clinically important changes in 13 of 21 progressive metastatic pNETs.", "content": "A prospective Uppsala study enrolled 21 people with progressive metastatic pNET for repeat clinical review, core biopsy, FDG PET and somatostatin-receptor PET. Adequate repeat tissue was obtained in 16. Ki-67 increased in 13 of 16 and 8 of 16 moved from grade 2 to grade 3. The combined review found a clinically significant change in 13 of 21 and led to a treatment change in seven. This is direct evidence that a diagnosis made years earlier may not describe current metastatic disease. It is not evidence that routine repeat biopsy improves survival: the study was single-centre, small, selected at progression and had no control arm. The operational rule is to repeat the least burdensome measurements when growth pattern changes, receptor-directed treatment is being considered or the old sample no longer represents the dominant lesion. The needed trial randomizes a locked re-characterization-and-action route, not imaging alone.", "sourceIds": [ "mollazadegan-2025-pnet-recharacterization", "grillo-2016-gepnet-grade-metastasis" ], "links": [ { "target": "pnet", "relation": "part-of" }, { "target": "model-pnet-causal-chain", "relation": "extends" }, { "target": "claim-panec-needs-separate-classification", "relation": "extends" }, { "target": "claim-prrt-pnet-randomized-control", "relation": "constrains" }, { "target": "unknown-pnet-reclassification-actionability", "relation": "motivates" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "moderate", "randomised": "none", "independentReplication": "moderate for grade discordance; none for outcome benefit from the full route", "sampleSize": "Prospective n=21; adequate repeat biopsy n=16", "effect": "Ki-67 rose in 13/16; grade 2 to 3 in 8/16; clinically significant change in 13/21; treatment changed in seven", "limits": [ "Single centre", "Very small cohort", "Selected at progression", "No control arm", "Treatment changes were not linked to better outcomes" ] }, "layout": { "x": 23, "y": 16 }, "updated": "2026-09-14" }, { "id": "claim-pnet-surgical-control-is-risk-stratified", "title": "Resected pNET can remain disease-free for years, but recurrence risk is strongly stratified", "shortTitle": "pNET surgical control", "type": "claim", "status": "supported", "scope": [ "pnet" ], "summary": "Two retrospective surgical cohorts reported recurrence in 12.3% and 15.1%; liver recurrence dominated and grade, size, nodes and vascular invasion separated risk.", "content": "Surgery is the only current route to remove a confined pancreatic neuroendocrine tumour, but the available evidence does not support one cure probability for all resected disease. A Verona cohort of 487 resections with 71 months median follow-up recorded recurrence in 12.3%, chiefly liver metastasis, and four recurrences after ten years. Size above 21 mm, G3 grade, nodal disease and vascular invasion were independent risk factors. A separate US multi-institutional cohort of 1,020 curative-intent resections found 15.1% recurrence; about half of those recurrences were liver-only. These similar crude rates provide some replication, but both studies selected surgical patients and neither compared resection with observation. The data support risk-adjusted surveillance and the possibility of durable surgical control in low-risk disease. They do not prove that every small non-functioning pNET benefits from surgery, nor that absence of recurrence at five years equals permanent eradication.", "sourceIds": [ "marchegiani-2019-pnet-recurrence", "dong-2020-pnet-recurrence" ], "links": [ { "target": "model-pnet-causal-chain", "relation": "supports" }, { "target": "pnet", "relation": "part-of" }, { "target": "residual-disease", "relation": "supports" }, { "target": "claim-prrt-pnet-randomized-control", "relation": "precedes" } ], "evidence": { "humanProspective": "none", "humanRetrospective": "moderate", "randomised": "none", "independentReplication": "moderate", "sampleSize": "487 single-centre and 1,020 multi-institutional curative-intent resections", "effect": "Recurrence in 12.3% and 15.1%; liver was the leading first recurrence site", "limits": [ "Surgical selection bias", "No observation comparator", "Long study eras with changing imaging and grading", "Subgroup cure statements are not randomised", "Late recurrence occurs" ] }, "layout": { "x": 41, "y": 19 }, "updated": "2026-09-14" }, { "id": "claim-pnet-targeted-control-differs-from-pdac", "title": "Advanced pNET can often be controlled for longer than PDAC with pathway and vascular drugs", "shortTitle": "pNET control", "type": "claim", "status": "supported", "scope": [ "pnet" ], "summary": "Everolimus and sunitinib roughly doubled median progression-free survival versus placebo in progressive well-differentiated pNET, showing a distinct treatment system.", "content": "In progressive advanced well-differentiated pNET, sunitinib increased median progression-free survival from 5.5 to 11.4 months in a 171-patient trial that stopped early. Everolimus increased it from 4.6 to 11.0 months in 410 patients, with crossover after progression. These gains are materially longer than many PDAC responses and fit pNET's different vascular and pathway biology. They are not cures: objective response with sunitinib was 9.3%, everolimus mainly slowed progression, and crossover or early stopping complicates overall-survival estimates. A pancreatic mass labelled only 'pancreatic cancer' would therefore make the treatment map wrong before therapy begins.", "sourceIds": [ "raymond-2011-sunitinib-pnet", "yao-2011-everolimus-pnet" ], "links": [ { "target": "pnet", "relation": "part-of" }, { "target": "claim-pnet-distinct-from-pdac", "relation": "supports" }, { "target": "treatment", "relation": "supports" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong", "independentReplication": "strong", "sampleSize": "171 randomised with sunitinib; 410 with everolimus", "effect": "Median PFS 11.4 versus 5.5 months and 11.0 versus 4.6 months", "limits": [ "Progressive well-differentiated pNET only", "Sunitinib trial stopped early", "Everolimus crossover", "PFS is not cure or overall survival", "Industry-funded" ] }, "layout": { "x": 4, "y": 12 }, "updated": "2026-09-14" }, { "id": "claim-ponsegromab-access-stack", "title": "A ponsegromab decision needs drug, GDF-15 assay, endpoint and data access, not only patent permission", "shortTitle": "Ponsegromab access stack", "type": "claim", "status": "supported", "scope": [ "pdac", "pan-cancer", "transfer-unproven" ], "summary": "Pfizer controls the medicine and trial while two patent families cover the antibody and cancer-cachexia use; independent utility review also needs biomarker and treatment-delivery data.", "content": "The access question is not solved by listing two patent publications. WO2020039321A2 points to the anti-GDF15 antibody family; the PCT record is marked ceased, but listed national family members include grants, so the family cannot be called dead. WO2025245200A1 is a pending PCT use-and-dose application covering ponsegromab with systemic chemotherapy in cancer cachexia including pancreatic cancer. Patent outcome ranges are drafting claims, not evidence. The clinical decision also depends on at least six operating layers: drug supply; the GDF-15 assay and threshold; chemotherapy dose and completion data; function and wearable data with missingness reasons; tumour and survival outcomes; and permission for independent analysis and negative publication. NCT06989437 is Pfizer-sponsored and recruits across 220 locations, which makes execution strong but centralizes those layers. A mission agreement should obtain the locked assay method, arm-level patient data, missing-data rules, reasons for chemotherapy change, drug access for a named follow-on setting and a publication timetable. No acquisition is justified by the current record: the needed action is a bounded data-and-supply agreement. Exact national claims, expiry, validity and freedom to operate require counsel before any follow-on trial.", "sourceIds": [ "clinicaltrials-nct06989437-ponsegromab-pdac", "pfizer-2026-ponsegromab-profile", "patent-wo2020039321-anti-gdf15", "patent-wo2025245200-ponsegromab-cachexia" ], "links": [ { "target": "patent-landscape", "relation": "extends" }, { "target": "company-pfizer-cachexia", "relation": "depends-on" }, { "target": "trial-pdac-cachexia-action-map", "relation": "constrains" }, { "target": "programme-500m-integrated-mission", "relation": "constrains" }, { "target": "risks", "relation": "supports" } ], "evidence": { "humanProspective": "not-applicable", "humanRetrospective": "not-applicable", "randomised": "not-applicable", "independentReplication": "moderate for public ownership and trial layers; none for contract access", "sampleSize": "Two priority patent families, one company development profile and one 982-person trial registry", "effect": "Six non-exclusive access layers identified; no efficacy effect", "limits": [ "Not a legal opinion", "No licence or data terms obtained", "Google legal status is not authoritative", "National claims differ", "The family search is not exhaustive", "Trial operating details may change" ], "dimensions": { "biologicalPlausibility": { "rating": "not-applicable", "reason": "This record maps operating and ownership control rather than biological truth." }, "animalEvidence": { "rating": "not-applicable", "reason": "Animal evidence cannot establish data, assay, drug-supply or legal access." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Analytical performance cannot establish independent operating access." }, "reproducibility": { "rating": "moderate", "reason": "Public patent and registry records establish the layers, but the mission has not tested a contract." }, "commercialConflicts": { "rating": "high", "reason": "One sponsor owns the product, patent position, pivotal trial and key data." }, "recency": { "rating": "current", "reason": "Patent, company and trial records were checked through September 2026." } } }, "layout": { "x": 147, "y": 71 }, "updated": "2026-09-14" }, { "id": "claim-ponsegromab-gdf15-cachexia-driver-not-survival", "title": "GDF-15 blockade can reverse weight loss and symptoms, but has not yet shown longer survival or more chemotherapy delivered", "shortTitle": "GDF-15 cachexia result", "type": "claim", "status": "supported", "scope": [ "pdac", "colorectal", "pan-cancer", "transfer-unproven" ], "summary": "Ponsegromab produced a dose-related 12-week weight gain in a randomized selected population; the result validates a cachexia driver, not cancer control.", "content": "In a 187-person randomized double-blind phase 2, patients with cancer cachexia and serum GDF-15 at least 1,500 pg/mL received ponsegromab 100, 200 or 400 mg or placebo every four weeks. Thirty-two per cent had pancreatic cancer. At twelve weeks the placebo-adjusted median weight differences were +1.22 kg, +1.92 kg and +2.81 kg. At 400 mg, appetite, cachexia symptoms and measured activity also improved. These results show that GDF-15 contributes causally to wasting in the selected population. They do not show that blocking it prolongs survival, keeps chemotherapy on dose, improves tumour response or works in a PDAC-only population. The primary endpoint was weight at twelve weeks, follow-up was short and Pfizer funded the trial. Any programme claim of durable control must require a chain longer than weight gain: function plus treatment delivered plus quality of life, with survival and toxicity measured. A positive body-weight endpoint with no treatment or survival gain would remain useful supportive care, but it would not count as cancer control.", "sourceIds": [ "groarke-2024-ponsegromab-cachexia", "clinicaltrials-nct06989437-ponsegromab-pdac", "pfizer-2026-ponsegromab-profile" ], "links": [ { "target": "claim-cachexia-is-parallel-lethal-system", "relation": "supports" }, { "target": "trial-pdac-cachexia-action-map", "relation": "informs" }, { "target": "unknown-cachexia-treatment-utility", "relation": "leads-to" }, { "target": "company-pfizer-cachexia", "relation": "developed-by" }, { "target": "treatment", "relation": "part-of" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong", "independentReplication": "none for ponsegromab", "sampleSize": "187 randomized across three cancer types; about 32% had pancreatic cancer", "effect": "Placebo-adjusted 12-week weight difference +1.22, +1.92 and +2.81 kg across ascending doses; symptoms and activity favoured 400 mg", "limits": [ "Weight rather than survival primary endpoint", "Twelve-week follow-up", "Mixed cancers", "GDF-15-selected population", "No PDAC subgroup effect reported in the abstract", "No chemotherapy-delivery endpoint", "Pfizer funded and employed several authors" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "A biomarker-selected randomized intervention changed weight, symptoms and activity in the expected direction." }, "animalEvidence": { "rating": "moderate", "reason": "The development route has model support, but the decision rests on the randomized human result." }, "inVitroEvidence": { "rating": "moderate", "reason": "Target binding supports the mechanism but cannot establish whole-body or treatment benefit." }, "reproducibility": { "rating": "none", "reason": "The exact randomized ponsegromab effect has one sponsor-run phase 2 result." }, "commercialConflicts": { "rating": "high", "reason": "Pfizer funded the study, employed several authors and owns the medicine and relevant patent families." }, "recency": { "rating": "current", "reason": "The result was published in 2024 and the PDAC trial record was updated in August 2026." } } }, "layout": { "x": 133, "y": 64 }, "updated": "2026-09-14" }, { "id": "claim-prrt-pnet-randomized-control", "title": "Radioligand therapy now has randomized evidence for longer pNET control, with late marrow risk", "shortTitle": "pNET radioligand control", "type": "claim", "status": "supported", "scope": [ "pnet" ], "summary": "OCLURANDOM produced 12-month PFS of 80.5% with lutetium-177 DOTATATE versus 41.9% with sunitinib; one treatment-related acute leukaemia occurred.", "content": "Pancreatic neuroendocrine tumours that express somatostatin receptors can be targeted with a radioactive peptide, a biology and treatment route distinct from PDAC. OCLURANDOM randomized 84 pretreated, progressive metastatic pNET patients to four cycles of lutetium-177 DOTATATE or daily sunitinib. Twelve-month progression-free survival was 80.5% versus 41.9%, an absolute difference of 38.6 percentage points and about 1.92 times the sunitinib rate. Grade 3–4 adverse events were 44% versus 72%, and quality of life during treatment favoured radioligand therapy by 10.3 points. Late grade 2-or-worse events occurred in 60% of the radioligand group, and one patient died from treatment-related acute leukaemia. COMPETE provides a separate phase 3 signal across 309 mixed gastroenteropancreatic NET patients: lutetium-177 edotreotide improved median PFS to 23.9 from 14.1 months versus everolimus, HR 0.67. CONTROL NETS then tested radioligand therapy plus capecitabine-temozolomide, but only 27 people had pNET. The 27-month PFS rates were 61.1% and 33.3%; the hazard ratio was 0.41, but its 95% confidence interval, 0.15-1.12, crossed no effect and p=0.08. Response was higher, while two of 63 radioligand-exposed participants developed treatment-related myeloid cancers. This is a combination signal needing phase 3, not established superiority. Together these trials establish a strong disease-control route, not cure or a pancreatic-specific survival gain. The programme question is sequencing: use receptor imaging, grade, growth rate, marrow reserve and later treatment needs to choose radioligand therapy rather than applying it uniformly.", "sourceIds": [ "baudin-2026-oclurandom", "walter-2026-compete", "chan-2026-control-nets", "raymond-2011-sunitinib-pnet", "yao-2011-everolimus-pnet" ], "links": [ { "target": "pnet", "relation": "part-of" }, { "target": "claim-pnet-distinct-from-pdac", "relation": "supports" }, { "target": "claim-pnet-targeted-control-differs-from-pdac", "relation": "extends" }, { "target": "claim-cabozantinib-pnet-pfs-not-cure", "relation": "competes-with" }, { "target": "experiments", "relation": "tests-sequencing-in" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong", "independentReplication": "moderate", "sampleSize": "OCLURANDOM pNET n=84; COMPETE mixed GEP-NET n=309; CONTROL NETS pNET n=27", "effect": "OCLURANDOM 12-month PFS 80.5% vs 41.9%; COMPETE median PFS 23.9 vs 14.1 months, HR 0.67; CONTROL NETS combination PFS HR 0.41, 95% CI 0.15-1.12", "limits": [ "OCLURANDOM phase 2 and non-comparative design", "Open-label", "Cross-over allowed", "No established OS advantage", "COMPETE pooled primary sites", "CONTROL NETS had only 27 pNET participants and a non-comparative phase 2 design", "CONTROL NETS confidence interval crossed no effect", "Somatostatin-receptor-positive patients only", "Late marrow toxicity", "Treatment-related myeloid cancers occurred", "Specialist-centre delivery" ] }, "layout": { "x": 45, "y": 82 }, "updated": "2026-09-14" }, { "id": "claim-rectal-anatomy-changes-local-treatment", "title": "Rectal anatomy creates a local-treatment problem that should not be merged with colon cancer", "shortTitle": "Rectal local treatment", "type": "claim", "status": "supported", "scope": [ "colorectal" ], "summary": "PROSPECT showed that selected sphincter-sparing candidates could receive neoadjuvant FOLFOX with selective radiation without worse five-year disease-free survival.", "content": "The pelvis makes local control, radiation toxicity and organ preservation central in rectal cancer, unlike most colon cancer. In the NCI-funded PROSPECT trial, 1,194 selected patients were randomized to preoperative FOLFOX with chemoradiotherapy reserved for poor response or intolerance, or routine chemoradiotherapy. Five-year disease-free survival was 80.8% versus 78.6%, meeting the declared noninferiority boundary; overall survival and local recurrence were similar, and only 9.1% of treated FOLFOX patients required preoperative radiation. This does not show that radiation is unnecessary for all rectal cancer: participants were candidates for sphincter-sparing surgery and excluded more locally advanced states. It does show why colorectal records must separate molecular systemic treatment from anatomy-specific local treatment.", "sourceIds": [ "schrag-2023-prospect-rectal" ], "links": [ { "target": "model-colorectal-causal-chain", "relation": "supports" }, { "target": "colorectal", "relation": "part-of" }, { "target": "treatment-modality-ranking", "relation": "extends" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong", "independentReplication": "moderate", "sampleSize": "1,194 randomized; 1,128 started assigned treatment", "effect": "Five-year DFS 80.8% versus 78.6%; HR 0.92, 90.2% CI 0.74-1.14; noninferior", "limits": [ "Selected sphincter-sparing candidates", "Open-label", "Noninferiority design", "Does not address metastatic liver disease", "Does not generalize to colon cancer" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Pelvic anatomy makes margin control, radiation exposure and organ preservation distinct decisions." }, "animalEvidence": { "rating": "not-applicable", "reason": "The local-treatment comparison is decided by human surgery and recurrence outcomes." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Laboratory response cannot represent pelvic control or radiation harm." }, "reproducibility": { "rating": "moderate", "reason": "PROSPECT is a large randomized trial, but this exact selective-radiation rule has limited direct replication." }, "commercialConflicts": { "rating": "low", "reason": "The NCI-funded comparison reduced use of a treatment rather than supporting a proprietary product." }, "recency": { "rating": "recent", "reason": "The randomized report was published in 2023." } } }, "layout": { "x": 69, "y": 29 }, "updated": "2026-09-14" }, { "id": "claim-routine-imaging-after-pdac-resection-not-yet-proven", "title": "Routine imaging after PDAC resection is associated with longer survival but is not yet proven to cause it", "shortTitle": "Routine imaging not yet proven", "type": "claim", "status": "supported", "scope": [ "pdac" ], "summary": "Two large observational comparisons favour scheduled imaging, while non-random choice and earlier diagnosis leave the survival effect unresolved.", "content": "The signal is consistent and potentially important. In an international study of 333 patients with diagnosed PDAC recurrence, routine imaging was associated with more recurrence-focused treatment, adjusted OR 2.57, and longer overall survival from surgery, adjusted HR 0.75; median survival was 28 versus 23 months. In a Dutch nationwide cohort of 1,311 resected patients, only 139, 11%, received routine imaging. It was associated with asymptomatic recurrence OR 3.79, recurrence treatment OR 2.50 and overall survival HR 0.56; median survival was 43 versus 22 months. Those effects are too large to accept as a scan effect without randomization. Healthier patients, trial participants, better-resourced centres and people already considered treatable may have been more likely to receive imaging. The international study also included only people whose recurrence was diagnosed. Finding recurrence earlier moves its recorded date even if death is unchanged, although both studies measured survival from surgery to reduce simple lead-time error. A 2025 Pan-Asian ESMO panel used the international study to support regular follow-up at level III, grade B; that is a consensus response to non-randomized evidence, not clinical utility proof. The live NCT04875325 RADAR-PANC trial randomizes scheduled CA19-9 plus CT every three months against mainly symptom-led care and uses overall survival as its primary endpoint. Until it reports, routine imaging is a plausible action under test, not a proven survival treatment.", "sourceIds": [ "andel-2025-routine-imaging", "van-goor-2025-nationwide-imaging", "tai-2025-pan-asian-esmo-pancreatic", "daamen-2024-radar-panc-protocol", "clinicaltrials-nct04875325-radar-panc", "nice-2018-ng85-pancreatic" ], "links": [ { "target": "residual-disease", "relation": "part-of" }, { "target": "claim-pdac-surgery-is-a-complete-treatment-route", "relation": "extends" }, { "target": "unknown-postresection-surveillance-utility", "relation": "leads-to" }, { "target": "trial-pdac-postresection-surveillance-action-map", "relation": "informs" }, { "target": "trial-pdac-ctdna-action-map", "relation": "sets-comparator-for" }, { "target": "person-lois-daamen", "relation": "studied-by" }, { "target": "lab-dpcg-radar-panc-network", "relation": "tested-by" } ], "evidence": { "humanProspective": "moderate — non-randomized international recurrence cohort", "humanRetrospective": "strong — nationwide observational cohort", "randomised": "none reported", "independentReplication": "moderate", "sampleSize": "International recurrence cohort n=333; Dutch nationwide resection cohort n=1,311; live randomized plan n=306", "effect": "Adjusted survival HR 0.75 and 0.56 in two non-randomized comparisons", "limits": [ "Follow-up strategy not randomized", "Confounding by fitness, centre and treatment eligibility", "International cohort conditioned on diagnosed recurrence", "Earlier detection changes recorded recurrence time", "Imaging protocols varied", "No completed randomized utility result" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "Earlier recurrence detection helps only when a tolerable local or systemic treatment works better at lower burden." }, "animalEvidence": { "rating": "not-applicable", "reason": "Human follow-up and treatment outcomes decide this claim." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Laboratory models cannot estimate scan utility, anxiety or treatment access." }, "reproducibility": { "rating": "moderate", "reason": "Two independent observational designs point in the same direction, but share non-random treatment-selection problems." }, "commercialConflicts": { "rating": "low", "reason": "The studies test care schedules rather than a proprietary product." }, "recency": { "rating": "current", "reason": "The observational results and guideline response are from 2025, with live trial status checked in September 2026." } } }, "updated": "2026-09-14" }, { "id": "claim-satricel-pdac-early-signal", "title": "Satri-cel produces some pancreatic responses but has no controlled PDAC benefit estimate", "shortTitle": "Satri-cel early PDAC signal", "type": "claim", "status": "mixed", "scope": [ "pdac", "pan-cancer", "transfer-unproven" ], "summary": "Four of 24 previously treated patients responded and median response duration was 9.5 months, but the pooled early-phase cohort had no concurrent control.", "content": "Engineering model. CAR-T gives a patient's own T cells a synthetic surface sensor. Unlike a normal T-cell receptor, it can read CLDN18.2 directly without the patient's HLA tissue type presenting a protein fragment. The cell must still be manufactured, infused after lymphocyte-depleting treatment, reach each lesion, keep its sensor, remain active and avoid normal stomach tissue that carries the same address. Advanced-disease signal. Two early-phase trials pooled 24 infused pancreatic patients. Four responded, ORR 16.7%; disease control was 17/24=70.8%; median PFS was 3.3 months, OS 10.0 months and response duration 9.5 months. There was no concurrent control, and separating responders or stable disease from patients with progression after the outcome cannot estimate treatment effect. Adjuvant signal. Six people with abnormal CA19-9 but no visible recurrence received satri-cel 6-10 months after surgery. One recurred by a median 6.05 months of follow-up. All six had cytokine-release syndrome after the first infusion; five had severe low white-cell counts, and the one person given a second infusion had grade 3 cytokine-release syndrome. These patients had already remained recurrence-free long enough to reach cell infusion, so five remaining recurrence-free cannot be compared with patients counted from surgery. Conclusion. Satri-cel has the strongest current CAR-T response signal in PDAC, not proof of longer useful life. The next claim requires a concurrent control from the same decision time, every screen-to-infusion denominator, fixed CLDN18.2 testing, quality-adjusted survival and complete toxicity.", "sourceIds": [ "qi-2024-ct041-pdac", "clinicaltrials-nct03874897-ct041", "clinicaltrials-nct04581473-ct041", "clinicaltrials-nct05911217-satricel-adjuvant", "carsgen-2025-satricel-pdac-adjuvant" ], "links": [ { "target": "treatment", "relation": "part-of" }, { "target": "treatment-modality-ranking", "relation": "ranked-by" }, { "target": "transfer-car-t-lineage-access", "relation": "extends" }, { "target": "failure-cell-therapy-route-and-denominator-pdac", "relation": "bounded-by" }, { "target": "unknown-engineered-cell-reproducible-benefit-pdac", "relation": "leaves-open" }, { "target": "trial-pdac-engineered-cell-action-map", "relation": "mapped-by" }, { "target": "company-carsgen-therapeutics", "relation": "developed-by" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "not-applicable", "randomised": "none", "independentReplication": "none", "sampleSize": "Advanced pooled n=24 infused; adjuvant n=6 of 20 planned", "effect": "Advanced ORR 16.7%, DCR 70.8%, median PFS 3.3 months, OS 10.0 months and response duration 9.5 months; uncontrolled", "limits": [ "Open-label early phase", "No concurrent PDAC control", "Pooled trials", "Infused denominator", "Short adjuvant follow-up", "Conditioning toxicity", "Company-controlled product" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "A directly measured surface protein can be recognized without HLA presentation, and pancreatic responses occurred." }, "animalEvidence": { "rating": "moderate", "reason": "Models support CLDN18.2-directed killing but do not solve human access, persistence or normal-tissue risk." }, "inVitroEvidence": { "rating": "strong", "reason": "Binding, killing, sensor density and product potency can be measured before infusion." }, "reproducibility": { "rating": "weak", "reason": "The PDAC evidence pools small uncontrolled cohorts and has no independent product replication." }, "commercialConflicts": { "rating": "high", "reason": "CARsgen owns the product and controls the trials, data and manufacturing route." }, "recency": { "rating": "current", "reason": "Uses the 2024 pooled paper, 2025 adjuvant poster and current registries." } } }, "layout": { "x": 226, "y": 142 }, "updated": "2026-09-15" }, { "id": "claim-screening-base-rate-maths", "title": "Population screening requires extreme specificity or strong risk enrichment", "shortTitle": "Base-rate problem", "type": "claim", "status": "supported", "scope": [ "pancreatic-all" ], "summary": "At an annual incidence near 13.9 per 100,000, even a 99% specific annual test would create roughly 72 false positives per true case before confirmatory testing.", "content": "Illustrative calculation using annual incidence as the pretest rate and assuming 100% sensitivity: among 100,000 people, expect about 13.9 true cases. At 99% specificity, about 999.9 people without cancer test positive, so positive predictive value is 13.9 ÷ (13.9 + 999.9) = 1.37%, or about 72 false positives per true positive. At 99.9% specificity the positive predictive value is about 12.2%; at 99.99% it is about 58.2%. A multi-year detectable window would raise pretest prevalence, and real sensitivity is lower. The calculation does not set a final specification; it proves why risk enrichment and a safe confirmatory pathway are central engineering requirements.", "sourceIds": [ "seer-pancreas-statfacts", "uspstf-2019-pancreatic-screening" ], "links": [ { "target": "detection", "relation": "supports" }, { "target": "unknowns", "relation": "supports" }, { "target": "experiments", "relation": "tests" } ], "evidence": { "humanProspective": "not-applicable", "humanRetrospective": "strong", "randomised": "not-applicable", "independentReplication": "not-applicable", "sampleSize": "Population incidence estimate", "effect": "Illustrative PPV: 1.37% at 99% specificity; 12.2% at 99.9%; 58.2% at 99.99%", "limits": [ "Uses one-year incidence as approximate pretest rate", "Assumes 100% sensitivity", "Does not model age or risk strata", "Confirmatory testing can alter downstream harm" ], "dimensions": { "biologicalPlausibility": { "rating": "not-applicable", "reason": "This is a transparent base-rate calculation, not a biological mechanism claim." }, "animalEvidence": { "rating": "not-applicable", "reason": "Animal evidence cannot resolve population positive predictive value." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Laboratory assay performance does not remove the stated arithmetic constraint." }, "reproducibility": { "rating": "strong", "reason": "The result follows directly from stated incidence, sensitivity and specificity assumptions." }, "commercialConflicts": { "rating": "none", "reason": "The calculation uses public incidence and guideline inputs rather than company performance claims." }, "recency": { "rating": "mixed", "reason": "The arithmetic is stable while incidence and screening guidance require periodic updating." } } }, "layout": { "x": 40, "y": 103 }, "updated": "2026-09-14" }, { "id": "claim-setidegrasib-g12d-degradation", "title": "KRAS-G12D degradation produced PDAC responses but short median control in a 21-patient phase 1 subgroup", "shortTitle": "G12D degradation", "type": "claim", "status": "open", "scope": [ "pdac" ], "summary": "Setidegrasib produced a 24% response rate in 21 heavily treated PDAC patients, with median PFS 3.0 months and OS 10.3 months; a masked first-line phase 3 combination trial is recruiting.", "content": "Setidegrasib is an intravenous degrader built to remove KRAS G12D, the most common single KRAS allele in PDAC. In the phase 1 trial, 21 metastatic PDAC patients received the selected 600 mg weekly dose as second or third line; 67% were in third line. Five responded, 24%, with a wide 95% confidence interval of 8% to 47%. Median progression-free survival was 3.0 months and overall survival 10.3 months. The result proves a second physical way to attack KRAS, degradation rather than inhibition, but the small uncontrolled subgroup and short median PFS do not show superiority to another RAS drug. A recruiting 614-person phase 3 trial tests setidegrasib plus mFOLFIRINOX or NALIRIFOX against placebo plus the same chemotherapy in untreated KRAS-G12D metastatic disease. Its masking and overall-survival primary endpoint make it a strong test of combination value. The key comparison with daraxonrasib is not cross-trial response rate; it is whether allele-specific protein removal adds depth or durability without making first-line chemotherapy intolerable.", "sourceIds": [ "park-2026-setidegrasib", "nct07409272-setidegrasib-first-line", "oreilly-2026-daraxonrasib" ], "links": [ { "target": "claim-daraxonrasib-common-ras-survival-breakthrough", "relation": "competes-with" }, { "target": "trial-daraxonrasib-expansion-programme", "relation": "competes-with" }, { "target": "claim-kras-mutant-dosage-may-shape-resistance", "relation": "tests" }, { "target": "treatment", "relation": "part-of" }, { "target": "companies", "relation": "part-of" } ], "evidence": { "humanProspective": "weak", "humanRetrospective": "not-applicable", "randomised": "none", "independentReplication": "none", "sampleSize": "21 previously treated PDAC patients at the selected phase 2 dose; 203 solid-tumour patients overall", "effect": "PDAC ORR 24%, median PFS 3.0 months, median OS 10.3 months", "limits": [ "Phase 1", "Single arm", "Small PDAC subgroup", "Wide response confidence interval", "Most patients in third line", "Manufacturer-funded", "Cross-trial comparison is invalid", "Phase 3 has no results" ] }, "layout": { "x": 95, "y": 72 }, "updated": "2026-09-14" }, { "id": "claim-sil204-trial-scales-before-human-knockdown", "title": "The SIL204 trial scales before the new RNA drug has reported human KRAS knockdown", "shortTitle": "SIL204 proof order", "type": "claim", "status": "open", "scope": [ "pdac", "colorectal", "pan-cancer", "transfer-unproven" ], "summary": "A staged 403-person trial is authorised, but its public release gates emphasise safety, plasma exposure and dose rather than direct mutant-KRAS reduction in primary and distant tumour.", "content": "Small interfering RNA, or siRNA, is a short matching instruction that can load into the cell's RISC cutting system and destroy a chosen messenger RNA before it is translated into protein. SIL204 aims to cut mutant KRAS messenger RNA. It is a new soluble, chemically modified product derived from the older siG12D-LODER implant, not the same product with a new name. The integrated route injects SIL204 into at least three sites in the primary tumour through endoscopic ultrasound and also gives it under the skin to seek distant disease. The authorised CTIS protocol plans 403 participants. Segment 1 uses 15 to 21 people for safety, three subcutaneous doses, a fixed intratumour dose, plasma pharmacokinetics and cardiac checks. Segment 2 selects a dose; Segment 3 is confirmatory, with overall survival primary in the G12D/V subset. Staging is therefore real. The proof order is still incomplete. The public objectives do not make paired tumour mutant-KRAS RNA or protein reduction a release gate, and the sponsor says individual participant data will not be shared. The first-generation randomized study did not extend survival overall, tested only 37 in its randomized cohort, identified mutation status in 31/59 overall, and found no measurable siRNA in plasma after local implants. It did not validate systemic coverage for this new dual route. As of 29 July 2026, one Israeli site was cleared to begin screening and first dosing was still anticipated. CTIS says German recruitment pending. Site activation is not patient exposure. The company release describes an 18-person run-in and about 166 randomized people, while CTIS and the later filing describe about 403 total; the public records must reconcile those denominators. The mission should not fund the company or its trial. It should offer independent tumour pharmacology and statistics only under a contract that makes target, dose, participant flow and negative results public before expansion.", "sourceIds": [ "alagesan-2026-sig12d-loder", "ctis-2026-sil204-phase23", "silexion-2026-sil204-site-initiation", "silexion-2026-halfyear-filing" ], "links": [ { "target": "treatment", "relation": "part-of" }, { "target": "failure-sig12d-loder-randomized-os-not-shown", "relation": "follows" }, { "target": "company-silexion-therapeutics", "relation": "developed-by" }, { "target": "claim-pdac-delivery-is-variable-not-single-barrier", "relation": "tests" }, { "target": "unknown-sil204-human-target-delivery", "relation": "leaves" }, { "target": "trial-sil204-action-map", "relation": "mapped-by" }, { "target": "hypothesis-multicompartment-kras-sirna-delivery", "relation": "constrains" } ], "evidence": { "humanProspective": "weak — SIL204 had not yet reported a first treated person at the public cutoff", "humanRetrospective": "not-applicable", "randomised": "moderate for the failed first-generation contribution; none for SIL204", "independentReplication": "none", "sampleSize": "First-generation study n=59 with randomized cohort n=37; SIL204 Segment 1 plans 15-21 and integrated trial plans about 403", "effect": "No human SIL204 target or efficacy result; authorised protocol and one activated site only", "limits": [ "Sponsor filings and registry describe plans", "Israeli recruitment status is not independently registered on the cited CTIS page", "Protocol details may change", "No patient-level data-sharing plan", "First-generation and second-generation products are not interchangeable" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "RNA cutting can suppress KRAS if enough guide strand reaches the tumour-cell cytosol." }, "animalEvidence": { "rating": "moderate", "reason": "Sponsor models support local and subcutaneous routes but do not reproduce human tumour distribution." }, "inVitroEvidence": { "rating": "moderate", "reason": "Cell systems show mutant-KRAS RNA and growth effects; delivery is easier than in human PDAC." }, "reproducibility": { "rating": "weak", "reason": "No independent human SIL204 result exists and the first-generation overall result was negative." }, "commercialConflicts": { "rating": "high", "reason": "The new-product evidence, trial design and operational updates are sponsor-controlled." }, "recency": { "rating": "current", "reason": "Uses the June 2026 authorisation, July site update and August filing." } } }, "layout": { "x": 177, "y": 96 }, "updated": "2026-09-15" }, { "id": "claim-smad4-failure-route-marker-contested", "title": "SMAD4 is prognostic, but its use to choose local versus systemic treatment is not reliable", "shortTitle": "SMAD4 route dispute", "type": "claim", "status": "contested", "scope": [ "pdac" ], "summary": "The rapid-autopsy cohort linked SMAD4 loss with widespread metastasis, while a 127-patient resection cohort did not confirm the predicted recurrence pattern.", "content": "The 2009 rapid-autopsy study found DPC4/SMAD4 loss was associated with widespread metastatic rather than locally destructive end-stage disease. This led to a plausible treatment rule: intensify local therapy when SMAD4 is intact and systemic therapy when it is lost. A later 127-patient resection series did not find the predicted local-versus-distant failure split. Other cohorts have reported varying associations with worse survival or recurrence. The current defensible claim is narrower: SMAD4 loss carries biological and prognostic information, but it is not a validated stand-alone treatment-selection test. A prospective stratified trial or locked multimarker model is required before using it to withhold local or systemic treatment.", "sourceIds": [ "iacobuzio-donahue-2009-autopsy-patterns", "winter-2013-smad4-failure-patterns" ], "links": [ { "target": "claim-pdac-death-can-be-local-or-metastatic", "relation": "supports" }, { "target": "unknowns", "relation": "part-of" }, { "target": "treatment", "relation": "depends-on" } ], "evidence": { "humanProspective": "none", "humanRetrospective": "moderate", "randomised": "none", "independentReplication": "weak", "sampleSize": "76 autopsies in original study; 127 resected patients in negative pattern study", "effect": "Original association P=0.007; later cohort did not confirm predicted failure pattern", "limits": [ "Retrospective cohorts", "Different disease stages and treatments", "Immunohistochemistry scoring varies", "Prognosis is not treatment prediction", "No locked prospective treatment rule" ] }, "layout": { "x": 89, "y": 62 }, "updated": "2026-09-14" }, { "id": "claim-small-nfpnet-surveillance-surgery-tradeoff", "title": "Small non-functioning pNET needs a surgery-harm rule, not a two-centimetre slogan", "shortTitle": "Small pNET surgery trade-off", "type": "claim", "status": "supported", "scope": [ "pnet" ], "summary": "Prospective observation looks safe over about two years for selected tumours, while pancreatic surgery causes major harm often enough that size alone cannot settle the decision.", "content": "ASPEN prospectively enrolled 500 people with sporadic, asymptomatic, non-functioning pancreatic neuroendocrine neoplasms no larger than 2 cm. Treating centres chose observation for 406 and surgery for 94. Assignment was not comparable: surgery was associated with age 64 or below, size above 10 mm, main-duct diameter above 3 mm, enrolment at a surgical centre and anxiety score. Patient preference accounted for 42 of 94 operations and centre preference for 37. After median 25 months, nine of 406 observed patients crossed to surgery and none developed distant metastasis during follow-up. Four people had distant metastases at entry and all entered surgery; 4/500 = 0.8%, not evidence that observation caused spread. Among the operated group, 19 of 94 had at least one aggressive pathological feature and severe complications occurred in 13%. Seventeen of those 19 aggressive specimens were above 10 mm; the two below 10 mm had duct dilation. This supports duct change and the 1-2 cm range as warning states, not a universal operation threshold. PANDORA separately observed 76 people for median 17 months: eight, or 10.5%, met its growth rule, two chose surgery and none developed metastases. These short, selected cohorts support guarded observation, especially below 1 cm without duct change. They do not prove lifetime safety or that surgery is useless. A 2026 multicentre series of 460 resections supplies the opposing harm bound: 25.0% clinically relevant pancreatic fistula, 14.8% major morbidity and 0.9% 90-day mortality. Operation type changed harm and lymph-node assessment. The decision must therefore estimate future metastatic or functional harm without surgery, the harm of the procedure actually required by tumour location, and the patient's ability to remain under long follow-up.", "sourceIds": [ "partelli-2022-aspen-interim", "clinicaltrials-nct03084770-aspen", "heidsma-2021-pandora-pnet", "maekawa-2026-small-pnet-surgery-morbidity" ], "links": [ { "target": "pnet", "relation": "supports" }, { "target": "claim-pnet-surgical-control-is-risk-stratified", "relation": "extends" }, { "target": "failure-aspen-cannot-estimate-surgery-benefit", "relation": "limited-by" }, { "target": "unknown-small-pnet-surgery-selector", "relation": "exposes" }, { "target": "hypothesis-pnet-time-updated-surgery-rule", "relation": "supports" } ], "evidence": { "humanProspective": "strong — two prospective observation cohorts", "humanRetrospective": "moderate — current procedure-specific harm cohort", "randomised": "none", "independentReplication": "moderate — short-term observation findings agree across ASPEN and PANDORA", "sampleSize": "ASPEN n=500, including 406 observed and 94 operated; PANDORA n=76 observed; 2026 surgical morbidity n=460", "effect": "No distant metastasis emerged during short prospective observation; ASPEN severe surgical complications 13%, current multicentre surgical major morbidity 14.8%", "limits": [ "Non-random treatment assignment", "Median prospective follow-up only 17-25 months", "Rare metastatic events", "Pathological risk is available mainly after surgery", "Procedure choice changes harm", "No mortality or quality-adjusted net-benefit comparison" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Small well-differentiated tumours often grow slowly, while duct change, grade and spread mark a more dangerous state." }, "animalEvidence": { "rating": "not-applicable", "reason": "The clinical surgery decision requires human natural-history and harm data." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Laboratory behaviour cannot set a pancreatic operation threshold." }, "reproducibility": { "rating": "moderate", "reason": "Two prospective cohorts agree on short-term safety, but neither resolves lifetime metastatic risk." }, "commercialConflicts": { "rating": "low", "reason": "The main studies are investigator-led, though specialist-centre preferences can still bias treatment selection." }, "recency": { "rating": "current", "reason": "The conclusion includes a 2026 procedure-specific morbidity report and a live registry check." } } }, "updated": "2026-09-14" }, { "id": "claim-smart-delivery-and-safety-without-added-survival-proof", "title": "MRI-guided adaptive radiation can deliver five high-dose treatments, but added survival is unproved", "shortTitle": "SMART delivery, no survival proof", "type": "claim", "status": "supported", "scope": [ "pdac", "locally-advanced", "borderline-resectable", "radiation" ], "summary": "A 136-person phase 2 study supports precise short-course treatment in a selected group; without chemotherapy-only control it cannot assign survival or surgery gains to radiation.", "content": "Stereotactic MRI-guided adaptive radiation therapy, called SMART, images the tumour and nearby bowel while a patient is on the treatment table. The team redraws structures and recalculates the plan before each dose. That can protect the stomach and bowel while delivering 50 Gy in five fractions; Gy is the unit of absorbed radiation energy. The largest prospective phase 2 study enrolled 136 people only after at least three months of chemotherapy without distant progression and with CA19-9 at most 500 U/mL. This is a strong selection filter: people whose disease spread early never entered. The initial safety report found no acute grade 3 or worse gastrointestinal toxicity definitely attributed to SMART. With longer follow-up, late grade 3 or worse toxicity was 0% when labelled definitely related, 4.6% when probably related and 11.5% when possibly related. Attribution changes the number and must not be hidden. Two-year survival was 53.6% from diagnosis and 40.5% from SMART; 34.6% later underwent surgery. There was no chemotherapy-only arm, so these outcomes combine induction response, selection, radiation, surgery and later treatment. CROSSFIRE randomized SMART against irreversible electroporation, not against continued chemotherapy, and found no survival difference between local techniques. SMART has passed a delivery-and-safety gate for selected patients. It has not proved added overall or quality-adjusted survival over current systemic treatment.", "sourceIds": [ "parikh-2023-smart-phase2-safety", "chuong-2024-smart-phase2-survival", "clinicaltrials-nct03621644-smart", "timmer-2024-crossfire" ], "links": [ { "target": "failure-smart-trial-and-vendor-dependency", "relation": "bounded-by" }, { "target": "unknown-smart-added-survival-and-surgery-harm", "relation": "leaves-open" }, { "target": "hypothesis-smart-vendor-independent-randomized-gate", "relation": "motivates" }, { "target": "trial-pdac-smart-action-map", "relation": "mapped-by" }, { "target": "failure-local-intensification-pdac-survival-not-shown", "relation": "extends" }, { "target": "treatment-modality-ranking", "relation": "updates" }, { "target": "treatment", "relation": "part-of" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "weak", "independentReplication": "moderate", "sampleSize": "SMART phase 2 n=136 single arm; CROSSFIRE n=68 randomized between two local techniques", "effect": "Five-fraction delivery and selected-cohort safety established; added survival over chemotherapy unknown", "limits": [ "No chemotherapy-only control in phase 2", "Entry required survival without distant progression through induction treatment", "CA19-9 selection", "Open-label", "Post-SMART surgery and later treatment varied", "Late severe-toxicity rate depends on attribution category", "Registry and paper enrollment counts differ", "CROSSFIRE compared two local treatments" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Daily imaging and replanning directly address moving tumour and bowel geometry." }, "animalEvidence": { "rating": "not-applicable", "reason": "The delivery claim is measured directly in people." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Cell killing does not identify the added clinical value of image guidance." }, "reproducibility": { "rating": "moderate", "reason": "Several prospective centres reproduce delivery and local control, not added survival." }, "commercialConflicts": { "rating": "high", "reason": "ViewRay sponsored the pivotal single-arm study and supplied its named MRIdian system." }, "recency": { "rating": "current", "reason": "Mature phase 2 and current trial-status records were checked through September 2026." } } }, "layout": { "x": 676, "y": 600 }, "updated": "2026-09-15" }, { "id": "claim-sonochemotherapy-randomized-signal", "title": "The first randomized sonoporation survival signal did not reproduce overall", "shortTitle": "Sonoporation replication", "type": "claim", "status": "mixed", "scope": [ "pdac" ], "summary": "A 78-person Chinese trial reported longer survival, but an accepted abstract from a second randomized trial reports no overall progression or survival benefit among 97 enrolled people.", "content": "Sonoporation is a delivery method: ultrasound pressure makes injected gas microbubbles oscillate and can briefly increase the permeability of nearby blood vessels and cells. It is a transport layer for chemotherapy, not a separate cancer-killing payload. The first uncontrolled feasibility study treated ten people and compared them with 63 historical controls; its 17.6 versus 8.9-month survival difference could be caused by selection, calendar time or care differences. A Chinese multicentre open-label trial then randomized 78 people with stage III or IV PDAC. Adding 30 minutes of ultrasound and microbubbles produced median overall survival of 9.1 versus 6.1 months, P=0.037, while progression-free survival was 5.5 versus 3.5 months, P=0.080. That justified replication. NCT04821284 is the independent randomized test with current gemcitabine/nab-paclitaxel or FOLFIRINOX backbones. Its public registry remains active not recruiting with estimated enrollment 120 and no posted results after an actual October 2025 primary completion. An accepted 2026 meeting abstract reports 97 enrolled and no significant overall PFS or OS difference, both P>0.36. It also reports median OS of 407.5 versus 234 days in the gemcitabine/nab-paclitaxel subgroup, a 173.5-day difference, but gives no subgroup denominator, confidence interval, interaction test, baseline balance, prespecification or adjustment for multiple searches. That subgroup cannot reverse the overall result. The honest state is failed overall replication with one incomplete subgroup, not a validated treatment and not proof that physical delivery never works.", "sourceIds": [ "dimcevski-2016-sonoporation-phase1", "wang-2023-sonochemotherapy-pdac", "clinicaltrials-nct04821284-sonoporation", "ius-2026-sonoporation-nct04821284-abstract", "adler-2026-sonoporation-perfusion" ], "links": [ { "target": "treatment", "relation": "part-of" }, { "target": "failure-local-intensification-pdac-survival-not-shown", "relation": "contradicts" }, { "target": "claim-panova3-ttfields-small-os-benefit", "relation": "compares-with" }, { "target": "treatment-long-tail-audit", "relation": "ranked-by" }, { "target": "failure-sonoporation-overall-replication-subgroup-rescue", "relation": "creates" }, { "target": "trial-pdac-sonoporation-action-map", "relation": "mapped-by" }, { "target": "unknown-pdac-sonoporation-treatment-effect", "relation": "leaves" }, { "target": "hypothesis-measured-fixed-backbone-sonoporation", "relation": "constrains" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "not-applicable", "randomised": "moderate", "independentReplication": "weak — negative overall result available only as an accepted abstract", "sampleSize": "First randomized trial n=78; second trial abstract n=97 enrolled against registry estimate n=120", "effect": "First trial OS 9.1 versus 6.1 months, P=0.037; second trial overall PFS and OS both P>0.36; incomplete gemcitabine/nab-paclitaxel subgroup OS 407.5 versus 234 days", "limits": [ "Both trials open-label", "Mixed disease stages", "Second trial uses two chemotherapy backbones", "Second result is an abstract without participant flow, hazard ratios or uncertainty", "Subgroup denominator and interaction are missing", "Perfusion is not direct tumour drug exposure" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "Ultrasound-driven microbubble motion can change local permeability, but this does not guarantee more active drug reaches every tumour cell." }, "animalEvidence": { "rating": "moderate", "reason": "Model work supports delivery, but the branch is now governed by two randomized human trials." }, "inVitroEvidence": { "rating": "weak", "reason": "A physical delivery effect depends on tissue, perfusion and acoustic geometry that simplified cultures do not reproduce." }, "reproducibility": { "rating": "weak", "reason": "The survival benefit did not reproduce in the overall second-trial abstract." }, "commercialConflicts": { "rating": "moderate", "reason": "The intervention uses commercial ultrasound and contrast-agent systems; full sponsor and device ties require the final paper." }, "recency": { "rating": "current", "reason": "Includes the 2026 accepted abstract and the latest public registry state." } } }, "layout": { "x": 134, "y": 92 }, "updated": "2026-09-15" }, { "id": "claim-stage-shift-insufficient", "title": "Earlier diagnosis alone does not meet the mission", "shortTitle": "Stage shift limit", "type": "claim", "status": "supported", "scope": [ "pancreatic-all" ], "summary": "Even the current localised-stage five-year survival of 43.6% remains far below reliable cure or durable control.", "content": "A deliberately simple ceiling calculation makes the gap visible. Overall five-year relative survival is 13.7%. If every patient somehow received the current aggregate localised-stage outcome, it would be 43.6%, a 3.18-fold increase: 43.6 ÷ 13.7 = 3.18. That would be a major gain but still leaves a 56.4-point deficit from 100% five-year relative survival. This is not a causal forecast; it shows that detection must be paired with better eradication of local and occult systemic disease.", "sourceIds": [ "seer-pancreas-statfacts" ], "links": [ { "target": "detection", "relation": "supports" }, { "target": "treatment", "relation": "supports" }, { "target": "residual-disease", "relation": "supports" }, { "target": "bottlenecks", "relation": "supports" } ], "evidence": { "humanProspective": "none", "humanRetrospective": "strong", "randomised": "not-applicable", "independentReplication": "not-applicable", "sampleSize": "Population registry, SEER 21", "effect": "Illustrative ceiling: 43.6% versus 13.7%, ratio 3.18", "limits": [ "Not a causal stage-shift estimate", "Mixed histologies", "Five-year survival is not cure" ] }, "layout": { "x": 39, "y": 61 }, "updated": "2026-09-14" }, { "id": "claim-staging-laparoscopy-avoids-incision-not-proven-survival", "title": "Staging laparoscopy can avoid a futile abdominal incision; survival benefit is not proved", "shortTitle": "Hidden-spread staging boundary", "type": "claim", "status": "supported", "scope": [ "pdac" ], "summary": "A camera check through small ports finds scan-hidden PDAC spread and can stop an open operation, but existing comparisons do not prove longer survival.", "content": "Staging laparoscopy means inspecting the abdomen through small camera ports before committing to the large incision needed for pancreatic resection. An occult metastasis is cancer spread that existed but was hidden on the scan. In the 2026 PREOPANC-2 analysis, 322 people reached surgical exploration after treatment. Hidden spread caused 90.5% of aborted operations. Starting with laparoscopy was associated with non-therapeutic laparotomy falling from 17.1% to 4.5%: 17.1 - 4.5 = 12.6 percentage points, or one avoided open exploration for about every eight people checked. The staging method was not randomized, so this is strong route evidence rather than a survival experiment. The earlier PREOPANC analysis found hidden spread in 13/133, 9.8%, at laparoscopy and 9/106, 8.5%, at direct open exploration; 76.9% versus 30.0% later received palliative chemotherapy, but route and treatment assignment were mixed. Mayo found visible spread or malignant wash cytology in 180/1,004, 18%. Among 762 separate procedures, nine, 1.2%, caused a complication and one, 0.1%, a major complication. A wash cytology result means tumour cells were seen under a microscope in saline recovered from the abdominal cavity. It is not the same evidence as a visible lesion confirmed by biopsy: 40/96 Mayo cytology-positive cases had no visible spread. The signal predicts poor prognosis, but no randomized study proves that denying or delaying resection solely for cytology improves useful life. Decision: starting a planned operation with direct inspection has a clear mechanical benefit when biopsy confirms spread; early separate staging, risk-based omission and cytology-only action remain open decisions. No survival claim should be made from detection yield alone.", "sourceIds": [ "gudmundsdottir-2023-staging-laparoscopy", "van-dongen-2023-preopanc-staging", "rompen-2026-preopanc2-staging", "theijse-2024-nontherapeutic-laparotomy", "kikuchi-2026-staging-laparoscopy", "nice-2018-ng85-pancreatic" ], "links": [ { "target": "treatment", "relation": "part-of" }, { "target": "claim-neoadjuvant-therapy-improves-selection-not-cure", "relation": "extends" }, { "target": "unknown-pdac-staging-laparoscopy-rule", "relation": "leaves-open" }, { "target": "failure-staging-yield-without-patient-utility", "relation": "bounded-by" }, { "target": "trial-pdac-occult-metastasis-action-map", "relation": "tested-by" }, { "target": "hypothesis-pdac-two-timepoint-staging-route", "relation": "motivates" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "strong", "randomised": "weak — parent trials randomized treatment, not staging method", "independentReplication": "moderate", "sampleSize": "PREOPANC-2 exploration n=322; PREOPANC n=239; Mayo n=1,004; TAPS n=663", "effect": "Associated non-therapeutic laparotomy 17.1% to 4.5%, absolute difference 12.6 points; Mayo positive yield 18%", "limits": [ "Staging not randomized", "Expert centres", "Gross spread and cytology often combined", "Selection before laparoscopy", "No randomized survival or quality-of-life effect", "Timing varies" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Direct inspection can reveal surface liver or abdominal spread below scan resolution before a large incision is made." }, "animalEvidence": { "rating": "not-applicable", "reason": "The claim concerns human staging and surgical route decisions." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "A laboratory model cannot measure avoided incision, delay or patient recovery." }, "reproducibility": { "rating": "moderate", "reason": "Several centres and trial cohorts find hidden spread, but yield and selection rules vary." }, "commercialConflicts": { "rating": "low", "reason": "The standard camera and pathology route is not tied to a single proprietary product." }, "recency": { "rating": "current", "reason": "The record includes a February 2026 PREOPANC-2 analysis and current England guidance." } } }, "updated": "2026-09-15" }, { "id": "claim-statistical-significance-can-hide-trivial-benefit", "title": "A statistically significant PDAC trial can still deliver a clinically small gain", "shortTitle": "0.33-month gain", "type": "claim", "status": "supported", "scope": [ "pancreatic-all" ], "summary": "Erlotinib plus gemcitabine reduced the death hazard in a 569-patient trial, but median survival improved by only 0.33 months.", "content": "The erlotinib plus gemcitabine phase 3 trial met its overall-survival endpoint: hazard ratio 0.82, 95% confidence interval 0.69 to 0.99, P=0.038. Median survival was 6.24 versus 5.91 months, an absolute difference of 0.33 months, about ten days. One-year survival was 23% versus 17%. The result is not false; the hazard ratio and tail difference contain information the median alone misses. But it is far below the mission target and illustrates why P values cannot rank treatments without absolute survival, duration, toxicity and reach. A programme should set a minimum worthwhile effect before running a pivotal trial, rather than treating any significant result as material progress.", "sourceIds": [ "moore-2007-erlotinib-gemcitabine" ], "links": [ { "target": "treatment", "relation": "supports" }, { "target": "failures", "relation": "supports" }, { "target": "bottlenecks", "relation": "supports" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong", "independentReplication": "moderate", "sampleSize": "569 randomised patients", "effect": "Median OS 6.24 versus 5.91 months; HR 0.82; P=0.038; one-year survival 23% versus 17%", "limits": [ "Advanced pancreatic cancer included locally advanced and metastatic disease", "Median difference does not describe the survival curve tail", "Older treatment era", "Clinical value also depends on toxicity and patient preference" ] }, "layout": { "x": 85, "y": 72 }, "updated": "2026-09-14" }, { "id": "claim-tamp-blood-exposure-not-tumour-delivery", "title": "TAMP changes gemcitabine in blood but has not proved greater active drug inside human tumours", "shortTitle": "Blood exposure, not tumour delivery", "type": "claim", "status": "open", "scope": [ "pdac", "trial-delivery", "transfer-unproven" ], "summary": "A 16-person randomized sub-study found less active gemcitabine in blood after artery delivery, but sampled neither tumour concentration nor the active drug forms inside cancer cells.", "content": "TAMP, or trans-arterial micro-perfusion, temporarily blocks a short artery segment with two balloons and pushes gemcitabine across the artery wall toward the pancreas. The engineering claim is plausible: raise local pressure, reduce washout and lower whole-body exposure. The human measurement remains one layer short of that claim. In 16 TIGeR-PaC participants across six sites, 11 received intra-arterial gemcitabine and five received it through a vein. Mean blood exposure to active parent gemcitabine, measured as area under the concentration-time curve, was 4.9 versus 8.8 hour-micrograms per millilitre, P=0.018. The inactive breakdown product dFdU was 46.3 versus 37.1, P=0.097. Lower active drug in peripheral blood supports reduced systemic exposure; higher inactive product was not statistically clear at the 0.05 rule. Among eight artery-treated people with raised baseline CA19-9, a blood marker released by many pancreatic cancers, dFdU exposure correlated with a two-week marker fall, r=-0.75 and P=0.034. The paper flags a visible outlier. This correlation cannot locate where conversion happened, and a marker fall is not survival. No tumour biopsy measured gemcitabine, the active intracellular forms dFdCDP or dFdCTP, or spatial distribution. The study therefore supports altered blood handling, not the stronger statement that more active drug reached cancer cells. Final survival and toxicity can still validate the treatment package without proving that proposed mechanism; extending the device to other drugs or cancers needs direct local measurement.", "sourceIds": [ "novelli-2026-tigerpac-pk", "clinicaltrials-nct03257033-tigerpac" ], "links": [ { "target": "treatment", "relation": "qualifies" }, { "target": "treatment-long-tail-audit", "relation": "expands" }, { "target": "failure-tigerpac-component-and-interim-inference", "relation": "explains" }, { "target": "unknown-tigerpac-package-survival-and-delivery", "relation": "leaves-open" }, { "target": "trial-pdac-intraarterial-delivery-action-map", "relation": "informs" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "not-applicable", "randomised": "moderate", "independentReplication": "weak", "sampleSize": "Randomized blood sub-study n=16: intra-arterial n=11 and intravenous n=5; marker correlation n=8", "effect": "Active-gemcitabine blood AUC 4.9 versus 8.8, P=0.018; no human tumour concentration or active intracellular metabolite measurement", "limits": [ "Very small groups", "Blood samples only", "No tumour drug measurement", "Marker analysis excludes three normal-baseline patients", "One reported outlier", "Sponsor-funded and company-linked authors" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Pressure-mediated artery delivery can change local exchange and blood escape." }, "animalEvidence": { "rating": "moderate", "reason": "A swine vessel study supports wall transport but is not a pancreatic tumour measurement." }, "inVitroEvidence": { "rating": "strong", "reason": "Gemcitabine activation and deactivation chemistry is well defined." }, "reproducibility": { "rating": "weak", "reason": "The direct human comparison contains 16 people and no independent tumour assay." }, "commercialConflicts": { "rating": "high", "reason": "RenovoRx funded the work and an employee-shareholder and paid consultant were authors." }, "recency": { "rating": "current", "reason": "The full peer-reviewed sub-study appeared in August 2026." } } }, "layout": { "x": 520, "y": 444 }, "updated": "2026-09-15" }, { "id": "claim-transmet-selected-liver-only-crc", "title": "Replacing the liver produced long survival in selected unresectable liver-only colorectal metastasis", "shortTitle": "TransMet survival", "type": "claim", "status": "supported", "scope": [ "colorectal" ], "summary": "In TransMet, five-year survival was 56.6% with transplantation plus chemotherapy and 12.6% with chemotherapy alone in 94 highly selected patients.", "content": "TransMet randomised 94 patients aged 18 to 65 with permanently unresectable, BRAF-non-mutated, chemotherapy-responsive colorectal metastases confined to the liver. Five-year overall survival was 56.6% with transplant plus chemotherapy and 12.6% with chemotherapy alone; hazard ratio 0.37. In the per-protocol population it was 73.3% versus 9.3%. This is unusually large randomised evidence that complete organ-level removal can change the course of metastatic solid cancer when disease remains geographically confined. It applies to a tiny selected subgroup and uses a scarce organ, lifelong immune suppression and major surgery; 80% of transplanted patients had a serious adverse event and three required another transplant. The transferable question is whether better tests can identify truly organ-confined metastatic biology and whether local replacement or ablation can be delivered without an organ donor.", "sourceIds": [ "adam-2024-transmet" ], "links": [ { "target": "colorectal-liver-metastasis", "relation": "part-of" }, { "target": "treatment", "relation": "supports" }, { "target": "transfer", "relation": "supports" }, { "target": "claim-mcrc-early-seeding", "relation": "succeeds-where" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong", "independentReplication": "weak", "sampleSize": "94 randomised; 74 per protocol", "effect": "Five-year OS 56.6% versus 12.6%; HR 0.37; per-protocol 73.3% versus 9.3%", "limits": [ "Highly selected liver-only disease", "Small trial", "Open-label", "Scarce donor organs", "Lifelong immunosuppression", "Serious adverse events common", "Does not transfer to disseminated PDAC" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Replacing the only involved organ can remove all visible disease when systemic spread is absent." }, "animalEvidence": { "rating": "not-applicable", "reason": "The survival claim comes from randomized human transplantation." }, "inVitroEvidence": { "rating": "weak", "reason": "Laboratory response cannot establish geographic confinement or transplant net benefit." }, "reproducibility": { "rating": "weak", "reason": "The exact transplant strategy has one small randomized trial." }, "commercialConflicts": { "rating": "low", "reason": "The dominant conflicts are organ scarcity, selection and procedure harm rather than product sales." }, "recency": { "rating": "recent", "reason": "TransMet was reported in 2024." } } }, "layout": { "x": 90, "y": 31 }, "updated": "2026-09-14" }, { "id": "claim-ukri-title-visible-pancreatic-capital", "title": "UKRI exposes at least £24.106m of title-visible pancreatic-cancer awards since 2015", "shortTitle": "UKRI title-visible floor", "type": "claim", "status": "supported", "scope": [ "pancreatic-all", "pdac", "pan-cancer" ], "summary": "A reproducible UKRI award query finds 74 explicitly titled projects since 2015, of which 43 have positive displayed commitments totalling £24,106,508.", "content": "This is a lower-bound ledger built from award records, not a national spending estimate. The UKRI Gateway to Research query for the exact phrase pancreatic cancer returned 147 projects. Restricting to starts from 1 January 2015 and titles that themselves contain pancreas or pancreatic plus cancer or tumour leaves 74 records. Forty-three show positive funded values totalling £24,106,508; 31 display zero and contribute nothing to the sum. Visible awards span direct treatment, stroma and immune biology, drug delivery, organoid models, blood tests, AI diagnosis, endoscopy and surgical imaging. The largest displayed project is £3.773m for preclinical development of an oncolytic vaccinia virus; other large records include £2.049m for microenvironment work, £1.540m for the STAR-PAC2 stromal-targeting trial, £1.522m for minimally invasive treatment technology and £1.310m for an early-detection endoscope. This filter intentionally misses grants titled around RAS, general cancer, shared infrastructure, colorectal, liver or biliary disease; the 31 zero-value records also mean the real title-visible commitment is higher. Conversely, multi-cancer grants contribute their full displayed value, so £24.106m is not a clean PDAC-only attribution. The useful conclusion is scale: an auditable UK public-award slice is tens of millions of pounds, not enough evidence for a global total and not comparable directly with the $2.0067bn US federal ledgers.", "sourceIds": [ "ukri-gtr-pancreatic-title-query-2026" ], "links": [ { "target": "investment", "relation": "part-of" }, { "target": "claim-observed-ten-year-research-capital", "relation": "extends" }, { "target": "treatment-long-tail-audit", "relation": "supports" }, { "target": "detection-modality-ranking", "relation": "supports" } ], "evidence": { "humanProspective": "not-applicable", "humanRetrospective": "not-applicable", "randomised": "not-applicable", "independentReplication": "moderate", "sampleSize": "147 retrieved records; 74 exact title-rule matches; 43 positive displayed award values", "effect": "£24,106,508 in displayed UKRI commitments for title-visible projects starting from 2015", "limits": [ "Title filter omits relevant broad grants", "Thirty-one matched records display zero funding", "Full multi-cancer award values are included", "Commitment is not annual expenditure", "Search index can change at later refreshes", "Not PDAC-only" ] }, "layout": { "x": 94, "y": 106 }, "updated": "2026-09-14" }, { "id": "claim-universal-germline-testing-needs-action", "title": "Germline testing finds an actionable minority only when results reach treatment and relatives", "shortTitle": "A test is not the action", "type": "claim", "status": "supported", "scope": [ "pdac", "pancreatic-all" ], "summary": "Family-history rules miss carriers, but a completed test is useful only if the result changes care for the patient or an at-risk relative.", "content": "The inherited subset is real and cannot be found reliably from family history alone. In a three-centre Spanish series, 14/179 newly diagnosed PDAC patients, 7.8%, had a pathogenic or likely pathogenic variant and 7/14 would not have met the study's clinical criteria. A larger Mayo case-control study found six established predisposition genes in 5.5% of 3,030 pancreatic-cancer cases, including 5.2% without a pancreatic-cancer family history. ASCO's 2024 panel guideline therefore treats exocrine pancreatic cancer as an all-patient germline-testing indication. The possible actions are narrower than the test: platinum choice or selected targeted treatment for some patients, gene-specific breast, ovarian, colorectal, endometrial, melanoma or prostate risk management for relatives, and specialist pancreatic surveillance for eligible high-risk carriers. Testing itself does not prove that any of those actions improves survival. Real services show that action can occur: a Johns Hopkins route changed chemotherapy in 28/992 patients seen, while a Dana-Farber cohort reported matched treatment in 13/26 pathogenic-variant carriers with incurable disease. Neither was a randomized treatment comparison. Variants of uncertain significance must not trigger treatment, surgery or family testing. The valid endpoint is the chain completed on time: offer, result, expert disclosure, patient decision, relative testing and a documented gene-specific action.", "sourceIds": [ "hu-2018-germline-risk", "llach-2024-universal-germline", "asco-2024-germline-panel-guideline", "asco-2019-pancreatic-susceptibility", "sinan-2025-pdac-germline-implementation", "chittenden-2021-systematic-germline", "chen-2023-fdr-cancer-risk" ], "links": [ { "target": "claim-germline-risk-subset", "relation": "extends" }, { "target": "treatment", "relation": "part-of" }, { "target": "prevention-ranking", "relation": "part-of" }, { "target": "claim-high-risk-surveillance", "relation": "routes-to" }, { "target": "failure-pdac-germline-route-dropout", "relation": "not-fully-delivered-by" }, { "target": "system-england-r367-germline-criteria", "relation": "contrasts-with" }, { "target": "unknown-pdac-germline-family-action-route", "relation": "leads-to" }, { "target": "trial-pdac-germline-cascade-action-map", "relation": "informs" }, { "target": "hypothesis-pdac-germline-family-route-completion", "relation": "supports" } ], "evidence": { "humanProspective": "moderate for testing delivery; none for survival", "humanRetrospective": "strong", "randomised": "none for test-versus-no-test clinical outcome", "independentReplication": "moderate", "sampleSize": "Risk series n=3,030 and n=179; delivery cohorts n=1,214 consultations and n=992 consultations", "effect": "Known high-risk variants in a minority; 7 of 14 Spanish carriers missed by clinical criteria; downstream care changes observed but not randomized", "limits": [ "Different panels and variant definitions", "Referral and ancestry selection", "No randomized survival effect", "Management availability differs by gene and country", "Family outcomes incompletely counted" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The implicated genes have established inherited cancer functions and can alter specific treatment or risk-management choices." }, "animalEvidence": { "rating": "not-applicable", "reason": "Human carrier detection and care delivery decide this claim." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Laboratory assays classify variants but cannot prove that the care route is completed." }, "reproducibility": { "rating": "moderate", "reason": "Several cohorts find an inherited minority and incomplete delivery, but yields and actions vary with panel, population and service." }, "commercialConflicts": { "rating": "moderate", "reason": "Some delivery studies use commercial panels or covered tests, while the main risk and guideline records are not product-performance studies." }, "recency": { "rating": "current", "reason": "The record includes 2024 guidance and 2025-2026 delivery evidence." } } }, "updated": "2026-09-14" }, { "id": "claim-vcn01-randomized-signal-not-survival-proof", "title": "VCN-01 reaches human tumours and has a randomized signal, not confirmed survival benefit", "shortTitle": "VCN-01 signal boundary", "type": "claim", "status": "mixed", "scope": [ "pdac" ], "summary": "Biopsies show virus replication and VIRAGE favoured VCN-01, but a 96-person sponsor analysis, conflicting alpha rules and no full all-enrolled result stop the claim short of proof.", "content": "Mechanism. VCN-01 is an engineered adenovirus: it is designed to copy itself preferentially in cancer cells whose RB1 growth-control circuit is broken, kill those cells, and make the PH20 enzyme that cuts hyaluronan, a gel-like part of the tumour barrier. This is a combined payload and delivery system, not the same product as pelareorep. Human mechanism. Intratumour treatment produced virus in later tumour biopsies, secondary blood peaks, rising PH20 and lower tumour stiffness. After intravenous treatment, viral genomes were found in five of six day-eight biopsies. That is direct evidence that the virus can cross the delivery boundary and replicate in some human tumours. It does not show survival benefit. Reach is also narrowed: 26% screened for the intravenous phase 1 study were excluded for high neutralising antibodies against adenovirus type 5. Randomized signal. VIRAGE registered 112 actual participants and has no posted results. The sponsor reported a 96-person full analysis set, 48 per arm: median survival was 10.8 versus 8.6 months, HR 0.57, 95% CI 0.34-0.96, log-rank P=0.0546; median progression-free survival was 7.0 versus 4.6 months, HR 0.55, P=0.0105. The direction is coherent, and the confidence interval excludes 1. Analysis boundary. The public protocol is internally inconsistent. Its sample-size section uses one-sided alpha 0.05, equivalent to two-sided 0.10; its main test is described as two-sided; and its ordered-testing section says OS must have P<0.05 before secondary endpoints can be tested. The sponsor says P=0.0546 met the main endpoint, while the official ESMO report gives adjusted Cox P=0.034 and log-rank P=0.055. The dated final analysis plan is not public. The protocol also calls only randomized-and-treated people its ITT set and applies further first-cycle dose rules to its FAS. Therefore the result is a strong phase 2 signal, not confirmed survival benefit. A full all-randomized analysis, mature events, quality of life and a double-blind phase 3 are required.", "sourceIds": [ "bazan-peregrino-2021-vcn01-intratumour", "garcia-carbonero-2022-vcn01-intravenous", "clinicaltrials-nct05673811-virage", "theriva-2023-virage-protocol", "theriva-2025-virage-topline", "esmo-2025-virage-2216mo" ], "links": [ { "target": "treatment", "relation": "part-of" }, { "target": "failure-oncolytic-virus-pelareorep", "relation": "differs-from" }, { "target": "failure-virage-analysis-and-survivor-selection", "relation": "bounded-by" }, { "target": "unknown-vcn01-all-enrolled-survival", "relation": "creates" }, { "target": "trial-vcn01-action-map", "relation": "mapped-by" }, { "target": "hypothesis-vcn01-replication-to-survival-gate", "relation": "supports" }, { "target": "company-theriva-biologics", "relation": "developed-by" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "not-applicable", "randomised": "moderate", "independentReplication": "weak", "sampleSize": "Intravenous phase 1 n=42 including 26 PDAC; biopsy n=6; VIRAGE registry n=112 and reported FAS n=96", "effect": "OS 10.8 vs 8.6 months, HR 0.57, 95% CI 0.34-0.96; log-rank P=0.0546; PFS 7.0 vs 4.6 months, HR 0.55, P=0.0105", "limits": [ "No posted registry results", "No peer-reviewed full VIRAGE paper", "Sixteen-person enrollment-to-FAS gap", "Conflicting public alpha and analysis-set rules", "Open label", "Liver-metastasis imbalance", "Sponsor-controlled data" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "Tumour biopsy and blood data show delivery and replication, while the link to durable control remains unproved." }, "animalEvidence": { "rating": "moderate", "reason": "Models support lysis and stromal disruption but cannot settle the human survival question." }, "inVitroEvidence": { "rating": "moderate", "reason": "Cell-selective replication and PH20 activity are engineered and measurable." }, "reproducibility": { "rating": "weak", "reason": "The randomized survival signal has not been independently repeated." }, "commercialConflicts": { "rating": "high", "reason": "The sponsor controls the product, trial, main analysis and current disclosures." }, "recency": { "rating": "current", "reason": "Uses the completed registry state, ESMO 2025 report and August 2026 filing." } } }, "layout": { "x": 170, "y": 90 }, "updated": "2026-09-15" }, { "id": "claim-zanidatamab-her2-biliary-selection", "title": "Zanidatamab shows why expression threshold matters in HER2-amplified biliary cancer", "shortTitle": "HER2 threshold matters", "type": "claim", "status": "supported", "scope": [ "cholangiocarcinoma", "gallbladder", "hepatobiliary" ], "summary": "Final single-arm data showed response 51.6% and median OS 18.1 months for IHC 3+, versus 5.6% and 5.2 months for IHC 2+; FDA approval is accelerated and confirmatory survival evidence remains due.", "content": "HERIZON-BTC-01 enrolled 80 previously treated ERBB2-amplified biliary cancers. At 33.4 months of follow-up, confirmed response was 41.3%, median response duration 14.9 months and median overall survival 15.5 months. The critical split was protein expression: among 62 IHC 3+ tumours, response was 51.6% and median survival 18.1 months; among 18 IHC 2+ tumours, one responded, 5.6%, and median survival was 5.2 months. The FDA accelerated approval is therefore restricted to IHC 3+ disease. This is a practical classification lesson: gene amplification alone did not define equal drug dependence. It is not proof of survival benefit because the trial had no control group and post-treatment survival can be influenced by selection and later therapy. A confirmatory randomized first-line trial must establish clinical benefit and report gallbladder, intrahepatic and extrahepatic sites separately.", "sourceIds": [ "pant-2026-zanidatamab-final", "fda-2024-zanidatamab-approval" ], "links": [ { "target": "claim-biliary-cancers-not-one-disease", "relation": "supports" }, { "target": "model-gallbladder-causal-chain", "relation": "supports" }, { "target": "model-cholangiocarcinoma-causal-chain", "relation": "supports" }, { "target": "treatment", "relation": "part-of" }, { "target": "companies", "relation": "part-of" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "not-applicable", "randomised": "none", "independentReplication": "weak", "sampleSize": "80 ERBB2-amplified patients; 62 IHC 3+ and 18 IHC 2+", "effect": "IHC 3+: ORR 51.6%, median OS 18.1 months. IHC 2+: ORR 5.6%, median OS 5.2 months", "limits": [ "Single-arm phase 2", "Accelerated approval", "Response-based approval", "No randomised survival effect", "Small IHC 2+ group", "Mixed biliary sites", "Company-funded", "Confirmatory trial pending" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "High HER2 protein expression marks greater dependence and matches the observed response split." }, "animalEvidence": { "rating": "moderate", "reason": "HER2 bispecific activity is supported in models but cannot prove survival benefit." }, "inVitroEvidence": { "rating": "strong", "reason": "The drug has direct HER2-binding and signalling evidence." }, "reproducibility": { "rating": "weak", "reason": "The decisive biliary result is single-arm and confirmatory survival evidence is pending." }, "commercialConflicts": { "rating": "high", "reason": "The drug, trial and accelerated-approval pathway are sponsor dependent." }, "recency": { "rating": "current", "reason": "Final follow-up and regulatory status were checked through September 2026." } } }, "layout": { "x": 73, "y": 74 }, "updated": "2026-09-14" }, { "id": "colorectal-liver-metastasis", "title": "Colorectal cancer with liver metastasis", "shortTitle": "CRC liver disease", "type": "cancer", "status": "supported", "scope": [ "colorectal" ], "summary": "A focused branch for liver-limited and liver-dominant metastatic colorectal cancer, including resection, ablation, infusion, transplant and systemic treatment.", "content": "Colorectal liver metastasis is not one state. Some people have a few technically removable liver lesions and no detected disease elsewhere; others have liver-dominant disease that is already broadly systemic. Anatomy, interval from the primary tumour, response to chemotherapy, molecular subtype and repeated whole-body staging all alter the probability that local treatment can matter.\n\nRandomized evidence supports a narrow route. In CLOCC, 119 people with fewer than ten unresectable liver metastases and no extrahepatic disease were assigned systemic treatment with or without radiofrequency ablation, with resection permitted. Eight-year survival was 35.9% with local treatment versus 8.9% without it. TransMet reported a larger transplant effect in an even smaller, intensely selected group with permanently unresectable liver-only disease. These results do not justify local treatment for unselected metastatic colorectal cancer.\n\nThe failure boundary is equally important. In EPOCH, adding yttrium-90 radioembolization in broader second-line liver disease improved hepatic progression by 1.9 months but not overall survival and increased severe adverse events. A local procedure changes survival only if occult extrahepatic disease is sufficiently controlled and procedure harm is smaller than the systemic benefit. This branch is useful to PDAC as a selection and trial-design example, not evidence that pancreatic liver metastases should be ablated or transplanted.", "sourceIds": [ "ruers-2017-clocc-crc-liver-ablation", "adam-2024-transmet", "mulcahy-2021-epoch-radioembolization", "hu-2019-crc-early-seeding", "tie-2022-dynamic-colon" ], "links": [ { "target": "colorectal", "relation": "part-of" }, { "target": "model-colorectal-causal-chain", "relation": "contains" }, { "target": "residual-disease", "relation": "part-of" }, { "target": "treatment", "relation": "part-of" }, { "target": "transfer", "relation": "supports" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "strong", "randomised": "strong", "independentReplication": "weak — mixed across different selection rules and local modalities", "sampleSize": "Randomized ablation and radioembolization studies plus a small randomized transplant trial and human phylogenetic work", "effect": "Long-survival benefit exists in tightly selected liver-only disease; broader liver-directed treatment can improve local progression without survival", "limits": [ "Selected groups are small", "Transplant consumes scarce organs", "Occult extrahepatic disease remains hard to exclude", "Results do not transfer directly to PDAC" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Complete organ control can matter when systemic spread is absent or controlled; early seeding makes that state uncommon." }, "animalEvidence": { "rating": "moderate", "reason": "Models support metastatic routes but cannot establish which human disease remains organ confined." }, "inVitroEvidence": { "rating": "weak", "reason": "Laboratory response does not measure occult extrahepatic disease or procedure harm." }, "reproducibility": { "rating": "moderate", "reason": "CLOCC and TransMet support the principle under different narrow rules, while EPOCH marks the broader failure boundary." }, "commercialConflicts": { "rating": "moderate", "reason": "Device interests affect radioembolization evidence; transplant and ablation results have different funding structures." }, "recency": { "rating": "recent", "reason": "The branch combines long follow-up with a 2024 transplant trial." } } }, "layout": { "x": 91, "y": 23 }, "updated": "2026-09-14" }, { "id": "colorectal", "title": "Colorectal cancer with liver involvement", "shortTitle": "Colorectal", "type": "cancer", "status": "supported", "scope": [ "colorectal" ], "summary": "Accessible precursors make prevention possible, but population uptake, residual-disease action and selection for liver-only cure remain limiting steps.", "content": "This page covers primary colon and rectal cancer, metastatic colorectal cancer, colorectal liver involvement and findings that may transfer to pancreatic or hepatobiliary disease. Population screening can remove accessible precursors, yet NordICC shows the difference between a procedure and an invitation: 42.0% attended, cancer incidence fell by 0.34 percentage points over 13 years and colorectal-cancer mortality did not fall significantly. MSI-high, BRAF V600E and HER2-positive metastatic disease require different systemic routes, while rectal anatomy changes local therapy. Postoperative ctDNA strongly separates risk, but randomized utility is action specific. Stage II treatment omission worked; stage III chemotherapy rules and trifluridine/tipiracil at molecular relapse did not; earlier CT in FIND raised curative-intent treatment but has no mature survival result. Liver-only metastasis creates a narrow route to long survival after ablation, resection or transplant. Transfer remains unproven until target-disease evidence exists.", "sourceIds": [ "mandel-2000-fobt-incidence", "atkin-2010-flexible-sigmoidoscopy", "kaminski-2026-nordicc-13y", "tie-2022-dynamic-colon", "tie-2025-dynamic3", "bando-2026-altair", "mo-2026-find-crc-surveillance", "andre-2025-keynote177-five-year", "tabernero-2021-beacon-crc", "sartore-bianchi-2016-heracles", "schrag-2023-prospect-rectal", "ruers-2017-clocc-crc-liver-ablation", "adam-2024-transmet" ], "links": [ { "target": "disease-system", "relation": "part-of" }, { "target": "model-colorectal-causal-chain", "relation": "contains" }, { "target": "trial-crc-action-map", "relation": "contains" }, { "target": "claim-crc-ctdna-utility-is-action-specific", "relation": "contains" }, { "target": "claim-crc-classification-produces-different-treatment-routes", "relation": "contains" }, { "target": "claim-rectal-anatomy-changes-local-treatment", "relation": "contains" }, { "target": "detection", "relation": "part-of" }, { "target": "transfer", "relation": "may-transfer-to" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "strong", "randomised": "strong", "independentReplication": "strong for screening and major treatment classes; mixed for rare routes", "sampleSize": "Population screening trials, large molecularly selected randomized trials and narrow liver-directed studies", "effect": "Screening prevents some cancers and deaths; classification and anatomy create several treatment routes, with cure limited to selected disease", "limits": [ "Colon and rectal cancer cannot always be pooled", "Rare molecular subsets do not cover most disease", "Liver transplant applies to very few people", "PDAC transfer is methodological, not biological proof", "Screening uptake limits population effect", "ctDNA utility differs by action" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Accessible precursors, molecular treatment groups and portal spread link measured state to distinct actions." }, "animalEvidence": { "rating": "moderate", "reason": "Models support routes but the branch conclusions depend on human screening and treatment studies." }, "inVitroEvidence": { "rating": "moderate", "reason": "Laboratory dependency work supports subgroups but cannot establish population or treatment utility." }, "reproducibility": { "rating": "strong", "reason": "Multiple randomized studies cover prevention, treatment routing, residual disease and local control." }, "commercialConflicts": { "rating": "moderate", "reason": "Commercial interests affect several drug and assay records but not all prevention evidence." }, "recency": { "rating": "current", "reason": "The branch includes randomized reports through 2026." } } }, "layout": { "x": 80, "y": 23 }, "updated": "2026-09-14" }, { "id": "companies", "title": "Companies and competitive landscape", "shortTitle": "Companies", "type": "company", "status": "supported", "scope": [ "pdac", "pnet", "cholangiocarcinoma", "gallbladder", "hcc", "colorectal", "pan-cancer" ], "summary": "Company claims, financing and partnerships are kept separate from independent clinical evidence.", "content": "Rank 1, proved human survival: Revolution Medicines is alone in PDAC. Daraxonrasib has randomized overall-survival proof and approval in previously treated RAS-G12 PDAC; durability and earlier use remain open. Rank 2, randomized signals requiring clean confirmation: Actuate has a peer-reviewed elraglusib survival result that persists in all randomized patients, but no matching PFS effect, full data audit or phase 3 funding; Theriva has direct tumour replication and a favourable VCN-01 phase 2 estimate, but incomplete participant accounting and no funded phase 3. Randomized answer pending: Astellas is testing setidegrasib first-line; Hengrui is testing HRS-4642 first-line; BioNTech and Roche/Genentech are testing an individual vaccine after resection. Rank 3, human signal without controlled contribution: Tango has the largest current selected combination signal and a $1.0bn balance, but no RAS-alone component control; Verastem has a small four-drug response signal; Alpha Tau has invasive local alpha delivery in a small uncontrolled study; PAQ and Jacobio have entered early human broad-KRAS programmes without PDAC efficacy results. Rank 4, useful measurement awaiting an action trial: Natera has large retrospective residual-disease separation; Valar has locked external pathology-AI validation; Cosomil has put a blood-enzyme workflow into a national feasibility study. Rank 5, commercial or early detection without utility: GRAIL has unmatched scale but missed the NHS-Galleri main stage endpoint; Exact Sciences has commercial multi-signal testing without pancreatic-specific utility; Mainz Biomed has early markers and severe financing limits. Negative-control companies are as important as active ones: Elicio missed its randomized disease-free-survival endpoint, Silexion's first implant did not show a significant randomized survival gain, and FibroGen ended pamrevlumab after two randomized failures. Related-cancer competition is now explicit. Jazz commercialises Zymeworks' HER2 bispecific after response-based biliary approval, but randomized survival confirmation is due. Relay designed selective FGFR2 inhibitor lirafugratinib and licensed it to Elevar; the 490-person study completed without posted registry results. Elevar also co-funded the positive CARES-310 HCC phase 3 result with Hengrui. Roche–Genentech's immune–vascular HCC evidence is separate from its PDAC vaccine role. Each detailed page separates founding and scientific origin where public, technology, target cancers, development stage, trial evidence, disclosed finance, partnerships, likely advantage and likely weakness. Missing or undisclosed facts stay missing. The competitive conclusion is narrow: broad RAS treatment is crowded and now clinically valid in PDAC; HER2 and FGFR2 selection are valid only in biliary subsets; HCC has several survival-positive immune–vascular combinations; degradation, residual-disease action, resistance routing and real-workflow detection are earlier; population screening sales remain ahead of proof.", "contentSections": [ "IGEA supplies the Cliniporator generator, electrodes and planning software across the pancreatic electrochemotherapy programme. Two protocol authors worked for IGEA and company funding covered part of the programme and trial insurance. Invest £0 in equity or equipment; any data-recovery contract requires raw pulse and plan export, independent analysis, negative-result publication, service and consumable pricing and permission to compare systems.", "ViewRay is a trial-continuity failure, not an investment case. Its MRIdian system delivered the 136-person SMART phase 2 and the company sponsored the dormant LAP-ABLATE phase 3 plan. At March 2023 it had $81.3m unrestricted cash, $77.8m debt and $53.6m quarterly operating cash use: $81.3m / $53.6m × 3 = 4.55 months of simple runway. It filed Chapter 11 in July. Invest £0 in equity or machinery; require service continuity, raw export, software escrow and trial-transfer rights from every future equipment route.", "Panbela Therapeutics is a data-custody case, not an equity case. Its last linked full quarter reported $142,000 cash, a $15.0m working-capital deficit and contractor termination on ASPIRE after unpaid invoices; a later filing disclosed $12.0m note default. Invest £0 in shares, licence or another trial. A capped £250,000 contract may buy verified assignment and data custody, paid only by deliverable.", "AstraZeneca adds an important disease-context control. Durvalumab combinations have mature randomized survival tails in mixed biliary cancer and HCC, while durvalumab with or without tremelimumab was almost inactive in unselected PDAC. Its licensed Jacobio pan-KRAS programme is a separate route with no posted PDAC efficacy. The same company therefore supplies both positive and negative evidence: execution scale is real, but neither an immune drug nor a company reputation transfers a treatment effect across cancers.", "Pfizer leads the cachexia route. Ponsegromab has a randomized weight, symptom and activity result and a recruiting 982-person PDAC trial, while Pfizer also controls the medicine and two relevant patent families. It ranks as a strong execution and access partner, not as a proved survival company: the completed trial lasted twelve weeks and the live trial keeps weight and anorexia as primary endpoints.", "Quercis adds a high-uncertainty oral clot-prevention route. Its 57-person sequential-dose CATIQ study changed D-dimer, PDI activity and other coagulation markers but had no control and no clinical event comparison. NCT06861088 now recruits an estimated 480 advanced-PDAC patients for Kinisoquin versus placebo. Two patent families cover an isoquercetin-vitamin composition and cancer-thrombosis use, while the exact formulation, licence chain and manufacturing controls are not public in the checked records. Quercis ranks as monitor and data-access work, not an investment recommendation; generic anticoagulants already have randomized clinical efficacy, so Kinisoquin must prove comparative event, bleeding, adherence and cost value.", "Boston Scientific adds mature device execution and a direct independence risk. Its WallFlex system was used in the 284-person metal-drainage versus surgery trial and the 119-person covered-versus-uncovered trial. The 2026 report states that the company helped design and execute the study, build and maintain the database, monitor data, analyse, interpret and write; employees were authors. That does not invalidate the result, but it requires independent adjudication, patient-level data, fixed device specifications, negative-result publication and alternative suppliers. The company ranks as a contracted device partner, not as a cancer-control or investment thesis.", "Novartis owns the active LuMIERE FAP-radioligand route. It has proved multicentre dose escalation and is recruiting a non-randomized estimated 222-person programme with pancreatic monotherapy and first-line combination groups. It has not proved added PDAC benefit, and the phase 1 paper does not publish tumour-lesion dose. Company value to the mission is therefore conditional data and drug access under independent-analysis and negative-publication terms, not a current investment conclusion. POINT/Lilly terminated FRONTIER for a stated business decision after 20 people and posts no result; do not relabel that as scientific failure.", "Silexion has moved SIL204 from a planned asset to an authorised but not yet outcome-producing programme. One Israeli site was cleared to screen in July 2026; CTIS lists German recruitment pending and 403 planned. The first-generation overall randomized result was negative, SIL204 has no reported human target result, public sample descriptions conflict and CTIS says participant-level data will not be shared. At 31 March the company had $2.413m cash after $3.646m first-quarter operating cash use; later financing can change that snapshot. Mission investment is £0.", "Theriva owns the strongest current engineered-virus route in PDAC. It has human tumour-replication evidence and a favourable randomized FAS estimate, but its 112-person registry record becomes 96 in the main report and the public protocol conflicts on alpha and analysis population. Cash was $9.6m in early August 2026; at the $5.541m six-month operating cash-use average, the simple runway is 9.6 ÷ (5.541 ÷ 6) = 10.4 months, while company guidance says Q1 2027 and explicitly excludes further VCN-01 trials beyond VIRAGE2. Rank it as monitor and data-access work, not an equity or phase 3 investment.", "Tango Therapeutics owns the leading MTAP-selected PRMT5 route. Vopimetostat monotherapy produced a 15% response rate in a mature 39-person pancreatic set; two RAS combinations produced 25 responses among 39 evaluable people. The company had $1.0bn of cash, equivalents and marketable securities at June 2026 after a large financing, so mission capital adds little. Rank it as a well-funded human-signal company that must prove component contribution, not as a survival-positive company.", "Actuate Therapeutics owns the strongest newly found phase 2 survival signal and one of the weakest balance sheets. Elraglusib improved all-randomized median survival from 7.2 to 8.9 months, HR 0.68, without significant PFS or response improvement. At 30 June cash was $4.426m against six-month operating use of $9.422m; 4.426/(9.422/6)=2.82 months under an unchanged-burn calculation, and no later company financing filing was found by 15 September. Rank it as urgent data-access and partner diligence, not an equity or phase 3 investment.", "Innovent Biologics owns IBI343 in Greater China and Takeda holds exclusive rights outside Greater China. The programme is not capital constrained: Innovent reported RMB30.2bn in cash and short-term financial assets at 31 July 2026, with first-half revenue of RMB8.6bn and profit of RMB1.3bn. The mission should therefore spend £0 on its 201-person phase 3 and instead require protocol, assay, power and complete outcome transparency.", "CARsgen Therapeutics owns satri-cel and has a working commercial autologous-cell manufacturing route. Its PDAC evidence remains 4/24 uncontrolled advanced responses and six short-followed adjuvant patients. At 30 June 2026 it reported RMB1.400bn cash and expects operations funded into 2030. The mission invests £0 and requests complete participant routes, marker maps, hospital time and a controlled PDAC design.", "Alligator Bioscience is not a conventional equity candidate. Cash fell to SEK16.6m and it stopped independent mitazalimab development in July 2026. The underlying 57-person CD40 signal remains uncontrolled and the registry counts 94 without final results. Mission value lies in a staged asset, data and trial-rights transaction: £0 before complete records; then at most £25m only with a solvent owner, drug supply, step-in rights, full publication and at least equal outside capital.", "Arcus Biosciences needs no mission subsidy. It reported $775m cash and investments at June 2026, has Gilead and Taiho programme partners, and completed PRISM-1 enrollment in September 2025. A crude unchanged operating-use estimate is 18.1 months, while the company formally guides funding into at least the second half of 2028. The decisive survival result is expected in the first half of 2027. Invest £0 in equity and £0 in the trial before that result; seek participant-level data access after lock.", "RenovoRx is a data-access case, not an equity or new-trial case. It reported $9.479m cash and $6.576m six-month operating cash use, or 8.65 months of simple unchanged-use coverage. Enrollment is closed and 78 of 86 deaths were reported in August 2026, so mission clinical capital cannot accelerate recruitment. Spend £0 on equity and a new cohort; offer at most £250,000 after lock for complete participant access, independent analysis and public negative-result rights.", "AngioDynamics owns NanoKnife and controls the unreported DIRECT phase 3 record. It reported $320.2m annual revenue, $53.9m cash, no debt and $3.1m operating cash provided; NanoKnife sales rose $8.6m. The business can fund its own product evidence. Invest £0 and fund no new cohort. Offer at most £250,000 only for complete participant-level access, independent analysis and public negative-result rights.", "Novocure owns Optune Pax and the first randomized pancreatic device result to improve survival: 16.2 versus 14.2 months in PANOVA-3. FDA approval is narrow and the evidence remains single-sponsor, burdensome and expensive. At June 2026 Novocure held $440.6m cash and short investments, had 285 active U.S. pancreatic users and recognized $1.6m of second-quarter Optune Pax revenue. It can fund replication and metastatic testing. Invest £0, provide £0 product subsidy and cap independent data and care-route work at £500,000 after access and publication rights.", "Adenocyte is a newly funded detection company, not a validated screen. iGan led a $10m first close in May 2026 for LINFU, which combines microbubbles, secretin and low-intensity ultrasound to enrich pancreatic-fluid cytology. The investor reports four unsuspected lesions among 90 high-risk people, but the two 500-person registries are invitation-only, observational and post no results. Invest £0; request the complete participant table, negative follow-up, central pathology, procedure harms and independent analysis." ], "sourceIds": [ "oreilly-2026-daraxonrasib", "park-2026-setidegrasib", "cui-2026-hrs4642-pdac", "nct05968326-imcode003", "sarno-2025-avatar", "sasieni-2026-nhs-galleri", "fibrogen-2024-pamrevlumab-winddown", "elicio-2026-amplify7p-results", "pant-2026-zanidatamab-final", "qin-2025-cares310-final", "subbiah-2023-rly4008", "oh-2025-topaz1-three-year", "rimassa-2025-himalaya-five-year", "groarke-2024-ponsegromab-cachexia", "clinicaltrials-nct06989437-ponsegromab-pdac", "pfizer-2026-ponsegromab-profile", "zwicker-2019-isoquercetin-catiq", "clinicaltrials-nct06861088-catiq-p3", "quercis-2026-kinisoquin-profile", "patent-us10391096-kinisoquin", "patent-wo2020112828-isoquercetin-cancer", "frere-2020-pdac-thromboprophylaxis-meta", "costamagna-2026-sems-early-surgery-rct", "seo-2019-covered-uncovered-sems-rct", "chen-2023-element-biliary-rct", "baum-2022-fap2286-firsthuman", "mcconathy-2026-lumiere-phase1", "clinicaltrials-nct04939610-lumiere", "baum-2026-3bp3940", "clinicaltrials-nct05432193-frontier", "clinicaltrials-nct07229768-ctr-fapi", "alagesan-2026-sig12d-loder", "ctis-2026-sil204-phase23", "silexion-2026-sil204-site-initiation", "silexion-2026-halfyear-filing", "bazan-peregrino-2021-vcn01-intratumour", "garcia-carbonero-2022-vcn01-intravenous", "clinicaltrials-nct05673811-virage", "theriva-2023-virage-protocol", "theriva-2025-virage-topline", "esmo-2025-virage-2216mo", "ctis-2026-virage2", "theriva-2026-q2-10q", "richards-2006-ci994-pdac", "heumann-2022-azacitidine-pdac", "sohal-2020-thu-decitabine-pdac", "baretti-2024-entinostat-nivolumab-pdac", "clinicaltrials-nct03250273-entinostat-nivolumab", "safyan-2026-azacitidine-pembrolizumab-pdac", "clinicaltrials-nct03264404-azacitidine-pembrolizumab", "rodon-2024-amg193-prmt5", "matsumoto-2026-mtap-pancreatic", "pavlick-2026-mtap-genomics", "clinicaltrials-nct05732831-vopimetostat", "tango-2025-vopimetostat-monotherapy", "clinicaltrials-nct06922591-vopimetostat-ras", "tango-2026-vopimetostat-ras-data", "tango-2026-q2-10q", "mahalingam-2026-elraglusib-randomized", "clinicaltrials-nct03678883-elraglusib", "pathak-2026-elraglusib-folfirinox", "clinicaltrials-nct05077800-elraglusib-folfirinox", "actuate-2026-q2-10q", "sec-2026-actuate-submissions", "innovent-2026-interim-results", "takeda-2025-ibi343-license", "clinicaltrials-nct07066098-ibi343-phase3", "qi-2024-ct041-pdac", "clinicaltrials-nct03874897-ct041", "clinicaltrials-nct04581473-ct041", "clinicaltrials-nct05911217-satricel-adjuvant", "carsgen-2025-satricel-pdac-adjuvant", "carsgen-2026-interim-results", "alligator-2026-q2", "clinicaltrials-nct04888312-optimize1", "van-laethem-2025-optimize1-biomarkers", "padron-2022-prince", "clinicaltrials-nct06608927-prism1", "arcus-2026-q2-10q", "wainberg-2026-quemliclustat-arc8", "clinicaltrials-nct05254171-aspire", "ctis-2024-514714-12-00-aspire", "panbela-2024-q3-10q", "panbela-2025-note-default", "sec-2025-panbela-registration-abandoned", "clinicaltrials-nct03257033-tigerpac", "hatoum-2024-tamp-rr1-rr2", "novelli-2026-tigerpac-pk", "renovorx-2023-tigerpac-interim", "renovorx-2025-10k", "renovorx-2026-q2-10q", "renovorx-2026-full-enrollment", "angiodynamics-2026-10k", "clinicaltrials-nct03899636-direct-rct", "clinicaltrials-nct03899649-direct-registry", "martin-2024-direct-registry-safety", "martin-2026-direct-registry-survival-preprint", "timmer-2024-crossfire", "macarulla-2025-panova3", "clinicaltrials-nct03377491-panova3", "fda-2026-optune-pax-ssed", "clinicaltrials-nct06390059-panova4", "novocure-2026-panova4-topline", "novocure-2026-q2-10q", "guzauskas-2026-ttfields-cost", "liu-2026-ttfields-cost", "parikh-2023-smart-phase2-safety", "chuong-2024-smart-phase2-survival", "clinicaltrials-nct03621644-smart", "clinicaltrials-nct05585554-lap-ablate", "viewray-2023-q1-10q", "viewray-2023-chapter11-8k", "izzo-2021-pancreatic-ect-phase12", "izzo-2021-laparoscopic-ect-protocol", "eudract-2018-003925-27-ect", "igea-2026-cliniporator-vitae", "igan-2026-adenocyte-financing", "clinicaltrials-nct04793633-linfu", "clinicaltrials-nct06276764-linfu-ipmn" ], "links": [ { "target": "investment", "relation": "part-of" }, { "target": "company-revolution-medicines", "relation": "contains" }, { "target": "company-astellas", "relation": "contains" }, { "target": "company-hengrui", "relation": "contains" }, { "target": "company-biontech", "relation": "contains" }, { "target": "company-roche-genentech", "relation": "contains" }, { "target": "company-verastem", "relation": "contains" }, { "target": "company-alpha-tau", "relation": "contains" }, { "target": "company-paq-therapeutics", "relation": "contains" }, { "target": "company-jacobio", "relation": "contains" }, { "target": "company-natera", "relation": "contains" }, { "target": "company-valar-labs", "relation": "contains" }, { "target": "company-cosomil", "relation": "contains" }, { "target": "company-grail", "relation": "contains" }, { "target": "company-exact-sciences", "relation": "contains" }, { "target": "company-mainz-biomed", "relation": "contains" }, { "target": "company-elicio-therapeutics", "relation": "contains" }, { "target": "company-silexion-therapeutics", "relation": "contains" }, { "target": "company-fibrogen", "relation": "contains" }, { "target": "company-jazz-pharmaceuticals", "relation": "contains" }, { "target": "company-relay-therapeutics", "relation": "contains" }, { "target": "company-elevar-therapeutics", "relation": "contains" }, { "target": "company-astrazeneca", "relation": "contains" }, { "target": "company-pfizer-cachexia", "relation": "contains" }, { "target": "treatment", "relation": "supports" }, { "target": "detection", "relation": "supports" }, { "target": "failures", "relation": "supports" }, { "target": "company-quercis-pharma", "relation": "contains" }, { "target": "company-boston-scientific-biliary", "relation": "contains" }, { "target": "company-novartis-fap-radioligand", "relation": "contains" }, { "target": "trial-pdac-fap-radioligand-action-map", "relation": "supports" }, { "target": "trial-sil204-action-map", "relation": "bounded-by" }, { "target": "claim-vcn01-randomized-signal-not-survival-proof", "relation": "bounded-by" }, { "target": "failure-virage-analysis-and-survivor-selection", "relation": "learns-from" }, { "target": "unknown-vcn01-all-enrolled-survival", "relation": "leaves-open" }, { "target": "trial-vcn01-action-map", "relation": "implements" }, { "target": "hypothesis-vcn01-replication-to-survival-gate", "relation": "tests" }, { "target": "company-theriva-biologics", "relation": "contains" }, { "target": "failure-broad-epigenetic-treatment-pdac", "relation": "learns-from" }, { "target": "claim-mtap-prmt5-selected-human-signal", "relation": "bounded-by" }, { "target": "failure-vopimetostat-ras-contribution-unknown", "relation": "learns-from" }, { "target": "unknown-mtap-prmt5-ras-contribution", "relation": "leaves-open" }, { "target": "trial-mtap-prmt5-action-map", "relation": "implements" }, { "target": "hypothesis-mtap-prmt5-ras-deeper-control", "relation": "tests" }, { "target": "company-tango-therapeutics", "relation": "contains" }, { "target": "claim-elraglusib-randomized-survival-signal", "relation": "bounded-by" }, { "target": "failure-elraglusib-analysis-and-survival-discordance", "relation": "learns-from" }, { "target": "unknown-elraglusib-survival-replication", "relation": "leaves-open" }, { "target": "trial-elraglusib-action-map", "relation": "implements" }, { "target": "hypothesis-elraglusib-survival-confirmation-gate", "relation": "tests" }, { "target": "company-actuate-therapeutics", "relation": "contains" }, { "target": "company-innovent-biologics", "relation": "contains" }, { "target": "company-carsgen-therapeutics", "relation": "contains" }, { "target": "company-alligator-bioscience", "relation": "contains" }, { "target": "company-arcus-biosciences", "relation": "contains" }, { "target": "trial-pdac-adenosine-action-map", "relation": "funds" }, { "target": "company-panbela-therapeutics", "relation": "contains" }, { "target": "trial-pdac-polyamine-action-map", "relation": "constrains" }, { "target": "company-renovorx", "relation": "contains" }, { "target": "trial-pdac-intraarterial-delivery-action-map", "relation": "funds" }, { "target": "hypothesis-tigerpac-existing-final-gate", "relation": "constrains" }, { "target": "company-angiodynamics", "relation": "contains" }, { "target": "trial-pdac-irreversible-electroporation-action-map", "relation": "funds" }, { "target": "hypothesis-direct-randomized-data-recovery-gate", "relation": "constrains" }, { "target": "company-novocure", "relation": "contains" }, { "target": "trial-pdac-ttfields-action-map", "relation": "funds" }, { "target": "hypothesis-optune-pax-independent-value-gate", "relation": "constrains" }, { "target": "company-viewray", "relation": "contains" }, { "target": "trial-pdac-smart-action-map", "relation": "failed-to-fund" }, { "target": "hypothesis-smart-vendor-independent-randomized-gate", "relation": "constrains" }, { "target": "company-igea", "relation": "contains" }, { "target": "trial-pdac-electrochemotherapy-action-map", "relation": "supplies" }, { "target": "hypothesis-electrochemotherapy-randomized-data-gate", "relation": "constrains" }, { "target": "company-adenocyte", "relation": "contains" }, { "target": "trial-linfu-action-map", "relation": "constrains" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "strong", "randomised": "strong", "independentReplication": "moderate", "sampleSize": "Forty detailed company records spanning approved treatment, randomized trials, early human work, diagnostics, supportive treatment and failed programmes", "effect": "Competitive ranking by strongest human evidence; no pooled treatment effect", "limits": [ "Many company disclosures are unaudited programme announcements", "Private funding and valuations are often undisclosed", "Company-wide R&D cannot be assigned cleanly to one disease", "Trial sponsors control much source data", "Rankings will change when current randomized trials report" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "The ranking follows human treatment and measurement evidence, while several company mechanisms remain unproved." }, "animalEvidence": { "rating": "not-applicable", "reason": "Animal activity is not sufficient to place a company in the main clinical ranking." }, "inVitroEvidence": { "rating": "weak", "reason": "Laboratory evidence informs early companies but carries little weight against patient outcomes." }, "reproducibility": { "rating": "moderate", "reason": "Several treatment classes repeat, while many company-specific claims remain single-sponsor results." }, "commercialConflicts": { "rating": "high", "reason": "Most source data, finance and pipeline status are reported or controlled by the ranked companies." }, "recency": { "rating": "current", "reason": "The landscape includes company and clinical evidence checked through September 2026." } } }, "layout": { "x": 77, "y": 105 }, "updated": "2026-09-15" }, { "id": "company-actuate-therapeutics", "title": "Actuate Therapeutics — randomized survival signal, weak balance sheet", "shortTitle": "Actuate Therapeutics", "type": "company", "status": "mixed", "scope": [ "pdac", "pan-cancer" ], "summary": "Actuate owns a credible phase 2 PDAC survival signal but cannot finance clean confirmation from its disclosed balance sheet.", "content": "Product. Actuate develops elraglusib, an intravenous GSK-3β inhibitor, and is also developing an oral form. The product has one peer-reviewed randomized phase 2 PDAC survival signal. In the all-randomized weekly-versus-control set, OS was 8.9 versus 7.2 months, HR 0.68. The larger 10.1-versus-7.2-month headline uses only people who began treatment. PFS, response and disease control did not significantly improve. A separate 49-person FOLFIRINOX combination study was too small and explicitly non-comparator; its authors said outcomes did not exceed historical standards. Evidence assets. Actuate controls the NCT03678883 participant data, drug supply and phase 3 route. The paper promises access to qualified researchers, while the registry says no participant-data sharing. That contradiction is a direct diligence test. Finance. At 30 June 2026 the company had $4.426m cash and $0.943m working capital. Six-month operating cash use was $9.422m, or $1.570m a month. The simple unchanged-burn ratio is 4.426/1.570=2.82 months; it is not company guidance. Management said cash would not satisfy operations beyond September without new capital and reported substantial going-concern doubt. SEC submissions checked through 15 September contain no later company financing filing. Nasdaq also gave the company until 11 January 2027 to restore a $50m listed-securities value; the notice had no immediate listing effect. Business position. Scientific upside and financing risk are coupled: a weak balance sheet may force dilution, licensing or trial delay, while urgency can also pressure the evidence threshold. Mission decision. Invest £0 in the company and £0 in phase 3 now. Request data under the paper's stated process and run an independent Stage 0 audit. If the all-randomized result survives, require a solvent strategic partner to fund manufacturing and most of a blinded phase 3. Any later mission contribution buys public data, negative-result publication and fixed stop rules; it does not rescue the company.", "sourceIds": [ "mahalingam-2026-elraglusib-randomized", "clinicaltrials-nct03678883-elraglusib", "pathak-2026-elraglusib-folfirinox", "clinicaltrials-nct05077800-elraglusib-folfirinox", "actuate-2026-q2-10q", "sec-2026-actuate-submissions" ], "links": [ { "target": "companies", "relation": "part-of" }, { "target": "investment", "relation": "part-of" }, { "target": "claim-elraglusib-randomized-survival-signal", "relation": "develops" }, { "target": "failure-elraglusib-analysis-and-survival-discordance", "relation": "must-repair" }, { "target": "unknown-elraglusib-survival-replication", "relation": "must-resolve" }, { "target": "trial-elraglusib-action-map", "relation": "controls-data-for" }, { "target": "hypothesis-elraglusib-survival-confirmation-gate", "relation": "could-test" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "moderate", "independentReplication": "none", "sampleSize": "Randomized phase 2 relevant ITT n=269; separate non-comparator study n=49", "effect": "Promising ITT OS HR 0.68 with no significant PFS effect and severe-event excess of 10.2 points", "limits": [ "Finance snapshot changes", "Constant-burn ratio is not company guidance", "No later financing filing does not prove no private financing", "Participant data unavailable so far", "Single-product concentration", "Nasdaq compliance risk" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "The company owns a defined inhibitor with human survival evidence but an unresolved mechanism and output mismatch." }, "animalEvidence": { "rating": "moderate", "reason": "Model data support the programme but should not drive company value after a randomized trial." }, "inVitroEvidence": { "rating": "moderate", "reason": "Target activity is measurable, while clinical usefulness remains the main asset question." }, "reproducibility": { "rating": "weak", "reason": "No independent randomized trial confirms the survival gain." }, "commercialConflicts": { "rating": "high", "reason": "Actuate controls the sole product, trial data and development narrative." }, "recency": { "rating": "current", "reason": "Clinical, registry and finance records were checked through 15 September 2026." } } }, "layout": { "x": 208, "y": 126 }, "updated": "2026-09-15" }, { "id": "company-adenocyte", "title": "Adenocyte — funded pancreatic-fluid sampling before public utility evidence", "shortTitle": "Adenocyte", "type": "company", "status": "open", "scope": [ "pdac", "pancreatic-all" ], "summary": "A $10m first close funds a plausible sampling system, but its public accuracy claim is sponsor-controlled and two observational registries have no results.", "content": "Company and origin: Adenocyte states that it was founded in Israel in 2018 by Mark Rutenberg and developed LINFU from ultrasound, microbubble and cytology work at the Technion. Technology: intravenous microbubbles and secretin are combined with low-intensity non-focused ultrasound to increase release of duct cells into pancreatic fluid, followed by cytology and possible computer-assisted analysis. This aims to improve the specimen rather than infer cancer from blood. Stage and trials: NCT04793633 is an industry-sponsored, invitation-only prospective observational cohort targeting 500 people at elevated risk or with possible disease, with estimated completion in 2030 and no posted results. NCT06276764 targets 500 people with IPMN, is also invitation-only and observational, and estimates completion in 2034. Financing and investors: iGan Partners led the first close of a $10m round in May 2026. The investor announcement claims more than 150 procedures and four unsuspected neoplasias among the first 90 high-risk people, including one T1N0 cancer, with 100% specimen adequacy and specificity. That report does not expose the complete denominator, locked analysis, adverse events, pathology review, interval cancers or peer-reviewed paper, so it is commercial evidence. Patent position: US20240131366A1 describes cavitation control and sampling; one publication does not establish the full family, claim validity or freedom to operate. Likely advantage: it attacks a real bottleneck—too few abnormal cells in a pancreatic-fluid sample—and has a defined clinical procedure and two follow-up registries. Likely weakness: secretin, contrast microbubbles, ultrasound and upper endoscopy create a multi-step specialist route; cytology can still miss unsampled lesions; an observational registry cannot assign benefit from acting on a positive result; and the evidence controller is the company. Decision: monitor and request the full 90-person dataset, assay lock, complications, negative follow-up and central pathology. Invest £0 and do not include it in average-risk screening until a consecutive intended-use study shows net benefit at a fixed work-up capacity.", "sourceIds": [ "adenocyte-company-linfu", "igan-2026-adenocyte-financing", "clinicaltrials-nct04793633-linfu", "clinicaltrials-nct06276764-linfu-ipmn", "patent-us20240131366-linfu" ], "links": [ { "target": "companies", "relation": "part-of" }, { "target": "detection", "relation": "competes-with" }, { "target": "trial-linfu-action-map", "relation": "sponsors" }, { "target": "patent-landscape", "relation": "owns" }, { "target": "investor-igan-adenocyte", "relation": "funded-by" } ], "evidence": { "humanProspective": "weak — sponsor reports procedures but registries post no results", "humanRetrospective": "none", "randomised": "none", "independentReplication": "none", "sampleSize": "Sponsor claims >150 procedures and reports 90 high-risk people; two registries each estimate n=500", "effect": "Investor-reported four unsuspected neoplasias among 90; no peer-reviewed complete result or utility effect", "limits": [ "Performance claim is investor and company controlled", "No complete public participant table", "No posted registry results", "Invitation-only cohorts can select participants", "Multi-step specialist procedure", "Patent family and operating rights not established" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "Increasing duct-cell release could improve cytology yield, but lesion sampling and benign atypia remain limiting." }, "animalEvidence": { "rating": "weak", "reason": "The company reports porcine development, but the checked public record does not expose a complete independent study." }, "inVitroEvidence": { "rating": "weak", "reason": "Cavitation and sample-enrichment engineering are plausible, but public analytical validation is incomplete." }, "reproducibility": { "rating": "none", "reason": "No independent team or posted registry result reproduces the sponsor performance claim." }, "commercialConflicts": { "rating": "high", "reason": "Company and lead investor provide the available performance and financing claims, while the company sponsors both registries." }, "recency": { "rating": "current", "reason": "Financing and trial status were checked in September 2026." } } }, "layout": { "x": 150, "y": 94 }, "updated": "2026-09-15" }, { "id": "company-alligator-bioscience", "title": "Alligator Bioscience", "shortTitle": "Alligator Bioscience", "type": "company", "status": "active", "scope": [ "pdac", "pan-cancer" ], "summary": "Alligator stopped independent mitazalimab development with SEK16.6m cash; the asset is a conditional data-and-rights opportunity, not an equity investment.", "content": "Asset. Alligator owns mitazalimab, the CD40 agonist behind the uncontrolled OPTIMIZE-1 PDAC signal. The current registry counts 94 participants, while the mature published efficacy set remains 57 and no registry result is posted. Finance. At 30 June 2026 cash was SEK16.6m after first-half cash flow of minus SEK45.5m. On 23 July the company stopped independent mitazalimab development and phase 3 support, began reducing staff to the minimum needed for another asset, and sought to sell or license mitazalimab. It also proposed a SEK125.6m rights issue with only 47% covered by commitments and guarantees and took SEK19m bridge loans. This is not a funded developer. Decision. Buy no listed equity and do not fund company survival. The possible mission transaction is limited to mitazalimab data, samples, manufacturing material, supply and trial rights. Pay nothing before all 94 participant records and manufacturing files are inspected. If they pass, require a solvent partner to supply the drug and at least half of a roughly £50m controlled trial. Mission exposure can then rise in stages to £25m, with full publication, step-in rights and asset access if the owner fails. The business logic is option value: £1 of early diligence can prevent £25 of blind rescue capital, while ownership terms prevent public money from paying once for evidence and again for access.", "sourceIds": [ "alligator-2026-q2", "clinicaltrials-nct04888312-optimize1", "van-laethem-2024-optimize1", "van-laethem-2025-optimize1-biomarkers", "padron-2022-prince" ], "links": [ { "target": "companies", "relation": "part-of" }, { "target": "claim-cd40-signal-without-contribution-proof", "relation": "develops" }, { "target": "unknown-cd40-added-survival-pdac", "relation": "owns-data-for" }, { "target": "hypothesis-mitazalimab-controlled-survival-gate", "relation": "must-enable" }, { "target": "trial-immune-reprogramming-action-map", "relation": "monitored-by" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "not-applicable", "randomised": "weak", "independentReplication": "weak", "sampleSize": "Published efficacy n=57, registry n=94; cash SEK16.6m at 30 June 2026", "effect": "Distressed proprietary asset with an unresolved controlled survival question", "limits": [ "Issuer-authored finance", "Post-report cash can change", "Rights and liabilities are unknown", "Final participant data are not public", "Cost and partner terms are estimates" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The product has human activity and a measurable immune mechanism." }, "animalEvidence": { "rating": "strong", "reason": "Model work supports the product class but does not value the company." }, "inVitroEvidence": { "rating": "strong", "reason": "Product activity and manufacturing quality can be audited." }, "reproducibility": { "rating": "weak", "reason": "No controlled mitazalimab survival result exists." }, "commercialConflicts": { "rating": "high", "reason": "The seller needs financing and controls the asset records." }, "recency": { "rating": "current", "reason": "Finance and strategy were checked through August 2026." } } }, "layout": { "x": 310, "y": 254 }, "updated": "2026-09-15" }, { "id": "company-alpha-tau", "title": "Alpha Tau — intratumoral alpha radiation in a 30-patient pancreatic safety study", "shortTitle": "Alpha Tau", "type": "company", "status": "open", "scope": [ "pdac", "pan-cancer" ], "summary": "An Israeli radiotherapy company has brought local alpha-emitter placement into pancreatic patients, but the active study cannot establish added survival benefit.", "content": "Founded in 2016 by Uzi Sofer around Alpha DaRT invented by Itzhak Kelson and Yona Keisari at Tel Aviv University. Technology: short-range alpha-emitting sources inserted into a tumour to generate diffusing radioactive atoms locally. Pancreatic stage: NCT06698458 is an open-label study of up to 30 newly diagnosed locally advanced or metastatic patients receiving Alpha DaRT during full-dose chemotherapy; safety is primary and survival is secondary. Advantage: high local energy with short physical range and an Israeli nuclear-medicine origin distinct from external radiation. Weakness: invasive placement, local reach in a systemic disease, no control arm, tiny sample and company-reported conference signals without a full paper. The decisive test would randomize only a local-dominant, systemically controlled group and measure survival and complications against identical systemic therapy.", "sourceIds": [ "nct06698458-alpha-dart-pdac" ], "links": [ { "target": "companies", "relation": "part-of" }, { "target": "failure-local-intensification-pdac-survival-not-shown", "relation": "must-overcome" }, { "target": "claim-panova3-ttfields-small-os-benefit", "relation": "shares-local-thesis-with" }, { "target": "hypothesis-selected-local-consolidation", "relation": "could-test" } ], "evidence": { "humanProspective": "none — planned study, no result", "humanRetrospective": "not-applicable", "randomised": "none", "independentReplication": "none", "sampleSize": "Target up to 30", "effect": "No controlled pancreatic efficacy result", "limits": [ "Open-label safety study", "Tiny cohort", "Local therapy in systemic disease", "Invasive implantation", "Company sponsored", "No mature peer-reviewed pancreatic results" ] }, "layout": { "x": 113, "y": 116 }, "updated": "2026-09-14" }, { "id": "company-angiodynamics", "title": "AngioDynamics", "shortTitle": "AngioDynamics", "type": "company", "status": "contested", "scope": [ "pdac", "local-control", "device-owner" ], "summary": "AngioDynamics owns NanoKnife, funds the evidence and has ample operating capacity; mission value lies in independent access to its randomized record, not subsidy or equity.", "content": "AngioDynamics owns and sells the NanoKnife system and sponsors both DIRECT studies. The evidence and business incentives therefore share one controller. In the 2024 registry paper, the company funded the study and participated in design, monitoring, data collection, statistics, interpretation and manuscript work; authors disclosed consulting, travel, speaking or research ties. The 2026 survival preprint says the company supplied financial and logistical support while academic investigators controlled interpretation and submission. Both statements should remain visible because governance changed or is described differently. Financial capacity is not the bottleneck. For the year to 31 May 2026, the company reported $320.2m revenue, $36.7m net loss, $3.1m cash provided by operations, $53.9m cash and no outstanding debt. NanoKnife sales rose by $8.6m and the Med Tech segment containing it grew 18.4%. In business terms, the product already has a funded owner, a growing revenue line and a missing proof asset. Mission money should buy decision-quality access, not company risk. Invest £0 in equity and fund £0 of a new device cohort. Offer at most £250,000 from the existing data and statistics line only for the complete NCT03899636 assignment and outcome record, original and changed plans, independent analysis and public negative-result rights. If access is refused, treat that as adverse diligence and allocate nothing.", "sourceIds": [ "angiodynamics-2026-10k", "clinicaltrials-nct03899636-direct-rct", "clinicaltrials-nct03899649-direct-registry", "martin-2024-direct-registry-safety", "martin-2026-direct-registry-survival-preprint", "timmer-2024-crossfire" ], "links": [ { "target": "companies", "relation": "part-of" }, { "target": "trial-pdac-irreversible-electroporation-action-map", "relation": "sponsors" }, { "target": "failure-direct-selection-and-randomized-reporting-gap", "relation": "controls" }, { "target": "hypothesis-direct-randomized-data-recovery-gate", "relation": "constrained-by" }, { "target": "investment", "relation": "constrains" } ], "evidence": { "humanProspective": "not-applicable", "humanRetrospective": "not-applicable", "randomised": "not-applicable", "independentReplication": "strong", "sampleSize": "$320.2m revenue; $53.9m cash; $3.1m operating cash provided; NanoKnife sales up $8.6m", "effect": "Equity and new-cohort allocation £0; conditional independent data recovery capped at £250,000", "limits": [ "Net loss and cash flow use different accounting measures", "Product-specific NanoKnife revenue is not disclosed", "Phase 3 enrollment is estimated", "Current result timing is not disclosed", "This is programme allocation, not personal investment advice" ], "dimensions": { "biologicalPlausibility": { "rating": "not-applicable", "reason": "Linked pages assess device biology and patient evidence." }, "animalEvidence": { "rating": "not-applicable", "reason": "This page assesses ownership, capacity and evidence control." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "This page assesses ownership, capacity and evidence control." }, "reproducibility": { "rating": "strong", "reason": "Trial registries, papers and the SEC filing identify the same product owner and evidence route." }, "commercialConflicts": { "rating": "high", "reason": "AngioDynamics owns NanoKnife, funds the studies and controls the unreported randomized record." }, "recency": { "rating": "current", "reason": "The capital decision uses the July 2026 annual filing and September 2026 trial state." } } }, "layout": { "x": 586, "y": 510 }, "updated": "2026-09-15" }, { "id": "company-arcus-biosciences", "title": "Arcus Biosciences", "shortTitle": "Arcus", "type": "company", "status": "active", "scope": [ "pdac", "pan-cancer", "transfer-unproven" ], "summary": "Arcus owns and funds PRISM-1 with Gilead and Taiho rights; its cash position removes any case for mission subsidy before the 2027 result.", "content": "Arcus develops quemliclustat, the small-molecule CD73 inhibitor in the 610-person PRISM-1 phase 3 trial. Gilead licenses quemliclustat and shares U.S. promotion rights if approved; Taiho holds rights in parts of Asia and participates in the global study. Arcus reports that enrollment completed in September 2025 and results are expected in the first half of 2027. At 30 June 2026 it held $775m in cash, cash equivalents and marketable securities, used $257m in operating cash during six months and said its plan is funded into at least the second half of 2028. A crude unchanged-use calculation is $257m / 6 = $42.8m per month and $775m / $42.8m = 18.1 months. That deliberately ignores future revenue, partner reimbursements, programme wind-down, debt and changed spending; the company's formal runway estimate is therefore the better planning source. The programme is financed and the decisive trial is complete in enrollment. Invest £0 in Arcus equity and £0 in PRISM-1 before the masked result. The information asset worth securing is post-lock participant-level access, not exposure to a binary share-price event.", "sourceIds": [ "clinicaltrials-nct06608927-prism1", "arcus-2026-q2-10q", "wainberg-2026-quemliclustat-arc8" ], "links": [ { "target": "companies", "relation": "part-of" }, { "target": "trial-pdac-adenosine-action-map", "relation": "funds" }, { "target": "hypothesis-quemliclustat-existing-phase3-gate", "relation": "depends-on" }, { "target": "investment", "relation": "constrains" } ], "evidence": { "humanProspective": "not-applicable", "humanRetrospective": "not-applicable", "randomised": "not-applicable", "independentReplication": "moderate", "sampleSize": "$775m cash and investments; $257m six-month operating cash use; phase 3 estimated n=610", "effect": "Company and trial need £0 mission capital before readout", "limits": [ "Cash and spending change", "Result timing is company guidance", "Licensing terms are complex", "This is a programme decision, not investment advice", "No phase 3 result exists" ], "dimensions": { "biologicalPlausibility": { "rating": "not-applicable", "reason": "Biology is assessed in the linked drug records." }, "animalEvidence": { "rating": "not-applicable", "reason": "This page assesses ownership and financing." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "This page assesses ownership and financing." }, "reproducibility": { "rating": "moderate", "reason": "Trial status, licenses and cash are supported by registry and SEC records." }, "commercialConflicts": { "rating": "high", "reason": "The company reports its own runway and expected result timing." }, "recency": { "rating": "current", "reason": "Financial evidence is current to June 2026 and filed in August 2026." } } }, "layout": { "x": 476, "y": 402 }, "updated": "2026-09-15" }, { "id": "company-astellas", "title": "Astellas — mutation-selective KRAS G12D degradation", "shortTitle": "Astellas", "type": "company", "status": "open", "scope": [ "pdac", "colorectal", "pan-cancer" ], "summary": "Setidegrasib is the first KRAS G12D protein degrader in phase 3, but its PDAC evidence remains a 21-person phase 1 subgroup with 24% response.", "content": "Technology: setidegrasib binds KRAS G12D and an E3 ligase so the mutant protein is destroyed rather than merely blocked. Astellas states the molecule came from its own research. Clinical position: at 600 mg, the phase 1 PDAC subgroup had 5 responses among 21 patients, median PFS 3.0 months and OS 10.3 months. The more decisive test is a randomized, double-blind, placebo-controlled phase 3 adding setidegrasib to first-line mFOLFIRINOX or NALIRIFOX in more than 600 G12D-mutant patients. Likely advantage: mutation specificity, confirmed protein degradation and the operating capacity of a large global pharmaceutical company. Main weakness: the current response estimate has a wide interval and no control, degrader delivery and toxicity may limit exposure, and G12D-only coverage leaves other PDAC RAS variants outside the product. Company-wide R&D spending cannot be treated as setidegrasib investment without a programme-level disclosure.", "sourceIds": [ "astellas-2026-setidegrasib-phase3", "park-2026-setidegrasib", "nct07409272-setidegrasib-first-line" ], "links": [ { "target": "companies", "relation": "part-of" }, { "target": "claim-setidegrasib-g12d-degradation", "relation": "developed" }, { "target": "company-revolution-medicines", "relation": "competes-with" }, { "target": "investment", "relation": "part-of" }, { "target": "experiments", "relation": "tests" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "not-applicable", "randomised": "none", "independentReplication": "none", "sampleSize": "Phase 1 PDAC n=21; phase 3 plans more than 600", "effect": "Phase 1 PDAC ORR 24%, PFS 3.0 months, OS 10.3 months", "limits": [ "Uncontrolled phase 1", "Very small PDAC subgroup", "Company-funded", "G12D only", "Phase 3 not read out", "Programme-specific spending not disclosed" ] }, "layout": { "x": 82, "y": 110 }, "updated": "2026-09-14" }, { "id": "company-astrazeneca", "title": "AstraZeneca — survival-proved immune treatment in biliary cancer and HCC, but not in unselected PDAC", "shortTitle": "AstraZeneca", "type": "company", "status": "open", "scope": [ "pdac", "cholangiocarcinoma", "gallbladder", "hcc", "pan-cancer" ], "summary": "Durvalumab programmes produced mature randomized survival tails in biliary cancer and HCC; the same checkpoint route failed in unselected PDAC, while a licensed pan-KRAS programme remains unproved there.", "content": "Founding and origin: AstraZeneca formed in 1999 from the merger of Astra AB and Zeneca Group; the cited company history, not a scientific paper, supports that fact. Core technology in this map: the PD-L1 antibody durvalumab, the CTLA-4 antibody tremelimumab and clinical development infrastructure; a newer PDAC route is the licensed pan-KRAS inhibitor JAB-23E73 from Jacobio. Cancer focus and stage: durvalumab combinations are established treatments in advanced biliary-tract cancer and unresectable HCC. In TOPAZ-1, median survival was 12.9 versus 11.3 months and three-year survival 14.6% versus 6.9%. In HIMALAYA, five-year survival was 19.6% with STRIDE versus 9.4% with sorafenib. In unselected metastatic PDAC, durvalumab with or without tremelimumab produced only one response among 64 treated patients and failed its expansion rule. JAB-23E73 entered a Chinese first-line pancreatic trial without a posted efficacy result.\n\nFunding and ownership: AstraZeneca is a public company financed through its operations and capital markets; a stable list of current investors was not reconstructed because share ownership changes. The 2025 annual report confirms company-wide reporting, but group R&D cannot be assigned to durvalumab, biliary cancer, HCC or JAB-23E73. The Jacobio agreement included $100m upfront and up to $1.915bn more in contingent value; the contingent amount is not deployed research cash. Patents and access: exact live durvalumab and JAB-23E73 families were not audited here, so programme access and combination rights remain a counsel-led task. Partnerships: the map records Jacobio for pan-KRAS; TOPAZ-1 and HIMALAYA were AstraZeneca-sponsored global trials. Latest scientific evidence is the 2025 mature follow-up from both programmes, not company sales material. Competitive advantage: two distinct randomized gastrointestinal-cancer survival programmes, global trial scale and an entry into broad RAS. Main weakness: sponsor control creates correlated evidence, broad company spend cannot reveal disease-level efficiency, mixed biliary sites blur exact benefit and checkpoint success in HCC or biliary cancer did not translate to unselected PDAC. Mission role: use AstraZeneca as a possible trial and compound partner only with independent analysis, specimen access, negative-result publication and no assumption that its immune or RAS assets will transfer across diseases.", "sourceIds": [ "astrazeneca-2025-history", "astrazeneca-2025-annual-report", "oh-2025-topaz1-three-year", "rimassa-2025-himalaya-five-year", "oreilly-2019-durvalumab-tremelimumab", "jacobio-2026-annual-results" ], "links": [ { "target": "companies", "relation": "part-of" }, { "target": "company-jacobio", "relation": "partners-with" }, { "target": "claim-biliary-immunochemotherapy-tail", "relation": "sponsors" }, { "target": "claim-hcc-stride-five-year-tail", "relation": "sponsors" }, { "target": "failure-checkpoint-blockade-unselected-pdac", "relation": "sponsors" }, { "target": "claim-current-capital-frontier-2026", "relation": "funds" }, { "target": "treatment", "relation": "supports" }, { "target": "risks", "relation": "depends-on" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong", "independentReplication": "moderate — two diseases, sponsor-correlated", "sampleSize": "TOPAZ-1 n=685; HIMALAYA comparison n=782; PDAC checkpoint study 64 treated patients", "effect": "Company capability record; biliary and HCC survival effects are stated in linked claims, while PDAC checkpoint activity was near zero", "limits": [ "Company status and ownership can change", "Disease-level R&D spend unavailable", "Exact patent families not audited", "Sponsor controls much source data", "Biliary sites were pooled", "JAB-23E73 lacks PDAC efficacy evidence" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "Human survival proves immune combinations can work in selected biliary and HCC settings, not across cancers." }, "animalEvidence": { "rating": "not-applicable", "reason": "The company ranking rests on human trials and corporate records rather than animal activity." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Laboratory activity is not used to rank AstraZeneca's mission capability." }, "reproducibility": { "rating": "moderate", "reason": "Two randomized disease programmes support execution, though both are sponsor-controlled and biologically distinct." }, "commercialConflicts": { "rating": "high", "reason": "The company funds and controls the medicines, trials and much of the resulting evidence." }, "recency": { "rating": "current", "reason": "The record includes mature 2025 evidence and company reporting available in 2026." } } }, "layout": { "x": 126, "y": 95 }, "updated": "2026-09-14" }, { "id": "company-biontech", "title": "BioNTech — individualised mRNA vaccine design and manufacture", "shortTitle": "BioNTech", "type": "company", "status": "open", "scope": [ "pdac", "colorectal", "pan-cancer" ], "summary": "BioNTech supplies the individualised RNA-vaccine platform in the first randomized post-resection PDAC efficacy trial, but a related colorectal monotherapy study was terminated.", "content": "Technology: each patient's tumour sequence is used to select mutation-derived targets and manufacture autogene cevumeran, an individualised mRNA cancer vaccine. PDAC position: BioNTech collaborates on the recruiting 260-person IMCODE003 phase 2, which adds vaccine and atezolizumab to mFOLFIRINOX after resection. Competitive advantage: an established RNA manufacturing platform and the ability to industrialise a patient-specific product across 89 sites. Main weakness: product creation takes time, antigen prediction can be wrong, low-mutation tumours offer few targets, and the phase 1 PDAC survival association is selected by immune response after treatment. In August 2026 BioNTech terminated a related ctDNA-positive resected colorectal monotherapy phase 2 while continuing PDAC; the public statement did not disclose enough result detail to decide whether the failure was disease, product, monotherapy or trial execution. Mission role: negotiate manufacturing-time and failure data, then judge the randomized PDAC trial by all assigned patients and recurrence—not only T-cell expansion.", "sourceIds": [ "nct05968326-imcode003", "rojass-2025-neoantigen-followup", "biontech-2026-cevumeran-crc-termination" ], "links": [ { "target": "companies", "relation": "part-of" }, { "target": "claim-neoantigen-vaccine-frontier", "relation": "developed" }, { "target": "company-roche-genentech", "relation": "partners-with" }, { "target": "person-vinod-balachandran", "relation": "partners-with" }, { "target": "hypothesis-low-burden-ras-immune-extinction", "relation": "could-test" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "moderate", "randomised": "none", "independentReplication": "weak", "sampleSize": "PDAC phase 1 vaccinated n=16; recruiting randomized phase 2 plans n=260", "effect": "Durable vaccine-induced T cells in some patients; clinical efficacy unproved", "limits": [ "Company-partnered evidence", "Individual manufacturing failures and timing not fully reported", "PDAC randomized readout expected years away", "Related colorectal monotherapy trial terminated", "No current PDAC treatment-effect estimate" ] }, "layout": { "x": 107, "y": 90 }, "updated": "2026-09-14" }, { "id": "company-boston-scientific-biliary", "title": "Boston Scientific — mature biliary-device execution with linked trial-control risk", "shortTitle": "Boston Scientific biliary", "type": "company", "status": "supported", "scope": [ "pdac", "cholangiocarcinoma", "pan-cancer", "transfer-unproven" ], "summary": "WallFlex metal stents have randomized pancreatic and malignant-obstruction evidence, while the manufacturer had major roles in the decisive modern trial.", "content": "Evidence: the WallFlex covered and uncovered self-expanding metal stent system appears in NCT01774019, the 284-person metal-drainage versus early-surgery trial, and NCT02238847, the 119-person covered versus uncovered trial during neoadjuvant therapy. The first found serious adverse events 29.0% versus 26.5% within a 15-point noninferiority margin; the second found sustained drainage 72.2% versus 72.9% with different migration and ingrowth failures. Capability: a standardized supplied device, international trial operations and a mature regulatory and manufacturing route. Independence limit: the 2026 paper states that Boston Scientific participated in study design, execution, database building and maintenance, monitoring, analysis, interpretation and writing; employees were authors, and multiple investigators disclosed company relationships. Mission role: a device supplier under fixed specifications, price and replacement terms, not the sole data owner or route designer. Required contract: investigator-controlled protocol and analysis, patient-level event data, independent adjudication, negative-result publication, device traceability, supply continuity and permission to compare any suitable metal or EUS route. No company investment case follows from noninferiority or stent patency.", "sourceIds": [ "costamagna-2026-sems-early-surgery-rct", "seo-2019-covered-uncovered-sems-rct", "chen-2023-element-biliary-rct", "esge-2018-biliary-stenting-guideline" ], "links": [ { "target": "companies", "relation": "part-of" }, { "target": "trial-pdac-biliary-drainage-action-map", "relation": "supplies-device-to" }, { "target": "claim-biliary-drainage-is-route-not-cancer-treatment", "relation": "informs" }, { "target": "unknown-pdac-biliary-drainage-route", "relation": "could-support" }, { "target": "risks", "relation": "constrained-by" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong", "independentReplication": "moderate for metal-stent class; weak for product-specific independence", "sampleSize": "Randomized device studies n=284 and n=119 plus mixed malignant-obstruction trials", "effect": "Strong device and trial execution; no anticancer or investment effect", "limits": [ "Manufacturer-controlled inputs", "No company financial analysis", "Named-device evidence", "Expert centres", "No survival gain", "No acquisition recommendation" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Larger-diameter internal drainage has a direct route to longer patency and fewer occlusions." }, "animalEvidence": { "rating": "not-applicable", "reason": "The company ranking rests on patient device outcomes." }, "inVitroEvidence": { "rating": "moderate", "reason": "Engineering and patency testing support the device but do not predict clinical route benefit alone." }, "reproducibility": { "rating": "moderate", "reason": "Metal-stent advantages repeat across studies, while product-specific independent results remain limited." }, "commercialConflicts": { "rating": "high", "reason": "The manufacturer supplied devices and held extensive roles in key trials and reports." }, "recency": { "rating": "current", "reason": "The company-linked evidence includes a 2026 randomized report." } } }, "updated": "2026-09-14" }, { "id": "company-carsgen-therapeutics", "title": "CARsgen Therapeutics", "shortTitle": "CARsgen", "type": "company", "status": "active", "scope": [ "pdac", "pan-cancer" ], "summary": "CARsgen can fund satri-cel development from RMB1.400bn cash; its PDAC evidence remains small, uncontrolled and largely not reusable.", "content": "Product. CARsgen owns satricabtagene autoleucel, or satri-cel, an autologous CLDN18.2 CAR-T product. Its pooled advanced-PDAC evidence is four responses among 24 infused people; its adjuvant evidence is six selected people with one recurrence after short follow-up. Neither estimates added survival. Manufacturing position. The company sells another autologous CAR-T product and now has a commercial cell-manufacturing route; this lowers delivery risk but does not prove pancreatic tumour access or benefit. Finance. At 30 June 2026 it reported RMB1.400bn cash and cash equivalents, RMB62m first-half revenue and expected at least RMB1.2bn year-end cash, with operations funded into 2030. It does not need mission capital for a 20-person pancreatic study. Evidence access. NCT03874897 says data may be shared, NCT04581473 and NCT05911217 say they will not be, and no registry posts results. Decision. Invest £0 in the company and £0 in a duplicate satri-cel cohort. Request the full 134-person NCT03874897 record, every pancreatic screen-to-outcome denominator, exact marker maps and a controlled adjuvant design. Commercial approval and gastric-cancer results do not raise the PDAC evidence tier.", "sourceIds": [ "qi-2024-ct041-pdac", "clinicaltrials-nct03874897-ct041", "clinicaltrials-nct04581473-ct041", "clinicaltrials-nct05911217-satricel-adjuvant", "carsgen-2025-satricel-pdac-adjuvant", "carsgen-2026-interim-results" ], "links": [ { "target": "companies", "relation": "part-of" }, { "target": "claim-satricel-pdac-early-signal", "relation": "develops" }, { "target": "failure-cell-therapy-route-and-denominator-pdac", "relation": "must-repair" }, { "target": "trial-pdac-engineered-cell-action-map", "relation": "monitored-by" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "not-applicable", "randomised": "none", "independentReplication": "none", "sampleSize": "PDAC advanced n=24 infused; adjuvant n=6 reported of n=20 planned; company cash RMB1.400bn", "effect": "Solvent developer with manufacturing capacity; no controlled PDAC survival effect", "limits": [ "Issuer-authored finance", "Cash can change", "PDAC cohorts are small and uncontrolled", "Most participant data are not promised", "Other-product approval does not establish PDAC value" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The product has a measurable surface target and human pancreatic responses." }, "animalEvidence": { "rating": "moderate", "reason": "Models support the platform but are not the company-value gate." }, "inVitroEvidence": { "rating": "strong", "reason": "Cell potency, target binding and manufacture are measurable and operational." }, "reproducibility": { "rating": "weak", "reason": "No controlled or independent PDAC efficacy result exists." }, "commercialConflicts": { "rating": "high", "reason": "CARsgen owns the product, funds the studies and controls most data." }, "recency": { "rating": "current", "reason": "Finance and clinical status were checked through September 2026." } } }, "layout": { "x": 262, "y": 190 }, "updated": "2026-09-15" }, { "id": "company-clinical-laserthermia-systems", "title": "Clinical Laserthermia Systems: pancreatic studies stalled while the business moved elsewhere", "shortTitle": "Clinical Laserthermia Systems", "type": "company", "status": "active", "scope": [ "pdac", "laser-ablation", "company", "device-lifecycle" ], "summary": "CLS supplied and sponsored the imILT pancreatic programme; the clinical record is incomplete and unsatisfactory, while its current investment case starts with neurosurgery and repeat disposable revenue.", "content": "Clinical Laserthermia Systems AB develops the TRANBERG Thermal Therapy System and the imILT method. It sponsored the eight-person randomized pancreatic study and the five-person Marseille study, and collaborated on the Verona programme. The combined surgical report found three late pancreatic fistulas among 13 treatments and called safety, feasibility and handling unsatisfactory. The randomized study stopped for slow recruitment and never posted results. The company's current investor page describes sales of systems plus repeat procedure disposables, claims gross margin above 60% and says commercial focus begins with neurosurgery before adjacent areas. It does not name pancreatic cancer as a current focus. This creates a data-custody and continuity question, not an investment case. The mission should buy no shares, equipment or consumables. It should offer a capped data-recovery contract only if CLS supplies the assignment file, treatment logs, temperature records, device failures, all adverse events and unrestricted publication rights. A negative result must be publishable. The value, if any, lies in learning why the pancreatic programme failed and whether the heat-control method can be tested independently.", "sourceIds": [ "paiella-2021-imilt-pancreas", "clinicaltrials-nct02702986-imilt-verona", "clinicaltrials-nct02973217-imilt-marseille", "clinicaltrials-nct03187587-imilt-randomized", "cls-2026-investment-case" ], "links": [ { "target": "claim-pancreatic-laser-heat-feasibility-not-benefit", "relation": "develops" }, { "target": "failure-laser-fistula-accrual-and-device-continuity", "relation": "exposed-to" }, { "target": "unknown-laser-ablation-delivery-and-net-benefit", "relation": "constrained-by" }, { "target": "hypothesis-laser-existing-study-recovery-gate", "relation": "data-requested-by" }, { "target": "trial-pdac-laser-ablation-action-map", "relation": "mapped-by" }, { "target": "companies", "relation": "part-of" } ], "evidence": { "humanProspective": "weak", "humanRetrospective": "none", "randomised": "weak — sponsored trial stopped at eight", "independentReplication": "weak", "sampleSize": "Company-linked pancreatic studies n=15 assigned and n=8 randomized", "effect": "No established pancreatic benefit; one unsatisfactory safety report and one terminated randomized study", "limits": [ "Current company page may omit dormant research", "Private data may be more complete", "Market approvals vary by country and indication", "Gross margin is a company statement", "Pancreatic work used more than one access route", "No current investment outreach was made" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The platform can deliver thermal energy; pancreatic selectivity is unproved." }, "animalEvidence": { "rating": "moderate", "reason": "Animal feasibility does not repair the human programme." }, "inVitroEvidence": { "rating": "strong", "reason": "Laser heating is established, but has low value for this investment decision." }, "reproducibility": { "rating": "weak", "reason": "The pancreatic clinical chain did not complete a useful controlled test." }, "commercialConflicts": { "rating": "high", "reason": "The company sponsored studies and earns from systems and per-procedure disposables." }, "recency": { "rating": "current", "reason": "The company investment case was checked in September 2026." } } }, "layout": { "x": 1000, "y": 924 }, "updated": "2026-09-15" }, { "id": "company-cosomil", "title": "Cosomil — Japanese single-molecule enzyme liquid biopsy in population feasibility testing", "shortTitle": "Cosomil", "type": "company", "status": "open", "scope": [ "pdac", "pancreatic-all", "pan-cancer" ], "summary": "A 2022 University of Tokyo/RIKEN spinout has moved its enzyme-activity assay into a 10,000-person real-workflow study; clinical utility remains unproved.", "content": "Founded in Tokyo in April 2022 by Yu Kagami after work by Toru Komatsu at the University of Tokyo and Rikiya Watanabe at RIKEN. Technology: fluorescence chemistry and single-molecule measurement of blood enzyme activity, developed as the Enzeavour pancreatic-cancer assay. Funding and partners: NEDO lists a ¥499m grant decision for 2024-2025. A pre-Series A named Coral Capital, ANRI and Green Core, while a 2026 Series A1 was led by Delight Ventures with Animal Spirits, Gogin Capital, ANRI and Green Core; both private amounts and valuation were undisclosed. Stage: an in-vitro diagnostic manufacturing and marketing business licence was obtained in 2025, but the specific test remained aimed at product approval; a nationwide 10,000-person feasibility study began enrolment in 2026. Advantage: it tests a functional protein signal and a full Japanese screening workflow rather than another mutation-only case-control set. Weakness: the assay's independent accuracy, specificity, sensitivity and net benefit are not yet established; the single-arm study may contain too few cancers and incomplete verification to do so. Company claims remain separate from registry endpoints.", "sourceIds": [ "nedo-2024-cosomil-profile", "jrct-2026-enzeavour-feasibility", "cosomil-2026-series-a1" ], "links": [ { "target": "companies", "relation": "part-of" }, { "target": "trial-enzeavour-japan-screening", "relation": "sponsors-product-for" }, { "target": "company-grail", "relation": "competes-with" }, { "target": "claim-screening-base-rate-maths", "relation": "constrained-by" }, { "target": "investment", "relation": "funded-by" }, { "target": "investor-cosomil-public-private-capital", "relation": "funded-by" } ], "evidence": { "humanProspective": "none — planned study, no result", "humanRetrospective": "not-applicable — not assessed in this record", "randomised": "none", "independentReplication": "none", "sampleSize": "Prospective feasibility target n=10,000", "effect": "No prospective accuracy or outcome result yet", "limits": [ "Company-originated technology", "Funding amount and valuation undisclosed", "No randomized utility trial", "Partial-verification risk", "Regulatory status of the specific assay remains incomplete" ] }, "layout": { "x": 101, "y": 114 }, "updated": "2026-09-14" }, { "id": "company-elevar-therapeutics", "title": "Elevar Therapeutics — HCC survival evidence plus acquired FGFR2 development risk", "shortTitle": "Elevar Therapeutics", "type": "company", "status": "open", "scope": [ "hcc", "cholangiocarcinoma", "hepatobiliary" ], "summary": "Elevar co-funded a positive HCC phase 3 trial and now controls global lirafugratinib development, joining proven execution to an unreported CCA programme.", "content": "Origin and ownership: the company's own history says it began as LSK BioPharma in 2005, received a $1.3m HLB investment in 2007 and was acquired by HLB for $440m in 2020. Technology and stage: Elevar holds ex-China and Korea rights to rivoceranib and co-developed it with Hengrui's camrelizumab. CARES-310 randomized 543 first-line unresectable HCC patients; median survival was 23.8 versus 15.2 months against sorafenib, hazard ratio 0.64. This is high-grade human evidence, but 83% of patients were Asian, the trial was open-label and treatment-related serious adverse events occurred in 25%. Elevar also acquired worldwide lirafugratinib rights from Relay. The 490-person REFOCUS study completed without posted registry results by February 2026, so an announced filing path is not clinical proof. Funding and investors: parent ownership and the $440m acquisition are visible; programme-specific R&D is not. The Relay deal is mostly contingent milestone value, not deployed capital. Advantage: late-stage regulatory and commercial capability plus a randomized HCC survival result. Weakness: two-company dependencies, incomplete public CCA results, geographic transport and liver-toxicity constraints. Mission role: a credible HCC and CCA execution partner only under patient-level data access and disease-specific evidence gates.", "sourceIds": [ "elevar-2024-company-overview", "qin-2025-cares310-final", "relay-elevar-2024-license", "nct04526106-refocus" ], "links": [ { "target": "companies", "relation": "part-of" }, { "target": "company-relay-therapeutics", "relation": "licenses-from" }, { "target": "company-hengrui", "relation": "partners-with" }, { "target": "claim-hcc-cares310-survival", "relation": "funds" }, { "target": "hcc", "relation": "treats-subgroup" }, { "target": "cholangiocarcinoma", "relation": "treats-subgroup" }, { "target": "investment", "relation": "contains" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong", "independentReplication": "moderate", "sampleSize": "HCC phase 3 n=543; CCA and solid-tumour phase 1/2 registry n=490; company ownership records", "effect": "HCC median OS 23.8 versus 15.2 months, HR 0.64; CCA comparative effect unknown", "limits": [ "Company-funded open-label HCC trial", "Eighty-three percent Asian HCC population", "CCA registry results absent", "Company deck used only for ownership", "Acquisition price is not R&D", "Deal value mostly contingent" ] }, "updated": "2026-09-14" }, { "id": "company-elicio-therapeutics", "title": "Elicio Therapeutics — lymph-node-targeted KRAS vaccine with a failed randomized endpoint", "shortTitle": "Elicio Therapeutics", "type": "company", "status": "contested", "scope": [ "pdac", "colorectal", "pan-cancer" ], "summary": "ELI-002 reliably induces KRAS-specific T cells, but AMPLIFY-7P missed its prespecified DFS endpoint and the company needs more capital before phase 3.", "content": "Technology: amphiphile-modified KRAS peptides and adjuvant designed to traffic to lymph nodes. Clinical position: phase 1 established immune activity, but randomized phase 2 AMPLIFY-7P in 144 resected PDAC patients missed its main disease-free-survival endpoint. The sponsor now proposes an R0-only phase 3 with more dosing and a small metastatic combination study based partly on three post-study complete responses after checkpoint therapy. Capital: the June 2026 filing reported $23.5 million cash, first-half R&D of $13.6 million, substantial doubt about continuing as a going concern, runway into first-quarter 2027, and that phase 3 requires another capital raise. Revolution Medicines held about $3.938 billion in cash and securities at the same date, roughly 168 times Elicio's cash. Likely advantage: off-the-shelf multi-variant immune priming with low observed treatment toxicity. Main weakness: immune response has not translated into a positive randomized main endpoint, subgroup rescue is vulnerable to chance, and financing can force underpowered or rapidly changing development plans. The mission should fund a clean confirmatory mechanism test only if full patient-level phase 2 data support the fixed selection rule.", "sourceIds": [ "elicio-2026-amplify7p-results", "wainberg-2025-eli002-final", "revolution-medicines-2026-q2-10q" ], "links": [ { "target": "companies", "relation": "part-of" }, { "target": "claim-eli002-mrd-frontier", "relation": "developed" }, { "target": "failure-eli002-phase2-missed-dfs", "relation": "sponsors" }, { "target": "company-revolution-medicines", "relation": "could-combine-with" }, { "target": "funding", "relation": "depends-on" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong", "independentReplication": "none", "sampleSize": "Phase 1 n=25; randomized phase 2 n=144", "effect": "Immune activity established; phase 2 prespecified ITT DFS endpoint missed", "limits": [ "Sponsor-reported phase 2 topline data", "Main effect estimate not published", "Post-hoc R0 subgroup", "No mature OS", "Going-concern warning", "Phase 3 depends on financing" ] }, "layout": { "x": 85, "y": 111 }, "updated": "2026-09-14" }, { "id": "company-erbe-elektromedizin", "title": "ERBE Elektromedizin: HybridTherm product and trial-control dependency", "shortTitle": "ERBE Elektromedizin", "type": "company", "status": "supported", "scope": [ "pdac", "local-treatment", "medical-device", "device-governance" ], "summary": "ERBE made HybridTherm, held its randomized allocation list, monitored the trial and withdrew the probe; any reuse of the evidence requires independent record custody.", "content": "ERBE Research Elektromedizin made the HybridTherm probe, an internally carbon-dioxide-cooled bipolar radiofrequency device. In NCT02336672, ERBE held the computer-generated randomization list and was responsible for monitoring data accuracy, completeness and reliability. Two paper authors were ERBE employees. The paper says the manufacturer withdrew the probe in 2020 after recruitment problems, helping stop the trial. These facts do not show misconduct, and the paper reports no severe procedure event. They show one company occupying four control points: the product, allocation record, data monitoring and technical authorship. The later 66-participant registry count conflicts with the paper's 40 randomized participants, so independent access to the dated screening, consent, assignment and monitoring trail is mandatory. The mission decision is £0 in ERBE equity and £0 in equipment. A capped evidence-recovery contract is possible only if the sponsor and company grant full raw export, data definitions, audit trail and publication rights, including a public explanation of the count mismatch. Because the probe was withdrawn, a clinical effect from this device cannot be assumed to transfer to another geometry.", "sourceIds": [ "testoni-2021-hybridtherm-rct", "clinicaltrials-nct02336672-hybridtherm" ], "links": [ { "target": "companies", "relation": "part-of" }, { "target": "claim-hybridtherm-randomized-no-patient-benefit", "relation": "supplies" }, { "target": "failure-hybridtherm-product-accrual-and-custody", "relation": "involves" }, { "target": "unknown-eus-rfa-added-benefit-after-hybridtherm", "relation": "constrained-by" }, { "target": "hypothesis-eus-rfa-existing-study-gate", "relation": "enables" }, { "target": "trial-pdac-eus-rfa-action-map", "relation": "involves" }, { "target": "patent-landscape", "relation": "requires-review-by" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "moderate", "independentReplication": "weak", "sampleSize": "One randomized programme: 40 reported assignments; registry says 66 enrolled", "effect": "Operational dependency map; no company valuation or efficacy claim", "limits": [ "Ownership and finance were not reviewed", "Contracts were not obtained", "The reason for withdrawal is described only in the paper", "The enrollment mismatch may be administrative", "No allegation of wrongdoing", "Transfer to current ERBE products was not assessed" ], "dimensions": { "biologicalPlausibility": { "rating": "not-applicable", "reason": "This record maps product and trial control." }, "animalEvidence": { "rating": "not-applicable", "reason": "Animal findings do not reduce the custody dependency." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Bench function does not reconcile the human record." }, "reproducibility": { "rating": "weak", "reason": "The main dataset has not been independently reproduced." }, "commercialConflicts": { "rating": "high", "reason": "One company controlled the device, assignment list, monitoring and employed authors." }, "recency": { "rating": "current", "reason": "Current status is limited to the linked completed trial record checked in 2026." } } }, "layout": { "x": 802, "y": 726 }, "updated": "2026-09-15" }, { "id": "company-exact-sciences", "title": "Exact Sciences — installed screening business moving into multicancer blood testing", "shortTitle": "Exact Sciences", "type": "company", "status": "open", "scope": [ "pdac", "colorectal", "hcc", "pan-cancer" ], "summary": "Cancerguard combines several blood-signal classes, but its headline aggressive-cancer sensitivity pools six organs and key validation reports remain company data on file.", "content": "Technology and position: Exact Sciences commercialises Cancerguard, a multi-biomarker blood test descended in part from the CancerSEEK/Thrive work, alongside a large established colorectal-screening and oncology-testing operation. Competitive advantage: several signal classes, laboratory and reimbursement experience, and an installed screening channel. The company brochure reports 67.9% sensitivity in 106 samples pooled across pancreatic, oesophageal, liver, lung, stomach and ovarian cancers. That number is not pancreatic sensitivity. The brochure cites key 2025 development and validation reports as data on file and notes that no established imaging guideline follows a positive result. Main weakness: commercial availability, pooled case-control performance and a plausible work-up are not proof of stage shift or mortality benefit. Mission role: seek organ-specific locked results, intended-population positive predictive value, benign-pancreas controls and downstream-harm data; do not fund a second small case-control panel.", "sourceIds": [ "exact-2025-cancerguard-brochure", "cohen-2018-cancerseek", "lennon-2020-detect-a" ], "links": [ { "target": "companies", "relation": "part-of" }, { "target": "claim-case-control-blood-tests-overstate-screening-readiness", "relation": "must-overcome" }, { "target": "claim-mced-workflow-feasible-utility-unproved", "relation": "connects-to" }, { "target": "company-grail", "relation": "competes-with" }, { "target": "programme-25m-prospective-observatory", "relation": "could-partner" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "moderate", "randomised": "none", "independentReplication": "weak", "sampleSize": "Company-reported aggressive-cancer subgroup n=106; DETECT-A n=10,006 used an earlier test workflow", "effect": "Pooled six-cancer sensitivity 67.9%; pancreatic-specific utility unreported", "limits": [ "Company brochure", "Pooled organs", "Key reports are data on file", "No established positive-result imaging guideline", "No mortality result", "Commercial availability can outrun evidence" ] }, "layout": { "x": 116, "y": 91 }, "updated": "2026-09-14" }, { "id": "company-fibrogen", "title": "FibroGen — two late-stage PDAC failures retired the CTGF programme", "shortTitle": "FibroGen", "type": "company", "status": "refuted", "scope": [ "pdac" ], "summary": "Pamrevlumab failed overall survival in both locally advanced and metastatic randomized settings; FibroGen ended the programme and cut about 75% of its US workforce.", "content": "Technology: pamrevlumab was an antibody against connective-tissue growth factor, intended to change the fibrotic tumour environment and make chemotherapy or surgery more effective. The 284-person double-blind LAPIS phase 3 produced median overall survival 17.3 versus 17.9 months, HR 1.08, with no improvement in event-free survival, progression-free survival or response. A separate PanCAN Precision Promise metastatic arm also missed overall survival. FibroGen then ended pamrevlumab R&D and announced an approximately 75% U.S. workforce reduction. What this teaches: a plausible stromal target, encouraging early study and progression through an adaptive platform did not survive decisive randomized testing. Main company weakness is now historical, not an investable PDAC platform. Mission role: obtain specimens and full negative results to learn whether target engagement failed, the target was wrong or no subgroup benefited; do not revive CTGF blockade without a human selector and direct target-engagement proof.", "sourceIds": [ "picozzi-2026-lapis", "fibrogen-2024-pamrevlumab-winddown" ], "links": [ { "target": "companies", "relation": "part-of" }, { "target": "failure-bulk-stroma-removal", "relation": "extends" }, { "target": "failures", "relation": "part-of" }, { "target": "company-verastem", "relation": "warns" }, { "target": "funding", "relation": "failed-because-of" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong", "independentReplication": "strong", "sampleSize": "LAPIS n=284 plus separate Precision Promise arm", "effect": "LAPIS OS 17.3 vs 17.9 months, HR 1.08, 95% CI 0.83-1.41; second trial also negative", "limits": [ "Full Precision Promise effect estimate not public", "Company ended the programme", "No validated responsive subgroup", "Failure of CTGF blockade does not refute every stromal intervention" ] }, "layout": { "x": 119, "y": 93 }, "updated": "2026-09-14" }, { "id": "company-genmab-partner-bizengri", "title": "Genmab and Partner Therapeutics: an approved rare-driver asset without separated pancreatic economics", "shortTitle": "Genmab / Partner Therapeutics", "type": "company", "status": "active", "scope": [ "pdac", "cholangiocarcinoma", "nrg1", "precision-oncology", "commercial", "capital-allocation" ], "summary": "Genmab acquired the asset with Merus and receives US royalties from Partner Therapeutics; the mission should buy route data, not company exposure.", "content": "Genmab acquired Merus and now holds zenocutuzumab in its royalty portfolio. Partner Therapeutics commercializes BIZENGRI in the United States and pays Genmab royalties on US net sales. The public first-quarter 2026 filing does not separate BIZENGRI sales, pancreatic sales, NRG1 testing completion or confirmatory-study spending. The asset has a direct target, an approved pancreatic indication and a 40% response signal, making it more mature than most pancreatic start-up programmes. It also has a narrow addressable group, an overdue public follow-up record, no proven comparative survival gain and no established UK route in the sources reviewed. Capital decision: buy £0 of either company and £0 of speculative drug inventory. Pay at most £150,000 for independent custody and publication of the completed pancreatic trial record, only with all-screened denominators, patient-level timing, negative publication rights and no promotional control. The separate £1.5m testing-route budget belongs to public laboratories and centres, not the drug owner. Enterprise value or royalty exposure is not a substitute for finding eligible patients or verifying benefit.", "contentSections": [ "FDA added an NRG1-fusion cholangiocarcinoma indication in May 2026. The single-arm efficacy set had 19 people and seven responses. This broadens the same testing-and-referral asset across two rare diseases; it does not create comparative survival proof or a broad biliary market." ], "sourceIds": [ "genmab-2026-q1-bizengri", "fda-2024-bizengri-multidisciplinary-review", "fda-2026-ongoing-bizengri-accelerated-approval", "fda-2026-bizengri-cholangiocarcinoma", "clinicaltrials-nct02912949-enrgy" ], "links": [ { "target": "companies", "relation": "part-of" }, { "target": "claim-nrg1-approved-response-not-durable-control", "relation": "owns-asset-for" }, { "target": "failure-nrg1-testing-and-confirmation-route", "relation": "bounded-by" }, { "target": "hypothesis-nrg1-rna-route-completion", "relation": "excluded-from-funding" }, { "target": "trial-pdac-nrg1-action-map", "relation": "mapped-by" }, { "target": "regulatory-bizengri-nrg1-cholangiocarcinoma", "relation": "involves" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "weak", "randomised": "none", "independentReplication": "none for comparative benefit", "sampleSize": "FDA pancreatic efficacy set n=30; required follow-up at least n=50; no separated product economics", "effect": "Commercial ownership of an approved response-producing treatment, without evidence that company investment repairs route loss or proves survival", "limits": [ "Sales not separated", "Pancreatic revenue unknown", "Royalty terms not fully quantified here", "US commercialization only", "UK access unresolved", "Confirmatory result overdue", "Acquisition value covered many assets" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The owned asset directly blocks the fusion's receptor pair." }, "animalEvidence": { "rating": "not-applicable", "reason": "Human response data drive the asset decision." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "The remaining commercial questions are access, durability and scale." }, "reproducibility": { "rating": "weak", "reason": "One clinical programme supports the pancreatic indication." }, "commercialConflicts": { "rating": "high", "reason": "The company owns the asset and receives product royalties." }, "recency": { "rating": "current", "reason": "Ownership and FDA status are current through 2026." } } }, "layout": { "x": 1240, "y": 1164 }, "updated": "2026-09-15" }, { "id": "company-grail", "title": "GRAIL — commercial multicancer screening with a negative main randomized endpoint", "shortTitle": "GRAIL", "type": "company", "status": "contested", "scope": [ "pan-cancer", "pdac", "hepatobiliary", "colorectal" ], "summary": "Galleri has the largest randomized MCED dataset and broad commercial use, but NHS-Galleri missed its combined stage III/IV endpoint and mortality benefit remains unknown.", "content": "Technology: a blood methylation test that reports a cancer signal and predicted origin across many cancers. Position: Galleri has moved far beyond case-control validation into prospective return-of-results studies, commercial use and the roughly 142,250-person randomized NHS-Galleri trial. That trial missed its main combined stage III/IV endpoint but produced a secondary signal for fewer stage IV diagnoses. Capital: at 30 June 2026 GRAIL reported $55.6 million cash and $806.0 million short-term securities, $861.6 million combined. First-half R&D was $95.45 million, operating cash use $167.7 million and net loss $203.4 million. Likely advantage: unmatched linked samples, population-scale operations, a commercial laboratory and the only completed randomized MCED stage trial. Main weakness: the assay misses much stage I disease, the main stage endpoint failed, downstream workups carry harm and cost, and the company has incentives to foreground secondary results while seeking approval and reimbursement. The mission should seek access to cancer-specific NHS data and long-term mortality follow-up rather than funding another small test-performance study.", "sourceIds": [ "grail-2026-q2-10q", "sasieni-2026-nhs-galleri", "liu-2020-ccga-methylation", "schrag-2023-pathfinder" ], "links": [ { "target": "companies", "relation": "part-of" }, { "target": "failure-nhs-galleri-primary-stage-shift", "relation": "sponsors" }, { "target": "claim-mced-early-stage-sensitivity-limit", "relation": "developed-test-in" }, { "target": "claim-mced-workflow-feasible-utility-unproved", "relation": "developed-test-in" }, { "target": "investment", "relation": "part-of" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "strong", "randomised": "strong", "independentReplication": "weak", "sampleSize": "NHS-Galleri approximately 142,250; PATHFINDER 6,621 analysable; CCGA validation 6,689", "effect": "Main stage III/IV endpoint not met; secondary stage IV IRR 0.86; mortality unknown", "limits": [ "Sponsor-linked evidence base", "Main randomized endpoint missed", "No mortality result", "Cancer-specific effects not established", "Commercial incentives", "Programme spending not reported by cancer type" ] }, "layout": { "x": 91, "y": 113 }, "updated": "2026-09-14" }, { "id": "company-hengrui", "title": "Jiangsu Hengrui — intravenous KRAS-G12D inhibition in phase 3", "shortTitle": "Hengrui", "type": "company", "status": "supported", "scope": [ "pdac", "pan-cancer" ], "summary": "Hengrui has moved HRS-4642 from a 30-patient response signal into a blinded first-line pancreatic phase 3 trial.", "content": "Technology and origin: HRS-4642 is a high-affinity non-covalent KRAS-G12D inhibitor delivered intravenously in a liposomal formulation intended to increase tumour exposure and reduce the oral gastrointestinal constraint. Disease and stage: advanced KRAS-G12D PDAC; phase 1b/2 NCT06520488 and phase 3 NCT07232875. Evidence: 30 untreated phase 2 patients had a 63.3% response rate, but all also received gemcitabine and nab-paclitaxel; 90.3% experienced grade 3 or worse treatment-related events. Competitive advantage: a different molecule, route and national trial network from the common-RAS and degrader programmes. Weakness: mutation-only reach, intravenous delivery, high combined toxicity and no controlled contribution or survival result yet. The phase 3 placebo-plus-identical-chemotherapy design can isolate contribution. Ownership and access: Hengrui is the assignee on WO2022268051A1 for fused tetracyclic KRAS-G12D compounds and WO2024022507A1 for liposomal compositions using a compound from that family. Its prospectus calls HRS-4642 a proprietary liposomal injectable. This makes the two families the probable control stack, but the checked public primary records do not directly print HRS-4642 beside the patent publications. Exact structure, live national claims, combination rights and freedom to operate require counsel. Company employment and sponsor control are explicit conflicts. Programme-specific expenditure, licence terms and reliable valuation allocation are not public enough to estimate.", "sourceIds": [ "cui-2026-hrs4642-pdac", "nct07232875-hrs4642-phase3", "patent-wo2022268051-hengrui-kras-g12d", "patent-wo2024022507-hengrui-liposome", "hengrui-2025-hk-prospectus" ], "links": [ { "target": "companies", "relation": "part-of" }, { "target": "claim-hrs4642-g12d-chemo-signal", "relation": "develops" }, { "target": "patent-landscape", "relation": "protected-by" }, { "target": "company-astellas", "relation": "competes-with" }, { "target": "company-revolution-medicines", "relation": "competes-with" }, { "target": "person-liwei-wang", "relation": "collaborates-with" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "not-applicable", "randomised": "none — ongoing trial, no result", "independentReplication": "none", "sampleSize": "Phase 1b/2 n=31; phase 3 ongoing", "effect": "Uncontrolled ORR 63.3% in 30 treatment-naive patients", "limits": [ "Sponsor-linked evidence", "No controlled efficacy result", "Funding by programme not disclosed", "G12D only", "High combined grade 3+ toxicity", "Patent-family match is probable rather than directly product-code-labelled", "Exact national claim scope and freedom to operate require counsel" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "KRAS-G12D is a direct driver target and the clinical response signal is consistent with target action, but the drug contribution cannot yet be separated from chemotherapy." }, "animalEvidence": { "rating": "moderate", "reason": "The development package and patent families report non-human support, but this company record does not rely on an independently reproduced animal effect." }, "inVitroEvidence": { "rating": "strong", "reason": "The compound and formulation families report target and ERK-pathway activity, although sponsor-generated laboratory results do not establish comparative clinical delivery." }, "reproducibility": { "rating": "weak", "reason": "The clinical signal is small, single arm and not independently repeated; the ongoing blinded phase 3 trial is the first decisive check." }, "commercialConflicts": { "rating": "high", "reason": "Hengrui sponsors the programme, employs contributors and appears to control the probable compound and formulation families." }, "recency": { "rating": "current", "reason": "The clinical report, phase 3 record, company filing and relevant patent applications are from 2024 to 2026." } } }, "layout": { "x": 94, "y": 111 }, "updated": "2026-09-14" }, { "id": "company-histosonics", "title": "HistoSonics: Edison system and pancreatic evidence dependency", "shortTitle": "HistoSonics", "type": "company", "status": "supported", "scope": [ "pdac", "liver-cancer", "focused-ultrasound", "medical-device" ], "summary": "HistoSonics controls the Edison device and sponsors the first pancreatic trial; independent imaging, pulse and screening data are the price of mission support.", "content": "HistoSonics makes the Edison System, which combines a therapy transducer, ultrasound imaging, robotic positioning, treatment planning and a water coupling path. FDA's De Novo classification authorizes non-invasive mechanical destruction of liver tumours. The same FDA document says the device was not evaluated for a specific disease and its cancer-treatment safety and effectiveness were not established. HistoSonics supplied and funded the nine-pig pancreatic safety study; several authors were employees or consultants. It is also the lead sponsor of GANNON, the first registered pancreatic adenocarcinoma feasibility study. This is not evidence of wrongdoing. It is a single-owner dependency across hardware, software, training, raw pulse records, imaging, animal translation, trial design and publication. The mission decision is £0 in company equity and £0 in equipment. Up to £400,000 of independent monitoring support can unlock only after continuous imaging and device logs, the full screened denominator, treatment plans, harms, outcomes, audit trail, pause authority and negative-result publication rights are contracted. The public language must preserve three boundaries: FDA liver authorization is not pancreas authorization; tissue destruction is not cancer control; a selected single-arm safety study is not evidence of added survival.", "sourceIds": [ "gannon-2026-pancreas-histotripsy-safety", "clinicaltrials-nct06282809-gannon", "fda-den220087-edison" ], "links": [ { "target": "companies", "relation": "part-of" }, { "target": "claim-histotripsy-pancreas-visibility-limited", "relation": "supplies" }, { "target": "failure-histotripsy-acoustic-window-and-translation", "relation": "involves" }, { "target": "unknown-histotripsy-human-coverage-and-benefit", "relation": "constrained-by" }, { "target": "hypothesis-gannon-independent-safety-gate", "relation": "enables" }, { "target": "trial-pdac-histotripsy-action-map", "relation": "involves" }, { "target": "patent-landscape", "relation": "requires-review-by" } ], "evidence": { "humanProspective": "weak — liver feasibility and recruiting pancreas trial", "humanRetrospective": "none for pancreatic treatment", "randomised": "none", "independentReplication": "weak", "sampleSize": "Pancreas animal n=9; pancreatic human target n=50; FDA liver clinical study n=44", "effect": "Operational dependency map; no company valuation or pancreatic efficacy claim", "limits": [ "Ownership and finance were not reviewed", "Contracts and raw logs were not obtained", "Liver and pancreas are different operating regions", "No human pancreas result", "FDA indication wording is narrow", "The independent-monitoring offer is not negotiated" ], "dimensions": { "biologicalPlausibility": { "rating": "not-applicable", "reason": "This record maps control of the device and evidence chain." }, "animalEvidence": { "rating": "moderate", "reason": "The company-supported pig study exposes both promise and failure." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Bench performance does not settle data access or clinical benefit." }, "reproducibility": { "rating": "weak", "reason": "Independent pancreatic data are not yet available." }, "commercialConflicts": { "rating": "high", "reason": "One company controls the platform and sponsors the first pancreas trial." }, "recency": { "rating": "current", "reason": "FDA, animal and July 2026 trial records were checked in this pass." } } }, "layout": { "x": 838, "y": 762 }, "updated": "2026-09-15" }, { "id": "company-igea", "title": "IGEA: pancreatic electrochemotherapy device and data dependency", "shortTitle": "IGEA", "type": "company", "status": "supported", "scope": [ "pdac", "pan-cancer", "drug-delivery", "device-governance" ], "summary": "IGEA supplies the Cliniporator generator, electrodes and planning software used across the pancreatic programme; that concentration makes independent device data and publication rights mandatory.", "content": "IGEA S.p.A. makes the Cliniporator VITAE pulse generator, compatible electrodes and PULSAR planning software used in the pancreatic electrochemotherapy studies. Its technical sheet says the system measures voltage and current waveforms and stores treatment records internally. Those files are the closest available check that the planned field was actually delivered; image response alone cannot substitute for them. The 25-person study used IGEA fixed and variable electrodes and its planning software. In the 90-person randomized protocol, two authors worked for IGEA, the device and software were IGEA products, and project funding included IGEA. An official 2024 institutional report says IGEA funded trial insurance. This is not evidence of misconduct. It is a concentrated control path across device, consumables, planning, waveform data, trial support and interpretation. The mission investment decision is £0 in company equity and £0 in equipment. Any analysis support requires export of raw pulse and planning files, a complete data dictionary, independent imaging and statistics, publication of negative results, service and consumable pricing, and permission to compare other delivery systems. If those rights are refused, the evidence route stops.", "sourceIds": [ "izzo-2021-pancreatic-ect-phase12", "izzo-2021-laparoscopic-ect-protocol", "eudract-2018-003925-27-ect", "igea-2026-cliniporator-vitae" ], "links": [ { "target": "companies", "relation": "part-of" }, { "target": "claim-electrochemotherapy-local-effect-without-patient-benefit", "relation": "supplies" }, { "target": "failure-electrochemotherapy-endpoint-and-trial-drift", "relation": "involves" }, { "target": "unknown-electrochemotherapy-added-benefit-and-delivery", "relation": "constrained-by" }, { "target": "hypothesis-electrochemotherapy-randomized-data-gate", "relation": "enables" }, { "target": "trial-pdac-electrochemotherapy-action-map", "relation": "involves" }, { "target": "patent-landscape", "relation": "requires-review-by" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "not-applicable", "randomised": "weak — company-linked protocol only", "independentReplication": "weak", "sampleSize": "Technology used in pancreatic cohorts n=25 and n=10 and planned randomized n=90", "effect": "Operational dependency map; no valuation or efficacy claim", "limits": [ "Private-company ownership and finance were not established", "No complete contract or patent-family review", "Device logs have not been obtained", "Trial support does not prove sponsor control", "Cliniporator use outside pancreas does not establish pancreatic benefit", "Service and consumable prices are not public" ], "dimensions": { "biologicalPlausibility": { "rating": "not-applicable", "reason": "This record maps ownership and operating control." }, "animalEvidence": { "rating": "not-applicable", "reason": "Animal evidence does not reduce device and data dependency." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Bench performance does not settle access or governance." }, "reproducibility": { "rating": "weak", "reason": "Independent access to raw pulse and planning files has not been shown." }, "commercialConflicts": { "rating": "high", "reason": "One supplier spans generator, electrodes, planning, records, authors and trial support." }, "recency": { "rating": "current", "reason": "Current manufacturer material and transitioned trial records were checked in 2026." } } }, "layout": { "x": 766, "y": 690 }, "updated": "2026-09-15" }, { "id": "company-innovent-biologics", "title": "Innovent Biologics — funded CLDN18.2 platform with an unanswered PDAC survival test", "shortTitle": "Innovent Biologics", "type": "company", "status": "mixed", "scope": [ "pdac", "pan-cancer" ], "summary": "Innovent owns Greater China IBI343 development and IBI389, has ample disclosed capital, and already funds the required randomized IBI343 answer.", "content": "Products. Innovent developed IBI343, a CLDN18.2-addressed exatecan antibody-drug carrier, and IBI389, a CLDN18.2/CD3 antibody that recruits T cells. IBI343 has a phase 1 marker-response gradient in pancreatic cancer and a recruiting 201-person masked phase 3 survival test. IBI389 has a preliminary 7/23 pancreatic response result, but the later full paper omits mature efficacy for its 73 pancreatic patients. Rights. Innovent retains IBI343 rights in Greater China. Takeda holds exclusive development, manufacturing and commercial rights outside Greater China, creating a solvent global production route but splitting data and decision control across two companies. Finance. Innovent reported first-half 2026 revenue of RMB8.6bn, IFRS profit about RMB1.3bn and cash reserves of RMB30.2bn at 31 July. The company does not need mission capital to complete an existing 201-person test. Evidence risk. The phase 3 registry does not state its marker cutoff, power assumptions or data-sharing route and currently lists one Chinese site. The IBI389 registry is stale after planned completion. Commercial strength therefore coexists with an evidence-access problem. Decision. Invest £0 in the company and £0 in a duplicate cohort. Request the IBI343 protocol, dated analysis plan, assay rule and participant-data route before outcomes, plus every mature pancreatic IBI389 result. Company or product value rises only if the all-randomized IBI343 survival and quality result passes and full negative as well as positive data become reusable.", "sourceIds": [ "yu-2025-ibi343-pdac", "clinicaltrials-nct05458219-ibi343", "clinicaltrials-nct07066098-ibi343-phase3", "hao-2024-ibi389-pdac", "xu-2025-ibi389-phase1", "clinicaltrials-nct05164458-ibi389", "innovent-2026-interim-results", "takeda-2025-ibi343-license" ], "links": [ { "target": "companies", "relation": "part-of" }, { "target": "investment", "relation": "part-of" }, { "target": "claim-cldn182-adc-selected-response", "relation": "develops" }, { "target": "claim-cldn18-bispecific-early", "relation": "develops" }, { "target": "unknown-cldn182-adc-survival-effect", "relation": "must-resolve" }, { "target": "trial-cldn182-antibody-action-map", "relation": "controls-data-for" }, { "target": "hypothesis-cldn182-adc-survival-test", "relation": "tests" }, { "target": "company-astellas", "relation": "learns-from" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "not-applicable", "randomised": "none", "independentReplication": "none", "sampleSize": "IBI343 PDAC phase 1 n=83; active phase 3 planned n=201; IBI389 mature PDAC exposure n=73", "effect": "Strong capital and development capacity; IBI343 survival effect unreported and IBI389 mature PDAC efficacy missing", "limits": [ "Finance can change", "Issuer-authored finance source", "Data split across Innovent and Takeda", "Phase 3 design details unavailable", "No independent PDAC replication" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The lead product joins a measurable address with targeted payload delivery and a human response gradient." }, "animalEvidence": { "rating": "moderate", "reason": "Models support both products but should not determine company value now." }, "inVitroEvidence": { "rating": "strong", "reason": "The company can measure binding, uptake, payload effect and T-cell engagement." }, "reproducibility": { "rating": "weak", "reason": "Neither pancreatic product has independent controlled efficacy." }, "commercialConflicts": { "rating": "high", "reason": "Innovent owns the Greater China programme and controls the main clinical records." }, "recency": { "rating": "current", "reason": "Clinical, rights and finance records were checked through 15 September 2026." } } }, "layout": { "x": 228, "y": 146 }, "updated": "2026-09-15" }, { "id": "company-jacobio", "title": "Jacobio — Chinese pan-KRAS programme licensed to AstraZeneca before PDAC proof", "shortTitle": "Jacobio", "type": "company", "status": "open", "scope": [ "pdac", "colorectal", "pan-cancer" ], "summary": "JAB-23E73 entered a Chinese first-line pancreatic phase Ib/III; the AstraZeneca deal has $100m upfront and up to $2.015bn total potential value.", "content": "Founded in 2015 and chaired by Yinxiang Wang, Jacobio has an approved Chinese KRAS-G12C product and a separate oral pan-KRAS inhibitor, JAB-23E73. A Chinese phase Ib/III trial combines JAB-23E73 with gemcitabine and nab-paclitaxel in first-line KRAS-mutant pancreatic cancer; the first-patient announcement provides no pancreatic efficacy. AstraZeneca licensed and collaborates on the asset. The company reports $100m upfront and $2.015bn total potential deal value; the $1.915bn difference is contingent and must not be counted as deployed capital. Advantage: Chinese clinical network, oral broad-KRAS route and a large pharmaceutical partner. Weakness: no peer-reviewed PDAC result or registered effect yet, company-only early data and reliance on milestone success. This is a good example of capital moving ahead of contribution proof, though unlike GRAIL the decisive randomized treatment design can arrive sooner.", "sourceIds": [ "jacobio-2026-annual-results", "jacobio-2026-jab23e73-first-patient" ], "links": [ { "target": "companies", "relation": "part-of" }, { "target": "company-revolution-medicines", "relation": "competes-with" }, { "target": "company-hengrui", "relation": "competes-with" }, { "target": "investment", "relation": "licensed-to-astrazeneca" }, { "target": "unknown-durable-ras-control", "relation": "may-address" } ], "evidence": { "humanProspective": "none — planned study, no result", "humanRetrospective": "not-applicable", "randomised": "none — planned trial, no result", "independentReplication": "none", "sampleSize": "42 pan-cancer later-line patients reported by company; no PDAC efficacy cohort", "effect": "No pancreatic efficacy result", "limits": [ "Company sources", "Trial identifier and full protocol not independently captured", "Milestones are contingent", "No peer-reviewed PDAC data", "Combination contribution will require the randomized phase" ] }, "layout": { "x": 110, "y": 113 }, "updated": "2026-09-14" }, { "id": "company-jazz-pharmaceuticals", "title": "Jazz Pharmaceuticals — licensed HER2 bispecific with response-based biliary approval", "shortTitle": "Jazz Pharmaceuticals", "type": "company", "status": "supported", "scope": [ "cholangiocarcinoma", "gallbladder", "colorectal", "hepatobiliary", "pan-cancer" ], "summary": "Jazz turned Zymeworks' zanidatamab into an approved biliary product, but the approval rests on single-arm response and confirmatory survival proof remains due.", "content": "Technology and origin: zanidatamab binds two HER2 epitopes and was created by Zymeworks. Jazz licensed most global rights in 2022. Disclosed deal terms were $50m upfront, a $325m opt-in and up to $1.385bn further milestones; the $1.76bn total potential value is not research spend. Stage and trial: Ziihera received accelerated US approval for previously treated unresectable or metastatic IHC 3+ HER2-positive biliary cancer. In the 80-person single-arm HERIZON-BTC-01 study, the IHC 3+ group had 51.6% response and median survival 18.1 months, while IHC 2+ disease had 5.6% response and 5.2-month survival. This is strong selection evidence but not a randomized survival effect. Commercial position: Jazz reported $15m Ziihera biliary sales in Q2 2026 and company-wide 2026 R&D guidance of $818m–$873m; neither figure reveals biliary programme cost. Investors and financing: Jazz is a public company, and the licence rather than a disease-specific venture round is the observable capital event. Advantage: approved product, commercial reach and a marker with a large response split. Weakness: mixed biliary primary sites, extensive sponsor interests, accelerated approval and a confirmatory trial still needed. Mission role: use it as proof that protein-expression threshold can sharpen classification, not as proof that HER2 treatment cures biliary cancer.", "sourceIds": [ "pant-2026-zanidatamab-final", "fda-2024-zanidatamab-approval", "jazz-zymeworks-2022-license", "jazz-2026-q2" ], "links": [ { "target": "companies", "relation": "part-of" }, { "target": "claim-zanidatamab-her2-biliary-selection", "relation": "commercialises" }, { "target": "cholangiocarcinoma", "relation": "treats-subgroup" }, { "target": "gallbladder", "relation": "treats-subgroup" }, { "target": "investment", "relation": "contains" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "weak", "randomised": "none", "independentReplication": "weak", "sampleSize": "80 in pivotal single-arm biliary cohort; company licensing, approval and Q2 2026 commercial records", "effect": "IHC 3+ response 51.6% and median OS 18.1 months; Q2 2026 biliary sales $15m", "limits": [ "Single-arm response-based approval", "Confirmatory survival benefit pending", "Mixed biliary sites", "Company-funded", "Deal milestones are mostly contingent", "Company-wide R&D not disease-specific" ] }, "updated": "2026-09-14" }, { "id": "company-mainz-biomed", "title": "Mainz Biomed — a financing-constrained pivot into pancreatic detection", "shortTitle": "Mainz Biomed", "type": "company", "status": "open", "scope": [ "pdac", "colorectal", "pan-cancer" ], "summary": "The company licensed pancreatic mRNA markers and retained an early stool test while winding down colorectal operations; $6m mainly preserved runway.", "content": "Technology: a March 2025 licence from Liquid Biosciences for mRNA biomarkers and an algorithm for a pancreatic blood test, plus an earlier PCR multiplex stool candidate called PancAlert. Stage: early development, with no prospective pancreatic screening result in the filing. Capital and condition: a February 2026 $6m convertible preferred placement in two $3m tranches was described as funding operations, liabilities and strategic alternatives. The 2025 annual report says the company wound down colorectal products and its German subsidiary and had roughly 13 staff in March 2026. Advantage: access to two non-invasive sample routes and an exclusive blood-test licence. Weakness: no clinical utility, no independently validated pancreatic accuracy, limited organisation and financing pressure. This is not evidence of a coming screening product; it is a case where commercial survival and a scientific programme can be mistaken for one another.", "sourceIds": [ "mainz-2026-10k", "mainz-2026-private-placement" ], "links": [ { "target": "companies", "relation": "part-of" }, { "target": "company-cosomil", "relation": "competes-with" }, { "target": "company-grail", "relation": "competes-with" }, { "target": "claim-case-control-blood-tests-overstate-screening-readiness", "relation": "must-overcome" }, { "target": "investment", "relation": "financed-by" } ], "evidence": { "humanProspective": "none", "humanRetrospective": "not-applicable — not assessed in this record", "randomised": "none", "independentReplication": "none", "sampleSize": "No pancreatic validation cohort reported in the cited filing", "effect": "No human utility result", "limits": [ "Company filing only", "Early development", "Financing pressure", "Small workforce", "No prospective accuracy", "No pancreatic clinical utility" ] }, "layout": { "x": 116, "y": 119 }, "updated": "2026-09-14" }, { "id": "company-natera", "title": "Natera — scaled MRD detection with action-specific utility and concentrated operating control", "shortTitle": "Natera", "type": "company", "status": "supported", "scope": [ "pdac", "colorectal", "pan-cancer" ], "summary": "Signatera has scale and strong risk separation, but randomized colorectal results show that utility depends on the action, while PDAC still has no completed action proof.", "content": "Technology: Signatera builds a personalized blood assay from mutations in each patient's tumour and is marketed for residual-disease monitoring. Competitive advantage: commercial scale, longitudinal testing, a large linked tumour-and-blood dataset and a low-frequency assay designed for residual rather than bulky disease. The 2025 annual report says the product is a laboratory-developed test, is not a therapy selector and has broad but indication-specific coverage. PDAC evidence: a 298-patient retrospective series found very large recurrence-risk separation, but company employees and equity holders were involved. Colorectal evidence now tests actions: DYNAMIC stage II supports one treatment-omission rule, DYNAMIC-III and ALTAIR reject two other treatment routes, FIND supports earlier curative-intent treatment without mature survival, and CIRCULATE-US remains open. Capital: Natera reported $320.678m, $404.138m and $624.110m of company-wide R&D in 2023-2025, $1.348926bn in total. The filing does not isolate oncology, Signatera, PDAC or colorectal cancer, so this cannot be counted as disease-specific MRD investment. Operating control is concentrated. Natera reports more than 650 patents across the company, clinical Signatera processing in its own laboratories, molecular tests validated only on Illumina sequencing, sole-sourced sequencers and reagents through August 2033, and companion-diagnostic submissions still under review. Main weakness: a negative result may reflect low shedding, while a positive result has no proven treatment action in PDAC. Commercial testing can spread faster than utility evidence and create surveillance anxiety or toxic treatment without benefit. Mission role: contract for the full access stack and use the platform only inside a randomized action trial with an independent assay comparator, frozen timing and algorithm rules, supply continuity, and complete capture of false negatives, lead time, quality of life and survival.", "sourceIds": [ "natera-2025-10k", "natera-2026-signatera-pdac", "patent-wo2019200228-natera-mrd", "botta-2024-tumor-informed-ctdna", "dynamic-pancreas-registry", "tie-2022-dynamic-colon", "tie-2025-dynamic3", "bando-2026-altair", "mo-2026-find-crc-surveillance", "clinicaltrials-nct05174169-circulate-us" ], "links": [ { "target": "companies", "relation": "part-of" }, { "target": "unknown-ctdna-actionability", "relation": "developed-test-in" }, { "target": "hypothesis-multimodal-mrd-action", "relation": "could-test" }, { "target": "company-grail", "relation": "differs-from" }, { "target": "programme-100m-randomized-platform", "relation": "could-partner" }, { "target": "claim-commercial-capital-ledger-boundaries", "relation": "part-of" }, { "target": "claim-mrd-platform-access-is-more-than-patent-license", "relation": "controls" }, { "target": "claim-crc-ctdna-utility-is-action-specific", "relation": "constrained-by" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "strong", "randomised": "none", "independentReplication": "moderate", "sampleSize": "PDAC series n=298 with 1,329 samples; randomized colorectal action studies n=243 to 968; one audited annual filing", "effect": "Postoperative or surveillance positivity strongly predicts recurrence; colorectal utility is action specific and PDAC action benefit is unproved", "limits": [ "Retrospective PDAC evidence", "Company employees and equity interests", "Commercial laboratory-developed test", "Low shedding can create false reassurance", "No completed randomized PDAC action trial", "Company-wide R&D and patent totals are not oncology or PDAC estimates", "Supply and regulatory statements are company-reported", "No mission licence terms obtained" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Tumour-informed residual DNA can mark hidden disease, but the effective action and low-shedding states remain separate questions." }, "animalEvidence": { "rating": "not-applicable", "reason": "Company ranking depends on human assay, utility and operating evidence." }, "inVitroEvidence": { "rating": "moderate", "reason": "Analytical sensitivity supports the platform but does not establish clinical utility." }, "reproducibility": { "rating": "moderate", "reason": "Risk separation is widely reported, while randomized action results are mixed and PDAC utility is absent." }, "commercialConflicts": { "rating": "high", "reason": "Natera owns the test, employs study authors, operates the laboratories and benefits from expanded testing." }, "recency": { "rating": "current", "reason": "The assessment includes the 2026 annual filing, action results and September 2026 trial status." } } }, "layout": { "x": 113, "y": 94 }, "updated": "2026-09-14" }, { "id": "company-novartis-fap-radioligand", "title": "Novartis — active FAP-radioligand developer, not yet a PDAC-efficacy company", "shortTitle": "Novartis FAP radiation", "type": "company", "status": "active", "scope": [ "pdac", "pan-cancer", "transfer-unproven" ], "summary": "Novartis sponsors the only large active PDAC-specific FAP-radioligand programme found, while current evidence establishes dose and safety rather than controlled benefit.", "content": "Technology and stage. Novartis sponsors LuMIERE, which pairs 68Ga-FAP-2286 PET selection with 177Lu-FAP-2286 treatment. The phase 1 paper established a 9.25-GBq recommended phase 2 dose after 27 treated people. The live non-randomized registry estimates 222 participants and includes pancreatic-cancer monotherapy after one or two advanced-disease regimens plus a first-line combination with modified FOLFIRINOX. Evidence boundary. Nine phase 1 patients had pancreatic cancer; four had stable disease and no pancreatic partial response was reported. Tumour-lesion dosimetry was not reported, and scan uptake could not be robustly related to response. The sponsor funded the study and participated in design, collection, analysis, publication decisions and manuscript work. Likely advantage: a running multicentre programme, isotope operations, paired scan-and-treatment asset and an explicit patient-data request route. Likely weakness: no randomized control, no proved pancreatic treatment effect and sponsor control over the most useful dose-response data. Mission stance: seek patient- and lesion-level access under a contract that permits independent analysis and negative publication. Treat Novartis as a possible data and drug partner, not an investment conclusion or source of independent validation.", "sourceIds": [ "mcconathy-2026-lumiere-phase1", "clinicaltrials-nct04939610-lumiere" ], "links": [ { "target": "companies", "relation": "part-of" }, { "target": "trial-pdac-fap-radioligand-action-map", "relation": "sponsors" }, { "target": "hypothesis-fapi-theranostic-pdac", "relation": "controls-data-for" }, { "target": "claim-fap-radioligand-feasible-pdac-efficacy-unproven", "relation": "supports" }, { "target": "failure-fapi-scan-dose-benefit-collapse", "relation": "must-repair" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "none needed for company-stage claim", "randomised": "none", "independentReplication": "weak", "sampleSize": "LuMIERE n=27 treated phase 1; n=222 estimated full programme", "effect": "Recommended phase 2 dose established; no controlled PDAC benefit", "limits": [ "No valuation or investment terms assessed", "No independent patient-level analysis", "Trial remains active", "Company role is inferred only from named sponsorship and paper disclosures" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "Paired imaging and therapy establish target reach but not patient benefit." }, "animalEvidence": { "rating": "moderate", "reason": "Model support is no longer the main stage limit." }, "inVitroEvidence": { "rating": "moderate", "reason": "Binding supports the asset but cannot rank the company clinically." }, "reproducibility": { "rating": "weak", "reason": "The company has one current platform without repeated PDAC efficacy." }, "commercialConflicts": { "rating": "high", "reason": "The sponsor has direct product, data and publication interests." }, "recency": { "rating": "current", "reason": "Registry verified July 2026 and checked 15 September 2026." } } }, "layout": { "x": 89, "y": 153 }, "updated": "2026-09-15" }, { "id": "company-novocure", "title": "Novocure", "shortTitle": "Novocure", "type": "company", "status": "supported", "scope": [ "pdac", "device-owner", "post-approval" ], "summary": "Novocure owns the only physical device with randomized pancreatic survival evidence and has enough cash to fund its own replication; the mission should buy independent data and value, not company risk.", "content": "Novocure was incorporated in 2000 and develops wearable tumour-treating field devices for solid cancers. Optune Pax delivers 150 kHz alternating electric fields across the abdomen. FDA approved it in February 2026 for adults with locally advanced pancreatic cancer together with gemcitabine and nab-paclitaxel; a European CE certificate followed in June. The company's defensible asset is not the general claim that electric fields affect dividing cells. It is the device, array supply and training route, regulatory approvals, operating data, patents and pivotal trial record. Its pancreatic advantage is unusually strong for a device: PANOVA-3 produced a randomized overall-survival gain in 571 people. Its weaknesses are one sponsor-run pivotal result, no established benefit on another chemotherapy backbone, high skin and daily-use burden, and value models far above normal payer thresholds. PANOVA-4 does not repair the evidence boundary because it added both the device and atezolizumab without a current randomized control. Capital is not the bottleneck. At 30 June 2026 Novocure reported $440.6m in cash, cash equivalents and short-term investments, $200m borrowed principal, $357.639m six-month revenue, $86.796m six-month net loss and $5.385m six-month operating cash use. The launch was early: the United States had 285 active Optune Pax patients at quarter end, 586 first-half prescriptions and $1.6m recognized product revenue in the second quarter; one German prescription made 587 worldwide. The filing says available cash can fund at least twelve months under the business plan. Novocure also has a Greater China licence and collaboration with Zai Lab. Mission decision: invest £0 in equity, provide £0 product subsidy and fund no company efficacy cohort. Offer at most £500,000 only for independent PANOVA-3 data analysis and all-eligible care-route measurement with public negative-result rights. Require the company to fund modern-backbone replication and any label expansion. Procurement should be outcome- and price-gated, not an investment thesis.", "sourceIds": [ "fda-2026-optune-pax-ssed", "clinicaltrials-nct03377491-panova3", "macarulla-2025-panova3", "clinicaltrials-nct06390059-panova4", "novocure-2026-panova4-topline", "novocure-2026-q2-10q", "guzauskas-2026-ttfields-cost", "liu-2026-ttfields-cost" ], "links": [ { "target": "companies", "relation": "part-of" }, { "target": "claim-panova3-ttfields-small-os-benefit", "relation": "commercializes" }, { "target": "regulatory-fda-optune-pax-2026", "relation": "holds" }, { "target": "trial-pdac-ttfields-action-map", "relation": "sponsors" }, { "target": "failure-panova4-component-and-historical-control", "relation": "controls" }, { "target": "hypothesis-optune-pax-independent-value-gate", "relation": "constrained-by" }, { "target": "investment", "relation": "constrains" } ], "evidence": { "humanProspective": "not-applicable", "humanRetrospective": "not-applicable", "randomised": "not-applicable", "independentReplication": "strong", "sampleSize": "$440.6m liquid resources; $357.639m six-month revenue; 285 U.S. Optune Pax active patients at quarter end", "effect": "Equity and product subsidy £0; independent analysis and route work capped at £500,000", "limits": [ "Product-specific cost and price are not fully disclosed", "Operating cash use is affected by working capital and non-cash pay", "Launch revenue is immature", "Patent-by-patent freedom to operate was not audited", "Company filing is self-reported", "This is programme allocation, not personal investment advice" ], "dimensions": { "biologicalPlausibility": { "rating": "not-applicable", "reason": "Linked pages assess the device mechanism and patient effect." }, "animalEvidence": { "rating": "not-applicable", "reason": "This page assesses ownership, capital and evidence control." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "This page assesses ownership, capital and evidence control." }, "reproducibility": { "rating": "strong", "reason": "FDA, trial, paper and SEC records agree on ownership, approval and commercial state." }, "commercialConflicts": { "rating": "high", "reason": "Novocure owns the product, sponsored the trials and funded one value model." }, "recency": { "rating": "current", "reason": "The capital decision uses the July 2026 filing and September 2026 registry state." } } }, "layout": { "x": 616, "y": 540 }, "updated": "2026-09-15" }, { "id": "company-oncosil-medical", "title": "OncoSil Medical: phosphorus-32 product and randomized-data dependency", "shortTitle": "OncoSil Medical", "type": "company", "status": "supported", "scope": [ "pdac", "brachytherapy", "medical-device", "device-governance" ], "summary": "OncoSil Medical makes the microparticle device, funded and staffed the pilot and sponsors TRIPP-FFX; access to all 88 assignments is the only current investment case.", "content": "OncoSil Medical makes OncoSil, silicon microparticles carrying phosphorus-32 for endoscopic implantation into pancreatic tumours. The company funded the ten-centre PanCO pilot and its medical writing; three authors were employees and shareholders, and other investigators disclosed consulting relationships. The company is the sponsor of the 88-person randomized TRIPP-FFX study. These facts do not invalidate the data and do not show misconduct. They create a concentrated control path across product supply, dosimetry, training, trial sponsorship, data and publication. The pilot supports technical capability: 40/42 early scans localized the radiation, and multicentre teams delivered the procedure. It does not establish survival benefit. The mission decision is £0 in equity, £0 in device purchase and at most £350,000 for independent analysis only after the full assignment key, dose files, images, chemotherapy records, quality of life, harms, surgery and survival are exportable. Contracts must grant a frozen independent analysis, publication of null or harmful results, a later survival update and freedom to compare other local-radiation systems. If the company refuses, this route stops; if the result passes, commercial diligence starts only after the clinical effect survives independent review.", "sourceIds": [ "ross-2022-panco-p32", "clinicaltrials-nct05466799-tripp-ffx", "lim-2026-p32-landmark" ], "links": [ { "target": "companies", "relation": "part-of" }, { "target": "claim-oncosil-local-control-signal-awaits-randomized-result", "relation": "supplies" }, { "target": "failure-p32-liquefaction-survival-and-reporting-gap", "relation": "involves" }, { "target": "unknown-oncosil-randomized-survival-and-delivery", "relation": "constrained-by" }, { "target": "hypothesis-tripp-ffx-independent-analysis-gate", "relation": "enables" }, { "target": "trial-pdac-p32-action-map", "relation": "involves" }, { "target": "patent-landscape", "relation": "requires-review-by" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "moderate", "randomised": "moderate — sponsored dataset unreported", "independentReplication": "weak", "sampleSize": "Pilot n=50; adjusted comparison n=86; sponsored randomized n=88", "effect": "Operational dependency map; no valuation or efficacy claim", "limits": [ "Ownership and balance sheet were not reviewed", "Raw trial data were not obtained", "Commercial approval scope was not independently mapped", "Trial result may properly await final follow-up", "No contract has been negotiated", "Company ties do not prove bias in results" ], "dimensions": { "biologicalPlausibility": { "rating": "not-applicable", "reason": "This record maps control of the product and evidence chain." }, "animalEvidence": { "rating": "not-applicable", "reason": "Animal evidence does not reduce sponsor and data dependency." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Bench dose measurements cannot establish access or benefit." }, "reproducibility": { "rating": "moderate", "reason": "Multicentre delivery is shown, but comparative analysis is not public." }, "commercialConflicts": { "rating": "high", "reason": "One company spans product, pilot funding, authorship and randomized sponsorship." }, "recency": { "rating": "current", "reason": "Company role is documented in the current randomized registry." } } }, "layout": { "x": 934, "y": 858 }, "updated": "2026-09-15" }, { "id": "company-panbela-therapeutics", "title": "Panbela Therapeutics", "shortTitle": "Panbela", "type": "company", "status": "contested", "scope": [ "pdac", "trial-delivery", "transfer-unproven" ], "summary": "Panbela owns ivospemin but lacked cash to maintain its trial contractor; the mission should buy data diligence, not equity or the product.", "content": "Panbela sponsored the 50-person ivospemin phase 1 and the global ASPIRE survival trial. Its last linked quarterly filing reported $142,000 cash at 30 September 2024, $12.5m operating cash use over nine months, a $15.0m working-capital deficit and an $18.2m shareholder deficit. Simple unchanged-use math gives $12.5m / 9 = $1.39m monthly use and $0.142m / $1.39m = 0.102 months, about 3.1 days of cash. This is only a stress indicator: financing arrives unevenly, liabilities are not cash use and the company later borrowed. The operational fact is stronger: its trial contractor began ending the contract after unpaid invoices and Panbela took direct responsibility. In January 2025 Panbela disclosed default on $12.0m of convertible notes; while default continued, interest became monthly SOFR plus 12% and the lender could demand immediate payment. In September 2025 the SEC declared a separate registration statement abandoned after no response to notice. These records do not prove insolvency, drug failure or loss of the data. They do remove any rational case for buying public equity before custody and liabilities are known. Invest £0 in Panbela equity, £0 in an ivospemin licence and £0 in a new trial. Offer at most £250,000 under an independent data-custody contract, paid by verified deliverable. Any asset rescue needs a clean title, itemized site and vendor claims, regulator agreement, retained assignment blind, public negative-result rights and equal displacement within the £500m ceiling.", "sourceIds": [ "panbela-2024-q3-10q", "panbela-2025-note-default", "sec-2025-panbela-registration-abandoned", "clinicaltrials-nct05254171-aspire", "ctis-2024-514714-12-00-aspire" ], "links": [ { "target": "companies", "relation": "part-of" }, { "target": "failure-aspire-trial-operational-collapse", "relation": "owns" }, { "target": "trial-pdac-polyamine-action-map", "relation": "sponsors" }, { "target": "hypothesis-aspire-data-rescue-gate", "relation": "constrained-by" }, { "target": "investment", "relation": "constrains" } ], "evidence": { "humanProspective": "not-applicable", "humanRetrospective": "not-applicable", "randomised": "not-applicable", "independentReplication": "strong", "sampleSize": "$142,000 cash; $12.5m nine-month operating use; $15.0m working-capital deficit; $12.0m note default", "effect": "Equity, licence and new-trial decision £0; data diligence capped at £250,000", "limits": [ "Latest linked full financial quarter ended September 2024", "Later private financing may exist", "SEC registration abandonment is not a bankruptcy finding", "Trial data ownership and vendor claims are unknown", "This is programme allocation, not personal investment advice" ], "dimensions": { "biologicalPlausibility": { "rating": "not-applicable", "reason": "Drug biology is assessed in linked evidence." }, "animalEvidence": { "rating": "not-applicable", "reason": "This page assesses operating capacity and ownership risk." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "This page assesses operating capacity and ownership risk." }, "reproducibility": { "rating": "strong", "reason": "SEC filings and two trial registries show the financing break and public-state conflict." }, "commercialConflicts": { "rating": "high", "reason": "Panbela owns the product and controls or controlled material trial records." }, "recency": { "rating": "current", "reason": "The linked record extends through the September 2025 SEC order and passed trial milestones." } } }, "layout": { "x": 514, "y": 438 }, "updated": "2026-09-15" }, { "id": "company-paq-therapeutics", "title": "PAQ Therapeutics — $77m financed pan-KRAS degradation before human efficacy", "shortTitle": "PAQ Therapeutics", "type": "company", "status": "open", "scope": [ "pdac", "colorectal", "pan-cancer" ], "summary": "PT0511 entered a 195-person phase 1 across KRAS-altered tumours; the company has financing and a trial, but no human response data.", "content": "Technology: targeted protein degradation of multiple KRAS forms by intravenous PT0511, with a separate G12D degrader PT0253. Stage: first patient dosed in late 2025; NCT07300150 is an open-label phase 1 with estimated enrolment 195 across pancreatic, colorectal, lung and other tumours and a 2028 primary completion. Capital: the company reported that its Series B and extension totalled $77m; exact allocation by asset and investor terms are not public in the announcement. Advantage: degradation may produce deeper or different pharmacology than occupancy inhibitors, and the study includes multiple KRAS forms. Weakness: no human pharmacodynamic or efficacy result, intravenous burden, pan-cancer expansion can hide weak PDAC numbers, and financing is company-reported. It belongs in the newly possible but unproved tier, not beside randomized daraxonrasib evidence.", "sourceIds": [ "paq-2025-series-b-investors", "paq-2026-seriesb-pt0511", "nct07300150-pt0511" ], "links": [ { "target": "companies", "relation": "part-of" }, { "target": "company-astellas", "relation": "competes-with" }, { "target": "company-revolution-medicines", "relation": "competes-with" }, { "target": "investment", "relation": "financed-by" }, { "target": "unknown-durable-ras-control", "relation": "may-address" }, { "target": "investor-paq-series-b-syndicate", "relation": "funded-by" } ], "evidence": { "humanProspective": "none — planned study, no result", "humanRetrospective": "not-applicable", "randomised": "none", "independentReplication": "none", "sampleSize": "Phase 1 estimated n=195 across several tumours", "effect": "No human efficacy result", "limits": [ "Company financing source", "No human efficacy", "No PDAC-specific cohort size disclosed", "Intravenous product", "Completion estimated 2028" ] }, "layout": { "x": 107, "y": 110 }, "updated": "2026-09-14" }, { "id": "company-pfizer-cachexia", "title": "Pfizer — leading the GDF-15 cachexia programme, with benefit beyond weight still unresolved", "shortTitle": "Pfizer cachexia", "type": "company", "status": "supported", "scope": [ "pdac", "colorectal", "pan-cancer" ], "summary": "Pfizer owns ponsegromab, the key antibody and use families, the phase 2 result and a 982-person PDAC trial; this is a strong execution position and a concentrated evidence-access risk.", "content": "Technology: ponsegromab, PF-06946860, is a humanized antibody that binds circulating GDF-15 and blocks its interaction with the hindbrain GFRAL receptor. Evidence: Pfizer funded the 187-person randomized phase 2 that changed weight, symptoms and activity over twelve weeks in a selected mixed-cancer population. Development: NCT06989437 is a recruiting randomized phase 2b/3 in an estimated 982 metastatic-PDAC patients at 220 locations across 21 countries, with primary completion estimated in January 2028. Ownership: WO2020039321A2 is a core anti-GDF15 antibody family with 2018 priority; its PCT entry is marked ceased while national family members include grants. WO2025245200A1 is a pending use-and-dose PCT application with 2024 priority covering ponsegromab in cancer cachexia including pancreatic cancer and systemic chemotherapy. Competitive advantage: the company joins target biology, drug supply, biomarker selection and a global pivotal-scale trial. Main weakness: the strongest completed result is weight at twelve weeks, while the PDAC trial keeps weight and anorexia as primary endpoints and excludes some frail patients. Mission role: do not acquire the company or duplicate the trial. Negotiate timely arm-level data, assay specification, drug access for a narrowly defined follow-on state, negative-result publication and independent analysis. This is an access map, not a legal opinion, valuation or forecast.", "sourceIds": [ "groarke-2024-ponsegromab-cachexia", "clinicaltrials-nct06989437-ponsegromab-pdac", "pfizer-2026-ponsegromab-profile", "patent-wo2020039321-anti-gdf15", "patent-wo2025245200-ponsegromab-cachexia" ], "links": [ { "target": "companies", "relation": "part-of" }, { "target": "claim-ponsegromab-gdf15-cachexia-driver-not-survival", "relation": "developed" }, { "target": "trial-pdac-cachexia-action-map", "relation": "runs-trial-in" }, { "target": "claim-ponsegromab-access-stack", "relation": "controls" }, { "target": "patent-landscape", "relation": "protects" } ], "evidence": { "humanProspective": "strong for weight and symptoms", "humanRetrospective": "not-applicable", "randomised": "strong for intermediate endpoints", "independentReplication": "none for ponsegromab", "sampleSize": "Completed randomized phase 2 n=187; recruiting PDAC phase 2b/3 estimated n=982; two priority patent families", "effect": "Strong target and execution position; survival and treatment-delivery benefit unresolved", "limits": [ "Company sources and company-funded trial", "No completed PDAC-specific efficacy result", "No acquisition or valuation work", "No contract terms", "Patent map is not exhaustive", "National claim scope requires counsel" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "A selected randomized population showed target-consistent weight, symptom and activity changes." }, "animalEvidence": { "rating": "moderate", "reason": "Target development has model support, but company rank is based mainly on human and operational evidence." }, "inVitroEvidence": { "rating": "moderate", "reason": "Antibody binding supports the product mechanism but not the clinical value claim." }, "reproducibility": { "rating": "none", "reason": "No independent randomized ponsegromab replication has reported." }, "commercialConflicts": { "rating": "high", "reason": "Pfizer owns the product, sponsors the studies and controls the key supply and data routes." }, "recency": { "rating": "current", "reason": "The company profile was current in May 2026 and the trial record was updated in August 2026." } } }, "layout": { "x": 145, "y": 69 }, "updated": "2026-09-14" }, { "id": "company-quercis-pharma", "title": "Quercis Pharma — a real PDAC phase 3 built on an uncontrolled marker study", "shortTitle": "Quercis Pharma", "type": "company", "status": "supported", "scope": [ "pdac", "pan-cancer", "transfer-unproven" ], "summary": "Quercis owns or controls the Kinisoquin programme and a 480-person phase 3; its advantage is oral dosing, while efficacy, comparator value and the licence chain remain unproved.", "content": "Technology: Kinisoquin is an oral isoquercetin formulation with vitamins B3 and C in the phase 3 registry, intended to reduce thromboembolic events through PDI and related coagulation routes. Human evidence: the 57-person CATIQ study changed D-dimer, PDI, thrombin generation and P-selectin, with no primary VTE or major hemorrhage, but used sequential open-label doses and no control. Development: NCT06861088 is recruiting an estimated 480 advanced-PDAC patients at three listed US sites, assigns 1,000 mg, 2,000 mg or placebo for 16 weeks, selects one dose after 26 events and estimates primary completion in October 2027. Ownership: US10391096B2 covers an isoquercetin-vitamin composition for venous thrombosis and names Quercis in its history. WO2020112828A1 and US11872241B2 arise from Beth Israel work on cancer thrombotic events; the licence connection is not public in the checked records. Quercis claims patent protection through 2045 and an FDA Special Protocol Assessment; those are company statements. Competitive advantage: oral use without conventional anticoagulation could matter if randomized clot prevention and bleeding safety are real. Main weakness: generic LMWH and direct factor-Xa inhibitors already have randomized VTE-prevention evidence, so Kinisoquin must show clinical value, adherence and cost against available care rather than marker activity. Mission role: monitor and request protocol, supply, licence and patient-level data; do not acquire or fund the company from phase 2 or company claims alone.", "sourceIds": [ "zwicker-2019-isoquercetin-catiq", "clinicaltrials-nct06861088-catiq-p3", "quercis-2026-kinisoquin-profile", "patent-us10391096-kinisoquin", "patent-wo2020112828-isoquercetin-cancer", "frere-2020-pdac-thromboprophylaxis-meta" ], "links": [ { "target": "companies", "relation": "part-of" }, { "target": "claim-kinisoquin-phase2-marker-not-clinical-proof", "relation": "developed" }, { "target": "trial-pdac-vte-action-map", "relation": "runs-trial-in" }, { "target": "claim-kinisoquin-access-stack", "relation": "controls" }, { "target": "person-jeffrey-zwicker", "relation": "works-with" }, { "target": "lab-bidmc-thrombosis-pdi", "relation": "originated-with" }, { "target": "patent-landscape", "relation": "depends-on" } ], "evidence": { "humanProspective": "moderate for marker activity; none for randomized efficacy", "humanRetrospective": "not-applicable", "randomised": "none for completed Kinisoquin evidence", "independentReplication": "none", "sampleSize": "Completed open-label n=57; recruiting phase 3 estimated n=480; two patent families checked", "effect": "Clinical competitive value unresolved; randomized phase 3 in progress", "limits": [ "Private company", "No audited financing", "No completed randomized efficacy", "Company claims", "Licence chain unclear", "No acquisition or valuation work" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "Human marker changes support target activity but do not prove prevention." }, "animalEvidence": { "rating": "moderate", "reason": "Thrombus models support the route but cannot rank the company." }, "inVitroEvidence": { "rating": "strong", "reason": "PDI inhibition and coagulation assays support a measurable pharmacological route." }, "reproducibility": { "rating": "none", "reason": "No independent randomized Kinisoquin clinical effect has reported." }, "commercialConflicts": { "rating": "high", "reason": "The company sponsors the pivotal study and publishes the strongest programme claims." }, "recency": { "rating": "current", "reason": "Company, registry and patent records were checked in September 2026." } } }, "layout": { "x": 164, "y": 88 }, "updated": "2026-09-14" }, { "id": "company-rakuten-medical", "title": "Rakuten Medical: one approved setting, one untested pancreatic transfer", "shortTitle": "Rakuten Medical", "type": "company", "status": "active", "scope": [ "pdac", "photoimmunotherapy", "company", "regulatory" ], "summary": "Rakuten owns the ASP-1929 drug-light platform and collaborates on the first pancreatic study; its Japanese head-and-neck approval is conditional and supplies no pancreatic outcome.", "content": "Rakuten Medical develops ASP-1929 and the Alluminox light platform. The drug links cetuximab, an EGFR-binding antibody, to IR700 dye and requires a matched laser system. Japan approved the pair in 2020 for unresectable locally advanced or recurrent head and neck cancer under conditional early approval. The regulator required a risk plan, surveillance of every treated patient because local study numbers were very limited, later phase 3 data and trained users. Rakuten states that the product remains unapproved outside Japan. It now collaborates on Brown University's first listed pancreatic study, planned for 30 people. This creates one-source dependence across drug, device, training and much of the operating knowledge. The pancreatic investment case is presently zero: no treated outcome, no randomized design and no proven light coverage. The mission should not buy shares, drug or device inventory. A maximum £300,000 measurement contract can be considered only if Brown controls the protocol, an independent monitor holds the data, every enrolled participant is reported and negative results can be published. Head-and-neck sales or response cannot be counted as pancreatic validation.", "sourceIds": [ "rakuten-medical-2026-asp1929", "pmda-2020-akalux-review", "clinicaltrials-nct07698613-asp1929-pancreas" ], "links": [ { "target": "claim-asp1929-pancreatic-human-test-not-started", "relation": "develops" }, { "target": "failure-asp1929-selection-light-and-endpoint-design", "relation": "exposed-to" }, { "target": "unknown-asp1929-pancreatic-target-coverage-and-benefit", "relation": "constrained-by" }, { "target": "hypothesis-asp1929-first-ten-measurement-gate", "relation": "data-requested-by" }, { "target": "trial-pdac-asp1929-action-map", "relation": "mapped-by" }, { "target": "companies", "relation": "part-of" } ], "evidence": { "humanProspective": "strong in a different cancer; none in pancreatic cancer", "humanRetrospective": "none in pancreatic cancer", "randomised": "none in pancreatic cancer", "independentReplication": "none in pancreatic cancer", "sampleSize": "Planned pancreatic n=30; current pancreatic outcome n=0", "effect": "Operational cross-cancer experience with no pancreatic safety or benefit estimate", "limits": [ "Private company data unavailable", "Japanese indication differs", "Conditional approval is not full pancreatic approval", "Regulatory state can change", "Trial has not begun", "Company source describes its own platform" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "The platform works operationally in head and neck tissue; pancreatic delivery is unproved." }, "animalEvidence": { "rating": "moderate", "reason": "Animal work supports the light mechanism but not the pancreatic business case." }, "inVitroEvidence": { "rating": "moderate", "reason": "Cell-level killing cannot price clinical pancreatic risk." }, "reproducibility": { "rating": "none", "reason": "There is no pancreatic result to reproduce." }, "commercialConflicts": { "rating": "high", "reason": "Rakuten owns the drug-device platform and supplies the first trial." }, "recency": { "rating": "current", "reason": "Company and trial states were checked in 2026." } } }, "layout": { "x": 1036, "y": 960 }, "updated": "2026-09-15" }, { "id": "company-relay-therapeutics", "title": "Relay Therapeutics — FGFR2-selective cholangiocarcinoma drug licensed before definitive utility", "shortTitle": "Relay Therapeutics", "type": "company", "status": "open", "scope": [ "cholangiocarcinoma", "pan-cancer" ], "summary": "Relay designed lirafugratinib to spare FGFR1 and FGFR4 and hit resistance mutations, then licensed global development to Elevar.", "content": "Technology: Relay combined structural modelling and experiments to design lirafugratinib, an irreversible FGFR2 inhibitor with more than 250-fold and 5,000-fold laboratory selectivity over FGFR1 and FGFR4. Early clinical cases showed tumour regression, but the 2023 paper mixed cell, animal and early human evidence and was led by company scientists. Stage: the open-label REFOCUS phase 1/2 trial enrolled 490 people with FGFR2-driven cholangiocarcinoma and other tumours and completed in September 2025; no registry results were posted by February 2026. Ownership and capital: Relay licensed worldwide rights to Elevar in December 2024 for up to $75m in upfront and regulatory milestones, up to $425m commercial milestones and royalties. Those headline totals are contingent and cannot be treated as received cash or R&D. Advantage: target selectivity directly addresses known off-target toxicity and some kinase-domain resistance. Weakness: cross-study response cannot prove superiority over approved FGFR inhibitors, later resistance still emerges, and Relay no longer controls clinical execution. Mission role: require full cohort denominators, prior-inhibitor strata, progression genotypes and a comparative trial before ranking the product above established FGFR2 drugs.", "sourceIds": [ "subbiah-2023-rly4008", "nct04526106-refocus", "relay-elevar-2024-license" ], "links": [ { "target": "companies", "relation": "part-of" }, { "target": "company-elevar-therapeutics", "relation": "licenses-to" }, { "target": "claim-fgfr2-targeting-valid-but-not-durable", "relation": "competes-with" }, { "target": "cholangiocarcinoma", "relation": "treats-subgroup" }, { "target": "investment", "relation": "contains" } ], "evidence": { "humanProspective": "weak", "humanRetrospective": "not-applicable", "randomised": "none", "independentReplication": "weak", "sampleSize": "Mechanism paper with early clinical cases; completed phase 1/2 registry n=490 without posted results", "effect": "Selective FGFR2 inhibition and early responses; comparative survival effect unknown", "limits": [ "Company-led discovery", "No randomized comparison", "Registry results absent", "Mixed tumour types", "Programme licensed out", "Most deal value contingent" ] }, "updated": "2026-09-14" }, { "id": "company-renovorx", "title": "RenovoRx", "shortTitle": "RenovoRx", "type": "company", "status": "contested", "scope": [ "pdac", "trial-delivery", "transfer-unproven" ], "summary": "RenovoRx owns the catheter and controls the enrolled phase 3 record; mission value is independent data access, not equity or another trial.", "content": "RenovoRx developed and sells RenovoCath, a two-balloon artery catheter cleared for temporary vessel blockage and drug infusion, and is testing the drug-device package with gemcitabine in locally advanced pancreatic cancer. The company is both platform claimant and evidence controller. It funded the early pooled study and 16-person blood study; company-linked authors disclosed shares, employment, a board role or consultancy. Its phase 3 plan changed after launch and its second interim estimate remains unpublished, so final access matters more than another sponsor summary. The operating picture does not support mission subsidy. At 30 June 2026, RenovoRx reported $9.479m cash and $6.576m operating cash use over six months. Simple unchanged-use coverage is $9.479m / ($6.576m / 6) = 8.65 months. This is not a forecast because sales, financing and trial payments vary. By August the company reported enrollment closed and 78 of 86 deaths, so new clinical capital cannot accelerate recruitment. It also reported $1.472m six-month revenue, showing that the device business and cancer evidence must be assessed separately. Spend £0 on equity and £0 on a new clinical trial. Offer at most £250,000 from the existing data and statistics line after database lock, only for participant-level access, original and changed plans, independent analysis and public negative-result rights. Do not value other-drug or other-cancer expansion as proved until target-tissue delivery and clinical benefit repeat.", "sourceIds": [ "renovorx-2025-10k", "renovorx-2026-q2-10q", "renovorx-2026-full-enrollment", "clinicaltrials-nct03257033-tigerpac", "novelli-2026-tigerpac-pk", "hatoum-2024-tamp-rr1-rr2" ], "links": [ { "target": "companies", "relation": "part-of" }, { "target": "trial-pdac-intraarterial-delivery-action-map", "relation": "sponsors" }, { "target": "failure-tigerpac-component-and-interim-inference", "relation": "controls" }, { "target": "hypothesis-tigerpac-existing-final-gate", "relation": "constrained-by" }, { "target": "investment", "relation": "constrains" } ], "evidence": { "humanProspective": "not-applicable", "humanRetrospective": "not-applicable", "randomised": "not-applicable", "independentReplication": "moderate", "sampleSize": "$9.479m cash; $6.576m six-month operating cash use; $1.472m six-month revenue; phase 3 at 78 of 86 deaths", "effect": "Equity and new-trial allocation £0; conditional independent analysis capped at £250,000", "limits": [ "Cash coverage is a crude unchanged-use ratio", "Later financing and sales may change the position", "Final randomized count is not in the enrollment release", "Company controls the unread result", "This is programme allocation, not personal investment advice" ], "dimensions": { "biologicalPlausibility": { "rating": "not-applicable", "reason": "Linked pages assess drug transport and patient evidence." }, "animalEvidence": { "rating": "not-applicable", "reason": "This page assesses company capacity, incentives and data access." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "This page assesses company capacity, incentives and data access." }, "reproducibility": { "rating": "moderate", "reason": "Registry, peer-reviewed papers, SEC filings and releases expose the same ownership and trial route." }, "commercialConflicts": { "rating": "high", "reason": "RenovoRx owns the product, funds studies and controls the decisive trial readout." }, "recency": { "rating": "current", "reason": "The page uses filings and trial records through August 2026." } } }, "layout": { "x": 550, "y": 474 }, "updated": "2026-09-15" }, { "id": "company-revolution-medicines", "title": "Revolution Medicines — the first common-RAS survival proof in PDAC", "shortTitle": "Revolution Medicines", "type": "company", "status": "supported", "scope": [ "pdac", "colorectal", "pan-cancer" ], "summary": "Daraxonrasib doubled median survival after prior PDAC therapy and became the first approved common-RAS drug; the company is now testing earlier disease and mutation-specific combinations.", "content": "Technology: oral inhibitors that bind active RAS, led by multi-selective daraxonrasib and mutation-selective zoldonrasib, elironrasib and RMC-5127. Clinical position: daraxonrasib produced randomized overall-survival HR 0.40 in previously treated metastatic PDAC and received FDA approval in August 2026. The first-line, post-resection and G12D-combination phase 3 programme now asks whether that gain can become longer and earlier control. Its paired phase 1/2 resistance study found new RAS-pathway changes in 26/44 progressing patients, including mutant KRAS amplification in 16/44. This is strategically valuable because the company owns both the broad and G12D-selective agents needed for one proposed deeper-blockade route. It is also a concentration risk: the main efficacy, resistance, compounds, patents and follow-on trials sit inside one sponsor. Origin and capital: Third Rock Ventures formed the company with a $45m Series A in 2015 around licensed University of Illinois chemistry; that original platform was not the later RAS programme and the money cannot be traced asset by asset. At 30 June 2026 the company reported $815.4 million cash and $3.123 billion marketable securities, $3.938 billion combined. It spent $738.9 million on R&D in the first half of 2026; $227.8 million, 30.8%, was identified as third-party daraxonrasib cost and $117.9 million as zoldonrasib cost. Likely advantage: a proven drug, variant breadth, companion molecules and enough capital to run several global trials. Main weakness: resistance is frequent, toxicity is material, the pivotal and resistance papers were company-funded, and one correlated programme cannot establish the best independent combination. The mission should buy or negotiate phase 3 data and specimens with independent analysis rights, not duplicate its late-stage trials or accept its combination choice without reproduction.", "sourceIds": [ "third-rock-2015-revolution-medicines-launch", "revolution-medicines-2026-q2-10q", "oreilly-2026-daraxonrasib", "aronchik-2026-daraxonrasib-resistance", "fda-2026-rasonque-approval" ], "links": [ { "target": "companies", "relation": "part-of" }, { "target": "claim-daraxonrasib-common-ras-survival-breakthrough", "relation": "developed" }, { "target": "trial-daraxonrasib-expansion-programme", "relation": "sponsors" }, { "target": "claim-kras-mutant-dosage-may-shape-resistance", "relation": "must-resolve" }, { "target": "claim-daraxonrasib-kras-amplification-resistance", "relation": "reported" }, { "target": "investment", "relation": "part-of" }, { "target": "investor-third-rock-revolution-medicines", "relation": "funded-by" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "moderate", "randomised": "strong", "independentReplication": "weak", "sampleSize": "Pivotal PDAC n=500; paired phase 1/2 resistance n=44; three further phase 3 programmes plan about 1,800 participants combined", "effect": "Second-line OS 13.2 vs 6.6 months in RAS G12 disease, HR 0.40", "limits": [ "Sponsor-funded pivotal and resistance studies", "Many company-employed authors and named patent inventors", "Resistance frequency lacks independent phase 3 reproduction", "Matched-combination benefit unknown", "First-line and post-resection efficacy unknown", "Financial spending is company-wide, not PDAC-only except named programme costs" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Randomized survival and paired resistance data establish human RAS dependence and recurrent pathway reactivation." }, "animalEvidence": { "rating": "strong", "reason": "A broad company and academic model programme supports the compounds and several candidate resistance combinations." }, "inVitroEvidence": { "rating": "strong", "reason": "Cell and human-derived systems reproduce multiple resistance routes and deeper-blockade concepts." }, "reproducibility": { "rating": "weak", "reason": "The pivotal and resistance results are each substantial but remain dominated by one sponsor and lack independent repeat." }, "commercialConflicts": { "rating": "high", "reason": "The company controls the drugs, patents, trials, main resistance dataset and much of the relevant capital." }, "recency": { "rating": "current", "reason": "Efficacy, approval, resistance and financial records were reported in 2026." } } }, "layout": { "x": 79, "y": 109 }, "updated": "2026-09-14" }, { "id": "company-roche-genentech", "title": "Roche–Genentech — global trial engine across PDAC vaccination and HCC treatment", "shortTitle": "Roche–Genentech", "type": "company", "status": "open", "scope": [ "pdac", "hcc", "pan-cancer" ], "summary": "Genentech sponsors the 89-site PDAC vaccine trial and has randomized HCC survival evidence for atezolizumab–bevacizumab, giving it broad execution power and control over correlated evidence.", "content": "PDAC role: Genentech sponsors IMCODE003, combines BioNTech's autogene cevumeran with atezolizumab and chemotherapy, and plans qualified access to patient-level trial data. HCC role: Roche–Genentech funded IMbrave150, in which atezolizumab plus bevacizumab improved survival against sorafenib in 501 first-line unresectable HCC patients. The HCC result does not transfer to PDAC and requires suitable liver function and bleeding-risk control. Competitive advantage: global regulatory, manufacturing and trial infrastructure can test combinations at a scale academia cannot easily reach. Main weakness: sponsor control of drugs, operations and analysis creates correlated evidence. In the PDAC vaccine trial, three interventions sit around an already toxic chemotherapy backbone, so a positive result may not isolate the active part. Mission role: do not duplicate these trials. Negotiate specimens, versioned antigen-selection data, manufacturing attrition and independent endpoint reproduction; use the HCC programme as evidence that disease context determines checkpoint value, not as a general immune claim.", "sourceIds": [ "nct05968326-imcode003", "rojass-2025-neoantigen-followup", "finn-2020-imbrave150" ], "links": [ { "target": "companies", "relation": "part-of" }, { "target": "company-biontech", "relation": "partners-with" }, { "target": "claim-neoantigen-vaccine-frontier", "relation": "sponsors" }, { "target": "programme-100m-randomized-platform", "relation": "could-inform" }, { "target": "risks", "relation": "depends-on" }, { "target": "claim-hcc-systemic-combination-improves-not-cures", "relation": "sponsors" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "not-applicable", "randomised": "strong", "independentReplication": "none", "sampleSize": "Recruiting PDAC randomized phase 2 plans n=260 at 89 sites; completed HCC phase 3 n=501", "effect": "HCC survival benefit established; PDAC vaccine randomized result absent", "limits": [ "Sponsor controls the trial", "Open-label", "Three-part experimental addition obscures active component", "Primary completion estimated 2031", "Manufacturing attrition can bias treated-only analyses" ] }, "layout": { "x": 110, "y": 92 }, "updated": "2026-09-14" }, { "id": "company-sichuan-enray-pharmaceutical", "title": "Sichuan Enray: patented iron-particle platform without a pancreatic mechanism result", "shortTitle": "Sichuan Enray Pharmaceutical", "type": "company", "status": "active", "scope": [ "pdac", "colorectal", "pan-cancer", "ferroptosis", "nanoparticle-delivery", "capital-allocation" ], "summary": "A listed-parent subsidiary owns CNSI-Fe and sponsors its trials; one pancreatic case cannot support investment or resistance-reversal claims.", "content": "Sichuan Enray Pharmaceutical Sciences grew from an innovation laboratory inside Chongqing Lummy Pharmaceutical and remains its controlled subsidiary. Enray develops CNSI-Fe, a suspension of carbon particles and ferrous iron injected directly into tumours. The US application US20250144252A1 in the PCT/CN2024/082864 family claims composition ranges, injection, MRI-guided distribution and use; its existence establishes possible access control, not validity or benefit. The company sponsored a completed 19-person dose study and a recruiting 54-person expansion across many solid tumours. Only two reported first-study participants had pancreatic cancer. The paper's one long-surviving pancreatic case did not show human ferroptosis or controlled drug resensitization. Enray's current site also promotes a 52-person phase II combination with radiation for sarcoma and head-and-neck disease, not a pancreatic efficacy programme. Chongqing Lummy's 2025 annual report confirms continuing phase II and heat-platform development but does not separate a CNSI-Fe valuation or pancreatic budget. Business decision: invest £0 in Enray or the listed parent, buy £0 of product and do not duplicate its trials. Offer at most the £400,000 independent mechanism package only for custody of existing tissue, raw spatial data, negative publication rights and component testing. If those access terms or tumour-selectivity gates fail, the mission exits. The potentially reusable asset is a validated local iron-delivery map, not the company claim that later chemotherapy sensitivity was restored.", "sourceIds": [ "enray-2026-company", "lummy-2025-annual-report-cnsi-fe", "patent-us20250144252a1-cnsi-fe", "clinicaltrials-nct06048367-cnsi-fe", "clinicaltrials-nct07433283-cnsi-fe-expansion", "xie-2026-cnsi-fe-first-human" ], "links": [ { "target": "claim-ferroptosis-pdac-selectivity-not-class-effect", "relation": "develops" }, { "target": "failure-cnsi-fe-resensitization-and-mechanism-claim", "relation": "bounded-by" }, { "target": "unknown-ferroptosis-cell-type-coverage-and-human-benefit", "relation": "constrained-by" }, { "target": "hypothesis-cnsi-fe-current-study-mechanism-gate", "relation": "data-requested-by" }, { "target": "trial-pdac-ferroptosis-action-map", "relation": "mapped-by" }, { "target": "companies", "relation": "part-of" } ], "evidence": { "humanProspective": "weak", "humanRetrospective": "weak", "randomised": "none", "independentReplication": "none", "sampleSize": "First study n=19 with n=2 pancreatic; current expansion target n=54; one pending US patent family", "effect": "Owns and advances the only identified human ferroptosis-labelled product, without a pancreatic mechanism or benefit result", "limits": [ "Subsidiary finances not separated", "Patent application is not efficacy", "Sponsor controls trials", "Two pancreatic cases", "No randomization", "No human mechanism biomarker", "English public company detail is limited" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "Local iron chemistry is plausible, while tumour selectivity and systemic value remain unmeasured." }, "animalEvidence": { "rating": "moderate", "reason": "Company models support local effects but independent PDAC context warns of opposite immune outcomes." }, "inVitroEvidence": { "rating": "moderate", "reason": "The product components can be tested independently, but a full public component result is missing." }, "reproducibility": { "rating": "none", "reason": "No independent human product study exists." }, "commercialConflicts": { "rating": "high", "reason": "Product, patent, trials and leading human paper are linked to Enray and its collaborators." }, "recency": { "rating": "current", "reason": "Parent and trial records are current through 2026." } } }, "layout": { "x": 1198, "y": 1122 }, "updated": "2026-09-15" }, { "id": "company-silexion-therapeutics", "title": "Silexion Therapeutics — local and systemic KRAS siRNA after an inconclusive implant trial", "shortTitle": "Silexion Therapeutics", "type": "company", "status": "open", "scope": [ "pdac", "colorectal", "pan-cancer" ], "summary": "The first-generation LODER implant missed overall randomized survival; SIL204 has an authorised 403-person phase 2/3 and one activated site, but no reported treated person or human target result.", "content": "Origin and ownership: the operating Israeli company, formerly Silenseed, became a wholly owned subsidiary of a Nasdaq-listed holding company through a 2024 SPAC transaction. Technology: first-generation siG12D-LODER placed extended-release RNA-silencing rods into tumour. Second-generation SIL204 changes the RNA chemistry and formulation and combines endoscopic injection at at least three tumour sites with under-skin dosing intended to reach dispersed disease. Clinical position: the 59-person LODER study randomized 37 and showed 22.7 versus 21.9 months median survival overall, not significant. Mutation status was known for 31/59; the G12D/V randomized subgroup was 11 versus five and produced HR 0.59, 95% CI 0.18-1.96, P=0.39. CTIS now authorises protocol SIL204SL-P001, an adaptive integrated phase 2/3 with 403 planned people. Segment 1 is a 15-to-21-person safety and plasma-exposure stage; later segments select dose and compare survival. One Israeli site was activated on 29 July 2026 and cleared to begin screening, but the release still anticipated first dosing. German CTIS status remains recruitment pending. The company release describes about 166 randomized people while CTIS and the later filing describe 403 total. No human SIL204 target or efficacy result exists. Access and finance: CTIS says individual participant data will not be shared. At 31 March 2026 the company reported $2.413m cash and $3.646m cash used in operations over three months, a simple unchanged-burn runway of $2.413m ÷ ($3.646m/3) = 1.99 months before new financing; the filing stated substantial doubt and later financing can change that snapshot. Likely advantage: one molecule and two routes try to cover primary and distant compartments. Main weaknesses: tumour-wide injection coverage, cell entry, endosomal escape, mutation-specific cutting, repeat EUS harm, under-skin exposure, a new product inheriting an old subgroup story, closed participant data and financing dependence. Mission decision: invest £0. Offer independent multi-region pharmacology only under a public-data contract, and stop before expansion if human knockdown, spatial agreement, systemic exposure or chemotherapy delivery fails.", "sourceIds": [ "silexion-2026-halfyear-filing", "alagesan-2026-sig12d-loder", "ctis-2026-sil204-phase23", "silexion-2026-sil204-site-initiation" ], "links": [ { "target": "companies", "relation": "part-of" }, { "target": "failure-sig12d-loder-randomized-os-not-shown", "relation": "developed" }, { "target": "company-revolution-medicines", "relation": "competes-with" }, { "target": "experiments", "relation": "requires-target-engagement-before" }, { "target": "investment", "relation": "part-of" }, { "target": "claim-sil204-trial-scales-before-human-knockdown", "relation": "develops" }, { "target": "unknown-sil204-human-target-delivery", "relation": "must-resolve" }, { "target": "trial-sil204-action-map", "relation": "runs" }, { "target": "hypothesis-multicompartment-kras-sirna-delivery", "relation": "could-test" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "not-applicable", "randomised": "moderate", "independentReplication": "none", "sampleSize": "First-generation n=59 with randomized n=37; second-generation trial plans 15-21 in Segment 1 and about 403 total", "effect": "No significant first-generation overall survival gain; no human SIL204 target or efficacy result", "limits": [ "Company filings and sponsor registry", "Second-generation product has no reported human dose", "First-generation subgroup was exploratory and very small", "Public sample descriptions conflict", "Participant-level data sharing set to no", "March cash snapshot may be stale after financing" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "The target is validated and RNA cutting can work if tumour-cell delivery succeeds." }, "animalEvidence": { "rating": "moderate", "reason": "Sponsor models support both routes but do not settle human distribution." }, "inVitroEvidence": { "rating": "moderate", "reason": "SIL204 reduces target and growth in cell systems." }, "reproducibility": { "rating": "weak", "reason": "No independent product-specific or human SIL204 result exists." }, "commercialConflicts": { "rating": "high", "reason": "Silexion owns the product, sponsors the trial and frames the old non-significant subgroup as validation." }, "recency": { "rating": "current", "reason": "Status checked through the August 2026 filing." } } }, "layout": { "x": 88, "y": 112 }, "updated": "2026-09-15" }, { "id": "company-sonire-therapeutics", "title": "Sonire Therapeutics: well-funded HIFU development awaiting its randomized result", "shortTitle": "Sonire Therapeutics", "type": "company", "status": "active", "scope": [ "pdac", "focused-ultrasound", "company", "capital-allocation" ], "summary": "Sonire reports $31 million of new private and public funding in 2026, completed enrollment in a 90-person Japanese randomization and began a ten-person U.S. feasibility study; no result yet validates patient benefit.", "content": "Sonire Therapeutics develops Suizenji, an ultrasound-guided high-intensity focused-ultrasound system intended to heat pancreatic tumours without an incision. It grew from the academic device work behind the MS-2 used in the K-912 study, but its current trials test thermal HIFU without K-912 and must remain a separate evidence branch. ClinicalTrials.gov lists SUNRISE-I as an open-label 90-person randomization of chemotherapy with or without Suizenji, with overall survival primary and no posted result. Sonire says enrollment is complete. SUNRISE-II is a recruiting ten-person Stanford feasibility study; its primary endpoint is more than 50% ablation of the treated volume at three months, not survival. In 2026 the company announced $18 million private financing and about $13 million of Japanese government support, a combined $31 million before currency and timing adjustments. That capital can fund the device programme without mission equity. The mission's leverage is data access: invest £0 in Sonire and release at most the existing £250,000 HIFU analysis cap only after the 90-person database locks, conditional on all assignments, device logs, chemotherapy delivery, pain, quality of life, harm, survival and independent publication. Breakthrough or grant status speeds development; it does not establish benefit.", "sourceIds": [ "clinicaltrials-nct05601323-suizenji", "clinicaltrials-nct07033689-sunrise2", "sonire-2026-series-a", "sonire-2026-nedo-award", "sonire-2026-sunrise2-first-patient", "muragaki-2026-k912-sonodynamic" ], "links": [ { "target": "claim-hifu-pain-signal-without-survival-proof", "relation": "develops" }, { "target": "unknown-hifu-pain-survival-and-system-transfer", "relation": "constrained-by" }, { "target": "hypothesis-hifu-existing-randomized-gate", "relation": "data-requested-by" }, { "target": "trial-pdac-hifu-action-map", "relation": "mapped-by" }, { "target": "claim-k912-first-human-local-signal-no-component-proof", "relation": "descends-from" }, { "target": "failure-k912-component-and-local-endpoint", "relation": "bounded-by" }, { "target": "trial-pdac-k912-sonodynamic-action-map", "relation": "mapped-by" }, { "target": "companies", "relation": "part-of" } ], "evidence": { "humanProspective": "moderate — randomized enrollment complete but result unavailable", "humanRetrospective": "none", "randomised": "moderate — n=90 dataset exists without public result", "independentReplication": "weak", "sampleSize": "SUNRISE-I target n=90; SUNRISE-II target n=10; reported new funding $31 million", "effect": "No public controlled patient-benefit estimate; company has enough announced capital to reach the next result", "limits": [ "Funding figures are company announcements", "Randomized data are unreported", "Chemotherapy choice varies", "U.S. study is single group", "Ablated volume is not patient benefit", "K-912 is a separate package", "Regulatory status can change" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Focused ultrasound can heat tissue; selective clinical value remains unresolved." }, "animalEvidence": { "rating": "moderate", "reason": "Device development has animal and early human roots, while randomization now matters most." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "The company decision turns on the completed randomized human dataset." }, "reproducibility": { "rating": "weak", "reason": "A new U.S. study has begun, but it is small and uncontrolled." }, "commercialConflicts": { "rating": "high", "reason": "Sonire sponsors both current device studies and reports its own funding and operating claims." }, "recency": { "rating": "current", "reason": "Financing, grant and U.S. trial start were checked in 2026." } } }, "layout": { "x": 1072, "y": 996 }, "updated": "2026-09-15" }, { "id": "company-tango-therapeutics", "title": "Tango Therapeutics — strong funding, selected biology and an unresolved component claim", "shortTitle": "Tango Therapeutics", "type": "company", "status": "mixed", "scope": [ "pdac", "pan-cancer" ], "summary": "Tango has a large MTAP-selected PDAC programme and enough capital to run it; the mission should demand the right comparator rather than subsidise the company.", "content": "Product. Tango develops vopimetostat, an oral MTA-cooperative PRMT5 inhibitor for cancers with MTAP deletion. It runs NCT05732831 as monotherapy and with pembrolizumab, and NCT06922591 with Revolution Medicines RAS inhibitors and chemotherapy. Evidence. Monotherapy produced 15% ORR and 71% disease control in a mature 39-person pancreatic set. The two early RAS combinations produced 25 responses among 39 evaluable PDAC patients, 19 confirmed at cutoff. This is an unusually large early signal in a biomarker that covers about 28% of PDAC, but no concurrent arm measures vopimetostat's addition to RAS inhibition. Finance. At 30 June 2026 Tango reported $1.0bn in cash, equivalents and marketable securities. Six-month operating cash use was $83.226m and direct vopimetostat programme expense was $21.677m. The simple constant-burn ratio is 1000/(83.226/6)=72.1 months. That is not company guidance, but it shows the next proof step is not capital constrained. June financing supplied $651.4m net. Business position. Tango owns the biomarker-selected layer; Revolution Medicines supplies the RAS layer. If the package works, value depends on exact MTAP testing, reliable drug supply, a clean contribution claim and whether two branded drugs earn their combined harm and price. Main weaknesses. Interim sponsor-only data, small response sets, no controlled component estimate, multiple doses, short durability and dependence on a partner's RAS agents. Mission decision. Invest £0. Require all-treated data, exact deletion reporting, independent analysis and a RAS-alone contribution arm. The mission can add more value by enforcing evidence quality than by adding money to a company with roughly twelve times its current annualized operating cash use on hand.", "sourceIds": [ "matsumoto-2026-mtap-pancreatic", "pavlick-2026-mtap-genomics", "clinicaltrials-nct05732831-vopimetostat", "tango-2025-vopimetostat-monotherapy", "clinicaltrials-nct06922591-vopimetostat-ras", "tango-2026-vopimetostat-ras-data", "tango-2026-q2-10q" ], "links": [ { "target": "companies", "relation": "part-of" }, { "target": "investment", "relation": "part-of" }, { "target": "claim-mtap-prmt5-selected-human-signal", "relation": "develops" }, { "target": "failure-vopimetostat-ras-contribution-unknown", "relation": "must-repair" }, { "target": "unknown-mtap-prmt5-ras-contribution", "relation": "must-resolve" }, { "target": "trial-mtap-prmt5-action-map", "relation": "runs" }, { "target": "hypothesis-mtap-prmt5-ras-deeper-control", "relation": "could-test" }, { "target": "company-revolution-medicines", "relation": "collaborates-with" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "strong", "randomised": "none", "independentReplication": "weak", "sampleSize": "Monotherapy PDAC mature set n=39; combination PDAC n=54 treated and n=39 response evaluable", "effect": "Early selected activity without controlled added effect", "limits": [ "Finance snapshot changes", "Constant-burn ratio is not company guidance", "No peer-reviewed full combination report", "No participant-level data", "Partner dependence" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The product targets a defined deletion-created dependency." }, "animalEvidence": { "rating": "moderate", "reason": "Models support the programme but human evidence sets valuation." }, "inVitroEvidence": { "rating": "strong", "reason": "Drug selectivity and target inhibition are measurable." }, "reproducibility": { "rating": "weak", "reason": "The exact pancreatic combination signal has no independent confirmation." }, "commercialConflicts": { "rating": "high", "reason": "Tango controls the agent, trial and interim disclosures." }, "recency": { "rating": "current", "reason": "Finance and trial state are checked through August 2026." } } }, "layout": { "x": 194, "y": 115 }, "updated": "2026-09-15" }, { "id": "company-theriva-biologics", "title": "Theriva Biologics — a real randomized VCN-01 signal with an unfunded proof step", "shortTitle": "Theriva Biologics", "type": "company", "status": "mixed", "scope": [ "pdac", "pan-cancer" ], "summary": "Theriva owns the strongest current oncolytic-virus signal in PDAC, but its 96-person analysis is incomplete against 112 enrolled and its cash does not fund phase 3.", "content": "Origin and ownership. Theriva Biologics, Inc. acquired VCN Biosciences in 2022 and develops the VCN oncolytic-adenovirus platform through its Spanish subsidiary. Technology. VCN-01 is an Ad5 virus designed to copy in tumour cells with broken RB1 control, rupture them and express PH20 to cut the hyaluronan barrier. Stage. Two phase 1 studies show tumour delivery and replication. VIRAGE is a completed randomized open-label phase 2b in first-line metastatic PDAC. VIRAGE2 began recruiting six people in Spain in July 2026 to test three doses 56 days apart. FDA and EMA discussions support a future double-blind placebo-controlled adaptive phase 3 design, but no funded trial has begun. Evidence. The reported 96-person FAS produced OS 10.8 versus 8.6 months, HR 0.57, log-rank P=0.0546, and PFS 7.0 versus 4.6 months, HR 0.55, P=0.0105. The registry records 112 actual participants and no results. The public protocol contains conflicting alpha and analysis-set statements, and the favourable repeat-dose subgroup selects survivors through four chemotherapy cycles. Finance. Cash was $11.337m on 30 June 2026 and $9.6m in early August. Six-month operating cash use was $5.541m, an average $0.924m per month; $9.6m divided by $0.924m is 10.4 months under a deliberately simple unchanged-burn model. The company says runway reaches Q1 2027, because planned spending and obligations differ from that average, and says current cash does not fund VCN-01 trials beyond VIRAGE2. A deferred $5m Grifols milestone could materially reduce cash. Likely advantage. This is one of the few PDAC platforms with direct tumour-replication evidence plus a randomized directionally favourable result. Main weaknesses. Sponsor-controlled incomplete data, no independent confirmation, uncertain reach in people with Ad5 antibodies, severe acute toxicity, manufacturing scale, and a phase 3 financing gap. Mission decision. Invest £0 now. Offer independent data recovery and phase 3 design review only under complete-public-data terms. Do not buy equity, licence rights or the claim that repeat dosing works. Reconsider cofunding only after the all-randomized analysis and six-person dosing gate pass, a partner funds most phase 3 cost, and equal capital leaves a lower-ranked programme.", "sourceIds": [ "bazan-peregrino-2021-vcn01-intratumour", "garcia-carbonero-2022-vcn01-intravenous", "clinicaltrials-nct05673811-virage", "theriva-2023-virage-protocol", "theriva-2025-virage-topline", "esmo-2025-virage-2216mo", "ctis-2026-virage2", "theriva-2026-q2-10q" ], "links": [ { "target": "companies", "relation": "part-of" }, { "target": "investment", "relation": "part-of" }, { "target": "claim-vcn01-randomized-signal-not-survival-proof", "relation": "develops" }, { "target": "failure-virage-analysis-and-survivor-selection", "relation": "must-repair" }, { "target": "unknown-vcn01-all-enrolled-survival", "relation": "must-resolve" }, { "target": "trial-vcn01-action-map", "relation": "runs" }, { "target": "hypothesis-vcn01-replication-to-survival-gate", "relation": "could-test" }, { "target": "company-fibrogen", "relation": "shares-stroma-risk-with" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "not-applicable", "randomised": "moderate", "independentReplication": "weak", "sampleSize": "VCN-01 has been given to 144 people across cancers; VIRAGE registry n=112 and reported FAS n=96", "effect": "One randomized PDAC signal; no confirmed phase 3 benefit", "limits": [ "Cash snapshot changes with financing", "Simple burn model is not company guidance", "No full VIRAGE report", "No public participant-level data", "Future phase 3 unfinanced" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "The product reaches and replicates in sampled human tumours." }, "animalEvidence": { "rating": "moderate", "reason": "Model evidence supports the platform but does not determine company value." }, "inVitroEvidence": { "rating": "moderate", "reason": "The engineered functions are measurable in laboratory systems." }, "reproducibility": { "rating": "weak", "reason": "Mechanism repeats across early studies, while clinical benefit does not." }, "commercialConflicts": { "rating": "high", "reason": "Company filings are primary for ownership, finance and plans but not independent efficacy proof." }, "recency": { "rating": "current", "reason": "Finance and trial status are checked through August 2026." } } }, "layout": { "x": 180, "y": 100 }, "updated": "2026-09-15" }, { "id": "company-valar-labs", "title": "Valar Labs — computational pathology with a causal-evidence gap", "shortTitle": "Valar Labs", "type": "company", "status": "open", "scope": [ "pdac", "pan-cancer" ], "summary": "A $22m-backed pathology-AI company has produced a locked external PDAC chemotherapy biomarker; prospective randomized use remains the key test.", "content": "Founded by Anirudh Joshi, Viswesh Krishna and Damir Vrabac after work linked to Andrew Ng's Stanford AI laboratory. Capital: a $22m Series A in 2024 co-led by DCVC and Andreessen Horowitz, with Pear VC participation; total capital is not reconstructed here. Technology: proprietary computational histology from routine H&E slides. PDAC stage: a 2026 Journal of Clinical Oncology paper used 178 development and 299 external-validation patients to predict differential benefit between two chemotherapy families. Competitive advantage: an existing slide, a near-universal first-line decision and a locked external cohort make the product cheaper and more actionable than many multi-omics models. Weakness: observational treatment choice cannot prove that model-directed switching improves survival; the model is proprietary, company scientists hold patents and equity, and the G-pref group did not show an OS difference. Best mission role: fund an independent biomarker-stratified randomized trial with escrowed model version and accessible audit outputs, not another company-controlled retrospective cohort.", "sourceIds": [ "valar-2024-series-a", "hendifar-2026-ai-chemo-pathology" ], "links": [ { "target": "companies", "relation": "part-of" }, { "target": "technology-ai-pathology-chemo-selection", "relation": "developed" }, { "target": "ai-programme", "relation": "part-of" }, { "target": "investment", "relation": "part-of" }, { "target": "programme-25m-prospective-observatory", "relation": "may-partner" }, { "target": "investor-valar-series-a-syndicate", "relation": "funded-by" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "moderate", "randomised": "none", "independentReplication": "weak", "sampleSize": "PDAC development n=178; external validation n=299", "effect": "$22m Series A; significant biomarker-treatment interaction but no model-directed trial", "limits": [ "Company announcement used for funding and founders", "No audited total funding disclosed", "Clinical paper includes company employees and patent holders", "Proprietary model limits reproduction", "No randomized utility evidence" ] }, "layout": { "x": 86, "y": 111 }, "updated": "2026-09-14" }, { "id": "company-verastem", "title": "Verastem — RAF/MEK plus FAK blockade with an uncontrolled first-line signal", "shortTitle": "Verastem", "type": "company", "status": "open", "scope": [ "pdac", "colorectal", "pan-cancer" ], "summary": "RAMP 205 reports 52% response and 86% six-month survival in 29 patients, but four-drug treatment and no control make the extra benefit unknown.", "content": "Technology: avutometinib blocks RAF/MEK signalling while defactinib blocks FAK, added to gemcitabine and nab-paclitaxel to suppress both KRAS output and a proposed resistance route. The same two-drug product is FDA-approved in KRAS-mutant recurrent low-grade serous ovarian cancer, proving manufacturing and regulatory capacity outside PDAC. PDAC position: sponsor-reported RAMP 205 results in 29 patients at the selected dose showed 52% confirmed response, 68% six-month progression-free survival and 86% six-month overall survival after 9.8 months median follow-up. Competitive advantage: pathway combination already has human dosing and an approved component pair. Main weakness: no randomized control, a small selected cohort, immature survival and a four-drug regimen whose active contribution and cumulative toxicity are unclear. Mission role: require a randomized chemotherapy control with dose intensity, patient-reported burden, duration and molecular escape measured before treating the response rate as a breakthrough.", "sourceIds": [ "nct05669482-ramp205", "verastem-2026-ramp205-update", "wainberg-2023-napoli3" ], "links": [ { "target": "companies", "relation": "part-of" }, { "target": "claim-modern-metastatic-chemotherapy-remains-short-control", "relation": "must-improve" }, { "target": "company-revolution-medicines", "relation": "competes-with" }, { "target": "hypothesis-prospective-ras-resistance-routing", "relation": "could-test" }, { "target": "company-fibrogen", "relation": "warned-by" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "not-applicable", "randomised": "none", "independentReplication": "none", "sampleSize": "Recommended-dose cohort n=29; registry plans n=40", "effect": "ORR 52%, six-month PFS 68%, six-month OS 86%; uncontrolled and immature", "limits": [ "Sponsor announcement", "No control", "Small cohort", "Four drugs obscure contribution", "Survival immature", "Dose intensity and quality-of-life comparison unavailable" ] }, "layout": { "x": 122, "y": 90 }, "updated": "2026-09-14" }, { "id": "company-viewray", "title": "ViewRay", "shortTitle": "ViewRay", "type": "company", "status": "failed", "scope": [ "radiation", "medical-device", "business-continuity" ], "summary": "The MRIdian machine maker sponsored the main pancreatic SMART studies, then entered Chapter 11, leaving the decisive survival trial dormant.", "content": "Role. ViewRay built MRIdian, the named MRI-guided radiation system used in the 136-person pancreatic SMART phase 2 study, and was the registered sponsor of both that study and the planned 267-person LAP-ABLATE phase 3 trial. Evidence position. The phase 2 established that selected centres could deliver five high-dose adaptive fractions with low definitely attributed acute severe gastrointestinal toxicity. It did not establish added survival. Financial position before failure. At 31 March 2023 ViewRay reported $81.3m unrestricted cash, $4.2m restricted cash, $77.8m debt outstanding, a $29.6m quarterly net loss and $53.6m operating cash use over three months. A simple constant-burn calculation gives $81.3m / $53.6m x 3 = 4.55 months of unrestricted runway. The company warned that it did not expect enough liquidity to fund operations or repay accelerated debt. On 16 July 2023 ViewRay and its operating subsidiary filed voluntary Chapter 11 petitions. Trial consequence. LAP-ABLATE still names ViewRay as sponsor and remains not yet recruiting on a record last updated in October 2022. This is evidence of dependency, not proof that bankruptcy alone caused the non-start. Mission decision. Invest £0 in equity, £0 in a machine and £0 in a vendor-owned trial. Pay at most £500,000 only for independent data recovery and a brand-independent performance and data-export specification. Any future equipment purchase must include service continuity, escrowed software and calibration material, raw-data export, trial transfer rights and a funded exit plan. The business value is a warning: a technically useful device can destroy clinical information if its company is the only keeper of the machine, protocol and data route.", "sourceIds": [ "clinicaltrials-nct03621644-smart", "clinicaltrials-nct05585554-lap-ablate", "viewray-2023-q1-10q", "viewray-2023-chapter11-8k", "parikh-2023-smart-phase2-safety", "chuong-2024-smart-phase2-survival" ], "links": [ { "target": "companies", "relation": "part-of" }, { "target": "claim-smart-delivery-and-safety-without-added-survival-proof", "relation": "sponsored" }, { "target": "failure-smart-trial-and-vendor-dependency", "relation": "caused" }, { "target": "unknown-smart-added-survival-and-surgery-harm", "relation": "constrains" }, { "target": "hypothesis-smart-vendor-independent-randomized-gate", "relation": "constrained-by" }, { "target": "trial-pdac-smart-action-map", "relation": "mapped-by" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "none", "independentReplication": "none", "sampleSize": "Phase 2 n=136; planned phase 3 n=267 not started; finance from Q1 2023 filing", "effect": "Vendor sponsored delivery evidence but failed before the planned survival test", "limits": [ "Bankruptcy does not disprove the technology", "Registry does not state why LAP-ABLATE did not start", "Successor ownership and service arrangements are not mapped here", "Simple runway assumes constant burn", "Restricted cash is excluded from runway", "No current investable equity remains in scope" ], "dimensions": { "biologicalPlausibility": { "rating": "not-applicable", "reason": "This record concerns ownership, solvency and trial continuity." }, "animalEvidence": { "rating": "not-applicable", "reason": "Animal evidence cannot establish business continuity." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Laboratory evidence cannot repair sponsor failure." }, "reproducibility": { "rating": "none", "reason": "The company did not deliver the randomized survival test." }, "commercialConflicts": { "rating": "high", "reason": "ViewRay supplied the system and sponsored the central studies." }, "recency": { "rating": "current", "reason": "SEC failure records are final for the 2023 event and trial status was checked in 2026." } } }, "layout": { "x": 706, "y": 630 }, "updated": "2026-09-15" }, { "id": "cross-disciplinary-transfer-map", "title": "Outside oncology: five engineering ideas worth transferring and the conditions they need", "shortTitle": "Outside-field transfers", "type": "system", "status": "supported", "scope": [ "pdac", "pan-cancer", "transfer-unproven" ], "summary": "Control, circuit design, microfluidics, materials physics and nuclear medicine contribute testable tools; none bypasses missing human measurement or an effective treatment.", "content": "1. Control theory and tumour ecology. Adaptive prostate treatment uses repeated PSA measurements to turn a drug on and off, trying to retain drug-sensitive cells that suppress resistant cells. The 11-person pilot used less than half the drug and appeared to delay progression against 16 nonrandomized controls. Transfer to PDAC requires a fast quantitative state sensor, at least one active drug, reversible response and enough mixing for sensitive and resistant cells to compete. PDAC currently lacks that complete sensor-actuator loop, so adaptive dosing is a hypothesis rather than a treatment recommendation. 2. Programmable cell circuits. Synthetic-receptor T cells can implement an AND gate: antigen A arms the cell and antigen B triggers killing. In models this reduced attack on tissue carrying only one antigen. PDAC transfer requires an antigen pair present across lethal tumour cells but absent together from essential normal tissue, plus cell entry and persistence. The circuit solves decision logic, not delivery. 3. Microfluidics and test automation. Small channels can standardize organoid size, reduce tissue demand and expose a biopsy-derived model to several drugs and immune cells in parallel. That can shorten the laboratory step. It does not repair failed biopsy acquisition, culture failure, a result arriving after treatment starts, missing in-body stroma or a recommended drug that cannot be obtained. The programme must count every patient from enrollment through treatment delivered. 4. Materials physics. Engineered matrices show that stiffness can switch drug-efflux behaviour in PDAC organoids and that the change can reverse when stiffness falls. This makes force and matrix composition measurable state variables. It does not rescue broad stroma removal, which has failed clinically, and three patient-derived lines cannot establish a human target. The next useful test measures matrix mechanics, drug concentration and cell state in the same human samples before choosing an intervention. 5. Nuclear-medicine theranostics. A matched imaging and radioactive-treatment pair can first show where a target is, calculate absorbed dose and then deliver treatment to the same target. This works in selected neuroendocrine tumours. FAP imaging and lutetium treatment make the route technically plausible in PDAC, but first-in-human FAP treatment included five pancreatic cases without a pancreatic stable-disease response and did not measure dose in pancreatic tumours. The transferable object is the closed measurement-and-delivery loop, not the efficacy result. Common rejection rule: do not fund an analogy because its language sounds useful. Measure the property it depends on in PDAC, state the smallest falsifying experiment and require a human outcome before promotion.", "sourceIds": [ "zhang-2017-evolutionary-adaptive-therapy", "roybal-2016-combinatorial-t-cell-circuit", "choi-2024-microfluidic-pdac-organoids", "lesavage-2024-engineered-matrix-chemoresistance", "baum-2022-fap2286-firsthuman", "baudin-2026-oclurandom" ], "links": [ { "target": "transferability-matrix", "relation": "extends" }, { "target": "treatment-long-tail-audit", "relation": "constrains" }, { "target": "claim-pdac-organoid-correlation-not-routing-utility", "relation": "depends-on" }, { "target": "failure-bulk-stroma-removal", "relation": "warned-by" }, { "target": "transfer-radioligand-selection", "relation": "extends" }, { "target": "research-frontier-ranking", "relation": "constrains" } ], "evidence": { "humanProspective": "weak — one 11-person nonrandomized adaptive-treatment pilot and early theranostic use", "humanRetrospective": "weak — limited pancreatic theranostic and organoid correlations", "randomised": "strong outside PDAC for neuroendocrine radioligand treatment; none for the proposed PDAC transfers", "independentReplication": "weak — each transfer depends on a different unproved PDAC property", "sampleSize": "Five transfer classes anchored by primary control, cell-circuit, microfluidic, materials and theranostic studies", "effect": "Property map; no claimed PDAC survival effect", "limits": [ "Most evidence is model or platform evidence", "The adaptive-treatment result is small and nonrandomized", "Synthetic circuits need tumour-wide safe antigen pairs", "Organoids omit whole-body delivery and immunity", "Matrix softening can have opposing biological effects", "FAP treatment lacks a pancreatic efficacy result" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "Each idea targets a measured problem, but the needed property is incompletely established in human PDAC." }, "animalEvidence": { "rating": "moderate", "reason": "Cell circuits and matrix mechanisms work in models; adaptive control and theranostics have limited human evidence outside or near PDAC." }, "inVitroEvidence": { "rating": "strong", "reason": "Microfluidic, organoid, matrix and circuit studies establish engineering function under controlled conditions." }, "reproducibility": { "rating": "weak", "reason": "No complete transfer has independent randomized PDAC outcome evidence." }, "commercialConflicts": { "rating": "moderate", "reason": "The cited papers are mainly academic, while eventual devices, cell products and radioligands create product interests." }, "recency": { "rating": "current", "reason": "The map combines established engineering papers with human evidence checked through September 2026." } } }, "layout": { "x": 62, "y": 154 }, "updated": "2026-09-15" }, { "id": "datasets", "title": "Dataset map: public atlases are rich in tumour description and poor in treatment decision histories", "shortTitle": "Datasets", "type": "system", "status": "supported", "scope": [ "pdac", "cholangiocarcinoma", "colorectal", "pan-cancer" ], "summary": "CPTAC, IDC, ICGC ARGO and HTAN provide multi-omic, imaging and precursor data; none supplies the complete serial exposure-to-resistance chain the mission needs.", "content": "CPTAC-PDAC connects genomic, transcriptomic, proteomic, glycoproteomic, clinical and imaging data. The PDC000272 glycoproteome study lists 166 cases and the wider programme reports 140 cancers, 67 adjacent tissues and nine normal duct tissues. IDC lists 195 CPTAC-PDA imaging subjects and 103 tumour-annotation subjects. ICGC ARGO Release 14 adds international programmes including Australia, Canada and SCRUM-Japan with ctDNA and microbiome work. HTAN phase 2 includes an OHSU pancreatic pre-cancer atlas, HTA201, under grant 1U01CA294548-01. Regional delivery data are much thinner but now have measured bounds. LACOG GASPAR reports 79 advanced pancreatic cases across Brazil, Argentina and Colombia, with private-care concentration and a registry-count conflict. LATAM-CCA harmonized 309 eligible cases from five expert centres and recorded treatment for 276, but excluded mandatory-data failures and lacks a population denominator. An African pancreatic review contains 26,850 cases yet nearly 40% lacked stage and fewer than half of the 20 studies reported complete treatment paths. The Africa HPB Cancer Consortium has a roadmap rather than a common dataset. These sources can find missing care, but none supplies the full route from population catchment to outcome. Surgical and referral cohorts overrepresent people who reach specialist care; public archives often lack serial treatment exposure, toxicity, repeated blood, progression biopsy and complete recurrence outcomes. AI cannot recover a missing counterfactual, a biopsy never taken or a patient never counted. The £25m observatory should collect the missing biological chain. The global data line should separately collect the dated care route: first contact → imaging → pathology → stage and organ reserve → recommendation → delivered treatment or reason for loss → survival. Patient-level data should remain governed; dictionaries, missingness, derived models and negative benchmark results should be public.", "sourceIds": [ "cptac-pdac-pdc000272", "idc-cptac-pda", "icgc-argo-2026-release14", "htan2-pancreas-hta201", "asombang-2026-africa-hpb-registry", "clinicaltrials-nct05924789-gaspar", "peixoto-2025-gaspar-results", "gismalla-2026-africa-pancreatic-review", "sobnach-2024-ssa-hcc-systematic-review", "da-fonseca-2024-latam-cca" ], "links": [ { "target": "ai-programme", "relation": "supplies-training-data-to" }, { "target": "programme-25m-prospective-observatory", "relation": "informs" }, { "target": "unknown-durable-ras-control", "relation": "insufficient-for" }, { "target": "unknown-dangerous-precursor-selection", "relation": "partly-addresses" }, { "target": "people", "relation": "used-by" }, { "target": "lab-africa-hpb-cancer-consortium", "relation": "could-be-built-by" }, { "target": "trial-gaspar-lacog-0222", "relation": "includes" }, { "target": "lab-latam-cca-registry", "relation": "includes" }, { "target": "unknown-regional-route-loss", "relation": "insufficient-for" } ], "evidence": { "humanProspective": "not-applicable", "humanRetrospective": "strong", "randomised": "none", "independentReplication": "strong", "sampleSize": "CPTAC/IDC 195 imaging subjects; PDC glycoproteome 166 cases; GASPAR 79 pancreatic cases; LATAM-CCA 309 eligible cases; African pancreatic review 26,850 incompletely described cases; African consortium roadmap without a common cohort", "effect": "Data-coverage assessment; no clinical effect", "limits": [ "Counts refer to different linked collections", "Controlled access", "Surgical selection", "Incomplete longitudinal exposure and progression tissue", "Demographic coverage not fully audited", "GASPAR is selected and its reported enrolment totals conflict", "The African consortium source is a roadmap rather than a complete dataset", "LATAM-CCA starts at selected expert centres and excludes incomplete records", "Large African review counts lack route fields and population linkage" ] }, "layout": { "x": 82, "y": 98 }, "updated": "2026-09-14" }, { "id": "detection-dna-cell-methods", "title": "Detection audit: liquid biopsy, DNA fragments and circulating cells", "shortTitle": "DNA and cell detection", "type": "system", "status": "supported", "scope": [ "pdac", "pancreatic-all", "hepatobiliary", "colorectal", "pan-cancer", "transfer-unproven" ], "summary": "DNA and cell assays can detect or follow some pancreatic cancers, but stage-I shedding, lead time and intended-population utility remain the limiting evidence.", "content": "Liquid biopsy is an umbrella, not one test. The useful output must specify whether it detects tumour mutations, methylation, fragmentomics, whole cells, RNA, proteins or vesicles, because each has a different burden threshold and error route. No cited liquid-biopsy product has shown that screening an average-risk population reduces pancreatic-cancer death. Regulatory clearance for a cancer-monitoring or tissue-profiling use would not establish a pancreatic screening indication. Cost and operating burden range from a routine blood draw plus central sequencing to rapid local assays; published accuracy rarely includes the downstream MRI, EUS, biopsy and surgery created by false positives.\n\nctDNA mutation detection is specific when a tumour alteration is found, but early PDAC releases little DNA and common mutations can come from age-related blood-cell clones. Detectable tumour burden and stage-I sensitivity depend on plasma volume, panel, error correction and whether the tumour is known. Tumour-informed ctDNA is strongly prognostic after PDAC treatment, but this is residual-disease monitoring rather than presymptomatic detection; its useful lead time and whether acting early helps survival are unresolved. Current evidence is mainly retrospective or observational, repeatability varies by assay, and no regulated PDAC screening use is established.\n\nMethylation raises the number of detectable signals beyond mutations. CCGA's locked multi-cancer classifier reported 99.3% specificity, but sensitivity across cancers was only 18% at stage I; pancreatic results came from diagnosed cases, not a low-prevalence screening population. The assay can suggest tissue of origin, but detectable burden, pancreatic lead time, interval cancers and false-positive work-up were not established. NHS-Galleri subsequently provided randomized workflow evidence but missed its combined late-stage primary endpoint, so multi-cancer early detection remains unproved as a route to fewer pancreatic deaths. Central sequencing, calibration and repeat sampling make cost and operations substantial.\n\nFragmentomics measures genome-wide cell-free-DNA fragment patterns. The 2019 case-control study reported 57% to more than 99% sensitivity across seven cancers at 98% specificity and 91% sensitivity when fragment and mutation signals were combined. It did not establish pancreatic stage-I sensitivity, lead time or utility. At annual PDAC incidence of 13.9 per 100,000, 98% specificity would generate about 2,000 false positives per 100,000 tests before repeat testing, so that operating point cannot support average-risk screening. Independent intended-population repeatability, locked processing, cost and regulatory status remain open.\n\nCirculating tumour cells require intact rare-cell capture and expert processing. One 100-person presentation-stage study reported 75.0% sensitivity and 96.4% specificity, while a 32-person resectable cohort showed prognostic association around surgery. Neither measured presymptomatic burden, stage-I screening sensitivity, lead time or clinical utility. At 96.4% specificity, the base-rate burden would be thousands of false positives per 100,000 average-risk tests. Repeatability across capture platforms is mixed, cost and laboratory complexity are high, and there is no established pancreatic screening indication. CTCs remain more credible for biology, staging and monitoring than population screening.", "sourceIds": [ "botta-2024-tumor-informed-ctdna", "liu-2020-ccga-methylation", "sasieni-2026-nhs-galleri", "cristiano-2019-fragmentomics", "poruk-2016-ctc-pdac-diagnosis", "song-2021-ctc-resectable-pdac", "seer-pancreas-statfacts" ], "links": [ { "target": "detection", "relation": "part-of" }, { "target": "detection-modality-ranking", "relation": "extends" }, { "target": "claim-case-control-blood-tests-overstate-screening-readiness", "relation": "depends-on" }, { "target": "unknown-ctdna-actionability", "relation": "constrained-by" }, { "target": "claim-screening-base-rate-maths", "relation": "constrained-by" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "strong", "randomised": "moderate", "independentReplication": "weak — mixed evidence", "sampleSize": "Evidence ranges from 32-person PDAC cell studies to 6,689-person CCGA validation and randomized NHS-Galleri screening", "effect": "No method has pancreatic mortality proof; low stage-I shedding and specificity at low prevalence remain decisive", "limits": [ "Assay families contain many non-equivalent products", "Cancer-type pooling obscures pancreatic performance", "Residual-disease use is not screening", "Costs are platform and health-system dependent", "Regulatory position can change and must be checked product by product" ] }, "layout": { "x": 109, "y": 48 }, "updated": "2026-09-14" }, { "id": "detection-imaging-methods", "title": "Detection audit: ultrasound, CT, MRI, PET and AI radiology", "shortTitle": "Imaging detection", "type": "system", "status": "supported", "scope": [ "pdac", "pnet", "cholangiocarcinoma", "gallbladder", "hcc", "colorectal", "pan-cancer" ], "summary": "Imaging can find actionable disease in selected patients, but access, radiation, anatomy, incidental findings and absence of a mortality trial prevent population pancreatic screening.", "content": "Ultrasound is inexpensive and widely available, but ordinary transabdominal ultrasound does not reproducibly see the full pancreas in average-risk people. Bowel gas, body habitus, operator dependence and small iso-echoic lesions limit detectable burden and repeatability. It has no established PDAC stage-I sensitivity, lead time or mortality utility. Ultrasound screening is more credible where liver or biliary anatomy and a very high-risk population change the test: a nonrandomized Thai cholangiocarcinoma programme found far more stage 0-II disease than walk-in diagnosis, but selection prevents a mortality claim.\n\nEndoscopic ultrasound reaches the pancreas at high resolution and permits sampling, but sedation, operator dependence, pancreatitis, bleeding, cost and specialist capacity prevent low-risk population use. MRI/MRCP avoids ionising radiation and maps cysts and ducts. In inherited or familial high-risk cohorts, MRI/MRCP and EUS surveillance are associated with stage shift and longer observed survival, but the studies are nonrandomized and do not give one transportable modality-specific stage-I sensitivity or specificity. Their current clinical use is selected high-risk surveillance and lesion work-up, not average-risk screening.\n\nCT is fast and available, and pancreas-protocol contrast CT is central after symptoms or a risk trigger. Repeated ionising radiation makes new CT scans a poor average-risk screen. Opportunistic analysis of CT already acquired for another reason removes the added-radiation cost. PANDA reported 92.9% lesion sensitivity and adjusted 99.9% specificity in 20,530 consecutive clinical scans, but some false positives were reclassified after review. Separate retrospective prediagnosis systems found about 69% to 73% sensitivity; one external-control specificity range was 81.3% to 87.5%. AI radiology therefore has a plausible lead time of months in selected existing scans, but no randomized stage-shift, harm or mortality result. Operations require scanner transport testing, calibrated alerts, radiologist review and a controlled downstream pathway.\n\nPET measures tracer uptake rather than screening anatomy. FDG PET is used for selected staging problems, while FAPI PET and other novel tracers may make fibroblast-rich PDAC more visible. Small selected studies reported better lesion detection and occasional upstaging, but inflammation also takes up tracer. Radiation, tracer production, scan cost and false-positive tissue make repeated population use implausible. The credible use is staging or an imaging-selected radiopharmaceutical test, not early screening. No cited PET or novel-tracer study establishes PDAC screening sensitivity, lead time, regulatory screening status or clinical utility.\n\nAcross imaging, detectable tumour burden is not a single millimetre threshold: contrast, duct obstruction, lesion location, scanner, protocol and reader alter it. Every proposed screening route needs consecutive intended-population enrolment, blinded reads, complete negative follow-up, stage-specific sensitivity and specificity, interval cancers, incidental findings, invasive work-up, cost and cancer outcomes. Availability of a regulated scanner does not mean the screening workflow has a regulated or evidence-based indication.", "sourceIds": [ "blackford-2024-high-risk-surveillance", "dbouk-2022-caps5", "cao-2023-panda-ct-ai", "ygberg-2025-pancanai-validation", "mukherjee-2026-redmod", "pang-2022-fapi-pet", "lee-2024-prospective-fapi-resectable-pdac", "evans-2026-fapi-feasibility", "khuntikeo-2020-thailand-cca-screening" ], "links": [ { "target": "detection", "relation": "part-of" }, { "target": "detection-modality-ranking", "relation": "extends" }, { "target": "technology-opportunistic-ct-ai-pancreas", "relation": "depends-on" }, { "target": "technology-fapi-pet-staging-not-screening", "relation": "depends-on" }, { "target": "claim-cca-surveillance-depends-on-risk-population", "relation": "may-transfer-to" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "strong", "randomised": "none for pancreatic imaging screening", "independentReplication": "moderate", "sampleSize": "From 20-person tracer studies to a 20,530-scan AI evaluation and multi-centre high-risk cohorts", "effect": "Selected high-risk surveillance shows stage shift; no modality has average-risk pancreatic mortality proof", "limits": [ "Modalities are often combined in surveillance", "Retrospective AI cohorts have enriched prevalence", "Scanner and operator effects matter", "Staging performance does not establish screening", "Cost and specialist capacity vary by health system" ] }, "layout": { "x": 110, "y": 66 }, "updated": "2026-09-14" }, { "id": "detection-integrated-workflows", "title": "Detection audit: longitudinal biomarkers, multi-modal combinations and complete screening workflows", "shortTitle": "Combined detection workflows", "type": "system", "status": "supported", "scope": [ "pdac", "pancreatic-all", "hepatobiliary", "colorectal", "pan-cancer", "transfer-unproven" ], "summary": "Combining tests helps only when a locked sequence improves net benefit in the intended population; it can also multiply false alarms and operating cost.", "content": "Longitudinal biomarkers use within-person change rather than one population threshold. CA19-9 trajectories can begin rising about two years before PDAC diagnosis, and routine-laboratory or glycaemic change may define a temporary risk window. The current evidence does not establish one detectable-burden threshold or prospective stage-I sensitivity. A credible study must lock sampling frequency, missing-data rules and an action threshold before outcomes are seen; measure specificity per testing round and per person over years; report lead time against aggressive cancers; and include people with obstruction, pancreatitis and diabetes. Repeated low-cost blood or health-record processing is operationally feasible, but every trigger consumes imaging and specialist capacity.\n\nMulti-modal combinations include mutation plus fragmentomics, CA19-9 plus proteins, or a record-based risk trigger followed by blood and imaging. CancerSEEK showed that a multi-analyte blood test could combine mutations and proteins in diagnosed nonmetastatic cancers, but its healthy-control case-control design did not establish screening utility. Combining weak signals can improve fitted AUC while worsening calibration or false-positive burden. Promotion therefore requires locked external validation at a fixed work-up capacity and proof that the added channel improves net benefit over the best single channel. PANCAID is spending €9.845m on a multi-marker and AI route, but its funded plan uses banked samples before designing a future prospective study. No cited multi-modal combination has a regulated average-risk PDAC screening indication.\n\nMulti-cancer early detection uses one test across cancer types. CCGA reported high specificity but low stage-I sensitivity across cancers. NHS-Galleri was a stronger randomized workflow test and missed its combined late-stage primary endpoint. Pancreatic subgroup estimates are too small to infer a mortality effect, and a tissue-of-origin call still needs imaging, endoscopy or biopsy. Central assay cost, repeat testing, false-positive resolution and capacity must be counted across every cancer, not charged only to pancreatic cases.\n\nThe closest complete PDAC routes are risk-enriched rather than average-risk. Expert surveillance uses inherited or familial risk followed by MRI/MRCP and EUS. A scalable alternative could use new-onset diabetes, pancreatitis, a cyst or an existing CT warning as the population gate, then one blood or record model, pancreas imaging and specialist confirmation. New-onset-diabetes evidence shows why every step must be measured. Prospective surveillance puts three-year risk at 0.62%. A 93-person CT pilot found 52 people with extrapancreatic incidental findings and only one PDAC, already stage IV. PANDOME found one stage-I PDAC among 109 MRI participants after four biopsies. EDI enrolled 8,869 and will measure stage and incidental findings, but it uses non-randomized site pathways and has no posted result. The Onomichi Method supplies broader operating evidence: 29,642 people were assessed in primary care and 925 cancers diagnosed; stage 0/I rose from 13.9% to 22.1% across reported eras. It has no concurrent control and did not improve survival within stage, so it shows a stage shift rather than fewer deaths. Japan's 10,000-person ENZEAVOUR study is testing liquid biopsy followed by investigator-selected MRI/MRCP, EUS or contrast CT. It can measure detection and positive predictive value, but mainly verifies positive tests and is too small for reliable sensitivity or mortality at the ordinary base rate. UK VAPOR2 adds a prospective 6,079-person breath-test cohort in symptomatic or referred patients; it has no posted results and is not a utility trial.\n\nThe missing screening programme is therefore a controlled action loop, not an assay: risk trigger → locked second test → imaging → safe confirmation → treatment that changes outcome. Required outputs are detectable tumour burden, stage-specific sensitivity, specificity, lead time, false-positive procedures, intended population, prospective design, cross-site repeatability, complete negative follow-up, cost, operating capacity, product-level regulatory status and randomized clinical utility. Until one route passes all of them, the honest regulatory and clinical status is no proven general-population pancreatic screening programme.", "sourceIds": [ "fahrmann-2021-ca19-9-trajectory", "cohen-2018-cancerseek", "cristiano-2019-fragmentomics", "liu-2020-ccga-methylation", "sasieni-2026-nhs-galleri", "blackford-2024-high-risk-surveillance", "dbouk-2022-caps5", "chari-2026-prospective-new-onset-diabetes", "sharma-2018-endpac", "wu-2022-nod-imaging-pilot", "frank-2025-pandome", "edi-nct04662879", "hanada-2026-onomichi-method", "jrct-2026-enzeavour-feasibility", "nct07243262-vapor2", "cordis-pancaid-101096309", "seer-pancreas-statfacts" ], "links": [ { "target": "detection", "relation": "part-of" }, { "target": "detection-modality-ranking", "relation": "extends" }, { "target": "trial-enzeavour-japan-screening", "relation": "depends-on" }, { "target": "hypothesis-one-time-risk-triggered-workup", "relation": "tests" }, { "target": "trial-new-onset-diabetes-workup-map", "relation": "depends-on" }, { "target": "claim-screening-base-rate-maths", "relation": "constrained-by" }, { "target": "claim-onomichi-community-pathway-stage-shift", "relation": "supported-by" }, { "target": "lab-imperial-vapor2", "relation": "tested-by" }, { "target": "lab-cnio-genetic-epidemiology", "relation": "developed-by" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "strong", "randomised": "moderate for multi-cancer workflow; none proving pancreatic mortality benefit", "independentReplication": "weak — mixed evidence", "sampleSize": "Includes large multi-cancer studies, several hundred inherited-risk participants, EDI n=8,869, an Onomichi pathway with 29,642 assessments, a planned 10,000-person Japanese assay workflow and VAPOR2 target n=6,079", "effect": "No complete pancreatic screening action loop has shown net benefit or fewer deaths", "limits": [ "Combined routes inherit every component's errors", "Pancreatic subgroup counts are small", "Completed diabetes imaging studies contain one pancreatic cancer each", "Work-up harms are poorly reported", "Costs depend on test order and capacity", "Regulatory status is product and intended-use specific" ] }, "layout": { "x": 109, "y": 75 }, "updated": "2026-09-14" }, { "id": "detection-modality-ranking", "title": "Detection modality map: promising signals still fail the screening-utility test", "shortTitle": "Detection ranking", "type": "system", "status": "supported", "scope": [ "pdac", "pancreatic-all", "hepatobiliary", "colorectal", "pan-cancer", "transfer-unproven" ], "summary": "No blood, cell, RNA, protein, sugar, metabolite, exosome, image or mixed test has yet shown fewer pancreatic-cancer deaths in a population screening trial.", "content": "Tier A — usable now in selected risk groups, not general screening: MRI/MRCP and endoscopic ultrasound surveillance in inherited or familial high-risk people; diagnostic CT/MRI/EUS after symptoms, a risk trigger or a lesion; CA19-9 for disease management with known sensitivity and Lewis-antigen limits. Tier B — closest to a scalable prospective workflow: opportunistic CT AI, new-onset-diabetes enrichment, longitudinal CA19-9 plus companion proteins, tumour-informed DNA for recurrence, and the Japanese ENZEAVOUR feasibility route. None has pancreatic mortality proof. Tier C — technically credible but still case-control or small nested-cohort evidence: methylation, fragment patterns, circulating tumour DNA, extracellular vesicles, protein panels, glycan panels, metabolites, platelet RNA, microRNA and circulating RNA. Examples show the gap. Platelet PLA2Sig reached stage I/II AUC 0.812 across case-control cohorts, but healthy controls do not represent pancreatitis and incidental-lesion clinics. A glycan plus CA19-9 model reached AUC 0.919 and stage-I sensitivity 89.7% in 29 stage-I cases, without external prospective screening. A Swedish prediagnostic metabolite model improved internal AUC from 0.681 to 0.784 in 82 future cases and 82 controls. EPIC microRNAs differed mainly within two years of diagnosis in 225 case-control pairs, not as a validated action rule. Tier D — poor fit for population screening despite biological value: circulating tumour cells are scarce and processing-heavy. At presentation, one 100-patient study reported 75.0% sensitivity and 96.4% specificity; in a 32-patient Korean surgical cohort, CTCs were prognostic, but sampling occurred around operation and did not test presymptomatic detection. Tumour-educated platelets and circulating cells may be useful for biology, staging or monitoring before they are screening tests. Tier E — staging rather than screening: FDG or FAPI PET and novel tracers; radiation, uptake in inflammation and cost make repeated low-risk use unsuitable. Common requirements before promotion: consecutive intended-use population; aggressive-stage-specific sensitivity; specificity high enough for the base rate; complete negative follow-up; positive predictive value; lead time; downstream MRI/EUS/biopsy/surgery harms; locked assay; cross-site reproducibility; cost; regulatory product match; and randomized clinical utility. Combining weak markers can increase overfitting and false alarms; the test must improve net benefit at a fixed work-up capacity.", "sourceIds": [ "blackford-2024-high-risk-surveillance", "botta-2024-tumor-informed-ctdna", "kulemann-2025-platelet-rna-pdac", "wang-2024-nglycan-pdac", "stepien-2024-prediagnostic-metabolomics", "duell-2017-epic-mirna-pdac", "poruk-2016-ctc-pdac-diagnosis", "song-2021-ctc-resectable-pdac", "jrct-2026-enzeavour-feasibility", "wang-2026-cfdna-prospective-detection", "oberg-2026-pcdc-biomarker-bakeoff", "wang-2026-itih3-ceacam1-panel" ], "links": [ { "target": "detection", "relation": "part-of" }, { "target": "claim-screening-base-rate-maths", "relation": "constrained-by" }, { "target": "claim-case-control-blood-tests-overstate-screening-readiness", "relation": "depends-on" }, { "target": "claim-2026-blood-tests-improve-signal-not-screening-utility", "relation": "constrained-by" }, { "target": "technology-opportunistic-ct-ai-pancreas", "relation": "ranks" }, { "target": "biomarker-ca19-9-longitudinal-anchor", "relation": "ranks" }, { "target": "failure-gpc1-exosome-perfect-classifier", "relation": "warned-by" }, { "target": "trial-enzeavour-japan-screening", "relation": "ranks" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "strong", "randomised": "weak", "independentReplication": "weak — mixed evidence", "sampleSize": "Ranking across population, high-risk, case-control, nested-cohort and workflow studies", "effect": "No cited method has pancreatic-cancer mortality proof in population screening", "limits": [ "Not every assay variant has its own node", "Comparisons use different intended populations", "Several studies report AUC rather than fixed operating points", "Technology changes quickly", "Clinical work-up capacity differs by country" ] }, "layout": { "x": 118, "y": 56 }, "updated": "2026-09-14" }, { "id": "detection-rna-protein-metabolite-methods", "title": "Detection audit: RNA, vesicles, proteins, glycans and metabolites", "shortTitle": "RNA and molecular panels", "type": "system", "status": "supported", "scope": [ "pdac", "pancreatic-all", "hepatobiliary", "colorectal", "pan-cancer", "transfer-unproven" ], "summary": "Several blood signals discriminate diagnosed PDAC from selected controls; none yet supplies a repeatable low-prevalence action rule with mortality evidence.", "content": "Tumour-educated platelets are assessed through platelet RNA rather than a tumour-specific cell. A 673-sample multicentre case-control study reported stage-I/II AUC 0.812 for a two-RNA signature. AUC does not reveal sensitivity at the extreme specificity needed for screening, and healthy controls do not reproduce pancreatitis, biliary obstruction or incidental-lesion clinics. Detectable burden, lead time, population repeatability, fixed cost, regulatory indication and clinical utility remain unknown. Platelet isolation and RNA handling add operating steps.\n\nMicroRNA and broader circulating RNA are attractive because signals may be amplified without tumour-DNA shedding. In 225 nested EPIC case-control pairs, plasma microRNA differences were strongest within two years of diagnosis. That provides a possible lead-time range but not a locked stage-I operating point. Independent assays, collection tubes and normalisation differ, so reproducibility is mixed. No pancreatic screening indication or utility trial exists. Central RNA processing is feasible but more complex than routine chemistry.\n\nExosomes and other extracellular vesicles can contain tumour RNA and protein, but isolation method and benign inflammation strongly change results. The original GPC1 report claimed 100% sensitivity and specificity; an independent benign-disease comparison found 26.7% sensitivity and 87.5% specificity for GPC1-positive vesicles. This is direct evidence that a curated case-control result can collapse in the intended differential diagnosis. Detectable burden, lead time, cost, cross-site repeatability, regulatory status and clinical utility remain unresolved for newer multi-analyte vesicle assays.\n\nProteomics includes CA19-9 and companion protein panels. In stored PLCO samples, CA19-9 within six months reached 60% sensitivity at 99% specificity overall and 50% for early-stage cases; against benign pancreatic disease, early-stage sensitivity fell to 30%. Adding LRG1 and TIMP1 recovered some cases below the CA19-9 threshold. The marker can rise roughly two years before diagnosis in some people, but Lewis non-secretors and benign obstruction create false negatives and positives. Laboratory cost is relatively low, operations are simple and CA19-9 is clinically available for management, but it is not an average-risk screening test.\n\nGlycomics measures altered sugar structures on circulating molecules. One N-glycan plus CA19-9 model reported AUC 0.919 and 89.7% sensitivity among only 29 stage-I cases, without external prospective screening. Metabolomics may capture host and tumour metabolism: a Swedish nested study of 82 future cases and 82 controls improved internal AUC from 0.681 to 0.784. Both approaches need locked assays, extreme-specificity operating points, benign-disease controls, useful lead time, cross-site repeatability, cost studies and a clinical action test. Neither has a regulated pancreatic screening indication. Immune-response signatures remain earlier: systemic inflammation is biologically plausible but nonspecific, and the current primary-evidence set does not support a stable stage-I sensitivity, specificity or lead-time estimate. That absence is a result, not permission to infer accuracy.", "sourceIds": [ "kulemann-2025-platelet-rna-pdac", "duell-2017-epic-mirna-pdac", "melo-2015-gpc1-exosomes", "karam-2019-gpc1-nonreplication", "fahrmann-2021-ca19-9-trajectory", "wang-2024-nglycan-pdac", "stepien-2024-prediagnostic-metabolomics" ], "links": [ { "target": "detection", "relation": "part-of" }, { "target": "detection-modality-ranking", "relation": "extends" }, { "target": "failure-gpc1-exosome-perfect-classifier", "relation": "warned-by" }, { "target": "biomarker-ca19-9-longitudinal-anchor", "relation": "depends-on" }, { "target": "claim-screening-base-rate-maths", "relation": "constrained-by" } ], "evidence": { "humanProspective": "weak", "humanRetrospective": "moderate", "randomised": "none", "independentReplication": "weak — mixed evidence", "sampleSize": "Largest cited modality cohort contains 673 platelet samples; stage-I subsets and prediagnostic cohorts are much smaller", "effect": "Promising case-control or stored-sample discrimination without pancreatic screening utility", "limits": [ "Different analytes and platforms are not interchangeable", "Several reports give AUC rather than fixed sensitivity and specificity", "Few stage-I cases", "Benign pancreatic disease is often underrepresented", "Regulatory and cost claims require product-level checks" ] }, "layout": { "x": 111, "y": 57 }, "updated": "2026-09-14" }, { "id": "detection", "title": "Prevention and early detection", "shortTitle": "Early detection", "type": "system", "status": "supported", "scope": [ "pdac", "pnet", "cholangiocarcinoma", "gallbladder", "hcc", "colorectal" ], "summary": "Stage-specific sensitivity, specificity, lead time, false-positive burden, cost and clinical utility are tracked separately.", "content": "There is no proven general-population PDAC screening programme. At an annual incidence of 13.9 per 100,000, a hypothetical 100%-sensitive test with 99% specificity creates about 1,000 false positives for 13.9 true cancers: roughly 72 false positives per true positive and a positive predictive value of 1.37%. At 99.9% specificity the positive predictive value is only about 12.2%; around 99.99% specificity is needed to reach 58.2% before losses in sensitivity and work-up are counted. The strongest current practice is MRI/MRCP and endoscopic ultrasound in selected inherited-risk cohorts at expert centres. Surveillance cohorts show more stage I disease and better observed survival, but are nonrandomized, predominantly White and cover a minority of future cases. New-onset diabetes, pancreatitis, cysts, family history and changing routine laboratory data can enrich risk; no single feature is selective enough, so the plausible workflow is temporary risk trigger → inexpensive second-stage blood or record model → pancreas imaging → specialist confirmation. New-onset diabetes raises rough three-year risk from about 41.7 to 620 per 100,000, but completed work-up evidence contains only one pancreatic cancer per study. A 93-person CT pilot found 56% with extrapancreatic incidental findings and one stage-IV PDAC. PANDOME found one stage-I PDAC among 109 MRI participants after four biopsies. EDI enrolled 8,869, assigns pathways by site, uses ENDPAC above 0 and has no posted result. Blood evidence is much weaker than commercial language suggests. CA19-9 is useful longitudinally but misses Lewis-negative patients and rises in benign obstruction. Methylation, fragmentomics, proteins, glycans, metabolites, platelet RNA, microRNA, exosomes, circulating tumour cells and multi-cancer tests mostly rely on diagnosed case-control samples, have limited stage I counts, uncertain lead time or no pancreatic-specific utility. CCGA reported 99.3% specificity but only 18% stage I sensitivity across all cancers; NHS-Galleri later missed its randomized combined stage III/IV endpoint. AI on existing abdominal CT is attractive because it adds no scan radiation and may recover overlooked warning signs, but it needs silent prospective validation across scanners, hospitals and pre-diagnosis intervals. MRI, CT, EUS, PET and FAPI imaging remain confirmation or staging tools rather than cheap population screens. A useful programme must report stage-specific sensitivity, specificity, lead time, positive predictive value, false-positive procedures, complications, overdiagnosis, cost and cancer mortality. The missing capability is not another high case-control AUC. It is a validated risk-enriched action path that finds dangerous disease early enough for an intervention that changes outcome.", "contentSections": [ "Three 2026 blood-test reports sharpen the validation ladder. A prospective cfDNA model found 6 of 8 PDAC cases in 1,926 people with diabetes or obesity, but also returned 40 false positives: PPV 13.0%, with no test of outcome benefit. Two larger blinded assay studies used selected known cancer cases. One PCDC panel reached AUC 0.963 against mainly healthy controls; another three-protein model reached AUC 0.917, but specificity was only 53.3% at roughly 95% sensitivity. These are useful signal and assay results, not evidence for population screening.", "Pancreatic precursor detection is a separate action path. Visible IPMN requires three decisions: whether HGD or cancer is already present, whether a low-grade cyst will progress later, and whether a separate PDAC will arise elsewhere. In a 2,104-person prospective Japanese cohort, the five-year routes were similar in size—1.90% HGD or invasive disease inside the IPMN and 2.11% separate PDAC—but had different predictors. A cyst-centred model therefore misses part of the risk.", "PancreaSeq GC detected current advanced neoplasia with 86.6% sensitivity and 97.9% specificity in 241 sampled patients, 97 of whom already had advanced disease. That does not establish future-progression prediction in a routine low-risk cohort. The operational gate is net benefit, not AUC: fewer harmful operations without moving resection so late that invasion replaces HGD as the dominant finding.", "Gallbladder prevention needs the same separation. POLYP found one HGD and no cancer among 68 operations indicated by the revised European polyp rule. GBCseeker and GAIA-MIL classify existing visible lesions, not future risk. At 2% prevalence, transporting their external sensitivity and specificity without loss would still create about 12 and 17 false positives per true positive, respectively. Those calculations show the base-rate problem; they are not observed screening results.", "EULAT, Chile BiLS and P-iGBC create useful recruitment and sample networks, but their case-control, operated or after-removal designs answer different decisions. A country-specific prevention study must join future cancer incidence to local elective surgery harm, then randomize the action after silent external validation.", "pNET detection must separate inherited surveillance, current grade and future danger. DutchMEN data support pancreatic imaging from adolescence in MEN1, but five clinically relevant tumours among 350 young people and two later nodal metastases do not define an optimal schedule. A prospective 110-person circulating mRNA assay failed to predict new tumour, growth or metastasis. For an incidental 1-2 cm sporadic tumour, the useful output is not another tumour label; it is a time-updated estimate of metastatic risk compared with the harm of the required operation.", "Cholangiocarcinoma detection has three different denominators. In PSC, annual MRI/MRCP generated severe or progressive duct findings in 23.8% of 512 people and roughly nine non-malignant such findings per malignancy, without durable curable detection. In liver-fluke regions, ultrasound shifted stage among diagnosed cancers but has no randomized population-mortality result. After resection, ctDNA can identify recurrence months before imaging but has no proven action. These are risk surveillance, population screening and residual-disease monitoring; their yields and harms cannot be pooled.", "HCC shows why detection must be judged as a complete route. Six-month ultrasound plus AFP reduced HCC mortality in one older randomized programme, but newer test comparisons have not repeated that endpoint. Noncontrast MRI reduced false referrals and shifted stage without showing fewer deaths; three-month ultrasound found more tiny lesions without more small HCC; adding AFP, AFP-L3 and DCP did not improve early-stage detection. TRACER tests late-stage disease and work-up harm for GALAD, while PREMIUM tests HCC mortality for abbreviated MRI. Both are recruiting and have no results.", "Colorectal residual-disease detection now has an action result and a survival gap. FIND randomized 584 eligible patients and used positive methylation ctDNA to trigger immediate CT. Among patients who recurred, curative-intent treatment rose from 23.6% to 48.1% and liver or lung-confined curative resection rose from 18.2% to 42.3%. Median follow-up was 23.3 months, so earlier detection and more surgery cannot yet be called longer survival. DYNAMIC-III and ALTAIR show that using a positive result to give more systemic treatment can fail even when the test strongly predicts recurrence." ], "sourceIds": [ "uspstf-2019-pancreatic-screening", "seer-pancreas-statfacts", "blackford-2024-high-risk-surveillance", "dbouk-2022-caps5", "chari-2026-prospective-new-onset-diabetes", "sharma-2018-endpac", "wu-2022-nod-imaging-pilot", "frank-2025-pandome", "edi-nct04662879", "fahrmann-2021-ca19-9-trajectory", "liu-2020-ccga-methylation", "sasieni-2026-nhs-galleri", "cao-2023-panda-ct-ai", "ohtsuka-2024-bd-ipmn-surveillance", "mirzaian-2026-low-risk-cysts", "singhi-2026-pancreaseq-gc", "pea-2026-post-surveillance-resection", "henn-2023-cyst-surgery-registry", "van-dijk-2025-polyp-study", "yang-2025-gbcseeker", "gupta-2026-gaia-mil", "clinicaltrials-nct06192719-eulat-gbc", "clinicaltrials-nct06531408-pigbc", "wong-2025-ambrose-cholecystectomy", "klein-haneveld-2021-men1-young-screening", "van-vliembergen-2026-men1-mnet-assay", "partelli-2022-aspen-interim", "villard-2023-psc-surveillance", "khuntikeo-2020-thailand-cca-screening", "yu-2025-btc-ctdna-recurrence", "yoo-2025-extrahepatic-cca-ctdna", "rhee-2025-miracle-hcc", "hirode-2026-hcc-biomarker-rct", "trinchet-2011-hcc-surveillance-interval", "singal-2024-tracer-protocol", "clinicaltrials-nct06084234-tracer", "ioannou-2026-premium-protocol", "clinicaltrials-nct05486572-premium", "tie-2025-dynamic3", "bando-2026-altair", "mo-2026-find-crc-surveillance", "wang-2026-cfdna-prospective-detection", "oberg-2026-pcdc-biomarker-bakeoff", "wang-2026-itih3-ceacam1-panel" ], "links": [ { "target": "bottlenecks", "relation": "supports" }, { "target": "failures", "relation": "supports" }, { "target": "experiments", "relation": "tests" }, { "target": "detection-modality-ranking", "relation": "contains" }, { "target": "claim-2026-blood-tests-improve-signal-not-screening-utility", "relation": "contains" }, { "target": "detection-dna-cell-methods", "relation": "contains" }, { "target": "detection-rna-protein-metabolite-methods", "relation": "contains" }, { "target": "detection-imaging-methods", "relation": "contains" }, { "target": "detection-integrated-workflows", "relation": "contains" }, { "target": "prevention-ranking", "relation": "contains" }, { "target": "trial-new-onset-diabetes-workup-map", "relation": "contains" }, { "target": "claim-ipmn-has-two-cancer-routes", "relation": "contains" }, { "target": "claim-cyst-classifier-current-disease-not-future-risk", "relation": "contains" }, { "target": "trial-pancreatic-precursor-action-map", "relation": "contains" }, { "target": "hypothesis-dynamic-cyst-interception-rule", "relation": "tests" }, { "target": "trial-gallbladder-prevention-detection-action-map", "relation": "contains" }, { "target": "claim-gallbladder-diagnostic-models-existing-lesions-only", "relation": "contains" }, { "target": "hypothesis-gallbladder-two-stage-interception-rule", "relation": "tests" }, { "target": "trial-small-pnet-action-map", "relation": "contains" }, { "target": "claim-men1-surveillance-finds-lesions-not-danger", "relation": "contains" }, { "target": "hypothesis-pnet-time-updated-surgery-rule", "relation": "tests" }, { "target": "trial-cca-action-map", "relation": "contains" }, { "target": "claim-cca-psc-surveillance-workload", "relation": "contains" }, { "target": "claim-cca-ctdna-prognostic-not-action-guiding", "relation": "contains" }, { "target": "claim-hcc-surveillance-test-improvement-not-mortality", "relation": "contains" }, { "target": "trial-hcc-action-map", "relation": "contains" }, { "target": "unknown-hcc-surveillance-modality-utility", "relation": "contains" }, { "target": "claim-crc-ctdna-utility-is-action-specific", "relation": "contains" }, { "target": "trial-crc-action-map", "relation": "contains" }, { "target": "unknown-crc-ctdna-positive-action", "relation": "contains" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "strong", "randomised": "moderate", "independentReplication": "strong", "sampleSize": "Population incidence, high-risk surveillance cohorts, new-onset-diabetes risk and work-up studies, large multi-cancer studies and modality-specific validation series", "effect": "At 13.9 per 100,000 annual incidence and 99% specificity, about 72 false positives occur per true positive even at 100% sensitivity", "limits": [ "Incidence varies by age and country", "High-risk pancreatic surveillance is nonrandomized", "Completed diabetes imaging studies contain one pancreatic cancer each", "EDI is non-randomized by site and unfinished", "Multi-cancer results do not equal pancreatic performance", "Clinical utility and mortality evidence are missing for most biomarkers", "Modern HCC test comparisons lack mortality results", "Gallbladder diagnostic models start after lesions are visible", "Gallbladder prevention lacks a validated country-specific action rule", "Cost estimates vary by health system" ] }, "layout": { "x": 31, "y": 78 }, "updated": "2026-09-14" }, { "id": "disease-system", "title": "Disease systems", "shortTitle": "Disease systems", "type": "system", "status": "supported", "scope": [ "pdac", "pnet", "cholangiocarcinoma", "gallbladder", "hcc", "colorectal" ], "summary": "Mechanistic chains from susceptibility and initiation through dissemination, resistance, residual disease, recurrence and death.", "content": "PDAC is the central lethal system. Germline susceptibility, smoking, excess body weight, abnormal glucose and pancreatitis alter risk, but none explains most cases. PanIN and some IPMN acquire KRAS early; later loss of CDKN2A, TP53 and SMAD4 and changing gene control allow invasion. These mutations define lineage better than they predict exactly which microscopic lesion will become dangerous. Cancer cells occupy shifting classical, basal-like and stress states rather than fixed diseases. Treatment can select a reversible state without a new resistance mutation, and model lineage work weakens the idea that a permanent rare stem-cell minority alone maintains disease. Dense fibroblast and matrix programmes, irregular vessels, myeloid suppression, nerve interaction and redundant nutrient use restrict delivery and immune attack without supplying one safe stromal target. Human tumour transport varies greatly; phase III hyaluronan removal increased response but not survival. Schwann cells can respond to tumour state and tissue force, but causal treatment evidence is still confined to models. Dissemination timing varies: some tumours seed early, some late, and some patients die mainly from uncontrolled local disease. Surgery removes visible local tissue but not occult systemic clones; treatment selects resistant states; residual disease becomes radiographic recurrence. Dormant-cell persistence and reawakening are causal in PDAC models but not yet directly defined in people. Host wasting is another branch: rapid muscle loss marks about twice the death hazard in one longitudinal cohort, yet blocking IL-6 preserved muscle without improving the main survival endpoint. That chain explains why detection, local control, muscle preservation or tumour shrinkage alone cannot be the mission endpoint. The related cancers must stay separate. Pancreatic neuroendocrine tumours more often use MEN1, DAXX/ATRX and mTOR biology, are usually more vascular and can be controlled with somatostatin, targeted drugs and radioligands. Cholangiocarcinoma differs by intrahepatic, perihilar and distal origin, with targetable FGFR2 fusions and IDH1 mutations concentrated in subsets. Gallbladder cancer often arises after long inflammation and gallstone exposure but lacks a safe population-selection rule for prophylactic surgery. HCC usually evolves in chronically injured liver, so viral prevention and liver function remain part of the cancer model. Colorectal cancer often offers a visible removable precursor and a longer interception window; liver metastasis can remain locally actionable in a selected minority. The shared engineering problem is a chain of observation and action: identify risk or minimal disease, classify the lethal state repeatedly, deliver a tolerable treatment at measured exposure, preserve host capacity, measure extinction rather than temporary response, and intervene again before resistant burden becomes clinically large. Each linked disease page records which link is biological, measurable, treatable or still unknown. Inherited susceptibility is also a family route: the tumour-bearing patient is the fastest entry point for finding a familial variant, but the biological finding produces no preventive effect unless consenting relatives receive testing and a gene-specific action.", "sourceIds": [ "gudmundsdottir-2023-staging-laparoscopy", "rompen-2026-preopanc2-staging", "theijse-2024-nontherapeutic-laparotomy", "tcga-2017-pdac", "chan-seng-yue-2020-evolution-subtypes", "singhal-2024-classical-kras-resistance", "lodestijn-2021-environment-clonogenicity", "koay-2014-human-pdac-drug-transport", "hingorani-2020-halo301", "fujii-nishimura-2018-schwann-met", "shi-2026-chemo-dormant-dtc", "babic-2019-muscle-loss-survival", "chen-2025-tocilizumab-cachexia", "yachida-2010-metastatic-timing", "iacobuzio-donahue-2009-autopsy-patterns", "scarpa-2017-pnet", "jusakul-2017-cholangiocarcinoma", "tcga-2017-hcc", "hu-2019-crc-early-seeding", "chen-2023-fdr-cancer-risk", "wang-2022-oncology-germline-cascade", "npaca-2026-state-of-nation", "nhs-england-2024-hpb-pancreatic-service", "nice-2018-ng85-pancreatic", "de-wilde-2012-centralization", "lemmens-2011-centralization", "latenstein-2021-centralization", "coupland-2016-england-resection-volume", "labori-2016-adjuvant-completion", "henry-2023-complications-adjuvant", "kollbeck-2025-complications-chemo", "van-hilst-2019-leopard2", "korrel-2023-diploma", "bruna-2025-diploma-long-term", "klotz-2024-europa", "liu-2024-robotic-open-pd", "jin-2026-portal", "effenberger-2012-pdac-bone-marrow-dtc", "nordgard-2022-unresectable-dtc", "chalabi-hajkarim-2025-organotropism", "pei-2025-spatial-metastatic-pdac", "mcclellan-2026-quiescent-pdac-car-t", "tanaka-2012-microscopic-local-recurrence", "dotan-2024-giant", "dotan-2025-giant-ga", "clinicaltrials-nct04233866-giant", "betge-2022-grantpax", "clinicaltrials-nct02143219-pamela70", "mohile-2021-gap70", "li-2021-gain", "ihorst-2026-geriatric-comanagement", "clinicaltrials-nct06040801-frail-gi", "stoop-2026-preopanc4-surgery", "dekker-2026-taps-restaging", "verkolf-2026-taps-resection-prediction", "hartlapp-2022-neolap-ca199", "guggenberger-2023-neolap-ct" ], "links": [ { "target": "claim-staging-laparoscopy-avoids-incision-not-proven-survival", "relation": "includes" }, { "target": "unknown-pdac-staging-laparoscopy-rule", "relation": "leaves-open" }, { "target": "bottlenecks", "relation": "supports" }, { "target": "unknowns", "relation": "supports" }, { "target": "model-pdac-causal-chain", "relation": "contains" }, { "target": "model-pnet-causal-chain", "relation": "contains" }, { "target": "model-cholangiocarcinoma-causal-chain", "relation": "contains" }, { "target": "model-gallbladder-causal-chain", "relation": "contains" }, { "target": "model-hcc-causal-chain", "relation": "contains" }, { "target": "model-colorectal-causal-chain", "relation": "contains" }, { "target": "claim-pdac-resistance-uses-reversible-cell-states", "relation": "contains" }, { "target": "claim-pdac-stemness-is-dynamic", "relation": "contains" }, { "target": "claim-pdac-delivery-is-variable-not-single-barrier", "relation": "contains" }, { "target": "claim-pdac-dormancy-model-proven-human-uncertain", "relation": "contains" }, { "target": "claim-cachexia-is-parallel-lethal-system", "relation": "contains" }, { "target": "system-england-r367-germline-criteria", "relation": "contains" }, { "target": "claim-universal-germline-testing-needs-action", "relation": "contains" }, { "target": "claim-pdac-surgery-is-a-complete-treatment-route", "relation": "links-to" }, { "target": "system-england-pancreatic-surgery-network", "relation": "links-to" }, { "target": "failure-pdac-postoperative-treatment-loss", "relation": "links-to" }, { "target": "failure-leopard2-laparoscopic-whipple-safety", "relation": "links-to" }, { "target": "unknown-pdac-surgical-route-completion", "relation": "links-to" }, { "target": "trial-pdac-surgical-route-action-map", "relation": "links-to" }, { "target": "hypothesis-pdac-surgical-route-completion", "relation": "links-to" }, { "target": "claim-human-pdac-dtc-prognostic-not-dormancy", "relation": "bounded-by" }, { "target": "failure-pdac-dormancy-sample-gap", "relation": "bounded-by" }, { "target": "trial-pdac-residual-cell-state-action-map", "relation": "bounded-by" }, { "target": "claim-age-alone-cannot-route-pdac-treatment", "relation": "contains" }, { "target": "failure-grantpax-geriatric-route", "relation": "contains" }, { "target": "unknown-older-pdac-treatment-route", "relation": "contains" }, { "target": "trial-older-pdac-action-map", "relation": "contains" }, { "target": "claim-lapc-response-warrants-surgical-reassessment", "relation": "contains" }, { "target": "failure-conversion-surgery-selection-bias", "relation": "bounded-by" }, { "target": "unknown-lapc-surgery-causal-benefit", "relation": "contains" }, { "target": "trial-lapc-conversion-surgery-action-map", "relation": "maps" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "strong", "randomised": "moderate", "independentReplication": "strong", "sampleSize": "Synthesis across genomic atlases, human autopsy and evolution studies, cohorts and treatment evidence", "effect": "Mechanistic causal synthesis; no pooled effect estimate", "limits": [ "Several transitions are inferred from cross-sectional tissue", "Dissemination timing varies between patients", "Model systems incompletely reproduce human treatment and immunity", "Related cancers share anatomy but not one molecular system" ] }, "layout": { "x": 21, "y": 52 }, "updated": "2026-09-14", "contentSections": [ "Hidden metastatic disease creates a measurement boundary before treatment and surgery. Pancreas-protocol CT and PET can label a tumour localized while millimetre-scale liver or abdominal-surface deposits already exist. A camera can see some surface disease; biopsy can confirm a lesion; wash cytology can find shed cells; none can prove that no systemic cells remain. These states must not be collapsed into one localized-versus-metastatic switch.", "The PDAC surgery branch adds the care system between local disease and systemic control. A patient must reach an expert review, receive the justified operation from a proficient team, survive and recover, then receive the planned drug treatment. Current England data report 23 specialist centres, 3.6% 90-day mortality after Whipple and 62% postoperative treatment within 14 weeks. The graph therefore treats resection as a route step, not a cure label.", "Dissemination, persistence and dormancy now stay separate. A tumour-cell signal outside the pancreas proves neither that the cell is alive nor that it is resting; late recurrence proves neither continuous growth nor reawakening. Human marrow studies support occult dissemination and risk. Organ-linked primary states and end-stage spatial plasticity support multiple routes. Only joined ancestry and time can call a residual state.", "Host reserve is now explicit in the PDAC system. The tumour state and the treatment system interact with nutrition, function, cognition, other illness, medicines, mood, transport and caregiver capacity. These are not background details: they can determine whether treatment starts, how much is delivered, whether harm causes admission and whether any tumour response becomes useful life.", "Locally advanced disease now has a state transition rather than a fixed inoperable label. Systemic treatment acts as both treatment and a time test: rapid spread exits the local route, while continued confinement, falling CA19-9, stable or smaller tumour and preserved physical reserve can trigger expert reassessment. Technical removal, expected survival and added benefit from surgery are three different functions. Existing models estimate the first two; the third still requires random allocation." ] }, { "id": "evidence-quality", "title": "Evidence quality: show the reasons, not one confidence number", "shortTitle": "Evidence quality", "type": "system", "status": "supported", "scope": [ "pdac", "pnet", "cholangiocarcinoma", "gallbladder", "hcc", "colorectal", "pan-cancer", "transfer-unproven" ], "summary": "488 decision records separate human study design, sample, effect, biological support, reproducibility, conflicts, age and limits.", "content": "A single confidence score would hide the reasons a result could fail. The atlas therefore keeps prospective and retrospective human evidence, randomization, independent repeat, sample size, observed effect and explicit limits separate. 488 decision records now also score six dimensions requested by the mission: biological plausibility, animal evidence, in-vitro evidence, reproducibility, commercial conflicts and recency. Ratings are not interchangeable. A large randomized trial can have strong human evidence and weak independent reproduction. A mechanism can be biologically convincing but clinically unproved. A current result can still be sponsor-dependent. Not-applicable means the claim does not rely on that evidence class; it does not mean favourable evidence. Conflict ratings describe risk of bias, not proof of misconduct. Recency describes when the decisive evidence was reported, not its truth. The scored set covers the mission conclusion, coverage audit, bottleneck ranking, experiment portfolio, recommended strategy, every programme and timeline page, AI and company rankings, common-RAS phase 3 and paired resistance evidence, durable-RAS uncertainty, adjuvant mFOLFIRINOX, inherited-risk surveillance, population-screening maths, colorectal MRD transfer, HCC and biliary decisions, the GRAIL utility and capital failure, Natera's operating-access stack, leading RAS-resistance, MRD and low-burden immune hypotheses, regional people and delivery networks, HRS-4642 clinical and patent access, the PDAC ctDNA action map, multicell immune failure, germline and pancreatitis risk, the modifiable-risk ceiling, the pancreatic primary-prevention action gap, the selected-local-consolidation hypothesis, the new-onset-diabetes claim, its work-up study map, the revised two-stage hypothesis, the full prevention action audit, IPMN's two cancer routes, present-disease versus future-risk classification, the surgery-timing trade-off, the precursor trial map, two trial-delivery failures, the dangerous-precursor unknown and the dynamic interception hypothesis. The cachexia pass adds the host-wasting and IL-6 claims, the GDF-15 result, the tocilizumab failure, the utility unknown, the live-trial action map, Pfizer and its access stack. The pancreatic-enzyme pass adds one bounded claim, two failures, one utility unknown and one action map. The clot-prevention pass adds eleven records: a pooled treatment claim, two failures, one selection-and-delivery unknown, one trial map, a new-product claim, its company and access stack, one delivery hypothesis, one investigator and one academic network. The biliary-obstruction pass adds seven records: a bounded treatment-route claim, two failures, one whole-route unknown, an action map, one route-completion hypothesis and one device company. The germline-delivery pass adds six records: an action-limited claim, a two-stage route failure, the current England R367 system boundary, one delivery unknown, one completed-evidence map and one family-route hypothesis. The latest regional pass adds the African pancreatic data-quality claim, the sub-Saharan HCC route failure, the Latin American CCA route claim, a shared uncertainty, a tested hypothesis design, an action map, an outcomes lead and a five-country registry route. The programme-level gap is closed: no proposed capital release can omit these reasons. The next expansion is every remaining claim that can change care or justify a new experiment. Opposing findings remain linked to the claim rather than averaged away.", "contentSections": [ "The hidden-spread staging pass adds six scored records. It keeps direct view, biopsy and abdominal-wash cytology separate; distinguishes an avoided incision from survival; and treats procedure timing, delay and false-negative disease as part of the evidence. PREOPANC-2 supplies the strongest route signal but did not randomize staging itself. The proposed 500-person check is inside the existing English surgical-route study and adds no capital.", "The current exact count is 488 fully scored decision records. The gallbladder additions cover polyp yield, operation harm, existing-lesion models, the action map and future-risk selector. Seven pNET additions cover the small-tumour observation and surgery trade-off, inherited-risk surveillance, failed mRNA prediction, the unresolved operation selector, study map and time-updated action hypothesis. Nine cholangiocarcinoma additions and seven older cholangiocarcinoma records now separate PSC workload, liver-fluke exposure control, postoperative ctDNA prognosis, three design failures, the residual-disease action unknown, the study map and a route-ranking hypothesis.", "The HCC scoring pass covers 22 records added to or completed in the scored set: the disease and causal-chain pages, viral and alcohol prevention, surveillance mortality and test limits, local treatment, recurrence routes, transplant, repeat treatment, systemic combinations, tumour-versus-liver failure, four failures, two uncertainties, the action map and the liver-reserve hypothesis. Each now states biological support, model limits, repeat evidence, commercial interests and age.", "The colorectal pass adds six action records and completes ten older branch records. It separates one-colonoscopy invitation from attendance, stage II treatment omission from stage III rules, molecular-relapse treatment from earlier imaging, and narrow liver-only control from broad liver-directed treatment.", "The ownership pass scores Natera and a separate MRD access record. It distinguishes a patent family from the laboratories, assay and algorithm version, data, supply continuity, turnaround, regulatory state and publication rights needed to run an independent action trial.", "The regional delivery pass scores eight new records. It separates proof of a large route failure from proof of its cause, treats post-treatment survival as selected rather than causal, distinguishes referral registries from population capture and makes the first £5m data stage block the conditional £15m intervention stage.", "The cachexia pass scores eight records. It distinguishes weight or muscle change from function, chemotherapy delivered, quality-adjusted survival and OS, and it prevents a duplicate 982-person trial unless a named gap remains.", "The pancreatic-enzyme pass scores five records. It separates maldigestion relief from an unproved survival claim, preserves two negative randomized nutrition results, records the prescription and supply failures, and blocks a duplicate broad trial while three existing studies remain incomplete.", "The PDAC clot-prevention pass scores eleven records. It separates a reproduced reduction in VTE from the failed survival claim, distinguishes bleeding uncertainty from no detected increase, records two delivery and reporting failures, keeps Kinisoquin marker activity below randomized clinical evidence and funds a route-completion experiment rather than another broad efficacy trial.", "The biliary-obstruction pass scores seven records. It preserves the 2010 randomized harm from routine delayed plastic drainage, bounds the 2026 metal-stent noninferiority result, separates patency from cancer control, records treatment interruption without making it causal survival proof and directs new money to whole-route delivery rather than another broad device comparison.", "The surgical-route pass adds seven scored records: one complete-route claim, the current English network boundary, two failures, one uncertainty, one trial map and one randomized delivery hypothesis. It separates centre volume from capability, head resection from left-sided resection, short recovery from cancer control, and treatment receipt from full-course completion. Current English audit values anchor the programme: 96.4% 90-day survival after Whipple, 62% postoperative treatment within 14 weeks and centre IQR 54-68%. Procedure trials keep their own operation, platform, learning-curve and funding limits.", "The post-resection follow-up pass adds five scored records: the bounded routine-imaging claim, one utility uncertainty, a non-duplicate trial map, the RADAR-PANC network and its linked investigator. Two observational studies associate scheduled imaging with more recurrence treatment and longer survival, while RADAR-PANC already tests overall survival and quality of life. The evidence score stays moderate until randomization reports; the £500m plan creates no new trial or budget line.", "The perioperative-timing pass adds eight scored records: a revised bounded claim, the NORPACT-1 delivery failure, one route uncertainty, one action map, two trial leaders and two delivery networks. Nine randomized trials show a stable event-free-survival gain but not a conclusive overall-survival gain. PREOPANC, NORPACT-1 and CISPD-1 point in different survival directions because patient state, regimen and delivery differ. A021806 is the live modern gate, so the £500m plan adds no trial or budget line.", "The human residual-cell pass adds eight scored records: five previously open timing, dormancy and experiment records plus a human disseminated-cell claim, a sample-gap failure and an action map. Bone-marrow signals reproduce prognosis, not dormancy. Primary and rapid-autopsy tissue map route and end-stage state, not a cell during the disease-free interval. The 250-person test sits inside the existing £25m observatory and cannot release an intervention without ancestry, timing, cross-laboratory and added-prediction gates.", "The older-and-vulnerable PDAC pass adds seven fully scored records. It separates chronological age, the clinician activity score, a structured function-and-needs inventory, treatment choice and completed support. GIANT, GrantPax, PAMELA-70 and three randomized mixed-cancer action studies keep their different questions and limits visible.", "The locally advanced surgery pass adds five fully scored decision records and eleven primary records. It separates technical removal, prognosis and treatment effect. PREOPANC-4 supplies current feasibility and harm; NEOLAP and TAPS supply response and selection evidence; PELICAN supplies a randomized local-treatment warning. No modern randomized resection-versus-continued-treatment result was found, so the proposed 540-person test remains an unfunded portfolio candidate rather than a clinical claim.", "Organoid evidence adds a denominator rule. Sensitivity and specificity among successful cultures answer conditional prediction; they do not measure all-patient utility. Report enrolled, adequate tissue, malignant culture, testable culture, on-time result, available recommendation, treatment delivered and outcome as separate counts. HOPE, ORGANOPREDICT, PASS-01 and AVATAR show that deleting the failed steps can turn a low-delivery system into an apparently accurate assay.", "Passes 50 to 62 add scored records for radiofrequency heat, histotripsy, freezing, microwave heat, phosphorus-32, light-activated treatment, laser heat, antibody-light treatment, sonodynamic treatment, pressurised and heated abdominal chemotherapy, nerve ablation and ferroptosis. Each keeps technical delivery separate from complete tumour coverage, patient benefit, survival and harm.", "Passes 63 and 64 add the NRG1 rare-fusion route and early palliative care. The first separates response from durable control and test access; the second separates quality of life and care near death from survival. Their programme gates use the same scored records as the claims and action maps.", "Pass 65 adds five scored records for DPYD and UGT1A1 treatment safety. Prospective carrier evidence and regulator action support DPYD testing now; historical controls, mixed cancers, small carrier groups, ancestry coverage and a failed pancreatic tolerability bound keep the whole pancreatic route open.", "Pass 66 adds five scored records for KRAS-wild-type driver routing. Repeated profiling supports the 10.7% branch and rare target sensitivity, but case-level treatment, tumour-wide approvals and a selected matched-treatment survival comparison cannot become one pooled efficacy estimate.", "Pass 67 adds five scored exercise and rehabilitation records. Seven randomized studies support safe activity and some function change, but primary quality, clinical or complication endpoints are negative or incomplete. Completer analyses and secondary timepoints are labelled rather than promoted.", "Pass 68 adds one fully scored interpretation record and a contextual medical-language layer, now covering 65 repeated terms. NCI, FDA and CONSORT anchor the definitions; software, physics and business comparisons are teaching aids and do not raise clinical confidence.", "Pass 69 recorded its own checkpoint: 673 connected pages, 839 used source records, 785 directly accessible URLs, 52 access-controlled responses, two network or server failures and no persistent not-found target. Later passes expanded the final release to the counts reported in the coverage page. Link status measures access, not whether a claim is correct.", "Pass 70 adds one fully scored detection claim from three 2026 studies. It separates prospective enrollment from the number of cancer events, AUC from a fixed operating point, and assay performance from clinical utility. The strongest prospective denominator is 6 true positives, 2 false negatives and 40 false positives, not the headline 75% sensitivity alone.", "Pass 71 adds one fully scored outside-field transfer map. It treats control, cell circuits, microfluidics, matrix mechanics and nuclear medicine as tools with measurable PDAC entry conditions, not as inherited evidence of patient benefit.", "Pass 72 adds three scored Adenocyte records: company, investor and trial map. All separate financed product claims, protocol facts, patent publication and patient benefit; none promotes the sponsor's 4-of-90 claim to independent detection evidence.", "Passes 73 to 77 add the randomized hepatobiliary stress test, a bounded nutrition signal, a blinded complementary-care failure and the NRG1-cholangiocarcinoma approval. Existing scored pages absorb the pNET combination, postoperative radiation, human paricalcitol tissue study, family-testing pilot, AI reader result and EU daraxonrasib access boundary. The same rule holds: response, progression, tissue change, reader accuracy, regulator status and survival are separate outputs." ], "sourceIds": [ "fda-2026-bizengri-cholangiocarcinoma", "ema-2026-daraxonrasib-phased-review", "sun-2026-talentop-hcc", "abou-alfa-2026-proof301", "kudo-2026-emerald3", "chan-2026-control-nets", "abrams-2026-rtog0848", "perez-2026-paricalcitol-pdac", "uemura-2026-elemental-diet-pdac", "wode-2026-mistral", "aguirre-2026-ignite-tx", "tanaka-2026-ai-eus-reader", "jameson-2026-ketogenic-diet-pdac", "adenocyte-company-linfu", "igan-2026-adenocyte-financing", "clinicaltrials-nct04793633-linfu", "clinicaltrials-nct06276764-linfu-ipmn", "patent-us20240131366-linfu", "zhang-2017-evolutionary-adaptive-therapy", "roybal-2016-combinatorial-t-cell-circuit", "choi-2024-microfluidic-pdac-organoids", "lesavage-2024-engineered-matrix-chemoresistance", "wang-2026-cfdna-prospective-detection", "oberg-2026-pcdc-biomarker-bakeoff", "wang-2026-itih3-ceacam1-panel", "nci-cancer-terms-dictionary", "fda-2018-oncology-endpoints", "hopewell-2025-consort", "gudmundsdottir-2023-staging-laparoscopy", "rompen-2026-preopanc2-staging", "theijse-2024-nontherapeutic-laparotomy", "oreilly-2026-daraxonrasib", "conroy-2022-prodige24-five-year", "blackford-2024-high-risk-surveillance", "tie-2022-dynamic-colon", "sasieni-2026-nhs-galleri", "ohtsuka-2024-bd-ipmn-surveillance", "singhi-2026-pancreaseq-gc", "henn-2023-cyst-surgery-registry", "clinicaltrials-nct04239573-ea2185", "van-dijk-2025-polyp-study", "gupta-2026-gaia-mil", "yang-2025-gbcseeker", "wong-2025-ambrose-cholecystectomy", "clinicaltrials-nct03762837-gbc-risk-cohort", "partelli-2022-aspen-interim", "maekawa-2026-small-pnet-surgery-morbidity", "klein-haneveld-2021-men1-young-screening", "van-vliembergen-2026-men1-mnet-assay", "villard-2023-psc-surveillance", "charoensuk-2024-one-health-opisthorchis", "clinicaltrials-nct05321992-echec", "yu-2025-btc-ctdna-recurrence", "yoo-2025-extrahepatic-cca-ctdna", "lim-2025-attention-hbv", "trinchet-2011-hcc-surveillance-interval", "hirode-2026-hcc-biomarker-rct", "rhee-2025-miracle-hcc", "singal-2024-tracer-protocol", "clinicaltrials-nct06084234-tracer", "ioannou-2026-premium-protocol", "clinicaltrials-nct05486572-premium", "bretthauer-2022-nordicc-10y", "kaminski-2026-nordicc-13y", "tie-2025-dynamic3", "bando-2026-altair", "mo-2026-find-crc-surveillance", "clinicaltrials-nct05174169-circulate-us", "gismalla-2026-africa-pancreatic-review", "sobnach-2024-ssa-hcc-systematic-review", "sobnach-2025-south-africa-hcc-disparities", "sobnach-2026-global-hcc-disparities", "da-fonseca-2024-latam-cca", "babic-2019-muscle-loss-survival", "chen-2025-tocilizumab-cachexia", "groarke-2024-ponsegromab-cachexia", "clinicaltrials-nct06989437-ponsegromab-pdac", "sandhya-2023-olanzapine-anorexia", "pfizer-2026-ponsegromab-profile", "patent-wo2020039321-anti-gdf15", "patent-wo2025245200-ponsegromab-cachexia", "woo-2016-pert-rct", "saito-2018-pert-rct", "picozzi-2025-pert-advanced-pdac", "dugic-2026-pert-mortality", "clinicaltrials-nct04098237-pancax3", "clinicaltrials-nct05642962-pertzye-dose", "clinicaltrials-nct06099119-pertseverance", "lemanska-2023-opensafely-pert-audit", "dhsc-2024-pert-shortage-alert", "maraveyas-2012-fragem", "pelzer-2015-conko004", "vadhan-raj-2020-cassini-pdac", "frere-2020-pdac-thromboprophylaxis-meta", "heffley-2024-pdac-vte-uptake", "noor-2026-pdac-vte-implementation", "ash-2023-primary-vte-prophylaxis-guide", "overvad-2022-khorana-validation", "clinicaltrials-nct05178628-impact-pro", "clinicaltrials-nct06861088-catiq-p3", "clinicaltrials-nct05245877-preposterous-pancreas", "clinicaltrials-nct07196020-pinpoint-vte", "zwicker-2019-isoquercetin-catiq", "quercis-2026-kinisoquin-profile", "patent-us10391096-kinisoquin", "patent-wo2020112828-isoquercetin-cancer", "van-der-gaag-2010-pbd-rct", "costamagna-2026-sems-early-surgery-rct", "tamura-2021-fcsems-plastic-rct", "seo-2019-covered-uncovered-sems-rct", "lyu-2023-metal-plastic-nat-meta", "vehvilainen-2024-cholangitis-nat-interruption", "chen-2023-element-biliary-rct", "paik-2018-eus-ercp-biliary-rct", "barbosa-2024-eus-ercp-meta", "esge-2018-biliary-stenting-guideline", "asge-2024-therapeutic-eus-guideline", "nhs-england-2026-r367", "asco-2024-germline-panel-guideline", "llach-2024-universal-germline", "chittenden-2021-systematic-germline", "walker-2021-embedded-germline-station", "wang-2022-oncology-germline-cascade", "ramsey-2023-mainstreaming-germline", "rodriguez-2024-generate", "mcdonald-2024-germline-protocol", "sinan-2025-pdac-germline-implementation", "grewal-2026-pancan-germline-survey", "katz-2026-gift-cascade", "chen-2023-fdr-cancer-risk", "npaca-2026-state-of-nation", "nhs-england-2024-hpb-pancreatic-service", "nice-2018-ng85-pancreatic", "de-wilde-2012-centralization", "lemmens-2011-centralization", "latenstein-2021-centralization", "coupland-2016-england-resection-volume", "labori-2016-adjuvant-completion", "henry-2023-complications-adjuvant", "kollbeck-2025-complications-chemo", "van-hilst-2019-leopard2", "korrel-2023-diploma", "bruna-2025-diploma-long-term", "klotz-2024-europa", "liu-2024-robotic-open-pd", "jin-2026-portal", "andel-2025-routine-imaging", "van-goor-2025-nationwide-imaging", "daamen-2024-radar-panc-protocol", "clinicaltrials-nct04875325-radar-panc", "tai-2025-pan-asian-esmo-pancreatic", "labori-2024-norpact1", "bai-2025-cispd1", "tan-2026-neoadjuvant-ipd-meta", "ghaneh-2023-espac5", "clinicaltrials-nct04340141-a021806", "effenberger-2012-pdac-bone-marrow-dtc", "nordgard-2022-unresectable-dtc", "chalabi-hajkarim-2025-organotropism", "pei-2025-spatial-metastatic-pdac", "mcclellan-2026-quiescent-pdac-car-t", "tanaka-2012-microscopic-local-recurrence", "dotan-2024-giant", "dotan-2025-giant-ga", "clinicaltrials-nct04233866-giant", "betge-2022-grantpax", "clinicaltrials-nct02143219-pamela70", "mohile-2021-gap70", "li-2021-gain", "ihorst-2026-geriatric-comanagement", "clinicaltrials-nct06040801-frail-gi", "stoop-2026-preopanc4-surgery", "dekker-2026-taps-restaging", "verkolf-2026-taps-resection-prediction", "seelen-2026-pelican", "kunzmann-2021-neolap", "hartlapp-2022-neolap-ca199", "guggenberger-2023-neolap-ct", "napoli-2025-surgery-versus-oncology", "imamura-2004-locally-invasive-surgery-rct", "clinicaltrials-nct06714604-duration", "clinicaltrials-nct06132087-proactive", "testoni-2021-hybridtherm-rct", "gannon-2026-pancreas-histotripsy-safety", "pusceddu-2026-percutaneous-cryoablation", "zhou-2024-pancreatic-mwa", "ross-2022-panco-p32", "huggett-2014-verteporfin-pdt", "dimatteo-2018-eus-laser-ablation", "clinicaltrials-nct07698613-asp1929-pancreas", "muragaki-2026-k912-sonodynamic", "clinicaltrials-nct05371223-nab-pipac", "padilla-valverde-2024-adjuvant-hipec-rct", "levy-2019-celiac-ganglia-versus-plexus-rct", "xie-2026-cnsi-fe-first-human", "fda-2024-bizengri-multidisciplinary-review", "maltoni-2016-systematic-palliative-qol", "henricks-2018-dpyd-guided-dosing", "philip-2022-kras-wildtype-pdac", "ngohuang-2023-pancfit-rct" ], "links": [ { "target": "claim-pdac-organoid-correlation-not-routing-utility", "relation": "scores" }, { "target": "failure-organoid-treatment-chain-attrition", "relation": "scores" }, { "target": "unknown-pdac-organoid-guided-treatment-utility", "relation": "scores" }, { "target": "trial-pdac-organoid-action-map", "relation": "scores" }, { "target": "hypothesis-organoid-guided-early-switch", "relation": "scores" }, { "target": "claim-staging-laparoscopy-avoids-incision-not-proven-survival", "relation": "scores" }, { "target": "failure-staging-yield-without-patient-utility", "relation": "scores" }, { "target": "unknown-pdac-staging-laparoscopy-rule", "relation": "scores" }, { "target": "trial-pdac-occult-metastasis-action-map", "relation": "scores" }, { "target": "hypothesis-pdac-two-timepoint-staging-route", "relation": "scores" }, { "target": "person-marc-besselink", "relation": "scores" }, { "target": "ranking-provisional-pdac-bottlenecks", "relation": "scores" }, { "target": "recommended-strategy", "relation": "scores" }, { "target": "claim-daraxonrasib-common-ras-survival-breakthrough", "relation": "scores" }, { "target": "claim-adjuvant-mfolfirinox-improves-but-recurrence-remains", "relation": "scores" }, { "target": "claim-high-risk-surveillance", "relation": "scores" }, { "target": "claim-screening-base-rate-maths", "relation": "scores" }, { "target": "claim-colon-ctdna-has-clinical-utility-proof", "relation": "scores" }, { "target": "failure-grail-capital-scaled-before-utility", "relation": "scores" }, { "target": "hypothesis-prospective-ras-resistance-routing", "relation": "scores" }, { "target": "hypothesis-multimodal-mrd-action", "relation": "scores" }, { "target": "claim-pdac-primary-prevention-action-gap", "relation": "scores" }, { "target": "trial-new-onset-diabetes-workup-map", "relation": "scores" }, { "target": "hypothesis-one-time-risk-triggered-workup", "relation": "scores" }, { "target": "unknown-dangerous-precursor-selection", "relation": "scores" }, { "target": "trial-pancreatic-precursor-action-map", "relation": "scores" }, { "target": "hypothesis-dynamic-cyst-interception-rule", "relation": "scores" }, { "target": "claim-gallbladder-polyp-guidelines-still-overtreat", "relation": "scores" }, { "target": "claim-gallbladder-diagnostic-models-existing-lesions-only", "relation": "scores" }, { "target": "claim-gallbladder-prevention-must-beat-surgery-harm", "relation": "scores" }, { "target": "failure-polyp-surveillance-low-yield", "relation": "scores" }, { "target": "failure-nct03762837-risk-cohort-never-started", "relation": "scores" }, { "target": "trial-gallbladder-prevention-detection-action-map", "relation": "scores" }, { "target": "hypothesis-gallbladder-two-stage-interception-rule", "relation": "scores" }, { "target": "unknown-gallbladder-prevention-selector", "relation": "scores" }, { "target": "claim-small-nfpnet-surveillance-surgery-tradeoff", "relation": "scores" }, { "target": "failure-aspen-cannot-estimate-surgery-benefit", "relation": "scores" }, { "target": "claim-men1-surveillance-finds-lesions-not-danger", "relation": "scores" }, { "target": "failure-men1-mnet-assay-no-progression-prediction", "relation": "scores" }, { "target": "unknown-small-pnet-surgery-selector", "relation": "scores" }, { "target": "trial-small-pnet-action-map", "relation": "scores" }, { "target": "hypothesis-pnet-time-updated-surgery-rule", "relation": "scores" }, { "target": "claim-cca-psc-surveillance-workload", "relation": "scores" }, { "target": "claim-fluke-control-prevents-exposure-not-yet-cancer", "relation": "scores" }, { "target": "claim-cca-ctdna-prognostic-not-action-guiding", "relation": "scores" }, { "target": "failure-psc-surveillance-no-curable-detection", "relation": "scores" }, { "target": "failure-echec-cancer-outcome-mismatch", "relation": "scores" }, { "target": "failure-cca-mrd-prediction-without-action", "relation": "scores" }, { "target": "unknown-cca-postoperative-mrd-action", "relation": "scores" }, { "target": "trial-cca-action-map", "relation": "scores" }, { "target": "hypothesis-cca-route-specific-action-platform", "relation": "scores" }, { "target": "claim-hcc-surveillance-test-improvement-not-mortality", "relation": "scores" }, { "target": "failure-hcc-three-month-ultrasound-no-small-cancer-gain", "relation": "scores" }, { "target": "failure-hcc-biomarker-addition-no-early-stage-gain", "relation": "scores" }, { "target": "claim-hcc-early-hbv-treatment-signal-immature", "relation": "scores" }, { "target": "unknown-hcc-surveillance-modality-utility", "relation": "scores" }, { "target": "trial-hcc-action-map", "relation": "scores" }, { "target": "hypothesis-hcc-liver-reserve-co-control", "relation": "scores" }, { "target": "claim-crc-colonoscopy-invitation-benefit-limited-by-participation", "relation": "scores" }, { "target": "failure-dynamic3-ctdna-guided-chemo-no-rfs-gain", "relation": "scores" }, { "target": "failure-altair-mrd-treatment-no-dfs-gain", "relation": "scores" }, { "target": "claim-crc-ctdna-utility-is-action-specific", "relation": "scores" }, { "target": "unknown-crc-ctdna-positive-action", "relation": "scores" }, { "target": "trial-crc-action-map", "relation": "scores" }, { "target": "hypothesis-selected-local-consolidation", "relation": "scores" }, { "target": "company-natera", "relation": "scores" }, { "target": "claim-mrd-platform-access-is-more-than-patent-license", "relation": "scores" }, { "target": "claim-africa-pdac-case-count-not-decision-data", "relation": "scores" }, { "target": "failure-ssa-hcc-curative-route-not-reaching-patients", "relation": "scores" }, { "target": "claim-latam-cca-referral-cohort-shows-route-loss", "relation": "scores" }, { "target": "unknown-regional-route-loss", "relation": "scores" }, { "target": "hypothesis-regional-route-completion", "relation": "scores" }, { "target": "trial-regional-route-completion-action-map", "relation": "scores" }, { "target": "person-sanju-sobnach", "relation": "scores" }, { "target": "lab-latam-cca-registry", "relation": "scores" }, { "target": "claim-cachexia-is-parallel-lethal-system", "relation": "scores" }, { "target": "claim-il6-cachexia-signal", "relation": "scores" }, { "target": "claim-ponsegromab-gdf15-cachexia-driver-not-survival", "relation": "scores" }, { "target": "failure-tocilizumab-cachexia-muscle-without-survival", "relation": "scores" }, { "target": "unknown-cachexia-treatment-utility", "relation": "scores" }, { "target": "trial-pdac-cachexia-action-map", "relation": "scores" }, { "target": "company-pfizer-cachexia", "relation": "scores" }, { "target": "claim-ponsegromab-access-stack", "relation": "scores" }, { "target": "claim-pert-survival-observational-randomized-unproved", "relation": "scores" }, { "target": "failure-pert-short-randomized-trials", "relation": "scores" }, { "target": "failure-pert-prescribing-and-supply-gap", "relation": "scores" }, { "target": "unknown-pert-pdac-utility", "relation": "scores" }, { "target": "trial-pert-pdac-action-map", "relation": "scores" }, { "target": "claim-anticoagulation-prevents-pdac-vte-not-survival", "relation": "scores" }, { "target": "failure-pdac-vte-prophylaxis-delivery-gap", "relation": "scores" }, { "target": "failure-impact-pro-status-unknown", "relation": "scores" }, { "target": "unknown-pdac-vte-prophylaxis-rule", "relation": "scores" }, { "target": "trial-pdac-vte-action-map", "relation": "scores" }, { "target": "claim-kinisoquin-phase2-marker-not-clinical-proof", "relation": "scores" }, { "target": "company-quercis-pharma", "relation": "scores" }, { "target": "claim-kinisoquin-access-stack", "relation": "scores" }, { "target": "hypothesis-pdac-vte-route-completion", "relation": "scores" }, { "target": "person-jeffrey-zwicker", "relation": "scores" }, { "target": "lab-bidmc-thrombosis-pdi", "relation": "scores" }, { "target": "claim-biliary-drainage-is-route-not-cancer-treatment", "relation": "scores" }, { "target": "failure-routine-plastic-pbd", "relation": "scores" }, { "target": "failure-biliary-stent-treatment-interruption", "relation": "scores" }, { "target": "unknown-pdac-biliary-drainage-route", "relation": "scores" }, { "target": "trial-pdac-biliary-drainage-action-map", "relation": "scores" }, { "target": "hypothesis-pdac-biliary-route-completion", "relation": "scores" }, { "target": "company-boston-scientific-biliary", "relation": "scores" }, { "target": "claim-universal-germline-testing-needs-action", "relation": "scores" }, { "target": "failure-pdac-germline-route-dropout", "relation": "scores" }, { "target": "system-england-r367-germline-criteria", "relation": "scores" }, { "target": "unknown-pdac-germline-family-action-route", "relation": "scores" }, { "target": "trial-pdac-germline-cascade-action-map", "relation": "scores" }, { "target": "hypothesis-pdac-germline-family-route-completion", "relation": "scores" }, { "target": "claim-pdac-surgery-is-a-complete-treatment-route", "relation": "scores" }, { "target": "system-england-pancreatic-surgery-network", "relation": "scores" }, { "target": "failure-pdac-postoperative-treatment-loss", "relation": "scores" }, { "target": "failure-leopard2-laparoscopic-whipple-safety", "relation": "scores" }, { "target": "unknown-pdac-surgical-route-completion", "relation": "scores" }, { "target": "trial-pdac-surgical-route-action-map", "relation": "scores" }, { "target": "hypothesis-pdac-surgical-route-completion", "relation": "scores" }, { "target": "claim-routine-imaging-after-pdac-resection-not-yet-proven", "relation": "scores" }, { "target": "unknown-postresection-surveillance-utility", "relation": "scores" }, { "target": "trial-pdac-postresection-surveillance-action-map", "relation": "scores" }, { "target": "person-lois-daamen", "relation": "scores" }, { "target": "lab-dpcg-radar-panc-network", "relation": "scores" }, { "target": "claim-neoadjuvant-therapy-improves-selection-not-cure", "relation": "scores" }, { "target": "failure-norpact1-neoadjuvant-delivery", "relation": "scores" }, { "target": "unknown-pdac-perioperative-treatment-route", "relation": "scores" }, { "target": "trial-pdac-perioperative-sequencing-action-map", "relation": "scores" }, { "target": "person-cristina-ferrone", "relation": "scores" }, { "target": "lab-alliance-a021806-network", "relation": "scores" }, { "target": "person-tingbo-liang", "relation": "scores" }, { "target": "lab-zhejiang-pancreatic-centre", "relation": "scores" }, { "target": "claim-pdac-dormancy-model-proven-human-uncertain", "relation": "scores" }, { "target": "unknown-dormant-cell-actionability", "relation": "scores" }, { "target": "hypothesis-human-dormancy-reawakening-gate", "relation": "scores" }, { "target": "unknown-actionable-interception-window", "relation": "scores" }, { "target": "claim-pdac-progression-timing-uncertain", "relation": "scores" }, { "target": "claim-human-pdac-dtc-prognostic-not-dormancy", "relation": "scores" }, { "target": "failure-pdac-dormancy-sample-gap", "relation": "scores" }, { "target": "trial-pdac-residual-cell-state-action-map", "relation": "scores" }, { "target": "claim-age-alone-cannot-route-pdac-treatment", "relation": "scores" }, { "target": "failure-grantpax-geriatric-route", "relation": "scores" }, { "target": "unknown-older-pdac-treatment-route", "relation": "scores" }, { "target": "trial-older-pdac-action-map", "relation": "scores" }, { "target": "hypothesis-older-pdac-fast-support-route", "relation": "scores" }, { "target": "person-efrat-dotan", "relation": "scores" }, { "target": "lab-ecog-acrin-geriatric-oncology-network", "relation": "scores" }, { "target": "claim-lapc-response-warrants-surgical-reassessment", "relation": "scores" }, { "target": "failure-conversion-surgery-selection-bias", "relation": "scores" }, { "target": "unknown-lapc-surgery-causal-benefit", "relation": "scores" }, { "target": "trial-lapc-conversion-surgery-action-map", "relation": "scores" }, { "target": "hypothesis-randomized-lapc-resection-after-response", "relation": "scores" }, { "target": "trial-pdac-eus-rfa-action-map", "relation": "scores" }, { "target": "trial-pdac-histotripsy-action-map", "relation": "scores" }, { "target": "trial-pdac-cryoablation-action-map", "relation": "scores" }, { "target": "trial-pdac-mwa-action-map", "relation": "scores" }, { "target": "trial-pdac-p32-action-map", "relation": "scores" }, { "target": "trial-pdac-pdt-action-map", "relation": "scores" }, { "target": "trial-pdac-laser-ablation-action-map", "relation": "scores" }, { "target": "trial-pdac-asp1929-action-map", "relation": "scores" }, { "target": "trial-pdac-k912-sonodynamic-action-map", "relation": "scores" }, { "target": "trial-pdac-pipac-action-map", "relation": "scores" }, { "target": "trial-pdac-hipec-action-map", "relation": "scores" }, { "target": "trial-pdac-celiac-pain-action-map", "relation": "scores" }, { "target": "trial-pdac-ferroptosis-action-map", "relation": "scores" }, { "target": "trial-pdac-nrg1-action-map", "relation": "scores" }, { "target": "trial-pdac-palliative-care-action-map", "relation": "scores" }, { "target": "trial-pdac-pharmacogenetic-safety-action-map", "relation": "scores" }, { "target": "trial-pdac-kras-wildtype-action-map", "relation": "scores" }, { "target": "trial-pdac-exercise-action-map", "relation": "scores" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "strong", "randomised": "strong", "independentReplication": "moderate", "sampleSize": "488 fully scored decision records", "effect": "Evidence-accounting method; it makes disagreement and weak support visible but does not improve the underlying studies", "limits": [ "Not yet every care-changing or company-ranking claim", "Ratings require judgement", "Conflict disclosure can be incomplete", "Recency does not measure truth" ] }, "updated": "2026-09-15" }, { "id": "experiments", "title": "Highest-information experiments", "shortTitle": "Experiments", "type": "experiment", "status": "hypothesis", "scope": [ "pdac", "pnet", "cholangiocarcinoma", "gallbladder", "hcc", "colorectal", "pan-cancer" ], "summary": "65 generated hypotheses state predictions, falsifying results, cost ranges, time, safety risks and stop rules.", "content": "An experiment earns priority when either result changes a large decision. 1, prospective RAS resistance routing: paired phase 1/2 plasma has already found mutant KRAS amplification in 36% of progressing patients, crossing the 25% frequency gate. The next experiment must reproduce the class independently, locate it across lesions, detect it early enough to route within 14 days and show that a feasible combination suppresses it at clinical exposure; stop if the class fails to repeat, laboratories disagree above 20% or coexisting escape remains active. 2, multimodal residual-disease action: ctDNA, CA19-9, imaging and immune state should define a recurrence state that supports random treatment assignment; stop for low assay concordance, no safe action or excess serious harm. 3, organoid-guided early switching: a blinded result must arrive before cycle two and predict differential human response; retire it if turnaround or prospective concordance fails. 4, a human liver-niche gate: tissue and blood before metastasis must show a reproducible, druggable niche signal; mouse exosome biology alone does not pass. 5, FAP-targeted radiation: human delivery and a phase 2 dose now exist. Recover lesion-level LuMIERE dose data and prove that absorbed Gy predicts control better than scan brightness before any new trial; ongoing NCT04939610 already owns the uncontrolled PDAC response question. 6, selected local consolidation: a systemic-risk classifier must identify oligometastatic patients whose overall survival, treatment-free time and quality of life improve—not merely local control. 7, low-burden RAS plus immune extinction: common-RAS suppression should create tumour killing and improve antigen-specific function without increasing suppressive myeloid state; stop if only T-cell infiltration changes. 8, one-time risk-triggered work-up: first obtain complete EDI and PANDOME operating data. Then test only a distinct locked second-stage selector before pancreas imaging, using individual or cluster randomization and complete negative follow-up. At 0.62% three-year risk, an END-PAC-like 78%-sensitive, 85%-specific rule models to 30.8 unaffected positives per cancer-bearing positive; correlated test errors, incidental procedures and interval cancers are explicit stop routes. 9, human dormancy-reawakening gate: serial host-injury and tumour measurements must identify liver recurrence before imaging and add information beyond ctDNA; no anti-NET intervention proceeds from mouse evidence alone. 10, HCC liver-reserve co-control: a 24-centre comparison should reduce death or first hepatic decompensation without losing tumour dose intensity or response. A separate near-term mechanism check follows the republished CNIO result: independently reproduce RAS, EGFR/HER2 and STAT3 blockade at exposures attainable in humans; this remains a frontier claim until clinical-quality drugs exist. Cost ranges are planning judgements, not quotes. The £5m programme runs cheap failure tests; the £25m observatory creates the longitudinal human evidence; only reproduced rules enter the £100m randomized platform. The eighteenth generated experiment closes the inherited-risk route: test 3,000 consecutive English PDAC patients, report pathogenic yield and actions inside versus outside R367, and family-randomize a standard letter against consented direct contact plus free home testing. At an 8% carrier yield, two eligible relatives and 70% consent and reach, 3,000 × 0.08 × 2 × 0.70 = 336 relatives; the 31% to 50% comparison needs about 271 after a 1.3 family-clustering allowance. Stop for privacy breach, action on an uncertain variant, slow results, unequal completion or no net benefit outside R367. The nineteenth generated experiment tests surgical-route completion across England's existing 23-centre network. One pilot centre fixes the fields; 22 centres cross to a named-owner, joined surgery-oncology, recovery and proficiency bundle in 11 matched pairs. The national benchmark is postoperative treatment within 14 weeks rising from 62% to 72%; a 1,200-person design assumes intracluster correlation 0.01 and can expand to 1,700 after a masked 300-person check. Full-course completion must rise ten points without higher 90-day mortality or serious complications.", "contentSections": [ "Experiment 46, randomized electrochemotherapy data gate: enroll nobody, buy no equipment and spend at most £250,000 only after independent access and publication rights. Recover at least 80 of 90 planned assignments, 95% vital status and 90% pulse-log and image coverage. Count death as response failure, retain chemotherapy and operation harm, and stop without patient-valued or survival benefit.", "Experiment 45, existing HIFU randomized-data gate: enroll nobody and commit £0 before database lock. Then spend at most £250,000 to recover the 40-person HIFU-PC15 assignment and the 90-person Suizenji trial after its expected 2027 completion. Require at least 95% and 99% assignment accounting, 90% device-log coverage, preserved chemotherapy and pain plus opioid benefit or positive quality-adjusted survival. A new trial receives £0 now.", "Experiment 44, PELICAN specimen salvage: enroll nobody and spend at most £150,000 to reuse the randomized trial's stored tissue and blood only if at least 60 paired samples survive, clinical coverage reaches 80% and two laboratories agree on at least 80% of the main measurements. Test whether ablation caused a repeatable local or immune change; the result may explain failure but cannot reopen radiofrequency ablation as treatment. The treatment route is stopped because survival was unchanged while severe serious events rose by 16 percentage points.", "Experiment 43, vendor-independent SMART recovery: enroll nobody and spend at most £500,000 to reconcile all 136 phase 2 records, recover the dormant LAP-ABLATE package and test archived cases across at least two maintained system families. Go requires at least ten committed centres, 95% plan-test passage, cross-system target and bowel-dose differences within 5%, full harm release and transferable trial rights. The 267-person randomized trial receives £0 now.", "Experiment 42, particle delivered-dose and systemic-risk gate: enroll nobody. Spend at most £500,000 to publish completed PACK and proton records, reconstruct daily delivered rather than planned dose, and apply one cross-centre calculation. Require at least 95% participant, 90% fraction and 80% serial-anatomy recovery, plus a five-point severe-harm or material treatment-delivery gain and a locked group with two-year distant-failure risk below 40%. A future roughly 500-person randomized trial receives £0 now.", "Experiment 41, regional-hyperthermia data recovery: enroll nobody. Spend at most £250,000 to recover HEATPAC and complete HEAT records. Require at least 95% of random assignments and deaths reconciled and temperature traces for at least 80% of delivered heat sessions. Publish the result even if recovery fails. A new trial remains blocked unless the all-assigned record shows a credible heat-specific signal and a current-care comparison changes heat alone.", "Experiment 37, ASPIRE data-rescue gate: enroll nobody. Spend at most £250,000 over three months to locate the assignment key, hash and copy raw records, reconcile at least 400 assignments, obtain all protocol and analysis-plan versions, and show at least 90% 12-month vital status with no more than a ten-point arm difference in missingness. Only a passing audit may price follow-up and analysis rescue inside the fixed £500m ceiling.", "Experiment 21 is a 500-person hidden-spread staging check nested inside the existing 23-centre English surgical route. It compares current selective practice with a locked two-timepoint route and targets non-therapeutic laparotomy falling from 17.1% to 7.0%. It stops for major procedure harm above 2%, material treatment delay, false-positive denial of surgery, cytology-only automatic exclusion or wider inequality. Its £0.4m is carved from the existing £2.5m surgical-route budget.", "10, HCC liver-reserve co-control: release £6m for a three-year randomized readiness stage in 240 patients at 12 centres. It must deliver required liver-care actions in at least 85% while losing no more than five percentage points in on-time cancer treatment or dose intensity. Stop for excess 90-day death or decompensation, no measurable ALBI or action-completion separation, or tumour-treatment loss. Only after every gate passes may a 600-person, 24-centre trial seek the further £12m-£19m needed to test a 25% reduction in death or first decompensation. The stage does not duplicate TRACER or PREMIUM, which test surveillance routes.", "11, dynamic cyst interception: separate current HGD or cancer, future within-cyst progression and PDAC elsewhere into three locked models. Enroll 6,000 consecutive surgery-fit patients across at least 20 systems, validate outside the training systems and then randomize the action. At a 2% five-year event rate this yields about 120 events before loss and competing death. Stop if positive operations contain HGD or invasive disease in fewer than 25%, external calibration shifts by more than five percentage points, interval cancers cross the limit, or procedure harm erases benefit. Estimated cohort cost is £28m–£40m over at least five years; it is a conditional large programme, not a £5m experiment.", "12, country-specific gallbladder interception: use up to £5m to align EULAT, Chile BiLS, P-iGBC and North Indian fields and run silent external tests; release £15m–£20m for a 40,000-person prospective stage only after calibration and outcome-linkage gates pass. At 0.2% target prevalence that gives about 80 events; at 1% it gives about 400. Stop for country calibration error above five points, sub-10-mm sensitivity below 80%, more than ten false-positive invasive work-ups per HGD or stage-I cancer, or fewer than five expected prevented cancer deaths per operative death.", "13, time-updated small-pNET action: first spend £3m to obtain complete ASPEN and PANDORA follow-up, freeze a size-plus-duct baseline and define procedure-specific harm. Release £12m–£18m for 3,000 consecutive 1-2 cm tumours with embedded random allocation only if external calibration error is below five points and outcome linkage exceeds 85%. At a 2% five-year metastatic or invasive-progression rate, 3,000 people yield about 60 events. Stop if post-resection pathology drives the score, the rule misses more than 1.5 additional metastatic progressions per 100, or major procedure harm is not reduced by at least 25%.", "14, route-specific cholangiocarcinoma action: spend £6m over three years to join 6,000 PSC records, 10,000 endemic-area prevention participants and 600 resected biliary cancers under one field set while preserving three separate analyses. At the Swedish target-state rate, 6,000 people with four years average follow-up would yield about 122 dysplasia or cancer events; at 59.4% recurrence, 600 resected cases would yield about 356 recurrences. Stop PSC expansion for more than ten benign invasive procedures per additional resectable target, fluke expansion for less than 50% infection reduction at 24 months, and residual-disease expansion for panel success below 90% or result return before action below 85%. Only one passing route may use up to £6m more.", "15, regional route completion: spend £5m first on one year of consecutive route registration at at least five African and five Latin American sites. Required capture is 80% or more, required-field missingness 20% or less and twelve-month follow-up 80% or more. Release £15m only if one repairable step accounts for at least 15 percentage points of route loss at two sites or 100 patients. Randomize clinics or districts in a stepped sequence to the measured repair and require at least a 10-point absolute increase in completed recommended treatment. Stop below a five-point gain, for serious excess harm, if delay only moves downstream, or if pooled results hide a failed cancer or site. Overall and quality-adjusted survival remain the final outcomes.", "Colorectal action gate: do not add a fifteenth generic ctDNA hypothesis. DYNAMIC-III and ALTAIR show that a strong risk marker can feed a failed treatment rule, while FIND shows earlier curative-intent treatment without mature survival. CIRCULATE-US already tests omission and intensification. The information gap is the effective positive-MRD action and the selector for organ-confined recurrence. Any use of the existing multimodal-MRD or local-consolidation hypotheses must name the action, avoid the live chemotherapy questions and retain overall or quality-adjusted survival as the release endpoint.", "16, PDAC clot-route completion: use £3m from the existing £15m patient-safety, access and health-economics line for a roughly 2,500-person stepped cluster trial across about 30 clinics. A 12% untreated risk and RR .31 gives 3.72% treated risk; moving use from 30% to 70% changes expected risk from 9.52% to 6.20%, a 3.31-point difference. A simple calculation needs about 2,070 people and a provisional 1.2 cluster penalty gives 2,485, rounded to 2,500. Require a documented decision within seven days in at least 80%, use up at least 30 points and symptomatic VTE down at least three points. Stop for major bleeding up two points, fatal-bleeding imbalance, no symptomatic benefit, incidental-only benefit or drug access failure above 10%.", "17, PDAC biliary-route completion: spend £4m from the £12m remaining patient-safety, access and health-economics line on a roughly 1,300-person stepped cluster trial across about 30 centres. Reducing stent-failure treatment interruption from 16% to 10% needs about 983 people before clustering; 983 × 1.3 = 1,278, rounded to 1,300. Require a documented indication and plan within 48 hours in at least 90%, metal-stent use in at least 90% when drainage spans neoadjuvant treatment, and interruption down six points. Stop for serious procedure harm up five points, death imbalance, surgery delayed seven additional days, no interruption gain or no increase in planned treatment delivered.", "19, PDAC surgical-route completion: start from every potentially curable scan, lock the expert-MDT plan, record proficiency and rescue, and follow every planned dose. Do not buy a robot or repeat a broad device race. At 62% versus 72%, individual allocation needs 692 people; centre clustering and 10% incomplete data raise the plan to 1,200, with a 1,700 cap if intracluster correlation is near 0.02. Stop above 5% 90-day mortality, for a greater than one-point mortality rise, five-point serious-complication rise, five-point unexplained operation-rate fall or ten-point inequality widening.", "Human residual-cell gate: the existing dormancy hypothesis is not a twentieth experiment. It is the 250-person nested subset already named inside the £25m observatory. Require at least 25 pre-imaging recurrence cases with the state, ancestry support in 80% of pairs, two-laboratory agreement in 80%, at least eight weeks lead time and added prediction beyond stage plus ctDNA. Stop before treatment if any condition fails or the target is needed for repair or antimicrobial defence.", "Experiment 20 is a 72-hour assessment-to-action route for older or otherwise vulnerable metastatic-PDAC patients. A six-site, 150-person run-in precedes a 900-treated-person randomized comparison. It uses the final £3m in the existing patient-safety line and stops for treatment delay, higher emergency care or death, no patient-valued gain, coercive treatment selection or wider inequality.", "Candidate 22, locally advanced responder surgery: enrol at diagnosis, but randomize 540 people only after at least four months without distant progression, two independent expert panels agree complete visible removal is technically possible and both routes remain acceptable. The calculation targets death HR .75: about 379 deaths, divided by 80% event capture and 90% evaluability, gives 527, rounded to 540. Overall and quality-adjusted survival decide; R0 and resection rate do not. A 120-person feasibility stage stops for allocation failure, crossover above 20%, 90-day surgical death above 5% or more than three points above control, or persistent quality-of-life loss. Planning cost is £22m, currently unfunded; it can enter the £500m ceiling only by replacing at least £22m after an explicit portfolio review.", "Experiment 22, organoid route validation: first recover NCT04931381, NCT04931394 and NCT05351983 and test the assay-by-treatment interaction in randomized PASS-01. Then use the existing £2.5m organoid line for shadow testing in 200 consecutive advanced-PDAC patients, including 50 split across two laboratories. Go only with at least 80% tumour-confirmed results within 14 days, at least 80% agreement on the top-two regimen order, at least 90% drug access and no inequality gap above ten points. If those pass and existing randomized data cannot answer utility, a 480-person £18m trial is an unfunded option requiring equal portfolio displacement.", "Experiment 5, FAP dose-response recovery: obtain every LuMIERE PDAC scan, serial therapy image, lesion dose, normal-organ dose, response and failed screen. Require at least 80% linked lesions, cross-centre calibration error no greater than 10%, and external dose-response AUC at least 0.70 with lower confidence bound above 0.55. Stop if SUV performs equally well, dose has no ordered relation with control or organ limits block useful exposure. This is analysis of an existing programme, not a new treatment cohort. Any later randomized treatment test receives £0 now and requires equal portfolio displacement.", "Experiment 23, tumour-confirmed autophagy blockade: first recover all 91 NCT04892017 records and determine whether paired pancreatic-tumour measurements exist. A replacement drug reaches a 20-person measurement stage only after two laboratories reproduce combination killing at human-achievable exposure. Go requires at least 80% usable paired biopsies, direct target suppression plus blocked-recycling cargo in at least 70% of exposed tumours, at least 80% of planned RAS-drug dose delivered in at least 75% of patients, and four-month control in at least 4/20 response-evaluable patients. Failure at any gate stops expansion. The later randomized add-on test receives £0 now and must displace equal capital.", "Experiment 24, measured fixed-backbone sonoporation: release and independently analyse NCT04821284 before exposing another person. Go requires a planned treatment-by-backbone interaction with uncertainty, complete intended-to-treat flow and missingness unable to explain the result. Only then can a fixed-backbone measurement stage require at least 90% follow-up, at least 90% of planned device sessions in at least 80% of patients, a validated tumour-delivery measure in at least 80% of the nested sample, a 25% or greater median delivery increase, preserved chemotherapy dose and no excess severe harm. The current cost is £0; any later test must displace equal capital.", "Experiment 25, multi-compartment KRAS siRNA delivery: use the live 15-to-21-person SIL204 Segment 1 and add no cohort. Before expansion, require paired primary-tumour results in at least 80%, 50% or greater mutant-RNA reduction with concordant protein reduction in at least 70%, agreement across two regions in at least 70%, active systemic exposure, preserved chemotherapy and severe procedure harm no higher than 5%. Plasma concentration or one injected site cannot pass. Current cost is £0; support stops if the sponsor refuses complete public data.", "Experiment 26, VCN-01 survival gate: spend £0 first to recover all 112 VIRAGE records, the final dated analysis plan and every outcome. Go requires at least 99% randomized accounting, an independently reproduced all-randomized OS HR no higher than 0.75 with its 95% interval excluding 1.0 or a documented regulator-agreed phase 2 rule fixed before outcome access, no missing-outcome scenario moving HR above 0.85, severe-toxicity excess no higher than ten points and chemotherapy dose loss no higher than ten points. Use the sponsor-funded six-person VIRAGE2 only to require at least five of six complete three doses, at least 80% planned chemotherapy and no treatment-related death. A blinded phase 3 remains conditional and must displace equal capital.", "Experiment 27, MTAP-selected PRMT5 plus RAS contribution: first mature all 54 treated PDAC records. Then compare the identical RAS inhibitor alone with RAS inhibitor plus vopimetostat. Go requires at least a 15-point confirmed-response gain, PFS HR no higher than 0.80, preserved response duration and at least 90% RAS dose intensity. Stop for a confidence interval including no PFS effect, severe-harm excess above ten points, RAS dose loss above ten points or benefit outside complete MTAP loss. Mission cost is £0; Tango funds it.", "Experiment 28, elraglusib survival confirmation: use £0 first to recover all 269 weekly-or-control randomized records, all protocol and analysis-plan versions and at least 99% survival accounting. Continue only if ITT HR is at most 0.75, its 95% interval excludes 1.0 and plausible missing outcomes cannot move HR above 0.85. A later blinded placebo-controlled 600-to-650-person phase 3 targets OS HR at most 0.80 and positive quality-adjusted survival. It stops for an interval including 1.0, uncompensated severe-harm excess above ten points, chemotherapy dose loss above ten points or a post-outcome-only subgroup. Mission phase 3 cost is £0 now; a solvent partner must fund manufacturing and most execution.", "Experiment 29, human replication-stress selector: recruit nobody. Recover all 12 NCT06015659 records and require at least ten usable paired tumours, checkpoint plus DNA-damage change in at least eight, six-month PFS in at least 5/12, two-method marker agreement and no treatment-related death. Hair, blood, plasma exposure and organoid sensitivity cannot substitute. Failure retires the PDAC WEE1 branch; success permits design of a same-chemotherapy randomized contribution test. Current mission cost is £0.", "Experiment 30, IBI343 randomized survival test: recruit nobody outside NCT07066098. Require the protocol and analysis plan, one fixed central assay and cut, every randomized person, overall and quality-adjusted survival, treatment exposure and complete toxicity. With 2:1 allocation, two-sided alpha 0.05 and 80% power, an assumed HR 0.70 needs about 278 deaths while HR 0.65 needs about 191; 201 planned people can therefore test only a large effect if nearly all supply an event. Stop the product claim for an OS interval including 1.0, a lower-marker escape, payload harm that erases useful life or unexplained denominator loss. Current mission cost is £0 because the sponsor and partner are funded.", "Experiment 31, KRAS-TCR human persistence gate: recruit nobody outside NCT04146298. Preserve the full mutation-HLA screen, collection, manufacture and infusion route. Among at least 25 infused people, require at least five confirmed responses, four lasting six months, paired tumour measurement in at least 80% and one fixed presentation-entry-persistence measurement that separates durable response from failure. Stop for treatment-related death, normal-tissue attack, collection-to-infusion below 70%, no durable response or a selector invented after outcomes. Current mission cost is £0.", "Experiment 32A, angiogenin selector retest: expose no patient. Apply the published assay and unchanged cut to blinded stored daNIS-2 baseline plasma linked to all 490 randomized outcomes. Require treatment-by-marker P below 0.01, high-group OS HR at most 0.75 with its interval below 1, matching PFS direction, missing samples below 20% and no harmful trade. Cap mission cost at £1m only with complete linkage and publication rights.", "Experiment 32B, mitazalimab survival gate: first release all 94 OPTIMIZE-1 records. If the signal survives, a solvent partner supplies drug and at least £25m before mission exposure can rise from £5m startup to £25m total. Randomize about 450 people to mitazalimab or placebo on identical mFOLFIRINOX; require OS HR at most 0.75, its interval below 1, no greater than ten-point severe-harm rise, no greater than two-point treatment-death rise and non-worse patient function.", "Experiment 33, localized losartan existing-trial gate: recruit nobody. Release all 168 NCT03563248 assignments and estimate losartan and nivolumab contributions separately. Require at least a ten-point complete-route gain or PFS HR at most 0.75, preserved OS, severe-harm rise no greater than five points and usable flow or oxygen measurement in at least 80%. Stop for no controlled effect, R0-only change, missing outcome above 5%, harm or an outcome-created subgroup. Mission cost is £0.", "Experiment 34, bacterial gemcitabine-deactivation gate: use every PANORAMIX Step 2 assignment. Reconcile the 142-versus-206 whole-trial denominator, freeze the tumour bacterial-enzyme assay before unmasking, and join target change to gemcitabine/dFdU exposure, six-month survival, PFS, severe harm and resistant organisms. Stop if any link fails or benefit appears only after subgroup selection. Mission cost is at most £1m for assays and independent analysis and £0 for new participants.", "Experiment 35, personalized RNA vaccine package gate: use every IMCODE003 assignment. Require a disease-free-survival HR at most 0.75, at least ten points of three-year absolute benefit, a 95% confidence interval excluding no effect and no more than five points excess severe treatment harm. Publish screening, tissue, manufacture, delivery and dose denominators. Stop on a missed primary endpoint or a result that exists only among manufactured, treated or immune-responder subsets. Cost is £0 for patients and product and at most £250,000 for independent analysis after data access.", "Experiment 36, quemliclustat phase 3 gate: use every PRISM-1 assignment. Require OS HR at most 0.80, at least two median months, a 95% confidence interval excluding no effect, aligned PFS and bounded serious harm. Publish the protocol, analysis plan, final assigned count, later treatment and every harm. Stop on a missed primary endpoint or a marker cut chosen after results. Cost is £0 before readout and at most £250,000 for independent analysis after data access.", "Experiment 38, TIGeR-PaC final gate: enrol nobody new. Reconcile at least 114 randomized patients and analyse all of them under the fixed final alpha after two interim looks. Pass for adoption only if the trial's locked primary rule succeeds and either survival HR is at most 0.75 with its 95% interval below 1.00, or the upper survival-HR limit is at most 1.20 with at least 15 points less severe harm and better fixed-time quality of life. Count catheter, infection, vessel and organ harm with chemotherapy harm. Cost is £0 before lock and at most £250,000 after participant-level access.", "Experiment 39, DIRECT randomized-data gate: enrol nobody new. Reconcile every NCT03899636 assignment and recover at least 90% 24-month vital status with no more than five points arm difference in missingness. Adoption requires the locked primary test, survival HR at most 0.80 with its 95% interval below 1.00, at least ten points more alive at two years, procedure death at most 3%, no more than ten points excess severe harm and no important quality-of-life loss. Cost is £0 before access and at most £250,000 after independent analysis and publication rights are signed.", "Experiment 40, Optune Pax independent value gate: enrol no new mission-funded trial participant. Reconcile all 571 PANOVA-3 assignments, preserve the all-randomized survival estimate, and extend patient-experience measurement through progression. Require survival HR at most 0.85 with its 95% interval below 1.00, at least five points more alive at one year, no more than ten points excess severe harm and no important quality-of-life loss. Measure every eligible, offered, accepting, starting and stopping patient. Treat hours worn as behaviour after assignment, not a randomized dose. Mission cost is at most £500,000 after data, analysis and publication rights; Novocure funds replication and any expansion.", "The sixty-third hypothesis tests a joined pancreatic chemotherapy-safety loop. It keeps required DPYD testing, separates UGT1A1, captures 2,000 consecutive treatment intentions and spends at most £800,000. It stops for late results, wrong doses, severe toxicity, treatment loss, unequal panel coverage or an upward early-death signal.", "The sixty-fourth hypothesis broadens the existing NRG1 route without adding money. In 5,000 consecutive cases, about 535 should be KRAS wild type at 10.7%; an assumed 20% fusion yield suggests about 107 candidates. The route stops for tissue, RNA, timing, disagreement, access or treatment-delay failure and never pools efficacy across targets.", "The sixty-fifth hypothesis enrolls nobody. It spends at most £250,000 to recover at least 80% of assignments from five exercise trials and tests whether a low-function rule repeats with at least five extra quality-weighted days or ten points more chemotherapy delivery. If it passes, the route enters already funded care work; added trial capital stays £0." ], "sourceIds": [ "gudmundsdottir-2023-staging-laparoscopy", "rompen-2026-preopanc2-staging", "theijse-2024-nontherapeutic-laparotomy", "oreilly-2026-daraxonrasib", "aronchik-2026-daraxonrasib-resistance", "ho-2025-cxcr4-pd1-pdac", "bever-2026-vaccine-checkpoint-pdac", "botta-2024-tumor-informed-ctdna", "beutel-2021-pdac-organoids", "costa-silva-2015-liver-niche", "lee-2024-prospective-fapi-resectable-pdac", "ludmir-2024-extend", "rojass-2025-neoantigen-followup", "chari-2026-prospective-new-onset-diabetes", "sharma-2018-endpac", "wu-2022-nod-imaging-pilot", "frank-2025-pandome", "edi-nct04662879", "ohtsuka-2024-bd-ipmn-surveillance", "mirzaian-2026-low-risk-cysts", "singhi-2026-pancreaseq-gc", "pea-2026-post-surveillance-resection", "henn-2023-cyst-surgery-registry", "clinicaltrials-nct06192719-eulat-gbc", "clinicaltrials-nct06531408-pigbc", "yang-2025-gbcseeker", "gupta-2026-gaia-mil", "wong-2025-ambrose-cholecystectomy", "partelli-2022-aspen-interim", "clinicaltrials-nct03084770-aspen", "heidsma-2021-pandora-pnet", "maekawa-2026-small-pnet-surgery-morbidity", "shi-2026-chemo-dormant-dtc", "seif-el-dahan-2026-early-hcc-causes-death", "storandt-2024-atezo-bev-liver-function", "liaki-2026-triple-combination-republished", "villard-2023-psc-surveillance", "charoensuk-2024-one-health-opisthorchis", "clinicaltrials-nct05321992-echec", "yu-2025-btc-ctdna-recurrence", "yoo-2025-extrahepatic-cca-ctdna", "clinicaltrials-nct06084234-tracer", "clinicaltrials-nct05486572-premium", "tie-2025-dynamic3", "bando-2026-altair", "mo-2026-find-crc-surveillance", "clinicaltrials-nct05174169-circulate-us", "ruers-2017-clocc-crc-liver-ablation", "adam-2024-transmet", "mulcahy-2021-epoch-radioembolization", "gismalla-2026-africa-pancreatic-review", "sobnach-2024-ssa-hcc-systematic-review", "sobnach-2025-south-africa-hcc-disparities", "sobnach-2026-global-hcc-disparities", "da-fonseca-2024-latam-cca", "maraveyas-2012-fragem", "pelzer-2015-conko004", "vadhan-raj-2020-cassini-pdac", "frere-2020-pdac-thromboprophylaxis-meta", "heffley-2024-pdac-vte-uptake", "noor-2026-pdac-vte-implementation", "ash-2023-primary-vte-prophylaxis-guide", "overvad-2022-khorana-validation", "van-der-gaag-2010-pbd-rct", "costamagna-2026-sems-early-surgery-rct", "tamura-2021-fcsems-plastic-rct", "seo-2019-covered-uncovered-sems-rct", "lyu-2023-metal-plastic-nat-meta", "vehvilainen-2024-cholangitis-nat-interruption", "chen-2023-element-biliary-rct", "paik-2018-eus-ercp-biliary-rct", "barbosa-2024-eus-ercp-meta", "esge-2018-biliary-stenting-guideline", "asge-2024-therapeutic-eus-guideline", "nhs-england-2026-r367", "asco-2024-germline-panel-guideline", "llach-2024-universal-germline", "chittenden-2021-systematic-germline", "walker-2021-embedded-germline-station", "wang-2022-oncology-germline-cascade", "ramsey-2023-mainstreaming-germline", "rodriguez-2024-generate", "mcdonald-2024-germline-protocol", "sinan-2025-pdac-germline-implementation", "grewal-2026-pancan-germline-survey", "katz-2026-gift-cascade", "chen-2023-fdr-cancer-risk", "npaca-2026-state-of-nation", "nhs-england-2024-hpb-pancreatic-service", "nice-2018-ng85-pancreatic", "de-wilde-2012-centralization", "lemmens-2011-centralization", "latenstein-2021-centralization", "coupland-2016-england-resection-volume", "labori-2016-adjuvant-completion", "henry-2023-complications-adjuvant", "kollbeck-2025-complications-chemo", "van-hilst-2019-leopard2", "korrel-2023-diploma", "bruna-2025-diploma-long-term", "klotz-2024-europa", "liu-2024-robotic-open-pd", "jin-2026-portal", "effenberger-2012-pdac-bone-marrow-dtc", "nordgard-2022-unresectable-dtc", "chalabi-hajkarim-2025-organotropism", "pei-2025-spatial-metastatic-pdac", "mcclellan-2026-quiescent-pdac-car-t", "tanaka-2012-microscopic-local-recurrence", "dotan-2024-giant", "dotan-2025-giant-ga", "clinicaltrials-nct04233866-giant", "betge-2022-grantpax", "clinicaltrials-nct02143219-pamela70", "mohile-2021-gap70", "li-2021-gain", "ihorst-2026-geriatric-comanagement", "clinicaltrials-nct06040801-frail-gi", "stoop-2026-preopanc4-surgery", "dekker-2026-taps-restaging", "napoli-2025-surgery-versus-oncology", "imamura-2004-locally-invasive-surgery-rct", "seelen-2026-pelican", "clinicaltrials-nct06714604-duration", "baum-2022-fap2286-firsthuman", "mcconathy-2026-lumiere-phase1", "clinicaltrials-nct04939610-lumiere", "baum-2026-3bp3940", "clinicaltrials-nct05432193-frontier", "clinicaltrials-nct07229768-ctr-fapi", "alistar-2017-devimistat-phase1", "philip-2024-avenger500", "surana-2025-erk-hcq-pdac", "clinicaltrials-nct05221320-ulixertinib-hcq", "manji-2023-mekiauto", "clinicaltrials-nct04214418-mekiauto", "clinicaltrials-nct04892017-inlexisertib", "deciphera-2022-dcc3116-phase1", "clinicaltrials-nct04524702-paricalcitol-hcq", "dimcevski-2016-sonoporation-phase1", "wang-2023-sonochemotherapy-pdac", "clinicaltrials-nct04821284-sonoporation", "ius-2026-sonoporation-nct04821284-abstract", "adler-2026-sonoporation-perfusion", "alagesan-2026-sig12d-loder", "ctis-2026-sil204-phase23", "silexion-2026-sil204-site-initiation", "silexion-2026-halfyear-filing", "bazan-peregrino-2021-vcn01-intratumour", "garcia-carbonero-2022-vcn01-intravenous", "clinicaltrials-nct05673811-virage", "theriva-2023-virage-protocol", "theriva-2025-virage-topline", "esmo-2025-virage-2216mo", "ctis-2026-virage2", "theriva-2026-q2-10q", "richards-2006-ci994-pdac", "heumann-2022-azacitidine-pdac", "sohal-2020-thu-decitabine-pdac", "baretti-2024-entinostat-nivolumab-pdac", "clinicaltrials-nct03250273-entinostat-nivolumab", "safyan-2026-azacitidine-pembrolizumab-pdac", "clinicaltrials-nct03264404-azacitidine-pembrolizumab", "rodon-2024-amg193-prmt5", "matsumoto-2026-mtap-pancreatic", "pavlick-2026-mtap-genomics", "clinicaltrials-nct05732831-vopimetostat", "tango-2025-vopimetostat-monotherapy", "clinicaltrials-nct06922591-vopimetostat-ras", "tango-2026-vopimetostat-ras-data", "tango-2026-q2-10q", "mahalingam-2026-elraglusib-randomized", "clinicaltrials-nct03678883-elraglusib", "pathak-2026-elraglusib-folfirinox", "clinicaltrials-nct05077800-elraglusib-folfirinox", "actuate-2026-q2-10q", "sec-2026-actuate-submissions", "chiorean-2023-abemaciclib-pdac", "baghdadi-2019-palbociclib-cdkn2a", "hidalgo-2022-palbociclib-nabpaclitaxel", "laquente-2017-chk1-pdac", "huffman-2023-ly2880070-pdac", "cuneo-2019-adavosertib-pdac", "huffman-2024-azenosertib-pdac-design", "clinicaltrials-nct06015659-azenosertib", "yu-2025-ibi343-pdac", "clinicaltrials-nct05458219-ibi343", "clinicaltrials-nct07066098-ibi343-phase3", "innovent-2026-interim-results", "takeda-2025-ibi343-license", "xu-2026-kras-g12v-tcr", "clinicaltrials-nct04146298-kras-g12v-tcr", "leidner-2022-kras-tcr", "aznar-2025-mesothelin-car-t-resistance", "pietrobono-2026-angiogenin-alk5", "clinicaltrials-nct04935359-danis2", "van-laethem-2025-optimize1-biomarkers", "clinicaltrials-nct04888312-optimize1", "alligator-2026-q2", "clinicaltrials-nct03563248-losartan-localized", "murphy-2019-losartan-lapc", "ramaswamy-2025-afpac-losartan", "geller-2017-bacterial-gemcitabine", "corty-2020-antibiotic-gemcitabine-toxicity", "ctis-panoramix-2024-517766-41", "clinicaltrials-nct07028424-panoramix", "middleton-2014-telovac", "le-2019-eclipse", "nct05968326-imcode003", "wainberg-2026-quemliclustat-arc8", "coveler-2024-oleclumab-pdac", "clinicaltrials-nct06608927-prism1", "arcus-2026-q2-10q", "clinicaltrials-nct05254171-aspire", "ctis-2024-514714-12-00-aspire", "panbela-2024-q3-10q", "panbela-2025-note-default", "singhal-2021-sbp101-phase1", "carla-2024-ivospemin-retinal-atrophy", "clinicaltrials-nct03257033-tigerpac", "hatoum-2024-tamp-rr1-rr2", "novelli-2026-tigerpac-pk", "renovorx-2023-tigerpac-interim", "renovorx-2025-10k", "renovorx-2026-full-enrollment", "timmer-2024-crossfire", "clinicaltrials-nct03899636-direct-rct", "clinicaltrials-nct03899649-direct-registry", "martin-2024-direct-registry-safety", "martin-2026-direct-registry-survival-preprint", "rai-2022-lappie-protocol", "isrctn-14986389-lappie", "nice-2017-ire-pancreatic", "macarulla-2025-panova3", "clinicaltrials-nct03377491-panova3", "fda-2026-optune-pax-ssed", "clinicaltrials-nct06390059-panova4", "novocure-2026-panova4-topline", "novocure-2026-q2-10q", "guzauskas-2026-ttfields-cost", "liu-2026-ttfields-cost", "issels-2023-heat-rct", "clinicaltrials-nct01077427-heat", "eudract-2008-004802-14-heat", "datta-2017-heatpac-protocol", "clinicaltrials-nct02439593-heatpac", "shinoto-2016-carbon-ion-gemcitabine", "kawashiro-2018-jcros-carbon-ion", "lai-2026-carbon-ion-risk-adapted", "rapp-2022-proton-phase2", "eckstein-2023-proton-registry", "clinicaltrials-nct02598349-proton", "clinicaltrials-nct03536182-cipher", "clinicaltrials-nct04194268-pack", "parikh-2023-smart-phase2-safety", "chuong-2024-smart-phase2-survival", "clinicaltrials-nct03621644-smart", "clinicaltrials-nct05585554-lap-ablate", "clinicaltrials-nct05114213-maspac", "viewray-2023-q1-10q", "viewray-2023-chapter11-8k", "clinicaltrials-nct03690323-pelican", "li-2015-hifu-s1", "marinova-2018-hifu-prospective", "marinova-2024-hifu-pc-protocol", "drks-00012367-hifu-pc", "clinicaltrials-nct05262452-hifu-folfirinox", "clinicaltrials-nct05601323-suizenji", "clinicaltrials-nct06211933-puls-hifu", "clinicaltrials-nct07325214-sonopan3", "izzo-2021-pancreatic-ect-phase12", "izzo-2021-laparoscopic-ect-protocol", "eudract-2018-003925-27-ect", "cebron-2023-panect-preliminary", "clinicaltrials-nct04281290-panect-results", "rudno-rudzinska-2021-irec-protocol", "igea-2026-cliniporator-vitae", "glewis-2024-pacific-pgx", "fusco-2021-kras-wildtype-fusions", "neuzillet-2023-apacap-rct" ], "links": [ { "target": "trial-pdac-organoid-action-map", "relation": "constrained-by" }, { "target": "unknown-pdac-organoid-guided-treatment-utility", "relation": "tests" }, { "target": "hypothesis-pdac-two-timepoint-staging-route", "relation": "contains" }, { "target": "programme", "relation": "part-of" }, { "target": "hypothesis-prospective-ras-resistance-routing", "relation": "contains" }, { "target": "claim-daraxonrasib-kras-amplification-resistance", "relation": "uses-human-gate-from" }, { "target": "hypothesis-multimodal-mrd-action", "relation": "contains" }, { "target": "hypothesis-organoid-guided-early-switch", "relation": "contains" }, { "target": "hypothesis-human-liver-niche-gate", "relation": "contains" }, { "target": "hypothesis-fapi-theranostic-pdac", "relation": "contains" }, { "target": "hypothesis-selected-local-consolidation", "relation": "contains" }, { "target": "hypothesis-low-burden-ras-immune-extinction", "relation": "contains" }, { "target": "hypothesis-one-time-risk-triggered-workup", "relation": "contains" }, { "target": "hypothesis-dynamic-cyst-interception-rule", "relation": "contains" }, { "target": "hypothesis-gallbladder-two-stage-interception-rule", "relation": "contains" }, { "target": "hypothesis-pnet-time-updated-surgery-rule", "relation": "contains" }, { "target": "hypothesis-human-dormancy-reawakening-gate", "relation": "contains" }, { "target": "hypothesis-hcc-liver-reserve-co-control", "relation": "contains" }, { "target": "trial-hcc-action-map", "relation": "constrained-by" }, { "target": "hypothesis-cca-route-specific-action-platform", "relation": "contains" }, { "target": "hypothesis-regional-route-completion", "relation": "contains" }, { "target": "trial-regional-route-completion-action-map", "relation": "constrained-by" }, { "target": "trial-crc-action-map", "relation": "constrained-by" }, { "target": "unknown-crc-ctdna-positive-action", "relation": "constrained-by" }, { "target": "unknown-crc-liver-local-cure-selector", "relation": "constrained-by" }, { "target": "research-frontier-ranking", "relation": "ranked-by" }, { "target": "hypothesis-pdac-vte-route-completion", "relation": "contains" }, { "target": "trial-pdac-vte-action-map", "relation": "constrained-by" }, { "target": "hypothesis-pdac-biliary-route-completion", "relation": "contains" }, { "target": "trial-pdac-biliary-drainage-action-map", "relation": "constrained-by" }, { "target": "hypothesis-pdac-germline-family-route-completion", "relation": "contains" }, { "target": "trial-pdac-germline-cascade-action-map", "relation": "constrained-by" }, { "target": "unknown-pdac-germline-family-action-route", "relation": "constrained-by" }, { "target": "claim-pdac-surgery-is-a-complete-treatment-route", "relation": "links-to" }, { "target": "system-england-pancreatic-surgery-network", "relation": "links-to" }, { "target": "failure-pdac-postoperative-treatment-loss", "relation": "links-to" }, { "target": "failure-leopard2-laparoscopic-whipple-safety", "relation": "links-to" }, { "target": "unknown-pdac-surgical-route-completion", "relation": "links-to" }, { "target": "trial-pdac-surgical-route-action-map", "relation": "links-to" }, { "target": "hypothesis-pdac-surgical-route-completion", "relation": "links-to" }, { "target": "trial-pdac-residual-cell-state-action-map", "relation": "bounded-by" }, { "target": "hypothesis-older-pdac-fast-support-route", "relation": "contains" }, { "target": "trial-older-pdac-action-map", "relation": "constrained-by" }, { "target": "hypothesis-randomized-lapc-resection-after-response", "relation": "contains" }, { "target": "trial-lapc-conversion-surgery-action-map", "relation": "constrained-by" }, { "target": "unknown-lapc-surgery-causal-benefit", "relation": "resolves" }, { "target": "claim-fap-radioligand-feasible-pdac-efficacy-unproven", "relation": "bounded-by" }, { "target": "failure-fapi-scan-dose-benefit-collapse", "relation": "learns-from" }, { "target": "unknown-pdac-fap-radioligand-treatment-effect", "relation": "tests" }, { "target": "trial-pdac-fap-radioligand-action-map", "relation": "constrained-by" }, { "target": "hypothesis-tumour-verified-selective-autophagy-blockade", "relation": "contains" }, { "target": "trial-pdac-autophagy-action-map", "relation": "constrained-by" }, { "target": "unknown-selective-autophagy-pdac-contribution", "relation": "tests" }, { "target": "hypothesis-measured-fixed-backbone-sonoporation", "relation": "contains" }, { "target": "trial-pdac-sonoporation-action-map", "relation": "constrained-by" }, { "target": "unknown-pdac-sonoporation-treatment-effect", "relation": "tests" }, { "target": "failure-sonoporation-overall-replication-subgroup-rescue", "relation": "learns-from" }, { "target": "hypothesis-multicompartment-kras-sirna-delivery", "relation": "contains" }, { "target": "trial-sil204-action-map", "relation": "constrained-by" }, { "target": "unknown-sil204-human-target-delivery", "relation": "tests" }, { "target": "claim-sil204-trial-scales-before-human-knockdown", "relation": "bounded-by" }, { "target": "claim-vcn01-randomized-signal-not-survival-proof", "relation": "bounded-by" }, { "target": "failure-virage-analysis-and-survivor-selection", "relation": "learns-from" }, { "target": "unknown-vcn01-all-enrolled-survival", "relation": "leaves-open" }, { "target": "trial-vcn01-action-map", "relation": "implements" }, { "target": "hypothesis-vcn01-replication-to-survival-gate", "relation": "tests" }, { "target": "company-theriva-biologics", "relation": "contains" }, { "target": "failure-broad-epigenetic-treatment-pdac", "relation": "learns-from" }, { "target": "claim-mtap-prmt5-selected-human-signal", "relation": "bounded-by" }, { "target": "failure-vopimetostat-ras-contribution-unknown", "relation": "learns-from" }, { "target": "unknown-mtap-prmt5-ras-contribution", "relation": "leaves-open" }, { "target": "trial-mtap-prmt5-action-map", "relation": "implements" }, { "target": "hypothesis-mtap-prmt5-ras-deeper-control", "relation": "tests" }, { "target": "company-tango-therapeutics", "relation": "contains" }, { "target": "claim-elraglusib-randomized-survival-signal", "relation": "bounded-by" }, { "target": "failure-elraglusib-analysis-and-survival-discordance", "relation": "learns-from" }, { "target": "unknown-elraglusib-survival-replication", "relation": "leaves-open" }, { "target": "trial-elraglusib-action-map", "relation": "implements" }, { "target": "hypothesis-elraglusib-survival-confirmation-gate", "relation": "tests" }, { "target": "company-actuate-therapeutics", "relation": "contains" }, { "target": "failure-broad-cell-cycle-checkpoint-pdac", "relation": "learns-from" }, { "target": "unknown-wee1-replication-stress-pdac", "relation": "leaves-open" }, { "target": "trial-cell-cycle-checkpoint-action-map", "relation": "implements" }, { "target": "hypothesis-human-replication-stress-selector", "relation": "tests" }, { "target": "hypothesis-cldn182-adc-survival-test", "relation": "tests" }, { "target": "hypothesis-kras-tcr-human-persistence-gate", "relation": "tests" }, { "target": "hypothesis-angiogenin-tgfb-existing-data-test", "relation": "tests" }, { "target": "hypothesis-mitazalimab-controlled-survival-gate", "relation": "tests" }, { "target": "hypothesis-localized-losartan-existing-trial-gate", "relation": "tests" }, { "target": "hypothesis-bacterial-gemcitabine-deactivation-gate", "relation": "tests" }, { "target": "unknown-microbiome-treatment-contribution-pdac", "relation": "tests" }, { "target": "trial-pdac-microbiome-action-map", "relation": "mapped-by" }, { "target": "hypothesis-autogene-cevumeran-existing-trial-gate", "relation": "tests" }, { "target": "unknown-pdac-vaccine-regimen-contribution", "relation": "resolves" }, { "target": "trial-pdac-vaccine-action-map", "relation": "mapped-by" }, { "target": "hypothesis-quemliclustat-existing-phase3-gate", "relation": "tests" }, { "target": "unknown-quemliclustat-survival-contribution", "relation": "resolves" }, { "target": "trial-pdac-adenosine-action-map", "relation": "mapped-by" }, { "target": "hypothesis-aspire-data-rescue-gate", "relation": "tests" }, { "target": "unknown-aspire-assigned-denominator-and-results", "relation": "resolves" }, { "target": "trial-pdac-polyamine-action-map", "relation": "mapped-by" }, { "target": "hypothesis-tigerpac-existing-final-gate", "relation": "tests" }, { "target": "unknown-tigerpac-package-survival-and-delivery", "relation": "resolves" }, { "target": "trial-pdac-intraarterial-delivery-action-map", "relation": "mapped-by" }, { "target": "hypothesis-direct-randomized-data-recovery-gate", "relation": "tests" }, { "target": "unknown-ire-added-survival-after-modern-chemotherapy", "relation": "resolves" }, { "target": "trial-pdac-irreversible-electroporation-action-map", "relation": "mapped-by" }, { "target": "hypothesis-optune-pax-independent-value-gate", "relation": "tests" }, { "target": "unknown-ttfields-survival-replication-and-access", "relation": "resolves" }, { "target": "trial-pdac-ttfields-action-map", "relation": "mapped-by" }, { "target": "hypothesis-heat-existing-randomization-recovery-gate", "relation": "tests" }, { "target": "unknown-regional-hyperthermia-added-benefit-modern-care", "relation": "resolves" }, { "target": "trial-pdac-regional-hyperthermia-action-map", "relation": "mapped-by" }, { "target": "hypothesis-particle-therapy-existing-data-gate", "relation": "tests" }, { "target": "unknown-particle-therapy-added-patient-benefit", "relation": "resolves" }, { "target": "trial-pdac-particle-therapy-action-map", "relation": "mapped-by" }, { "target": "hypothesis-smart-vendor-independent-randomized-gate", "relation": "tests" }, { "target": "unknown-smart-added-survival-and-surgery-harm", "relation": "resolves" }, { "target": "trial-pdac-smart-action-map", "relation": "mapped-by" }, { "target": "hypothesis-pelican-specimen-salvage-gate", "relation": "tests" }, { "target": "unknown-rfa-human-mechanism-after-negative-trial", "relation": "resolves" }, { "target": "trial-pdac-rfa-action-map", "relation": "mapped-by" }, { "target": "hypothesis-hifu-existing-randomized-gate", "relation": "tests" }, { "target": "unknown-hifu-pain-survival-and-system-transfer", "relation": "resolves" }, { "target": "trial-pdac-hifu-action-map", "relation": "mapped-by" }, { "target": "hypothesis-electrochemotherapy-randomized-data-gate", "relation": "tests" }, { "target": "unknown-electrochemotherapy-added-benefit-and-delivery", "relation": "resolves" }, { "target": "trial-pdac-electrochemotherapy-action-map", "relation": "mapped-by" }, { "target": "hypothesis-pdac-pharmacogenetic-route-completion", "relation": "contains" }, { "target": "hypothesis-kras-wildtype-driver-route-completion", "relation": "contains" }, { "target": "hypothesis-pdac-exercise-existing-data-gate", "relation": "contains" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "strong", "randomised": "moderate", "independentReplication": "moderate", "sampleSize": "65 fully specified hypothesis records plus one drug-readiness frontier claim", "effect": "Generated experiment portfolio; no efficacy claim", "limits": [ "Predictions, costs and thresholds are programme judgements", "Compound and specimen access are not negotiated", "Some experiments require assay development before randomization", "A negative result may retire only the implementation, not the whole mechanism" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "Each experiment starts from a human observation, but the tested causal rules remain hypotheses." }, "animalEvidence": { "rating": "weak", "reason": "Model support varies and is deliberately insufficient for several human decisions in the portfolio." }, "inVitroEvidence": { "rating": "moderate", "reason": "Organoid and perturbation evidence supports some routes but not the whole ranked portfolio." }, "reproducibility": { "rating": "weak", "reason": "Several inputs repeat across studies, while the proposed decision rules have not been reproduced." }, "commercialConflicts": { "rating": "moderate", "reason": "Some experiments depend on company-controlled compounds or assays; others can run independently." }, "recency": { "rating": "current", "reason": "The ranking includes results, trials and failure signals checked through September 2026." } } }, "layout": { "x": 70, "y": 94 }, "updated": "2026-09-15" }, { "id": "failure-adenosine-control-and-marker-pdac", "title": "Adenosine programmes confused three comparisons: drug effect, checkpoint effect and marker response", "shortTitle": "Adenosine comparison failure", "type": "failure", "status": "supported", "scope": [ "pdac", "transfer-unproven" ], "summary": "ARC-8 randomized the checkpoint component while estimating quemliclustat from other trials; its promising marker was chosen after treatment and did not repeat as a baseline selector.", "content": "Original belief: blocking CD73 would reduce tumour adenosine, restore nearby T cells and make chemotherapy more effective. What ARC-8 actually randomized: quemliclustat plus chemotherapy with versus without zimberelimab. The 29-person arm without zimberelimab had longer median survival, 19.4 versus 14.6 months, and higher confirmed response, 38% versus 25%. This does not prove that checkpoint treatment harms patients because the trial was small, open-label and not designed for that efficacy comparison. It does show that pooling the arms cannot attribute outcome to either immune component. What produced the headline: 122 quemliclustat-treated patients were matched after the trial to 122 people from four historical trials. The OS difference was large, 15.7 versus 9.8 months, but response was 39% versus 41% and PFS was not significantly different. A matched external control is a useful stress test, not randomization. Marker failure: high baseline NR4A predicted better outcome inside ARC-8 but not in the independent PRINCE or MORPHEUS cohorts. Larger NR4A falls after treatment tracked better survival in 37 paired samples, but a post-treatment change cannot select patients before treatment and can reflect exposure, health or tumour response. Class failure: direct randomization of oleclumab showed no overall response, progression or survival benefit, and oleclumab without checkpoint had OS HR 1.26. Recovery rule: keep molecule effect, checkpoint addition, baseline selection and on-treatment mechanism as four separate questions. Do not scale from an external control or a marker that exists after dosing.", "sourceIds": [ "wainberg-2026-quemliclustat-arc8", "coveler-2024-oleclumab-pdac" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "claim-cd73-blockade-signal-without-concurrent-control", "relation": "supports" }, { "target": "failure-immune-recruitment-without-response", "relation": "extends" }, { "target": "unknown-quemliclustat-survival-contribution", "relation": "constrains" }, { "target": "trial-pdac-adenosine-action-map", "relation": "constrains" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "moderate", "randomised": "moderate", "independentReplication": "moderate", "sampleSize": "ARC-8 n=122 at the chosen dose, randomized subset n=90 and paired tissue n=37; oleclumab randomized n=170", "effect": "Large external-control OS signal without response or PFS confirmation; independent CD73 programme missed its primary endpoint", "limits": [ "Quemliclustat and oleclumab differ", "ARC-8 randomization addressed checkpoint addition", "Historical matching may be carefully done but cannot remove unmeasured confounding", "The biomarker analyses were exploratory" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The inference failure occurs despite a measurable human pathway." }, "animalEvidence": { "rating": "strong", "reason": "Model success did not settle component contribution in people." }, "inVitroEvidence": { "rating": "strong", "reason": "NR4A responds to adenosine, but that does not validate its use as a treatment selector." }, "reproducibility": { "rating": "moderate", "reason": "The overall clinical effect failed with oleclumab and the baseline marker did not repeat externally." }, "commercialConflicts": { "rating": "high", "reason": "Both programmes were sponsored and included company authors." }, "recency": { "rating": "current", "reason": "The complete quemliclustat report appeared in March 2026." } } }, "layout": { "x": 444, "y": 378 }, "updated": "2026-09-15" }, { "id": "failure-adjuvant-gemox-biliary", "title": "Adjuvant GEMOX did not reduce recurrence or death after resection of mixed biliary cancers", "shortTitle": "GEMOX adjuvant failure", "type": "failure", "status": "refuted", "scope": [ "cholangiocarcinoma", "gallbladder", "hepatobiliary" ], "summary": "In PRODIGE 12, 196 randomised patients had RFS HR 0.88 and OS HR 1.08, with much more grade 3–4 toxicity on chemotherapy.", "content": "Original belief: the gemcitabine–oxaliplatin activity used in advanced biliary cancer would clear occult disease after surgery. What happened: 196 patients with resected biliary cancers were randomised to twelve cycles of GEMOX or surveillance. Median recurrence-free survival was 30.4 versus 18.5 months, but the hazard ratio was 0.88 with a wide 95% CI of 0.62 to 1.25 and P=0.48. Median overall survival was 75.8 versus 50.8 months, HR 1.08, P=0.74. Maximum grade 3 and grade 4 events were 62% and 11% on GEMOX versus 18% and 3% on surveillance. Why it probably failed: advanced-disease activity did not translate to eradication, site and molecular groups were pooled, and toxicity constrained treatment. The result rejects unselected adjuvant GEMOX. It does not show that all postoperative therapy fails; fluoropyrimidine trials give a different, still qualified signal and site-specific targeted interception remains open.", "sourceIds": [ "edeline-2019-prodige12", "primrose-2019-bilcap" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "claim-bilcap-primary-endpoint-uncertainty", "relation": "contradicts" }, { "target": "model-cholangiocarcinoma-causal-chain", "relation": "supports" }, { "target": "model-gallbladder-causal-chain", "relation": "supports" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong", "independentReplication": "moderate", "sampleSize": "196 randomised patients", "effect": "RFS HR 0.88, P=0.48; OS HR 1.08, P=0.74; grade 3/4 maximum toxicity 62%/11% versus 18%/3%", "limits": [ "Mixed biliary sites", "Open-label", "Underpowered for site subgroups", "Does not test molecular selection", "Regimen-specific failure" ] }, "layout": { "x": 19, "y": 86 }, "updated": "2026-09-14" }, { "id": "failure-algenpantucel-l-vaccine-phase3", "title": "Algenpantucel-L added immune complexity and toxicity without improving PDAC survival", "shortTitle": "Algenpantucel-L failure", "type": "failure", "status": "refuted", "scope": [ "pdac" ], "summary": "In 303 randomised patients, median OS was 14.3 months with vaccine plus standard treatment and 14.9 months with standard treatment alone; HR 1.02.", "content": "Original belief: allogeneic pancreatic-cancer cells engineered to express murine alpha-gal would recruit a strong pre-existing antibody and immune response that could broaden into tumour killing. What happened: 303 borderline-resectable or locally advanced patients were randomised across 32 sites. Both groups received chemotherapy and chemoradiation. Median overall survival was 14.3 months with algenpantucel-L and 14.9 months without it, hazard ratio 1.02 and P=0.98. Median progression-free survival was 12.4 versus 13.4 months. Grade 3 or worse events occurred in 81% versus 75%. Why it probably failed: an inflammatory response to foreign alpha-gal did not reliably create T cells that recognised and entered each patient's tumour; the product was not matched to patient antigens, and it was layered onto intensive treatment. The result rejects this whole-cell construct in this setting. It does not refute vaccines that measure patient-specific or shared-driver T cells, but it raises the evidence bar: immune activation alone is insufficient, and phase 3 survival must decide.", "sourceIds": [ "hewitt-2022-algenpantucel-l" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "claim-neoantigen-vaccine-frontier", "relation": "contradicts" }, { "target": "claim-eli002-mrd-frontier", "relation": "contradicts" }, { "target": "treatment", "relation": "part-of" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong", "independentReplication": "none", "sampleSize": "303 randomised patients at 32 sites", "effect": "Median OS 14.3 versus 14.9 months, HR 1.02, P=0.98; grade 3+ events 81% versus 75%", "limits": [ "Open-label", "Locally advanced and borderline-resectable population", "Whole-cell construct-specific", "Intensive background therapy", "Does not test personalised neoantigen vaccines" ] }, "layout": { "x": 11, "y": 93 }, "updated": "2026-09-14" }, { "id": "failure-altair-mrd-treatment-no-dfs-gain", "title": "ALTAIR found molecular relapse earlier but its chosen drug did not significantly delay recurrence", "shortTitle": "ALTAIR treatment failure", "type": "failure", "status": "refuted", "scope": [ "colorectal" ], "summary": "Trifluridine/tipiracil after ctDNA recurrence gave median disease-free survival 9.30 versus 5.55 months, HR 0.79 with a confidence interval crossing 1, and caused severe blood toxicity in 73.0%.", "content": "Original belief: starting trifluridine/tipiracil at molecular recurrence, before radiological disease, would improve disease-free survival after colorectal-cancer treatment. What happened: ALTAIR randomized 243 ctDNA-positive, scan-negative patients to six months of drug or placebo. Median disease-free survival was 9.30 versus 5.55 months, HR 0.79, 95% CI 0.60 to 1.05, P=0.107; the primary endpoint was not met. Grade 3 or worse hematologic adverse events were 73.0% versus 3.3%, an absolute increase of 69.7 percentage points. Why it probably failed: finding resistant residual disease earlier does not make a later-line cytotoxic drug curative, and finite treatment may only shift the date of visible relapse in a subset. Six authors were Natera employees and many investigators reported assay or drug-company relationships, so the negative result and the toxicity must remain visible beside the promising point estimate. The mechanism that survives is molecular detection. The tested drug action does not. A successor must show that a chosen treatment clears ctDNA durably and improves survival or treatment-free time enough to exceed its harm.", "sourceIds": [ "bando-2026-altair", "nakamura-2024-galaxy-ctdna" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "claim-crc-ctdna-utility-is-action-specific", "relation": "supports" }, { "target": "unknown-crc-ctdna-positive-action", "relation": "leads-to" }, { "target": "trial-crc-action-map", "relation": "informs" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong", "independentReplication": "weak for this exact drug action", "sampleSize": "243 randomized: 122 drug and 121 placebo", "effect": "DFS HR 0.79, 95% CI 0.60-1.05, P=0.107; grade 3 or worse hematologic adverse events 73.0% versus 3.3%", "limits": [ "Stage 0-IV resected population", "ctDNA-positive and scan-negative selected state", "Disease-free rather than overall survival primary endpoint", "Finite six-month drug course", "High commercial involvement", "A confidence interval crossing 1 does not prove zero effect" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "Blood-positive residual disease is real, but it may already contain clones resistant to this cytotoxic drug." }, "animalEvidence": { "rating": "not-applicable", "reason": "Human randomized utility decides whether acting on the result helps." }, "inVitroEvidence": { "rating": "weak", "reason": "Drug activity does not establish benefit in molecular relapse." }, "reproducibility": { "rating": "weak", "reason": "This exact molecular-relapse treatment has one randomized phase 3 result." }, "commercialConflicts": { "rating": "high", "reason": "Six authors were assay-company employees and many investigators disclosed related company ties." }, "recency": { "rating": "current", "reason": "The phase 3 result was published in 2026." } } }, "updated": "2026-09-14" }, { "id": "failure-antibody-target-and-payload-transfer-pdac", "title": "A tumour address does not make every antibody or payload work in PDAC", "shortTitle": "Antibody transfer failures", "type": "failure", "status": "supported", "scope": [ "pdac", "pan-cancer", "transfer-unproven" ], "summary": "Controlled and early human results reject class-wide claims for surface-targeted antibodies, drug carriers and immune engagers.", "content": "System model. A surface marker is an address, not a treatment. The result also depends on the carrier, payload, internalisation, drug release, immune partner, dose, spatial coverage and background treatment. Plain CLDN18.2 antibody. GLEAM screened 2,113 people, found 585 high-marker tumours and randomized 393 in a 2:1 ratio. Zolbetuximab plus gemcitabine/nab-paclitaxel produced OS 13.70 versus 13.57 months, HR 0.999 (95% CI 0.776-1.286), P=.4987. ORR rose by 7.3 points and response duration by two months, yet disease control, PFS and survival did not improve. Serious treatment-emergent events were 73.7% versus 53.9%, a 19.8-point excess, and 37.1% permanently stopped the combination; nausea and vomiting were common stop reasons. Mesothelin drug carrier. Anetumab ravtansine treated 18 selected people. Four never reached a follow-up scan, none of 14 evaluable patients responded, median time to progression was 63.5 days and all 18 died before three years. CEACAM5/6 drug carrier. EBC-129 produced four unconfirmed responses among 21 selected people, PFS 12 weeks and no visible relation between marker score and response. At 2.2 mg/kg, severe neutropenia affected 9/11=81.8%. CLDN18.2 immune-costimulation. SHR-3821 produced zero confirmed responses among 11 pancreatic patients at 15 or 20 mg/kg, despite the abstract calling the combined gastric and pancreatic activity promising. Conclusion. The broad antibody-drug and two-target-antibody class has failed the transfer test. IBI343 and preliminary IBI389 remain product-specific signals; they cannot borrow benefit from gastric cancer, another target or another carrier. Future claims must preserve every treated denominator, lock the exact staining rule, name the component being tested and use a concurrent control with survival and quality-of-life outcomes.", "sourceIds": [ "park-2026-gleam-zolbetuximab", "astellas-2026-gleam-medical-information", "clinicaltrials-nct03816163-gleam", "clinicaltrials-nct03023722-anetumab", "lentz-2025-ebc129-pdac", "gou-2026-shr3821", "abou-alfa-2006-exatecan-pdac", "yu-2025-ibi343-pdac", "hao-2024-ibi389-pdac" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "treatment", "relation": "bounds" }, { "target": "claim-cldn182-adc-selected-response", "relation": "bounds" }, { "target": "claim-cldn18-bispecific-early", "relation": "bounds" }, { "target": "unknown-cldn182-adc-survival-effect", "relation": "motivates" }, { "target": "trial-cldn182-antibody-action-map", "relation": "mapped-by" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong", "independentReplication": "moderate", "sampleSize": "Zolbetuximab randomized n=393; anetumab n=18; EBC-129 n=21; SHR-3821 PDAC at main doses n=11; free exatecan randomized n=349", "effect": "Zolbetuximab OS HR 0.999; anetumab response 0/14; SHR-3821 response 0/11; EBC-129 unconfirmed response 4/21 with 12-week PFS", "limits": [ "Different targets and mechanisms", "Different disease lines", "Several small single-arm studies", "GLEAM was open-label", "Complete GLEAM paper not yet public", "Cross-product comparison is not a head-to-head trial" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The address-carrier-payload decomposition explains why products aimed at the same marker can diverge." }, "animalEvidence": { "rating": "strong", "reason": "Each product had model support, making the failed human transfers more informative." }, "inVitroEvidence": { "rating": "strong", "reason": "Binding and cell killing are measurable but did not reliably predict patient benefit." }, "reproducibility": { "rating": "strong", "reason": "Failure to infer a class effect repeats across a randomized antibody test and several selected carrier or engager products." }, "commercialConflicts": { "rating": "moderate", "reason": "Most results are sponsor linked, but negative and zero-response outcomes are public." }, "recency": { "rating": "current", "reason": "Includes randomized 2026 results and 2025-2026 early product reports." } } }, "layout": { "x": 220, "y": 138 }, "updated": "2026-09-15" }, { "id": "failure-asp1929-selection-light-and-endpoint-design", "title": "The first pancreatic ASP-1929 design cannot separate selection, light delivery and benefit", "shortTitle": "ASP-1929 first-study design failure", "type": "failure", "status": "supported", "scope": [ "pdac", "photoimmunotherapy", "trial-design", "selection" ], "summary": "The planned study treats only people found inoperable during surgery, uses no control and leads with six-month progression, leaving drug uptake, illumination and patient selection entangled.", "content": "The planned flow splits after ASP-1929 infusion. Everyone undergoes surgical exploration. People found resectable have surgery and no light. People found inoperable receive five minutes of 690 nm illumination. This makes the illuminated group conditional on what surgeons discover inside the abdomen, while the non-illuminated group has different anatomy and receives a potentially curative operation. It is not a control. The first ten form a safety run-in; if accepted, total target enrollment is 30 and six-month progression-free survival becomes primary. Without random assignment, that endpoint mixes prior chemotherapy response, anatomy, surgical selection, time off systemic treatment, ASP-1929 uptake, fibre placement and natural disease speed. The eligibility plan requires ASP-1929 four to ten weeks after chemotherapy or radiation and excludes planned non-protocol cancer treatment for at least eight weeks after it, making systemic-treatment delay part of net harm. The registry labels one condition as metastatic but excludes distant metastases, an internal record inconsistency. External red light reaches about 2 cm in mouse work, so deep pancreatic targets need inserted fibres and a spatial light map. EGFR expression cannot close the gap: 90% positivity did not select benefit from ordinary cetuximab in a large trial. Unless all-enrolled flow, tumour binding, light coverage, treatment interruption and surgery outcomes are joined, a six-month rate will be impossible to interpret.", "sourceIds": [ "clinicaltrials-nct07698613-asp1929-pancreas", "maruoka-2018-interstitial-photoimmunotherapy", "philip-2010-swog-s0205-cetuximab", "kim-2015-egfr-expression-pdac" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "claim-asp1929-pancreatic-human-test-not-started", "relation": "limits" }, { "target": "unknown-asp1929-pancreatic-target-coverage-and-benefit", "relation": "creates" }, { "target": "hypothesis-asp1929-first-ten-measurement-gate", "relation": "repaired-by" }, { "target": "trial-pdac-asp1929-action-map", "relation": "mapped-by" }, { "target": "company-rakuten-medical", "relation": "involves" }, { "target": "failure-pdt-geometry-selection-and-bundling", "relation": "supports" } ], "evidence": { "humanProspective": "none in pancreatic cancer", "humanRetrospective": "none in pancreatic cancer", "randomised": "none", "independentReplication": "none", "sampleSize": "Planned n=30 with first-ten safety run-in; zero pancreatic outcomes", "effect": "The registered flow cannot estimate added ASP-1929 benefit", "limits": [ "The protocol may add analysis detail not summarized in the registry", "Safety can still be learned without a control", "Surgical findings are clinically useful", "No participant has yet tested actual flow", "Light may be delivered internally", "Six-month progression may support later design but cannot prove efficacy" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "Target binding and local activation are plausible, but the design does not measure the whole delivery chain." }, "animalEvidence": { "rating": "moderate", "reason": "Mouse light-depth work exposes the geometry problem rather than resolving the human design." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Selection, treatment delay and endpoint attribution occur only in clinical delivery." }, "reproducibility": { "rating": "none", "reason": "The first pancreatic study has not started." }, "commercialConflicts": { "rating": "high", "reason": "The product owner supplies the platform and collaborates on a design without a control." }, "recency": { "rating": "current", "reason": "The criticism applies to the July 2026 registry record." } } }, "layout": { "x": 1012, "y": 936 }, "updated": "2026-09-15" }, { "id": "failure-aspen-cannot-estimate-surgery-benefit", "title": "ASPEN observes practice but cannot estimate the benefit of surgery", "shortTitle": "ASPEN comparison failure", "type": "failure", "status": "supported", "scope": [ "pnet" ], "summary": "The largest prospective small-pNET cohort lets centre and patient preference choose treatment, follows patients briefly and still lacks the planned final public result.", "content": "Original belief: a 1,000-person prospective comparison of surveillance and surgery would define optimal management for asymptomatic sporadic non-functioning PanNEN no larger than 2 cm. Observed result: the interim ASPEN report included 500 people, with 406 observed and 94 operated, but treatment was selected rather than randomized. Forty-two operations followed patient preference and 37 followed centre preference. Enrolment at a surgical centre independently doubled the odds of surgery, while younger age, size above 10 mm and duct dilation also drove assignment. After only 25 months median follow-up, no observed patient developed distant metastasis, nine crossed to surgery, 13% of operated patients had severe complications and 20% had at least one aggressive pathology feature. Those facts cannot estimate the counterfactual: what would have happened to the same person under the other choice. The registry targeted 1,000, estimated completion in December 2024, was last verified in January 2024, now shows unknown status and has no posted results. The cause of the missing final public result is not stated. This is a design and reporting failure, not evidence against observation or surgery. Recovery: obtain the complete final cohort, publish all country and centre assignment factors, separate synchronous spread from later progression, retain every crossover, and embed random allocation for the 1-2 cm group in which both choices are acceptable. Stop claiming comparative safety from an observational hazard ratio until follow-up and measured confounding make the two routes exchangeable enough for the stated endpoint.", "sourceIds": [ "partelli-2022-aspen-interim", "clinicaltrials-nct03084770-aspen", "heidsma-2021-pandora-pnet" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "claim-small-nfpnet-surveillance-surgery-tradeoff", "relation": "failed-because-of" }, { "target": "trial-small-pnet-action-map", "relation": "constrains" }, { "target": "hypothesis-pnet-time-updated-surgery-rule", "relation": "succeeds-where" } ], "evidence": { "humanProspective": "strong — prospective cohort and current registry", "humanRetrospective": "not-applicable", "randomised": "none", "independentReplication": "weak — PANDORA repeats short observation safety but not the comparative design", "sampleSize": "ASPEN interim n=500 against planned n=1,000; PANDORA n=76", "effect": "Treatment choice was strongly centre- and preference-dependent; the public registry has unknown status and no final result", "limits": [ "Final unpublished data may exist", "Registry lag does not prove the study stopped", "Confounding direction is complex", "Short follow-up does not measure lifetime safety" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The decision concerns a real trade-off between slow malignant progression and immediate pancreatic-surgery harm." }, "animalEvidence": { "rating": "not-applicable", "reason": "Animal studies cannot repair treatment-selection bias in a human cohort." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Laboratory assays do not resolve the missing treatment comparison." }, "reproducibility": { "rating": "moderate", "reason": "Short-term observation safety repeats, while comparative benefit remains unreproduced." }, "commercialConflicts": { "rating": "low", "reason": "The study is academic, but surgeon and centre preferences are direct non-financial sources of bias." }, "recency": { "rating": "current", "reason": "Registry status and passed completion dates were checked in September 2026." } } }, "updated": "2026-09-14" }, { "id": "failure-aspire-trial-operational-collapse", "title": "ASPIRE's public experiment state diverged after its operating chain broke", "shortTitle": "ASPIRE operational failure", "type": "failure", "status": "observed", "scope": [ "pdac", "trial-delivery", "transfer-unproven" ], "summary": "A global survival trial passed three safety reviews, then lost its contractor for non-payment and now has conflicting U.S. and European status records with no result.", "content": "Original plan: randomize 600 untreated metastatic-PDAC patients, keep patients and trial staff masked, and compare ivospemin plus gemcitabine and nab-paclitaxel with placebo plus the same chemotherapy. Delivery record: the sponsor reported more than 50% enrollment, 89 open sites and a 395-person safety database by mid-2024. Its data monitoring board recommended continuation three times. Financial break: at 30 September 2024 Panbela had $142,000 cash, a $15.0m working-capital deficit and $12.5m nine-month operating cash use. Simple cash-only coverage was $0.142m / ($12.5m / 9) = 0.102 months, about 3.1 days. That ratio omits financing and payment timing, but it shows the scale mismatch. The trial contractor began terminating the contract on 19 August because invoices were unpaid; Panbela took direct control and sought a replacement. By January 2025 the company disclosed default on $12.0m of notes. State conflict: ClinicalTrials.gov still says recruiting, estimates 600 and was last updated in November 2024; the European record says ended on 28 March 2025 and lists 547 planned participants. The projected early-2025 interim analysis and August-2026 main completion passed without a posted result in either linked registry. A later SEC order says a securities registration was abandoned after Panbela did not answer notice. None of these facts proves that ivospemin failed. They prove that the experiment's public control plane is unreliable. Recovery must start with legal custody, actual assignment count, data completeness and patient follow-up, not a new efficacy story.", "sourceIds": [ "clinicaltrials-nct05254171-aspire", "ctis-2024-514714-12-00-aspire", "panbela-2024-q3-10q", "panbela-2025-note-default", "sec-2025-panbela-registration-abandoned" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "company-panbela-therapeutics", "relation": "caused-by" }, { "target": "unknown-aspire-assigned-denominator-and-results", "relation": "creates" }, { "target": "hypothesis-aspire-data-rescue-gate", "relation": "recovered-by" }, { "target": "trial-pdac-polyamine-action-map", "relation": "changes" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong", "independentReplication": "moderate", "sampleSize": "Planned n=600 in the U.S. registry versus 547 in Europe; sponsor-reported safety database n=395; actual randomized count unknown", "effect": "The pivotal experiment may contain valuable survival data, but its operating and public-reporting chain failed before transparent completion", "limits": [ "No public assignment log", "Ended can mean national trial closure rather than completed analysis", "Current legal data controller is unverified", "Sponsor statements do not establish data completeness" ], "dimensions": { "biologicalPlausibility": { "rating": "not-applicable", "reason": "This record tests trial delivery and data custody." }, "animalEvidence": { "rating": "not-applicable", "reason": "Animal evidence cannot repair a broken human experiment." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Cell evidence cannot determine whether trial data remain usable." }, "reproducibility": { "rating": "moderate", "reason": "SEC, U.S. registry and European registry records independently expose the break and status conflict." }, "commercialConflicts": { "rating": "high", "reason": "The distressed sponsor controlled trial operations and the expected interim analysis." }, "recency": { "rating": "current", "reason": "The European end date, note default and SEC order extend into 2025; the U.S. record remained stale in 2026." } } }, "layout": { "x": 490, "y": 414 }, "updated": "2026-09-15" }, { "id": "failure-avatar-precision-medicine-delivery", "title": "A randomized precision-medicine system failed because results rarely reached patients in time", "shortTitle": "AVATAR delivery failure", "type": "failure", "status": "supported", "scope": [ "pdac", "transfer-unproven" ], "summary": "Only 4 of 81 patients assigned to the sequencing-and-model arm received matched treatment; overall survival was unchanged.", "content": "AVATAR tested a full precision-oncology chain rather than one assay: biopsy, whole-exome sequencing, patient-derived organoids, mouse models, a molecular board and matched second-line treatment. Of 81 patients assigned to that system, sequencing succeeded in 65, potentially actionable changes appeared in 14, experimental models were generated for 16, and only four received the recommended personalized treatment. Median overall survival was 8.6 months in the system arm versus 8.7 months under physician-choice care, P=0.849. The four treated patients lived a median 19.3 months, but that selected subset cannot establish benefit because patients had to remain well enough and possess a match. The failure identifies a delivery constraint: 4/81 = 4.9% of assigned patients received the intended output. A useful future system must return a validated treatment decision before clinical deterioration, cover common PDAC biology rather than rare actionable changes, and prove benefit by assigned group.", "sourceIds": [ "sarno-2025-avatar" ], "links": [ { "target": "hypothesis-organoid-guided-early-switch", "relation": "contradicts" }, { "target": "technology-ai-organoid-active-learning", "relation": "constrains" }, { "target": "treatment", "relation": "part-of" }, { "target": "failures", "relation": "part-of" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong", "independentReplication": "none", "sampleSize": "125 randomized; 81 assigned to precision medicine; 4 received personalized treatment", "effect": "Median OS 8.6 versus 8.7 months; P=0.849", "limits": [ "Complex multi-component intervention", "Few patients received its output", "Subset survival is severely selected", "Turnaround and access may improve with newer technology" ] }, "layout": { "x": 24, "y": 119 }, "updated": "2026-09-14" }, { "id": "failure-axitinib-vascular-targeting", "title": "VEGF-receptor inhibition failed a blinded 632-patient pancreatic phase 3 trial", "shortTitle": "Axitinib failure", "type": "failure", "status": "supported", "scope": [ "pdac" ], "summary": "Axitinib plus gemcitabine produced median survival of 8.5 months versus 8.3 with placebo plus gemcitabine and crossed the futility boundary.", "content": "A randomized phase 2 signal suggested that adding the VEGF-receptor inhibitor axitinib to gemcitabine might improve survival. The double-blind phase 3 randomized 632 patients and stopped for futility. Median overall survival was 8.5 versus 8.3 months, HR 1.014, 95% CI 0.786–1.309, one-sided P=0.5436. Hypertension and several other adverse events were more common. The result invalidates this anti-VEGF drug-plus-gemcitabine implementation and weakens simple vessel-starvation logic in unselected PDAC. It does not prove that vascular state is irrelevant: drug delivery, immune trafficking and vessel normalization could require different dose, timing or selection. Any revival must show a measured human vascular change tied to better delivery, not reuse angiogenesis as a generic rationale.", "sourceIds": [ "kindler-2011-axitinib-pdac" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "treatment-long-tail-audit", "relation": "supports" }, { "target": "claim-pdac-metabolic-adaptation-is-redundant", "relation": "adjacent-to" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong", "independentReplication": "moderate", "sampleSize": "632 randomized", "effect": "Median OS 8.5 versus 8.3 months; HR 1.014, 95% CI 0.786–1.309", "limits": [ "Old gemcitabine backbone", "Axitinib-specific", "Unselected population", "Does not test vessel normalization or delivery-guided dosing" ] }, "layout": { "x": 30, "y": 124 }, "updated": "2026-09-14" }, { "id": "failure-biliary-stent-treatment-interruption", "title": "Cholangitis and stent failure interrupt pancreatic treatment before the tumour question is tested", "shortTitle": "Biliary route interruption", "type": "failure", "status": "supported", "scope": [ "pdac" ], "summary": "One surgical cohort recorded cholangitis in 20% and stent-failure treatment interruption in 16%; the survival association is serious but not causal proof.", "content": "Original operating belief: inserting a stent completes biliary management and permits uninterrupted neoadjuvant treatment. What happened: in a retrospective cohort of 162 pancreatic-head cancers that all received drainage and neoadjuvant treatment before surgery, 33, 20%, developed cholangitis. Their median OS was 26 rather than 36 months and PFS 8 rather than 17 months. Twenty-six, 16%, had a treatment interruption caused by stent failure; median OS was 26 rather than 35 months and PFS 7 rather than 17. These are selected associations, not proof that the interruption caused the survival difference: only eventual surgical patients entered, and tumour burden, infection, tolerance and treatment choice can differ. The route failure itself is direct. A mixed-design meta-analysis found metal rather than plastic stents associated with less reintervention, OR .04, treatment delay, OR .17, recurrent obstruction, OR .09, and cholangitis, OR .16. Covered and uncovered metal stents then produced similar sustained drainage, 72.2% versus 72.9%, but failed differently: covered devices migrated and uncovered devices had tumour ingrowth. What fails: treating insertion as completion or choosing by purchase price alone. What survives: a dated route that owns patency, symptoms, laboratory recovery, infection, chemotherapy dates, reintervention and final surgery.", "sourceIds": [ "vehvilainen-2024-cholangitis-nat-interruption", "lyu-2023-metal-plastic-nat-meta", "tamura-2021-fcsems-plastic-rct", "seo-2019-covered-uncovered-sems-rct" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "claim-biliary-drainage-is-route-not-cancer-treatment", "relation": "blocks-delivery-of" }, { "target": "unknown-pdac-biliary-drainage-route", "relation": "leads-to" }, { "target": "hypothesis-pdac-biliary-route-completion", "relation": "motivates" }, { "target": "risks", "relation": "supports" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "strong", "randomised": "moderate for device function; none for the survival association", "independentReplication": "moderate", "sampleSize": "Retrospective surgical n=162; meta-analytic outcomes up to n=463; randomized metal-design n=119 and n=22", "effect": "Cholangitis 20% and stent-failure interruption 16%; metal stents reduce route failures but have no proved survival effect", "limits": [ "Surgical survivor selection", "Observational survival comparison", "Small trials", "Different devices", "Different treatment regimens", "Outcome definitions vary" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Infection and obstruction can directly force admission, treatment delay and dose loss." }, "animalEvidence": { "rating": "not-applicable", "reason": "The failure concerns human devices and care delivery." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Bench patency cannot establish treatment completion or infection." }, "reproducibility": { "rating": "moderate", "reason": "Several studies repeat lower route failure with metal stents, while survival causality is unproved." }, "commercialConflicts": { "rating": "moderate", "reason": "Device studies include manufacturer links, while the survival cohort was academic." }, "recency": { "rating": "current", "reason": "The route and meta-analytic evidence reported in 2023-2024." } } }, "updated": "2026-09-14" }, { "id": "failure-broad-cell-cycle-checkpoint-pdac", "title": "Common cell-cycle faults do not make broad checkpoint drugs effective in PDAC", "shortTitle": "Cell-cycle drug failures", "type": "failure", "status": "supported", "scope": [ "pdac", "cholangiocarcinoma" ], "summary": "CDK4/6 and CHK1 inhibitors repeatedly failed human PDAC efficacy gates despite common pathway changes, selected patients, target exposure and strong models.", "content": "System model. A cell cycle is a state machine for copying DNA and dividing. CDK4 and CDK6 help move the cell through the first gate; the CDKN2A gene normally supplies a brake. CHK1 and WEE1 are damage checkpoints: they pause a stressed cell so DNA can be repaired before division. A common broken brake does not prove the tumour still depends on one remaining switch. CDK4/6 evidence. TAPUR selected CDKN2A loss or mutation, yet none of ten evaluable pancreatic and ten biliary patients responded or stayed stable for 16 weeks on palbociclib; pancreatic PFS was 7.2 weeks and OS 12.4 weeks. Palbociclib plus nab-paclitaxel enrolled 76 people and used a 65% one-year-survival gate at the maximum tolerated dose. The observed estimate was 50% in 27 people, 95% CI 29.9-67.2, so it failed its own gate. The decisive randomized study assigned 99 people equally to abemaciclib, abemaciclib plus a PI3K/mTOR inhibitor, or chemotherapy. Disease control was 15.2%, 12.1% and 36.4%; median PFS 1.7, 1.8 and 3.3 months. The OS HRs were 1.60 and 1.53 versus chemotherapy, both in the wrong direction. No experimental arm advanced. CHK1 evidence. LY2603618 reached the exposure predicted for target effect, but randomized OS was 7.8 versus 8.3 months in 65 versus 34 people; posterior probability of superiority was 0.3 against a required 0.8, and PFS and response did not improve. A newer CHK1 inhibitor plus low-dose gemcitabine produced no response in 11 PDAC patients; every person stopped by 3.2 months, although organoids made from two participant biopsies were strongly sensitive. Decision. Put broad CDK4/6 and CHK1 treatment in Tier D. Do not use CDKN2A loss, model synergy, surrogate-tissue target effect or participant-derived organoid killing as sufficient selection. A future cell-cycle route must show a different, directly measured human dependency and beat the same treatment without the added drug.", "sourceIds": [ "chiorean-2023-abemaciclib-pdac", "baghdadi-2019-palbociclib-cdkn2a", "hidalgo-2022-palbociclib-nabpaclitaxel", "laquente-2017-chk1-pdac", "huffman-2023-ly2880070-pdac" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "treatment", "relation": "bounds" }, { "target": "unknown-wee1-replication-stress-pdac", "relation": "differs-from" }, { "target": "trial-cell-cycle-checkpoint-action-map", "relation": "mapped-by" }, { "target": "hypothesis-human-replication-stress-selector", "relation": "constrains" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong", "independentReplication": "strong", "sampleSize": "Abemaciclib randomized n=99; CHK1 randomized n=99; palbociclib phase 1b n=76; TAPUR pancreatic n=12 and biliary n=10; newer CHK1 PDAC n=11", "effect": "Abemaciclib arms had PFS 1.7-1.8 versus 3.3 months and OS HR 1.53-1.60; randomized CHK1 OS 7.8 versus 8.3 months", "limits": [ "Several studies used older chemotherapy", "Abemaciclib trial was not biomarker selected", "Small selected and combination cohorts", "WEE1 is a related but distinct checkpoint", "Drug schedules differ" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "Cell-cycle faults are common, but repeated human failures show that pathway presence is not drug dependence." }, "animalEvidence": { "rating": "strong", "reason": "Multiple models showed activity and synergy, which makes the human failure informative." }, "inVitroEvidence": { "rating": "strong", "reason": "Target and DNA-damage effects were measurable, including strong participant-organoid sensitivity." }, "reproducibility": { "rating": "strong", "reason": "Failure repeats across two checkpoint families, several agents, selected and unselected cohorts, and randomized comparisons." }, "commercialConflicts": { "rating": "moderate", "reason": "Industry sponsored several studies, but negative results were published." }, "recency": { "rating": "current", "reason": "The main trials are older, but the 2023 randomized and model-to-patient results confirm the current boundary." } } }, "layout": { "x": 210, "y": 128 }, "updated": "2026-09-15" }, { "id": "failure-broad-epigenetic-treatment-pdac", "title": "Broad epigenetic treatment changes cell state without changing PDAC outcomes", "shortTitle": "Broad epigenetic failure", "type": "failure", "status": "refuted", "scope": [ "pdac" ], "summary": "Two randomized studies found no survival gain, while three small single-arm studies produced little tumour control despite measurable biological changes.", "content": "What epigenetic means here. Cells use chemical marks on DNA and its packing proteins as configuration flags: the code is unchanged, but which instructions can run changes. Broad epigenetic drugs try to alter many flags at once. The engineering risk is a high-dimensional configuration write with no address for the cancer dependency. Randomized tests. Adding the histone deacetylase inhibitor CI-994 to gemcitabine in 174 advanced pancreatic cancer patients reduced rather than increased median survival, 194 versus 214 days, P=.908, worsened two-month quality of life and increased low blood counts. Oral azacitidine maintenance after surgery and chemotherapy produced PFS 9.2 versus 8.9 months, HR 0.94, and OS HR 0.98 in 49 randomized patients. Neither study supports benefit. Single-arm tests. THU was meant to stop rapid decitabine breakdown, but five of 13 refractory patients progressed before week eight, seven of eight scanned had progression and median survival was 3.1 months. Entinostat plus nivolumab produced three responses among 27, PFS 1.89 months, OS 2.729 months and grade 3 or worse related harm in 19; measured immune activation did not make the main endpoint pass. Azacitidine plus pembrolizumab produced three responses among 31 efficacy patients, PFS 1.51 months and OS 4.83 months. Why the class label fails. The drugs, stages and combinations differ, so one meta-effect would be false precision. The shared failure is uncontrolled breadth: no locked tumour dependency, incomplete target measurement and biological readouts used as substitutes for useful life. What survives. A narrow genetic deletion can create a specific drug dependency. MTAP-loss PRMT5 treatment is assessed separately because it is a selected failure mode, not another attempt to reset the whole configuration.", "sourceIds": [ "richards-2006-ci994-pdac", "heumann-2022-azacitidine-pdac", "sohal-2020-thu-decitabine-pdac", "baretti-2024-entinostat-nivolumab-pdac", "clinicaltrials-nct03250273-entinostat-nivolumab", "safyan-2026-azacitidine-pembrolizumab-pdac", "clinicaltrials-nct03264404-azacitidine-pembrolizumab" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "treatment", "relation": "part-of" }, { "target": "failure-epigenetic-azacitidine-maintenance", "relation": "includes" }, { "target": "claim-mtap-prmt5-selected-human-signal", "relation": "differs-from" }, { "target": "failure-immune-recruitment-without-response", "relation": "repeats" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong", "independentReplication": "moderate", "sampleSize": "Randomized n=174 and n=49; single-arm n=13, n=27 and n=36", "effect": "No randomized survival gain; single-arm response rates 11% and 9.7% in the checkpoint combinations", "limits": [ "Different drugs and disease stages", "Old chemotherapy comparator in CI-994 study", "Small single-arm cohorts", "Does not refute biomarker-selected chromatin dependencies" ], "dimensions": { "biologicalPlausibility": { "rating": "weak", "reason": "Changing many control marks does not identify which state change a human tumour requires." }, "animalEvidence": { "rating": "moderate", "reason": "Models motivated several trials but did not predict useful human outcomes." }, "inVitroEvidence": { "rating": "moderate", "reason": "Cell-state and immune effects are measurable in laboratory systems." }, "reproducibility": { "rating": "moderate", "reason": "Negative or limited clinical results recur across five products or combinations." }, "commercialConflicts": { "rating": "moderate", "reason": "The evidence includes academic and commercial studies with primary reports and registries." }, "recency": { "rating": "current", "reason": "Includes the 2026 azacitidine-pembrolizumab report." } } }, "layout": { "x": 182, "y": 103 }, "updated": "2026-09-15" }, { "id": "failure-bulk-stroma-removal", "title": "Removing PDAC stroma wholesale can release aggressive disease in mouse models", "shortTitle": "Stroma-removal warning", "type": "failure", "status": "supported", "scope": [ "pdac" ], "summary": "Two independent 2014 mouse studies found that deleting stromal programmes or myofibroblasts produced more aggressive tumours and shorter survival.", "content": "The simple belief was that dense PDAC stroma is uniformly tumour-supporting and that removing it should improve treatment. Two studies challenged that model. Genetic Sonic-hedgehog deletion reduced stroma but produced more vascular, proliferative and aggressive mouse tumours; Smoothened inhibition reproduced the phenotype. In a separate transgenic model, depleting alpha-SMA-positive myofibroblasts increased invasive, poorly differentiated disease, immune suppression and shortened survival. Human tumour correlations in the latter study pointed in the same direction but do not prove causality. The lesson is not that stroma should never be targeted. It is that stroma contains different cell states and functions; bulk removal can destroy restraint as well as barriers. A valid intervention needs a named harmful cell state or signal, a way to spare protective functions, and human target-engagement evidence.", "sourceIds": [ "rhim-2014-stroma-restrains", "ozdemir-2014-fibroblast-depletion" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "treatment", "relation": "supports" }, { "target": "pdac", "relation": "part-of" } ], "evidence": { "humanProspective": "none", "humanRetrospective": "weak", "randomised": "none", "independentReplication": "moderate", "sampleSize": "Two independent genetically engineered mouse studies; human correlative cohort in one study", "effect": "Reduced stroma accompanied more aggressive disease and shorter mouse survival", "limits": [ "Intervention evidence is from mouse models", "Different stromal perturbations are not interchangeable", "Human correlations do not establish causation", "Does not test selective stromal reprogramming" ] }, "layout": { "x": 8, "y": 70 }, "updated": "2026-09-14" }, { "id": "failure-cca-mrd-prediction-without-action", "title": "Biliary residual-disease studies repeatedly stop at prediction", "shortTitle": "CCA MRD action failure", "type": "failure", "status": "supported", "scope": [ "cholangiocarcinoma", "gallbladder", "hepatobiliary" ], "summary": "Strong recurrence separation and months of lead time have not been connected to a randomized treatment or surveillance action.", "content": "Original belief: earlier molecular recurrence detection would itself improve postoperative outcome. Observed result: a 56-person real-world cohort found 15 of 16 recurrences and an average 3.7-month lead, while an 89-person STAMP biomarker cohort found strong disease-free-survival separation from serial ctDNA. Neither study assigned treatment from the result. A separate prospective registry, NCT05743959, is observational, planned 100 participants, was last updated in May 2023, now has unknown status, retains passed December 2024 and December 2025 completion estimates and has no posted results. A prospective 18-person feasibility study lost four cases before its main analysis and produced only 12 postoperative comparisons. Likely cause: rare mixed biliary cancers make accrual slow, assay construction can fail, and there is no settled low-toxicity treatment with which to act during molecular-only disease. Mechanism status: ctDNA as a recurrence marker survives; ctDNA-guided care is unproved. Recovery: pool by anatomical site and adjuvant regimen, require result return within 21 days, and randomize a named escalation among ctDNA-positive patients. Keep intensified imaging and systemic treatment as separate interventions. Stop if panel success is below 90%, fewer than 85% receive results before the treatment decision, or no tolerable intervention has a plausible absolute benefit larger than its serious harm.", "sourceIds": [ "yu-2025-btc-ctdna-recurrence", "yoo-2025-extrahepatic-cca-ctdna", "park-2026-btc-ctdna-feasibility", "clinicaltrials-nct05743959-btc-mrd" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "claim-cca-ctdna-prognostic-not-action-guiding", "relation": "failed-because-of" }, { "target": "unknown-cca-postoperative-mrd-action", "relation": "constrains" }, { "target": "hypothesis-cca-route-specific-action-platform", "relation": "succeeds-where" }, { "target": "failure-avatar-precision-medicine-delivery", "relation": "resembles" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "moderate", "randomised": "none for an MRD-directed action", "independentReplication": "moderate for prognosis; none for action", "sampleSize": "Three reported cohorts with 56, 89 and 18 enrolled plus an observational registry planned for 100", "effect": "Repeated recurrence-risk separation but zero randomized ctDNA-directed treatment comparisons", "limits": [ "No failed action trial has occurred", "The record describes a design gap rather than a disproved biomarker", "Biliary sites are mixed", "Registry lag does not prove abandonment", "Treatment options are changing" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Residual tumour DNA should mark remaining malignant material, but a marker cannot specify an effective treatment." }, "animalEvidence": { "rating": "not-applicable", "reason": "The missing evidence is a human treatment decision." }, "inVitroEvidence": { "rating": "moderate", "reason": "Assay construction works in many samples but failed in three of seventeen operable cases in the small feasibility study." }, "reproducibility": { "rating": "moderate", "reason": "Prognostic separation repeats; clinical utility has not been tested." }, "commercialConflicts": { "rating": "high", "reason": "Commercial assay ownership and selected real-world ordering can favour positive prognostic reports." }, "recency": { "rating": "current", "reason": "Published and registry evidence was checked through September 2026." } } }, "updated": "2026-09-14" }, { "id": "failure-ccr2-pulmonary-toxicity", "title": "CCR2 monocyte targeting changed blood cells but added lung toxicity without an efficacy signal", "shortTitle": "CCR2 toxicity", "type": "failure", "status": "supported", "scope": [ "pdac" ], "summary": "PF-04136309 plus chemotherapy produced 24% pulmonary toxicity and a 23.8% response rate in 21 patients, no better signal than chemotherapy alone.", "content": "Researchers expected CCL2–CCR2 blockade to reduce immunosuppressive inflammatory monocytes and improve chemotherapy. In a 21-patient first-line phase 1b with gemcitabine and nab-paclitaxel, blood CCR2-positive monocytes fell, but 24% developed pulmonary toxicity and the response rate was 23.8%, 95% CI 8.2–47.2, without a signal above the chemotherapy backbone. The result weakens this molecule and combination, not the entire myeloid mechanism: pharmacology reached blood cells but may not have created the right tumour change, and overlapping lung effects limited exposure. A successor requires tumour myeloid target engagement, lung-specific preclinical safety and a randomized contribution design; another blood-monocyte shift is insufficient.", "sourceIds": [ "noel-2020-ccr2-pdac" ], "links": [ { "target": "claim-immune-exclusion-is-multicellular", "relation": "constrains" }, { "target": "failures", "relation": "part-of" }, { "target": "treatment-long-tail-audit", "relation": "supports" } ], "evidence": { "humanProspective": "weak", "humanRetrospective": "not-applicable", "randomised": "none", "independentReplication": "none", "sampleSize": "21", "effect": "ORR 23.8%; pulmonary toxicity 24%", "limits": [ "Phase 1b", "No control", "Small", "Combination toxicity", "Blood pharmacology may not reflect tumour macrophages" ] }, "layout": { "x": 27, "y": 121 }, "updated": "2026-09-14" }, { "id": "failure-cell-therapy-route-and-denominator-pdac", "title": "PDAC cell-therapy reports lose patients before infusion and confuse cell detection with control", "shortTitle": "Cell-therapy route failure", "type": "failure", "status": "supported", "scope": [ "pdac", "pan-cancer" ], "summary": "One terminated mesothelin registry counts 54 enrolled people while its main paper reports nine infused; local delivery found cells in tissue but produced no objective response.", "content": "System failure. A living-cell treatment route begins with marker and HLA testing, then collection, manufacture, release, conditioning, infusion, tumour entry, expansion, persistence, killing and recovery. Papers commonly begin at infusion and erase earlier loss. Registry-to-paper gap. NCT03323944 now reports actual enrollment 54 and termination for administrative reasons. Its main paper describes nine infused people and the registry posts no results, leaving 45 participants unreconciled. Among the nine treated patients, one had stable disease at day 28 and progressed by month three; the other eight had progression at their first stated scan. Median PFS was 3.9 weeks and OS 6.7 weeks. Local injection into the abdomen or liver artery did not solve the system: cells were sometimes detected in tumour or fluid, but expansion was low, surface receptor was lost and functional exhaustion appeared. Other route loss. A multiantigen T-cell programme collected cells from 56 people and infused 37, so 19/56=33.9% did not reach treatment. Satri-cel's 24-person pooled efficacy starts at infusion and its adjuvant report starts 6-10 months after surgery, after early recurrences have already selected themselves out. Conclusion. Cell detection is a delivery measurement, not tumour control, and an infused response rate is not a patient-route success rate. Every programme must publish the screen-to-outcome flow and state where each patient was lost.", "sourceIds": [ "clinicaltrials-nct03323944-meso-cart", "aznar-2025-mesothelin-car-t-resistance", "musher-2026-multiantigen-t-cells", "qi-2024-ct041-pdac", "carsgen-2025-satricel-pdac-adjuvant" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "failure-mesothelin-car-t-pdac", "relation": "extends" }, { "target": "claim-multiantigen-t-cells-feasible-not-effective-yet", "relation": "bounds" }, { "target": "claim-satricel-pdac-early-signal", "relation": "bounds" }, { "target": "trial-pdac-engineered-cell-action-map", "relation": "changes" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "not-applicable", "randomised": "none", "independentReplication": "moderate", "sampleSize": "Mesothelin registry n=54 enrolled versus n=9 reported infused; multiantigen route n=56 collected and n=37 infused; satri-cel n=24 infused", "effect": "Mesothelin CAR-T zero objective responses, median PFS 3.9 weeks and OS 6.7 weeks; 45 registry participants unreconciled", "limits": [ "Registry enrollment may include people never infused", "Reasons for the 45-person gap are not public", "Different products and targets", "Early-phase cohorts", "No randomized efficacy estimate" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Each lost route step and the distinction between cell presence and function are directly measurable." }, "animalEvidence": { "rating": "moderate", "reason": "Mouse engineering can alter persistence but cannot repair missing human denominators." }, "inVitroEvidence": { "rating": "moderate", "reason": "Product potency is measurable but does not prove collection-to-benefit completion." }, "reproducibility": { "rating": "moderate", "reason": "Attrition, weak persistence and uncontrolled infused-only reporting recur across distinct cell products." }, "commercialConflicts": { "rating": "moderate", "reason": "Several records are product-owner controlled; the largest denominator gap is academic." }, "recency": { "rating": "current", "reason": "The terminated registry was updated in January 2026 and is compared with the 2025 paper." } } }, "layout": { "x": 238, "y": 158 }, "updated": "2026-09-15" }, { "id": "failure-checkpoint-blockade-unselected-pdac", "title": "Checkpoint blockade alone has almost no activity in unselected advanced PDAC", "shortTitle": "Checkpoint failure", "type": "failure", "status": "supported", "scope": [ "pdac" ], "summary": "Ipilimumab produced no standard responses in 27 patients; durvalumab with or without tremelimumab produced one response among 64 treated patients and failed its expansion rule.", "content": "The original transfer logic came from melanoma and other immune-responsive cancers: release CTLA-4 or PD-L1 inhibition and existing T cells may control PDAC. A 27-patient ipilimumab phase 2 study found no RECIST responses, although one person had a delayed regression after initial progression. A later randomised phase 2 lead-in found objective response of 3.1% for durvalumab plus tremelimumab and 0% for durvalumab alone; the planned expansion did not open because neither arm reached the 10% response threshold. The narrow conclusion is decisive for unselected advanced PDAC: checkpoint release alone is insufficient. It does not refute checkpoint treatment in rare mismatch-repair-deficient tumours or as one component after antigen, trafficking and myeloid barriers are changed.", "sourceIds": [ "royal-2010-ipilimumab", "oreilly-2019-durvalumab-tremelimumab", "feig-2013-cxcl12-immune-exclusion", "bayne-2012-gmcsf-myeloid" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "claim-immune-exclusion-is-multicellular", "relation": "supports" }, { "target": "treatment", "relation": "part-of" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "not-applicable", "randomised": "moderate", "independentReplication": "strong", "sampleSize": "27 ipilimumab patients; 65 randomised in durvalumab lead-in and 64 treated", "effect": "Standard response 0/27 with ipilimumab; 1/32 combination and 0/32 monotherapy with durvalumab study", "limits": [ "Small early trials", "Previously treated advanced disease", "Response endpoint rather than overall survival", "Does not apply to biomarker-selected mismatch-repair-deficient disease", "Does not test multi-layer immune combinations" ] }, "layout": { "x": 8, "y": 90 }, "updated": "2026-09-14" }, { "id": "failure-cnsi-fe-resensitization-and-mechanism-claim", "title": "The first CNSI-Fe study did not measure ferroptosis or reversal of drug resistance", "shortTitle": "CNSI-Fe mechanism claim outruns data", "type": "failure", "status": "supported", "scope": [ "pdac", "pan-cancer", "ferroptosis", "nanoparticle-delivery", "trial-design" ], "summary": "Local scan necrosis and later chemotherapy response were renamed as ferroptosis and resensitization without a biomarker, control or pancreatic cohort.", "content": "NCT06048367 was a 19-person, five-dose safety study across many cancers. Fifteen completed, seven received a second dose and two had pancreatic cancer. The later paper selected four long survivors and called later treatment responses resensitization or hypersensitization. For hypersensitization, the operational rule itself required more than 24 months survival after CNSI-Fe plus later systemic therapy. That definition selects the outcome it then presents as evidence. The pancreatic example had 12% growth of the injected lesion, classed stable disease, then responded to liposomal irinotecan and S-1. Without a same-treatment control, response after a drug break can reflect ordinary sensitivity, clone change, timing or chance. No patient-level GPX4, glutathione, lipid-peroxide or cell-type measurement established ferroptosis. Scan necrosis is non-specific. Among 14 with paired growth-rate data, mean change was -5.85 percentage points per month with 95% CI from -26.62 to 14.92 and p=0.554. Fourteen of 19 had grade 3-plus events and eight had serious events, although none was reported as a treatment-related death. The company-led 54-person expansion remains one arm, many cancers and safety-primary. It may establish delivery and harm. It cannot validate resistance reversal without a defined subsequent-treatment comparison.", "sourceIds": [ "xie-2026-cnsi-fe-first-human", "clinicaltrials-nct06048367-cnsi-fe", "clinicaltrials-nct07433283-cnsi-fe-expansion", "enray-2026-company" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "claim-ferroptosis-pdac-selectivity-not-class-effect", "relation": "limits" }, { "target": "unknown-ferroptosis-cell-type-coverage-and-human-benefit", "relation": "creates" }, { "target": "hypothesis-cnsi-fe-current-study-mechanism-gate", "relation": "repaired-by" }, { "target": "trial-pdac-ferroptosis-action-map", "relation": "mapped-by" }, { "target": "company-sichuan-enray-pharmaceutical", "relation": "involves" }, { "target": "failure-local-control-without-survival-lap07", "relation": "supports" } ], "evidence": { "humanProspective": "weak", "humanRetrospective": "weak — post-trial selected case analysis", "randomised": "none", "independentReplication": "none", "sampleSize": "n=19 mixed cancers; n=2 pancreatic; n=1 pancreatic long-survivor case", "effect": "Local necrosis and one later pancreatic response without measured mechanism or counterfactual", "limits": [ "Dose-escalation purpose", "Many cancers", "Two pancreatic cases", "No control", "Post-treatment therapy confounding", "No ferroptosis biomarker", "Two lost to survival follow-up", "Company-linked analysis" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "Injected ferrous iron can drive oxidation, but the human causal chain was not measured." }, "animalEvidence": { "rating": "moderate", "reason": "Company and academic models support possible local iron effects while exposing context risk." }, "inVitroEvidence": { "rating": "moderate", "reason": "Ferroptosis can be measured in cells, but those measurements were not joined to the treated people." }, "reproducibility": { "rating": "none", "reason": "No independent human cohort or controlled resensitization test exists." }, "commercialConflicts": { "rating": "high", "reason": "The product sponsor, authors, patent owner and expansion sponsor are closely linked." }, "recency": { "rating": "current", "reason": "The human analysis and expansion record appeared in 2026." } } }, "layout": { "x": 1174, "y": 1098 }, "updated": "2026-09-15" }, { "id": "failure-conversion-surgery-selection-bias", "title": "Conversion-surgery survival is repeatedly confused with the filter that selected the patient", "shortTitle": "Surgery selection trap", "type": "failure", "status": "supported", "scope": [ "pdac" ], "summary": "Large survival gaps after selected surgery remain non-causal because the same response and fitness fields choose surgery and predict survival.", "content": "Original belief: people who survive much longer after conversion surgery prove that removing the tumour caused the gain. What the data measure: treatment first removes people with early spread or decline, then expert teams choose younger, fitter responders with more favourable anatomy and CA19-9 change. In TAPS, median survival was 36.1 months after resection and 20.2 months without exploration. Its newer 2,338-person restaging model found 34.3 versus 11.2 months, but the authors explicitly limited that comparison to description because selection and immortal-time bias could explain some or all of it. Immortal-time bias means a person must remain alive and fit long enough to reach surgery, while early deaths remain in the non-surgery path. A 2025 international comparison retained a 42.6 versus 18.6-month gap after matching 75 pairs, yet hospitals and surgeons chose the route; unrecorded judgement and hidden disease remain. The rare direct randomized signal is a 42-person Japanese trial from 2004, after open exploration, before modern scans and combination treatment; it cannot settle the current state. PELICAN is the local-treatment warning: in 188 non-progressing patients, adding invasive radiofrequency ablation changed median survival by only 0.5 months, HR 1.07, while serious adverse events rose from 11% to 27% and quality of life worsened. Resection is not ablation, but the result proves that a selected-responder cohort plus an invasive local action is not enough. The repair is randomization at the point where both surgery and continued treatment remain acceptable, with survival counted from allocation and every crossover retained.", "contentSections": [ "PELICAN's exact median survival was 12.1 months with ablation and 11.6 months without it; serious adverse events were 27% and 11%." ], "sourceIds": [ "theijse-2024-nontherapeutic-laparotomy", "dekker-2026-taps-restaging", "napoli-2025-surgery-versus-oncology", "imamura-2004-locally-invasive-surgery-rct", "seelen-2026-pelican" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "unknown-lapc-surgery-causal-benefit", "relation": "creates" }, { "target": "claim-lapc-response-warrants-surgical-reassessment", "relation": "limits" }, { "target": "failure-local-intensification-pdac-survival-not-shown", "relation": "extends" }, { "target": "hypothesis-randomized-lapc-resection-after-response", "relation": "resolved-by" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "strong", "randomised": "strong for the ablation warning; weak and outdated for resection", "independentReplication": "strong for the observational survival gap; absent for a modern causal surgery effect", "sampleSize": "TAPS n=663 and n=2,338; matched cohort n=312 with 75 pairs; old surgery trial n=42; PELICAN n=188", "effect": "Large observational surgery gaps coexist with a randomized ablation HR 1.07 and excess serious harm of 16 percentage points", "limits": [ "Ablation is not tumour resection", "The old randomized trial used obsolete treatment", "Matching cannot remove hidden selection", "Expert-centre cohorts may not transport", "Quality of life was missing from major surgery cohorts" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Response, occult spread and physical reserve affect both surgical selection and survival." }, "animalEvidence": { "rating": "not-applicable", "reason": "The failure is in human treatment comparison." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Laboratory models cannot remove clinical selection bias." }, "reproducibility": { "rating": "strong", "reason": "Large survival gaps repeat across selected cohorts, and the same studies document why they are not causal." }, "commercialConflicts": { "rating": "moderate", "reason": "Operator and centre volume incentives can favour wider surgical indications even without a proprietary product." }, "recency": { "rating": "current", "reason": "Includes 2025 and 2026 comparisons and a 2026 randomized local-treatment result." } } }, "layout": { "x": 106, "y": 164 }, "updated": "2026-09-15" }, { "id": "failure-crc-radioembolization-pfs-without-os", "title": "Yttrium-90 radioembolization improved liver control but not survival in colorectal liver metastases", "shortTitle": "EPOCH no survival", "type": "failure", "status": "supported", "scope": [ "colorectal", "transfer-unproven" ], "summary": "EPOCH improved PFS by 0.8 months and liver PFS by 1.9 months but left median survival at 14.0 versus 14.4 months and increased severe events.", "content": "Original belief: adding targeted yttrium-90 radioembolization to second-line chemotherapy would control liver-dominant colorectal metastases strongly enough to extend life. What happened: EPOCH randomized 428 patients across 95 centres. Progression-free survival improved from 7.2 to 8.0 months, HR 0.69, and liver PFS from 7.2 to 9.1 months, HR 0.59. Response rose from 21.1% to 34.0%. Median overall survival was 14.0 with radioembolization and 14.4 months with chemotherapy alone, HR 1.07. Grade 3 adverse events rose from 49.3% to 68.4%. Why it probably failed: delaying one organ's visible progression was insufficient once systemic resistant disease and treatment burden determined death. This does not invalidate all local treatment: CLOCC and TransMet show survival tails in much narrower organ-confined groups. It rejects broad second-line addition based on liver involvement alone and strengthens the need for a systemic-risk selector before expensive or toxic local therapy.", "sourceIds": [ "mulcahy-2021-epoch-radioembolization", "ruers-2017-clocc-crc-liver-ablation", "adam-2024-transmet" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "colorectal-liver-metastasis", "relation": "part-of" }, { "target": "claim-crc-liver-ablation-selected-survival", "relation": "contradicts-broad-use" }, { "target": "claim-transmet-selected-liver-only-crc", "relation": "contradicts-broad-use" }, { "target": "hypothesis-selected-local-consolidation", "relation": "supports-selection-for" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong", "independentReplication": "strong", "sampleSize": "428 randomized at 95 centres", "effect": "PFS HR 0.69 and liver-PFS HR 0.59; OS HR 1.07; grade 3 events 68.4% versus 49.3%", "limits": [ "Second-line population", "Not restricted to truly organ-confined disease", "Open-label intervention", "Device-company ties", "Improved PFS may still matter for selected symptoms despite no OS gain" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Liver radiation can delay liver progression, but systemic resistant disease can still determine death." }, "animalEvidence": { "rating": "not-applicable", "reason": "The failed survival claim is established in a large randomized human study." }, "inVitroEvidence": { "rating": "weak", "reason": "Radiation response in cells cannot establish whole-body survival benefit." }, "reproducibility": { "rating": "strong", "reason": "The trial was multicentre and its failure boundary is consistent with the contrast between broad and narrowly selected local treatment." }, "commercialConflicts": { "rating": "high", "reason": "Device-company ties are material to a costly local procedure." }, "recency": { "rating": "historical", "reason": "The phase III report was published in 2021." } } }, "layout": { "x": 13, "y": 106 }, "updated": "2026-09-14" }, { "id": "failure-cryoablation-selection-attrition-and-route-harm", "title": "Cryoablation studies turn selected survivors and mixed procedures into broad benefit claims", "shortTitle": "Cryo evidence-system failure", "type": "failure", "status": "supported", "scope": [ "pdac", "pain", "trial-design", "local-treatment" ], "summary": "Percutaneous studies select safer anatomy and report surviving patients; open studies add major leakage risk; neither design isolates pain or survival benefit.", "content": "The 11-person percutaneous series excluded direct invasion of bowel or other hollow organs and chose people by fitness, symptom burden, access path and multidisciplinary judgment. Its later pain means used 10 survivors at six months and eight at twelve months. Removing people after death makes the remaining average look better even if treatment has no effect. Opioid use was not converted to a common daily morphine-equivalent dose, and quality of life was reconstructed from charts. The approach deliberately froze inside the tumour rather than adding a wide margin, trading completeness for safety. The open-surgery route has a different harm system. Eighteen of 24 people, 75%, leaked pancreatic digestive fluid; four had grade B fistulas requiring a change in care. Delayed bleeding, stomach-emptying delay and fluid collections also occurred. Calling that minor because no one died hides material morbidity. The 24 comparison patients were selected separately, systemic treatment varied and only age, sex, function and stage were reported as matched. The older 118-person study already found no significant prognosis difference. The new 20-person cryo-plus-immunotherapy study bundles two unproved changes in one arm. Across the field, pain, tumour size, immune markers and survival are repeatedly joined without a design that can assign cause.", "sourceIds": [ "pusceddu-2026-percutaneous-cryoablation", "kang-2025-laparotomic-cryoablation", "song-2014-cryo-bypass-comparison", "clinicaltrials-nct07306286-cryo-immunotherapy" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "claim-cryoablation-pain-signal-not-survival-proof", "relation": "limits" }, { "target": "unknown-cryoablation-randomized-pain-and-net-benefit", "relation": "creates" }, { "target": "hypothesis-cryoablation-pragmatic-pain-trial", "relation": "repaired-by" }, { "target": "trial-pdac-cryoablation-action-map", "relation": "mapped-by" }, { "target": "failure-hifu-randomization-and-trial-drift", "relation": "compares-with" }, { "target": "failure-rfa-selection-and-route-attrition", "relation": "supports" } ], "evidence": { "humanProspective": "weak", "humanRetrospective": "moderate", "randomised": "none", "independentReplication": "weak", "sampleSize": "Recent retrospective n=11 and n=24 plus 24 controls; older comparison total n=118; new combination target n=20", "effect": "Large uncontrolled pain change, 75% fluid leakage after open treatment, no causal survival estimate", "limits": [ "Some pancreatic leaks were biochemical only", "Percutaneous and open routes should not be pooled", "No standardized opioid dose", "No validated quality-of-life measure", "Matching details are sparse", "Small samples", "The combination study has no result" ], "dimensions": { "biologicalPlausibility": { "rating": "not-applicable", "reason": "This record concerns selection, missing outcomes and route-specific harm." }, "animalEvidence": { "rating": "not-applicable", "reason": "Animal freezing cannot repair biased pain or survival comparisons." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Cell death does not establish net patient benefit." }, "reproducibility": { "rating": "weak", "reason": "No controlled result survives differences in route, treatment and eligibility." }, "commercialConflicts": { "rating": "moderate", "reason": "Device dependence exists, though the recent papers report no author conflict." }, "recency": { "rating": "current", "reason": "The failure analysis includes two 2025-2026 reports and a recruiting trial." } } }, "layout": { "x": 850, "y": 774 }, "updated": "2026-09-15" }, { "id": "failure-direct-selection-and-randomized-reporting-gap", "title": "DIRECT registry survival is selected while the randomized answer is missing", "shortTitle": "DIRECT evidence split", "type": "failure", "status": "supported", "scope": [ "pdac", "local-control", "trial-reporting" ], "summary": "The public DIRECT story mixes a selected registry with a separate phase 3 trial, making an association look like the missing treatment effect.", "content": "DIRECT is two studies, not one body of interchangeable evidence. NCT03899649 is an observational registry. Its 2026 survival preprint contains 99 IRE patients and 38 controls, of whom only ten were enrolled prospectively and 28 came from contemporaneous past records. Patients reached IRE enrollment a median eight months after diagnosis versus four months for controls, P<0.0001. That is a twofold difference in the filter time: the IRE group had to remain alive, controlled and eligible for four extra months before its comparison clock began. The initial safety report also shows smaller IRE tumours, 2.2 versus 3.2 cm, P=0.0066, and prior radiation in 42.5% versus 0%. Median survival from enrollment was 18 versus 10 months and the unadjusted HR was 0.4, but regression cannot recover unmeasured biology, referral and treatment choices after such different routes. Restricting the comparison to the ten prospectively enrolled controls gave HR 0.5 with 95% CI 0.3-1.0 and P=0.1016. The registry is useful for harm: 2/87, 2.3%, died from procedure-related bleeding or vascular injury in the initial report; 29/99 had an IRE-related adverse event and 55/99 had a grade 3 or worse event in the later preprint. NCT03899636 is the separate clean experiment: same mFOLFIRINOX chemotherapy with or without added IRE after three months without progression. It lists 528 estimated participants and an estimated April 2025 completion, but still says active, not recruiting, with no posted result and no update after February 2025. Estimated enrollment is not actual enrollment. The failure is causal substitution and reporting delay: a selected association is visible while the randomized answer is not.", "sourceIds": [ "clinicaltrials-nct03899649-direct-registry", "martin-2024-direct-registry-safety", "martin-2026-direct-registry-survival-preprint", "clinicaltrials-nct03899636-direct-rct" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "claim-ire-ablation-without-proven-added-survival", "relation": "limits" }, { "target": "unknown-ire-added-survival-after-modern-chemotherapy", "relation": "creates" }, { "target": "hypothesis-direct-randomized-data-recovery-gate", "relation": "repaired-by" }, { "target": "company-angiodynamics", "relation": "controlled-by" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "moderate", "randomised": "strong", "independentReplication": "weak", "sampleSize": "Registry safety n=114; registry survival n=137; randomized target n=528 estimated, actual assignment unknown", "effect": "The registry cannot identify added survival; the direct randomized result is not public", "limits": [ "Survival paper is a preprint", "Registry control has only ten prospective participants", "Regression cannot remove unmeasured selection", "Randomized enrollment is estimated", "Randomized completion date is estimated", "Sponsor controls both records" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The mechanism warrants a trial but does not correct selection in the registry." }, "animalEvidence": { "rating": "not-applicable", "reason": "This failure concerns human comparison and reporting." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "This failure concerns human comparison and reporting." }, "reproducibility": { "rating": "weak", "reason": "The apparent survival effect lacks a published randomized reproduction." }, "commercialConflicts": { "rating": "high", "reason": "The device maker sponsored the registry, controlled the randomized trial and supported both registry reports." }, "recency": { "rating": "current", "reason": "The analysis uses the May 2026 survival preprint and registry state through September 2026." } } }, "layout": { "x": 562, "y": 486 }, "updated": "2026-09-15" }, { "id": "failure-dynamic3-ctdna-guided-chemo-no-rfs-gain", "title": "DYNAMIC-III did not validate either ctDNA-guided chemotherapy rule", "shortTitle": "DYNAMIC-III action failure", "type": "failure", "status": "refuted", "scope": [ "colorectal" ], "summary": "De-escalation reduced oxaliplatin and hospital admission but failed noninferiority; escalation did not improve recurrence-free survival in ctDNA-positive stage III colon cancer.", "content": "Original belief: postoperative ctDNA could safely reduce chemotherapy in blood-negative stage III colon cancer and identify blood-positive patients who benefit from more chemotherapy. What happened: 968 patients were evaluable and 702, or 72.5%, were ctDNA negative. In the negative group, the guided rule cut oxaliplatin use from 88.6% to 34.8% and treatment-related hospital admission from 13.2% to 8.5%, but three-year recurrence-free survival was 85.3% versus 88.1% and the rule failed its noninferiority test. In the positive group, escalation produced two-year recurrence-free survival of 51% versus 61% under standard management; it did not improve outcome. Persistent ctDNA after treatment separated a very high-risk group, with three-year recurrence-free survival 14% versus 79%, so prognosis was strong while the tested actions failed. Why it probably failed: assay-defined risk does not create sensitivity to more of the same drug class, and a low measured signal does not prove that conventional pathological risk can be ignored. This rejects these de-escalation and escalation rules in this stage III setting. It does not reject ctDNA as a risk marker or every future action.", "sourceIds": [ "tie-2025-dynamic3", "tie-2022-dynamic-colon" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "claim-crc-ctdna-utility-is-action-specific", "relation": "supports" }, { "target": "unknown-crc-ctdna-positive-action", "relation": "leads-to" }, { "target": "trial-crc-action-map", "relation": "informs" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong", "independentReplication": "moderate for prognosis; weak for either exact action rule", "sampleSize": "968 evaluable; 702 ctDNA negative", "effect": "Negative group RFS 85.3% versus 88.1% and noninferiority not met; positive escalation 2-year RFS 51% versus 61%; persistent-positive 3-year RFS 14% versus 79%", "limits": [ "Open-label strategy trial", "Escalation analysis was exploratory", "Stage III colon cancer only", "The result tests named chemotherapy choices, not all MRD treatment", "Assay and treatment effects cannot be separated without the randomized action comparison" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "ctDNA marks residual tumour, but residual tumour need not be sensitive to more oxaliplatin or irinotecan." }, "animalEvidence": { "rating": "not-applicable", "reason": "The failed claim concerns human treatment decisions." }, "inVitroEvidence": { "rating": "weak", "reason": "Drug sensitivity in models cannot validate a treatment rule based on postoperative blood status." }, "reproducibility": { "rating": "moderate", "reason": "The prognostic split repeats; this exact stage III decision strategy has one randomized test." }, "commercialConflicts": { "rating": "moderate", "reason": "Commercial assay interests were disclosed, although the important result is randomized and negative." }, "recency": { "rating": "current", "reason": "The report was published in late 2025." } } }, "updated": "2026-09-14" }, { "id": "failure-ea2185-randomized-question-closed", "title": "EA2185 no longer provides the planned randomized answer on cyst-surveillance intensity", "shortTitle": "EA2185 randomized question closed", "type": "failure", "status": "supported", "scope": [ "pdac" ], "summary": "A 4,606-person randomized comparison was replaced by a 770-person observational biomarker cohort, leaving the core utility question unresolved.", "content": "The 2020 protocol planned 4,606 asymptomatic patients randomized 1:1 to high- or low-intensity imaging for five years. It was powered at 90% to detect a 30% relative reduction in a composite unfavorable outcome and was designed to measure surgery, cancer, costs, incidental findings and patient reports. The ClinicalTrials.gov record updated 4 September 2026 now lists an observational prospective cohort with estimated enrolment 770 and completion in November 2034. Both randomized arms are marked closed to accrual from 13 August 2024. The active primary question is whether baseline blood and imaging markers predict later worrisome or high-risk features. Some quality-of-life and cost comparisons remain for people randomized before the change, but the planned clinical-effectiveness comparison no longer has its intended sample or current primary role. This is a programme failure even if the remaining cohort produces useful biomarkers: the high-value randomized surveillance question has been lost. The registry does not state why the arms closed, so no cause should be assigned without protocol, sponsor or monitoring-board records. Recovery action: obtain accrual, crossover, event and amendment data; determine whether the randomized subset can answer any bounded question; then design a new utility comparison only for the decision not recoverable from that subset.", "sourceIds": [ "weinberg-2020-ea2185-protocol", "clinicaltrials-nct04239573-ea2185" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "trial-pancreatic-precursor-action-map", "relation": "failed-because-of" }, { "target": "unknown-dangerous-precursor-selection", "relation": "supports" }, { "target": "programme-100m-randomized-platform", "relation": "constrains" } ], "evidence": { "humanProspective": "weak — randomized subset exists but current accrual and event counts are not public", "humanRetrospective": "not-applicable", "randomised": "weak — randomized subset exists but public accrual and event counts are missing", "independentReplication": "none", "sampleSize": "Published plan n=4,606; current observational estimate n=770; randomized subset size not reported in the live record", "effect": "The original randomized clinical-effectiveness question is no longer the current study design or primary outcome", "limits": [ "Reason for the amendment is unknown", "Randomized-subset accrual and events are not public", "Registry may change", "Observational biomarker work can still be useful" ], "dimensions": { "biologicalPlausibility": { "rating": "not-applicable", "reason": "This is a trial-delivery and evidence-loss claim, not a biological mechanism." }, "animalEvidence": { "rating": "not-applicable", "reason": "Animal evidence cannot answer surveillance intensity." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Laboratory performance cannot replace a randomized care-pathway comparison." }, "reproducibility": { "rating": "strong", "reason": "The change is directly visible by comparing the published protocol with the current federal registry." }, "commercialConflicts": { "rating": "low", "reason": "The programme is cooperative-group and NCI sponsored; the main concern is delivery, not product ownership." }, "recency": { "rating": "current", "reason": "The registry was checked after its 4 September 2026 update." } } }, "layout": { "x": 21, "y": 103 }, "updated": "2026-09-14" }, { "id": "failure-echec-cancer-outcome-mismatch", "title": "eCHEC can test exposure control but cannot show that cholangiocarcinoma was prevented", "shortTitle": "eCHEC outcome mismatch", "type": "failure", "status": "open", "scope": [ "cholangiocarcinoma", "hepatobiliary", "prevention" ], "summary": "The randomized prevention study ends at 12-month self-reported behaviour and stool infection, years before a cholangiocarcinoma result could be measured.", "content": "Original belief: a randomized community behaviour programme labelled as cholangiocarcinoma prevention would supply direct prevention evidence. Observed design: NCT05321992 randomizes an estimated 480 adults to eCHEC or usual care, but its eight primary outcomes at 12 months are raw-fish intake, fruit and vegetable intake, smoking, alcohol, praziquantel use, stool positivity and concentration, and stage of behaviour change. Cholangiocarcinoma incidence, stage and mortality are absent. The record still lists recruiting, an estimated April 2026 completion, last verification in November 2024 and no posted results. Likely cause: cancer develops too slowly and rarely for a small five-year implementation award to measure directly. Mechanism status: the trial can validly test whether the programme changes infection exposure; it cannot establish cancer prevention by itself. Recovery: lock stool infection and reinfection as the short-term causal endpoint, preserve cluster assignment, and add passive ten- to twenty-year linkage to cancer diagnosis, site, stage, resection and death. Report both results even if behaviour improves but infection does not. Stop calling the programme cancer-effective until the exposure change reproduces and linked cancer outcomes move in the expected direction.", "sourceIds": [ "clinicaltrials-nct05321992-echec", "iarc-2012-liver-flukes", "charoensuk-2024-one-health-opisthorchis" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "claim-fluke-control-prevents-exposure-not-yet-cancer", "relation": "failed-because-of" }, { "target": "trial-cca-action-map", "relation": "constrains" }, { "target": "hypothesis-cca-route-specific-action-platform", "relation": "succeeds-where" } ], "evidence": { "humanProspective": "weak — planned randomized community study", "humanRetrospective": "not-applicable", "randomised": "moderate — yes for short-term exposure outcomes", "independentReplication": "none for the registered programme", "sampleSize": "Estimated n=480 with eight 12-month primary outcomes", "effect": "No posted result and no cancer endpoint", "limits": [ "Registry status may lag", "The trial was not designed to measure long-latency cancer", "Stool infection remains a useful causal endpoint", "Long-term linkage may not be funded", "Self-report can overstate behaviour change" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Reducing chronic fluke infection should reduce a proven carcinogenic exposure." }, "animalEvidence": { "rating": "moderate", "reason": "Animal evidence supports the causal chain but not this delivery model." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "The gap is endpoint and follow-up design." }, "reproducibility": { "rating": "none", "reason": "No results are posted for this trial and long-term cancer outcomes are not planned." }, "commercialConflicts": { "rating": "low", "reason": "The intervention is academic and community-based." }, "recency": { "rating": "current", "reason": "The live registry record was checked in September 2026." } } }, "updated": "2026-09-14" }, { "id": "failure-electrochemotherapy-endpoint-and-trial-drift", "title": "Electrochemotherapy studies lose patients, compare chosen devices and leave the randomized answer unpublished", "shortTitle": "ECT evidence-system failure", "type": "failure", "status": "supported", "scope": [ "pdac", "trial-design", "local-treatment", "device-governance" ], "summary": "The main 25-person cohort confuses scan response with benefit and chosen electrode geometry with a treatment comparison; the 90-person randomized trial has no public result after transition to CTIS.", "content": "The 25-person phase 1/2 cohort accrued from November 2011 to December 2019, about 25 / 8 = 3.1 patients per year. Its one-month imaging appears strong: 24/25 partial response or stable disease. By six months, however, 11/25, 44%, had no scan because they had died. Counting 14/25 remaining controlled is honest; describing response only among survivors would not be. The surgeon chose fixed or variable electrode geometry. The paper reports median survival of 6 versus 12 months, P=0.18, then calls the improvement 4.5 months even though 12 - 6 = 6 months. Neither arithmetic nor non-random groups support superiority. The decisive 90-person Laparoscopic-ECT-LAPC study started in April 2019, planned 60 months of recruitment and a 40% to 70% objective-response comparison. It has transitioned from EudraCT to EU CT 2024-519067-17-00, but no public result is linked. Its main endpoint is scan response, while survival and quality of life are secondary. The IREC paper uses a Results heading to list future endpoints for three planned electroporation arms but gives no participants, result or registry identifier. PanECT planned expansion beyond safety, enrolled 10, left survival outcomes unposted and reports one serious fistula followed by death. The repeated failure is endpoint promotion plus weak custody: technical coverage or scan change becomes a patient-benefit story before randomization, all outcomes and harm are public.", "sourceIds": [ "izzo-2021-pancreatic-ect-phase12", "izzo-2021-laparoscopic-ect-protocol", "eudract-2018-003925-27-ect", "rudno-rudzinska-2021-irec-protocol", "clinicaltrials-nct04281290-panect-results" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "claim-electrochemotherapy-local-effect-without-patient-benefit", "relation": "limits" }, { "target": "unknown-electrochemotherapy-added-benefit-and-delivery", "relation": "creates" }, { "target": "hypothesis-electrochemotherapy-randomized-data-gate", "relation": "repaired-by" }, { "target": "trial-pdac-electrochemotherapy-action-map", "relation": "mapped-by" }, { "target": "company-igea", "relation": "involves" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "weak — protocol only", "independentReplication": "weak", "sampleSize": "Single arm n=25; PanECT n=10; randomized target n=90 without public result", "effect": "Forty-four percent missing six-month imaging through death; no public randomized result", "limits": [ "The transitioned trial may be active in CTIS", "Actual randomized enrollment is not public", "Missing scans caused by death are reported", "The 4.5-month statement may be a writing error", "PanECT cannot attribute its fistula to ECT alone", "Feasibility remains established" ], "dimensions": { "biologicalPlausibility": { "rating": "not-applicable", "reason": "This record concerns trial design, arithmetic and reporting." }, "animalEvidence": { "rating": "not-applicable", "reason": "Animal evidence cannot repair missing randomized outcomes." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Cell uptake cannot establish survival or recurrence benefit." }, "reproducibility": { "rating": "weak", "reason": "The needed same-care randomized effect has not been reported." }, "commercialConflicts": { "rating": "high", "reason": "IGEA technology, staff and funding touch the principal programme." }, "recency": { "rating": "current", "reason": "PanECT results and the transitioned EU trial were checked through September 2026." } } }, "layout": { "x": 742, "y": 666 }, "updated": "2026-09-15" }, { "id": "failure-eli002-phase2-missed-dfs", "title": "ELI-002 generated KRAS immunity but missed its randomized phase 2 disease-free-survival endpoint", "shortTitle": "ELI-002 phase 2 miss", "type": "failure", "status": "contested", "scope": [ "pdac" ], "summary": "AMPLIFY-7P randomized 144 resected KRAS-mutant PDAC patients to ELI-002 7P or observation and did not meet its prespecified intention-to-treat DFS endpoint.", "content": "AMPLIFY-7P is the test that the earlier uncontrolled vaccine study required. It enrolled 144 people with resected stage I–III mutant-KRAS PDAC at 24 U.S. sites and compared ELI-002 7P with observation after standard local treatment. The sponsor reported that the trial did not meet its prespecified disease-free-survival endpoint in the intention-to-treat population. The public filing did not give the main hazard ratio or P value. Instead, it highlighted about 14 percentage points of separation at three and six months, a strong association between vaccine-induced T-cell response and outcome, and a post-hoc result in 121 R0-resected patients: median DFS 23.8 versus 12.8 months, HR 0.65, P = 0.048. R1 resections were imbalanced, 19% in the vaccine arm versus 10% in observation. That imbalance may matter, but excluding it after seeing results breaks the protection of randomization and raises the risk of a chance subgroup finding. The trial weakens the claim that immune response predicts treatment benefit. A confirmatory trial is defensible only with the R0 rule fixed before enrollment, stratification for residual-risk variables, longer dosing justified by measured immune decay, complete main-effect reporting and an endpoint analysis protected from subgroup selection. Three later complete responses after checkpoint treatment are uncontrolled post-study observations and belong only as a hypothesis for a prospective combination study.", "sourceIds": [ "elicio-2026-amplify7p-results", "wainberg-2025-eli002-final" ], "links": [ { "target": "claim-eli002-mrd-frontier", "relation": "weakens" }, { "target": "residual-disease", "relation": "targets" }, { "target": "failure-checkpoint-blockade-unselected-pdac", "relation": "does-not-yet-overcome" }, { "target": "companies", "relation": "involves" }, { "target": "experiments", "relation": "resolves" }, { "target": "failures", "relation": "part-of" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong", "independentReplication": "none", "sampleSize": "144 randomized across 24 U.S. sites", "effect": "Prespecified ITT DFS endpoint missed; sponsor-reported post-hoc R0 subgroup HR 0.65, P=0.048, n=121", "limits": [ "Sponsor topline filing", "Main hazard ratio and P value not publicly reported", "No full peer-reviewed report", "Overall survival immature", "Post-hoc subgroup", "Baseline R1 imbalance", "Immune-response association is not a randomized treatment effect", "Future combination claim rests on three uncontrolled patients" ] }, "layout": { "x": 14, "y": 118 }, "updated": "2026-09-14" }, { "id": "failure-elraglusib-analysis-and-survival-discordance", "title": "Elraglusib survival is sensitive to who counts and is not matched by tumour control", "shortTitle": "Elraglusib evidence gap", "type": "failure", "status": "mixed", "scope": [ "pdac" ], "summary": "The survival direction remains in all randomized patients, but the larger headline estimate excludes unequal proportions before treatment and has no matching PFS, response or disease-control effect.", "content": "Claim being tested. The reported 2.9-month median survival gain is being used to justify phase 3 development. Counting failure. Among 269 people randomized to the selected weekly schedule or control, 16 of 171 assigned elraglusib and 20 of 98 controls never received treatment. The removal rates are 16/171=9.4% and 20/98=20.4%. The main modified intention-to-treat analysis therefore keeps 155 and 78 people and breaks some of randomization's protection. Fifteen controls withdrew consent and four started standard care closer to home, compared with five consent withdrawals in the elraglusib arm. This open-label imbalance can make the treated groups differ before any drug is given. The signal does not disappear: all-randomized OS is 8.9 versus 7.2 months, HR 0.68, P=0.022. The repair is to lead with that estimate and explain why the treated estimate is larger. Endpoint failure. The original primary comparison was one-year survival. After the study began, regulatory feedback led to a survival-curve log-rank test; sample size was not recalculated. The paper says the change occurred after study initiation but does not make the full dated decision trail easy to audit. The current registry still names one-year survival, estimates 350 total participants, lists an estimated June 2026 completion, posts no results and says participant data sharing is no. Output mismatch. PFS, time to treatment failure, response, disease control and response duration all missed significance. The control median OS of 7.2 months is below the 8-10 months the paper describes for modern randomized GnP trials. Safety costs also rose: grade 3 or worse events were 89.7% versus 79.5%; severe neutropenia 52.3% versus 30.8%; severe fatigue 16.8% versus 5.1%; and any-grade transient visual impairment 67.7% versus 9.0%. Repair. Obtain the frozen participant data, protocol versions and dated analysis plans; reproduce survival in every randomized person; test missing-outcome, site, region, early-death and later-treatment effects; then demand a blinded confirmation whose primary result cannot depend on treatment receipt.", "sourceIds": [ "mahalingam-2026-elraglusib-randomized", "clinicaltrials-nct03678883-elraglusib" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "claim-elraglusib-randomized-survival-signal", "relation": "bounds" }, { "target": "unknown-elraglusib-survival-replication", "relation": "creates" }, { "target": "trial-elraglusib-action-map", "relation": "repaired-by" }, { "target": "company-actuate-therapeutics", "relation": "must-repair" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "moderate", "independentReplication": "none", "sampleSize": "Relevant randomized ITT n=269; primary treated set n=233; 36 randomized people excluded before treatment", "effect": "mITT OS HR 0.62 versus ITT HR 0.68; no significant PFS, response, disease-control or response-duration effect", "limits": [ "No participant data", "Endpoint-change chronology incomplete", "Registry not reconciled", "Open-label withdrawals", "Phase 2", "Safety comparison uses treated population" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "A real target exists, but biology does not explain away trial-accounting problems or the survival-control mismatch." }, "animalEvidence": { "rating": "moderate", "reason": "Models cannot resolve allocation, missing data or survival analysis." }, "inVitroEvidence": { "rating": "moderate", "reason": "Laboratory activity cannot prove that the observed survival difference was caused by treatment." }, "reproducibility": { "rating": "none", "reason": "No independent randomized dataset tests the same survival claim." }, "commercialConflicts": { "rating": "high", "reason": "The sponsor controls the trial record and benefits from the larger treated-population headline." }, "recency": { "rating": "current", "reason": "Compares the 2026 paper with the April 2026 registry update." } } }, "layout": { "x": 200, "y": 118 }, "updated": "2026-09-15" }, { "id": "failure-epigenetic-azacitidine-maintenance", "title": "Epigenetic maintenance did not delay PDAC recurrence", "shortTitle": "Azacitidine failure", "type": "failure", "status": "refuted", "scope": [ "pdac" ], "summary": "After surgery and adjuvant treatment, oral azacitidine produced PFS 9.2 versus 8.9 months and OS HR 0.98 in a 49-person randomized study.", "content": "Original belief: DNA hypomethylation might reactivate suppressed programmes, delay metastatic recurrence and sensitize later chemotherapy when burden was low. What happened: 49 high-risk resected patients were randomized to oral azacitidine or observation after standard adjuvant treatment. Imaging-confirmed median PFS was 9.2 versus 8.9 months, HR 0.94, P=.85; OS HR was 0.98, P=.96. Four of 24 assigned drug stopped for adverse events. Why it probably failed: a broad epigenetic drug was not tied to a dependency or responder marker; treatment began a median 9.6 months after surgery; most recurrences may already have been biologically established; human target engagement was scarcely sampled. What survives: state-specific epigenetic vulnerabilities could still matter, but must be locked before enrolment and show tumour target engagement. The result rejects unselected CC-486 maintenance, not every chromatin target.", "sourceIds": [ "heumann-2022-azacitidine-pdac" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "treatment", "relation": "part-of" }, { "target": "residual-disease", "relation": "failed-because-of" }, { "target": "claim-pdac-subtypes-are-mixed-evolving-states", "relation": "depends-on" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong", "independentReplication": "weak", "sampleSize": "49 randomized", "effect": "PFS HR 0.94, P=.85; OS HR 0.98, P=.96", "limits": [ "Small phase 2", "Baseline imbalance", "Late post-surgery start", "Minimal paired pharmacology", "Specific drug failure does not refute every epigenetic target" ] }, "layout": { "x": 126, "y": 75 }, "updated": "2026-09-14" }, { "id": "failure-fapi-scan-dose-benefit-collapse", "title": "Failure mode: a visible FAP lesion is treated as proof of dose and benefit", "shortTitle": "Scan ≠ dose ≠ benefit", "type": "failure", "status": "observed", "scope": [ "pdac", "pan-cancer", "transfer-unproven" ], "summary": "The development route loses denominators and collapses three different measurements: tracer visibility, absorbed radiation and patient benefit.", "content": "What fails. Scan visibility, absorbed dose and useful-life gain are serial gates, not synonyms. A bright PET scan measures tracer concentration at one time. Absorbed tumour dose integrates activity over time. Patient benefit further depends on what cells are hit, disease outside measured lesions, toxicity, other treatment and comparison with the best alternative. Denominator loss makes the signal look larger. In LuMIERE, 35 were scanned, 27 treated and 25 had a response assessment: 25 / 35 = 71.4% reached that measurement. Nine treated people had pancreatic cancer; four had stable disease and no pancreatic shrinkage was reported. In the 3BP-3940 report, 88 people received treatment but best response was reported for 51, or 58.0%. The published 34 / 51 = 66.7% response rate becomes 34 / 88 = 38.6% when displayed against the original treated cohort, before accounting for 21 different cancers, three isotopes, tandem schedules and additional standard care. This is not a corrected causal effect; it shows how denominator choice changes the picture. Programme loss is also evidence. FRONTIER stopped after 20 people for a stated business decision and posts no result. That cannot be called biological failure, but unpublished data and a vanished development route reduce what the public can learn. Repair. Every report must preserve scanned, eligible, treated, dose-evaluable, response-evaluable and survival denominators by cancer. Store scan threshold, lesion absorbed dose, normal-organ dose, isotope, co-treatment and lesion response in one table. Model within-person lesion dose-response separately from patient survival. A future treatment claim requires a concurrent control and all randomized people in the analysis.", "sourceIds": [ "mcconathy-2026-lumiere-phase1", "clinicaltrials-nct04939610-lumiere", "baum-2026-3bp3940", "clinicaltrials-nct05432193-frontier", "baum-2022-fap2286-firsthuman" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "claim-fap-radioligand-feasible-pdac-efficacy-unproven", "relation": "limits" }, { "target": "unknown-pdac-fap-radioligand-treatment-effect", "relation": "creates" }, { "target": "hypothesis-fapi-theranostic-pdac", "relation": "repaired-by" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "moderate", "randomised": "none", "independentReplication": "moderate — missing causal controls and uneven denominators recur across agents", "sampleSize": "LuMIERE 35 to 27 to 25; 3BP-3940 88 treated and 51 response-evaluable; FRONTIER 20 before termination", "effect": "Published 3BP-3940 response fraction is 34/51 = 66.7%; displayed from all treated it is 34/88 = 38.6%, with causality still unresolved", "limits": [ "Reasons for missing response assessments may differ", "Original-cohort response display is conservative, not a causal estimate", "FRONTIER gives no scientific stopping reason", "Cross-agent comparisons are descriptive" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Tracer kinetics, radiation deposition and tumour control are physically distinct processes." }, "animalEvidence": { "rating": "moderate", "reason": "Models can link dose and response but do not repair missing human denominators." }, "inVitroEvidence": { "rating": "moderate", "reason": "Cell killing after radiation does not measure delivery across a person." }, "reproducibility": { "rating": "strong", "reason": "The measurement split and denominator loss are visible in more than one human programme." }, "commercialConflicts": { "rating": "high", "reason": "Selected response denominators and unpublished stopped programmes can favour apparent progress." }, "recency": { "rating": "current", "reason": "Uses 2026 papers and current registry records." } } }, "layout": { "x": 68, "y": 138 }, "updated": "2026-09-15" }, { "id": "failure-gpc1-exosome-perfect-classifier", "title": "The claimed perfect GPC1 exosome test did not survive independent benign-disease comparison", "shortTitle": "GPC1 exosome failure", "type": "failure", "status": "refuted", "scope": [ "pdac" ], "summary": "A 2015 study reported 100% sensitivity and specificity; an independent study found 26.7% sensitivity and 87.5% specificity and could not separate PDAC from benign pancreatic disease.", "content": "Original belief: counting glypican-1-positive circulating exosomes could identify early and late PDAC with perfect accuracy. The originating study reported 100% sensitivity and 100% specificity in both its discovery and validation sets. What happened: independent teams could not reproduce the simple classifier. One direct benign-disease comparison found GPC1-positive extracellular-vesicle sensitivity of 26.7% and specificity of 87.5%; adding glycoprotein 2 produced 23.3% and 90.0%. Another study found enrichment with tumour burden but could not distinguish PDAC from benign pancreatic disease. Why it probably failed: extracellular-vesicle isolation and antibody assays are highly sensitive to method, cancer and benign inflammation overlap, case-control spectra were narrow, and the originating result came from small curated samples. A later author correction fixed several method and figure errors but did not withdraw the conclusion. What survives: extracellular vesicles carry tumour information and newer RNA-plus-protein assays may work, but GPC1 count alone is not a universal diagnostic. Any replacement needs fixed collection, isolation and calibration, blinded multi-centre testing against pancreatitis and cysts, and prospective low-prevalence validation.", "sourceIds": [ "melo-2015-gpc1-exosomes", "karam-2019-gpc1-nonreplication" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "detection", "relation": "part-of" }, { "target": "claim-case-control-blood-tests-overstate-screening-readiness", "relation": "supports" }, { "target": "experiments", "relation": "resolves" } ], "evidence": { "humanProspective": "none", "humanRetrospective": "moderate", "randomised": "none", "independentReplication": "strong", "sampleSize": "Originating validation: 56 PDAC, 20 healthy and 6 benign; direct nonreplication included 93 patients", "effect": "Originating sensitivity/specificity 100%/100%; independent GPC1 result 26.7%/87.5%", "limits": [ "Assays and isolation methods differed", "Small cohorts", "Case-control designs", "Does not refute all extracellular-vesicle markers", "Later assays can combine RNA and protein", "No prospective screening study" ] }, "layout": { "x": 10, "y": 96 }, "updated": "2026-09-14" }, { "id": "failure-grail-capital-scaled-before-utility", "title": "GRAIL shows the cost of scaling a detection platform before clinical utility", "shortTitle": "GRAIL capital lesson", "type": "failure", "status": "supported", "scope": [ "pancreatic-all", "pdac", "hcc", "colorectal", "pan-cancer" ], "summary": "A $7.1bn acquisition and very large subsequent losses preceded a failed randomized stage-shift endpoint and still no mortality proof.", "content": "Illumina completed its $7.1bn GRAIL acquisition in 2021 despite active competition challenges and spun the company out again in June 2024. GRAIL's 2024 10-K reports $9.8bn of post-acquisition cumulative net loss, but $6.8bn was goodwill and intangible impairment and $461.1m was amortisation: those accounting charges must not be mislabeled as research cash. The clinical lesson is separate and stronger. The roughly 142,250-person NHS-Galleri trial later missed its main combined stage III/IV endpoint, and cancer mortality plus pancreatic-specific net benefit remain unproved. GRAIL nevertheless built rare population-scale trial, laboratory and linked-data capacity, so this is not a claim that all spending produced nothing. It is evidence that valuation, acquisition price, commercial reach and diagnostic yield cannot substitute for a randomized patient-benefit gate. Mission rule: do not release population-scale capital merely because analytical validity, case-control sensitivity or commercial demand looks strong; require a locked disease-relevant endpoint and preserve long-term mortality follow-up.", "sourceIds": [ "ftc-2024-illumina-grail", "grail-2025-10k-capital-history", "sasieni-2026-nhs-galleri", "grail-2026-q2-10q" ], "links": [ { "target": "investment", "relation": "part-of" }, { "target": "company-grail", "relation": "constrains" }, { "target": "failure-nhs-galleri-primary-stage-shift", "relation": "failed-because-of" }, { "target": "claim-mced-workflow-feasible-utility-unproved", "relation": "contradicts" }, { "target": "claim-stage-shift-insufficient", "relation": "supports" }, { "target": "failures", "relation": "part-of" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "moderate", "randomised": "strong", "independentReplication": "weak", "sampleSize": "NHS-Galleri about 142,250 participants; acquisition and public-company filings", "effect": "$7.1bn transaction; main stage III/IV endpoint IRR 1.03, 95% CI 0.92–1.14", "limits": [ "Acquisition price is not R&D spending", "Net loss contains $6.8bn non-cash impairment", "Trial was pan-cancer rather than pancreatic-specific", "Mortality follow-up incomplete", "Infrastructure and evidence generated still have option value" ], "dimensions": { "biologicalPlausibility": { "rating": "not-applicable", "reason": "This record tests a capital-allocation and clinical-utility claim rather than a molecular mechanism." }, "animalEvidence": { "rating": "not-applicable", "reason": "Animal results cannot establish population screening utility or capital efficiency." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Analytical assay performance is insufficient for the patient-benefit conclusion." }, "reproducibility": { "rating": "moderate", "reason": "Trial and transaction facts are auditable, but mortality follow-up and independent utility replication remain incomplete." }, "commercialConflicts": { "rating": "high", "reason": "Company filings and a company-built test supply part of the evidence, creating strong commercial incentives." }, "recency": { "rating": "current", "reason": "The failed primary endpoint and financial position were reported in 2026." } } }, "layout": { "x": 92, "y": 101 }, "updated": "2026-09-14" }, { "id": "failure-grantpax-geriatric-route", "title": "GrantPax: assessment without enforced action under-recruited and misrouted treatment", "shortTitle": "GrantPax routing failure", "type": "failure", "status": "supported", "scope": [ "pdac" ], "summary": "GrantPax enrolled 32 of 135 planned patients, and clinicians placed 15 of 32 on treatment inconsistent with the study's own assessment.", "content": "Original belief: classifying each person as Go-Go, Slow-Go or Frail would make the treatment choice follow naturally. What happened: GrantPax planned 45 people in each of three routes, 135 total, but enrolled only 32 across six German centres and stopped for poor recruitment. The structured assessment classified 29/32, 91%, as Slow-Go and three as Go-Go. Clinicians nevertheless allocated 15/32, 47%, to a route inconsistent with those scores. Fifteen received combination gemcitabine plus nab-paclitaxel, 15 gemcitabine alone and two best supportive care. Serious adverse events affected 11/14, 78.6%, in the combination safety group and 8/15, 53.3%, with gemcitabine. Median time until quality of life worsened was 13 versus 29 days. Those comparisons are not treatment effects because allocation was not randomized and the groups differed before treatment. Why it failed: the assessment generated information but did not own the decision; its three labels hid the exact impairments and actions; the recruitment route missed 103 of 135 planned places; and the sample was too small to recover a stable rule. Recovery: enroll consecutive patients before the treatment decision, retain the raw domains, record the clinician and patient plan before and after assessment, assign each impairment to a named action within 72 hours, include people choosing supportive care in the denominator, and test the action system rather than relabelling drug intensity. Stop if assessment delays treatment, if clinicians can silently override without a reason, or if the programme reports only treated survivors.", "sourceIds": [ "betge-2022-grantpax" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "claim-age-alone-cannot-route-pdac-treatment", "relation": "supports" }, { "target": "unknown-older-pdac-treatment-route", "relation": "leads-to" }, { "target": "hypothesis-older-pdac-fast-support-route", "relation": "recovered-by" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "not-applicable", "randomised": "none", "independentReplication": "weak", "sampleSize": "Planned n=135; enrolled n=32; safety n=31", "effect": "15/32, 47%, assigned inconsistently with assessment; serious adverse events 78.6% versus 53.3% in nonrandom groups", "limits": [ "Severe under-recruitment", "Nonrandom treatment allocation", "Six German centres", "Two supportive-care patients", "Unstable small-group survival", "Assessment labels required clinician interpretation" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "An assessment cannot change treatment harm unless its findings trigger timely care or a reasoned treatment change." }, "animalEvidence": { "rating": "not-applicable", "reason": "This is a human recruitment and decision-ownership failure." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Laboratory systems cannot test recruitment, consent or clinician override." }, "reproducibility": { "rating": "weak", "reason": "The exact misrouting estimate comes from one small failed study, although wider trials support acting on assessment findings." }, "commercialConflicts": { "rating": "moderate", "reason": "The study tested branded nab-paclitaxel plus gemcitabine, but the route failure does not depend on a manufacturer comparison." }, "recency": { "rating": "recent", "reason": "The 2022 result remains the only direct PDAC assessment-allocation failure of this form." } } }, "updated": "2026-09-14" }, { "id": "failure-hcc-biomarker-addition-no-early-stage-gain", "title": "Adding three serum markers to ultrasound did not improve early-stage HCC detection", "shortTitle": "Marker add-on did not improve detection", "type": "failure", "status": "refuted", "scope": [ "hcc", "hepatobiliary" ], "summary": "A 1,208-person randomized trial found 22 early HCCs with ultrasound plus AFP, AFP-L3 and DCP versus 30 with ultrasound alone, HR 0.81, with a wide confidence interval.", "content": "Original belief: adding AFP, AFP-L3 and DCP to six-month ultrasound would trigger confirmatory imaging early enough to increase early-stage HCC detection. What happened: 1,208 adults with cirrhosis or high-risk HBV were randomized, 603 to ultrasound and 605 to ultrasound plus the three markers. HCC was diagnosed in 35 versus 27 people, with 30 versus 22 early-stage cases. Early-stage detection did not differ, HR 0.81, 95% CI 0.47 to 1.40, P=0.45. The point estimate did not favour the marker arm, but the wide interval also means the trial cannot exclude useful benefit or harm. Why it probably failed: fixed thresholds can add alerts without reliably separating early cancer from chronic liver injury, and the event count was too small to compare marker combinations cleanly. Two authors were employed by Fujifilm companies and the registry names Wako Life Sciences as sponsor, so commercial incentives must be shown alongside the negative result. This refutes the tested three-marker add-on and threshold rule, not every longitudinal blood-marker strategy. TRACER tests the GALAD combination as the primary screening route rather than a simple add-on.", "sourceIds": [ "hirode-2026-hcc-biomarker-rct", "singal-2024-tracer-protocol", "clinicaltrials-nct06084234-tracer" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "claim-hcc-surveillance-test-improvement-not-mortality", "relation": "supports" }, { "target": "trial-hcc-action-map", "relation": "informs" }, { "target": "unknown-hcc-surveillance-modality-utility", "relation": "leads-to" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong", "independentReplication": "weak for the exact add-on and threshold rule", "sampleSize": "1,208 randomized; 62 HCCs and 52 early-stage HCCs", "effect": "Early-stage detection HR 0.81, 95% CI 0.47-1.40; 30 early cases under ultrasound alone and 22 with the marker add-on", "limits": [ "Only 62 HCCs occurred", "The authors state that comparisons beyond ultrasound plus AFP were underpowered", "The registry describes a broader observational study rather than the published random allocation in a clear one-to-one form", "Most participants had HBV and more than half were Asian", "Company-employed authors and an industry lead sponsor", "Mortality was not the primary endpoint" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "The markers track tumour or liver biology, but chronic injury can raise them and fixed thresholds may discard useful change over time." }, "animalEvidence": { "rating": "not-applicable", "reason": "Animal studies cannot establish surveillance utility in a human care route." }, "inVitroEvidence": { "rating": "moderate", "reason": "Analytical assays exist, but assay detection does not prove an early-stage clinical gain." }, "reproducibility": { "rating": "weak", "reason": "This exact add-on rule has one completed randomized result." }, "commercialConflicts": { "rating": "high", "reason": "The paper includes Fujifilm-company authors and the linked registry names Wako Life Sciences as lead sponsor." }, "recency": { "rating": "current", "reason": "The result was published in 2026 and checked in September 2026." } } }, "updated": "2026-09-14" }, { "id": "failure-hcc-three-month-ultrasound-no-small-cancer-gain", "title": "Three-month ultrasound found more tiny liver lesions but not more small HCC", "shortTitle": "More scans, no small-HCC gain", "type": "failure", "status": "refuted", "scope": [ "hcc", "hepatobiliary" ], "summary": "In 1,278 people with compensated cirrhosis, shortening ultrasound surveillance from six to three months raised detection of lesions 10 mm or smaller but not HCC incidence or the proportion of HCC 30 mm or smaller.", "content": "Original belief: if six-month surveillance misses fast-growing HCC, doubling scan frequency should find more tumours while they remain small. What happened: 1,278 people with compensated cirrhosis at 43 French and Belgian sites were randomized to ultrasound every three or six months. At least one focal lesion was found in 358/1,278 = 28.0%, while HCC was confirmed in 123/1,278 = 9.6%. The difference, 358 - 123 = 235 people, is about 1.91 people with a focal finding but no confirmed HCC for each confirmed HCC, although one person could have more than one work-up. Three-month scanning increased the share of detected lesions that were 10 mm or smaller from 28% to 41%, but did not significantly change HCC incidence or the share of HCC 30 mm or smaller, 79% versus 70%, P=0.30. Why it failed: scan frequency increased the supply of indeterminate small lesions faster than the recall route could classify useful cancer. This refutes three-month ultrasound as a general route to more small HCC in compensated cirrhosis. It does not refute a shorter interval for a tightly selected fast-risk state with a better confirmatory test.", "sourceIds": [ "trinchet-2011-hcc-surveillance-interval" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "claim-hcc-surveillance-test-improvement-not-mortality", "relation": "supports" }, { "target": "trial-hcc-action-map", "relation": "informs" }, { "target": "detection", "relation": "constrains" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong", "independentReplication": "weak for this exact interval comparison", "sampleSize": "1,278 randomized at 43 sites", "effect": "Lesions 10 mm or smaller 41% versus 28%; HCC 30 mm or smaller 79% versus 70%, P=0.30", "limits": [ "The trial compared intervals, not surveillance with no surveillance", "The primary endpoint was tumour size rather than mortality", "Diagnostic imaging and treatment have changed since 2011", "All participants had compensated histologically proven cirrhosis", "The calculation does not count repeated procedures per person" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Cirrhotic livers generate many benign or indeterminate nodules, so more frequent imaging can increase work-up without increasing useful cancer detection." }, "animalEvidence": { "rating": "not-applicable", "reason": "Animal evidence cannot resolve the human interval and recall-path question." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "The failure concerns clinical imaging frequency and work-up." }, "reproducibility": { "rating": "weak", "reason": "This is one large multicentre randomized interval trial." }, "commercialConflicts": { "rating": "low", "reason": "No commercial sponsor is needed to explain the negative interval result." }, "recency": { "rating": "historical", "reason": "The trial was published in 2011, so current imaging and recall routes may differ." } } }, "updated": "2026-09-14" }, { "id": "failure-heat-component-and-registry-drift", "title": "HEAT changed too many treatment parts to identify the value of heating", "shortTitle": "HEAT component failure", "type": "failure", "status": "supported", "scope": [ "pdac", "trial-design", "regional-hyperthermia" ], "summary": "One randomized contrast mixed heat, cisplatin, gemcitabine dose and schedule, and a control-regimen change; it cannot estimate a heat effect.", "content": "The trial name foregrounds hyperthermia, but the random assignment did not change heat alone. The registered control used gemcitabine plus capecitabine. The experimental arm used gemcitabine on different days and at a lower total planned dose, added cisplatin, and added regional hyperthermia. The publication then reports that only eight control participants actually received capecitabine. The observed difference can therefore be written as: treatment contrast = heat effect + cisplatin effect + gemcitabine schedule and dose effect + capecitabine or control-drift effect + interactions. One equation with at least five unknown terms cannot solve the heat term. Excluding the eight capecitabine recipients changed the overall-survival P value to 0.052, but a later subset change does not repair the trial architecture and still did not cross 0.05. The trial also allocated only 34.8% of its planned 336 people. This is a design and reporting failure, not proof that regional heating has no biological effect. The repair is to recover a randomized record where the arms differ only by heat, reconcile every assigned participant and preserve measured temperature as treatment-delivery data.", "sourceIds": [ "issels-2023-heat-rct", "clinicaltrials-nct01077427-heat", "eudract-2008-004802-14-heat" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "claim-heat-regional-hyperthermia-primary-endpoint-negative", "relation": "limits" }, { "target": "unknown-regional-hyperthermia-added-benefit-modern-care", "relation": "creates" }, { "target": "hypothesis-heat-existing-randomization-recovery-gate", "relation": "repaired-by" }, { "target": "trial-pdac-regional-hyperthermia-action-map", "relation": "mapped-by" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "not-applicable", "randomised": "moderate", "independentReplication": "none", "sampleSize": "117 randomized; 336 planned", "effect": "No isolated regional-hyperthermia estimate is identifiable", "limits": [ "The paper's control differs from the registered control", "Only eight controls received capecitabine", "Cisplatin and heat were bundled", "Gemcitabine timing and total dose differed", "Treatment interactions are unknown", "Early trial end reason is not stated in the regulator record", "No public participant-level dataset" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "Heating can alter blood flow and treatment sensitivity, but this comparison cannot measure its contribution." }, "animalEvidence": { "rating": "not-applicable", "reason": "The failure is in the human experiment's component isolation." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Cell experiments cannot separate the bundled clinical interventions." }, "reproducibility": { "rating": "none", "reason": "No second reported pancreatic randomized trial isolates heat." }, "commercialConflicts": { "rating": "low", "reason": "The trial was non-commercial; its central problem is design and record drift rather than ownership." }, "recency": { "rating": "mixed", "reason": "The result was published in 2023, but the chemotherapy comparison and registry record are old." } } }, "layout": { "x": 622, "y": 546 }, "updated": "2026-09-15" }, { "id": "failure-hifu-randomization-and-trial-drift", "title": "HIFU evidence mixes a retrospective comparison, a nine-year small trial and stopped device programmes", "shortTitle": "HIFU evidence-system failure", "type": "failure", "status": "supported", "scope": [ "pancreatic-all", "pdac", "trial-design", "device-governance" ], "summary": "The strongest old survival contrast is not demonstrably randomized, while the first prospective randomized study still reports recruiting after nine years and supplies no results or participant-data plan.", "content": "The 120-person HIFU plus S-1 paper presents itself two ways. Its abstract says patients were randomly assigned. Its methods say the investigators retrospectively collected medical records and that patients were divided into groups. It gives no sequence generation, allocation concealment, dated protocol or trial identifier. The adjusted death hazard ratio is 0.193 despite medians of 10.3 versus 6.6 months, an unusually large model result whose dependence on covariates cannot be checked from participant data. Treat it as a retrospective comparison. HIFU-PC15 was meant to repair this with a prospective randomized control. It began on 24 February 2017, targets only 40 people, still says recruiting on a registry last updated in February 2024 and has no final count, result table or planned participant-data sharing. Its 2024 protocol mixes locally advanced, metastatic and recurrent disease, first and second treatment lines and several chemotherapy regimens in 20 people per group. NCT05262452 planned 60 single-arm participants but has unknown status, a passed March 2023 primary date, a record stale since 2022 and no results. PULS planned 26, then suspended in September 2026 because its industrial partner decided to close recruitment permanently; no actual count or result is posted. The failure is not acoustic physics. It is the inability to preserve assignment, denominator, comparable systemic care, device delivery and public outcome data long enough to answer a patient question.", "sourceIds": [ "li-2015-hifu-s1", "marinova-2024-hifu-pc-protocol", "drks-00012367-hifu-pc", "clinicaltrials-nct05262452-hifu-folfirinox", "clinicaltrials-nct06211933-puls-hifu" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "claim-hifu-pain-signal-without-survival-proof", "relation": "limits" }, { "target": "unknown-hifu-pain-survival-and-system-transfer", "relation": "creates" }, { "target": "hypothesis-hifu-existing-randomized-gate", "relation": "repaired-by" }, { "target": "trial-pdac-hifu-action-map", "relation": "mapped-by" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "strong", "randomised": "weak", "independentReplication": "weak", "sampleSize": "Retrospective n=120; HIFU-PC15 target n=40 since 2017; stale single arm target n=60; suspended PULS target n=26", "effect": "No auditable randomized survival result; multiple incomplete or stale trial routes", "limits": [ "The old paper may have used an undocumented allocation process", "HIFU-PC15 could have unpublished data", "Registry status can lag site activity", "Actual enrollment in PULS is not public", "Failure of one industrial partner does not refute the device class", "Pain relief can remain useful without a survival effect" ], "dimensions": { "biologicalPlausibility": { "rating": "not-applicable", "reason": "This record concerns trial design, custody and reporting." }, "animalEvidence": { "rating": "not-applicable", "reason": "Model evidence cannot repair missing allocation or outcomes." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Cell death cannot answer the failed clinical comparison." }, "reproducibility": { "rating": "weak", "reason": "The claimed survival contrast has not been reproduced in an auditable randomized study." }, "commercialConflicts": { "rating": "high", "reason": "Device firms sponsor current studies, funded the protocol article charge or controlled whether recruitment continued." }, "recency": { "rating": "current", "reason": "The industrial-partner closure was posted in September 2026." } } }, "layout": { "x": 712, "y": 636 }, "updated": "2026-09-15" }, { "id": "failure-hipec-local-endpoint-and-package-selection", "title": "HIPEC evidence confuses local control with survival and whole packages with one component", "shortTitle": "HIPEC component and selection failure", "type": "failure", "status": "supported", "scope": [ "pdac", "hipec", "trial-design", "selection", "local-treatment" ], "summary": "The randomized result shows that fewer local recurrences need not extend life; metastatic series select long responders and change several treatments at once.", "content": "The adjuvant trial is a direct warning. HIPEC changed local recurrence by 42 percentage points but changed median survival by minus 0.9 months, 17.1 versus 18.0. It cannot support the argument that stronger local control automatically cures more people. It was also too small for stable harm or subgroup estimates, and fewer HIPEC participants received later chemotherapy, 12/21 versus 16/21. The metastatic route adds a larger attribution problem. Entry required response to at least six months of systemic therapy, no disease outside the abdomen, very low peritoneal burden and an operation judged likely to remove every visible deposit. The pilot then combined cytoreduction, HIPEC and one of three treatments for the pancreatic primary. A person must pass the response gate, anatomy gate, fitness gate and operative-completion gate before their survival enters the treated series. The live NCT04858009 study preserves the same narrow selection and has one arm. Its survival clock begins at cytoreduction and HIPEC, after the long-response interval used to qualify. A favourable result cannot tell whether survival came from tumour biology, prior chemotherapy, complete surgery, HIPEC or their interaction.", "sourceIds": [ "padilla-valverde-2024-adjuvant-hipec-rct", "grotz-2022-crs-hipec-pdac-pilot", "clinicaltrials-nct04858009-hipec-pdac", "yan-2024-crs-hipec-pdac-retrospective" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "claim-hipec-local-control-without-survival-proof", "relation": "limits" }, { "target": "unknown-hipec-added-value-peritoneal-pdac", "relation": "creates" }, { "target": "hypothesis-hipec-existing-trial-measurement-gate", "relation": "repaired-by" }, { "target": "trial-pdac-hipec-action-map", "relation": "mapped-by" }, { "target": "failure-conversion-surgery-selection-bias", "relation": "supports" }, { "target": "failure-pipac-repeat-procedure-selection-and-bundling", "relation": "compares-with" } ], "evidence": { "humanProspective": "strong — for identifying the design failure", "humanRetrospective": "moderate — for showing package selection", "randomised": "moderate", "independentReplication": "weak", "sampleSize": "Randomized n=42; prospective selected package n=18; active single-arm target n=40", "effect": "Large local-control change without survival gain; metastatic HIPEC contribution unidentifiable", "limits": [ "Adjuvant trial underpowered for survival", "Metastatic population differs", "Some package components may interact", "Exact screening denominator is missing", "Current study may improve data capture" ], "dimensions": { "biologicalPlausibility": { "rating": "not-applicable", "reason": "This is an attribution and endpoint failure." }, "animalEvidence": { "rating": "not-applicable", "reason": "Models cannot separate the clinical package." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Cell killing cannot establish net life gained after surgery." }, "reproducibility": { "rating": "moderate", "reason": "Local-control-without-survival is directly visible in the randomized data and consistent with wider PDAC local-treatment failures." }, "commercialConflicts": { "rating": "low", "reason": "The main pressure is procedure and centre commitment rather than a named product owner." }, "recency": { "rating": "current", "reason": "The active Mayo design was updated one day before this audit." } } }, "layout": { "x": 1114, "y": 1038 }, "updated": "2026-09-15" }, { "id": "failure-histotripsy-acoustic-window-and-translation", "title": "Histotripsy can fail because its controller cannot see through bowel gas", "shortTitle": "Histotripsy control-loop failure", "type": "failure", "status": "supported", "scope": [ "pdac", "local-treatment", "imaging", "device-governance" ], "summary": "The pancreas is behind gas-filled bowel; poor ultrasound visibility caused severe off-target injury in pigs, while the first human study can establish only selected-patient feasibility.", "content": "This is a sensing-and-control failure, not a weak tissue-destruction mechanism. The system needs an acoustic path that both sends therapy and shows the target and bubble cloud. Bowel gas scatters ultrasound. In the nine-pig safety study, pancreas targeting succeeded in 6/9, 66.7%, and the bubble cloud was visible in 5/9, 55.6%. Three treatments reached bowel or nearby tissue when visibility was poor. Two of nine animals, 22.2%, deteriorated enough to be killed early. The investigators intentionally treated under adverse visibility to expose the failure mode; the clinical lesson is a hard interlock, not an average safety claim. The GANNON human study correctly excludes targets without an adequate acoustic window or with less than one centimetre separation from stomach or bowel. That protection also selects an unusually favourable subset, so its safety rate cannot be generalized to everyone with pancreatic cancer. It enrolls 50 in one arm and lists only 30-day complications as primary. It will not show that histotripsy prolongs life. The FDA liver authorization is also narrower than promotional shorthand: it allows non-invasive destruction of liver tumours and explicitly says cancer-treatment safety and effectiveness were not established. Liver clearance is not pancreas proof.", "sourceIds": [ "gannon-2026-pancreas-histotripsy-safety", "clinicaltrials-nct06282809-gannon", "fda-den220087-edison" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "claim-histotripsy-pancreas-visibility-limited", "relation": "limits" }, { "target": "unknown-histotripsy-human-coverage-and-benefit", "relation": "creates" }, { "target": "hypothesis-gannon-independent-safety-gate", "relation": "repaired-by" }, { "target": "trial-pdac-histotripsy-action-map", "relation": "mapped-by" }, { "target": "company-histosonics", "relation": "involves" }, { "target": "failure-hifu-randomization-and-trial-drift", "relation": "compares-with" } ], "evidence": { "humanProspective": "none — feasibility trial recruiting", "humanRetrospective": "none", "randomised": "none", "independentReplication": "weak", "sampleSize": "Nine pigs; planned human n=50 selected for visible target and bowel clearance", "effect": "Two of nine pigs required early killing after mistargeting; human design cannot estimate survival benefit", "limits": [ "Adverse-condition treatments were deliberate", "Animal anatomy and depth differ", "Human exclusion rules may prevent the observed failures", "Clinical enrollment count is not posted", "No pancreas result yet", "FDA record concerns liver", "Long-term human harm is unknown" ], "dimensions": { "biologicalPlausibility": { "rating": "not-applicable", "reason": "The limiting issue is imaging and control, not whether cavitation fragments tissue." }, "animalEvidence": { "rating": "strong", "reason": "The loss-of-visibility failure caused direct severe outcomes in living animals." }, "inVitroEvidence": { "rating": "moderate", "reason": "Bench work can test acoustic paths but not moving human bowel and pancreas." }, "reproducibility": { "rating": "weak", "reason": "Independent pancreas-specific system tests are not established." }, "commercialConflicts": { "rating": "high", "reason": "The manufacturer sponsors the clinical translation and supported the animal work." }, "recency": { "rating": "current", "reason": "The safety report and trial status were checked in September 2026." } } }, "layout": { "x": 814, "y": 738 }, "updated": "2026-09-15" }, { "id": "failure-hybridtherm-product-accrual-and-custody", "title": "HybridTherm failed as a trial system before its clinical question was settled", "shortTitle": "HybridTherm system failure", "type": "failure", "status": "supported", "scope": [ "pdac", "trial-design", "local-treatment", "device-governance" ], "summary": "Slow recruitment, a withdrawn product, manufacturer-held allocation and conflicting enrollment records turned a randomized study into a small uncertain result.", "content": "The programme planned 33 people per arm for its phase 2 decision and 100 per arm for the full comparison. Other centres could not recruit and follow participants, so the study became single-centre. It randomized 40 people over about 4.6 years, 40 / 4.6 = 8.7 per year. The manufacturer withdrew the probe in 2020 and recruitment stopped. Based on the observed six-month rates, the authors calculated that 288 people per arm, 576 total, would be needed to test the originally desired 20-point gain; they judged that infeasible. Three people assigned to the probe arm were excluded after metastases were found, breaking a strict all-assigned comparison. The manufacturer held the randomization list, monitored accuracy, completeness and reliability, and employed two paper authors. The registry now says actual enrollment was 66 but posts no results; the paper says 40 were randomized and analyzes 37. That 26-person mismatch is 65% of the paper's randomized count. It may reflect registry maintenance rather than patient fabrication, but it must be reconciled from dated screening, consent and assignment logs. The product, recruitment network, custody design and public record all failed before a reliable patient-benefit answer was produced.", "sourceIds": [ "testoni-2021-hybridtherm-rct", "clinicaltrials-nct02336672-hybridtherm", "clinicaltrials-nct04990609-pancardinal1" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "claim-hybridtherm-randomized-no-patient-benefit", "relation": "limits" }, { "target": "unknown-eus-rfa-added-benefit-after-hybridtherm", "relation": "creates" }, { "target": "hypothesis-eus-rfa-existing-study-gate", "relation": "repaired-by" }, { "target": "trial-pdac-eus-rfa-action-map", "relation": "mapped-by" }, { "target": "company-erbe-elektromedizin", "relation": "involves" }, { "target": "failure-rfa-selection-and-route-attrition", "relation": "supports" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "moderate", "independentReplication": "weak", "sampleSize": "Planned phase 2 n=66; randomized n=40; reported analysis n=37; registry says actual n=66 without results", "effect": "Trial stopped for slow accrual and product withdrawal; records disagree by 26 people", "limits": [ "The reason for the registry count mismatch is unknown", "No allegation of misconduct follows from the mismatch", "Small size leaves moderate effects unresolved", "The device is no longer available", "The paper reports technical safety", "Successor probes are not technically identical" ], "dimensions": { "biologicalPlausibility": { "rating": "not-applicable", "reason": "This record concerns recruitment, product continuity and control of trial data." }, "animalEvidence": { "rating": "not-applicable", "reason": "Model evidence cannot repair a broken randomized evidence chain." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Bench ablation cannot reconcile participant counts or survival." }, "reproducibility": { "rating": "weak", "reason": "The sole randomized study stopped early and has no independent raw-data reproduction." }, "commercialConflicts": { "rating": "high", "reason": "The same manufacturer controlled the product, allocation list and monitoring." }, "recency": { "rating": "current", "reason": "The registry mismatch and active successor design were checked in September 2026." } } }, "layout": { "x": 778, "y": 702 }, "updated": "2026-09-15" }, { "id": "failure-imbrave050-interim-benefit-disappeared", "title": "The apparent IMbrave050 adjuvant HCC benefit disappeared with longer follow-up", "shortTitle": "IMbrave050 reversal", "type": "failure", "status": "refuted", "scope": [ "hcc" ], "summary": "Recurrence-free survival moved from HR 0.72 at the planned interim analysis to HR 0.90 later; overall survival HR was 1.26 and immature, so the benefit-risk profile no longer supported treatment.", "content": "Original belief: because atezolizumab plus bevacizumab prolongs survival in unresectable HCC, one year after resection or ablation might eliminate occult cells and prevent recurrence. What happened: among 668 randomised high-risk patients, the prespecified interim recurrence-free-survival result was positive at hazard ratio 0.72. With longer follow-up the hazard ratio weakened to 0.90, 95% CI 0.72 to 1.12. Overall survival remained immature and numerically favoured surveillance, hazard ratio 1.26, 95% CI 0.85 to 1.87. The investigators concluded the benefit-risk profile did not support adjuvant use. Why it probably failed: early and late HCC recurrence mix metastatic relapse with new primary tumours arising from a diseased liver; a treatment effective against measurable advanced disease may not clear either route durably, and early event separation can vanish. This refutes the broad regimen and schedule, not all immune interception. A better trial must classify recurrence route, continue cause control, use time-updated liver state and demand stable late benefit.", "sourceIds": [ "yopp-2026-imbrave050-update", "finn-2020-imbrave150", "bruix-2015-storm-sorafenib" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "failure-sorafenib-adjuvant-hcc", "relation": "replicates" }, { "target": "model-hcc-causal-chain", "relation": "supports" }, { "target": "experiments", "relation": "resolves" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong", "independentReplication": "moderate", "sampleSize": "668 randomised patients", "effect": "RFS HR changed from 0.72 at interim to 0.90 with longer follow-up; immature OS HR 1.26", "limits": [ "Later RFS analysis was post hoc", "OS immature", "Open-label treatment versus surveillance", "Cross-over after recurrence", "Sponsor-funded", "Does not refute other immune regimens or biomarker-selected subgroups" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "Immune and vascular treatment can attack occult tumour, but cannot prevent every new primary HCC in a damaged liver." }, "animalEvidence": { "rating": "moderate", "reason": "Models supported the combination but did not establish a stable adjuvant effect." }, "inVitroEvidence": { "rating": "moderate", "reason": "Target activity does not model recurrence route or long follow-up." }, "reproducibility": { "rating": "moderate", "reason": "The longer follow-up reverses the same trial's interim result and aligns with the earlier STORM adjuvant failure." }, "commercialConflicts": { "rating": "high", "reason": "The sponsor owned both medicines and the commercial adjuvant indication." }, "recency": { "rating": "current", "reason": "The reversal was reported with longer follow-up in 2026." } } }, "layout": { "x": 21, "y": 92 }, "updated": "2026-09-14" }, { "id": "failure-immune-recruitment-without-response", "title": "Moving T cells into PDAC did not produce useful response when suppressive myeloid cells moved with them", "shortTitle": "T-cell entry was insufficient", "type": "failure", "status": "supported", "scope": [ "pdac" ], "summary": "CXCR4 blockade plus PD-1 treatment produced no responses in 21 patients, while a vaccine and dual-checkpoint trial produced two responses among 57 despite tumour-specific T-cell recruitment.", "content": "The original belief was that PDAC checkpoint resistance could be overcome by bringing T cells into the tumour. In a phase 2 trial, plerixafor blocked CXCR4 while cemiplimab blocked PD-1. Twenty-one reported patients had no objective responses and only two best responses of stable disease. Paired samples showed more CD8 T cells, but also more granulocytes and macrophages; macrophages sat closer to T cells and an immunosuppressive macrophage state expanded. A separate randomized phase 2 study gave CRS-207, nivolumab and ipilimumab with or without GVAX to 57 treated patients. It generated mesothelin- and mutant-KRAS-specific T-cell clones inside tumours, yet produced only two partial responses, both without GVAX, and 39/57 = 68% had related grade 3 or worse events. Myeloid and regulatory T-cell signatures were enriched and tracked poor response. These studies do not prove that every immune combination will fail. They refute T-cell entry or expansion as a sufficient success marker. The surviving mechanism is a layered treatment rule: require antigen-specific function, favourable myeloid state, tumour killing and a clinical endpoint, and stop a programme that changes infiltration without tumour control.", "sourceIds": [ "ho-2025-cxcr4-pd1-pdac", "bever-2026-vaccine-checkpoint-pdac", "feig-2013-cxcl12-immune-exclusion", "bayne-2012-gmcsf-myeloid" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "claim-immune-exclusion-is-multicellular", "relation": "supports" }, { "target": "failure-checkpoint-blockade-unselected-pdac", "relation": "extends" }, { "target": "hypothesis-low-burden-ras-immune-extinction", "relation": "constrains" }, { "target": "treatment-modality-ranking", "relation": "updates" } ], "evidence": { "humanProspective": "strong — two interventional phase 2 studies with paired immune measurements", "humanRetrospective": "not-applicable", "randomised": "moderate — one randomized comparison between two experimental immune combinations", "independentReplication": "moderate", "sampleSize": "CXCR4 plus PD-1 paper n=21; vaccine plus dual-checkpoint trial n=57 treated", "effect": "No objective responses and two stable-disease best responses in the CXCR4 study; 2/57 partial responses and 39/57 grade 3+ related events in the vaccine/checkpoint study", "limits": [ "Small trials", "Previously treated metastatic setting", "No standard-care control in either efficacy comparison", "Different immune interventions", "The plerixafor registry lists 25 enrolled while the paper reports 21", "Mechanism measurements do not establish every resistance cause" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Paired human tissue shows that T-cell recruitment can coexist with suppressive macrophage and regulatory-cell expansion." }, "animalEvidence": { "rating": "moderate", "reason": "Parallel mouse work supports macrophage-mediated resistance, but human treatment selection still requires direct validation." }, "inVitroEvidence": { "rating": "moderate", "reason": "Antigen-specific T-cell expansion was measured, yet this laboratory activity failed to predict useful tumour response." }, "reproducibility": { "rating": "moderate", "reason": "Two different clinical programmes produced immune recruitment with little response and independently implicated suppressive myeloid states." }, "commercialConflicts": { "rating": "moderate", "reason": "Drug-company collaboration and investigator relationships exist, but both studies were institution-led and report negative efficacy plainly." }, "recency": { "rating": "current", "reason": "The decisive reports were published in 2025 and 2026." } } }, "layout": { "x": 13, "y": 98 }, "updated": "2026-09-14" }, { "id": "failure-immune-reprogramming-without-patient-benefit", "title": "PDAC immune programmes repeatedly change cells without proving that patients gain useful life", "shortTitle": "Immune-change endpoint failure", "type": "failure", "status": "supported", "scope": [ "pdac", "pan-cancer", "transfer-unproven" ], "summary": "TGF-beta, CD40 and macrophage programmes expose the same error: target engagement, cell movement and within-arm marker correlations are not controlled patient benefit.", "content": "System failure. These drugs try to alter the tumour's local immune network rather than directly poison every cancer cell. The chain is target binding, tissue change, immune-cell state, cancer-cell killing, durable control, useful life and acceptable harm. Most reports stop in the middle. TGF-beta. NIS793 reached and changed the intended tissue system, yet the 511-person phase 3 lost 2.0 median survival months and added 14.6 percentage points of serious adverse events. CD40. PRINCE found immune patterns associated with survival, but neither CD40 arm passed its historical survival gate and there was no chemotherapy-only arm. OPTIMIZE-1 produced a better-looking 57-person result on a different chemotherapy backbone but no control. Macrophages. The CSF1R antibody cabiralizumab was meant to remove or alter tumour-supporting macrophages. In a 205-person randomized phase 2 after prior chemotherapy, cabiralizumab plus nivolumab alone had worse central-review PFS than chemotherapy, HR 1.47. Adding the pair to gemcitabine chemotherapy gave PFS HR 1.00 and OS HR 1.04. A 5-FU chemotherapy arm gave PFS HR 0.78 and OS HR 0.81, but both confidence intervals included no effect and liver-test abnormalities were common. A separate randomized plan enrolled two people. Repair. Publish every participant, keep an identical chemotherapy-only arm, set survival and patient function above tissue markers, and require a fixed treatment-by-marker interaction before selected claims. Stop cabiralizumab and broad NIS793. Preserve only the cheap archived-sample test and the controlled mitazalimab question.", "sourceIds": [ "clinicaltrials-nct04935359-danis2", "pelletier-2026-nis793-biomarkers", "padron-2022-prince", "van-laethem-2025-optimize1-biomarkers", "clinicaltrials-nct03336216-cabiralizumab", "clinicaltrials-nct03697564-gemcan" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "claim-nis793-target-engagement-without-benefit", "relation": "supported-by" }, { "target": "claim-cd40-signal-without-contribution-proof", "relation": "bounds" }, { "target": "trial-immune-reprogramming-action-map", "relation": "repaired-by" }, { "target": "failure-immune-recruitment-without-response", "relation": "extends" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "moderate", "randomised": "strong", "independentReplication": "strong", "sampleSize": "NIS793 phase 3 n=511; cabiralizumab phase 2 n=205; PRINCE efficacy n=105; OPTIMIZE-1 efficacy n=57", "effect": "NIS793 harmed the outcome direction; cabiralizumab gave no controlled benefit; CD40 signal remains uncontrolled", "limits": [ "Different immune targets and treatment lines", "Cabiralizumab 5-FU confidence intervals are wide", "PRINCE cannot isolate immune-drug contribution", "Mitazalimab may differ from sotigalimab" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The full chain contains several measurable places where tissue change can fail to become killing or survival." }, "animalEvidence": { "rating": "strong", "reason": "Strong animal rationales repeatedly failed to predict broad patient benefit." }, "inVitroEvidence": { "rating": "strong", "reason": "Target action is measurable but is shown not to be sufficient." }, "reproducibility": { "rating": "strong", "reason": "Three different immune-reprogramming routes expose the same endpoint and control error." }, "commercialConflicts": { "rating": "high", "reason": "Product owners controlled most trials, assays and participant-level data." }, "recency": { "rating": "current", "reason": "Includes results posted through 2026." } } }, "layout": { "x": 280, "y": 214 }, "updated": "2026-09-15" }, { "id": "failure-impact-pro-status-unknown", "title": "imPaCT-PRO cannot answer its survival question while status and results remain unreported", "shortTitle": "imPaCT-PRO reporting gap", "type": "failure", "status": "supported", "scope": [ "pdac" ], "summary": "A planned 450-person phase 3 directly tested tinzaparin, VTE and progression, but its status became unknown and the passed completion date has no posted result.", "content": "Original belief: a larger current-regimen study could determine whether anticoagulation affects both thrombosis and tumour progression. NCT05178628 planned an open-label phase 3 in 450 advanced or metastatic PDAC patients at two Greek sites. It assigned therapeutic-dose tinzaparin, 175 anti-Xa IU/kg daily, with gemcitabine and nab-paclitaxel versus chemotherapy alone. Twelve-month PFS and VTE were co-primary, with bleeding, response and quality of life secondary. What happened publicly: the registry was last updated in March 2022, now reports unknown status, retains a December 2024 completion estimate and posts no result. This is not evidence that tinzaparin failed. It is an execution and reporting failure that leaves a patient, site and data-accounting question. Before funding another survival trial, contact the sponsor, recover actual enrolment and arm counts, adverse events, VTE, imaging, PFS and reasons for interruption. If fewer than 70% of planned participants entered or outcome completeness is below 80%, do not use the study for a survival claim; publish the operational failure and return to the already-settled VTE-prevention question.", "sourceIds": [ "clinicaltrials-nct05178628-impact-pro", "pelzer-2015-conko004" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "trial-pdac-vte-action-map", "relation": "constrains" }, { "target": "unknown-pdac-vte-prophylaxis-rule", "relation": "leaves-open" }, { "target": "programme-500m-integrated-mission", "relation": "blocks-duplication-by" } ], "evidence": { "humanProspective": "weak for registered intent; none for results", "humanRetrospective": "not-applicable", "randomised": "weak for registered intent; none for reported results", "independentReplication": "none", "sampleSize": "Estimated n=450 at two sites; actual enrolment is not posted", "effect": "No public result; status unknown after a passed December 2024 completion estimate", "limits": [ "Registry-only evidence", "Actual recruitment unknown", "Sponsor may hold unposted data", "Status can change", "No inference about efficacy" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "Tinzaparin prevents clotting, while a separate direct antitumour effect remains unproved." }, "animalEvidence": { "rating": "not-applicable", "reason": "The failure concerns trial execution and reporting." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Laboratory evidence cannot replace missing randomized outcomes." }, "reproducibility": { "rating": "none", "reason": "No result exists to reproduce." }, "commercialConflicts": { "rating": "moderate", "reason": "The named product and missing data require sponsor and supply disclosures before interpretation." }, "recency": { "rating": "current", "reason": "Unknown registry status and absence of results were checked in September 2026." } } }, "layout": { "x": 156, "y": 80 }, "updated": "2026-09-14" }, { "id": "failure-k912-component-and-local-endpoint", "title": "The K-912 study calls a two-component package sonodynamic without identifying either component", "shortTitle": "K-912 component failure", "type": "failure", "status": "supported", "scope": [ "pdac", "cholangiocarcinoma", "sonodynamic-therapy", "trial-design" ], "summary": "Heat, chemotherapy and any sound-triggered interaction are inseparable, while a one-month local scan is promoted despite body-wide spread and short survival.", "content": "The causal system has at least three terms: drug effect D, ultrasound heat or cavitation effect U and any interaction I. The observed package is D + U + I. Because every participant received both components, the study measures only the sum. Its four sequential groups of three used 75 or 150 watts and 30 or 80 mg/m2 K-912. Coagulation occurred in 4/6 at 75 watts and 5/6 at 150 watts; all four complete coagulations occurred at 150 watts. It occurred in 5/6 at the lower drug dose and 4/6 at the higher dose. These tiny non-random groups cannot estimate dose response, but they do not show that more K-912 drives the scan result. Heat power tracks complete coagulation more closely. The endpoint also breaks the disease model. Every participant had peritoneal spread, but the headline is local coagulation at one month. Five tumours shrank, fewer than the nine called coagulated. One complete-coagulation case had progressive disease at that same month. Survival ranged from 37 to 677 days among pancreatic participants and the median was 140. Safety follow-up highlighted 30 days and repeat treatment was not tested. The route may be a useful lower-power HIFU plus chemotherapy package. Calling it a distinct sound-activated drug mechanism requires direct release chemistry and component controls.", "sourceIds": [ "muragaki-2026-k912-sonodynamic" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "claim-k912-first-human-local-signal-no-component-proof", "relation": "limits" }, { "target": "unknown-k912-ultrasound-drug-interaction-and-benefit", "relation": "creates" }, { "target": "hypothesis-k912-component-identification-gate", "relation": "repaired-by" }, { "target": "trial-pdac-k912-sonodynamic-action-map", "relation": "mapped-by" }, { "target": "company-sonire-therapeutics", "relation": "involves" }, { "target": "failure-tigerpac-component-and-interim-inference", "relation": "supports" } ], "evidence": { "humanProspective": "weak", "humanRetrospective": "none", "randomised": "none", "independentReplication": "none", "sampleSize": "Four sequential dose cells of n=3; total n=12", "effect": "Complete coagulation appeared only at 150 W, with no positive K-912 dose pattern; local image response diverged from progression and survival", "limits": [ "Post-hoc dose comparison", "Very small cells", "No HIFU-only group", "No K-912-only group", "K-912 has ordinary chemotherapy activity", "One cholangiocarcinoma mixed in", "Long-term toxicity incompletely tested" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "A sound-drug interaction is possible, but the study cannot separate it from heat and chemotherapy." }, "animalEvidence": { "rating": "moderate", "reason": "Animal package results do not repair missing human component controls." }, "inVitroEvidence": { "rating": "weak", "reason": "The exact release and excess-kill relation needs matched-temperature testing." }, "reproducibility": { "rating": "none", "reason": "No independent clinical repeat exists." }, "commercialConflicts": { "rating": "moderate", "reason": "The manuscript received company medical-writing support despite public funding." }, "recency": { "rating": "current", "reason": "The failure is visible in the 2026 first-human report." } } }, "layout": { "x": 1048, "y": 972 }, "updated": "2026-09-15" }, { "id": "failure-kras-wildtype-matching-route-and-selection", "title": "Molecular matching loses patients before treatment and overstates selected survival", "shortTitle": "Rare-driver matching failure", "type": "failure", "status": "supported", "scope": [ "pdac", "precision-treatment", "gene-fusion", "care-delivery" ], "summary": "A promising matched-treatment subgroup can hide failed tissue, late reports, missing outcomes, unavailable drugs and patients who became too unwell.", "content": "Know Your Tumor shows both the promise and the denominator failure. Of 1,856 referred pancreatic patients, 1,082, or 58%, received a molecular report. Of those, 282 had an alteration labelled actionable. Outcomes were available for 677, 189 had an actionable alteration, and 46 received matched treatment. The matched group lived 2.58 versus 1.51 years from advanced diagnosis, HR 0.42, but treatment was not randomized. The group had to survive testing, remain fit, find a drug and be selected for it. In all-referred terms, only 46/1,856=2.5% reached the matched group. That does not negate their possible benefit; it prevents the 0.42 hazard ratio from becoming a profiling effect. The same route can fail technically. DNA panels can miss fusions, RNA can fail from poor tissue, a KRAS-wild-type result can be a test failure rather than true absence, and a tumour-wide drug response can be based on other cancers. England's pancreatic RNA panel names NTRK1/2/3 as essential and lists several other reportable targets but not NRG1; laboratory implementation must be checked rather than inferred. Repair. Begin the denominator at diagnostic tissue, confirm tumour content and KRAS result, reflex suitable cases to RNA, state which targets the panel can and cannot see, return the result before treatment fitness falls, and record whether a drug decision and actual treatment followed. Never pool unlike targets into a single matched-therapy success rate.", "sourceIds": [ "pishvaian-2020-know-your-tumor", "philip-2022-kras-wildtype-pdac", "fusco-2021-kras-wildtype-fusions", "nhs-england-2026-cancer-genomic-test-directory", "oreilly-2019-ntrk-pdac-response", "fda-2018-larotrectinib-approval" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "failure-avatar-precision-medicine-delivery", "relation": "replicates" }, { "target": "failure-nrg1-testing-and-confirmation-route", "relation": "extends" }, { "target": "unknown-kras-wildtype-driver-route-completion", "relation": "creates" }, { "target": "hypothesis-kras-wildtype-driver-route-completion", "relation": "succeeds-where" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "strong", "randomised": "weak — the observed survival comparison was treatment-selected", "independentReplication": "moderate — route loss repeats in precision programmes, with different systems", "sampleSize": "Know Your Tumor n=1,856 referred, 1,082 reports, 677 outcomes and 46 matched treatments", "effect": "Only 2.5% of all referrals reached the matched-treatment group; selected HR 0.42 cannot estimate profiling utility", "limits": [ "Retrospective outcomes", "Survivor and fitness selection", "Commercial profiling", "Assay changes", "Heterogeneous targets and drugs" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Target dependence can be real while the care route loses most candidates." }, "animalEvidence": { "rating": "not-applicable", "reason": "The failure is in human sampling, access, timing and selection." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Laboratory response cannot recover missing patient-route denominators." }, "reproducibility": { "rating": "moderate", "reason": "Know Your Tumor and AVATAR expose similar route loss in different systems." }, "commercialConflicts": { "rating": "high", "reason": "Several authors were affiliated with the profiling provider and products have commercial owners." }, "recency": { "rating": "current", "reason": "The failure is checked against the July 2026 English panel route." } } }, "layout": { "x": 2140, "y": 1540 }, "updated": "2026-09-15" }, { "id": "failure-laser-fistula-accrual-and-device-continuity", "title": "Pancreatic laser programmes failed at leakage, recruitment and device continuity", "shortTitle": "Laser programme execution failure", "type": "failure", "status": "supported", "scope": [ "pdac", "laser-ablation", "trial-design", "device-lifecycle" ], "summary": "The route's largest reported cohort produced three pancreatic-fluid leaks, its randomized study stopped at eight and a later endoscopic study stopped because its device disappeared.", "content": "This is a coupled tissue-and-delivery failure. Heat must cross a tumour volume without cooking bowel, vessels, nerves or the duct network that carries digestive enzymes. A nominal setting of power times seconds gives input energy, but local temperature also depends on absorption, fibre geometry, blood-flow cooling and the distance to fragile structures. The nine-person endoscopic study measured a dead zone after treatment but did not report real-time temperature throughout the target. The open surgical imILT programme used feedback control yet completed only 13/15 assigned procedures. Three treated people, 23%, developed late pancreatic fistulas and four treatment-related events occurred overall. Its own authors called the result unsatisfactory. The only randomized imILT study enrolled eight, stopped for slow recruitment and posted nothing; it also compared laser without simultaneous chemotherapy against chemotherapy, so it changed the whole treatment stack instead of isolating added laser value. A five-person endoscopic trial stopped because the device was unavailable. A ten-person planning study is still marked unknown after its December 2023 expected finish and mixes several tumour types. These are not isolated paperwork issues. Together they show that anatomy, workflow, recruitment, hardware continuity, data return and causal design all failed before efficacy could be tested.", "sourceIds": [ "paiella-2021-imilt-pancreas", "clinicaltrials-nct02702986-imilt-verona", "clinicaltrials-nct02973217-imilt-marseille", "clinicaltrials-nct03187587-imilt-randomized", "clinicaltrials-nct03784417-eus-laser", "clinicaltrials-nct05549960-laser-model" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "claim-pancreatic-laser-heat-feasibility-not-benefit", "relation": "limits" }, { "target": "unknown-laser-ablation-delivery-and-net-benefit", "relation": "creates" }, { "target": "hypothesis-laser-existing-study-recovery-gate", "relation": "repaired-by" }, { "target": "trial-pdac-laser-ablation-action-map", "relation": "mapped-by" }, { "target": "company-clinical-laserthermia-systems", "relation": "involves" }, { "target": "failure-hybridtherm-product-accrual-and-custody", "relation": "supports" } ], "evidence": { "humanProspective": "weak", "humanRetrospective": "none", "randomised": "weak — stopped at n=8 without result", "independentReplication": "weak", "sampleSize": "Surgical n=15 assigned; randomized n=8; terminated endoscopic n=5; unknown planning target n=10", "effect": "Three fistulas among 13 treatments; two programmes terminated; one registry stale", "limits": [ "Event grades are not stated in the abstract", "Small denominators make rates unstable", "Reasons for each randomized non-enrollment are unavailable", "Device unavailability may be temporary", "Mixed tumour planning study may contain no pancreatic adenocarcinoma", "Registry results may exist privately" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Thermal injury near enzyme-carrying ducts provides a direct route to leakage." }, "animalEvidence": { "rating": "moderate", "reason": "Animal feasibility did not prevent human leaks or programme failure." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Recruitment, hardware continuity and pancreatic fistulas are clinical system failures." }, "reproducibility": { "rating": "weak", "reason": "No human programme reached a stable controlled test." }, "commercialConflicts": { "rating": "high", "reason": "The manufacturer sponsored the failed randomized study and collaborated on the surgical studies." }, "recency": { "rating": "current", "reason": "Registry states and current commercial focus were checked in 2026." } } }, "layout": { "x": 976, "y": 900 }, "updated": "2026-09-15" }, { "id": "failure-leopard2-laparoscopic-whipple-safety", "title": "LEOPARD-2 found no recovery gain and an unsafe mortality signal after laparoscopic Whipple", "shortTitle": "LEOPARD-2 safety stop", "type": "failure", "status": "supported", "scope": [ "pancreatic-all" ], "summary": "Even trained surgeons at centres doing at least 20 Whipple procedures yearly had 5/50 versus 1/49 complication-related deaths, and the trial stopped early.", "content": "Original belief: a laparoscopic Whipple could shorten recovery after training and volume controls. Observed result: LEOPARD-2 required four Dutch centres to perform at least 20 pancreaticoduodenectomies yearly and surgeons to complete dedicated training and at least 20 laparoscopic procedures. After 105 randomizations and 99 operations, the monitoring board stopped the trial. Complication-related 90-day mortality was 5/50, 10%, after laparoscopic versus 1/49, 2%, after open surgery, RR 4.90 with a very wide 95% CI 0.59-40.44 and P=0.20. Median functional recovery was not better: 10 versus 8 days, P=0.80. Severe complications were 50% versus 39%. The low count makes the harm estimate imprecise, but a safety board need not wait for a conventional significance threshold when deaths rise and benefit is absent. The study included benign, premalignant and malignant indications, so it is not a PDAC survival comparison. Later robotic trials do not erase this result: they use a different platform, selected high-volume teams and mainly short-term endpoints. DIPLOMA supports minimally invasive left pancreatectomy in resectable PDAC; it does not support laparoscopic Whipple. PORTAL supports faster recovery with credentialled robotic Whipple in seven Chinese centres but costs more and has not yet supplied long-term PDAC control. Mechanism status: smaller access can help recovery in the right operation and team; unmeasured proficiency and transfer can kill. Recovery: introduce each operation and platform under procedure-specific training, independent case-level audit, fixed rescue standards and stopping rules.", "sourceIds": [ "van-hilst-2019-leopard2", "korrel-2023-diploma", "bruna-2025-diploma-long-term", "klotz-2024-europa", "liu-2024-robotic-open-pd", "jin-2026-portal" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "claim-pdac-surgery-is-a-complete-treatment-route", "relation": "constrains" }, { "target": "trial-pdac-surgical-route-action-map", "relation": "informs" }, { "target": "unknown-pdac-surgical-route-completion", "relation": "creates-safety-boundary-for" }, { "target": "hypothesis-pdac-surgical-route-completion", "relation": "constrains" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "none", "randomised": "strong for the tested procedure", "independentReplication": "weak — mixed across different operations and platforms", "sampleSize": "LEOPARD-2 n=105 randomized, 99 operated; comparison trials n=81, n=164, n=258 and n=268", "effect": "Complication-related 90-day mortality 10% versus 2%; no functional-recovery benefit", "limits": [ "Early stop", "Only six deaths", "Very wide confidence interval", "Mixed indications", "Laparoscopic Whipple only", "Later robotic trials are not direct repeats" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "A long learning curve and complex reconstruction can turn technical errors into leak, haemorrhage, organ failure and death." }, "animalEvidence": { "rating": "not-applicable", "reason": "Human operative safety is decisive." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "No bench model can establish surgical proficiency or rescue." }, "reproducibility": { "rating": "weak", "reason": "Later trials test robotic rather than laparoscopic Whipple or a different left-sided operation and show mixed short-term results." }, "commercialConflicts": { "rating": "moderate", "reason": "LEOPARD-2 had an investigator-initiated Johnson and Johnson Medical grant; DIPLOMA also had device-company funding." }, "recency": { "rating": "mixed", "reason": "The decisive safety stop is from 2019 and is interpreted beside randomized results through July 2026." } } }, "updated": "2026-09-14" }, { "id": "failure-liaki-undisclosed-conflict-retraction", "title": "An undisclosed company interest forced retraction of a high-profile pancreatic mouse study", "shortTitle": "Triple-therapy disclosure failure", "type": "failure", "status": "supported", "scope": [ "pdac", "pan-cancer" ], "summary": "PNAS retracted the first triple-therapy paper because three authors held undisclosed financial interests; a declared, directly reviewed replacement was later published.", "content": "This is a governance failure, not proof that the underlying experiments were false. PNAS retracted the December 2025 paper because Mariano Barbacid, Vasiliki Liaki and Carmen Guerra held financial interests in Vega Oncotargets that were not disclosed for a contributed submission. The journal published a replacement through its direct-submission route in June 2026 with a new DOI and expanded declarations. The chronology changes how the result should be weighted: its mouse data can be examined, but confidence is reduced by the original disclosure failure, founder ownership around the tested idea, small groups and no independent reproduction. It also shows why fundraising and media attention cannot be an evidence gate. Mission rule: every programme records company, patent, equity, research-contract and sponsor relationships before review; conflicted investigators do not vote on capital release; raw images and analysis code are escrowed; and an independent laboratory repeats any result used to justify a human combination. Retraction status must be linked to the exact version so a corrected replacement is neither hidden nor confused with the withdrawn record.", "sourceIds": [ "pnas-2026-liaki-retraction", "liaki-2026-triple-combination-republished", "cnio-experimental-oncology-2026" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "claim-kras-egfr-stat3-mouse-regression", "relation": "constrains" }, { "target": "risks", "relation": "supports" }, { "target": "lab-cnio-experimental-oncology", "relation": "constrains" } ], "evidence": { "humanProspective": "not-applicable", "humanRetrospective": "not-applicable", "randomised": "not-applicable", "independentReplication": "strong", "sampleSize": "One original article, one journal retraction notice and one replacement article", "effect": "Original DOI retracted for undisclosed relevant competing interests; replacement published with declarations", "limits": [ "Retraction was for disclosure and submission-route failure, not a stated data finding", "Replacement does not independently reproduce the experiments", "Current commercial interests still require management" ] }, "layout": { "x": 88, "y": 95 }, "updated": "2026-09-14" }, { "id": "failure-local-control-without-survival-lap07", "title": "Better local control did not extend survival in LAP07", "shortTitle": "Local-control limit", "type": "failure", "status": "supported", "scope": [ "pdac" ], "summary": "Chemoradiotherapy reduced local progression from 46% to 32% after induction chemotherapy, while median survival was 15.2 versus 16.5 months.", "content": "LAP07 asked whether adding chemoradiotherapy after four months of controlled locally advanced pancreatic cancer improved survival. Among 269 patients in the second randomisation, chemoradiotherapy reduced local progression from 46% to 32% but did not improve median overall survival: 15.2 versus 16.5 months, hazard ratio 1.03. This shows that local control and overall survival can separate. A reasonable inference is that disease outside the radiation field or later systemic failure can dominate mortality in this selected setting. The trial does not rule out modern radiation for pain, local complications, better-selected biology, conversion to surgery or newer systemic backbones. It rules out assuming that a local-progression gain automatically becomes a survival gain.", "sourceIds": [ "hammel-2016-lap07" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "residual-disease", "relation": "supports" }, { "target": "treatment", "relation": "part-of" }, { "target": "bottlenecks", "relation": "supports" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong", "independentReplication": "moderate", "sampleSize": "449 enrolled; 269 entered chemoradiotherapy versus chemotherapy randomisation", "effect": "Local progression 32% versus 46%; median OS 15.2 versus 16.5 months, HR 1.03", "limits": [ "Only disease controlled after induction entered second randomisation", "Older gemcitabine-era systemic treatment", "Open-label", "Does not test current radiation methods or surgery conversion", "Inference about distant disease is not a direct mechanism measurement" ] }, "layout": { "x": 30, "y": 80 }, "updated": "2026-09-14" }, { "id": "failure-local-intensification-pdac-survival-not-shown", "title": "More local treatment has repeatedly improved local measures without proving longer survival in PDAC", "shortTitle": "Local intensification limit", "type": "failure", "status": "supported", "scope": [ "pdac" ], "summary": "Six randomized radiation or ablation studies did not prove longer survival; RTOG 0848 and PELICAN added harm, while PANOVA-3 remains a small contradictory device result.", "content": "The recurring claim is that stronger local destruction will convert locally advanced or borderline resectable PDAC into durable control. Five randomized tests bound that claim. CROSSFIRE compared MRI-guided ablative radiotherapy with irreversible electroporation after FOLFIRINOX: median survival was 16.1 versus 12.5 months, but the comparison was not significant and the trial stopped for futility; both arms lacked a chemotherapy-only control. SABER directly added stereotactic radiotherapy to mFOLFIRINOX: 37 of 92 planned patients enrolled, median PFS was 14.0 versus 11.7 months and OS 28.9 versus 28.1 months, neither significant. In CONKO-007, chemoradiotherapy after induction chemotherapy improved the R0 rate among the selected patients who reached surgery, 69.4% versus 50.0%, but not the overall R0 rate or survival, HR 0.937. In A021501, the radiotherapy arm closed early after only 10 of its first 30 evaluable patients achieved R0 resection; 18-month survival was 47.3% versus 66.7% with mFOLFIRINOX alone. PELICAN randomized 188 non-progressing locally advanced patients: radiofrequency ablation changed median survival from 11.6 to 12.1 months, HR 1.07, raised grade 3 or worse serious events from 11% to 27% and worsened quality of life. PANOVA-3 is the important contradiction: tumour-treating fields added to gemcitabine and nab-paclitaxel improved median survival by 2.0 months, HR 0.82, even though PFS, local PFS and response did not improve. The combined evidence rejects both extreme claims. Local treatment is not universally futile, but local endpoints and selected-cohort survival are far from durable control. Surgery, radiation, ablation and fields are not interchangeable. Each future trial needs a current systemic control, a locked response state and survival plus quality of life.", "contentSections": [ "RTOG 0848 adds a postoperative boundary. Among 354 people who had pancreatic-head resection and remained progression-free through five chemotherapy cycles, adding chemoradiotherapy did not improve overall survival, HR 0.96, or disease-free survival, and raised grade 3 toxicity from 19% to 38%. A reported node-negative subgroup interaction needs confirmation with current systemic treatment; it is not a general radiation rule." ], "sourceIds": [ "abrams-2026-rtog0848", "timmer-2024-crossfire", "yoo-2026-saber", "fietkau-2025-conko007", "katz-2022-a021501", "hammel-2016-lap07", "seelen-2026-pelican", "macarulla-2025-panova3" ], "links": [ { "target": "failure-local-control-without-survival-lap07", "relation": "replicates" }, { "target": "claim-pdac-death-can-be-local-or-metastatic", "relation": "depends-on" }, { "target": "claim-smad4-failure-route-marker-contested", "relation": "requires-better-selector-than" }, { "target": "residual-disease", "relation": "depends-on" }, { "target": "experiments", "relation": "resolves" }, { "target": "claim-panova3-ttfields-small-os-benefit", "relation": "contradicted-by" }, { "target": "failure-conversion-surgery-selection-bias", "relation": "supports" }, { "target": "trial-lapc-conversion-surgery-action-map", "relation": "constrains" }, { "target": "failures", "relation": "part-of" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong", "independentReplication": "strong", "sampleSize": "RTOG 0848 n=354 randomized after resection and chemotherapy; CROSSFIRE n=68; SABER n=37; CONKO-007 randomized n=336; A021501 n=126; LAP07 n=442; PELICAN n=188", "effect": "Six cited radiation or ablation comparisons did not show an overall-survival benefit; RTOG 0848 doubled grade 3 toxicity and PELICAN added 16 points of serious harm; PANOVA-3 showed a 2.0-month benefit, HR 0.82", "limits": [ "Different disease stages", "Different systemic backbones", "RTOG 0848 used gemcitabine-era postoperative treatment", "RTOG 0848 node-negative subgroup signal requires confirmation", "Several trials stopped early", "CROSSFIRE lacked a chemotherapy-only control", "Local modalities differ", "PANOVA-3 used a burdensome device and did not improve PFS or local PFS", "May conceal benefit in a selected local-dominant subgroup", "Does not estimate resection benefit" ] }, "layout": { "x": 14, "y": 112 }, "updated": "2026-09-15" }, { "id": "failure-losartan-delivery-signal-not-replicated", "title": "Losartan's single-arm PDAC delivery signal did not reproduce in advanced randomized treatment", "shortTitle": "Losartan scale-up failure", "type": "failure", "status": "supported", "scope": [ "pdac", "pan-cancer", "transfer-unproven" ], "summary": "A 49-person combined route reported 31.4-month survival; AFPAC found no response, progression or survival gain and stopped at 88 of 264 planned.", "content": "Mechanism. Losartan is a blood-pressure drug. In PDAC models it can reduce solid pressure and collagen, open compressed vessels and improve drug movement. Initial human route. One centre treated 49 locally advanced patients with losartan, eight planned FOLFIRINOX cycles, short- or long-course radiation and attempted surgery. Thirty-nine completed eight cycles, 42 reached attempted surgery, 34/49=69% had a clean microscopic margin, and median survival was 31.4 months. Every treatment component and the selection of patients who stayed local could produce that outcome. Randomized cross-check. AFPAC assigned 44 advanced patients to mFOLFIRINOX and 44 to the same chemotherapy plus losartan. Median survival was 10.4 versus 9.1 months, progression 8.44 versus 7.36 months, and response 22% versus 23%. Plasma TGF-beta did not separate, and nine of 41 losartan-treated patients paused it. The planned 264-person study stopped after the unplanned 88-person efficacy look found no signal. Decision. Losartan does not improve broad advanced-PDAC treatment at the tested dose and schedule. It remains unresolved only in localized disease, where decompression could matter before radiation and surgery. NCT03563248 has already randomized 168 people across the needed losartan and no-losartan groups. Recover that result; do not start another cohort.", "sourceIds": [ "murphy-2019-losartan-lapc", "ramaswamy-2025-afpac-losartan", "clinicaltrials-nct03563248-losartan-localized" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "failure-vascular-starvation-pdac", "relation": "contrasts-with" }, { "target": "unknown-localized-losartan-contribution-pdac", "relation": "leaves-open" }, { "target": "hypothesis-localized-losartan-existing-trial-gate", "relation": "tested-by" }, { "target": "trial-pdac-vascular-hypoxia-action-map", "relation": "changes" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "moderate", "independentReplication": "moderate", "sampleSize": "Single-arm localized n=49; randomized advanced n=88 of 264 planned; localized randomized registry n=168", "effect": "AFPAC response 22% versus 23%, PFS 8.44 versus 7.36 months and OS 10.4 versus 9.1 months", "limits": [ "AFPAC stopped at one third of planned accrual", "Its efficacy analysis was unplanned", "Advanced and localized states differ", "Losartan dose may matter", "Localized randomized result is not posted" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Pressure reduction and vessel opening are physically coherent and measurable." }, "animalEvidence": { "rating": "strong", "reason": "Models support decompression but did not predict the broad advanced result." }, "inVitroEvidence": { "rating": "weak", "reason": "Cell culture cannot represent tissue pressure and blood flow." }, "reproducibility": { "rating": "moderate", "reason": "The broad uncontrolled signal failed a randomized advanced test; localized contribution awaits an existing trial." }, "commercialConflicts": { "rating": "low", "reason": "Losartan is generic and investigators reported the negative randomized stop." }, "recency": { "rating": "current", "reason": "AFPAC reported in 2025 and the localized registry was checked in 2026." } } }, "layout": { "x": 328, "y": 278 }, "updated": "2026-09-15" }, { "id": "failure-men1-mnet-assay-no-progression-prediction", "title": "The MEN1 circulating mRNA assay detected a tumour label but not future danger", "shortTitle": "MEN1 mRNA prediction failure", "type": "failure", "status": "supported", "scope": [ "pnet" ], "summary": "A recalibrated circulating gene-expression score separated some current disease groups but failed every tested development, growth and metastasis prediction.", "content": "Original belief: a circulating NET gene-expression assay could reduce repeated imaging or identify which MEN1-associated PanNET would grow or spread. Observed result: in a single-centre prospective cohort of 110 eligible MEN1 patients, 60% had PanNET at baseline and 9% developed one during follow-up. The assay distinguished people without manifestations from those with only PanNET at baseline, but future-action performance was poor. AUC was 0.65 for new PanNET, 0.49 for growth, 0.54 for increasing tumour count, 0.39 for new metastatic disease and 0.46 for progression of metastases. None supports a surgery, drug or imaging-interval threshold. Likely causes include a germline condition affecting circulating RNA across multiple organs, common microadenomas, multiple simultaneous tumours, small event counts and a blood signal that reflects current NET burden rather than the dangerous clone. The mechanism survives only as a current-state measurement to be retested against a simple clinical model. Recovery requires multicentre prospective samples linked lesion by lesion to imaging, resection pathology and metastasis, with one locked endpoint and an external site. Stop the assay route if it cannot improve calibration or net benefit over age, largest-lesion size, growth and duct change.", "sourceIds": [ "van-vliembergen-2026-men1-mnet-assay", "klein-haneveld-2021-men1-young-screening" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "claim-men1-surveillance-finds-lesions-not-danger", "relation": "failed-because-of" }, { "target": "unknown-small-pnet-surgery-selector", "relation": "supports" }, { "target": "hypothesis-pnet-time-updated-surgery-rule", "relation": "constrains" } ], "evidence": { "humanProspective": "strong — prospective cohort with longitudinal outcomes", "humanRetrospective": "not-applicable", "randomised": "none", "independentReplication": "none for the MEN1 recalibrated assay", "sampleSize": "110 eligible MEN1 patients; 60% with PanNET at baseline and 9% developing PanNET during follow-up", "effect": "Prediction AUCs ranged from 0.39 to 0.65 and did not support a future-risk action", "limits": [ "Single centre", "Small event counts", "One recalibrated assay", "Follow-up 12-48 months", "No direct comparison of an assay-guided care pathway" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "A circulating tumour signal is plausible, but germline MEN1 and multiple lesions weaken lesion-specific prediction." }, "animalEvidence": { "rating": "not-applicable", "reason": "Animal evidence is not relevant to the reported human prediction failure." }, "inVitroEvidence": { "rating": "moderate", "reason": "The transcript panel is technically measurable, but technical detection did not yield clinical prediction." }, "reproducibility": { "rating": "none", "reason": "No independent MEN1 cohort has reproduced a useful future-risk result for this assay." }, "commercialConflicts": { "rating": "moderate", "reason": "Gene-expression assays create a commercial incentive for repeated testing, requiring independent thresholds and code." }, "recency": { "rating": "current", "reason": "The negative longitudinal report was published in 2026." } } }, "updated": "2026-09-14" }, { "id": "failure-mesothelin-car-t-pdac", "title": "Mesothelin CAR T cells reached PDAC metastases but did not expand or control disease", "shortTitle": "Mesothelin CAR T failure", "type": "failure", "status": "refuted", "scope": [ "pdac" ], "summary": "Nine phase 1 patients had no objective responses, median PFS 3.9 weeks and median OS 6.7 weeks; local delivery did not solve poor expansion and exhaustion.", "content": "Original belief: mesothelin expression offered a surface target, and regional intraperitoneal or hepatic-arterial delivery might overcome poor trafficking. What happened: nine treated patients received intravenous or regional fully human mesothelin CAR T cells. One had stable disease at day 28 and progressed by month three; all regional-delivery patients had progressive disease at their first scan. Median progression-free survival was 3.9 weeks and overall survival 6.7 weeks. Cells could be detected in blood, fluid or metastases, but expanded poorly and rapidly lost persistence. Tumour samples showed low T-cell entry and exhaustion states including ID3 and SOX4. Why it probably failed: target expression was heterogeneous; trafficking, expansion and persistence were inadequate; chronic antigen and the local immune environment disabled the cells. Double knockout of ID3 and SOX4 prolonged control in mice, not patients. The clinical result invalidates this construct and delivery scheme. It does not invalidate all cell therapy, but the next version must show human tumour entry, expansion, multi-target coverage and maintained function before an efficacy trial. Reporting boundary: the registry now says 54 people enrolled before administrative termination, while the paper accounts for nine infused. Recover the other 45 routes and outcomes before treating the paper as a complete trial report.", "sourceIds": [ "aznar-2025-mesothelin-car-t-resistance", "clinicaltrials-nct03323944-meso-cart" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "claim-immune-exclusion-is-multicellular", "relation": "supports" }, { "target": "claim-kras-tcr-one-response-not-platform-proof", "relation": "contradicts" }, { "target": "experiments", "relation": "resolves" }, { "target": "failure-cell-therapy-route-and-denominator-pdac", "relation": "extended-by" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "not-applicable", "randomised": "none", "independentReplication": "moderate", "sampleSize": "Nine infused patients", "effect": "No objective responses; median PFS 3.9 weeks; median OS 6.7 weeks", "limits": [ "Phase 1", "Very small", "Heavily treated advanced disease", "All participants were White", "Multiple delivery routes", "Mouse knockout rescue not human evidence", "Construct-specific failure" ] }, "layout": { "x": 14, "y": 100 }, "updated": "2026-09-15" }, { "id": "failure-metabolic-repurposing-pdac", "title": "Metabolism and autophagy signals repeatedly failed when clinical contribution was tested", "shortTitle": "Metabolic translation failures", "type": "failure", "status": "refuted", "scope": [ "pdac" ], "summary": "Randomized PDAC studies rejected metformin, devimistat and hydroxychloroquine combinations; early response, blood-cell markers and model synergy did not predict useful patient control.", "content": "Four translation failures now point to one engineering defect: the programme accepted an output signal before proving that the intended dependency was blocked in the patient's tumour. Metformin moved from population associations and metabolic models into a 121-person double-blind trial. Median overall survival was 6.8 months with metformin and 7.6 with placebo, HR 1.056, P=.78. Hydroxychloroquine, a blunt inhibitor of the lysosome recycling compartment, moved from autophagy biology into a 112-person randomized trial. It raised response from 21.1% to 38.2% but did not improve one-year or median survival: 41% versus 49% and 11.1 versus 12.1 months. Devimistat then supplied the clearest scale-up failure. Its uncontrolled phase 1 reported 11 responses among 18 people at the selected dose, 61%. AVENGER 500 randomized 528 people and found median survival 11.10 months with devimistat plus modified FOLFIRINOX versus 11.73 with standard FOLFIRINOX, HR .95, P=.655; progression-free survival was also unchanged. Finally, a direct randomized contribution test gave the ERK inhibitor LY3214996 with hydroxychloroquine to 20 people and the ERK inhibitor alone to 19. Four-month disease control was 1/20 versus 1/19, while the combination required half the ERK-inhibitor dose. Eleven derived organoids showed no clear synergy, and blood-cell markers did not consistently show autophagy inhibition. Separate ERK or MEK combinations found 0/18 pancreatic responses or stopped for toxicity and lack of efficacy. The shared failure is not proof that metabolism or autophagy are irrelevant. It shows that response without a control, target measurement outside the tumour, and model synergy without reproduced patient pharmacology are unsafe escalation rules. A more selective molecule starts again at human tumour target engagement; potency claims do not inherit credit from the failed drugs.", "contentSections": [ "A 36-person paricalcitol run-in adds a useful measurement warning. The vitamin-D-receptor drug changed fibroblast signals and brought CD8 T cells closer to tumour cells in paired human biopsies, as models predicted. The study was built for safety and tissue pharmacology, not clinical efficacy; five of 12 people receiving oral paricalcitol had grade 2-4 high blood calcium requiring dose reduction. A measured microenvironment change is an intermediate result, not patient benefit." ], "sourceIds": [ "kordes-2015-metformin-pdac", "karasic-2019-hydroxychloroquine-pdac", "alistar-2017-devimistat-phase1", "philip-2024-avenger500", "surana-2025-erk-hcq-pdac", "clinicaltrials-nct05221320-ulixertinib-hcq", "manji-2023-mekiauto", "clinicaltrials-nct04214418-mekiauto", "clinicaltrials-nct04892017-inlexisertib", "deciphera-2022-dcc3116-phase1", "clinicaltrials-nct04524702-paricalcitol-hcq", "perez-2026-paricalcitol-pdac" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "treatment", "relation": "part-of" }, { "target": "claim-pdac-metabolic-adaptation-is-redundant", "relation": "supports" }, { "target": "claim-hcq-combinations-fail-pdac-not-autophagy", "relation": "supports" }, { "target": "unknown-selective-autophagy-pdac-contribution", "relation": "leaves-open" }, { "target": "trial-pdac-autophagy-action-map", "relation": "informs" }, { "target": "hypothesis-tumour-verified-selective-autophagy-blockade", "relation": "constrains" }, { "target": "investment", "relation": "warns" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "moderate", "randomised": "strong", "independentReplication": "moderate", "sampleSize": "Metformin n=121; perioperative hydroxychloroquine n=112; devimistat phase 1 n=20 and phase 3 n=528; randomized ERK plus hydroxychloroquine contribution n=39 after a 13-person safety run-in; paricalcitol tissue run-in n=36", "effect": "No survival gain from metformin, hydroxychloroquine or devimistat; four-month disease control 5.0% with ERK plus hydroxychloroquine versus 5.3% with ERK alone", "limits": [ "Each result is agent-, dose-, disease-state- and backbone-specific", "The ERK study was small and reduced ERK dose in the combination", "No reliable clinical measure established the magnitude of tumour autophagy blockade", "Paricalcitol measured tissue change but was not powered for clinical benefit", "Does not refute every metabolic or autophagy target", "Several studies were open-label" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "PDAC nutrient use and autophagy remain biologically plausible, but the tested interventions did not establish a patient tumour dependency." }, "animalEvidence": { "rating": "moderate", "reason": "Model rationale repeatedly failed to predict the tested human contribution." }, "inVitroEvidence": { "rating": "weak", "reason": "Earlier model synergy contrasts with absent synergy in eleven trial-derived organoids." }, "reproducibility": { "rating": "strong", "reason": "Negative clinical direction repeats across randomized metabolism and HCQ studies and separate ERK or MEK programmes." }, "commercialConflicts": { "rating": "moderate", "reason": "Several development programmes were sponsor-backed, while the strongest negative results were reported in full or posted registries." }, "recency": { "rating": "current", "reason": "Includes the 2025 randomized report and 2026 direct-inhibitor termination." } } }, "layout": { "x": 129, "y": 78 }, "updated": "2026-09-15" }, { "id": "failure-microbiome-proxy-without-patient-benefit", "title": "PDAC microbiome translation repeatedly loses the comparator or stops at a marker", "shortTitle": "Microbiome proxy and delivery failure", "type": "failure", "status": "observed", "scope": [ "pdac", "transfer-unproven" ], "summary": "A long chain from interesting microbes to patients has broken through weak endpoints, single arms and sponsor failure.", "content": "The failure is not that microbes are irrelevant. It is that the human test chain keeps dropping the part that can prove benefit. The randomized 1-kestose pilot enrolled 40, completed 38 and had no power calculation. It reported significant changes inside an intervention arm, although the causal quantity is difference-in-change: (after minus before with kestose) minus (after minus before without kestose). Its 70% versus 61% response and 95% versus 83% disease-control differences were not significant, and survival was not tested. The first antibiotic-plus-pembrolizumab study enrolled 0 before withdrawal. The MRx0518 live-bacterium plus radiation study stopped at 13 because its sponsor became insolvent and posted no result. The MD Anderson stool-transfer study remains suspended at investigator request with an estimated 10, not an actual 10, and no result. The two current US studies total only 45 planned people and are single arm: 25 receive chemotherapy, two antibiotics, pembrolizumab and surgery; 20 receive one stool transfer plus chemotherapy. Neither can isolate the microbial component. Observational antibiotic signals also conflict with harm: cohorts associate antibiotics with longer progression or survival, but a 430-person gemcitabine trial reanalysis associates exposure with 1.77 times the rate of severe adverse events, including blood-count and gut toxicity. Infection, drug exposure and patient state confound all of these non-randomized comparisons. Root causes: model effects were promoted before a human dose-to-tumour assay existed; marker endpoints replaced patient outcomes; multiple treatments made attribution impossible; small sponsors could not finish; and antibiotic resistance was treated as a side issue. Reusable lesson: keep assigned controls, count every enrolled person, quantify the target microbe or enzyme in tumour tissue, measure active drug exposure, record resistant organisms and make survival, function and serious harm the release gate.", "sourceIds": [ "nakaoka-2024-kestose-pdac", "jrct-kestose-pdac", "clinicaltrials-nct03891979-microbiome-pdac", "clinicaltrials-nct04193904-mrx0518-pdac", "clinicaltrials-nct04975217-fmt-pdac", "clinicaltrials-nct05462496-antibiotic-pembro-pdac", "clinicaltrials-nct06393400-chase-pdac", "nakano-2020-antibiotics-gnp", "gong-2025-antibiotics-adjuvant-pdac", "corty-2020-antibiotic-gemcitabine-toxicity" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "claim-pdac-microbiome-mechanism-without-human-benefit", "relation": "limits" }, { "target": "unknown-microbiome-treatment-contribution-pdac", "relation": "creates" }, { "target": "hypothesis-bacterial-gemcitabine-deactivation-gate", "relation": "informs" }, { "target": "trial-pdac-microbiome-action-map", "relation": "mapped-by" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "strong", "randomised": "weak", "independentReplication": "weak", "sampleSize": "Kestose n=40; failed delivery n=0, 13 and estimated 10; current single arms planned n=45; toxicity reanalysis n=430", "effect": "No proven microbial-treatment contribution; severe-adverse-event association HR 1.77 during antibiotics", "limits": [ "Trial registries may lag", "The adverse-event analysis did not randomize antibiotics", "No patient-level microbial and drug-exposure record was available", "Suspension reason is not explained beyond investigator request" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The failure concerns testing and delivery, not the plausibility of microbial drug metabolism." }, "animalEvidence": { "rating": "strong", "reason": "Model activity makes the weak human test chain more consequential." }, "inVitroEvidence": { "rating": "strong", "reason": "The enzyme and inactive drug product can be directly measured." }, "reproducibility": { "rating": "weak", "reason": "No controlled patient-benefit signal has repeated." }, "commercialConflicts": { "rating": "high", "reason": "One trial ended with sponsor insolvency; several interventions depend on a product owner or donor-manufacturing chain." }, "recency": { "rating": "current", "reason": "Registry states were checked on 15 September 2026." } } }, "layout": { "x": 362, "y": 318 }, "updated": "2026-09-15" }, { "id": "failure-mistletoe-pdac-no-quality-benefit", "title": "Mistletoe extract did not improve quality of life in a blinded pancreatic trial", "shortTitle": "Mistletoe negative trial", "type": "failure", "status": "refuted", "scope": [ "pdac", "supportive-care" ], "summary": "A nine-centre double-blind placebo trial in 290 people found no quality-of-life, weight, care-use or biomarker benefit from mistletoe extract.", "content": "Mistletoe extract is used as a complementary cancer treatment, often with a quality-of-life claim. MISTRAL tested that claim in 290 people with advanced pancreatic cancer at nine oncology centres. Participants received usual cancer treatment or best supportive care plus mistletoe extract or placebo. The double-blind phase 3 comparison found no significant improvement in quality of life or weight, no measured reduction in hospital, nutrition, home-palliative or symptom-medicine use, and no useful biomarker change. The failure matters because supportive care should be judged with the same controlled standard as anticancer treatment. A popular or natural product does not receive a weaker proof rule. Recovery means redirecting time and money to support routes with measurable nutrition, function, symptom or treatment-delivery effects and publishing negative results.", "sourceIds": [ "wode-2026-mistral" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "treatment", "relation": "part-of" }, { "target": "claim-pdac-palliative-care-improves-qol-not-survival", "relation": "constrains" }, { "target": "unknown-cachexia-treatment-utility", "relation": "informs" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong", "independentReplication": "moderate — the tested claim failed in a large blinded trial", "sampleSize": "n=290 at nine centres", "effect": "No significant difference in quality of life, weight, measured care use or biomarkers", "limits": [ "Advanced disease only", "Tested one extract schedule", "Does not assess every possible supportive-care intervention", "Survival was not the main reported question" ], "dimensions": { "biologicalPlausibility": { "rating": "weak", "reason": "The trial found no clinical or biomarker effect supporting the proposed route." }, "animalEvidence": { "rating": "not-applicable", "reason": "The quality-of-life claim was tested directly in people." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Laboratory activity cannot replace patient-reported benefit." }, "reproducibility": { "rating": "moderate", "reason": "The multicentre blinded result is internally strong, though it is one programme." }, "commercialConflicts": { "rating": "low", "reason": "The authors reported no relevant competing interests." }, "recency": { "rating": "current", "reason": "Published and checked in 2026." } } }, "updated": "2026-09-15" }, { "id": "failure-mwa-selection-bundling-and-missing-imaging", "title": "Microwave studies select easier tumours, bundle treatments and lose outcome data", "shortTitle": "MWA evidence-system failure", "type": "failure", "status": "supported", "scope": [ "pdac", "trial-design", "local-treatment", "data-custody" ], "summary": "The apparent survival advantage appears in a group chosen for lower burden, while nearly 30% lack imaging; prospective work changes four treatments together and includes a fatal fistula.", "content": "In the 150-person retrospective comparison, the tumour board generally offered microwave only to lower-burden cases. Mean tumour diameter was 3.4 versus 4.9 centimetres, a 1.5 / 4.9 = 30.6% smaller starting tumour. Metastatic disease was present in 34.3% versus 54.2%, a 19.9-point advantage. Those two variables strongly predict survival before treatment begins. Imaging existed for 47/67 microwave patients, 70.1%, versus all 83 controls. Among those with scans, distant progression was 61.7% versus 33.7% and local progression 10.6% versus 9.6%. The reported 14.6-versus-9.0-month survival gap therefore cannot be assigned to microwave energy. In the 12-person prospective trial, every participant received dual immune drugs, a direct tumour biopsy and gemcitabine; only eight also completed surgical microwave treatment after four hidden metastases were found. One of those eight died from bleeding shock after a pancreatic fistula, a leak of digestive fluid, following biopsy and ablation. The trial stopped below target for slow recruitment and efficacy concern. NCT06231160 planned 185 non-random sampled participants but has had unknown status since 2024. The repeated failure is adding multiple moving parts while leaving selection and missing records uncontrolled.", "sourceIds": [ "vogl-2026-tace-mwa", "zhou-2024-pancreatic-mwa", "topal-2025-mwa-dual-ici", "clinicaltrials-nct06231160-mwa-cohort" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "claim-mwa-pain-signal-no-added-benefit", "relation": "limits" }, { "target": "unknown-mwa-safe-coverage-and-patient-benefit", "relation": "creates" }, { "target": "hypothesis-mwa-existing-data-gate", "relation": "repaired-by" }, { "target": "trial-pdac-mwa-action-map", "relation": "mapped-by" }, { "target": "failure-rfa-selection-and-route-attrition", "relation": "supports" }, { "target": "failure-cryoablation-selection-attrition-and-route-harm", "relation": "compares-with" } ], "evidence": { "humanProspective": "weak", "humanRetrospective": "moderate", "randomised": "none", "independentReplication": "weak", "sampleSize": "Retrospective n=150 and n=72; prospective n=12; dormant observational target n=185", "effect": "Large measured baseline imbalance, 29.9% imaging missing in one treated group and one fatal fistula pathway", "limits": [ "The 150-person paper explicitly disclaims causal comparison", "Overall survival may have more complete follow-up than imaging", "Death cannot be attributed to microwave alone", "Historic care spans 13 years", "TACE is itself unproved", "The dormant cohort may have unposted enrollment" ], "dimensions": { "biologicalPlausibility": { "rating": "not-applicable", "reason": "This record concerns allocation, bundling and missing outcomes." }, "animalEvidence": { "rating": "not-applicable", "reason": "Animal heating cannot repair human selection bias." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Bench ablation does not identify the causal clinical effect." }, "reproducibility": { "rating": "weak", "reason": "The available estimates do not reproduce under comparable entry and care." }, "commercialConflicts": { "rating": "moderate", "reason": "Device logs and specifications still require supplier-independent custody." }, "recency": { "rating": "current", "reason": "The failure analysis includes the 2025 prospective and 2026 retrospective reports." } } }, "layout": { "x": 880, "y": 804 }, "updated": "2026-09-15" }, { "id": "failure-napabucasin-biomarker-did-not-rescue", "title": "Napabucasin failed phase 3, including in its pSTAT3-positive subgroup", "shortTitle": "Napabucasin failure", "type": "failure", "status": "supported", "scope": [ "pdac" ], "summary": "Adding napabucasin to gemcitabine and nab-paclitaxel produced median survival of 11.4 versus 11.7 months and failed in the proposed biomarker-positive group.", "content": "Napabucasin was developed around a broad cancer-stemness and STAT3 narrative and later described as an NQO1-activated reactive-oxygen generator. CanStem111P randomised 1,134 untreated metastatic PDAC patients. Median overall survival was 11.4 months with napabucasin and 11.7 months with control; hazard ratio 1.07. The trial stopped for futility. The exploratory pSTAT3-positive subgroup also showed no survival difference, and severe diarrhoea and abdominal pain were more frequent. The result rejects clinical efficacy for this regimen and rejects pSTAT3 positivity as a rescue marker in this trial. It does not establish that every STAT3 or redox intervention is invalid, but future programmes need a much tighter target-mechanism assay and proof that the drug changes that mechanism in human tumour tissue.", "sourceIds": [ "bekaii-saab-2023-napabucasin" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "treatment", "relation": "part-of" }, { "target": "unknowns", "relation": "supports" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong", "independentReplication": "moderate", "sampleSize": "1,134 randomised; 382 pSTAT3-positive patients across groups", "effect": "Median OS 11.4 versus 11.7 months; HR 1.07; no benefit in pSTAT3-positive subgroup", "limits": [ "Open-label", "Metastatic first-line setting", "Exploratory biomarker analysis", "Does not test every STAT3 or NQO1 strategy", "Company-funded" ] }, "layout": { "x": 23, "y": 86 }, "updated": "2026-09-14" }, { "id": "failure-nct03762837-risk-cohort-never-started", "title": "The planned 100,000-person gallbladder-risk cohort never established a live study record", "shortTitle": "100,000-person cohort stalled", "type": "failure", "status": "supported", "scope": [ "gallbladder" ], "summary": "NCT03762837 still says not yet recruiting seven years after its estimated start and retains a primary-completion date that passed in 2022.", "content": "NCT03762837 proposed the study the field most needs: 50,000 people with benign gallbladder disease and 50,000 healthy comparators, followed prospectively for gallbladder-cancer incidence. Its registry still says not yet recruiting, with an estimated start of 1 April 2019, estimated primary completion of 31 December 2022 and no results. Status was last verified in November 2018 and the record last updated in March 2019. On 14 September 2026, enrolment was therefore 0% of the planned 100,000 on the public record and the primary date was more than three years past. The reason is not public, so this is evidence of programme and reporting failure, not evidence that the cohort question is infeasible or scientifically wrong. Recovery should begin with contact and protocol-status verification. If the study did not start, preserve its intended endpoint but replace a single 100,000-person launch with linked feasibility stages inside active high-incidence and routine-surgery networks. No new full cohort should open until a 2,000-person stage reaches at least 85% complete baseline data, at least 90% outcome linkage and a verified event rate within a prespecified range.", "sourceIds": [ "clinicaltrials-nct03762837-gbc-risk-cohort" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "trial-gallbladder-prevention-detection-action-map", "relation": "failed-because-of" }, { "target": "unknown-gallbladder-prevention-selector", "relation": "supports" }, { "target": "hypothesis-gallbladder-two-stage-interception-rule", "relation": "constrains" } ], "evidence": { "humanProspective": "none — no public recruitment", "humanRetrospective": "not-applicable", "randomised": "none", "independentReplication": "none", "sampleSize": "Planned n=100,000; public record remains not yet recruiting", "effect": "Public recruitment reached 0% of target and the estimated 2022 primary-completion date passed without results", "limits": [ "Registry may be abandoned rather than maintained", "Unregistered recruitment cannot be excluded", "No public reason for non-start", "No contact outcome has been obtained" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "A prospective cohort can measure absolute risk, but the proposed risk groups and three-year horizon may still be weak." }, "animalEvidence": { "rating": "not-applicable", "reason": "The failure concerns execution of a human cohort." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "No assay result was reported." }, "reproducibility": { "rating": "none", "reason": "The study has no public participant or outcome data." }, "commercialConflicts": { "rating": "low", "reason": "The registry names an academic sponsor and no intervention product." }, "recency": { "rating": "historical", "reason": "The record was last verified in 2018 and updated in 2019." } } }, "layout": { "x": 27, "y": 113 }, "updated": "2026-09-14" }, { "id": "failure-nhs-galleri-primary-stage-shift", "title": "NHS-Galleri missed its prespecified late-stage cancer endpoint despite fewer stage IV diagnoses", "shortTitle": "NHS-Galleri main miss", "type": "failure", "status": "contested", "scope": [ "pan-cancer", "pdac", "hepatobiliary", "colorectal" ], "summary": "In about 142,250 randomized adults, three annual Galleri tests did not reduce combined stage III/IV diagnoses: incidence-rate ratio 1.03, 95% CI 0.92-1.14.", "content": "NHS-Galleri is the largest randomized test of a multicancer blood screen. About 142,250 adults aged 50–77 received three annual Galleri tests plus usual screening or usual screening alone. The prespecified main endpoint was combined stage III and IV diagnoses across 12 high-mortality cancers within one year of the final appointment. It was not met: incidence-rate ratio 1.03, 95% CI 0.92–1.14, P = 0.6324, meaning the point estimate was 3% higher rather than lower. A prespecified secondary analysis found 14% fewer stage IV cancers overall, incidence-rate ratio 0.86, 95% CI 0.744–0.998, described as nominally significant. Stage IV reductions grew from 9% in round one to 22% and 26% in rounds two and three, but the round-two confidence interval included no effect and multiple secondary comparisons weaken the certainty. More stage III disease may partly represent cancers moved back from stage IV, but that explanation is not directly identifiable from aggregate stage counts. The trial has not shown fewer cancer deaths, and the public result does not establish a pancreatic-specific effect. The correct reading is neither “the test failed completely” nor “stage shift is proven”: the main endpoint failed, a biologically plausible stage IV signal remains, and mortality plus longer follow-up must decide whether that signal benefits patients.", "sourceIds": [ "sasieni-2026-nhs-galleri", "grail-2026-q2-10q" ], "links": [ { "target": "claim-mced-workflow-feasible-utility-unproved", "relation": "extends" }, { "target": "claim-mced-early-stage-sensitivity-limit", "relation": "depends-on" }, { "target": "claim-stage-shift-insufficient", "relation": "supports" }, { "target": "company-grail", "relation": "involves" }, { "target": "unknowns", "relation": "supports" }, { "target": "failures", "relation": "part-of" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong", "independentReplication": "none", "sampleSize": "Approximately 142,250 randomized adults aged 50-77 in England", "effect": "Primary stage III/IV IRR 1.03, 95% CI 0.92-1.14, P=0.6324; secondary stage IV IRR 0.86, 95% CI 0.744-0.998", "limits": [ "Conference report rather than full paper", "Primary endpoint missed", "Secondary result nominal", "Multiple comparisons", "No mortality result", "Cancer types pooled", "No pancreatic-specific effect", "One-year follow-up after final screen", "Sponsor designed test and coauthored study" ] }, "layout": { "x": 14, "y": 124 }, "updated": "2026-09-14" }, { "id": "failure-norpact1-neoadjuvant-delivery", "title": "NORPACT-1: a sound treatment-first idea failed in delivery and did not improve survival", "shortTitle": "NORPACT-1 delivery failure", "type": "failure", "status": "supported", "scope": [ "pdac" ], "summary": "Only 61 of 77 assigned patients received neoadjuvant FOLFIRINOX, severe toxicity rose and survival favoured upfront surgery in a small Nordic phase 2 trial.", "content": "Original belief: four cycles of FOLFIRINOX before surgery would deliver systemic treatment early and improve survival in resectable pancreatic-head PDAC. What happened: NORPACT-1 randomized 77 patients to treatment first and 63 to surgery first. Only 61/77, 79%, received neoadjuvant therapy. Seventeen, 22%, were excluded from that arm's per-protocol analysis, including ten who did not receive the assigned treatment. Grade 3 or worse adverse events affected 42/73, 58%, in the treatment-first safety set versus 19/47, 40%, after surgery first. Resection occurred in 63/77, 82%, versus 56/63, 89%. At 18 months, 60% versus 73% were alive; median overall survival was 25.1 versus 38.5 months, HR 1.52, 95% CI 1.00-2.33. The study changed permitted postoperative chemotherapy during recruitment and was too small for a stable subgroup rule. Why it failed: assigning a treatment sequence did not ensure that patients received it, and the extra early toxicity did not buy a survival gain. Recovery rule: keep intention-to-treat as the main estimate; count every allocated patient, dose, delay, progression, operation and postoperative cycle; and do not explain away failure with a per-protocol responder subset. NORPACT-1 does not prove that every treatment-first regimen harms. It proves that four-cycle Nordic FOLFIRINOX under this delivery route cannot be assumed superior.", "sourceIds": [ "labori-2024-norpact1" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "claim-neoadjuvant-therapy-improves-selection-not-cure", "relation": "contradicts" }, { "target": "unknown-pdac-perioperative-treatment-route", "relation": "leads-to" }, { "target": "trial-pdac-perioperative-sequencing-action-map", "relation": "informs" }, { "target": "hypothesis-pdac-surgical-route-completion", "relation": "constrains" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong but small phase 2", "independentReplication": "weak — trials disagree", "sampleSize": "Randomized n=140; neoadjuvant n=77; upfront surgery n=63", "effect": "18-month survival 60% versus 73%; OS HR 1.52, 95% CI 1.00-2.33", "limits": [ "Small phase 2 trial", "Pancreatic-head tumours only", "WHO performance status 0-1", "Treatment delivery failure", "Adjuvant regimen amended during recruitment", "No universal harm conclusion" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "Early systemic treatment could suppress occult disease, but added toxicity and delay can remove benefit." }, "animalEvidence": { "rating": "not-applicable", "reason": "Human treatment delivery and survival decide this failure." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Cell systems cannot model the complete treatment and surgery route." }, "reproducibility": { "rating": "weak", "reason": "PREOPANC and CISPD-1 point toward benefit, while NORPACT-1 does not." }, "commercialConflicts": { "rating": "low", "reason": "Public and charitable funders supported the trial; one author reported unrelated company honoraria." }, "recency": { "rating": "current", "reason": "The randomized result was published in 2024 and remains directly relevant to modern sequence delivery." } } }, "updated": "2026-09-14" }, { "id": "failure-nrg1-testing-and-confirmation-route", "title": "An approved rare-driver drug fails if RNA testing and confirmation do not reach the patient", "shortTitle": "NRG1 route failure", "type": "failure", "status": "open", "scope": [ "pdac", "nrg1", "rna-testing", "regulatory", "care-delivery", "capital-allocation" ], "summary": "NRG1 is rare, DNA panels miss difficult fusions, England does not name NRG1 in its pancreatic test line, and the FDA follow-up deadline passed without a public result.", "content": "Failure 1, denominator loss. If 1% of PDAC carries NRG1, 100 unselected tumours must traverse sampling, testing, interpretation, referral and drug access to find one candidate. Failure 2, wrong data type. NRG1 has many possible partners and large breakpoint regions; in one 48-person series, RNA found 34 cases, DNA 11 and both only three. Failure 3, policy coverage. England's July 2026 M219.2 pancreatic route requires an RNA structural-variant panel for NTRK1/2/3 and permits additional reportable targets, but does not name NRG1. That is a coverage question, not proof that every laboratory omits it. Failure 4, result timing. A fusion result that arrives after fitness loss or death has no treatment value. Failure 5, approval overstatement. Response in a single arm can justify access for a rare disease but cannot show longer survival. FDA required at least 50 evaluable pancreatic patients with response follow-up and a final report by August 2026. On 15 September the FDA still listed the indication as ongoing, and the trial registry remained stale without posted results. Failure 6, business opacity. Genmab owns the acquired asset, Partner Therapeutics sells it in the US and Genmab receives royalties, but pancreatic sales and route completion are not separated. Buying the owner does not repair any of these care losses.", "sourceIds": [ "schram-2026-nrg1-fusion-detection", "philip-2022-kras-wildtype-pdac", "nhs-england-2026-cancer-genomic-test-directory", "fda-2024-bizengri-multidisciplinary-review", "fda-2026-ongoing-bizengri-accelerated-approval", "clinicaltrials-nct02912949-enrgy", "genmab-2026-q1-bizengri" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "claim-nrg1-approved-response-not-durable-control", "relation": "limits" }, { "target": "unknown-nrg1-identification-and-survival-benefit", "relation": "creates" }, { "target": "hypothesis-nrg1-rna-route-completion", "relation": "repaired-by" }, { "target": "company-genmab-partner-bizengri", "relation": "involves" } ], "evidence": { "humanProspective": "moderate for drug response; none for complete English route", "humanRetrospective": "moderate for test detection", "randomised": "none", "independentReplication": "weak", "sampleSize": "Drug n=30 FDA set; detection series n=48; molecular cohort n=2,483; current NHS policy workbook", "effect": "A roughly one-in-100 target can be lost at testing, reporting, referral or access before any drug decision", "limits": [ "No national NRG1 completion denominator", "NHS laboratories may report unlisted additional targets", "US approval does not equal UK access", "Registry may lag sponsor data", "No randomized survival estimate" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "No correctly targeted treatment can occur before the fusion is found." }, "animalEvidence": { "rating": "not-applicable", "reason": "This is a testing, regulation and delivery failure." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "The drug target is established; route completion is human operations." }, "reproducibility": { "rating": "moderate", "reason": "Multiple cohorts find targetable fusions enriched in KRAS-wild-type PDAC." }, "commercialConflicts": { "rating": "high", "reason": "Drug, pivotal evidence and US commercial rights are connected." }, "recency": { "rating": "current", "reason": "Policy and FDA status were checked in September 2026." } } }, "layout": { "x": 1216, "y": 1140 }, "updated": "2026-09-15" }, { "id": "failure-oncolytic-virus-pelareorep", "title": "Oncolytic reovirus changed immune markers but did not improve PDAC progression", "shortTitle": "Oncolytic virus failure", "type": "failure", "status": "refuted", "scope": [ "pdac" ], "summary": "Pelareorep plus chemotherapy produced PFS 4.9 months versus 5.2 months in a 73-patient randomized phase 2 despite measurable immune changes.", "content": "Original belief: an intravenously delivered reovirus would exploit activated RAS, lyse PDAC cells and inflame an immune-cold tumour. What happened: 73 evaluable untreated metastatic patients received carboplatin/paclitaxel with or without pelareorep; median PFS was 4.9 versus 5.2 months, P=.6, and KRAS status did not select benefit. Cytokines and immune-cell subsets changed, showing biological activity without a clinical effect. Why it probably failed: systemic virus delivery and intratumour spread were inadequate, the chemotherapy backbone may have shaped immunity, and virus-driven inflammation included suppressive as well as activating signals. What survives: direct tumour pharmacology, replication and immune-state measurement can determine whether a newer virus actually reaches and changes human lesions. VCN-01 now clears part of that gate: intravenous viral genomes were found in five of six day-eight tumour biopsies, and its 96-person VIRAGE FAS favoured virus plus chemotherapy for survival. But the registry records 112 enrolled, the public analysis sets exclude people after allocation, the primary alpha statements conflict and no independent trial confirms benefit. The pelareorep result rejects that virus-regimen pair, not every engineered virus. Next test: recover the full VCN-01 denominator and independently repeat all-randomized survival before a confirmatory phase 3 receives mission support.", "sourceIds": [ "noonan-2016-pelareorep-pdac", "bazan-peregrino-2021-vcn01-intratumour", "garcia-carbonero-2022-vcn01-intravenous", "clinicaltrials-nct05673811-virage", "theriva-2023-virage-protocol", "theriva-2025-virage-topline", "esmo-2025-virage-2216mo", "ctis-2026-virage2", "theriva-2026-q2-10q" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "treatment", "relation": "part-of" }, { "target": "claim-immune-exclusion-is-multicellular", "relation": "failed-because-of" }, { "target": "research-frontier-ranking", "relation": "warns" }, { "target": "claim-vcn01-randomized-signal-not-survival-proof", "relation": "compared-with" }, { "target": "failure-virage-analysis-and-survivor-selection", "relation": "warns" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong", "independentReplication": "moderate", "sampleSize": "73 evaluable", "effect": "PFS 4.9 vs 5.2 months, P=.6", "limits": [ "Phase 2", "Non-current chemotherapy backbone", "Does not test newer engineered viruses", "Immune correlative findings exploratory" ] }, "layout": { "x": 114, "y": 75 }, "updated": "2026-09-15" }, { "id": "failure-organoid-treatment-chain-attrition", "title": "Organoid accuracy is reported after the route has already lost many patients", "shortTitle": "Organoid denominator loss", "type": "failure", "status": "supported", "scope": [ "pdac", "transfer-unproven" ], "summary": "A high accuracy among successful cultures can coexist with a low chance that an enrolled patient receives a useful decision.", "content": "Original belief: if an organoid predicts response with high sensitivity and specificity, the test is ready to choose treatment. The missing denominator is the whole pipeline: tissue must contain cancer, the correct clone must grow, normal cells must not take over, the result must reproduce, at least two suitable drugs must be available, the answer must arrive before the decision, the person must remain fit, and the route must deliver the choice. These probabilities multiply rather than average. In HOPE, the all-enrolled delivery rate to drug testing was 12/76 = 15.8%, with median initial testing at 96 days. In ORGANOPREDICT, 54/87 = 62.1% produced organoids and 34/87 = 39.1% had complete evaluable follow-up. A paired randomized-order biopsy study generated any organoid in 17/50 = 34%; only 13 of those 17 had malignancy on histology, and the apparent advantage of core biopsy over aspiration was not statistically secure. PASS-01 generated organoids in 50%; only 75/140 = 53.6% of treated patients received any second line, and 33/140 = 23.6% received a choice guided by any molecular correlate. The 5.4 versus 4.4-month survival after correlate-guided versus standard second-line choice was not significant, P=.45, and was not an organoid randomization. AVATAR shows the system-level failure mode: only 4/81 = 4.9% assigned to its model-and-sequencing route received matched treatment, and assigned-group survival was unchanged. Recovery requires all-patient accounting, a decision deadline, tumour identity checks, split-sample reproduction, current drug access and randomized policy utility. Conditional accuracy cannot repair route failure.", "sourceIds": [ "grossman-2022-hope-organoids", "boileve-2024-organopredict", "wiessner-2024-organoid-biopsy-randomized", "knox-2025-pass01", "sarno-2025-avatar" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "failure-avatar-precision-medicine-delivery", "relation": "generalizes" }, { "target": "claim-pdac-organoid-correlation-not-routing-utility", "relation": "limits" }, { "target": "unknown-pdac-organoid-guided-treatment-utility", "relation": "creates" }, { "target": "hypothesis-organoid-guided-early-switch", "relation": "repaired-by" } ], "evidence": { "humanProspective": "strong — consecutive and prospective route denominators are available", "humanRetrospective": "moderate", "randomised": "moderate — biopsy order and chemotherapy were randomized, but guidance usually was not", "independentReplication": "strong — low or slow route completion repeats across countries and platforms", "sampleSize": "HOPE n=76; ORGANOPREDICT n=87; biopsy comparison n=50; PASS-01 n=160 randomized; AVATAR n=125 randomized", "effect": "Observed all-enrolled test or action completion ranges from 4.9% matched treatment in AVATAR to 39.1% evaluable follow-up in ORGANOPREDICT", "limits": [ "Pipelines are not directly interchangeable", "PASS-01 guidance combined several correlates", "Biopsy study tested acquisition order rather than clinical utility", "Newer rapid methods may improve timing", "Published denominators may not expose every loss" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Every lost pipeline step can prevent a correct laboratory result from changing treatment." }, "animalEvidence": { "rating": "not-applicable", "reason": "This is a human service-chain failure rather than an animal mechanism." }, "inVitroEvidence": { "rating": "moderate", "reason": "Culture failure, normal-cell growth and method variation are directly observed in laboratory workflows." }, "reproducibility": { "rating": "strong", "reason": "Attrition and delay recur in independent prospective and randomized clinical programmes." }, "commercialConflicts": { "rating": "moderate", "reason": "Conditional accuracy can favour platform promotion unless original denominators and failures are published." }, "recency": { "rating": "current", "reason": "Includes PASS-01 and live-trial checks through September 2026." } } }, "layout": { "x": 108, "y": 162 }, "updated": "2026-09-15" }, { "id": "failure-p32-liquefaction-survival-and-reporting-gap", "title": "Phosphorus-32 can change the tumour while harming patients or leaving survival unchanged", "shortTitle": "P-32 endpoint and reporting failure", "type": "failure", "status": "supported", "scope": [ "pdac", "trial-design", "brachytherapy", "data-custody" ], "summary": "An older randomized formulation produced dramatic tumour liquefaction with more serious events and no survival gain; the modern programme again leads with local control while its randomized data remain unreported.", "content": "The 2008 randomized study is a warning against treating a changed scan as a changed life. Repeated colloidal phosphorus-32 plus chemotherapy and external radiation liquefied 78% of tumours versus 8% without phosphorus-32. It did not shrink cross-sectional area. Serious adverse events averaged 4.2 per person versus 1.8, P=0.03. Mean survival was 7.4 versus 11.5 months, P=0.16: an uncertain but unfavourable point estimate. Modern silicon microparticles are a different product, schedule and delivery system, so the old result does not prove that OncoSil fails. It proves that local radiation response cannot serve as benefit. The PanCO pilot then reported local control and resection without a comparator, with company funding, company employees and medical writing support. Its 82% all-enrolled local-control rate is more honest than the 90.5% implanted-only rate, because eight people never received the device. The adjusted comparison improves on historical controls but starts at three months, removing early deaths and late implants before weighting. TRIPP-FFX fixes chemotherapy and randomizes the device, but keeps 16-week local control as a primary endpoint and has posted no result after primary completion. Until survival, quality of life, symptoms, chemotherapy delivery, surgery and all harm appear together, the programme remains vulnerable to endpoint promotion.", "sourceIds": [ "rosemurgy-2008-p32-randomized", "ross-2022-panco-p32", "lim-2026-p32-landmark", "clinicaltrials-nct05466799-tripp-ffx" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "claim-oncosil-local-control-signal-awaits-randomized-result", "relation": "limits" }, { "target": "unknown-oncosil-randomized-survival-and-delivery", "relation": "creates" }, { "target": "hypothesis-tripp-ffx-independent-analysis-gate", "relation": "repaired-by" }, { "target": "trial-pdac-p32-action-map", "relation": "mapped-by" }, { "target": "company-oncosil-medical", "relation": "involves" }, { "target": "failure-rfa-selection-and-route-attrition", "relation": "supports" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "moderate", "randomised": "strong for old formulation; modern result unavailable", "independentReplication": "weak", "sampleSize": "Old randomized n=30; modern pilot n=50; adjusted n=86; modern randomized n=88", "effect": "Old formulation: liquefaction +70 points, serious events +2.4 per person, survival point estimate -4.1 months", "limits": [ "Old colloid differs from modern microparticles", "Old sample is small", "Modern trial follow-up continues", "Local control may still have palliative value", "Adjusted analysis handles measured factors only", "Public result timing may follow the planned 2027 completion" ], "dimensions": { "biologicalPlausibility": { "rating": "not-applicable", "reason": "This record tests whether local response maps to patient benefit." }, "animalEvidence": { "rating": "not-applicable", "reason": "The failure is demonstrated in randomized humans." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Radiation killing cannot substitute for survival and harm." }, "reproducibility": { "rating": "weak", "reason": "Modern comparative results are not public." }, "commercialConflicts": { "rating": "high", "reason": "The modern evidence chain is concentrated around OncoSil." }, "recency": { "rating": "current", "reason": "The record includes the 2026 adjusted analysis and trial state." } } }, "layout": { "x": 910, "y": 834 }, "updated": "2026-09-15" }, { "id": "failure-panova4-component-and-historical-control", "title": "PANOVA-4 cannot tell which addition caused disease control", "shortTitle": "PANOVA-4 component failure", "type": "failure", "status": "supported", "scope": [ "pdac", "metastatic", "trial-design" ], "summary": "One single-arm cohort added both tumour-treating fields and atezolizumab, then compared disease control with an old chemotherapy trial; it cannot identify either component's effect.", "content": "PANOVA-4 enrolled 84 people with metastatic pancreatic ductal adenocarcinoma into one treatment arm: tumour-treating fields plus the checkpoint drug atezolizumab plus gemcitabine and nab-paclitaxel. ClinicalTrials.gov has no posted result table. A company release reports 78 analysed people, 74.4% disease control for at least 16 weeks versus 48% in the older MPACT trial, a one-sided P value below 0.001. It also reports a 34.6% objective response rate, 95% CI 24.2-46.2, and median overall survival of 9.7 months, 95% CI 7.9-12.7. The arithmetic exposes the design failure. One observed contrast is asked to estimate at least five terms: field effect + atezolizumab effect + their interaction + differences in patient mix + differences in time and care. One equation cannot solve five unknowns. A small P value against an old number does not repair that identification problem. Disease control also includes stable scans, which can change with scan timing and entry rules and is not durable survival. This trial can establish that the four-part route was deliverable and describe its harms. It cannot show that the device works in metastatic disease, that atezolizumab works, that the two additions interact, or that either improves survival over current treatment. No metastatic label extension, mission adoption or company valuation should use this comparison as component proof.", "sourceIds": [ "clinicaltrials-nct06390059-panova4", "novocure-2026-panova4-topline" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "claim-panova3-ttfields-small-os-benefit", "relation": "does-not-extend" }, { "target": "unknown-ttfields-survival-replication-and-access", "relation": "creates" }, { "target": "hypothesis-optune-pax-independent-value-gate", "relation": "repaired-by" }, { "target": "trial-pdac-ttfields-action-map", "relation": "mapped-by" }, { "target": "company-novocure", "relation": "reported-by" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "not-applicable", "randomised": "none", "independentReplication": "none", "sampleSize": "84 actual enrolled; 78 in company topline analysis; one arm", "effect": "DCR 74.4% versus 48% historical; ORR 34.6%; median OS 9.7 months; no causal component estimate", "limits": [ "Company topline only", "No posted registry results", "Six enrolled participants absent from the topline analysis", "Historical control", "Two experimental additions", "One-sided test", "Disease control is not survival", "Contemporary later treatment may differ" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "Electric fields and checkpoint blockade each have proposed mechanisms, but the cohort cannot test either contribution." }, "animalEvidence": { "rating": "not-applicable", "reason": "The failure is in human comparison design." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Cell experiments cannot identify the cause of a clinical cohort difference." }, "reproducibility": { "rating": "none", "reason": "There is no randomized metastatic comparison and no full public result." }, "commercialConflicts": { "rating": "high", "reason": "Novocure sponsored the study and released the positive topline interpretation." }, "recency": { "rating": "current", "reason": "The registry and March 2026 topline record were checked in September 2026." } } }, "layout": { "x": 592, "y": 516 }, "updated": "2026-09-15" }, { "id": "failure-particle-therapy-comparator-and-accrual", "title": "The particle-therapy programme has not delivered its decisive comparison", "shortTitle": "Particle comparison failure", "type": "failure", "status": "supported", "scope": [ "pdac", "trial-design", "proton", "carbon-ion" ], "summary": "CIPHER, the planned phase 3 carbon-versus-photon trial, enrolled nobody; the remaining studies rely on historical controls and mixed treatment packages.", "content": "CIPHER was meant to randomize locally advanced pancreatic cancer between carbon-ion and photon chemoradiation and test two-year survival. It was withdrawn after zero enrollment because of COVID and other recruitment problems. Even the planned contrast was not a clean particle test. Carbon treatment used 12 fractions with concurrent gemcitabine, while photon treatment used 28 fractions with gemcitabine or capecitabine; later chemotherapy could also vary. One contrast therefore mixed particle type + biological dose model + fraction number + concurrent drug + travel and centre + later treatment and their interactions. The uncontrolled alternatives do not repair this. The 15-person proton phase 2 study stopped early and achieved 47% one-year survival against a 75% target. Its registry remains recruiting toward 60 after starting in 2016 and now estimates primary completion in 2039, with no posted results. PACK completed 25 carbon-ion participants in November 2025 and updated its registry in March 2026, but has no posted results and no control. Comparing their survival with old cohorts cannot remove selection, induction-treatment response, calendar time, centre skill or later-treatment differences. Better beam geometry is real. The programme failure is that it has not connected geometry to net patient benefit in an identifiable comparison.", "sourceIds": [ "clinicaltrials-nct03536182-cipher", "rapp-2022-proton-phase2", "clinicaltrials-nct02598349-proton", "clinicaltrials-nct04194268-pack" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "claim-particle-therapy-local-control-without-survival-proof", "relation": "limits" }, { "target": "unknown-particle-therapy-added-patient-benefit", "relation": "creates" }, { "target": "hypothesis-particle-therapy-existing-data-gate", "relation": "repaired-by" }, { "target": "trial-pdac-particle-therapy-action-map", "relation": "mapped-by" }, { "target": "failure-local-intensification-pdac-survival-not-shown", "relation": "extends" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "moderate", "randomised": "none", "independentReplication": "none", "sampleSize": "CIPHER n=0; proton phase 2 n=15 reported and n=60 registry target; PACK n=25 completed", "effect": "No causal particle-versus-photon survival estimate", "limits": [ "Decisive trial enrolled zero", "Planned arms differed beyond particle type", "Small single-arm studies", "Historical targets cannot control selection", "Proton registry completion now extends to 2039", "PACK has no posted result", "Centre and travel effects are entangled with technology" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The physics can improve dose placement, but the missing comparison blocks a patient-benefit conclusion." }, "animalEvidence": { "rating": "not-applicable", "reason": "The failure is accrual and treatment-effect identification in people." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Cell killing cannot answer whether the full clinical route improves survival." }, "reproducibility": { "rating": "none", "reason": "No randomized comparative result exists to reproduce." }, "commercialConflicts": { "rating": "moderate", "reason": "Specialist centres benefit from use and expansion even without a single corporate sponsor." }, "recency": { "rating": "current", "reason": "Current registries confirm zero CIPHER accrual, completed PACK and the extended proton timeline." } } }, "layout": { "x": 652, "y": 576 }, "updated": "2026-09-15" }, { "id": "failure-pdac-dormancy-sample-gap", "title": "PDAC dormancy fails at the sample: the accessible human tissues are from the wrong time", "shortTitle": "Dormancy sample gap", "type": "failure", "status": "supported", "scope": [ "pdac", "transfer-unproven" ], "summary": "Primary tumours, one-time marrow aspirates and end-stage autopsies bracket the disease-free interval but do not observe the residual cell that later recurs.", "content": "Original belief: late recurrence, marrow tumour-cell detection or end-stage metastasis tissue would reveal a human dormant PDAC state. What happened: each sample answers a different time. Bone marrow at surgery finds cytokeratin or RNA signals and predicts poor outcome, but does not prove cell viability, inactivity or later ancestry. Primary-tumour sequencing can associate a state with later liver or lung recurrence, but in the 744-person cohort only 21 selected primaries entered single-nucleus analysis and none was a residual cell. Rapid autopsy maps extensive end-stage disease: the 2025 spatial study used 49 sections from 13 people and found multiple changing lineages within and between organs, after the dormant interval and often after treatment. A 14-autopsy surgical series found microscopic local cancer in 5/14, 36%, despite negative imaging, but again only at death. Why the approach fails: time, tissue and outcome are not joined in the same person. It confuses a marker in marrow, a predicted route in the primary and a state at death with the cell that survived treatment. Recovery rule: preserve a fixed patient timeline, bank the primary and serial blood, collect recurrence tissue before new therapy when safe, use rapid autopsy, and prove ancestry by shared private alterations before calling a state residual. A human dormancy claim additionally requires low proliferation at the disease-free time and later outgrowth. Stop the route if tissue handling creates the state, if the marker adds nothing beyond stage and ctDNA, or if sampling cannot link the proposed cell to recurrence.", "sourceIds": [ "effenberger-2012-pdac-bone-marrow-dtc", "nordgard-2022-unresectable-dtc", "chalabi-hajkarim-2025-organotropism", "pei-2025-spatial-metastatic-pdac", "tanaka-2012-microscopic-local-recurrence" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "claim-human-pdac-dtc-prognostic-not-dormancy", "relation": "failed-because-of" }, { "target": "unknown-dormant-cell-actionability", "relation": "leaves-open" }, { "target": "trial-pdac-residual-cell-state-action-map", "relation": "recovered-by" }, { "target": "programme-25m-prospective-observatory", "relation": "constrains" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "strong", "randomised": "not-applicable", "independentReplication": "moderate", "sampleSize": "Marrow cohorts n=175 and n=48; recurrence cohort n=744 with 21 sequenced primaries; rapid autopsy n=13 and n=14", "effect": "Sampling failure record; linked studies do not estimate a shared effect", "limits": [ "No direct disease-free organ sample", "Invasive repeat sampling is unsafe", "Autopsy selection", "Assays differ", "Ancestry not established" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The missing cell lies in an organ during a clinically silent interval that is rarely safe to biopsy." }, "animalEvidence": { "rating": "strong", "reason": "Models can label and recover residual cells at times inaccessible in people." }, "inVitroEvidence": { "rating": "weak", "reason": "Culture supplies access but changes quiescence, stromal contact and immune pressure." }, "reproducibility": { "rating": "moderate", "reason": "Different human study types repeat the same time-and-tissue gap." }, "commercialConflicts": { "rating": "low", "reason": "The failure concerns sample timing rather than a proprietary product." }, "recency": { "rating": "current", "reason": "The gap persists despite 2025 spatial and organ-linked human studies." } } }, "updated": "2026-09-14" }, { "id": "failure-pdac-exercise-endpoint-and-attrition", "title": "Exercise trials can turn completer activity into an apparent cancer-pathway benefit", "shortTitle": "Exercise endpoint and attrition failure", "type": "failure", "status": "supported", "scope": [ "pdac", "supportive-care", "surgery", "exercise", "trial-design" ], "summary": "Death, progression, non-completion, active controls and intermediate endpoints can make a safe programme look more effective than it is.", "content": "Failure 1, survivor analysis. APACaP randomized 313 but its week-16 main analysis used 172, or 55.0%. Advanced pancreatic disease removes questionnaires through progression, illness and death, so a completer score can improve while total useful days do not. Failure 2, actuator substitution. The pancreatectomy phone trial increased steps from 4,568 to 5,071 among completers, but severe complications were 22.9% versus 20.5%. Changing the actuator does not prove system output. Failure 3, converging controls. PancFit supplied nutrition counselling, trackers and activity encouragement to both groups. Recorded aerobic minutes were similar, so a negative group difference may mean the contrast was weak, not that movement has no value. Failure 4, secondary rescue. The 40-person multicomponent surgical trial missed overall complications but found delayed gastric emptying of 5.6% versus 40.9%, P=0.01. Nutrition, diabetes care, enzyme care, physical and breathing training all changed, so the component and multiplicity are unresolved. Failure 5, wrong comparator. Goal-directed walking did not beat conventional physical and nutritional support in 180 mixed operations; this does not compare support with none. Repair. Keep all assigned patients, count days alive and outside hospital, treatment dose and delay, function, complications and burden; record why activity was missed; and treat death as the worst missing state rather than deleting it.", "sourceIds": [ "neuzillet-2023-apacap-rct", "ngohuang-2023-pancfit-rct", "ausania-2019-pancreatic-prehab-rct", "zarate-rodriguez-2023-pancreatectomy-steps-rct", "mizuno-2024-hpb-prehab-rct", "li-2024-pancreatic-early-mobilization-rct" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "failure-pdac-palliative-timing-and-analysis", "relation": "replicates" }, { "target": "failure-pdac-postoperative-treatment-loss", "relation": "coordinates-with" }, { "target": "unknown-pdac-exercise-net-benefit-and-selector", "relation": "creates" }, { "target": "hypothesis-pdac-exercise-existing-data-gate", "relation": "succeeds-where" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "moderate", "randomised": "strong — the failures are visible in randomized trials", "independentReplication": "strong — endpoint and attrition limits recur across settings", "sampleSize": "APACaP 313 randomized and 172 primary analysable; phone trial 152 randomized and 83 completed; other trials n=40 to n=180", "effect": "Repeated separation between changed activity or secondary outcomes and negative main patient outcomes", "limits": [ "Reports do not expose every missing-data reason", "Interventions differ", "Mixed tumour and operation types", "No participant-level pooled analysis" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Function can help care, but the measurement hierarchy is causally distinct." }, "animalEvidence": { "rating": "not-applicable", "reason": "The bias arises from human trial participation and outcome handling." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "No laboratory assay resolves survivor analysis or programme adherence." }, "reproducibility": { "rating": "strong", "reason": "Several randomized records repeat the same endpoint and missingness boundary." }, "commercialConflicts": { "rating": "low", "reason": "Most programmes are academic and do not depend on a commercial device." }, "recency": { "rating": "current", "reason": "Multiple recent surgical trials confirm the boundary." } } }, "layout": { "x": 2360, "y": 1540 }, "updated": "2026-09-15" }, { "id": "failure-pdac-germline-route-dropout", "title": "The germline route loses people before testing and loses relatives after a positive result", "shortTitle": "Germline route dropout", "type": "failure", "status": "supported", "scope": [ "pdac", "pan-cancer", "transfer-unproven" ], "summary": "Guidelines and referral letters do not complete the route; low-friction testing works after entry, while invitation and family contact remain the larger losses.", "content": "Original belief: recommending universal testing and sending a genetics referral would be enough. Observed result: Dana-Farber automatic referral raised completion from 16.5% to 38.0%, but still left more than six in ten new consultations untested. Embedded or oncology-led routes did better but not uniformly: UCSF reached 71%, Johns Hopkins about 69.8% of all patients seen, and a physician-led protocol 81.5%. The route then broke again. Only 64.7% of positive patients in one mainstream service and 69.4% in the Hopkins service completed genetic counselling. In a prospective exocrine-pancreatic-cancer cohort, probands told 84% of first-degree relatives, but only 31% of informed relatives tested. A selected PanCAN survey found that 61.7% of mutation-positive respondents had at least one first-degree relative tested, not that 61.7% of all eligible relatives tested. Randomized family evidence separates reach from completion. In GENERATE, 90% of 601 motivated relatives completed covered home testing after joining; online education alone reached 93% versus 87% with an added live session and did not worsen mean anxiety, depression or cancer worry. In the pan-cancer GIFT trial, probands invited only 948/4,946 reported relatives, 19.2%. Free rather than $50 testing increased odds 2.5-fold, but the mean family fraction tested rose only four points from a 3% baseline; a navigator did not show a clear gain. Likely causes are clinician-dependent ordering, late testing in a lethal disease, language and access gaps, uncertain payment, family communication, privacy limits and no named owner after disclosure. Recovery requires one measured route with every eligible patient and adult relative in the denominator, consented family contact, rapid testing, expert handling of pathogenic and uncertain results, and documented actions. A raw test count can hide failure at every other step.", "sourceIds": [ "chittenden-2021-systematic-germline", "walker-2021-embedded-germline-station", "wang-2022-oncology-germline-cascade", "ramsey-2023-mainstreaming-germline", "rodriguez-2024-generate", "mcdonald-2024-germline-protocol", "sinan-2025-pdac-germline-implementation", "grewal-2026-pancan-germline-survey", "katz-2026-gift-cascade" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "claim-universal-germline-testing-needs-action", "relation": "limits" }, { "target": "unknown-pdac-germline-family-action-route", "relation": "creates" }, { "target": "trial-pdac-germline-cascade-action-map", "relation": "informs" }, { "target": "hypothesis-pdac-germline-family-route-completion", "relation": "repaired-by" } ], "evidence": { "humanProspective": "strong for delivery loss", "humanRetrospective": "strong", "randomised": "moderate for family-delivery components", "independentReplication": "strong", "sampleSize": "PDAC implementation cohorts up to 1,214 consultations; GENERATE n=601; GIFT 414 probands reporting 4,946 relatives", "effect": "Patient testing commonly 38%-82%; 31% of informed relatives tested in one prospective cohort; GIFT invitation 19.2%", "limits": [ "Different denominators", "Mostly academic North American centres", "Self-selection in family trials", "GIFT was pan-cancer", "Testing is not a health outcome" ], "dimensions": { "biologicalPlausibility": { "rating": "not-applicable", "reason": "This is a care-delivery failure rather than a biological mechanism." }, "animalEvidence": { "rating": "not-applicable", "reason": "Animals cannot model health-system and family-contact losses." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Test accuracy does not measure whether people receive and use results." }, "reproducibility": { "rating": "strong", "reason": "Several independent services and two family trials show losses at distinct steps." }, "commercialConflicts": { "rating": "moderate", "reason": "Several studies used commercial test providers or covered test costs, which can raise uptake compared with routine care." }, "recency": { "rating": "current", "reason": "The evidence runs through a 2026 population-based randomized family trial and survey." } } }, "updated": "2026-09-14" }, { "id": "failure-pdac-pain-death-and-evaluable-denominators", "title": "Pancreatic pain studies can improve the survivor score while losing the patient", "shortTitle": "Pain-study death and denominator failure", "type": "failure", "status": "supported", "scope": [ "pdac", "pain", "palliative-care", "trial-design", "missing-data" ], "summary": "Late pain endpoints exclude people who die or become too ill, and several procedures lower scores without quality-of-life or survival gain.", "content": "Pain near the end of life creates a competing-outcome problem. A three-month pain score exists only if the person is alive and able to answer at three months. Removing deaths from the analysis can make a treatment look helpful even if it trades pain for earlier death or hospitalization. The celiac-radiosurgery study shows the denominator gap: 48/90 evaluable people responded at three weeks, 53%, but only 90/125 treated people were evaluable. On the all-treated denominator, documented response was at most 48/125, 38.4%. Eleven serious grade 3 or worse events were at least possibly related, including two deaths judged possibly related. The ganglion and splanchnic randomized trials supply a stronger warning: small pain or opioid gains coexisted with death hazard ratios of 1.49 and 1.56. Different targets and secondary survival analyses prevent pooling those numbers as one causal harm estimate. They still invalidate pain-only expansion. The live 94-person trial uses three-month mean pain as primary, excludes people already using opioids and mixes direct ganglion treatment with plexus treatment according to visibility. Without exact procedure labels, all-randomized death handling and a survival stop rule, it can repeat both denominator and component failures. The planned UK radiation study is another single arm after a 125-person single-arm study and interrupts systemic treatment for at least 14 days around radiation.", "sourceIds": [ "lawrence-2024-celiac-radiosurgery-phase2", "levy-2019-celiac-ganglia-versus-plexus-rct", "dong-2021-splanchnic-neurolysis-rct", "clinicaltrials-nct06160323-upfront-cgn-cpn", "clinicaltrials-nct07689955-celiac-radiotherapy", "kanno-2020-eus-cpn-modern-opioid-rct" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "claim-pdac-nerve-ablation-pain-benefit-inconsistent", "relation": "limits" }, { "target": "unknown-pdac-nerve-ablation-net-benefit", "relation": "creates" }, { "target": "hypothesis-pdac-nerve-ablation-existing-trial-safety-gate", "relation": "repaired-by" }, { "target": "trial-pdac-celiac-pain-action-map", "relation": "mapped-by" }, { "target": "failure-cryoablation-selection-attrition-and-route-harm", "relation": "supports" }, { "target": "evidence-quality", "relation": "supports" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "none", "randomised": "strong", "independentReplication": "moderate — denominator and survival conflict recur", "sampleSize": "Randomized safety signals n=110 and n=96; radiosurgery n=125 treated and n=90 evaluable", "effect": "Pain response can be overstated and detached from survival and useful life", "limits": [ "Death attribution is difficult", "Survival was not always primary", "Techniques differ", "Some missing pain is unrelated to decline", "All-treated response treats unknowns conservatively", "Small event counts" ], "dimensions": { "biologicalPlausibility": { "rating": "not-applicable", "reason": "This is an outcome-accounting failure." }, "animalEvidence": { "rating": "not-applicable", "reason": "Only human follow-up can capture pain, death and missingness together." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "No bench system measures useful life." }, "reproducibility": { "rating": "moderate", "reason": "Multiple studies show pain endpoints separating from quality of life or survival." }, "commercialConflicts": { "rating": "low", "reason": "Reporting design matters more than a central commercial sponsor." }, "recency": { "rating": "current", "reason": "The same risks remain in 2026 trial designs." } } }, "layout": { "x": 1144, "y": 1068 }, "updated": "2026-09-15" }, { "id": "failure-pdac-palliative-timing-and-analysis", "title": "Palliative trials fail when care bundles vary and the sickest people disappear from measurement", "shortTitle": "Palliative route and analysis failure", "type": "failure", "status": "open", "scope": [ "pdac", "palliative-care", "quality-of-life", "trial-design", "care-delivery" ], "summary": "Referral labels hide different services, fixed-time quality scores select survivors, and uncontrolled survival comparisons reward time and selection bias.", "content": "Failure 1, undefined service. Early palliative care can mean a specialist physician, nurse screening, psychiatric referral, home visits, exercise, nutrition, telephone coaching or community navigation. A referral count does not identify the active part. Failure 2, survivor-only measurement. A 12-week questionnaire excludes people who die, decline or cannot respond unless the design assigns those outcomes explicitly. In a fast cancer, complete-case quality can improve while population useful life does not. Failure 3, wrong survival claim. The pancreatic randomized trial showed no survival difference. The 2026 8.0-versus-4.9-month home result compared later patients with an earlier cohort, so treatment improvements, epidemic periods, referral and fitness can explain it. Failure 4, mixed diseases. The 288-person trial joined pancreatic and biliary cancers and missed both main symptom endpoints. Failure 5, late referral. On-demand care can arrive after avoidable distress, repeated emergency use or non-beneficial chemotherapy. Failure 6, forced withdrawal. Palliative input must run beside active treatment and patient choice; using it to ration cancer therapy would create harm. Failure 7, research duplication. A 250-person Danish pancreatic randomized trial completed its main endpoint in 2024 without public results, while a new 60-person digital-navigation pilot remained recruiting after its estimated completion date. Recover existing answers before starting another generic referral study.", "sourceIds": [ "maltoni-2016-systematic-palliative-qol", "maltoni-2016-systematic-palliative-end-of-life", "woo-2019-pancreatobiliary-palliative-rct", "dufva-2026-home-palliative-pdac", "clinicaltrials-nct04632303-pdac-palliative", "clinicaltrials-nct07591571-pdac-community-worker" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "claim-pdac-palliative-care-improves-qol-not-survival", "relation": "limits" }, { "target": "unknown-pdac-palliative-route-net-benefit", "relation": "creates" }, { "target": "hypothesis-pdac-palliative-route-completion", "relation": "repaired-by" }, { "target": "failure-pdac-pain-death-and-evaluable-denominators", "relation": "shares-denominator-failure-with" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "moderate", "randomised": "moderate and mixed", "independentReplication": "moderate — mixed results", "sampleSize": "Two relevant randomized records n=207 and n=288; one historical-control record n=125; active completed-primary trial target n=250", "effect": "Heterogeneous packages and missing death-aware outcomes block one portable effect estimate", "limits": [ "Published reports use different endpoints", "Interventions cannot be blinded", "Care systems differ", "Danish result unavailable", "No patient-level pooled data", "Cause-specific service use difficult to assign" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Timing, component definition and death-aware analysis directly determine what a care trial measures." }, "animalEvidence": { "rating": "not-applicable", "reason": "The failure occurs in human services and analysis." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Service delivery cannot be tested in culture." }, "reproducibility": { "rating": "moderate", "reason": "Different trials reproduce some quality effects but not a common package or endpoint." }, "commercialConflicts": { "rating": "low", "reason": "No single commercial product controls the care model." }, "recency": { "rating": "current", "reason": "The failure includes 2026 trial and cohort status." } } }, "layout": { "x": 1252, "y": 1176 }, "updated": "2026-09-15" }, { "id": "failure-pdac-postoperative-treatment-loss", "title": "PDAC surgery often fails to deliver the planned systemic treatment", "shortTitle": "Treatment lost after surgery", "type": "failure", "status": "supported", "scope": [ "pdac" ], "summary": "Current English data show 38% without recorded systemic treatment within 14 weeks after Whipple, while older cohorts expose progression, complications and frailty behind the loss.", "content": "Original belief: once a resectable patient reaches surgery, adjuvant treatment can be added after recovery. Observed result: England's 2026 audit reports that only 62% of Whipple patients received chemotherapy or chemoradiotherapy within 14 weeks, with centre IQR 54-68%. It does not report full-course completion, so the complete-route rate is at most 62% under that time rule. In a 203-person surgery-first PDAC cohort, 85, 41.9%, completed chemotherapy, 41, 20.2%, started but did not complete, and 77, 37.9%, never started. The stated causes among failures were early progression 34.7%, complications or poor performance 32.2%, and age over 75 years 24.6%. A Dutch 1,071-person cohort associated major complications with disease-free interval HR 1.22 and overall-survival HR 1.25; organ-failure HRs were 1.86 and 1.89, and mediation modelling assigned much of these associations to adjuvant treatment. That cohort excluded deaths within 90 days and neoadjuvant therapy, so it understates the entire route and cannot prove mediation. A later 920-person PDAC cohort associated pancreatic and bile leaks with chemotherapy omission, OR 1.97 and 1.96, but found neither a delay beyond 90 days nor an overall-survival difference. Likely cause: biology, operation harm, frailty and handoff loss are mixed under one missing-treatment field. Recovery: record the intended route at first expert MDT, every delivered dose, every deviation and a fixed reason set; book postoperative oncology before discharge; audit nutrition, pancreatic enzymes, fitness and rescue; and report by centre, age, deprivation and travel time. Mechanism status: systemic treatment after resection remains effective; the failed belief is that surgery reliably hands patients to it.", "sourceIds": [ "npaca-2026-state-of-nation", "nice-2018-ng85-pancreatic", "labori-2016-adjuvant-completion", "henry-2023-complications-adjuvant", "kollbeck-2025-complications-chemo" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "claim-pdac-surgery-is-a-complete-treatment-route", "relation": "supports" }, { "target": "system-england-pancreatic-surgery-network", "relation": "occurs-within" }, { "target": "unknown-pdac-surgical-route-completion", "relation": "creates" }, { "target": "hypothesis-pdac-surgical-route-completion", "relation": "repaired-by" } ], "evidence": { "humanProspective": "moderate for national audit capture", "humanRetrospective": "strong", "randomised": "none for cause of loss", "independentReplication": "moderate", "sampleSize": "English national audit; retrospective PDAC cohorts n=203 and n=920; Dutch pancreatic-cancer cohort n=1,071", "effect": "62% received treatment within 14 weeks after Whipple; older complete-course rate 41.9%; leaks associated with omission around OR 1.96-1.97", "limits": [ "Current audit reports receipt rather than completion", "Different treatment eras and regimens", "Confounding by progression and fitness", "Mediation assumptions are unverified", "Selected operated populations" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Complications, malnutrition, frailty and progression can each prevent toxic systemic treatment after a major operation." }, "animalEvidence": { "rating": "not-applicable", "reason": "Human delivery records decide the failure." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Cell studies cannot separate clinical route-loss causes." }, "reproducibility": { "rating": "moderate", "reason": "Several cohorts show treatment loss, but exact rates and causes vary." }, "commercialConflicts": { "rating": "low", "reason": "No product claim drives the main result." }, "recency": { "rating": "current", "reason": "The route denominator is current to the September 2026 national audit, with older cohorts used to expose causes." } } }, "updated": "2026-09-14" }, { "id": "failure-pdac-vaccine-escalation-reversals", "title": "PDAC vaccine programmes repeatedly scaled immune or early survival signals that did not survive the next control", "shortTitle": "Vaccine escalation reversals", "type": "failure", "status": "supported", "scope": [ "pdac", "transfer-unproven" ], "summary": "TeloVac and ECLIPSE show two versions of the same failure: a biological response or small comparative signal was promoted before the clinical claim was stable.", "content": "Failure pattern 1, immune response became a stand-in for benefit. GV1001 had generated immune responses in early studies, yet TeloVac's 1,062-person phase 3 found no survival gain in either schedule. Failure pattern 2, the comparator could not answer the real question. In 90 treated patients, GVAX followed by CRS-207 beat GVAX alone, 6.1 versus 3.9 months, HR 0.59. That showed relative performance against another unproved vaccine route, not superiority to effective care. ECLIPSE then enrolled 303 and compared the combination with chemotherapy. In the 213-person primary cohort, survival was 3.7 versus 4.6 months, HR 1.17, 95% CI 0.84-1.64. Failure pattern 3, post-treatment responders were mistaken for treatment effect. The personalized RNA and ELI-002 studies both link larger vaccine-induced T-cell responses with longer recurrence-free survival, but those groups exist only after treatment. A patient able to mount a response may have better disease or host biology; randomization is what balances those causes. Failure pattern 4, a missed main endpoint was followed by a favourable subgroup. AMPLIFY-7P missed intention-to-treat disease-free survival, then highlighted an R0-resection subgroup selected after results. Recovery rule: do not scale on immune response, responder-versus-non-responder survival, a control made of another unproved vaccine, or a result-selected subgroup. Scale only an all-assigned comparison against the same standard care with a patient endpoint and complete harm reporting.", "sourceIds": [ "middleton-2014-telovac", "le-2015-gvax-crs207", "le-2019-eclipse", "rojass-2025-neoantigen-followup", "elicio-2026-amplify7p-results", "wainberg-2025-eli002-final" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "claim-pdac-vaccine-immunity-without-proven-benefit", "relation": "supports" }, { "target": "failure-algenpantucel-l-vaccine-phase3", "relation": "groups" }, { "target": "failure-eli002-phase2-missed-dfs", "relation": "groups" }, { "target": "trial-pdac-vaccine-action-map", "relation": "constrains" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong", "independentReplication": "strong", "sampleSize": "TeloVac n=1,062; GVAX/CRS-207 phase 2 n=90 treated; ECLIPSE n=303 with primary cohort n=213", "effect": "Early immune or comparative signal did not become survival benefit in the larger standard-care comparison", "limits": [ "Failure mechanisms can differ by product", "TeloVac tested advanced disease", "ECLIPSE tested previously treated metastatic disease", "The rule does not reject a new platform before a valid randomized test" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The failure concerns inference and trial design, not absence of immune biology." }, "animalEvidence": { "rating": "moderate", "reason": "Model evidence motivated the combinations but did not rescue the clinical results." }, "inVitroEvidence": { "rating": "strong", "reason": "Immune activity was measurable and still failed as a patient-benefit proxy." }, "reproducibility": { "rating": "strong", "reason": "The response-to-escalation error recurs across peptide, whole-cell, vector, RNA and KRAS programmes." }, "commercialConflicts": { "rating": "high", "reason": "Sponsor ownership and product-development incentives affect escalation and result framing." }, "recency": { "rating": "current", "reason": "The historical failures are interpreted against 2025-2026 results and registries." } } }, "layout": { "x": 404, "y": 348 }, "updated": "2026-09-15" }, { "id": "failure-pdac-vte-prophylaxis-delivery-gap", "title": "PDAC clot prevention fails when assessment and prescribing arrive after the highest-risk weeks", "shortTitle": "Late clot-prevention delivery", "type": "failure", "status": "supported", "scope": [ "pdac" ], "summary": "One cohort found prophylaxis in only 30% with a 43-day median delay while half of clots occurred within two months; education alone barely changed practice.", "content": "Original operating belief: published trials and guidelines would translate into an early offer of prophylaxis for eligible people starting systemic PDAC treatment. What happened: a 243-person pancreatic-adenocarcinoma cohort recorded VTE in 24%. Half of events occurred within two months of diagnosis; 52% of affected patients were hospitalized and 5% died from the clot. Only 30% received ambulatory prophylaxis and the median start was 43 days, using prescription rather than fill or adherence as exposure. Major bleeding occurred in 5.8%, and intestinal tumour invasion predicted bleeding. A 2026 single-centre quality project found that only 36% of 14 surveyed staff routinely offered prophylaxis, 29% never did and 64% cited eligibility uncertainty. Teaching changed eligible prescribing from 0/14 to 1/16. Adding an electronic prompt was followed by 9/16, 56%, receiving prophylaxis, but follow-up was too short and small for clot or bleeding effects. Why delivery fails: the decision lacks a named owner, a dated assessment, a usable bleeding and interaction check, a prescription and pharmacy route, and follow-up for persistence. What survives: routine early assessment and shared choice. What fails: education alone, counting a guideline as delivered care, or measuring prescriptions without fill, use, stop reason, clot and bleeding outcomes.", "sourceIds": [ "heffley-2024-pdac-vte-uptake", "noor-2026-pdac-vte-implementation", "ash-2023-primary-vte-prophylaxis-guide" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "claim-anticoagulation-prevents-pdac-vte-not-survival", "relation": "blocks-delivery-of" }, { "target": "unknown-pdac-vte-prophylaxis-rule", "relation": "leads-to" }, { "target": "hypothesis-pdac-vte-route-completion", "relation": "motivates" }, { "target": "risks", "relation": "supports" } ], "evidence": { "humanProspective": "weak", "humanRetrospective": "moderate", "randomised": "none", "independentReplication": "weak", "sampleSize": "One 243-person cohort and one small 46-patient quality project with a 14-person staff survey", "effect": "30% prophylaxis uptake with median start at day 43; electronic prompt followed by 56% uptake in 16 eligible patients", "limits": [ "Single centres", "Nonrandom exposure", "Prescription does not prove use", "Tiny quality project", "Short follow-up", "Modern local practice may differ" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "A preventive medicine cannot avert an early clot if assessment and treatment occur after the risk window." }, "animalEvidence": { "rating": "not-applicable", "reason": "This is a care-delivery question." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Laboratory tests cannot establish timely clinical delivery." }, "reproducibility": { "rating": "weak", "reason": "Two settings show low or delayed uptake, but neither is a large representative audit." }, "commercialConflicts": { "rating": "low", "reason": "The main evidence is academic observational and quality work rather than a product claim." }, "recency": { "rating": "current", "reason": "The cohort reported in 2024 and the quality project in 2026." } } }, "layout": { "x": 154, "y": 78 }, "updated": "2026-09-14" }, { "id": "failure-pdt-geometry-selection-and-bundling", "title": "PDT evidence fails at optical coverage, patient selection and bundled treatment", "shortTitle": "PDT coverage and attribution failure", "type": "failure", "status": "supported", "scope": [ "pdac", "photodynamic-therapy", "trial-design", "local-treatment" ], "summary": "The visible dead zone is a product of uneven drug, light and oxygen fields; highly selected cohorts and added chemotherapy then make scan and survival signals impossible to assign to PDT.", "content": "The physical system has three spatial inputs: photosensitizer concentration D(x), light energy L(x) and oxygen O(x). A useful treatment effect requires their product to exceed a killing threshold through the intended tumour while remaining below an injury threshold in bowel, vessels and normal pancreas. The studies measured input dose at the fibre and dead tissue later on a scan; they did not measure all three fields through the tumour. In the 15-person dose study, 40 joules produced dead zones consistently but their volumes varied substantially. With multiple fibres, about 1 cm of unintended light behind the diffuser tips overlapped and caused inflammation outside the pancreas. In the eight-person endoscopic study, 3/8 produced no visible dead zone. It screened 623 people, approached 54 and treated eight: 8/623 = 1.3%, a narrow path through anatomy, timing, travel and clinical fitness. Its registry records 13 starters, eight completers and five withdrawals, while the paper analyzes eight. The porfimer study then added chemotherapy after light treatment. The current 24-person study adds FOLFIRINOX chemotherapy, pembrolizumab immune treatment, biopsies and PDT in one group. A response could come from any component or selection. This is a measurement and design failure: delivered local chemistry is not reconstructed, non-completion is not integrated into efficacy and bundled care prevents causal attribution.", "sourceIds": [ "huggett-2014-verteporfin-pdt", "chandrasekhara-2021-eus-verteporfin-pdt", "clinicaltrials-nct03033225-vertpac02", "dewitt-2019-eus-porfimer-pdt", "clinicaltrials-nct01770132-porfimer-pdt", "clinicaltrials-nct06381154-pdt-pembrolizumab" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "claim-pdt-local-ablation-without-benefit-proof", "relation": "limits" }, { "target": "unknown-pdt-spatial-coverage-and-patient-benefit", "relation": "creates" }, { "target": "hypothesis-pdt-measured-spatial-gate", "relation": "repaired-by" }, { "target": "trial-pdac-pdt-action-map", "relation": "mapped-by" }, { "target": "failure-mwa-selection-bundling-and-missing-imaging", "relation": "supports" }, { "target": "failure-histotripsy-acoustic-window-and-translation", "relation": "compares-with" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "none", "randomised": "none", "independentReplication": "weak", "sampleSize": "n=15 dose study; n=12 porfimer study; 13 starters and eight completers in VERTPAC-02; current combination n=24", "effect": "Variable and incomplete dead zones, 38% non-response in the eight-person study, 38% non-completion among 13 starters, and no isolated treatment effect", "limits": [ "Imaging may understate or overstate cell killing", "Dead-zone shapes are irregular", "Attribution of individual withdrawals is limited", "Local treatment may still relieve obstruction", "Current outcomes are immature", "Drugs and delivery routes differ" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The failure follows directly from the need for drug, photons and oxygen to overlap in space." }, "animalEvidence": { "rating": "moderate", "reason": "Animal work confirms local injury but cannot remove human anatomy and selection constraints." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "A dish does not reproduce fibre placement, bowel clearance, tumour perfusion or withdrawal." }, "reproducibility": { "rating": "weak", "reason": "Small cohorts use different drugs, access routes, doses and later treatment." }, "commercialConflicts": { "rating": "moderate", "reason": "The key failure persists independent of vendor: treatment fields and patient flow are incompletely resolved." }, "recency": { "rating": "current", "reason": "The failure includes the active 24-person combination design checked in 2026." } } }, "layout": { "x": 946, "y": 870 }, "updated": "2026-09-15" }, { "id": "failure-pegph20-response-without-survival", "title": "PEGPH20 increased tumour response without improving progression or survival", "shortTitle": "PEGPH20 failure", "type": "failure", "status": "supported", "scope": [ "pdac" ], "summary": "Even after selecting hyaluronan-high metastatic PDAC, response rose from 36% to 47% but median survival was 11.2 versus 11.5 months.", "content": "HALO 301 tested enzymatic hyaluronan depletion to open PDAC stroma and improve chemotherapy delivery. It used the apparent mechanistic biomarker—hyaluronan-high tumour—and a double-blind phase 3 design. Among 492 people in the primary analysis, PEGPH20 increased objective response from 36% to 47%. Median progression-free survival was identical at 7.1 months and median overall survival was 11.2 versus 11.5 months; hazard ratio for death 1.00. Added fatigue, muscle spasm and hyponatraemia were observed. This is a clean example of tumour shrinkage not translating into control. It weakens this drug and bulk-hyaluronan strategy in metastatic PDAC. It does not prove that delivery is irrelevant; it shows that changing one matrix component and initial response was insufficient to change the lethal disease course.", "sourceIds": [ "hingorani-2020-halo301" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "failure-bulk-stroma-removal", "relation": "extends" }, { "target": "treatment", "relation": "part-of" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong", "independentReplication": "moderate", "sampleSize": "494 randomised; 492 in intention-to-treat analysis", "effect": "ORR 47% versus 36%; PFS 7.1 versus 7.1 months; OS 11.2 versus 11.5 months, HR 1.00", "limits": [ "Metastatic disease setting", "Tests one enzyme and matrix component", "Hyaluronan assay may not capture all delivery barriers", "Sponsor-developed biomarker", "Does not test selective fibroblast-state treatment" ] }, "layout": { "x": 16, "y": 92 }, "updated": "2026-09-14" }, { "id": "failure-pert-prescribing-and-supply-gap", "title": "PERT guidance cannot help patients who are not prescribed treatment or cannot obtain stock", "shortTitle": "PERT delivery failure", "type": "failure", "status": "supported", "scope": [ "pdac", "pancreatic-all", "transfer-unproven" ], "summary": "English prescribing reached about 48% against a stated 100% standard, then a national supply alert reported ingredient and manufacturing limits.", "content": "Original operating belief: once guidelines recommend PERT for unresectable pancreatic cancer, the low-cost supportive treatment will reach patients. What happened: an OpenSAFELY audit of 22,860 people with pancreatic cancer found prescribing rose from 41% in 2015 to 48% in early 2023, about one percentage point per year and still below the stated 100% standard. Regional rates varied, with the highest around 50% to 60%. A December 2024 UK national patient-safety alert then reported limited Creon 10,000 and 25,000 supply, intermittent alternatives that could not cover the gap, and out-of-stock periods caused by raw-ingredient and manufacturing-capacity limits. It said community supply was affected more than secondary care and required local mitigation. The audit cannot tell whether every unprescribed patient had a clinical need, filled a prescription or adhered; the alert is dated and current stock must be refreshed. The failure is still concrete: a care rule has a prescription, dispensing and continuity route. A trial or quality programme must record eligible patients, recommendation, prescription, pharmacy fill, dose, adherence, symptoms, product switches and treatment gaps. Counting prescriptions alone overstates delivery; counting a guideline overstates it further.", "sourceIds": [ "lemanska-2023-opensafely-pert-audit", "dhsc-2024-pert-shortage-alert" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "trial-pert-pdac-action-map", "relation": "constrains" }, { "target": "unknown-pert-pdac-utility", "relation": "constrains" }, { "target": "claim-pdac-delivery-is-variable-not-single-barrier", "relation": "supports" }, { "target": "risks", "relation": "supports" } ], "evidence": { "humanProspective": "not-applicable", "humanRetrospective": "strong for prescribing", "randomised": "none", "independentReplication": "moderate for the existence of delivery limits", "sampleSize": "22,860 English pancreatic-cancer records plus one national patient-safety alert", "effect": "Prescribing 41% to 48% versus a stated 100% standard; supply insufficient to meet full demand in the alert", "limits": [ "Prescription does not prove fill or use", "Cancer-stage eligibility is incompletely captured", "The safety alert is time-specific", "No patient-outcome effect from the shortage", "English system may not transfer", "Need current stock verification" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "A medicine has no effect when it is not prescribed, dispensed or taken." }, "animalEvidence": { "rating": "not-applicable", "reason": "This is a care-delivery and supply question." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Laboratory evidence cannot establish dispensing continuity." }, "reproducibility": { "rating": "moderate", "reason": "National prescribing variation and a separate safety alert support distinct access losses." }, "commercialConflicts": { "rating": "low", "reason": "The main sources are a public-data audit and a national safety alert." }, "recency": { "rating": "current", "reason": "The alert covered supply into 2026; live stock can change and must be checked before action." } } }, "layout": { "x": 153, "y": 77 }, "updated": "2026-09-14" }, { "id": "failure-pert-short-randomized-trials", "title": "Two randomized PERT trials did not prevent short-term weight loss or prove longer survival", "shortTitle": "PERT randomized failure", "type": "failure", "status": "refuted", "scope": [ "pdac", "pancreatic-all" ], "summary": "Trials of 67 and 88 patients missed their eight-week weight or BMI endpoints; opposite survival point estimates both remained statistically uncertain.", "content": "Original belief: giving pancreatic enzymes to unresectable pancreatic-cancer patients would prevent weight loss and thereby improve survival. What happened: the Korean double-blind phase 2 randomized 67 patients to PERT or placebo. Eight-week weight change was 1.49% versus 2.99%, P=.381; nutrition score, quality of life and OS did not improve, and median OS was 5.84 versus 8.13 months, P=.744. The Japanese open-label trial randomized 88 chemotherapy-treated patients. Ninety per cent had a low NBT-PABA test, but eight-week BMI ratios were 0.975 and 0.980, P=.780, and other nutrition measures were similar. Median OS favoured PERT numerically, 19.0 versus 12.0 months, P=.070, but the trial did not establish it. Why the tests may have failed: eight weeks may be too short; weight joins tumour wasting, intake, inflammation and malabsorption; enzyme dose and meal timing may be inadequate; and broad eligibility can dilute a treatment effect confined to proven maldigestion. What survives: treat symptoms and likely insufficiency. What does not survive: assuming that PERT prevents cachexia or extends survival in all PDAC. A successor must select the maldigestion state, verify dose and adherence, measure stool symptoms and absorption, and then test function, chemotherapy delivered and survival.", "sourceIds": [ "woo-2016-pert-rct", "saito-2018-pert-rct" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "claim-pert-survival-observational-randomized-unproved", "relation": "constrains" }, { "target": "unknown-pert-pdac-utility", "relation": "leads-to" }, { "target": "trial-pert-pdac-action-map", "relation": "informs" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong", "independentReplication": "moderate for failure of short weight endpoints", "sampleSize": "67-person double-blind phase 2 and 88-person open-label multicentre trial", "effect": "No significant eight-week weight or BMI benefit; neither survival comparison was significant", "limits": [ "Small studies", "Short endpoints", "Different eligibility and doses", "Possible under-treatment", "Enzyme-deficiency selection differed", "Failure does not refute symptom benefit in a selected patient" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "The therapy directly replaces digestive enzymes, but weight and survival also depend on tumour and inflammatory wasting." }, "animalEvidence": { "rating": "not-applicable", "reason": "The intervention replaces a human digestive function and is best tested directly in patients." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Bench digestion cannot settle patient weight or survival." }, "reproducibility": { "rating": "moderate", "reason": "Two randomized trials repeat failure of their short nutrition endpoint, though both are small." }, "commercialConflicts": { "rating": "low", "reason": "The decisive negative results were academic trials and report no dominant sponsor efficacy claim in the abstracts." }, "recency": { "rating": "historical", "reason": "The trials reported in 2016 and 2018; newer trials may alter the conclusion." } } }, "layout": { "x": 151, "y": 75 }, "updated": "2026-09-14" }, { "id": "failure-pharmacogenetic-panel-and-route-coverage", "title": "A small gene panel and a completed lab report can hide unsafe treatment delivery", "shortTitle": "Gene panel and route failure", "type": "failure", "status": "supported", "scope": [ "pdac", "pan-cancer", "treatment-safety", "pharmacogenetics", "health-equity" ], "summary": "Testing can fail through missed eligibility, late results, incomplete ancestry coverage, wrong dose action, absent escalation, or toxicity outside the tested variants.", "content": "Failure 1, denominator deletion. PACIFIC-PGx screened 487 people for DPYD eligibility and enrolled 462, or 95%, but only 50 of 109 eligible for UGT1A1, or 46%. Only 276/462 and 30/50 received the relevant medicine and entered the safety analysis. Reporting only treated carriers would hide each earlier loss. Failure 2, panel coverage. The standard four DPYD variants were validated mainly in European populations. A 32-study review found 53 variants in non-European patients and identifies c.557A>G as an African-ancestry candidate absent from the UK panel. Race is not a valid substitute for genotype, and adding a weakly validated variant is not automatically safer. The repair is direct variant and enzyme-performance research across populations plus outcome monitoring after every negative result. Failure 3, false safety. UK guidance explicitly warns that severe toxicity can occur despite a negative test. The test removes one failure class; it does not certify the whole regimen. Failure 4, dose-action drift. Starting lower can prevent exposure injury but permanent underdosing may lose cancer control. PACIFIC-PGx returned 96% of results before cycle one in five to seven days, yet later escalation was not the central randomized question. Failure 5, UGT1A1 transfer. UGT1A1 clears SN-38, the active irinotecan metabolite. In 199 pancreatic patients, heterozygous risk appeared with original but not modified FOLFIRINOX. A genotype-guided gastrointestinal study then failed its pancreatic and biliary tolerability bound: 6/19 and 4/19 had dose-limiting toxicity. The 2026 Dutch cohort supports a 70% poor-metabolizer start, but it was non-randomized and mixed-cancer. DPYD and UGT1A1 are two separate control loops and must not be collapsed into one positive-test rule.", "sourceIds": [ "glewis-2024-pacific-pgx", "mhra-2020-dpd-testing", "chan-2024-dpyd-non-european-review", "shirasu-2019-ugt1a1-folfirinox", "sharma-2019-ugt1a1-mfolfirinox", "heersche-2026-ugt1a1-guided-dosing" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "claim-dpyd-testing-prevents-avoidable-fluoropyrimidine-harm", "relation": "limits" }, { "target": "unknown-pdac-pharmacogenetic-route-net-benefit", "relation": "creates" }, { "target": "hypothesis-pdac-pharmacogenetic-route-completion", "relation": "succeeds-where" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "strong", "randomised": "weak — pancreatic utility and route completion were not randomized", "independentReplication": "moderate — the risk repeats but panel, regimen and population boundaries differ", "sampleSize": "PACIFIC-PGx 487 DPYD-eligible and 109 UGT1A1-eligible; pancreatic UGT1A1 cohorts n=199 and n=19", "effect": "Measured eligibility and analysis losses, residual variant coverage, and a failed pancreatic genotype-guided tolerability bound", "limits": [ "Mixed cancers", "Historical comparisons", "Small poor-metabolizer groups", "No national pancreatic route denominator", "Ancestry evidence is incomplete" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Both enzymes directly alter exposure to drugs used in pancreatic regimens." }, "animalEvidence": { "rating": "not-applicable", "reason": "The failure is visible in human route and toxicity records." }, "inVitroEvidence": { "rating": "moderate", "reason": "Variant function can support inclusion but cannot prove clinical route benefit." }, "reproducibility": { "rating": "moderate", "reason": "Exposure risk repeats, while dosing success varies by regimen and study." }, "commercialConflicts": { "rating": "low", "reason": "No single assay vendor is required for the proposed route audit." }, "recency": { "rating": "current", "reason": "The latest UGT1A1 multicentre outcome was published in June 2026." } } }, "layout": { "x": 1920, "y": 1540 }, "updated": "2026-09-15" }, { "id": "failure-pipac-repeat-procedure-selection-and-bundling", "title": "PIPAC studies turn survival and fitness into apparent treatment response", "shortTitle": "PIPAC repeat-procedure bias", "type": "failure", "status": "supported", "scope": [ "pdac", "peritoneal-metastases", "pipac", "trial-design", "selection" ], "summary": "Repeat biopsy and treatment require survival, stable disease and operative fitness, while the current trial bundles aerosol and systemic copies of the same drug.", "content": "A tissue-response comparison needs at least two operations: a starting biopsy and a later biopsy. A three-procedure analysis needs a person to remain alive, operable and sufficiently stable across roughly three treatment windows. In the 156-person registry, only 55 reached at least three PIPAC procedures; 77 stopped for progression and ten for poor condition. Completing treatment is therefore partly an output of slower disease, not just an input that could cause longer survival. A 120-day landmark analysis removes the guaranteed-survival period before three procedures, but cannot make fitter completers exchangeable with people whose disease or health stopped treatment. Choice of cisplatin-doxorubicin versus nab-paclitaxel was also not randomized. The current phase II plan adds another attribution failure: every participant receives six intravenous cycles of nab-paclitaxel-gemcitabine and three abdominal nab-paclitaxel aerosols. Its disease-control rate can describe the package but cannot estimate aerosol benefit. Eligibility further selects ECOG 0-1 fitness, no extra-abdominal spread, no bowel obstruction and no early progression on systemic treatment. A positive package rate would remain compatible with systemic chemotherapy plus selection.", "sourceIds": [ "digiorgio-2026-pipac-pdac-registry", "clinicaltrials-nct05371223-nab-pipac", "digiorgio-2024-nab-pipac-protocol", "graversen-2017-pipac-pdac" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "claim-pipac-local-delivery-signal-without-benefit-proof", "relation": "limits" }, { "target": "unknown-pipac-added-survival-quality-and-systemic-control", "relation": "creates" }, { "target": "hypothesis-pipac-existing-data-and-control-gate", "relation": "repaired-by" }, { "target": "trial-pdac-pipac-action-map", "relation": "mapped-by" }, { "target": "failure-conversion-surgery-selection-bias", "relation": "supports" }, { "target": "failure-local-control-without-survival-lap07", "relation": "supports" } ], "evidence": { "humanProspective": "weak", "humanRetrospective": "strong — for identifying the selection mechanism, not efficacy", "randomised": "none in pancreatic cancer", "independentReplication": "moderate — repeat-procedure selection recurs across reports", "sampleSize": "Registry n=156; 55 completed at least three procedures; planned single-arm trial n=38", "effect": "Completion, response and survival cannot be assigned to PIPAC", "limits": [ "Landmark adjustment is better than naive completer analysis", "Unmeasured fitness remains", "Registry details may permit further weighting", "Package activity is still measurable", "Severe peritoneal disease has few options" ], "dimensions": { "biologicalPlausibility": { "rating": "not-applicable", "reason": "This is a causal-design failure." }, "animalEvidence": { "rating": "not-applicable", "reason": "Animal models cannot repair survival-conditioned human sampling." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Cell studies cannot separate clinical selection or concurrent care." }, "reproducibility": { "rating": "moderate", "reason": "The same need for repeated surgery appears across centres and older cohorts." }, "commercialConflicts": { "rating": "moderate", "reason": "Procedure programmes benefit from completer-focused reporting, whether or not bias is intentional." }, "recency": { "rating": "current", "reason": "The largest 2026 report quantifies attrition and applies a partial correction." } } }, "layout": { "x": 1084, "y": 1008 }, "updated": "2026-09-15" }, { "id": "failure-polar-missed-primary-thresholds", "title": "POLAR missed both efficacy thresholds, but a long-survival tail keeps biomarker-guided PARP–checkpoint maintenance open", "shortTitle": "POLAR mixed result", "type": "failure", "status": "contested", "scope": [ "pdac" ], "summary": "The core 33-person HRD cohort had 35% response and 64% six-month progression-free survival, below the planned 43% and 77% thresholds, while estimated three-year survival was 44%.", "content": "Original belief: homologous-recombination deficiency could create both PARP sensitivity and enough mutation-derived immune visibility for olaparib plus pembrolizumab maintenance to produce unusually durable control after platinum. What happened: in the 33-person BRCA1, BRCA2 or PALB2 cohort, only 20 were measurable for response. Objective response was 35% against a 43% threshold, and six-month progression-free survival was 64% against a 77% threshold; both co-primary targets were missed. Median progression-free survival was 8.3 months and median overall survival 28 months; estimated three-year survival was 44%. Why it probably missed: the combination may help only a narrower biological subgroup, maintenance selected platinum-sensitive survivors, and the small single-arm trial cannot separate treatment effect from selection or indolent biology. What survives: deep molecular study of exceptional responders, frameshift neoantigens, immune infiltration and ctDNA clearance may define a testable subgroup. The failure invalidates a broad efficacy claim, not the entire HRD–immune interaction. Next test: randomise a biomarker locked before enrolment, stratify by biallelic core-HRD state and prior platinum response, and require an overall-survival or treatment-free-survival effect rather than a response-defined subgroup.", "sourceIds": [ "park-2026-polar-pdac", "kindler-2022-polo-overall-survival" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "failure-polo-os-not-demonstrated", "relation": "extends" }, { "target": "experiments", "relation": "resolves" }, { "target": "treatment", "relation": "part-of" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "not-applicable", "randomised": "none", "independentReplication": "none", "sampleSize": "63 total; 33 core BRCA1/2 or PALB2 HRD; 20 response-evaluable in the core cohort", "effect": "Core cohort ORR 35% versus planned 43%; six-month PFS 64% versus planned 77%; median OS 28 months; estimated three-year OS 44%", "limits": [ "Single-arm phase 2", "Small biomarker cohorts", "Maintenance population selected after platinum", "Only 20 core-cohort patients were response-evaluable", "Long-survival tail lacks a randomised counterfactual", "Commercial conflicts", "Exploratory biomarkers need locked validation" ] }, "layout": { "x": 17, "y": 89 }, "updated": "2026-09-14" }, { "id": "failure-polo-os-not-demonstrated", "title": "POLO delayed progression but did not demonstrate an overall-survival gain", "shortTitle": "POLO endpoint split", "type": "failure", "status": "supported", "scope": [ "pdac" ], "summary": "Maintenance olaparib met its progression-free endpoint in germline BRCA-mutated metastatic PDAC, but final median overall survival was 19.0 versus 19.2 months.", "content": "The phase 3 POLO trial tested maintenance olaparib against placebo in 154 people with germline BRCA1/2-mutated metastatic pancreatic adenocarcinoma whose disease had not progressed during at least 16 weeks of platinum chemotherapy. The earlier trial result established a progression-free survival benefit. At final analysis, median overall survival was 19.0 months with olaparib and 19.2 months with placebo; hazard ratio 0.83, 95% confidence interval 0.56 to 1.22, P=0.35. Three-year survival estimates were 33.9% and 17.8%, and time to later therapy improved. The failed claim is narrow: this trial did not demonstrate longer overall survival. It does not refute DNA-repair targeting, olaparib activity, or benefit on other endpoints. The endpoint split is a warning against treating progression-free survival as proof of cure or durable control.", "sourceIds": [ "kindler-2022-polo-overall-survival" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "treatment", "relation": "supports" }, { "target": "pdac", "relation": "part-of" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong", "independentReplication": "weak", "sampleSize": "154 randomised patients with germline BRCA-mutated metastatic pancreatic adenocarcinoma", "effect": "Median OS 19.0 versus 19.2 months; hazard ratio 0.83, 95% CI 0.56–1.22, P=0.35", "limits": [ "Small molecular subgroup", "Only platinum-responsive or stable patients entered maintenance", "Open-label later treatment can affect overall survival", "Industry involvement" ] }, "layout": { "x": 14, "y": 82 }, "updated": "2026-09-14" }, { "id": "failure-polyp-surveillance-low-yield", "title": "Polyp surveillance substitutes repeated measurement for a cancer-risk selector", "shortTitle": "Low-yield polyp surveillance", "type": "failure", "status": "supported", "scope": [ "gallbladder" ], "summary": "The current pathway repeatedly scans common, unstable ultrasound findings, yet most operations remove no neoplastic target and no trial has shown fewer deaths.", "content": "Original belief: polyp size, two-millimetre growth and age would concentrate a premalignant group enough to justify follow-up and cholecystectomy. Observed result: in POLYP, patients with repeat imaging received a median of three scans over about two years; the 2022 guideline changed only three operation indications versus 2017. Among 68 operations supported by the revised rule, six adenomas and one high-grade lesion were found, while 62 had no adenoma, 41 had non-neoplastic findings and 21 had no polyp or wall abnormality. No cancer was found. In the much larger Kaiser cohort, 507 polyps grew to 10 mm with no cancer recorded over 1,549 person-years. Likely causes are millimetre-scale ultrasound noise, cholesterol deposits and stones mislabelled as polyps, an action threshold tied to size rather than biology, and the low base rate of dangerous progression. The mechanism survives narrowly: a large solitary lesion, PSC or another strong feature can still justify action, and the one 17-mm solitary lesion in POLYP contained high-grade dysplasia. Recovery requires one locked confirmatory scan, a second independent risk signal and prospective outcome capture. Stop routine research follow-up for a small polyp if growth is the only new feature and a second test cannot raise expected benefit above surgery harm.", "sourceIds": [ "van-dijk-2025-polyp-study", "szpakowski-2020-gallbladder-polyps", "pawlik-2025-psc-gallbladder-polyps" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "claim-gallbladder-polyp-guidelines-still-overtreat", "relation": "failed-because-of" }, { "target": "claim-gallbladder-polyp-selector", "relation": "extends" }, { "target": "unknown-gallbladder-prevention-selector", "relation": "supports" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "strong", "randomised": "none", "independentReplication": "moderate for low yield; none for mortality benefit", "sampleSize": "Prospective polyp cohort n=302 with 68 revised-guideline operations; retrospective ultrasound cohort n=622,227", "effect": "One HGD and no cancer among 68 revised-guideline operations; no cancer after growth to 10 mm among 507 people followed for 1,549 person-years", "limits": [ "Not a randomized comparison of surveillance strategies", "One high-grade lesion creates wide uncertainty", "Follow-up ended for many people at surgery", "PSC and endemic populations require separate rules", "Failure of the current rule does not prove every polyp harmless" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The adenoma route exists, but ultrasound polyp labels pool several benign and neoplastic entities." }, "animalEvidence": { "rating": "not-applicable", "reason": "The failure is a human measurement and delivery problem." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Laboratory models do not validate a surveillance threshold." }, "reproducibility": { "rating": "moderate", "reason": "Prospective and large retrospective data agree that small-polyp pathways have low cancer yield." }, "commercialConflicts": { "rating": "low", "reason": "No product controls the core evidence." }, "recency": { "rating": "current", "reason": "The decisive practice study reported in 2025." } } }, "layout": { "x": 24, "y": 109 }, "updated": "2026-09-14" }, { "id": "failure-psc-surveillance-no-curable-detection", "title": "Annual MRI/MRCP plus CA19-9 did not create curable PSC detection", "shortTitle": "PSC surveillance failure", "type": "failure", "status": "supported", "scope": [ "cholangiocarcinoma", "hepatobiliary" ], "summary": "A prospective five-year national surveillance route detected cholangiocarcinoma but did not detect it early enough for long-term survival.", "content": "Original belief: annual contrast MRI/MRCP, CA19-9 and clinical review, followed by ERCP for severe or progressive duct change, would move PSC-related cholangiocarcinoma into a curable state. Observed result: among 512 people at 11 Swedish hospitals, 11 cholangiocarcinomas and two high-grade dysplasias were diagnosed, but median survival after cholangiocarcinoma was 13 months. Severe or progressive duct change occurred in 122 people, 23.8%, and only 10% of that subgroup had malignancy. Likely cause: inflammatory strictures create many suspicious images, while the interval from detectable malignant change to unresectable disease may be short and the escalation rule lacks enough specificity. The study did not randomize surveillance, so it cannot estimate what would have happened without the route. Mechanism status: risk-based PSC follow-up survives; this fixed annual test-and-escalation rule does not qualify as proven cancer interception. Recovery: retain the cohort and specimens, build a time-updated rule that separates benign duct change from malignant progression, and test the whole action route across centres. Stop if the new rule does not reduce benign invasive work or increase resectable high-grade dysplasia and cancer at an independently locked threshold.", "sourceIds": [ "villard-2023-psc-surveillance", "tan-2024-psc-mrcp-surveillance" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "claim-cca-psc-surveillance-workload", "relation": "failed-because-of" }, { "target": "unknown-cca-surveillance-utility", "relation": "constrains" }, { "target": "hypothesis-cca-route-specific-action-platform", "relation": "succeeds-where" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "moderate", "randomised": "none", "independentReplication": "weak — the independent retrospective association does not show post-cancer survival benefit", "sampleSize": "512 prospectively followed for five years; 298 in an independent retrospective cohort", "effect": "Prospective route found 11 cancers but median survival was 13 months; 23.8% had severe or progressive imaging change", "limits": [ "Failure applies to the tested route, not all PSC follow-up", "No no-surveillance arm", "ERCP and transplant decisions may differ", "Cancer is rare", "The exact source of the retrospective mortality association is unresolved" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Inflammatory duct change can hide or mimic malignant change and explains the specificity problem." }, "animalEvidence": { "rating": "not-applicable", "reason": "This is a human care-path failure." }, "inVitroEvidence": { "rating": "weak", "reason": "No laboratory result repairs a late or non-specific imaging trigger." }, "reproducibility": { "rating": "moderate", "reason": "Poor survival after cancer diagnosis persists across the prospective and retrospective reports, while all-cause results differ." }, "commercialConflicts": { "rating": "low", "reason": "No product claim drives the conclusion." }, "recency": { "rating": "current", "reason": "The relevant 2023 and 2024 human reports were checked in September 2026." } } }, "updated": "2026-09-14" }, { "id": "failure-rfa-selection-and-route-attrition", "title": "Selected observational survival hid the ablation route's losses and harms", "shortTitle": "RFA selection and attrition", "type": "failure", "status": "supported", "scope": [ "pdac", "locally-advanced", "radiofrequency-ablation", "trial-design" ], "summary": "Earlier 20-to-34-month observational survival did not survive randomization; PELICAN also showed that 26% assigned to ablation never received it.", "content": "Earlier pancreatic RFA cohorts reported median survival near 20 months and up to 34 months in complex multi-treatment groups. PELICAN's randomized ablation arm produced 12.1 months from randomization and no advantage. The difference is not evidence that the operation worsened every person's survival; it shows why survival measured only after successful procedure selection is biased. In PELICAN, 95 people were assigned to the RFA route but only 70, 74%, received it. Twenty-one, 22%, had hidden metastases found during operative inspection, two underwent resection instead, one progressed before ablation and one declined. Controls had 91/93, 98%, receive chemotherapy. A procedure-only analysis would remove precisely the people whose hidden spread or changing anatomy made the route fail. Random assignment correctly preserves them because detection, travel, surgery and treatment interruption are consequences of choosing the route. The harm burden also includes the operation: any serious adverse event was 33% versus 15%, grade 3 or worse serious events 27% versus 11%, and 81% of the RFA group had a recorded complication from laparotomy. The reusable lesson is that local-procedure evidence must start at assignment and retain discovery of spread, avoided or substituted treatment, delay, operation harm and quality of life. A technically successful burn is not route success.", "sourceIds": [ "seelen-2026-pelican", "clinicaltrials-nct03690323-pelican" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "claim-pelican-rfa-no-survival-and-more-harm", "relation": "supports" }, { "target": "unknown-rfa-human-mechanism-after-negative-trial", "relation": "bounds" }, { "target": "hypothesis-pelican-specimen-salvage-gate", "relation": "repaired-by" }, { "target": "trial-pdac-rfa-action-map", "relation": "mapped-by" }, { "target": "failure-direct-selection-and-randomized-reporting-gap", "relation": "parallels" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong", "independentReplication": "moderate", "sampleSize": "95 assigned to RFA route and 93 analysed controls; 70 versus 91 received assigned treatment", "effect": "RFA route completion 74% versus chemotherapy 98%; severe serious events 27% versus 11%", "limits": [ "Some non-delivery was clinically appropriate", "Laparoscopy discovered occult disease that may independently help avoid futile surgery", "The trial tested an open surgical route, not endoscopic RFA", "Induction treatment was shorter than many current pathways", "Two control withdrawals were excluded", "Older observational comparisons used different survival origins" ], "dimensions": { "biologicalPlausibility": { "rating": "not-applicable", "reason": "This failure concerns denominator and route selection." }, "animalEvidence": { "rating": "not-applicable", "reason": "Animal ablation cannot measure human route attrition." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Cell destruction cannot repair selection bias." }, "reproducibility": { "rating": "strong", "reason": "The randomized design directly exposes why selected treated cohorts can overstate benefit." }, "commercialConflicts": { "rating": "low", "reason": "The main conclusion opposes adoption and is not a sponsor efficacy claim." }, "recency": { "rating": "current", "reason": "Full route flow and harms were published in 2026." } } }, "layout": { "x": 718, "y": 642 }, "updated": "2026-09-15" }, { "id": "failure-routine-plastic-pbd", "title": "Routine delayed plastic-stent drainage increased complications before pancreatic surgery", "shortTitle": "Routine plastic drainage failed", "type": "failure", "status": "supported", "scope": [ "pdac" ], "summary": "The 2010 randomized rule doubled serious complications without improving mortality, so jaundice relief alone cannot justify delaying prompt surgery.", "content": "Original belief: lowering bilirubin for four to six weeks before pancreatic-head surgery would improve postoperative recovery. What happened: among 202 randomized patients, serious complications occurred in 39% with surgery within one week and 74% with drainage followed by surgery, a 35-point absolute increase. Drainage succeeded in 94%, yet 46% had a drainage complication; the full report recorded 27 cholangitis cases. Surgery-related complications were 37% versus 47%, and mortality did not differ. Why the rule failed: the route added ERCP, a small plastic device, occlusion, infection, reintervention and four to six weeks of exposure before an operation that could occur promptly. What survives: urgent drainage for cholangitis and selected drainage for severe symptoms, neoadjuvant treatment or unavoidable delay. Modern metal stents lower dysfunction and a 2026 trial found a different safety boundary, so this failure must not be stretched into a claim that every drainage device or indication is harmful. Recovery rule: record why drainage is needed, use a durable internal route when delay is real, and measure infection, reintervention, treatment completion and surgery—not bilirubin alone.", "sourceIds": [ "van-der-gaag-2010-pbd-rct", "costamagna-2026-sems-early-surgery-rct", "esge-2018-biliary-stenting-guideline" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "claim-biliary-drainage-is-route-not-cancer-treatment", "relation": "constrains" }, { "target": "unknown-pdac-biliary-drainage-route", "relation": "leads-to" }, { "target": "trial-pdac-biliary-drainage-action-map", "relation": "learned-by" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong", "independentReplication": "moderate for the route boundary", "sampleSize": "Randomized n=202; later randomized metal-stent comparator n=284", "effect": "Serious complications 39% versus 74%; drainage complications 46%; no mortality difference", "limits": [ "Mostly plastic stents", "Bilirubin limited to 40-250 micromol/L", "No neoadjuvant therapy", "Older ERCP practice", "Failure is route-specific" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "An invasive temporary device can add infection and occlusion during an avoidable delay." }, "animalEvidence": { "rating": "not-applicable", "reason": "Randomized human procedural evidence answers the question directly." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Device complications cannot be inferred from laboratory work." }, "reproducibility": { "rating": "moderate", "reason": "The large effect is internally strong, while modern metal-stent results define a different route." }, "commercialConflicts": { "rating": "low", "reason": "The original randomized failure was not built around a manufacturer efficacy claim." }, "recency": { "rating": "historical", "reason": "The 2010 result remains the routine-drainage anchor but device practice has changed." } } }, "updated": "2026-09-14" }, { "id": "failure-sig12d-loder-randomized-os-not-shown", "title": "A local KRAS siRNA implant was feasible, but its randomized trial did not show longer survival", "shortTitle": "siG12D-LODER uncertainty", "type": "failure", "status": "contested", "scope": [ "pdac" ], "summary": "The first-generation implant missed overall randomized survival; its second-generation successor is entering a staged 403-person trial without reported human tumour knockdown.", "content": "Original idea: place a slow-release small-interfering RNA implant directly into a locally advanced pancreatic tumour so KRAS G12D/V can be suppressed despite poor drug delivery. What happened: the phase 2 study enrolled 59 people across two cohorts without requiring the target mutation. In randomized cohort 1, modified-intention-to-treat median overall survival was 22.7 months with the implant plus gemcitabine and nab-paclitaxel versus 21.9 months with chemotherapy alone, with P greater than 0.05. The exploratory G12D/V subgroup produced 22.7 versus 13.5 months, but the hazard ratio was 0.59 with a 95% confidence interval from 0.18 to 1.96 and P = 0.39. That interval spans a large benefit and substantial harm. Cohort 2 was single-arm: response rate was 31.6% overall and 57.1% in its G12D/V subgroup. Procedure-related gastrointestinal events and more infections accompanied implantation. The study establishes that endoscopic local delivery is possible; it does not establish efficacy. A new trial would need prospective mutation selection, enough patients for survival, a current systemic KRAS inhibitor comparator, and measurements showing sustained tumour-wide KRAS knockdown. Local suppression also cannot treat occult distant disease unless paired with effective systemic control. The 2026 successor plan does not erase this result. SIL204 changes the RNA chemistry, formulation and delivery into combined intratumour and subcutaneous dosing, so first-generation efficacy cannot be inherited and first-generation safety cannot cover every new exposure. CTIS authorises 403 planned participants, with a 15-to-21-person safety segment before randomized dose selection and confirmation. That staging is useful, but the public objectives list plasma concentration rather than paired multi-region mutant-KRAS reduction, and participant-level data sharing is set to no. One Israeli site was cleared to screen in July 2026; first dosing was still anticipated, while German recruitment remained pending. Repair requires human target knockdown before expansion, complete mutation denominators and a public independent analysis.", "sourceIds": [ "alagesan-2026-sig12d-loder", "ctis-2026-sil204-phase23", "silexion-2026-sil204-site-initiation", "silexion-2026-halfyear-filing" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "claim-daraxonrasib-common-ras-survival-breakthrough", "relation": "contrasts-with" }, { "target": "claim-kras-mutant-dosage-may-shape-resistance", "relation": "depends-on" }, { "target": "companies", "relation": "involves" }, { "target": "experiments", "relation": "resolves" }, { "target": "claim-sil204-trial-scales-before-human-knockdown", "relation": "extends-to" }, { "target": "unknown-sil204-human-target-delivery", "relation": "leaves" }, { "target": "trial-sil204-action-map", "relation": "repaired-by" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "not-applicable", "randomised": "moderate", "independentReplication": "none", "sampleSize": "59 across two cohorts; the randomized mutation-selected subgroup was smaller", "effect": "Randomized unselected median OS 22.7 vs 21.9 months, not significant; exploratory G12D/V HR 0.59, 95% CI 0.18-1.96", "limits": [ "Open-label", "Modified intention-to-treat analysis", "Underpowered", "Mutation-unselected primary cohort", "Exploratory subgroup", "Wide confidence interval", "Single-arm second cohort", "Procedure-related harms", "Company-affiliated authors" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "Mutant-KRAS RNA cutting is plausible, but local and systemic cytosolic delivery remain unproved." }, "animalEvidence": { "rating": "moderate", "reason": "Second-generation sponsor models justify a small measurement stage, not inherited clinical benefit." }, "inVitroEvidence": { "rating": "moderate", "reason": "SIL204 shows cell-system activity after access." }, "reproducibility": { "rating": "weak", "reason": "The first-generation overall randomized signal was negative and no second-generation human result exists." }, "commercialConflicts": { "rating": "high", "reason": "Two paper authors were company-affiliated and all successor-product evidence is sponsor-controlled." }, "recency": { "rating": "current", "reason": "Includes the 2026 paper and authorised successor trial." } } }, "layout": { "x": 14, "y": 106 }, "updated": "2026-09-15" }, { "id": "failure-sipp-t3-accrual-collapse", "title": "SIPP-T3 stopped recruiting after four of 212 planned participants", "shortTitle": "SIPP-T3 accrual collapse", "type": "failure", "status": "supported", "scope": [ "pdac" ], "summary": "The randomized tocotrienol trial enrolled 1.9% of its protocol target and cannot answer its intended prevention question at the recorded size.", "content": "The posted SIPP-T3 protocol planned a double-masked randomized comparison of delta-tocotrienol with placebo in 212 people with surveillance-eligible IPMN, expected more than ten recruits per month and predicted full accrual within two years. The live registry, last updated 13 November 2025, records actual enrolment four and status active, not recruiting: 4/212 = 1.89% of target. No results are posted. Regardless of treatment effect, four participants cannot estimate a credible prevention effect or common harms. The primary composite is also weakly tied to cancer: it can count an increase in cyst size above 1 mm as progression, alongside worrisome features and high-risk stigmata. Measurement error or harmless growth can therefore create events without HGD or cancer. This is an execution and endpoint failure, not evidence that tocotrienol lacks biological activity. Recovery would require public accrual reasons, participant disposition and safety, followed by a new feasibility gate before another efficacy trial. Do not fund a restart unless at least 60 eligible participants can be enrolled across sites within 12 months, retention exceeds 85%, and the primary clinical endpoint is HGD, invasive cancer, pancreatic resection for locked criteria, or a validated bridge to those outcomes rather than 1-mm growth alone.", "sourceIds": [ "clinicaltrials-nct06519097-sipp-t3" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "trial-pancreatic-precursor-action-map", "relation": "failed-because-of" }, { "target": "programme-5m-decision-core", "relation": "constrains" }, { "target": "unknown-dangerous-precursor-selection", "relation": "supports" } ], "evidence": { "humanProspective": "none usable — four participants and no posted results", "humanRetrospective": "not-applicable", "randomised": "none usable — designed as randomized but actual n=4 cannot answer efficacy", "independentReplication": "none", "sampleSize": "Actual n=4 versus protocol target n=212", "effect": "Accrual reached 1.89% of target; no efficacy or safety result posted", "limits": [ "Registry was last updated in November 2025", "Reasons for poor accrual are not public", "Participant disposition is unknown", "This does not establish lack of drug activity" ], "dimensions": { "biologicalPlausibility": { "rating": "weak", "reason": "Cell, mouse and short human exposure work motivated the treatment, but none establishes long-term IPMN prevention." }, "animalEvidence": { "rating": "moderate", "reason": "The protocol cites mouse evidence, but the clinical programme never recruited enough people to test translation." }, "inVitroEvidence": { "rating": "moderate", "reason": "Laboratory activity supports a testable idea, not efficacy in people." }, "reproducibility": { "rating": "none", "reason": "There is no adequately sized randomized result to repeat." }, "commercialConflicts": { "rating": "moderate", "reason": "American River Nutrition supplied collaboration to a product trial, while NCI and academic centres also participated." }, "recency": { "rating": "recent", "reason": "The active-not-recruiting record is current to November 2025 and had not posted results by September 2026." } } }, "layout": { "x": 22, "y": 104 }, "updated": "2026-09-14" }, { "id": "failure-smart-trial-and-vendor-dependency", "title": "SMART evidence depended on one vendor and its decisive trial never started", "shortTitle": "SMART trial-system failure", "type": "failure", "status": "supported", "scope": [ "pdac", "trial-design", "radiation", "business-continuity" ], "summary": "The proposed 267-person survival trial remains not yet recruiting on a stale record; its named machine sponsor entered bankruptcy, while a separate randomized pain trial stopped at nine people.", "content": "LAP-ABLATE has the comparison the field needs: people with locally advanced pancreatic cancer who have not progressed after induction chemotherapy would be randomized to continued care with or without 50 Gy in five SMART fractions. Overall survival at two years is primary. But the registry remains not yet recruiting, still shows an estimated July 2023 start, and has not been updated since 19 October 2022. Its sponsor and named machine maker was ViewRay. ViewRay reported $81.3m unrestricted cash, $77.8m debt and $53.6m operating cash use in the first three months of 2023. If that burn repeated, simple cash runway was $81.3m / $53.6m x 3 = 4.55 months. Management warned of substantial going-concern doubt, and the company filed Chapter 11 on 16 July 2023. The clinical evidence plan therefore had a single-vendor failure domain. The phase 2 registry also lists 133 estimated participants while its papers report 136 and has no posted result table. MASPAC, an academic randomized test of SMART for pain in metastatic disease, terminated for low accrual after nine people. These are different failures—corporate continuity, registry integrity and patient recruitment—but they converge on one lesson: a mission cannot make a survival question depend on one machine brand, one insolvent sponsor or an unmaintained registry.", "sourceIds": [ "clinicaltrials-nct03621644-smart", "clinicaltrials-nct05585554-lap-ablate", "clinicaltrials-nct05114213-maspac", "viewray-2023-q1-10q", "viewray-2023-chapter11-8k" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "claim-smart-delivery-and-safety-without-added-survival-proof", "relation": "limits" }, { "target": "unknown-smart-added-survival-and-surgery-harm", "relation": "creates" }, { "target": "hypothesis-smart-vendor-independent-randomized-gate", "relation": "repaired-by" }, { "target": "trial-pdac-smart-action-map", "relation": "mapped-by" }, { "target": "company-viewray", "relation": "caused-by" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "weak", "independentReplication": "none", "sampleSize": "LAP-ABLATE target n=267 and not started on registry; MASPAC n=9 terminated; phase 2 n=136 paper versus 133 registry", "effect": "The planned survival comparison did not recruit; the randomized pain study stopped at nine", "limits": [ "LAP-ABLATE registry is stale", "No formal registry link states bankruptcy caused non-start", "MASPAC stopped for low accrual rather than vendor failure", "The machinery and intellectual property may have successor owners", "Phase 2 count mismatch needs reconciliation", "No posted result tables" ], "dimensions": { "biologicalPlausibility": { "rating": "not-applicable", "reason": "This is a business, recruitment and evidence-governance failure." }, "animalEvidence": { "rating": "not-applicable", "reason": "Model evidence cannot repair a failed trial network." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Cell evidence cannot answer operational continuity." }, "reproducibility": { "rating": "none", "reason": "The needed treatment-versus-no-treatment survival comparison never began." }, "commercialConflicts": { "rating": "high", "reason": "The machine maker sponsored both the single-arm phase 2 and dormant phase 3 plan before bankruptcy." }, "recency": { "rating": "current", "reason": "Trial records were checked after MASPAC's November 2025 closure." } } }, "layout": { "x": 682, "y": 606 }, "updated": "2026-09-15" }, { "id": "failure-sonoporation-overall-replication-subgroup-rescue", "title": "An incomplete subgroup cannot rescue a failed overall sonoporation replication", "shortTitle": "Sonoporation subgroup rescue", "type": "failure", "status": "supported", "scope": [ "pdac", "pan-cancer", "transfer-unproven" ], "summary": "The second randomized trial missed its overall PFS and OS comparisons; a large subgroup difference lacks the information required to distinguish treatment interaction from chance or imbalance.", "content": "The scale-up path looked persuasive: ten treated people outlived historical controls, then a 78-person randomized trial reported a three-month survival gain. The independent NCT04821284 abstract breaks that path. Among 97 enrolled people, neither overall progression-free survival nor overall survival differed, both P>0.36. The abstract then highlights gemcitabine/nab-paclitaxel median survival of 407.5 versus 234 days. That is 173.5 days, or 74.1% relative to 234 days. Size does not remove bias. The public report omits the subgroup denominator, confidence interval, stage balance, randomization balance within the subgroup, number of subgroup cuts, whether the cut was planned, and the treatment-by-chemotherapy interaction. A within-subgroup P value would still not prove that the treatment effect differs by backbone; the interaction is the required test. The registry also mixes stage II, III and IV disease and two chemotherapy systems. Repair means release the locked analysis plan and complete individual-level dataset, report the intended-to-treat flow, overall hazard ratios and confidence intervals, quality of life, toxicity, treatment exposure, stage and backbone strata, and the prespecified interaction. Do not start another trial or sell a device from the subgroup. If the interaction was not planned or does not reproduce, close the efficacy branch.", "sourceIds": [ "dimcevski-2016-sonoporation-phase1", "wang-2023-sonochemotherapy-pdac", "clinicaltrials-nct04821284-sonoporation", "ius-2026-sonoporation-nct04821284-abstract" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "claim-sonochemotherapy-randomized-signal", "relation": "corrects" }, { "target": "failure-local-intensification-pdac-survival-not-shown", "relation": "replicates-pattern-of" }, { "target": "unknown-pdac-sonoporation-treatment-effect", "relation": "leaves" }, { "target": "trial-pdac-sonoporation-action-map", "relation": "repaired-by" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "not-applicable", "randomised": "moderate — overall negative direction is reported, but only in an accepted abstract", "independentReplication": "weak — negative overall result is abstract-only", "sampleSize": "NCT04821284 abstract n=97 enrolled; registry estimates n=120", "effect": "Overall PFS and OS both P>0.36; highlighted subgroup median OS 407.5 versus 234 days without denominator or uncertainty", "limits": [ "Accepted abstract, not full final report", "Overall estimates and event counts missing", "Subgroup analysis details missing", "Registry and abstract enrollment differ", "No posted results record" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "A backbone-specific delivery effect is possible, but plausibility cannot identify a treatment interaction." }, "animalEvidence": { "rating": "weak", "reason": "Human randomized non-replication now outweighs model support for an overall benefit claim." }, "inVitroEvidence": { "rating": "weak", "reason": "The failure is statistical and clinical, not a claim that bubbles cannot affect permeability." }, "reproducibility": { "rating": "weak", "reason": "The first overall survival signal did not repeat overall in the independent trial." }, "commercialConflicts": { "rating": "moderate", "reason": "A large device-ready subgroup creates a strong incentive to foreground it after an overall negative result." }, "recency": { "rating": "current", "reason": "Based on the 2026 accepted abstract and current registry." } } }, "layout": { "x": 142, "y": 118 }, "updated": "2026-09-15" }, { "id": "failure-sorafenib-adjuvant-hcc", "title": "Sorafenib activity in advanced HCC did not prevent recurrence after local cure", "shortTitle": "STORM failure", "type": "failure", "status": "supported", "scope": [ "hcc" ], "summary": "In STORM, recurrence-free survival was 33.3 months with sorafenib and 33.7 with placebo after resection or ablation.", "content": "The transfer assumption was that a drug active in advanced HCC might eradicate occult disease after resection or ablation. STORM randomised 1,114 patients with complete radiological response. Median recurrence-free survival was 33.3 months with sorafenib and 33.7 months with placebo; hazard ratio 0.94. Dose modification was required in 89% on sorafenib, and hand-foot skin reaction was common. This rejects adjuvant sorafenib for this population. It does not mean residual HCC is untreatable. Recurrence can arise from surviving tumour cells or new cancers in a damaged liver field; an advanced-disease growth inhibitor may be poorly matched to either low-burden state. The next programme needs to classify recurrence origin and test an intervention with a mechanism active in that state.", "sourceIds": [ "bruix-2015-storm-sorafenib" ], "links": [ { "target": "hcc", "relation": "part-of" }, { "target": "failures", "relation": "part-of" }, { "target": "residual-disease", "relation": "supports" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong", "independentReplication": "moderate", "sampleSize": "1,114 randomised after resection or ablation", "effect": "Median recurrence-free survival 33.3 versus 33.7 months; HR 0.94", "limits": [ "Does not separate metastatic recurrence from new primary HCC", "High dose-modification rate", "Older treatment era", "Bayer-funded", "Failure of sorafenib is not failure of every residual-disease strategy" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "Activity against measurable advanced HCC did not guarantee clearance of occult clones or prevention of new primary tumours." }, "animalEvidence": { "rating": "moderate", "reason": "Models supported sorafenib activity but did not predict the adjuvant failure." }, "inVitroEvidence": { "rating": "moderate", "reason": "Target inhibition established drug activity but not postoperative recurrence prevention." }, "reproducibility": { "rating": "moderate", "reason": "IMbrave050 later reproduced the broader failure of moving an advanced-disease regimen into unselected adjuvant HCC." }, "commercialConflicts": { "rating": "high", "reason": "Bayer funded the trial and owned the drug interest." }, "recency": { "rating": "historical", "reason": "STORM was published in 2015 but remains a decisive negative adjuvant result." } } }, "layout": { "x": 37, "y": 75 }, "updated": "2026-09-14" }, { "id": "failure-ssa-hcc-curative-route-not-reaching-patients", "title": "Sub-Saharan HCC care fails before most patients reach a curative route", "shortTitle": "SSA HCC route failure", "type": "failure", "status": "supported", "scope": [ "hcc", "hepatobiliary" ], "summary": "Across 3,989 reported patients, 6% received curative treatment and 84% supportive care alone; Cape Town and Sweden comparisons show that stage, liver disease and access differ too much to call this a drug-only failure.", "content": "Original belief to reject: importing a high-income HCC treatment guideline is enough if clinicians know the standard. What happened: a registered systematic review found 39 reports from only 15 of 48 sub-Saharan African countries and 3,989 patients. Curative treatment reached 6%; 84% received best supportive care and few survived one year. In a Cape Town comparison, 384/551 = 69.7% of public patients versus 8/51 = 15.7% of private patients received supportive care alone, while median survival was 68 versus 703 days. Public patients were younger and had far more HBV, multifocal disease, vascular invasion, metastasis and late BCLC stage. Within the small selected resection and ablation groups, survival did not differ significantly, showing that some public-sector technical capability exists after selection; it does not prove equal care or a causal benefit from insurance. A later South Africa-Sweden comparison found curative-intent treatment in 9.2% versus 42.5% and one-year survival of 17.8% versus 59.3%; survival remained lower within treatment groups. Why the route fails: prevention, surveillance, diagnosis, liver-function preservation, referral, staging and treatment access can all remove eligibility before a drug choice is made. The two comparisons do not isolate their shares. The surviving action is to measure every lost step and test a route-completion bundle. Do not fund a new regional drug trial until consecutive registration can show who was eligible, what was recommended, what was delivered and why it was not delivered.", "sourceIds": [ "sobnach-2024-ssa-hcc-systematic-review", "sobnach-2025-south-africa-hcc-disparities", "sobnach-2026-global-hcc-disparities", "jonas-2025-sub-saharan-hcc-pathways" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "lab-uct-groote-schuur-hpb", "relation": "observed-at" }, { "target": "unknown-regional-route-loss", "relation": "leads-to" }, { "target": "hypothesis-regional-route-completion", "relation": "motivates" }, { "target": "model-hcc-causal-chain", "relation": "extends" } ], "evidence": { "humanProspective": "none for the proposed route intervention", "humanRetrospective": "strong for a large delivery gap; weak for its cause", "randomised": "none", "independentReplication": "moderate — a systematic review and two cohort comparisons point in the same direction", "sampleSize": "Systematic review n=3,989; South African public/private cohort n=602; South Africa/Sweden cohort n=959", "effect": "Curative treatment 6% and supportive care 84% across reviewed SSA reports; Cape Town median survival 68 versus 703 days; curative-intent care 9.2% versus 42.5% in South Africa versus Sweden", "limits": [ "Observational comparisons", "Referral-centre selection", "Confounding by stage, liver function, aetiology, age and calendar time", "Small private Cape Town cohort", "Treatment-group comparisons condition on selection", "Countries outside the observed sites may differ", "No tested route intervention" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Earlier stage and preserved liver function directly determine access to resection, ablation and transplantation." }, "animalEvidence": { "rating": "not-applicable", "reason": "The failure is in human prevention, diagnosis and care delivery." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Laboratory systems cannot establish route completion." }, "reproducibility": { "rating": "moderate", "reason": "Three human evidence sets show poor curative access, but none isolates the causal step." }, "commercialConflicts": { "rating": "low", "reason": "The comparative cohorts report foundation rather than product-company funding; the systematic evidence is not a product test." }, "recency": { "rating": "current", "reason": "The latest comparative cohort was published in 2026." } } }, "updated": "2026-09-14" }, { "id": "failure-staging-yield-without-patient-utility", "title": "Staging-laparoscopy studies often count discoveries without testing patient benefit", "shortTitle": "Staging yield is not utility", "type": "failure", "status": "supported", "scope": [ "pdac" ], "summary": "The field repeatedly combines visible spread with cytology and treats a positive result as success without comparing delay, harm, useful treatment or survival.", "content": "Original belief: if staging laparoscopy finds more hidden disease, the test has proved its value. Observed result: modern studies report yields from about 10% to 24%, but nearly all are single-arm or use staging by centre, surgeon or treatment route. The largest Mayo series combined visible or biopsy-confirmed metastases with positive abdominal-wash cytology in one 18% endpoint, although 40/96 cytology-positive people had no visible spread. PREOPANC-2 found a large 12.6-point association with fewer non-therapeutic laparotomies, but staging itself was not randomized. The live Japanese UMIN000061741 study plans 200 people in a single arm and makes hidden-metastasis detection its primary outcome. That repeats the measurement gap: it can estimate yield and immediate harm, but not whether the timing or resulting treatment improves survival, days outside hospital or quality of life. The 2026 global survey found that 82% of 617 surgeons use staging, yet timing, selection and adjuncts vary across regions. Why this fails: discovery, prognosis and treatment benefit are three different questions; visible biopsy-proven spread and cytology-only spread are different states; selected referral centres distort yield; patients never referred disappear from the denominator; and a separate procedure can delay treatment. Recovery: begin with every scan judged potentially treatable, state whether staging occurs before treatment or in the same anaesthetic as surgery, record contraindications and delay, separate direct view, biopsy and cytology, preserve negative people who later prove metastatic, and measure the treatment action and patient outcome caused by each result.", "sourceIds": [ "gudmundsdottir-2023-staging-laparoscopy", "rompen-2026-preopanc2-staging", "stoop-2026-global-staging-survey", "umin-000061741-staging-laparoscopy", "kikuchi-2026-staging-laparoscopy" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "claim-staging-laparoscopy-avoids-incision-not-proven-survival", "relation": "constrains" }, { "target": "unknown-pdac-staging-laparoscopy-rule", "relation": "leads-to" }, { "target": "hypothesis-pdac-two-timepoint-staging-route", "relation": "recovered-by" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "strong", "randomised": "weak", "independentReplication": "moderate", "sampleSize": "Mayo n=1,004; PREOPANC-2 explored n=322; survey n=617; live Japanese target n=200", "effect": "Detection yield is repeatedly measured; comparative patient utility and cytology-only action are not", "limits": [ "Failure is across study designs", "No direct randomized utility trial", "Practice survey is self-report", "Patient preferences poorly recorded", "Changing imaging and treatment" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "A test result creates benefit only through a treatment decision that improves outcomes after procedure harm and delay are counted." }, "animalEvidence": { "rating": "not-applicable", "reason": "The failure concerns human study design and care delivery." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Cell systems cannot test referral, delay, incision or decision value." }, "reproducibility": { "rating": "moderate", "reason": "The outcome gap is visible across several independent cohorts and the new single-arm registry." }, "commercialConflicts": { "rating": "low", "reason": "No single manufacturer controls the standard procedure, though local procedural incentives can remain." }, "recency": { "rating": "current", "reason": "A 2026 survey and live 2026 registry show that the design gap remains active." } } }, "updated": "2026-09-15" }, { "id": "failure-tigerpac-component-and-interim-inference", "title": "TIGeR-PaC cannot isolate delivery and its first interim estimate did not cross the stopping rule", "shortTitle": "TIGeR-PaC inference limits", "type": "failure", "status": "observed", "scope": [ "pdac", "trial-delivery", "transfer-unproven" ], "summary": "The trial compares two treatment packages, changed its event plan after launch and reported a near-threshold 45-person interim result without a hazard ratio or confidence interval.", "content": "Question claimed: does local artery delivery overcome poor pancreatic tumour access? Question randomized: after about four months of induction chemotherapy and radiotherapy, does eight fortnightly artery-delivered gemcitabine treatments beat four cycles of intravenous gemcitabine plus nab-paclitaxel in non-progressing people who still cannot have surgery? Route, second drug and schedule all change together. A positive answer proves package value, not the catheter's separate contribution. The first interim analysis used all 45 randomized people then available: 23 experimental and 22 control, with 13 deaths in each arm. Median survival from randomization was 16 versus 10 months, P=0.051. The sponsor release gives no hazard ratio or confidence interval and notes that both medians omit the four-to-five-month induction period. The result did not cross the early stopping rule. A monitoring committee recommendation to continue means the trial may proceed; it does not mean the device works. The plan also changed after launch: analyse only stereotactic-radiation induction, add a second interim look, reduce the randomized target from 200 to 114, reduce required deaths to 86 and reduce power from 90% to 80%. Such changes can be valid, but the final inference needs the original and changed plans, timing, regulator correspondence, the alpha-spending rule that controls false positives across repeated looks, and every randomized patient. Early uncontrolled safety also matters: among 23 people with available safety data there were 17 severe-or-worse events, 19 procedure-related events, sepsis in 4/23 = 17.4% and one ischemic injury requiring surgery. Lower chemotherapy toxicity cannot be presented without adding catheter and infection harm.", "sourceIds": [ "clinicaltrials-nct03257033-tigerpac", "renovorx-2023-tigerpac-interim", "renovorx-2025-10k", "hatoum-2024-tamp-rr1-rr2" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "claim-tamp-blood-exposure-not-tumour-delivery", "relation": "constrains" }, { "target": "unknown-tigerpac-package-survival-and-delivery", "relation": "creates" }, { "target": "hypothesis-tigerpac-existing-final-gate", "relation": "repaired-by" }, { "target": "trial-pdac-intraarterial-delivery-action-map", "relation": "changes" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "weak", "randomised": "moderate", "independentReplication": "weak", "sampleSize": "First interim n=45 with 26 deaths; early safety n=23; final plan targets n=114 randomized and 86 deaths", "effect": "Interim median survival 16 versus 10 months, P=0.051, without hazard ratio or confidence interval; component contribution remains unidentified", "limits": [ "Open-label trial", "Sponsor-only interim report", "Second interim result withheld", "Final analysis plan unavailable", "Final randomized denominator not disclosed", "Early safety studies were uncontrolled" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Local pressure delivery is physically plausible, but plausibility cannot isolate the package components." }, "animalEvidence": { "rating": "moderate", "reason": "Animal vessel transport does not resolve the human package comparison." }, "inVitroEvidence": { "rating": "strong", "reason": "Gemcitabine is active in pancreatic cancer systems, but the route contribution is the issue." }, "reproducibility": { "rating": "weak", "reason": "No independent modern randomized TAMP result exists and the final trial is unread." }, "commercialConflicts": { "rating": "high", "reason": "The device owner controls the interim releases and depends on the trial for expansion." }, "recency": { "rating": "current", "reason": "The plan and current completion state are traced through August 2026." } } }, "layout": { "x": 526, "y": 450 }, "updated": "2026-09-15" }, { "id": "failure-tocilizumab-cachexia-muscle-without-survival", "title": "Tocilizumab preserved muscle in PDAC but did not improve the main survival endpoint and increased severe toxicity", "shortTitle": "IL-6 cachexia failure", "type": "failure", "status": "refuted", "scope": [ "pdac" ], "summary": "The trial changed a host measure without changing six-month survival, median survival, progression or response; grade 3 or worse treatment toxicity rose by 24.7 points.", "content": "Original belief: selecting inflamed advanced-PDAC patients and adding IL-6 receptor blockade to gemcitabine plus nab-paclitaxel would reduce cachexia enough to improve survival. What happened: among 147 treated patients including the safety run-in, six-month survival was 68.6% with tocilizumab and 62.0% with placebo, P=.409. Median survival, progression-free survival and response did not differ. Several early muscle measures improved and an exploratory 18-month survival comparison favoured the intervention. Grade 3 or worse treatment-related events rose from 63.4% to 88.1%, an absolute increase of 24.7 percentage points. Why it probably failed: IL-6 is one part of a multi-route wasting system, the added immune toxicity may consume treatment capacity, and muscle quantity may not mediate enough of the death risk. The positive exploratory results do not overturn a failed primary endpoint. What survives is the need to measure host function. What does not survive is using muscle preservation alone as proof of cancer benefit. A successor needs a narrower biological group, lower added toxicity and a locked chain from function to treatment delivered to survival or patient-valued time.", "sourceIds": [ "chen-2025-tocilizumab-cachexia" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "claim-il6-cachexia-signal", "relation": "bounds" }, { "target": "claim-cachexia-is-parallel-lethal-system", "relation": "constrains" }, { "target": "trial-pdac-cachexia-action-map", "relation": "informs" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong", "independentReplication": "none for the exact intervention", "sampleSize": "147 treated including a 6-person safety run-in", "effect": "Six-month OS 68.6% versus 62.0%, P=.409; no median OS, PFS or response gain; grade 3+ treatment-related events 88.1% versus 63.4%", "limits": [ "Phase 2", "Inflammation-selected advanced disease", "Exploratory late-survival and muscle endpoints", "Multiple comparisons", "One IL-6 strategy", "No independent repeat" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "IL-6 can contribute to inflammatory wasting, but the human result shows it is not a sufficient low-toxicity route in this setting." }, "animalEvidence": { "rating": "moderate", "reason": "Model support cannot rescue the failed human survival endpoint." }, "inVitroEvidence": { "rating": "moderate", "reason": "Pathway effects do not establish whole-patient survival benefit." }, "reproducibility": { "rating": "none", "reason": "No independent randomized replication of this treatment strategy was identified." }, "commercialConflicts": { "rating": "moderate", "reason": "The trial used a commercial antibody, but the negative primary result is clearly reported." }, "recency": { "rating": "current", "reason": "The randomized result was published in 2025." } } }, "layout": { "x": 137, "y": 80 }, "updated": "2026-09-14" }, { "id": "failure-vascular-starvation-pdac", "title": "Four large PDAC trials reject simple starvation of tumour blood vessels", "shortTitle": "Vessel-starvation failure", "type": "failure", "status": "supported", "scope": [ "pdac", "pan-cancer", "transfer-unproven" ], "summary": "Bevacizumab, aflibercept and axitinib failed overall survival across four phase 3 tests containing 2,387 people.", "content": "System idea. VEGF tells tumours to grow blood vessels. Blocking VEGF or its receptor aims to starve the cancer, but pancreatic tumours already have compressed, uneven, poorly working vessels. Removing more vessel signal can lower perfusion and drug delivery as well as tumour supply. Human result. CALGB 80303 enrolled 602 people: bevacizumab plus gemcitabine gave median survival 5.8 months versus 5.9 with placebo, P=.95. A 607-person bevacizumab trial added the drug to gemcitabine/erlotinib: progression improved, HR 0.73, but survival did not, HR 0.89, 95% CI 0.74-1.07. VANILLA randomized 546: aflibercept produced survival 6.5 versus 7.8 months, HR 1.165, and adverse-event stopping rose from 12% to 23%. Axitinib randomized 632 and stopped for futility, with survival 8.5 versus 8.3 months, HR 1.014. Across 602+607+546+632=2,387 people, no trial established longer life. Decision. Stop unselected VEGF starvation in PDAC. The result does not say blood flow is irrelevant; it says lowering a vessel-growth signal without measuring perfusion, oxygen and drug exposure is not a delivery solution. A normalization claim must show that a named dose improves working flow and preserves useful life, not merely changes a scan or delays progression.", "sourceIds": [ "kindler-2010-bevacizumab-calgb80303", "van-cutsem-2009-bevacizumab-erlotinib", "rougier-2013-aflibercept-vanilla", "kindler-2011-axitinib-pdac" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "failure-axitinib-vascular-targeting", "relation": "extends" }, { "target": "claim-hypoxia-control-without-survival-pdac", "relation": "adjacent-to" }, { "target": "failure-losartan-delivery-signal-not-replicated", "relation": "contrasts-with" }, { "target": "trial-pdac-vascular-hypoxia-action-map", "relation": "changes" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong", "independentReplication": "strong", "sampleSize": "Four phase 3 trials totaling 2,387 enrolled or randomized participants", "effect": "No OS benefit; one PFS-only effect, two futility stops and added hypertension, proteinuria or discontinuation", "limits": [ "Older chemotherapy backbones", "Three drug designs", "No prospective perfusion selector", "Does not test vessel-opening doses or non-VEGF routes" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Compressed and sparse PDAC vessels make further starvation capable of reducing delivery." }, "animalEvidence": { "rating": "moderate", "reason": "Models support both starvation and normalization, which is why direct human measurement matters." }, "inVitroEvidence": { "rating": "weak", "reason": "Cell systems cannot represent blood flow, pressure and systemic drug delivery." }, "reproducibility": { "rating": "strong", "reason": "Four large randomized trials failed overall survival across related vessel-targeting drugs." }, "commercialConflicts": { "rating": "moderate", "reason": "Industry funded several trials, but all decisive negative results are public." }, "recency": { "rating": "historical", "reason": "The trials are old but large, blinded where possible and not displaced by a positive modern PDAC test." } } }, "layout": { "x": 316, "y": 262 }, "updated": "2026-09-15" }, { "id": "failure-virage-analysis-and-survivor-selection", "title": "VIRAGE can be over-read through changing denominators and post-randomization selection", "shortTitle": "VIRAGE analysis risk", "type": "failure", "status": "mixed", "scope": [ "pdac" ], "summary": "The favourable phase 2 estimate may be real, but 112 enrolled became a 96-person FAS and the strongest two-dose result selects people who survived and remained fit through four chemotherapy cycles.", "content": "Original belief. Random allocation would make the VCN-01 survival estimate and its apparent two-dose gain straightforward. What happened. ClinicalTrials.gov records 112 actual participants. The sponsor's main result contains 96, 48 per arm. The public 2023 protocol planned 92 and defines ITT as randomized people who received any study medication; its FAS further requires at least one third of planned chemotherapy in both arms and a complete first VCN-01 dose in the virus arm. Those are events after allocation, so the public analysis is not a clean all-randomized comparison. The public documents do not show how the 16-person enrollment-to-FAS gap is split by screening, allocation, arm, treatment, reason or outcome. Alpha is also unclear: the protocol powers at one-sided 0.05, describes a two-sided log-rank test, then says ordered secondary testing requires OS P<0.05. The sponsor's OS log-rank P=0.0546 passes a two-sided 0.10 rule but fails 0.05; adjusted Cox P=0.034 differs. Why the dose claim fails. The 14.8 versus 11.6-month comparison includes only 34 virus-arm and 29 control patients who reached four chemotherapy cycles, and only virus-arm patients who received the second dose. To enter that comparison a person had to stay alive, fit and on treatment for about three months. This selects after randomization and cannot isolate what the second dose caused. What survives. The overall direction, HR 0.57 confidence interval and PFS result justify confirmation, while the dose subgroup only supports a dosing question. Recovery. Release the allocation flow for all 112, true all-randomized survival, every protocol and analysis-plan version, unadjusted and adjusted models, missingness, event dates, treatment exposure, quality of life, harms and biomarker records. Until then, do not market the phase 2 result as confirmed survival benefit or the subgroup as proof of repeat dosing.", "sourceIds": [ "clinicaltrials-nct05673811-virage", "theriva-2023-virage-protocol", "theriva-2025-virage-topline", "esmo-2025-virage-2216mo", "theriva-2026-q2-10q" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "claim-vcn01-randomized-signal-not-survival-proof", "relation": "constrains" }, { "target": "unknown-vcn01-all-enrolled-survival", "relation": "creates" }, { "target": "trial-vcn01-action-map", "relation": "repaired-by" }, { "target": "failure-oncolytic-virus-pelareorep", "relation": "distinct-from" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "not-applicable", "randomised": "moderate", "independentReplication": "weak", "sampleSize": "Registry n=112; reported FAS n=96; post-randomization dose subgroup n=63", "effect": "Sixteen-person enrollment-to-FAS gap; OS log-rank P=0.0546 versus adjusted Cox P=0.034; selected dose subgroup HR 0.44", "limits": [ "Enrollment may include people not randomized", "Final analysis plan unavailable", "Participant-level data unavailable", "Some subgroup analysis was named in the protocol but remains post-randomization" ], "dimensions": { "biologicalPlausibility": { "rating": "not-applicable", "reason": "This record tests the analysis, not the virus mechanism." }, "animalEvidence": { "rating": "not-applicable", "reason": "Animal results cannot repair human denominator bias." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Laboratory results cannot identify excluded human outcomes." }, "reproducibility": { "rating": "weak", "reason": "Only sponsor summaries and an abstract are public." }, "commercialConflicts": { "rating": "high", "reason": "The sponsor chose and reported the available analysis sets." }, "recency": { "rating": "current", "reason": "The registry and current filing still do not post full results." } } }, "layout": { "x": 172, "y": 92 }, "updated": "2026-09-15" }, { "id": "failure-vopimetostat-ras-contribution-unknown", "title": "Vopimetostat combination data cannot identify the added treatment effect", "shortTitle": "PRMT5 contribution gap", "type": "failure", "status": "mixed", "scope": [ "pdac" ], "summary": "A 64% combined response calculation is striking, but the study lacks a RAS-inhibitor-only arm and joins two drugs, several doses and short follow-up.", "content": "Claim being tested. Tango describes the PRMT5 and RAS inhibitor results as support for rapid phase 3 development. Observed data. Fifty-four PDAC patients had started a combination by 28 May 2026: 20 with daraxonrasib and 34 with zoldonrasib. Response analysis required at least 14 weeks since first dose, leaving 12 and 27 evaluable. Daraxonrasib combination ORR was 11/12, with nine confirmed; zoldonrasib combination ORR was 14/27, with ten confirmed. The pooled calculation is 25/39=64.1%, but it is descriptive, not an estimate of added effect. Causal fault. There is no concurrent daraxonrasib-alone or zoldonrasib-alone arm. The two groups use different RAS inhibitors and several dose levels, and 15 started patients were not yet response evaluable. Historical single-agent rates cannot correct those differences. A good response may come from the RAS inhibitor, PRMT5 inhibitor, their interaction or selection of people well enough for trial entry. Safety boundary. Three dose-limiting events occurred in two people at the higher vopimetostat plus daraxonrasib dose; no grade 4 or 5 related events were reported at cutoff. Short exposure cannot settle cumulative low blood counts or duration. Repair. Publish all 54 PDAC records as they mature, with confirmed responses, duration, dose intensity, progression and harms. Then randomize each RAS drug against the same RAS drug plus vopimetostat inside exact MTAP-loss strata. For second-line RAS-G12 disease, daraxonrasib alone is the direct comparator; chemotherapy alone cannot isolate vopimetostat's contribution. A front-line phase 3 against chemotherapy may answer whether the package works, but not which component earns its toxicity and cost.", "sourceIds": [ "tango-2025-vopimetostat-monotherapy", "clinicaltrials-nct06922591-vopimetostat-ras", "tango-2026-vopimetostat-ras-data" ], "links": [ { "target": "failures", "relation": "part-of" }, { "target": "claim-mtap-prmt5-selected-human-signal", "relation": "bounds" }, { "target": "unknown-mtap-prmt5-ras-contribution", "relation": "creates" }, { "target": "trial-mtap-prmt5-action-map", "relation": "repaired-by" }, { "target": "claim-daraxonrasib-common-ras-survival-breakthrough", "relation": "requires-as-comparator" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "not-applicable", "randomised": "none", "independentReplication": "none", "sampleSize": "54 treated PDAC; 39 response evaluable after at least 14 weeks", "effect": "25 of 39 evaluable responded across two combination cohorts; 19 responses confirmed at cutoff", "limits": [ "No concurrent control", "Fifteen treated patients not yet response evaluable", "Short follow-up", "Mixed drugs and doses", "Sponsor report rather than peer-reviewed full paper" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The selected dependency and both drug targets are real, but plausibility cannot allocate the observed response." }, "animalEvidence": { "rating": "moderate", "reason": "Models support combination activity but cannot estimate the added human effect." }, "inVitroEvidence": { "rating": "strong", "reason": "MTAP loss, PRMT5 inhibition and RAS inhibition can each be measured." }, "reproducibility": { "rating": "none", "reason": "No controlled or independent study reproduces the combination estimate." }, "commercialConflicts": { "rating": "high", "reason": "Tango controls vopimetostat, the trial and interim analysis." }, "recency": { "rating": "current", "reason": "Uses the May 2026 cutoff disclosed in June." } } }, "layout": { "x": 186, "y": 107 }, "updated": "2026-09-15" }, { "id": "failures", "title": "Failure evidence", "shortTitle": "Failures", "type": "failure", "status": "supported", "scope": [ "pdac", "pnet", "cholangiocarcinoma", "gallbladder", "hcc", "colorectal", "pan-cancer" ], "summary": "102 failure records reveal repeated causes that future programmes must explicitly escape.", "content": "102 failure records reveal repeated causes, not unrelated disappointments. Wrong endpoint: PEGPH20 improved some response measures without survival; hydroxychloroquine raised response without extending life; ELI-002 foregrounded immune activity and a subgroup after missing its randomized disease-free-survival endpoint. Wrong biological simplification: broad fibroblast depletion worsened disease in models, hyaluronidase failed phase 3, CTGF blockade failed two randomized settings, and CCR2 inhibition added pulmonary toxicity. Wrong patient selection or unproved selector: PARP maintenance was confined to germline-BRCA disease and did not show overall-survival benefit; napabucasin's proposed biomarker did not rescue efficacy; unselected checkpoint blockade produced almost no responses. Wrong intermediate marker: CXCR4 plus PD-1 treatment increased T-cell infiltration without producing an objective response in 21 reported patients, while vaccine plus dual-checkpoint treatment generated tumour-specific T cells but only two responses among 57 and 68% severe related toxicity. Both programmes also enriched suppressive myeloid states. T-cell entry is therefore a measurement, not a success condition. Local treatment against systemic risk: LAP07 and other radiation intensification did not show a general survival solution, although PANOVA-3 and a small sonochemotherapy trial keep selected physical treatment open. EPOCH supplies the cross-cancer warning: adding yttrium-90 radioembolization to chemotherapy improved liver and overall progression measures in colorectal liver metastases but left median survival at 14.0 versus 14.4 months and increased grade 3 events. CLOCC and TransMet show why this does not invalidate all local treatment: much narrower organ-confined groups can gain a survival tail. Delivery failure: AVATAR assigned 81 patients to a complex precision system but only four received personalized treatment. A larger regional failure now sits beside it: across 3,989 reported sub-Saharan African HCC patients, 6% received curative treatment and 84% supportive care alone. Treatment guidance cannot work for patients lost before diagnosis, referral or eligibility. Non-replication: the near-perfect GPC1 exosome classifier did not reproduce. Commercial scaling ahead of utility: GRAIL's $7.1bn acquisition preceded a failed randomized combined stage III/IV endpoint and still no mortality result. Scientific-governance failure: a high-profile mouse triple-therapy paper was retracted for undisclosed company interests and then republished with declarations; the data remain testable, but independent reproduction is mandatory. Some mechanisms survive their products. PDAC still uses stroma and metabolism, but crude bulk depletion and weak repurposed drugs failed. Immune treatment may work at low burden or in a selected molecular state, but generic checkpoint addition or trafficking alone failed. Local control may help a selected subgroup, but systemic-risk selection is missing. Every new programme must state which prior failure it escapes by molecule, exposure, patient rule, measurement or delivery. If it cannot name that difference, it is repetition. Germline testing adds a two-stage delivery failure. A guideline or automated referral still left patient completion between 38% and 82% in several services, and the route then lost pathogenic-result counselling and relatives. In one prospective pancreatic cohort, 84% of first-degree relatives were informed but only 31% tested. GENERATE shows that covered home testing works after motivated relatives enter; GIFT shows why that is insufficient, because only 19.2% of all reported relatives were invited. Recovery must measure every eligible patient and relative, use consented contact, and prohibit action on uncertain variants.", "contentSections": [ "Electrochemotherapy adds endpoint and custody failure. Eleven of 25 selected patients lacked six-month imaging because they died; electrode geometry was chosen, not assigned; and a stated 4.5-month median-survival difference conflicts with 12 - 6 = 6 months. The 90-person randomized trial transitioned registries without a public result. PanECT enrolled ten without a control and left survival outcomes unposted.", "Focused ultrasound adds a reporting and trial-continuity failure. A 120-person paper calls its groups randomly assigned in the abstract but retrospective in the methods. The first prospective randomized study has remained recruiting since 2017 with no result; a 60-person study is stale and unknown; PULS suspended recruitment after an industrial-partner decision. The repair is custody of assignments, device logs, systemic treatment and patient outcomes, not another attractive local-control cohort.", "PELICAN closes a local-ablation selection loop. Observational series could make survival look favourable after excluding people whose hidden spread or technical state prevented treatment. Random assignment preserved that route: only 70 of 95 assigned to ablation received it, 21 had hidden metastases found at surgery, and 81% had a laparotomy complication. Median survival did not improve and severe serious events rose from 11% to 27%. Future device studies must keep every assigned patient and every failed delivery in the denominator.", "ASPIRE adds trial-operating failure. A sound masked placebo design reached a reported 395-person safety database, then its contractor began termination after unpaid invoices. Sponsor cash was $142,000 against $12.5m nine-month operating use; later a $12.0m note default was disclosed. The U.S. registry says recruiting while Europe says ended, and no actual assigned count or result is public. Preserve the comparison before telling another efficacy story.", "Hidden-spread staging adds a measurement failure. Large cohorts find disease hidden from scans and PREOPANC-2 links a camera check to fewer open explorations, but studies often merge visible biopsy-confirmed spread with positive wash cytology and stop at detection yield. The recovery route measures delay, harm, the exact action caused by each result and patient outcomes; the live Japanese yield-only study cannot close that gap.", "The failure database now contains 52 records. Two precursor failures concern trial delivery rather than drug biology. EA2185 closed both randomized surveillance arms and changed from a planned 4,606-person clinical-effectiveness comparison to an estimated 770-person observational biomarker cohort; the reason is not public. SIPP-T3 stopped recruiting after four of 212 planned participants and uses a progression composite that can count more than 1 mm of cyst growth without HGD or cancer. Neither failure proves its scientific idea false. Both prove that accrual, endpoint choice and preservation of the original decision question must be gated before large follow-on spending.", "Gallbladder prevention adds two different failures. The revised European polyp rule still produced only one HGD and no cancer among 68 indicated operations, showing that repeated size measurement is not a sufficient cancer-risk rule. NCT03762837 planned the needed 100,000-person incidence cohort but still says not yet recruiting, retains a passed 2022 primary date and has no results. The first is a selection failure; the second is an execution and reporting failure. Neither proves that a country-specific preventive-surgery rule is impossible.", "Small pNET adds two measurement failures. ASPEN is prospective but lets centre and patient preference choose observation or surgery, so it cannot estimate the counterfactual benefit of surgery; its 1,000-person registry now shows unknown status after a passed 2024 completion estimate and has no final posted result. A prospective MEN1 circulating mRNA assay separated some current disease groups but failed prediction of new tumour, growth and metastasis, with AUCs from 0.39 to 0.65. The first needs randomized action evidence and full follow-up. The second needs a different lesion-linked signal, not another cut point on the same score.", "Cholangiocarcinoma adds three route failures. Annual MRI/MRCP plus CA19-9 in PSC generated severe or progressive duct findings in 23.8% of 512 people but left median survival at 13 months after cancer. eCHEC can test 12-month behaviour and infection, not long-latency cancer incidence. Postoperative ctDNA repeatedly predicts recurrence, while every reported or registered study still stops before a randomized treatment or imaging action. These failures preserve the underlying mechanisms—risk follow-up, infection control and residual-disease detection—but reject claims of achieved cancer interception or utility.", "HCC adds two completed surveillance failures. Ultrasound every three months found more lesions 10 mm or smaller than every six months but did not increase the proportion of HCC 30 mm or smaller. AFP, AFP-L3 and DCP added to ultrasound did not improve early-stage HCC detection in 1,208 randomized adults. These results do not refute organised HCC surveillance. They refute the tested interval and marker-addition rules and show why late-stage disease, work-up harm and death must follow test performance.", "Colorectal residual disease adds two randomized action failures. DYNAMIC-III strongly separated risk but did not validate either its stage III de-escalation or escalation rule. ALTAIR found a promising disease-free-survival point estimate for trifluridine/tipiracil at molecular relapse, but the confidence interval crossed no effect, the primary endpoint failed and grade 3 or worse hematologic toxicity rose from 3.3% to 73.0%. These results preserve ctDNA as a detector and reject the named treatment rules.", "Regional delivery adds a system failure. Thirty-nine reports from only 15 of 48 sub-Saharan African countries contained 3,989 HCC patients; 6% received curative treatment and 84% supportive care alone. A Cape Town public/private comparison found median survival of 68 versus 703 days, while a South Africa/Sweden comparison found curative-intent treatment in 9.2% versus 42.5%. Stage, liver disease, age, aetiology and access differed sharply, so these are not causal treatment estimates. The failure is that most patients do not complete the route; the dominant repairable step is still unknown.", "Cachexia adds an endpoint-and-toxicity failure. Tocilizumab reduced early muscle loss in advanced PDAC but missed six-month survival, median survival, progression and response while grade 3 or worse treatment-related events rose from 63.4% to 88.1%. The result keeps host wasting open as a treatment target and rejects muscle preservation alone as proof of benefit.", "Pancreatic enzymes add two different failures. Randomized trials of 67 and 88 patients missed their eight-week weight or BMI endpoints and did not prove longer survival; opposite survival point estimates show why the small studies cannot settle utility. Care delivery also failed: prescribing in 22,860 English pancreatic-cancer records rose from 41% to 48% against a stated 100% standard, followed by a national alert that enzyme supply could not meet demand. The first rejects weight as a complete endpoint and broad treatment as a proven survival rule. The second shows that guidance is not delivery: recommendation, prescription, pharmacy fill, dose, use and product gaps must all be measured.", "PDAC clot prevention adds two execution failures, not a failed drug-class mechanism. In one 243-person cohort, 24% developed VTE, only 30% received ambulatory prophylaxis, treatment began a median 43 days after diagnosis and half the events occurred within two months. Teaching alone changed eligible prescribing from 0/14 to 1/16; an electronic prompt was followed by 9/16 receiving prophylaxis, but the project was too small to establish event benefit or safety. Separately, imPaCT-PRO planned 450 patients to test tinzaparin, VTE and PFS, yet its registry has unknown status after a December 2024 completion estimate and no result. The first failure requires a named seven-day assessment, pharmacy and persistence route. The second requires recovery of enrolment, outcomes and stop reasons before any duplicate survival study.", "Biliary obstruction adds two route failures. Routine four-to-six-week drainage with mainly plastic stents raised serious complications from 39% to 74% without lowering mortality; it rejects avoidable drainage delay, not urgent or neoadjuvant indications using modern devices. In a separate selected 162-person neoadjuvant cohort, cholangitis affected 20% and stent-failure interruption 16%, with much shorter survival associations that cannot prove causation. Metal stents lower reintervention, obstruction, treatment delay and cholangitis relative to plastic, while insertion alone still does not complete the route. The repair is a named indication-to-treatment owner and dated follow-up, not another device-only claim.", "The surgical-route pass adds two failures. LEOPARD-2 stopped after complication-related deaths reached 5/50 after laparoscopic Whipple versus 1/49 after open surgery, without faster recovery, despite training and centre-volume requirements. Separately, current England data show 38% of Whipple patients without recorded chemotherapy or chemoradiotherapy within 14 weeks, and older cohorts show additional loss during treatment. The first is a procedure-specific safety failure. The second is a whole-route delivery failure with progression, complications, frailty, patient choice and handoff loss still mixed.", "NORPACT-1 adds a treatment-route failure rather than a verdict against every neoadjuvant regimen. Only 61/77 assigned patients received treatment first, grade 3 or worse events were 58% versus 40%, and median survival favoured upfront surgery. The recovery rule is intention-to-treat accounting of every allocated patient, dose, delay, progression, operation and postoperative cycle; per-protocol responder subsets cannot erase route failure.", "The dormancy sample gap is a failure of time and tissue, not proof that dormancy is absent. One-time marrow, primary tumour and end-stage autopsy sit on different sides of the disease-free interval. None directly observes a residual cell, proves low cycling and ancestry, then follows it into recurrence. Recovery joins those states prospectively and forbids a treatment claim from a prognostic marker alone.", "GrantPax adds a care-routing failure. It planned 135 older metastatic-PDAC patients but enrolled 32, and clinicians placed 15/32, 47%, on treatment inconsistent with the study assessment. Information without a named decision owner and completed action did not create a reliable route.", "Conversion surgery adds a selection failure. TAPS, its 2026 restaging model and a 2025 matched cohort all show large survival gaps after resection, but response, fitness, hidden spread, centre and surgical judgement choose the operation and predict survival. PELICAN proves the practical danger of trusting such local-treatment associations: randomized ablation did not extend survival, raised serious events from 11% to 27% and worsened quality of life. The repair is allocation at the shared decision point, not a more elaborate matched cohort.", "Organoid treatment routing exposes a denominator failure. Reported sensitivity and specificity usually begin after a culture succeeds, while the patient route begins at biopsy. HOPE reached drug testing in 12/76 people, ORGANOPREDICT had complete evaluable follow-up in 34/87, PASS-01 generated organoids in 50% and AVATAR delivered matched treatment to 4/81 assigned people. A clinical claim must retain every failed, late, contaminated, unavailable and undelivered result.", "FAP-radioligand reports expose a three-gate collapse: PET visibility is treated as absorbed dose, and dose is treated as patient benefit. LuMIERE moved from 35 scanned to 27 treated to 25 response-assessed; nine treated people had pancreatic cancer, four were stable and none had a reported partial response. The 3BP-3940 response fraction was 34/51 = 66.7% among evaluable people but 34/88 = 38.6% against all treated, before mixed cancers, isotopes and other care are separated. FRONTIER stopped for a business decision and posts no result. Repair requires all denominators and a joined scan-dose-response table by cancer, followed by a concurrent control for useful-life claims.", "Autophagy and metabolism add a scale-up failure. Devimistat moved from 11 responses among 18 people at its phase 1 dose to no survival or progression gain in the 528-person AVENGER 500 randomized trial. ERK plus HCQ then produced four-month disease control in 1/20 people versus 1/19 with ERK inhibition alone, while its derived organoids showed no clear synergy and blood cells did not show consistent autophagy inhibition. A separate ERK-HCQ basket produced 0/18 pancreatic responses, and MEKiAUTO stopped for toxicity and lack of efficacy before phase 2. The repair is not to relabel a stronger molecule as progress. Recover the terminated 91-person direct-inhibitor dataset, prove blockade in tumour, preserve the active RAS-drug dose, and require a control.", "Sonoporation adds a replication and subgroup failure. A ten-person historical comparison and 78-person randomized survival signal justified a second randomized trial. Its accepted abstract reports no overall PFS or OS difference among 97 enrolled people, both P>0.36, then highlights gemcitabine/nab-paclitaxel survival of 407.5 versus 234 days without subgroup size, uncertainty, balance, prespecification or interaction. The repair is full intended-to-treat release and a planned treatment-by-backbone interaction, not another device cohort.", "The siG12D-LODER failure now has an active successor boundary. The first product missed overall survival in 37 randomized people and produced an exploratory G12D/V estimate from 11 treated and five controls. SIL204 changes chemistry, formulation and route, so it inherits neither efficacy nor complete safety. Its 403-person adaptive plan should not expand beyond 15-21 until multi-region tumour knockdown and systemic exposure are public; the CTIS no-data-sharing rule must be changed before mission support.", "VIRAGE adds an analysis failure risk, not a verdict that VCN-01 fails. The registry records 112 participants, while the main report analyses 96 under a protocol that removes people after allocation and conflicts on its alpha rule. The reported two-dose advantage further selects people who survived and stayed fit through four chemotherapy cycles. Keep the overall favourable signal, discard the causal dose claim, and recover every denominator plus the final dated analysis plan.", "Broad epigenetic treatment now has a multi-product clinical failure record. CI-994 and azacitidine maintenance produced no randomized survival gain; THU-decitabine produced one stable scan among eight; entinostat-nivolumab and azacitidine-pembrolizumab produced three responses each with short median progression. Measured cell-state or immune change cannot substitute for longer useful life.", "The vopimetostat-RAS result has a contribution failure, not an efficacy verdict. Twenty-five of 39 evaluable people responded, but the pooled number joins two RAS drugs and several doses, excludes 15 not yet evaluable treated people at cutoff, and has no RAS-alone control. Repair it with mature all-treated reporting and randomized component comparison.", "The elraglusib result is not a failed treatment; its evidence chain has three repairable failures. The main 2.9-month survival estimate removes 36 randomized people before treatment, and control removal is 20.4% versus 9.4%. The primary analysis changed after trial start without a new sample-size calculation. Survival improved while PFS, response, disease control and response duration did not. Lead with the all-randomized 1.7-month estimate, recover the dated analysis record and require independent blinded survival confirmation.", "Cell-cycle biology supplies a high-value negative control. CDKN2A-selected palbociclib produced no response or 16-week stability; palbociclib plus nab-paclitaxel missed its one-year-survival gate; randomized abemaciclib arms controlled less disease than chemotherapy; and randomized CHK1 inhibition did not improve survival despite planned exposure. Two participant organoids later predicted a CHK1 combination that produced no human response. Common pathway change, target exposure and model killing are not treatment effects.", "Antibody treatment adds an address-carrier-payload failure. High CLDN18.2 selected 393 people for GLEAM, yet zolbetuximab added no survival: HR 0.999, while serious treatment-emergent events rose 19.8 points. Mesothelin-directed anetumab produced zero responses among 14 evaluable people; CLDN18.2/4-1BB SHR-3821 produced zero confirmed responses among 11 pancreatic patients at its main doses; and four EBC-129 responses among 21 remain unconfirmed with severe neutropenia at the higher dose. A surface address, another cancer's benefit or model killing cannot be transferred across carriers and payloads. IBI343 and IBI389 keep only their own product-specific questions open.", "Engineered-cell treatment adds a route-and-denominator failure. NCT03323944 now counts 54 enrolled people before administrative termination, while its main mesothelin CAR-T paper accounts for nine infused people, zero objective responses and median survival of 6.7 weeks. Local injection found cells in tissue but did not create expansion, persistence or control. A separate multiantigen route infused 37/56 collected people, losing 33.9% before treatment. Satri-cel and KRAS TCR-T keep product-specific signals open, but every future claim must begin at marker and HLA screening, retain manufacture failure and distinguish cell presence from tumour killing.", "Immune reprogramming adds a target-to-patient failure. NIS793 reduced its intended TGF-beta, fibroblast and collagen signals, yet blinded phase 3 survival was 9.2 versus 11.2 months and serious harm rose 14.6 points. Cabiralizumab plus nivolumab gave worse progression than chemotherapy alone, while adding the pair to gemcitabine chemotherapy gave PFS HR 1.00 and OS HR 1.04. PRINCE and OPTIMIZE-1 measured immune changes or within-arm outcomes without a chemotherapy-only contribution test. Tissue change and marker correlation cannot substitute for cancer killing, durable control, useful life and acceptable harm.", "Tumour blood flow adds two failures. Four phase 3 vessel-starvation trials totaling 2,387 people found no overall-survival benefit from bevacizumab, aflibercept or axitinib. Opening vessels with losartan looked strong inside a 49-person chemotherapy-radiation-surgery route, then AFPAC found no response, progression or survival gain and stopped at 88 of 264 planned patients. These results reject unselected starvation and broad advanced losartan. They do not merge vessel opening with vessel removal, and they leave the existing 168-person localized losartan allocation to answer its own setting.", "Microbiome translation adds a comparator-and-endpoint failure. The 40-person randomized kestose pilot had no power calculation, found no significant response difference and leaned on within-arm marker changes. One antibiotic-checkpoint study enrolled zero; a live-bacterium study ended at 13 after sponsor insolvency; a planned 10-person stool-transfer study is suspended; and current 25- and 20-person single arms cannot isolate their microbial components. The repair is all-enrolled controlled analysis joining microbial function, active drug or immune change, survival, serious harm and resistant organisms.", "PDAC vaccination adds a repeated escalation failure. TeloVac converted early immune response into a 1,062-person phase 3 that found no survival gain. A 90-person GVAX/CRS-207 comparison against another unproved vaccine route reported HR 0.59; the 303-person ECLIPSE comparison against chemotherapy then reported HR 1.17 in its primary cohort. ELI-002 generated KRAS immunity but missed randomized disease-free survival and moved attention to a result-selected subgroup. The repair is an all-assigned standard-care comparison with patient outcome, complete manufacturing denominators and no responder-defined efficacy claim.", "Adenosine blockade adds a component-and-marker failure. ARC-8 randomized checkpoint addition while every arm received quemliclustat, then estimated the drug effect against patients from other trials. Its matched survival signal did not align with response or PFS. Baseline NR4A did not validate externally, and treatment-linked NR4A change defines groups after exposure. Oleclumab then missed the direct randomized response endpoint. The repair is the already enrolled PRISM-1 placebo comparison, not a new marker cut or cohort.", "TIGeR-PaC adds a package-and-interim inference failure. The trial changes route, removes nab-paclitaxel and changes schedule together, so it cannot isolate catheter delivery. Its first 45-person interim estimate was 16 versus 10 months, P=0.051, without a hazard ratio or confidence interval, and the second estimate is withheld. The repair is the complete final all-randomized package comparison plus a separate direct target-tissue measurement before platform transfer.", "Irreversible electroporation adds a causal-substitution and reporting failure. The DIRECT registry reports 18 versus 10 months after device patients survived and remained eligible for a median eight months from diagnosis versus four for controls; only ten of 38 controls were enrolled prospectively. That association cannot replace NCT03899636, the separate same-chemotherapy randomized trial. Its record lists 528 estimated participants and an estimated April 2025 completion but no actual count or result. Recover the existing assignments before another cohort.", "PANOVA-4 adds a component and historical-control failure. One metastatic cohort received tumour-treating fields, atezolizumab and two chemotherapies, then its disease-control rate was compared with an old chemotherapy trial. The observed difference combines at least the device effect, checkpoint effect, interaction, patient mix and time-period effects. One number cannot solve those terms. The 74.4% disease-control headline therefore cannot establish either component's effect or survival value. Require a current randomized control and one changed component per contrast before label or platform transfer.", "The newer local-treatment records add the same recurring failure in five forms. HybridTherm missed patient benefit and accrual; histotripsy can lose its acoustic window; cryoablation and microwave reports select survivors and bundle treatments; phosphorus-32 reports local change without proving survival. A device reaching a tumour is not the same as complete safe coverage or useful life.", "Light, heat and sound add geometry and component failures. Photodynamic and laser studies leave untreated margins around vulnerable structures, ASP-1929 has no pancreatic human result, and K-912 changes drug, ultrasound and chemotherapy together. The repair is measured target, energy and viable-tumour coverage plus one changed component per comparison.", "PIPAC and HIPEC add repeat-procedure and local-endpoint failures. Patients must remain well enough for another operation, making repeat recipients a selected group. HIPEC reduced local recurrence in a 42-person trial without improving median survival. Every analysis must keep all assigned patients, systemic progression, quality of life and total harm.", "Nerve ablation and ferroptosis add direct safety failures. Pain scores can improve while survival worsens, and killing by iron-driven membrane damage can affect tumour and non-tumour pancreatic cells in opposite ways. Patient benefit and cell-type coverage must be measured together.", "NRG1 and palliative care show that a useful treatment can still fail as a route. A rare fusion response is lost if RNA testing, confirmation and access break; supportive care is lost if referral comes late or if missing outcomes are removed from analysis. The whole path from eligibility to useful days must be counted.", "Pharmacogenetic dosing adds a panel-and-route failure. A four-variant DPYD result can prevent some 5-FU overexposure but does not certify safety, covers ancestries unevenly and can arrive too late; UGT1A1 risk changes with the irinotecan regimen. Count intended treatment, result, dose action, escalation, toxicity and tumour control as one route.", "KRAS-wild-type matching adds another denominator failure. Know Your Tumor returned reports to 1,082/1,856 referrals and only 46, 2.5%, entered the matched group; its survival advantage is selected. A valid route retains failed tissue, late RNA, unavailable drugs and fitness loss, then reports every target separately.", "Exercise adds a survivor-and-intermediate-endpoint failure. APACaP kept 172/313 randomized patients in its week-16 main analysis, while step and strength interventions changed movement without repeating a major complication or clinical benefit. Death, progression and missed activity must remain outcomes, not deletions." ], "sourceIds": [ "gudmundsdottir-2023-staging-laparoscopy", "rompen-2026-preopanc2-staging", "umin-000061741-staging-laparoscopy", "hingorani-2020-halo301", "rhim-2014-stroma-restrains", "fibrogen-2024-pamrevlumab-winddown", "noel-2020-ccr2-pdac", "kindler-2022-polo-overall-survival", "bekaii-saab-2023-napabucasin", "oreilly-2019-durvalumab-tremelimumab", "ho-2025-cxcr4-pd1-pdac", "bever-2026-vaccine-checkpoint-pdac", "hammel-2016-lap07", "mulcahy-2021-epoch-radioembolization", "ruers-2017-clocc-crc-liver-ablation", "adam-2024-transmet", "sarno-2025-avatar", "karam-2019-gpc1-nonreplication", "sasieni-2026-nhs-galleri", "pnas-2026-liaki-retraction", "weinberg-2020-ea2185-protocol", "clinicaltrials-nct04239573-ea2185", "clinicaltrials-nct06519097-sipp-t3", "van-dijk-2025-polyp-study", "clinicaltrials-nct03762837-gbc-risk-cohort", "partelli-2022-aspen-interim", "clinicaltrials-nct03084770-aspen", "van-vliembergen-2026-men1-mnet-assay", "villard-2023-psc-surveillance", "clinicaltrials-nct05321992-echec", "yu-2025-btc-ctdna-recurrence", "yoo-2025-extrahepatic-cca-ctdna", "clinicaltrials-nct05743959-btc-mrd", "trinchet-2011-hcc-surveillance-interval", "hirode-2026-hcc-biomarker-rct", "tie-2025-dynamic3", "bando-2026-altair", "sobnach-2024-ssa-hcc-systematic-review", "sobnach-2025-south-africa-hcc-disparities", "sobnach-2026-global-hcc-disparities", "chen-2025-tocilizumab-cachexia", "woo-2016-pert-rct", "saito-2018-pert-rct", "lemanska-2023-opensafely-pert-audit", "dhsc-2024-pert-shortage-alert", "heffley-2024-pdac-vte-uptake", "noor-2026-pdac-vte-implementation", "clinicaltrials-nct05178628-impact-pro", "van-der-gaag-2010-pbd-rct", "costamagna-2026-sems-early-surgery-rct", "lyu-2023-metal-plastic-nat-meta", "vehvilainen-2024-cholangitis-nat-interruption", "seo-2019-covered-uncovered-sems-rct", "chittenden-2021-systematic-germline", "walker-2021-embedded-germline-station", "wang-2022-oncology-germline-cascade", "rodriguez-2024-generate", "sinan-2025-pdac-germline-implementation", "katz-2026-gift-cascade", "npaca-2026-state-of-nation", "nhs-england-2024-hpb-pancreatic-service", "nice-2018-ng85-pancreatic", "de-wilde-2012-centralization", "lemmens-2011-centralization", "latenstein-2021-centralization", "coupland-2016-england-resection-volume", "labori-2016-adjuvant-completion", "henry-2023-complications-adjuvant", "kollbeck-2025-complications-chemo", "van-hilst-2019-leopard2", "korrel-2023-diploma", "bruna-2025-diploma-long-term", "klotz-2024-europa", "liu-2024-robotic-open-pd", "jin-2026-portal", "labori-2024-norpact1", "effenberger-2012-pdac-bone-marrow-dtc", "nordgard-2022-unresectable-dtc", "chalabi-hajkarim-2025-organotropism", "pei-2025-spatial-metastatic-pdac", "mcclellan-2026-quiescent-pdac-car-t", "tanaka-2012-microscopic-local-recurrence", "dotan-2024-giant", "dotan-2025-giant-ga", "clinicaltrials-nct04233866-giant", "betge-2022-grantpax", "clinicaltrials-nct02143219-pamela70", "mohile-2021-gap70", "li-2021-gain", "ihorst-2026-geriatric-comanagement", "clinicaltrials-nct06040801-frail-gi", "theijse-2024-nontherapeutic-laparotomy", "dekker-2026-taps-restaging", "napoli-2025-surgery-versus-oncology", "imamura-2004-locally-invasive-surgery-rct", "seelen-2026-pelican", "baum-2022-fap2286-firsthuman", "mcconathy-2026-lumiere-phase1", "clinicaltrials-nct04939610-lumiere", "baum-2026-3bp3940", "clinicaltrials-nct05432193-frontier", "clinicaltrials-nct07229768-ctr-fapi", "alistar-2017-devimistat-phase1", "philip-2024-avenger500", "surana-2025-erk-hcq-pdac", "clinicaltrials-nct05221320-ulixertinib-hcq", "manji-2023-mekiauto", "clinicaltrials-nct04214418-mekiauto", "clinicaltrials-nct04892017-inlexisertib", "deciphera-2022-dcc3116-phase1", "clinicaltrials-nct04524702-paricalcitol-hcq", "dimcevski-2016-sonoporation-phase1", "wang-2023-sonochemotherapy-pdac", "clinicaltrials-nct04821284-sonoporation", "ius-2026-sonoporation-nct04821284-abstract", "adler-2026-sonoporation-perfusion", "alagesan-2026-sig12d-loder", "ctis-2026-sil204-phase23", "silexion-2026-sil204-site-initiation", "silexion-2026-halfyear-filing", "bazan-peregrino-2021-vcn01-intratumour", "garcia-carbonero-2022-vcn01-intravenous", "clinicaltrials-nct05673811-virage", "theriva-2023-virage-protocol", "theriva-2025-virage-topline", "esmo-2025-virage-2216mo", "ctis-2026-virage2", "theriva-2026-q2-10q", "richards-2006-ci994-pdac", "heumann-2022-azacitidine-pdac", "sohal-2020-thu-decitabine-pdac", "baretti-2024-entinostat-nivolumab-pdac", "clinicaltrials-nct03250273-entinostat-nivolumab", "safyan-2026-azacitidine-pembrolizumab-pdac", "clinicaltrials-nct03264404-azacitidine-pembrolizumab", "rodon-2024-amg193-prmt5", "matsumoto-2026-mtap-pancreatic", "pavlick-2026-mtap-genomics", "clinicaltrials-nct05732831-vopimetostat", "tango-2025-vopimetostat-monotherapy", "clinicaltrials-nct06922591-vopimetostat-ras", "tango-2026-vopimetostat-ras-data", "tango-2026-q2-10q", "mahalingam-2026-elraglusib-randomized", "clinicaltrials-nct03678883-elraglusib", "pathak-2026-elraglusib-folfirinox", "clinicaltrials-nct05077800-elraglusib-folfirinox", "actuate-2026-q2-10q", "sec-2026-actuate-submissions", "chiorean-2023-abemaciclib-pdac", "baghdadi-2019-palbociclib-cdkn2a", "hidalgo-2022-palbociclib-nabpaclitaxel", "laquente-2017-chk1-pdac", "huffman-2023-ly2880070-pdac", "cuneo-2019-adavosertib-pdac", "huffman-2024-azenosertib-pdac-design", "clinicaltrials-nct06015659-azenosertib", "park-2026-gleam-zolbetuximab", "clinicaltrials-nct03816163-gleam", "yu-2025-ibi343-pdac", "clinicaltrials-nct07066098-ibi343-phase3", "hao-2024-ibi389-pdac", "xu-2025-ibi389-phase1", "lentz-2025-ebc129-pdac", "clinicaltrials-nct03023722-anetumab", "gou-2026-shr3821", "abou-alfa-2006-exatecan-pdac", "clinicaltrials-nct03323944-meso-cart", "aznar-2025-mesothelin-car-t-resistance", "musher-2026-multiantigen-t-cells", "qi-2024-ct041-pdac", "xu-2026-kras-g12v-tcr", "clinicaltrials-nct04935359-danis2", "pelletier-2026-nis793-biomarkers", "clinicaltrials-nct03336216-cabiralizumab", "padron-2022-prince", "van-laethem-2025-optimize1-biomarkers", "kindler-2010-bevacizumab-calgb80303", "van-cutsem-2009-bevacizumab-erlotinib", "rougier-2013-aflibercept-vanilla", "kindler-2011-axitinib-pdac", "murphy-2019-losartan-lapc", "ramaswamy-2025-afpac-losartan", "nakaoka-2024-kestose-pdac", "clinicaltrials-nct03891979-microbiome-pdac", "clinicaltrials-nct04193904-mrx0518-pdac", "clinicaltrials-nct04975217-fmt-pdac", "corty-2020-antibiotic-gemcitabine-toxicity", "middleton-2014-telovac", "le-2015-gvax-crs207", "le-2019-eclipse", "elicio-2026-amplify7p-results", "wainberg-2026-quemliclustat-arc8", "coveler-2024-oleclumab-pdac", "clinicaltrials-nct06608927-prism1", "clinicaltrials-nct05254171-aspire", "ctis-2024-514714-12-00-aspire", "panbela-2024-q3-10q", "panbela-2025-note-default", "sec-2025-panbela-registration-abandoned", "clinicaltrials-nct03257033-tigerpac", "hatoum-2024-tamp-rr1-rr2", "novelli-2026-tigerpac-pk", "renovorx-2023-tigerpac-interim", "renovorx-2025-10k", "renovorx-2026-full-enrollment", "clinicaltrials-nct03899636-direct-rct", "clinicaltrials-nct03899649-direct-registry", "martin-2024-direct-registry-safety", "martin-2026-direct-registry-survival-preprint", "nice-2017-ire-pancreatic", "macarulla-2025-panova3", "clinicaltrials-nct03377491-panova3", "fda-2026-optune-pax-ssed", "clinicaltrials-nct06390059-panova4", "novocure-2026-panova4-topline", "novocure-2026-q2-10q", "guzauskas-2026-ttfields-cost", "liu-2026-ttfields-cost", "issels-2023-heat-rct", "clinicaltrials-nct01077427-heat", "eudract-2008-004802-14-heat", "datta-2017-heatpac-protocol", "clinicaltrials-nct02439593-heatpac", "shinoto-2016-carbon-ion-gemcitabine", "kawashiro-2018-jcros-carbon-ion", "lai-2026-carbon-ion-risk-adapted", "rapp-2022-proton-phase2", "eckstein-2023-proton-registry", "clinicaltrials-nct02598349-proton", "clinicaltrials-nct03536182-cipher", "clinicaltrials-nct04194268-pack", "parikh-2023-smart-phase2-safety", "chuong-2024-smart-phase2-survival", "clinicaltrials-nct03621644-smart", "clinicaltrials-nct05585554-lap-ablate", "clinicaltrials-nct05114213-maspac", "viewray-2023-q1-10q", "viewray-2023-chapter11-8k", "clinicaltrials-nct03690323-pelican", "li-2015-hifu-s1", "marinova-2018-hifu-prospective", "marinova-2024-hifu-pc-protocol", "drks-00012367-hifu-pc", "clinicaltrials-nct05262452-hifu-folfirinox", "clinicaltrials-nct05601323-suizenji", "clinicaltrials-nct06211933-puls-hifu", "clinicaltrials-nct07325214-sonopan3", "izzo-2021-pancreatic-ect-phase12", "izzo-2021-laparoscopic-ect-protocol", "eudract-2018-003925-27-ect", "cebron-2023-panect-preliminary", "clinicaltrials-nct04281290-panect-results", "rudno-rudzinska-2021-irec-protocol", "igea-2026-cliniporator-vitae", "testoni-2021-hybridtherm-rct", "gannon-2026-pancreas-histotripsy-safety", "pusceddu-2026-percutaneous-cryoablation", "zhou-2024-pancreatic-mwa", "ross-2022-panco-p32", "huggett-2014-verteporfin-pdt", "dimatteo-2018-eus-laser-ablation", "clinicaltrials-nct07698613-asp1929-pancreas", "muragaki-2026-k912-sonodynamic", "clinicaltrials-nct05371223-nab-pipac", "padilla-valverde-2024-adjuvant-hipec-rct", "levy-2019-celiac-ganglia-versus-plexus-rct", "xie-2026-cnsi-fe-first-human", "fda-2024-bizengri-multidisciplinary-review", "maltoni-2016-systematic-palliative-qol", "glewis-2024-pacific-pgx", "pishvaian-2020-know-your-tumor", "neuzillet-2023-apacap-rct" ], "links": [ { "target": "failure-organoid-treatment-chain-attrition", "relation": "contains" }, { "target": "failure-staging-yield-without-patient-utility", "relation": "contains" }, { "target": "failure-liaki-undisclosed-conflict-retraction", "relation": "contains" }, { "target": "failure-crc-radioembolization-pfs-without-os", "relation": "contains" }, { "target": "failure-immune-recruitment-without-response", "relation": "contains" }, { "target": "failure-ea2185-randomized-question-closed", "relation": "contains" }, { "target": "failure-sipp-t3-accrual-collapse", "relation": "contains" }, { "target": "failure-polyp-surveillance-low-yield", "relation": "contains" }, { "target": "failure-nct03762837-risk-cohort-never-started", "relation": "contains" }, { "target": "failure-aspen-cannot-estimate-surgery-benefit", "relation": "contains" }, { "target": "failure-men1-mnet-assay-no-progression-prediction", "relation": "contains" }, { "target": "failure-psc-surveillance-no-curable-detection", "relation": "contains" }, { "target": "failure-echec-cancer-outcome-mismatch", "relation": "contains" }, { "target": "failure-cca-mrd-prediction-without-action", "relation": "contains" }, { "target": "failure-hcc-three-month-ultrasound-no-small-cancer-gain", "relation": "contains" }, { "target": "failure-hcc-biomarker-addition-no-early-stage-gain", "relation": "contains" }, { "target": "failure-dynamic3-ctdna-guided-chemo-no-rfs-gain", "relation": "contains" }, { "target": "failure-altair-mrd-treatment-no-dfs-gain", "relation": "contains" }, { "target": "failure-ssa-hcc-curative-route-not-reaching-patients", "relation": "contains" }, { "target": "failure-tocilizumab-cachexia-muscle-without-survival", "relation": "contains" }, { "target": "failure-pert-short-randomized-trials", "relation": "contains" }, { "target": "failure-pert-prescribing-and-supply-gap", "relation": "contains" }, { "target": "unknowns", "relation": "supports" }, { "target": "experiments", "relation": "resolves" }, { "target": "failure-pdac-vte-prophylaxis-delivery-gap", "relation": "contains" }, { "target": "failure-impact-pro-status-unknown", "relation": "contains" }, { "target": "failure-routine-plastic-pbd", "relation": "contains" }, { "target": "failure-biliary-stent-treatment-interruption", "relation": "contains" }, { "target": "failure-pdac-germline-route-dropout", "relation": "contains" }, { "target": "unknown-pdac-germline-family-action-route", "relation": "supports" }, { "target": "hypothesis-pdac-germline-family-route-completion", "relation": "resolved-by" }, { "target": "claim-pdac-surgery-is-a-complete-treatment-route", "relation": "links-to" }, { "target": "system-england-pancreatic-surgery-network", "relation": "links-to" }, { "target": "failure-pdac-postoperative-treatment-loss", "relation": "links-to" }, { "target": "failure-leopard2-laparoscopic-whipple-safety", "relation": "links-to" }, { "target": "unknown-pdac-surgical-route-completion", "relation": "links-to" }, { "target": "trial-pdac-surgical-route-action-map", "relation": "links-to" }, { "target": "hypothesis-pdac-surgical-route-completion", "relation": "links-to" }, { "target": "failure-norpact1-neoadjuvant-delivery", "relation": "contains" }, { "target": "failure-pdac-dormancy-sample-gap", "relation": "contains" }, { "target": "failure-grantpax-geriatric-route", "relation": "contains" }, { "target": "failure-conversion-surgery-selection-bias", "relation": "contains" }, { "target": "failure-fapi-scan-dose-benefit-collapse", "relation": "contains" }, { "target": "claim-fap-radioligand-feasible-pdac-efficacy-unproven", "relation": "bounded-by" }, { "target": "unknown-pdac-fap-radioligand-treatment-effect", "relation": "supports" }, { "target": "failure-metabolic-repurposing-pdac", "relation": "contains" }, { "target": "claim-hcq-combinations-fail-pdac-not-autophagy", "relation": "bounded-by" }, { "target": "unknown-selective-autophagy-pdac-contribution", "relation": "supports" }, { "target": "failure-sonoporation-overall-replication-subgroup-rescue", "relation": "contains" }, { "target": "claim-sonochemotherapy-randomized-signal", "relation": "bounded-by" }, { "target": "unknown-pdac-sonoporation-treatment-effect", "relation": "supports" }, { "target": "failure-sig12d-loder-randomized-os-not-shown", "relation": "contains" }, { "target": "claim-sil204-trial-scales-before-human-knockdown", "relation": "bounded-by" }, { "target": "unknown-sil204-human-target-delivery", "relation": "supports" }, { "target": "claim-vcn01-randomized-signal-not-survival-proof", "relation": "bounded-by" }, { "target": "failure-virage-analysis-and-survivor-selection", "relation": "learns-from" }, { "target": "unknown-vcn01-all-enrolled-survival", "relation": "leaves-open" }, { "target": "trial-vcn01-action-map", "relation": "implements" }, { "target": "hypothesis-vcn01-replication-to-survival-gate", "relation": "tests" }, { "target": "company-theriva-biologics", "relation": "contains" }, { "target": "failure-broad-epigenetic-treatment-pdac", "relation": "learns-from" }, { "target": "claim-mtap-prmt5-selected-human-signal", "relation": "bounded-by" }, { "target": "failure-vopimetostat-ras-contribution-unknown", "relation": "learns-from" }, { "target": "unknown-mtap-prmt5-ras-contribution", "relation": "leaves-open" }, { "target": "trial-mtap-prmt5-action-map", "relation": "implements" }, { "target": "hypothesis-mtap-prmt5-ras-deeper-control", "relation": "tests" }, { "target": "company-tango-therapeutics", "relation": "contains" }, { "target": "claim-elraglusib-randomized-survival-signal", "relation": "bounded-by" }, { "target": "failure-elraglusib-analysis-and-survival-discordance", "relation": "learns-from" }, { "target": "unknown-elraglusib-survival-replication", "relation": "leaves-open" }, { "target": "trial-elraglusib-action-map", "relation": "implements" }, { "target": "hypothesis-elraglusib-survival-confirmation-gate", "relation": "tests" }, { "target": "company-actuate-therapeutics", "relation": "contains" }, { "target": "failure-broad-cell-cycle-checkpoint-pdac", "relation": "learns-from" }, { "target": "unknown-wee1-replication-stress-pdac", "relation": "leaves-open" }, { "target": "trial-cell-cycle-checkpoint-action-map", "relation": "implements" }, { "target": "hypothesis-human-replication-stress-selector", "relation": "tests" }, { "target": "failure-antibody-target-and-payload-transfer-pdac", "relation": "contains" }, { "target": "trial-cldn182-antibody-action-map", "relation": "mapped-by" }, { "target": "failure-cell-therapy-route-and-denominator-pdac", "relation": "contains" }, { "target": "trial-pdac-engineered-cell-action-map", "relation": "mapped-by" }, { "target": "failure-immune-reprogramming-without-patient-benefit", "relation": "contains" }, { "target": "trial-immune-reprogramming-action-map", "relation": "mapped-by" }, { "target": "failure-vascular-starvation-pdac", "relation": "contains" }, { "target": "failure-losartan-delivery-signal-not-replicated", "relation": "contains" }, { "target": "trial-pdac-vascular-hypoxia-action-map", "relation": "mapped-by" }, { "target": "failure-microbiome-proxy-without-patient-benefit", "relation": "contains" }, { "target": "claim-pdac-microbiome-mechanism-without-human-benefit", "relation": "constrains" }, { "target": "trial-pdac-microbiome-action-map", "relation": "mapped-by" }, { "target": "failure-pdac-vaccine-escalation-reversals", "relation": "contains" }, { "target": "claim-pdac-vaccine-immunity-without-proven-benefit", "relation": "constrains" }, { "target": "trial-pdac-vaccine-action-map", "relation": "mapped-by" }, { "target": "failure-adenosine-control-and-marker-pdac", "relation": "contains" }, { "target": "claim-cd73-blockade-signal-without-concurrent-control", "relation": "constrains" }, { "target": "trial-pdac-adenosine-action-map", "relation": "mapped-by" }, { "target": "failure-aspire-trial-operational-collapse", "relation": "contains" }, { "target": "claim-ivospemin-signal-without-controlled-benefit", "relation": "constrains" }, { "target": "trial-pdac-polyamine-action-map", "relation": "mapped-by" }, { "target": "failure-tigerpac-component-and-interim-inference", "relation": "contains" }, { "target": "claim-tamp-blood-exposure-not-tumour-delivery", "relation": "constrains" }, { "target": "trial-pdac-intraarterial-delivery-action-map", "relation": "mapped-by" }, { "target": "failure-direct-selection-and-randomized-reporting-gap", "relation": "contains" }, { "target": "claim-ire-ablation-without-proven-added-survival", "relation": "constrains" }, { "target": "trial-pdac-irreversible-electroporation-action-map", "relation": "mapped-by" }, { "target": "failure-panova4-component-and-historical-control", "relation": "contains" }, { "target": "claim-panova3-ttfields-small-os-benefit", "relation": "bounded-by" }, { "target": "trial-pdac-ttfields-action-map", "relation": "mapped-by" }, { "target": "failure-heat-component-and-registry-drift", "relation": "contains" }, { "target": "claim-heat-regional-hyperthermia-primary-endpoint-negative", "relation": "bounded-by" }, { "target": "trial-pdac-regional-hyperthermia-action-map", "relation": "mapped-by" }, { "target": "failure-particle-therapy-comparator-and-accrual", "relation": "contains" }, { "target": "claim-particle-therapy-local-control-without-survival-proof", "relation": "bounded-by" }, { "target": "trial-pdac-particle-therapy-action-map", "relation": "mapped-by" }, { "target": "failure-smart-trial-and-vendor-dependency", "relation": "contains" }, { "target": "claim-smart-delivery-and-safety-without-added-survival-proof", "relation": "bounded-by" }, { "target": "trial-pdac-smart-action-map", "relation": "mapped-by" }, { "target": "failure-rfa-selection-and-route-attrition", "relation": "contains" }, { "target": "claim-pelican-rfa-no-survival-and-more-harm", "relation": "bounded-by" }, { "target": "trial-pdac-rfa-action-map", "relation": "mapped-by" }, { "target": "failure-hifu-randomization-and-trial-drift", "relation": "contains" }, { "target": "claim-hifu-pain-signal-without-survival-proof", "relation": "bounded-by" }, { "target": "trial-pdac-hifu-action-map", "relation": "mapped-by" }, { "target": "failure-electrochemotherapy-endpoint-and-trial-drift", "relation": "contains" }, { "target": "claim-electrochemotherapy-local-effect-without-patient-benefit", "relation": "bounded-by" }, { "target": "trial-pdac-electrochemotherapy-action-map", "relation": "mapped-by" }, { "target": "trial-pdac-eus-rfa-action-map", "relation": "mapped-by" }, { "target": "trial-pdac-histotripsy-action-map", "relation": "mapped-by" }, { "target": "trial-pdac-cryoablation-action-map", "relation": "mapped-by" }, { "target": "trial-pdac-mwa-action-map", "relation": "mapped-by" }, { "target": "trial-pdac-p32-action-map", "relation": "mapped-by" }, { "target": "trial-pdac-pdt-action-map", "relation": "mapped-by" }, { "target": "trial-pdac-laser-ablation-action-map", "relation": "mapped-by" }, { "target": "trial-pdac-asp1929-action-map", "relation": "mapped-by" }, { "target": "trial-pdac-k912-sonodynamic-action-map", "relation": "mapped-by" }, { "target": "trial-pdac-pipac-action-map", "relation": "mapped-by" }, { "target": "trial-pdac-hipec-action-map", "relation": "mapped-by" }, { "target": "trial-pdac-celiac-pain-action-map", "relation": "mapped-by" }, { "target": "trial-pdac-ferroptosis-action-map", "relation": "mapped-by" }, { "target": "trial-pdac-nrg1-action-map", "relation": "mapped-by" }, { "target": "trial-pdac-palliative-care-action-map", "relation": "mapped-by" }, { "target": "failure-pharmacogenetic-panel-and-route-coverage", "relation": "contains" }, { "target": "failure-kras-wildtype-matching-route-and-selection", "relation": "contains" }, { "target": "failure-pdac-exercise-endpoint-and-attrition", "relation": "contains" }, { "target": "failure-mistletoe-pdac-no-quality-benefit", "relation": "contains" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "moderate", "randomised": "strong", "independentReplication": "strong", "sampleSize": "102 linked failure records spanning randomized trials, low-yield care rules, delivery systems, host wasting, immune mechanism, replication and governance", "effect": "Failure classification; no pooled clinical effect", "limits": [ "Failure causes are partly inferred", "A failed product may leave its mechanism open", "Some trial data remain sponsor reports", "Negative subgroup analyses are underpowered", "Absence of efficacy is not proof of absence for every selected state" ] }, "layout": { "x": 7, "y": 76 }, "updated": "2026-09-15" }, { "id": "funding-route-stage-fit", "title": "Funding routes must match the evidence stage and cheque size", "shortTitle": "Funding route fit", "type": "funding", "status": "supported", "scope": [ "pdac", "pnet", "cholangiocarcinoma", "gallbladder", "hcc", "colorectal", "pan-cancer" ], "summary": "R01s and disease charities fit bounded work; £20m challenge awards fit rare, fixed calls; £100m-plus trials require a public–philanthropic–commercial syndicate.", "content": "Stage 1 — mechanism, assay and model work, roughly £0.2m–£3m: NIH R01/U01, Wellcome Discovery Awards and disease-charity grants fit. An R01 normally runs one to five years; Wellcome reports a £3.5m average Discovery Award over seven years. NCI's FY2025 competing RPG success rate was 9.4%, so grant funding is a competitive route, not committed mission capital. Stage 2 — prospective feasibility and shared specimen systems, roughly £3m–£25m: several grants, a SPORE, a lead charity or a challenge award can fit. The 2025 Cancer Grand Challenges round offered up to £20m but accepted only funder-defined challenges and is now closed; 5 teams were selected from 227 expressions of interest, 2.2%, after 12 were shortlisted. It is a useful scale benchmark, not an available PDAC cheque. Stage 3 — randomized action trials, roughly £25m–£150m: NCI cooperative groups, NIHR-like public trials, large philanthropy and drug-company supply must be combined. The £100m programme should require at least one anchor to underwrite 30–40%, then make the rest conditional on drug, assay and public-trial contributions. Stage 4 — a 40,000–60,000-person detection mortality or late-stage-incidence trial and parallel global treatment platform, £100m–£500m: only a health system, government, very large foundation or consortium can carry follow-up and liability. Venture capital alone has the wrong return clock for a negative-capable population trial. Funding likelihood: an excellent bounded project faces an observed NCI-wide base rate near 1 in 11 in one FY; a bespoke £20m challenge bid faced about 1 in 45 in the last round and only if the question matched. For a £100m–£500m mission, probability is not estimated from these rates because funders and decisions are correlated. The practical route is tranche zero from one lead donor, then negotiated matching commitments against locked stop rules—not dozens of small grants pretending to be a programme.", "sourceIds": [ "nih-2026-r01", "nci-2026-rpg-funding", "nci-2026-spore-funding", "wellcome-2026-discovery-awards", "cancer-grand-challenges-2026-round", "cancer-grand-challenges-2025-guidelines", "pancan-2025-grants", "lustgarten-2025-grants", "pcuk-2025-annual-report", "ukri-accelerated-cancer-diagnosis-2026", "cruk-therapeutic-catalyst-2026", "horizon-cancer-mission-calls-2026", "arpa-h-open-funding-2026" ], "links": [ { "target": "funding", "relation": "part-of" }, { "target": "programme-5m-decision-core", "relation": "funds" }, { "target": "programme-25m-prospective-observatory", "relation": "funds" }, { "target": "programme-100m-randomized-platform", "relation": "funds" }, { "target": "programme-500m-integrated-mission", "relation": "funds" }, { "target": "claim-observed-ten-year-research-capital", "relation": "depends-on" }, { "target": "claim-live-funding-routes-2026", "relation": "contains" } ], "evidence": { "humanProspective": "not-applicable", "humanRetrospective": "not-applicable", "randomised": "not-applicable", "independentReplication": "moderate", "sampleSize": "11,711 NCI competing RPG requests; 227 Cancer Grand Challenges expressions of interest", "effect": "NCI RPG success 9.4%; Cancer Grand Challenges 5/227 = 2.2% in the last round", "limits": [ "Rates are portfolio-wide, not applicant-specific", "Cancer Grand Challenges round is closed", "Wellcome publishes average award size but no scheme-specific probability here", "Large mission syndicate probability cannot be inferred from grant application rates", "Commercial terms and drug contributions require negotiation" ] }, "layout": { "x": 80, "y": 116 }, "updated": "2026-09-14" }, { "id": "funding", "title": "Funding routes", "shortTitle": "Funding", "type": "funding", "status": "supported", "scope": [ "pdac", "pnet", "cholangiocarcinoma", "gallbladder", "hcc", "colorectal", "pan-cancer" ], "summary": "Each programme stage is matched to capital that can tolerate its time, evidence and return profile; current public and charity anchors are recorded before commercial routes are added.", "content": "Match the money to the evidence clock. Stage 1, roughly £0.2m–£3m for mechanism, assay and model work: NIH R01 or U01 awards, Wellcome Discovery Awards and disease charities fit. An R01 normally runs one to five years; Wellcome reports a £3.5m average Discovery Award over seven years. NCI's FY2025 competing research-project-grant success rate was 9.4%, so grant submission is a route, not committed capital. Stage 2, roughly £3m–£25m for prospective feasibility, specimens and shared systems: several grants, an NCI SPORE, a lead charity or a fixed challenge can fit. Cancer Grand Challenges offered up to £20m in its 2025 round, but only five teams were selected from 227 expressions, 2.2%, the call is closed, and topics were funder-defined. Stage 3, £25m–£150m for randomized action trials: combine a public cooperative-trial network, large philanthropy, assay support and drug-company supply. One anchor should underwrite 30–40% of the £100m programme; remaining commitments become conditional on data rights, negative publication and the safety gate. Stage 4, £100m–£500m for a 40,000–60,000-person detection utility trial plus treatment platforms: only a government, health system, very large foundation or consortium can carry long follow-up and liability. Venture capital is suited to a defined product with a shorter value inflection—drug, assay, imaging software or manufacturing platform—not a negative-capable mortality trial. Pharmaceutical capital should supply compounds, safety and manufacturing while independent public or philanthropic capital protects comparison, specimens and publication. Charities such as PanCAN, Lustgarten and Pancreatic Cancer UK are well placed to fund careers, shared data, bounded pilots and patient participation; their disclosed annual and cumulative scales do not support a £100m–£500m mission alone. The current 2026 route map adds an Innovate UK £25.4m diagnostic-validation call, Wellcome's September Discovery Award round, Cancer Research UK's £250,000–£350,000 Therapeutic Catalyst, a tightly topic-bound Horizon Cancer Mission call and rolling ARPA-H mission-office concept routes. Their deadlines, eligibility and scientific scopes differ, so they fund separate work packages rather than one programme. The practical route is a lead-donor tranche for contracts and assay lock, then matched public, philanthropic and commercial commitments against explicit gates. Avoid dozens of small grants with incompatible consent, samples and endpoints. Funding terms are part of the experiment: the mission needs control of patient-level data, model versions, failed arms, stored specimens and independent analysis even when a company exits or a grant ends.", "sourceIds": [ "nih-2026-r01", "wellcome-2026-discovery-awards", "nci-2026-rpg-funding", "nci-2026-spore-funding", "cancer-grand-challenges-2025-guidelines", "cancer-grand-challenges-2026-round", "pancan-2025-grants", "lustgarten-2025-grants", "pcuk-2025-annual-report", "ukri-accelerated-cancer-diagnosis-2026", "cruk-therapeutic-catalyst-2026", "horizon-cancer-mission-calls-2026", "arpa-h-open-funding-2026", "arpa-h-adapt-2026" ], "links": [ { "target": "investment", "relation": "part-of" }, { "target": "programme", "relation": "supports" }, { "target": "risks", "relation": "depends-on" }, { "target": "claim-observed-ten-year-research-capital", "relation": "contains" }, { "target": "funding-route-stage-fit", "relation": "contains" }, { "target": "claim-live-funding-routes-2026", "relation": "contains" } ], "evidence": { "humanProspective": "not-applicable", "humanRetrospective": "not-applicable", "randomised": "not-applicable", "independentReplication": "moderate", "sampleSize": "Public funding rules, observed grant rates, one major challenge round and charity disclosures", "effect": "Stage-to-capital route map; no promise of funding", "limits": [ "Live calls and budgets change", "Success rates are funder-wide rather than PDAC-specific", "Commercial terms are private", "Cheque-size ranges are judgement", "Funders and decisions are correlated" ] }, "layout": { "x": 78, "y": 114 }, "updated": "2026-09-14" }, { "id": "gallbladder", "title": "Gallbladder cancer", "shortTitle": "Gallbladder", "type": "cancer", "status": "supported", "scope": [ "gallbladder" ], "summary": "A separate branch for gallstone-linked risk, geographic incidence, incidental diagnosis, surgery, recurrence and molecular treatment.", "content": "Gallbladder cancer shares some advanced biliary trials but differs in precursor sequence, sex distribution, geography, surgical anatomy and target frequency. Gallstones and chronic mucosal injury are common while cancer is rare, so prophylactic cholecystectomy cannot be justified by gallstones alone. South American surgical data supply candidate selectors, not a validated individual prevention rule. A 622,227-person ultrasound cohort found almost identical overall cancer occurrence with and without a recorded polyp, although large initial polyps carried more risk.\n\nThe central observability problem is geographical and clinical. In a prospective 1,500-patient North Indian hospital cohort, 80.2% arrived with metastatic disease and only 92 received definitive surgery. That is evidence of late presentation and access failure inside a referral population, not a population incidence estimate. Earlier ultrasound or surgery must therefore be tested in a defined high-risk population with harms and denominator visible.\n\nIncidental cancer after cholecystectomy and direct liver or nodal invasion create different surgical decisions. A Dutch national cohort found residual disease in 35% of re-resected incidental cancers, most often lymph nodes, but its survival comparison is strongly selected. A 163-patient randomized trial found that removing liver segments 4b/5 did not improve survival over a wedge during radical cholecystectomy and required more operative time and blood loss. HER2 can identify an advanced-treatment subgroup, but amplification and strong IHC expression cannot be collapsed into one marker. Mixed biliary immunochemotherapy is supportive background, not a gallbladder-specific effect estimate. The practical mission is to find a prospective risk rule among common gallbladder disease, diagnose before metastasis, standardise pathology and re-resection decisions, and test systemic treatment by gallbladder biology.", "contentSections": [ "Prevention and diagnosis are now split into four actions. Future cancer risk needs longitudinal operated and unoperated cohorts. Existing-lesion diagnosis may use expert imaging, cfDNA and AI. Optional cholecystectomy needs a local net-benefit threshold. Histology and re-resection begin only after removal. EULAT, Chile BiLS, POLYP, GBCseeker, GAIA-MIL, FANCY and P-iGBC each cover a fragment; none supplies the whole risk-to-mortality pathway.", "The operation is not harmless. In 21,706 contemporary benign-disease cholecystectomies across 57 countries, 8.0% had a 30-day complication, 0.4% died and 0.2% had a severe bile-duct injury. Because emergency and delayed cases are included, every prevention programme must measure its own elective harm rather than import these totals." ], "sourceIds": [ "nakamura-2015-biliary-genomics", "gonzalez-2025-cholecystectomy-chile", "lorenzo-bermejo-2025-gallbladder-specimens", "szpakowski-2020-gallbladder-polyps", "van-dijk-2025-polyp-study", "yang-2025-gbcseeker", "gupta-2026-gaia-mil", "clinicaltrials-nct06192719-eulat-gbc", "clinicaltrials-nct06531408-pigbc", "wong-2025-ambrose-cholecystectomy", "vineet-2026-north-india-gallbladder-cohort", "de-savornin-lohman-2020-incidental-gallbladder", "singh-2026-gallbladder-liver-resection-rct", "pant-2026-zanidatamab-final", "oh-2022-topaz1" ], "links": [ { "target": "hepatobiliary", "relation": "part-of" }, { "target": "model-gallbladder-causal-chain", "relation": "contains" }, { "target": "disease-system", "relation": "part-of" }, { "target": "detection", "relation": "part-of" }, { "target": "treatment", "relation": "part-of" }, { "target": "claim-gallbladder-polyp-selector", "relation": "contains" }, { "target": "claim-gallbladder-polyp-guidelines-still-overtreat", "relation": "contains" }, { "target": "claim-gallbladder-diagnostic-models-existing-lesions-only", "relation": "contains" }, { "target": "claim-gallbladder-prevention-must-beat-surgery-harm", "relation": "contains" }, { "target": "trial-gallbladder-prevention-detection-action-map", "relation": "contains" }, { "target": "hypothesis-gallbladder-two-stage-interception-rule", "relation": "tests" }, { "target": "claim-gallbladder-incidental-residual-disease", "relation": "contains" }, { "target": "claim-gallbladder-liver-resection-extent", "relation": "contains" }, { "target": "company-jazz-pharmaceuticals", "relation": "contains" }, { "target": "claim-biliary-immunochemotherapy-tail", "relation": "contains" }, { "target": "company-astrazeneca", "relation": "contains" } ], "evidence": { "humanProspective": "moderate for component decisions; none for risk-guided prevention benefit", "humanRetrospective": "strong", "randomised": "weak and mainly mixed biliary", "independentReplication": "weak — mixed evidence", "sampleSize": "Large South American and international surgery cohorts, 1,500-person North Indian prospective hospital cohort, 622,227-person polyp cohort, and smaller molecular and imaging studies", "effect": "Late metastatic presentation dominates one endemic referral cohort; no prospective prevention selector is validated; current revised polyp rules found one HGD among 68 indicated operations", "limits": [ "Referral cohorts do not measure population incidence", "The only direct randomized surgical comparison is single-centre with short follow-up and wide intervals", "Geographic causes may not transport", "No prospective prevention selector has cancer-mortality validation", "Re-resection comparisons remain nonrandomized", "Existing-lesion AI does not establish future-risk or screening utility", "Surgery harm and baseline cancer risk vary by country" ] }, "layout": { "x": 54, "y": 5 }, "updated": "2026-09-14" }, { "id": "hcc", "title": "Hepatocellular carcinoma", "shortTitle": "HCC", "type": "cancer", "status": "supported", "scope": [ "hcc" ], "summary": "HCC is modelled through chronic liver injury, cirrhosis, viral prevention, surveillance, liver function, local cure, recurrence and systemic control.", "content": "The host liver is part of HCC. Hepatitis B or C, alcohol and metabolic liver disease create chronic injury, cirrhosis and repeated clonal selection. Removing one tumour does not remove the damaged field, while treatment can be limited by liver reserve before tumour burden alone would be fatal.\n\nHCC proves that cancer prevention and surveillance can work when a causal exposure and a high-risk organ population are visible. Hepatitis-B vaccination and viral suppression reduce later HCC, and one randomized Chinese ultrasound programme reduced HCC mortality. In alcohol-related cirrhosis, a prospective cohort found abstinence associated with lower HCC incidence before previous decompensation but not after it; this is a plausible interception window, not randomized proof. None of these results transfers automatically to pancreatic screening.\n\nResection, ablation and transplant can eliminate visible disease in selected patients. In the randomized 302-patient SURF trial, resection did not beat ablation and fewer than half were recurrence-free at five years despite roughly 70% survival. Transplant can remove both tumour and diseased field, but the original 92% four-year recurrence-free result came from only 35 patients who met strict pathological tumour limits. When early HCC recurred after resection, a separate 240-patient randomized trial found no significant survival difference between repeat surgery and ablation: surgery improved local control but caused more complications. Multiregion sequencing shows that intrahepatic recurrence may arise from several clones or a new field tumour, while distant metastasis more often traces to one clone. Adjuvant sorafenib and atezolizumab-bevacizumab did not create durable low-burden benefit.\n\nThe tumour and liver are separate failure axes. In 1,336 people with early HCC, 114 deaths were attributed to non-cancer liver disease versus 220 to HCC. In a 322-person advanced-treatment cohort, median survival fell from 21.6 months with Child-Pugh A to 9.1 months with B7 and 4.7 months with B8-C12. Those observational differences do not prove that improving liver function will improve cancer survival, but they make a tumour-only endpoint inadequate. The mission therefore joins cause control, repeated surveillance, paired tumour and liver-function staging, recurrence-origin classification, repeatable local treatment and systemic therapy that preserves future liver reserve.", "contentSections": [ "Modern surveillance evidence separates a test gain from a survival gain. Noncontrast MRI lowered false-positive referrals from 3.1% to 0.7% and shifted stage in a 414-person randomized trial, but did not show a mortality effect. Adding AFP, AFP-L3 and DCP to ultrasound did not improve early-stage detection in 1,208 randomized adults. Ultrasound every three months found more lesions 10 mm or smaller than every six months but no more HCC 30 mm or smaller. TRACER now tests late-stage HCC and harm for GALAD; PREMIUM tests HCC mortality for abbreviated MRI. Neither has results.", "Prevention still has an unfinished boundary. ATTENTION moved tenofovir alafenamide into noncirrhotic HBV with moderate or high viraemia. HCC occurred in 2 of 369 treated people and 7 of 365 observed people after median follow-up of 17.7 months, but only 11 composite events occurred, the 97.5% confidence interval crossed no effect and the early-stop boundary was not crossed. The mature analysis, not the interim point estimate, should set any wider treatment rule.", "Three 2026 randomized reports sharpen treatment claims. TALENTOP selected 201 of 489 people after systemic induction and anatomical review; adding resection increased median time to treatment failure by 8.6 months, but caused more severe harm and two treatment-related deaths, and has not yet established longer overall survival. EMERALD-3 increased median progression-free survival by 3.2 months with STRIDE immunotherapy, lenvatinib and arterial chemoembolisation, while the reported overall-survival difference was not significant and serious adverse events were 64% versus 23%. TORCH produced the strongest patient-outcome result: among 241 people with liver-confined BCLC stage B HCC, selective radiofrequency ablation after arterial chemoembolisation increased median overall survival from 35.1 to 88.6 months, HR 0.50. Its two-centre Chinese setting and selected tumour geometry bound transfer. These are state-specific treatment routes, not proof of cure or a general local-treatment rule.", "The first new HCC investment is a £6m randomized readiness stage for liver-reserve co-control, not another broad surveillance trial. It tests whether 12 centres can deliver cause treatment, portal-hypertension care, nutrition, toxicity control and repeated liver-state measurement in 240 patients without delaying or weakening cancer treatment. A larger 600-person outcome trial is released only after delivery, safety, ALBI and tumour-treatment gates pass." ], "sourceIds": [ "tcga-2017-hcc", "chang-1997-hbv-vaccine-hcc", "liaw-2004-lamivudine-hbv", "rodriguez-2021-alcohol-abstinence-hcc", "zhang-2004-hcc-screening", "kawaguchi-2025-surf-hcc", "lyu-2026-torch-hcc", "mazzaferro-1996-hcc-transplant", "xia-2020-repeat-hcc-treatment", "zhang-2026-planet-hcc-recurrence", "bruix-2015-storm-sorafenib", "yopp-2026-imbrave050-update", "finn-2020-imbrave150", "qin-2025-cares310-final", "seif-el-dahan-2026-early-hcc-causes-death", "storandt-2024-atezo-bev-liver-function", "lim-2025-attention-hbv", "rhee-2025-miracle-hcc", "hirode-2026-hcc-biomarker-rct", "trinchet-2011-hcc-surveillance-interval", "singal-2024-tracer-protocol", "clinicaltrials-nct06084234-tracer", "ioannou-2026-premium-protocol", "clinicaltrials-nct05486572-premium", "sun-2026-talentop-hcc", "kudo-2026-emerald3" ], "links": [ { "target": "hepatobiliary", "relation": "part-of" }, { "target": "model-hcc-causal-chain", "relation": "contains" }, { "target": "disease-system", "relation": "part-of" }, { "target": "detection", "relation": "part-of" }, { "target": "treatment", "relation": "part-of" }, { "target": "claim-hcc-transplant-selection", "relation": "contains" }, { "target": "claim-hcc-repeat-treatment-sequence", "relation": "contains" }, { "target": "claim-hcc-alcohol-interception-window", "relation": "contains" }, { "target": "claim-hcc-tumour-and-liver-are-competing-failure-axes", "relation": "contains" }, { "target": "hypothesis-hcc-liver-reserve-co-control", "relation": "tests" }, { "target": "claim-hcc-cares310-survival", "relation": "contains" }, { "target": "company-elevar-therapeutics", "relation": "contains" }, { "target": "claim-hcc-stride-five-year-tail", "relation": "contains" }, { "target": "company-astrazeneca", "relation": "contains" }, { "target": "claim-hcc-early-hbv-treatment-signal-immature", "relation": "contains" }, { "target": "claim-hcc-surveillance-test-improvement-not-mortality", "relation": "contains" }, { "target": "trial-hcc-action-map", "relation": "contains" }, { "target": "trial-2026-randomised-result-stress-test", "relation": "contains" }, { "target": "unknown-hcc-surveillance-modality-utility", "relation": "contains" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "strong", "randomised": "strong", "independentReplication": "strong for viral prevention and advanced treatment; mixed for surveillance and recurrence routing", "sampleSize": "Vaccination, antiviral, screening, local-treatment and systemic randomized studies plus multiregion recurrence sequencing", "effect": "Viral prevention and selected surveillance reduce HCC burden; local and systemic control remain constrained by recurrence and liver function", "limits": [ "Metabolic liver-disease prevention remains incomplete", "Alcohol abstinence evidence is observational", "Early-HBV treatment evidence is interim", "The classic surveillance mortality trial predates modern treatment", "Modern comparative surveillance mortality results are pending", "Recurrence-origin classification is not ready to direct care", "Transplant evidence is highly selected and donor-limited", "Liver-preservation co-treatment has not been randomized" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The disease model joins causal liver injury, malignant clones, organ reserve, treatment and recurrence." }, "animalEvidence": { "rating": "moderate", "reason": "Models support several mechanisms, while the main prevention and treatment statements rest on human evidence." }, "inVitroEvidence": { "rating": "moderate", "reason": "Tumour and antiviral systems support parts of the chain but omit cirrhosis, surveillance and competing liver death." }, "reproducibility": { "rating": "moderate", "reason": "Viral prevention and advanced treatment repeat; modern surveillance mortality and liver co-control do not." }, "commercialConflicts": { "rating": "moderate", "reason": "The branch combines public prevention evidence with company-funded antiviral and cancer-treatment trials." }, "recency": { "rating": "current", "reason": "Completed and live evidence was checked through September 2026." } } }, "layout": { "x": 63, "y": 8 }, "updated": "2026-09-14" }, { "id": "hepatobiliary", "title": "Hepatobiliary cancers", "shortTitle": "Hepatobiliary", "type": "cancer", "status": "supported", "scope": [ "cholangiocarcinoma", "gallbladder", "hcc", "hepatobiliary" ], "summary": "Cholangiocarcinoma, gallbladder cancer and hepatocellular carcinoma are compared without merging materially different biology.", "content": "Hepatobiliary is a navigation and comparison layer, not one cancer. Hepatocellular carcinoma arises mainly from injured hepatocytes in a damaged liver. Cholangiocarcinoma arises along intrahepatic, perihilar or distal bile ducts. Gallbladder cancer follows its own mucosal, gallstone-linked and surgical routes. Pooling them can create a false average for prevention, imaging, molecular targets and treatment.\n\nThe shared engineering problems are real: the liver and ducts are difficult to sample repeatedly; jaundice and inflammation distort biomarkers; organ function constrains treatment; local recurrence and occult systemic disease compete; and several advanced trials enrol mixed biliary sites. The disease pages therefore preserve anatomical site, cause, liver reserve, molecular subgroup and line of treatment before allowing a comparison.\n\nWhat can transfer is a method. HCC shows that controlling a causal exposure plus surveillance in a high-risk organ population can reduce death. Cholangiocarcinoma shows that FGFR2, IDH1 and HER2 selection can create distinct treatment routes. Gallbladder cancer shows why common benign disease does not by itself justify preventive surgery. None of those facts proves the same intervention in PDAC or in another hepatobiliary cancer.", "sourceIds": [ "chang-1997-hbv-vaccine-hcc", "zhang-2004-hcc-screening", "jusakul-2017-cholangiocarcinoma", "nakamura-2015-biliary-genomics", "villard-2023-psc-surveillance", "gonzalez-2025-cholecystectomy-chile", "oh-2022-topaz1" ], "links": [ { "target": "disease-system", "relation": "part-of" }, { "target": "hcc", "relation": "contains" }, { "target": "cholangiocarcinoma", "relation": "contains" }, { "target": "gallbladder", "relation": "contains" }, { "target": "detection", "relation": "part-of" }, { "target": "treatment", "relation": "part-of" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "strong", "randomised": "strong but disease-specific", "independentReplication": "weak — mixed evidence", "sampleSize": "Cross-disease navigation built from prevention, surveillance, genomic and treatment studies", "effect": "Comparison framework; no pooled hepatobiliary effect estimate", "limits": [ "This page does not replace disease-specific evidence", "Mixed biliary trials can obscure site effects", "Causes and health-system pathways vary geographically", "No intervention transfers without target-disease evidence" ] }, "layout": { "x": 50, "y": 13 }, "updated": "2026-09-14" }, { "id": "hypothesis-angiogenin-tgfb-existing-data-test", "title": "A fixed high-angiogenin state predicts benefit from TGF-beta-pathway inhibition", "shortTitle": "Angiogenin selector retest", "type": "hypothesis", "status": "proposed", "scope": [ "pdac", "pan-cancer", "transfer-unproven" ], "summary": "Test the old galunisertib marker claim in stored NIS793 phase 3 samples before exposing another patient to TGF-beta blockade.", "content": "Rationale. The original 156-person galunisertib trial gave an uncertain overall result: OS 8.9 versus 7.1 months, HR 0.79 with a 95% credible interval of 0.59-1.09. A 2026 analysis reports that patients with high baseline angiogenin had OS 8.90 versus 5.65 months, HR 0.57, 95% CI 0.35-0.92. It links tumour angiogenin to macrophage EGFR, TGF-beta and TNF-alpha activity. Contrary evidence. NIS793 directly changed TGF-beta tissue signals but the 511-person phase 3 made OS, PFS, response and serious harm worse. The angiogenin result was selected years after the parent trial and has not been fixed and repeated. Assumptions. Stored baseline plasma is adequate, the published threshold is fully recoverable, missing samples do not select survivors, and antibody blockade tests enough of the claimed pathway to challenge the ALK5 result. Prediction. Apply the published angiogenin assay and threshold, unchanged, to stored baseline plasma from daNIS-2. A credible selector requires a treatment-by-marker interaction with two-sided P below 0.01, NIS793 OS HR at most 0.75 in the high group, no benefit claim in the low group, the same direction for PFS, and no severe-harm excess that removes the gain. Falsification experiment. Secure the analysis plan, code hash and threshold before group labels are revealed; assay blinded samples; retain all 490 randomized participants and document every missing sample. Falsify if the interaction misses, the high-group OS interval includes 1, missingness exceeds 20% or differs by group, harm removes benefit, or the threshold is changed after results. This is a data and sample study, not a new treatment cohort. Mission cost is capped at £1m only after Novartis grants complete outcome linkage, assay access and publication rights. Time is 12 months. Major risks are false confirmation through threshold drift, selected sample availability, broken outcome linkage and loss of publication rights; no patient receives NIS793.", "contentSections": [ "Supporting evidence is the retrospective high-angiogenin survival difference and its linked tumour-macrophage experiments; it is deliberately held below independent confirmation." ], "sourceIds": [ "melisi-2018-galunisertib", "pietrobono-2026-angiogenin-alk5", "clinicaltrials-nct04935359-danis2", "pelletier-2026-nis793-biomarkers" ], "links": [ { "target": "experiments", "relation": "part-of" }, { "target": "claim-nis793-target-engagement-without-benefit", "relation": "must-explain" }, { "target": "failure-immune-reprogramming-without-patient-benefit", "relation": "learns-from" }, { "target": "trial-immune-reprogramming-action-map", "relation": "implemented-by" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "moderate", "randomised": "strong", "independentReplication": "none", "sampleSize": "Original randomized n=156; potential independent randomized archive n=490 plus n=21 run-in", "effect": "Retrospective high-angiogenin OS HR 0.57; fixed independent interaction not yet tested", "limits": [ "NIS793 and galunisertib block different parts of the pathway", "Stored sample availability is unconfirmed", "Threshold details must be recoverable", "A negative antibody test may not fully rule out ALK5 inhibition" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The marker is linked to a measured tumour-macrophage resistance chain." }, "animalEvidence": { "rating": "strong", "reason": "Angiogenin-dependent combination effects repeated across mouse models." }, "inVitroEvidence": { "rating": "strong", "reason": "The EGFR-TGF-beta-TNF-alpha route was perturbed in cell and co-culture systems." }, "reproducibility": { "rating": "weak", "reason": "The human selector is retrospective and from one parent trial." }, "commercialConflicts": { "rating": "moderate", "reason": "The galunisertib developers performed the reuse analysis; Novartis controls the validation archive." }, "recency": { "rating": "current", "reason": "The marker paper and phase 3 result appeared in 2026." } } }, "layout": { "x": 292, "y": 230 }, "updated": "2026-09-15" }, { "id": "hypothesis-asp1929-first-ten-measurement-gate", "title": "Make the first ten ASP-1929 participants a binding, light and safety gate", "shortTitle": "ASP-1929 first-ten gate", "type": "hypothesis", "status": "testing", "scope": [ "pdac", "photoimmunotherapy", "trial-design", "dosimetry" ], "summary": "Before the first infusion, the planned safety run-in can be made decision-ready by measuring every link from EGFR binding through light coverage to systemic-treatment delay.", "content": "Biological rationale. ASP-1929 should bind EGFR-positive cells and rupture their membranes only where 690 nm light reaches the dye. Supporting evidence. EGFR appears in many pancreatic cancers; the platform has regulated head-and-neck use in Japan; inserted light improves depth in animal work. Contrary evidence. Pancreatic human treatment is zero, EGFR is not universal, ordinary cetuximab failed despite 90% expression and the planned study has no randomized control. Assumptions. Before enrollment, Brown University and Rakuten Medical will add an independent data monitor, all-enrolled flow, whole-section EGFR measurement, drug fluorescence where feasible, fibre coordinates, calibrated light output, spatial injury maps, chemotherapy interruption, symptoms and public participant-level data stripped of identity. Prediction. At least 8/10 safety-run-in participants will be illuminated as planned; at least 80% of their intended viable target will show both adequate drug binding and light coverage; no more than one will have grade 3 or worse treatment-related harm; planned systemic therapy will not be delayed by more than 14 extra days. Falsification experiment. Invest £0 in company equity and fund no expansion drug or device. Spend at most £300,000 on independent measurement and monitoring for the first ten only, with payment after raw-file release. Pause after each of the first three and again at ten. Stop if fewer than eight receive illumination, coverage falls below 80% in more than two, two or more have severe related harm, any resectable participant loses an operation because of drug or timing, or systemic care slips beyond 14 extra days. If the gate passes, finish the descriptive cohort under the amended plan. Do not call its six-month progression rate efficacy; a later comparison must hold systemic care constant and randomize illumination. Major safety risks are cetuximab infusion reaction, photosensitivity, bleeding, pancreatitis, bowel or vessel injury, infection, lost resection opportunity and delayed chemotherapy.", "sourceIds": [ "clinicaltrials-nct07698613-asp1929-pancreas", "pmda-2020-akalux-review", "rakuten-medical-2026-asp1929", "kim-2015-egfr-expression-pdac", "maruoka-2018-interstitial-photoimmunotherapy", "philip-2010-swog-s0205-cetuximab" ], "links": [ { "target": "experiments", "relation": "part-of" }, { "target": "unknown-asp1929-pancreatic-target-coverage-and-benefit", "relation": "tests" }, { "target": "failure-asp1929-selection-light-and-endpoint-design", "relation": "repairs" }, { "target": "trial-pdac-asp1929-action-map", "relation": "implemented-by" }, { "target": "company-rakuten-medical", "relation": "depends-on" }, { "target": "programme-5m-decision-core", "relation": "constrained-by" } ], "evidence": { "humanProspective": "none — planned first-human pancreatic study", "humanRetrospective": "none in pancreatic cancer", "randomised": "none", "independentReplication": "none", "sampleSize": "First ten planned participants; cap £300,000 for measurement and monitoring only", "effect": "Convert an uncontrolled safety cohort into a hard go or stop test of delivery and harm", "limits": [ "No efficacy control", "Some binding measurements may require resection tissue", "Fluorescence may not equal active drug", "Coverage threshold is a mission choice", "Ten people give wide harm uncertainty", "Company and sponsor agreement unconfirmed" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "The gate directly measures the drug-target-light chain in human pancreatic tissue." }, "animalEvidence": { "rating": "moderate", "reason": "Animal depth results set measurement needs but not release thresholds." }, "inVitroEvidence": { "rating": "moderate", "reason": "Targeted light killing supports the test but not the clinical decision." }, "reproducibility": { "rating": "weak", "reason": "Independent monitoring and raw release begin reproducibility; the sample remains small." }, "commercialConflicts": { "rating": "high", "reason": "Funds are conditional on independent custody and publication rights around a company-owned platform." }, "recency": { "rating": "current", "reason": "The gate can be inserted before the planned 2026 start." } } }, "layout": { "x": 1024, "y": 948 }, "updated": "2026-09-15" }, { "id": "hypothesis-aspire-data-rescue-gate", "title": "ASPIRE's existing randomized record can still produce a trustworthy decision", "shortTitle": "ASPIRE rescue gate", "type": "hypothesis", "status": "testing", "scope": [ "pdac", "trial-delivery", "transfer-unproven" ], "summary": "Audit custody and completeness before funding rescue; use the existing assignments, not another ivospemin cohort.", "content": "Biological rationale. Ivospemin changes a cancer growth supply system and produced an uncontrolled response signal, but it also produced serious liver and eye toxicity; benefit and harm require a direct comparison. Supporting evidence. ASPIRE used a masked placebo control, reported a 395-person safety database and operated across 89 sites before its contractor transition. Contrary evidence. The actual assigned denominator, arm-level outcomes, analysis plan, assignment-key custody and follow-up completeness are not public; the U.S. and European status records conflict. Assumptions. Patient consent permits survival follow-up and independent analysis, the blind and assignment key survive, raw records can be copied from sponsor, contractor and sites, and enough assigned patients have outcome data. Prediction. Within three months, independent diligence can reconcile at least 400 assigned patients and locate the assignment key, protocol versions, safety reports and at least 90% 12-month vital status with no more than ten percentage points arm difference in missingness. If that passes, a fixed all-randomized analysis completed within another nine months can estimate overall survival and severe eye and liver harm without enrolling anyone else. Falsification experiment. Spend at most £250,000 from the existing data and statistics line on legal custody, data hashes, assignment reconciliation and a blinded missingness report. Stop if consent blocks transfer, the key or raw database is missing, fewer than 400 assignments reconcile, vital status is below 90%, arm missingness differs by more than ten points, or an independent committee cannot control the analysis. Do not pay rescue invoices, license the drug, buy equity or expose another patient before this gate. Any later rescue budget must be itemized and remove the same amount from another line inside the £500m ceiling. Major safety risks are privacy breach, accidental unmasking, corrupt or selectively missing records, unpaid-site leverage, false reassurance from a monitoring-board continuation, and undercounted irreversible sight or liver injury.", "sourceIds": [ "clinicaltrials-nct05254171-aspire", "ctis-2024-514714-12-00-aspire", "panbela-2024-q3-10q", "panbela-2025-note-default", "singhal-2021-sbp101-phase1", "carla-2024-ivospemin-retinal-atrophy" ], "links": [ { "target": "experiments", "relation": "part-of" }, { "target": "unknown-aspire-assigned-denominator-and-results", "relation": "tests" }, { "target": "failure-aspire-trial-operational-collapse", "relation": "recovers" }, { "target": "trial-pdac-polyamine-action-map", "relation": "implemented-by" }, { "target": "programme-500m-integrated-mission", "relation": "constrained-by" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong", "independentReplication": "weak", "sampleSize": "Diligence gate at 400 reconciled assignments; follow-up gate at 90%; no new participants", "effect": "A recoverable randomized record could answer more than another uncontrolled cohort at far lower cost", "limits": [ "Thresholds are mission choices", "Current data controller is unknown", "Actual record completeness may be below the gate", "A rescued trial may still be underpowered", "No efficacy result is assumed" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "The drug question remains plausible enough to justify preserving randomized evidence, not enough to dose more patients." }, "animalEvidence": { "rating": "moderate", "reason": "Model evidence supports only the value of resolving the human comparison." }, "inVitroEvidence": { "rating": "strong", "reason": "Cell effects show target activity but do not lower the data-integrity gate." }, "reproducibility": { "rating": "weak", "reason": "No independent controlled efficacy result exists." }, "commercialConflicts": { "rating": "high", "reason": "Company distress, unpaid vendors and ownership claims can affect access and reporting." }, "recency": { "rating": "current", "reason": "The planned milestones have passed and rescue value decays as sites close and records scatter." } } }, "layout": { "x": 502, "y": 426 }, "updated": "2026-09-15" }, { "id": "hypothesis-autogene-cevumeran-existing-trial-gate", "title": "The autogene-cevumeran package reduces recurrence after PDAC surgery", "shortTitle": "RNA vaccine trial gate", "type": "hypothesis", "status": "testing", "scope": [ "pdac", "transfer-unproven" ], "summary": "Use IMCODE003's existing allocation; require a clinically useful all-assigned effect, full manufacture accounting and harm reporting before advancing the regimen.", "content": "Biological rationale. Surgery lowers the number of cancer cells, personalized RNA can teach T cells to recognize each tumour's mutations, and low burden gives those cells a more tractable search problem. Supporting evidence. In 16 vaccinated people, eight developed large, durable neoantigen-specific T-cell responses; their median recurrence-free survival was not reached versus 13.4 months among eight non-responders. Contrary evidence. This was a response-defined comparison in a one-arm combination study, and four other randomized PDAC vaccine programmes totaling 1,812 participants failed their main clinical endpoint. Assumptions. IMCODE003 preserves every randomized patient, manufactures and gives the vaccine quickly enough, maintains comparable chemotherapy, and reports every failure from tissue receipt to completed doses. Prediction. In all assigned patients, the complete autogene-cevumeran, atezolizumab and mFOLFIRINOX package produces a disease-free-survival hazard ratio at most 0.75, a three-year absolute disease-free-survival gain of at least ten percentage points, a 95% confidence interval excluding no effect, and no more than five percentage points excess grade 3 or worse treatment-related harm. Falsification experiment. Recruit no new mission cohort. Lock the existing 260-person primary analysis; publish the screening, tissue, manufacture, delivery and dose denominators by arm; report recurrence, new cancers, death, overall survival and harm for every assigned participant. Falsify if the registered primary endpoint misses, the effect is smaller than the mission threshold, the confidence interval includes no effect, harm crosses the stop line, or benefit appears only after excluding manufacture failures or selecting immune responders. A positive result establishes the package, not the vaccine's separate contribution, because atezolizumab is added in the same arm. Mission cost is £0 for patients and product and at most £250,000 after database lock for independent analysis if individual-level access is granted; time is set by the existing January 2031 primary-completion estimate. Major safety risks are immune inflammation from atezolizumab, chemotherapy toxicity, infusion reactions, treatment delay and recurrence while a personalized product is made.", "sourceIds": [ "rojass-2025-neoantigen-followup", "nct05968326-imcode003", "middleton-2014-telovac", "le-2019-eclipse", "hewitt-2022-algenpantucel-l", "elicio-2026-amplify7p-results" ], "links": [ { "target": "experiments", "relation": "part-of" }, { "target": "unknown-pdac-vaccine-regimen-contribution", "relation": "resolves" }, { "target": "failure-pdac-vaccine-escalation-reversals", "relation": "learns-from" }, { "target": "trial-pdac-vaccine-action-map", "relation": "implemented-by" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "moderate", "independentReplication": "none", "sampleSize": "Existing randomized phase 2 estimated n=260 across 87 sites", "effect": "Mission gate: DFS HR at most 0.75, at least ten-point three-year DFS gain and 95% CI excluding no effect", "limits": [ "Thresholds are mission choices", "Open-label trial", "Enrollment is still labelled estimated", "The design cannot isolate vaccine from atezolizumab", "Primary completion is estimated 2031", "Independent data access is conditional" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Human data show durable vaccine-induced T cells at minimal disease burden." }, "animalEvidence": { "rating": "strong", "reason": "Models support the route but are discounted after prior clinical reversals." }, "inVitroEvidence": { "rating": "strong", "reason": "Neoantigen recognition and clone persistence can be tested directly." }, "reproducibility": { "rating": "none", "reason": "The proposed patient benefit has not been reproduced or shown once in a randomized PDAC trial." }, "commercialConflicts": { "rating": "high", "reason": "The product and trial are controlled by Genentech and BioNTech." }, "recency": { "rating": "current", "reason": "The gate uses the September 2026 registry record." } } }, "layout": { "x": 420, "y": 360 }, "updated": "2026-09-15" }, { "id": "hypothesis-bacterial-gemcitabine-deactivation-gate", "title": "Ciprofloxacin improves gemcitabine benefit only when it removes bacterial drug deactivation", "shortTitle": "Bacterial gemcitabine gate", "type": "hypothesis", "status": "proposed", "scope": [ "pdac", "transfer-unproven" ], "summary": "Use the existing blinded PANORAMIX allocation to connect tumour bacteria, active drug, resistance, toxicity and survival.", "content": "Biological rationale. Some tumour bacteria express long-form cytidine deaminase, an enzyme that turns gemcitabine into inactive dFdU. Removing those bacteria should raise active drug exposure, but can also raise toxicity and select resistant organisms. Supporting evidence. Bacteria appeared in 86/113 human PDAC samples; ciprofloxacin reversed bacterial resistance in mice. Three non-randomized human cohorts associate antibiotics with longer progression or survival. Contrary evidence. Those cohorts assigned no antibiotic and could reflect infection, timing or treatment selection; one 99-person study did not show an OS difference, and a 430-person trial reanalysis linked antibiotics to severe adverse events HR 1.77. No randomized patient-benefit result exists. Assumptions. PANORAMIX Step 2 retains the placebo comparison, stores tumour and stool samples, measures gemcitabine and dFdU, and reconciles its planned denominator: CTIS says 142 across the whole trial while ClinicalTrials.gov says 206 and omits Step 2 outcomes. Prediction. Among all Step 2 randomized participants, ciprofloxacin reduces the fixed tumour bacterial cytidine-deaminase signal, increases the gemcitabine-to-dFdU exposure ratio, and improves six-month survival by at least ten points or PFS with HR at most 0.75, without more than five points excess severe toxicity or ten points excess resistant organisms. Falsification experiment. Recruit no separate cohort. Before unmasking, publish the exact Step 2 denominator, endpoint order and assay; analyse every assigned patient; test intervention-to-microbe, microbe-to-drug and drug-to-outcome links; report OS, PFS, response, function, severe harm, infection and resistance. Falsify if ciprofloxacin does not change the target function, target change does not change drug exposure, benefit intervals include no effect, serious harm crosses the stop line, or the signal appears only in a data-made subgroup. Estimated mission cost is at most £1m for fixed tumour assays and independent analysis inside the existing trial, and £0 for new participants; time is 30 months to the six-month endpoint after last Step 2 assignment. Major risks are antibiotic resistance, infection, tendon or vascular toxicity, confounded rescue antibiotics, sample contamination, weak tumour-sample coverage and a changing chemotherapy backbone.", "sourceIds": [ "geller-2017-bacterial-gemcitabine", "nakano-2020-antibiotics-gnp", "gong-2025-antibiotics-adjuvant-pdac", "corty-2020-antibiotic-gemcitabine-toxicity", "ctis-panoramix-2024-517766-41", "clinicaltrials-nct07028424-panoramix" ], "links": [ { "target": "experiments", "relation": "part-of" }, { "target": "unknown-microbiome-treatment-contribution-pdac", "relation": "resolves" }, { "target": "failure-microbiome-proxy-without-patient-benefit", "relation": "learns-from" }, { "target": "trial-pdac-microbiome-action-map", "relation": "implemented-by" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "moderate", "randomised": "weak", "independentReplication": "moderate", "sampleSize": "Existing phase 2 programme: CTIS n=142 across both steps; ClinicalTrials.gov n=206 across both steps; exact Step 2 denominator unresolved", "effect": "Existing-trial causal-chain gate; no assigned benefit yet", "limits": [ "The sample-size conflict must be resolved", "Tumour enzyme measurement is not listed as an endpoint", "The proposed thresholds are mission choices", "Six-month survival in only the experimental arm is the registered primary endpoint" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The enzyme, active drug and inactive product make a testable chain." }, "animalEvidence": { "rating": "strong", "reason": "Ciprofloxacin reversed bacteria-driven gemcitabine resistance in mice." }, "inVitroEvidence": { "rating": "strong", "reason": "Bacterial gemcitabine conversion was directly measured." }, "reproducibility": { "rating": "moderate", "reason": "Human antibiotic associations repeat, but causal benefit does not." }, "commercialConflicts": { "rating": "low", "reason": "Both drugs are established; specimen and analysis access remain controlled." }, "recency": { "rating": "current", "reason": "PANORAMIX began recruiting in February 2026." } } }, "layout": { "x": 378, "y": 330 }, "updated": "2026-09-15" }, { "id": "hypothesis-cca-route-specific-action-platform", "title": "Hypothesis: one data standard can rank three separate cholangiocarcinoma action routes", "shortTitle": "CCA route-specific action platform", "type": "hypothesis", "status": "hypothesis", "scope": [ "cholangiocarcinoma", "hepatobiliary" ], "summary": "A £6 million first stage should compare PSC interception, liver-fluke prevention and postoperative residual-disease action without pretending that their populations or endpoints are interchangeable.", "content": "Biological rationale: PSC-related duct cancer, liver-fluke-related carcinogenesis and postoperative residual disease arise at different points and require different actions, but all three can use a common record for exposure, anatomical site, test, work-up, treatment, harm and survival. Supporting evidence: PSC surveillance supplies a measured 23.8% abnormal-imaging workload; a Thai One Health programme supplies an infection-control signal; resected biliary cohorts supply a strong ctDNA recurrence signal; and surgery still leaves 59.4% recurrence in a large intrahepatic cohort. Evidence against: annual PSC MRI did not create durable curable detection, fluke programmes do not yet show fewer cancers, and no ctDNA-guided biliary treatment has improved survival. Assumptions: at least 6,000 PSC records, 10,000 participants in endemic control programmes and 600 resected biliary cancers can be linked or assembled across existing networks; anatomical site and cause remain separate; treatment results return within 21 days; and national linkage captures stage, surgery, serious harm and death. Prediction: a locked comparison will identify at most one route ready for a full patient-benefit trial and will prevent money being split across three weak expansions. Study and estimated cost: spend £6m over three years from the existing £20m related-cancer line—£1.5m for the PSC data and specimen federation, £1.5m for liver-fluke cluster follow-up and registry linkage, £2m for a 600-person site-stratified postoperative assay and action run-in, and £1m for independent statistics, patient governance and common data checks. If the Swedish target-state rate of 13/512 over five years held, 6,000 people with four years average follow-up would supply about 122 dysplasia or cancer events: 6,000 × 4 × (13 / 512 / 5) = 121.9. If 59.4% recurrence held, 600 resected cancers would supply about 356 recurrences: 600 × 0.594 = 356. These are planning estimates, not promised yields. Falsification: stop the shared platform if common fields force site or cause to be pooled, outcome linkage is below 85%, or independent teams cannot reproduce the locked calculations. PSC expansion stops if benign invasive work exceeds ten procedures per additional resectable high-grade dysplasia or cancer, action completion is below 80%, or stage does not move. Liver-fluke expansion stops if infection or reinfection is not reduced by at least 50% at 24 months or long-term cancer linkage cannot be secured. Residual-disease expansion stops if panel success is below 90%, results arrive before action in fewer than 85%, or no safe named treatment exists. Only the best passing route may seek up to £6m more from the uncommitted reserve; the other £6m of the related-cancer line remains for HCC. Major safety risks include benign ERCP or biopsy, false reassurance, delayed cancer, unnecessary systemic treatment, treatment toxicity, stigma around food practices, privacy loss and unequal access to surgery or transplant.", "sourceIds": [ "villard-2023-psc-surveillance", "tan-2024-psc-mrcp-surveillance", "iarc-2012-liver-flukes", "charoensuk-2024-one-health-opisthorchis", "clinicaltrials-nct05321992-echec", "hu-2020-icc-resection-recurrence", "yu-2025-btc-ctdna-recurrence", "yoo-2025-extrahepatic-cca-ctdna", "park-2026-btc-ctdna-feasibility", "clinicaltrials-nct05743959-btc-mrd", "clinicaltrials-nct02170090-acticca1" ], "links": [ { "target": "experiments", "relation": "part-of" }, { "target": "trial-cca-action-map", "relation": "depends-on" }, { "target": "unknown-cca-surveillance-utility", "relation": "tests" }, { "target": "unknown-cca-postoperative-mrd-action", "relation": "tests" }, { "target": "failure-psc-surveillance-no-curable-detection", "relation": "succeeds-where" }, { "target": "failure-echec-cancer-outcome-mismatch", "relation": "succeeds-where" }, { "target": "failure-cca-mrd-prediction-without-action", "relation": "succeeds-where" }, { "target": "programme-500m-integrated-mission", "relation": "constrains" } ], "evidence": { "humanProspective": "moderate — all three component routes have human prospective data", "humanRetrospective": "strong for care, recurrence and surgical burden", "randomised": "none for the proposed route-ranking platform", "independentReplication": "none for the combined design", "sampleSize": "Proposed linkage or run-in: 6,000 PSC records, 10,000 endemic-area participants and 600 resected biliary cancers", "effect": "Unobserved; designed to rank at most one route for expansion under explicit action and harm gates", "limits": [ "Existing records may not share fields", "Event-rate transport may fail", "£6m funds ranking and readiness, not definitive mortality trials", "Long-term fluke cancer outcomes exceed three years", "Rare anatomical subgroups reduce power", "Costs are planning estimates" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Each lane connects a distinct biological state to a specific action and endpoint." }, "animalEvidence": { "rating": "moderate", "reason": "Animal evidence supports fluke carcinogenesis and some treatment routes, but not platform utility." }, "inVitroEvidence": { "rating": "moderate", "reason": "Assays and molecular targets are workable components, while their care effects remain unproved." }, "reproducibility": { "rating": "none", "reason": "No independent programme has applied these shared gates across all three routes." }, "commercialConflicts": { "rating": "high", "reason": "The design decides whether imaging, assay and drug programmes expand, so independent ownership and analysis are essential." }, "recency": { "rating": "current", "reason": "The proposal responds to human and registry evidence checked through September 2026." } } }, "updated": "2026-09-14" }, { "id": "hypothesis-cldn182-adc-survival-test", "title": "Hypothesis: CLDN18.2-guided exatecan delivery extends useful life in selected PDAC", "shortTitle": "IBI343 survival test", "type": "hypothesis", "status": "hypothesis", "scope": [ "pdac" ], "summary": "The active masked phase 3 can test whether the IBI343 carrier turns a failed free payload into a survival-positive selected treatment.", "content": "Biological rationale. Free exatecan exposes healthy and tumour tissue without a reliable address. IBI343 should concentrate the same DNA-damaging payload in CLDN18.2-bearing cells after antibody binding and internalisation. Supporting evidence. At 6 mg/kg, ten of 44 high-marker PDACs responded versus none of 12 below the analysis cutoff; the carrier has a stated surface address and no treatment-emergent death occurred in 83 treated people. Contrary evidence. Free exatecan plus gemcitabine failed randomized survival; plain CLDN18.2 antibody produced OS HR 0.999; other antibody-drug carriers produced zero or weak uncontrolled response; and IBI343 itself has no concurrent PDAC control. Assumptions. The central assay is accurate across sites and lesions, the cutoff is fixed before allocation, enough marker-bearing cells internalise the carrier, payload reaches the nucleus, resistance does not erase the effect, supportive care is balanced, and masking survives recognizable toxicity. Prediction. In all 201 assigned participants, IBI343 produces OS HR at or below 0.75 with a 95% interval excluding 1.0, positive quality-adjusted survival, and a larger effect in a marker gradient fixed before outcomes without more than ten points of uncompensated severe harm. Falsification experiment. Use the existing G-HOPE-002 masked 2:1 comparison; recruit no mission cohort. Falsify a useful late-line treatment if the OS interval includes 1.0, missing outcomes can move HR above 0.85, quality-adjusted survival is not positive, severe harm rises more than ten points without compensation, the marker rule changes after outcomes, or benefit appears only in a post-outcome subgroup. Estimated cost is £0 to the mission because Innovent and Takeda fund development. Estimated time is about 22 months to the June 2028 registry completion estimate, with protocol repair required now. Major safety risks are low blood counts, infection, bleeding, fatigue, nausea, payload damage to healthy tissue, hospitalization and masking failure caused by distinctive toxicity.", "sourceIds": [ "yu-2025-ibi343-pdac", "clinicaltrials-nct07066098-ibi343-phase3", "abou-alfa-2006-exatecan-pdac", "park-2026-gleam-zolbetuximab", "astellas-2026-gleam-medical-information", "innovent-2026-interim-results", "takeda-2025-ibi343-license" ], "links": [ { "target": "experiments", "relation": "part-of" }, { "target": "claim-cldn182-adc-selected-response", "relation": "supported-by" }, { "target": "failure-antibody-target-and-payload-transfer-pdac", "relation": "constrained-by" }, { "target": "unknown-cldn182-adc-survival-effect", "relation": "tests" }, { "target": "trial-cldn182-antibody-action-map", "relation": "gated-by" }, { "target": "company-innovent-biologics", "relation": "funded-by" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "not-applicable", "randomised": "none", "independentReplication": "none", "sampleSize": "Existing randomized masked phase 3 planned n=201; no mission recruitment", "effect": "Predicted OS HR at most 0.75 with positive quality-adjusted survival and controlled harm", "limits": [ "Sponsor power assumptions unavailable", "Protocol unavailable", "Marker cutoff absent from registry", "Late-line setting", "One currently listed site", "Thresholds are mission decisions" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The hypothesis requires a defined surface address, carrier uptake and payload action in the same selected tumour." }, "animalEvidence": { "rating": "moderate", "reason": "Animal activity supports the carrier idea but cannot establish therapeutic ratio in people." }, "inVitroEvidence": { "rating": "strong", "reason": "Every link from binding through DNA damage can be measured and joined to outcome." }, "reproducibility": { "rating": "none", "reason": "The human response gradient has not repeated independently or randomly." }, "commercialConflicts": { "rating": "high", "reason": "The funders own the product, assay strategy and trial data." }, "recency": { "rating": "current", "reason": "The falsification experiment is the active phase 3 programme." } } }, "layout": { "x": 226, "y": 144 }, "updated": "2026-09-15" }, { "id": "hypothesis-cnsi-fe-current-study-mechanism-gate", "title": "Use the live CNSI-Fe study to test tumour-cell ferroptosis before testing efficacy", "shortTitle": "CNSI-Fe mechanism gate", "type": "hypothesis", "status": "testing", "scope": [ "pdac", "pan-cancer", "ferroptosis", "nanoparticle-delivery", "capital-allocation" ], "summary": "Fund no new cohort; require paired injected and uninjected tissue, exact spatial drug maps and cell-specific death measurements in the ongoing expansion.", "content": "Biological rationale. Carbon particles may retain ferrous iron inside an injected lesion, where iron can drive lipid-membrane oxidation. Supporting evidence. CNSI-Fe produces central scan necrosis in people; PDAC cells have measured ferroptosis dependencies; tumour-selective GPX4 loss works in mouse models. Contrary evidence. Human ferroptosis was never measured, growth-rate change was non-significant, only two pancreatic people were treated and pancreas-wide ferroptotic damage can promote KRAS tumours in mice. Assumptions. Enray and the hospitals running NCT07433283 will permit independent tissue custody, exact injection maps, negative publication and a pancreatic label only if enough pancreatic samples exist. Prediction. In at least 12 people with paired viable injected-rim and uninjected-lesion samples, the product will create a coherent chain: retained ferrous iron, at least two independent lipid-oxidation measures, loss of GPX4 or glutathione repair, ferroptosis-shaped membrane injury and tumour-cell death without equal injury in CD8, dendritic or natural-killer cells. Falsification experiment. Buy £0 of parent-company or Enray equity and fund no product or participant. Spend at most £400,000 for independent analysis of existing expansion samples and a four-way patient-derived co-culture test: vehicle, carbon particles, matched ferrous iron and CNSI-Fe, each with and without the proposed rescue reagent. Require at least three pancreatic models and three non-pancreatic models. Stop the pancreatic branch if fewer than three paired pancreatic samples exist, tumour coverage is below 70%, immune-cell injury reaches half the tumour-cell injury, the combined carrier effect does not exceed matched iron in at least four of six models, or a ferroptosis rescue fails in at least four of six. Estimated time. Twelve months from complete sample access. Do not call later drug response resensitization. That claim needs a later randomized same-drug rechallenge with uninjected disease and survival, which receives £0 until this gate passes. Major risks are injection injury, bleeding, iron toxicity, inflammatory tumour promotion, immune-cell loss, resistant viable rim and delayed systemic treatment.", "sourceIds": [ "xie-2026-cnsi-fe-first-human", "clinicaltrials-nct06048367-cnsi-fe", "clinicaltrials-nct07433283-cnsi-fe-expansion", "daher-2019-xct-pdac-ferroptosis", "dai-2020-ferroptotic-damage-pdac", "li-2023-tumour-selective-gpx4-degrader", "caneque-2025-fentomycin-lysosomal-iron", "patent-us20250144252a1-cnsi-fe" ], "links": [ { "target": "experiments", "relation": "part-of" }, { "target": "unknown-ferroptosis-cell-type-coverage-and-human-benefit", "relation": "tests" }, { "target": "failure-cnsi-fe-resensitization-and-mechanism-claim", "relation": "repairs" }, { "target": "trial-pdac-ferroptosis-action-map", "relation": "implemented-by" }, { "target": "company-sichuan-enray-pharmaceutical", "relation": "depends-on" }, { "target": "programme-5m-decision-core", "relation": "constrained-by" } ], "evidence": { "humanProspective": "weak", "humanRetrospective": "weak", "randomised": "none — mechanism gate only", "independentReplication": "weak — planned independent analysis only", "sampleSize": "At least 12 paired human cases and six patient-derived models; zero new people; cap £400,000", "effect": "Test whether the product produces selective human tumour-cell ferroptosis rather than non-specific local necrosis", "limits": [ "Paired biopsies may not exist", "Only three pancreatic models required", "Rescue reagents are ex-vivo tools", "Tissue timing varies", "Local mechanism cannot prove systemic benefit", "Company agreement unconfirmed", "Thresholds are mission choices" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The gate measures iron, lipid damage, repair failure, cell identity and rescue in one chain." }, "animalEvidence": { "rating": "strong", "reason": "Opposite mouse outcomes make selectivity and immune direction mandatory tests." }, "inVitroEvidence": { "rating": "strong", "reason": "Component and rescue tests can identify the claimed death mechanism before efficacy expansion." }, "reproducibility": { "rating": "moderate", "reason": "Independent custody and six models provide a first cross-system check." }, "commercialConflicts": { "rating": "high", "reason": "Payment is conditional on independent samples, raw data and negative publication around a patented company product." }, "recency": { "rating": "current", "reason": "The 54-person expansion is recruiting and should complete in 2026." } } }, "layout": { "x": 1186, "y": 1110 }, "updated": "2026-09-15" }, { "id": "hypothesis-cryoablation-pragmatic-pain-trial", "title": "Test percutaneous tumour cryoablation as pain care, not as an implied cure", "shortTitle": "Cryo pain trial", "type": "hypothesis", "status": "testing", "scope": [ "pdac", "unresectable", "pain", "randomized-trial" ], "summary": "A 100-person randomized palliative trial can test the strong pain signal with complete death handling, standard opioid units and days-at-home accounting.", "content": "Biological rationale. Freezing a painful pancreatic mass may reduce local pressure, inflammation and nerve activation while the visible ice ball allows controlled partial treatment. Supporting evidence. In 11 selected cases, one-month pain fell 3.27 points on a 10-point scale and all reduced opioids. Contrary evidence. There was no comparator, opioid doses were not standardized and later scores retained survivors only; open cryo caused pancreatic fluid leakage in 18/24. Assumptions. At least five centres can identify people with a safe percutaneous path, moderate-to-severe pain despite optimized care and expected survival beyond eight weeks. Prediction. Percutaneous intratumour cryoablation plus standard care will improve the four-week pain-interference score by at least two points more than standard interventional pain care, reduce daily oral morphine-equivalent dose by at least 25% and not reduce days alive at home. Sample logic. With standard deviation 2.5, alpha 0.05, power 90% and desired difference 2, a simple two-group approximation gives 2 × (1.96 + 1.28)^2 × 2.5^2 / 2^2 = 32.8 per arm; dividing by 0.8 for death and missingness gives 41. Enroll 50 per arm, 100 total. Falsification experiment. Spend up to £3 million only after a £250,000 protocol and site-readiness gate. Randomize against the locally accepted best pain route on identical cancer care. Count death before four weeks as failure. Stop for less than 20% safe-path eligibility, two treatment-related deaths, bowel injury above 3%, grade 3 or worse pancreatitis above 5%, median chemotherapy delay above seven days, pain difference below one point at the planned interim or fewer days alive at home. Do not claim tumour control from pain relief. Major risks are bleeding, infection, fistula, pancreatitis, stomach or bowel injury, incomplete freezing, opioid withdrawal error and diverted time near death.", "sourceIds": [ "pusceddu-2026-percutaneous-cryoablation", "kang-2025-laparotomic-cryoablation", "song-2014-cryo-bypass-comparison" ], "links": [ { "target": "experiments", "relation": "part-of" }, { "target": "unknown-cryoablation-randomized-pain-and-net-benefit", "relation": "tests" }, { "target": "failure-cryoablation-selection-attrition-and-route-harm", "relation": "repairs" }, { "target": "trial-pdac-cryoablation-action-map", "relation": "implemented-by" }, { "target": "programme-5m-decision-core", "relation": "constrained-by" } ], "evidence": { "humanProspective": "weak", "humanRetrospective": "moderate", "randomised": "none — proposed", "independentReplication": "weak", "sampleSize": "Protocol gate £250,000; proposed randomized n=100; trial cap £3 million", "effect": "Powered for a two-point between-arm pain difference at four weeks with death treated as failure", "limits": [ "Effect and variance come from a tiny uncontrolled series", "Comparator varies by centre", "Blinding is difficult", "Rare harms remain imprecise", "Eligibility may be low", "Budget and thresholds are mission choices", "The trial is palliative, not survival-powered" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The trial connects local freezing to pain and opioid outcomes." }, "animalEvidence": { "rating": "not-applicable", "reason": "The decision needs a comparative human palliative result." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Cell destruction cannot estimate pain interference or days at home." }, "reproducibility": { "rating": "weak", "reason": "Multicentre randomization and common outcome capture are designed to improve it." }, "commercialConflicts": { "rating": "moderate", "reason": "Procure by performance specification and keep analysis independent of any needle supplier." }, "recency": { "rating": "current", "reason": "The design responds to the strongest 2026 human signal." } } }, "layout": { "x": 862, "y": 786 }, "updated": "2026-09-15" }, { "id": "hypothesis-direct-randomized-data-recovery-gate", "title": "Recover the existing DIRECT randomization before exposing another patient", "shortTitle": "DIRECT data gate", "type": "hypothesis", "status": "testing", "scope": [ "pdac", "local-control", "data-recovery" ], "summary": "A capped independent recovery of NCT03899636 can decide whether the device adds useful life, whether the experiment is irretrievable, or whether a new trial is justified.", "content": "Biological rationale. IRE can destroy a bounded local volume, so it may help patients whose systemic disease remains controlled after chemotherapy; it cannot control hidden distant cells. Supporting evidence. The observational registry reports an 18-versus-10-month survival association and human procedure experience. The separate phase 3 design cleanly adds IRE to the same mFOLFIRINOX route after a fixed no-progression gate. Contrary evidence. CROSSFIRE stopped for futility with IRE survival numerically below radiotherapy, HR 1.39. The registry survival comparison began after median diagnosis-to-enrollment times of eight versus four months and mostly uses reconstructed controls. Procedure-related deaths were 2/87, or 2.3%, in the initial registry report. Assumptions. Randomization occurred, assignment records and dated plans survive, consent permits independent analysis, and vital status can be recovered without arm-dependent loss. Prediction. Within three months, independent diligence can identify the actual screened and assigned counts, hash the source tables, recover the original and changed protocol and analysis plans, and locate at least 90% 24-month vital status with no more than five percentage points difference in missingness between arms. Within another nine months, a fixed all-randomized analysis can estimate survival and total burden. Mission adoption requires the trial's locked primary rule to pass, overall-survival HR at most 0.80 with its 95% confidence interval below 1.00, at least ten percentage points more alive at two years, procedure-related 90-day mortality at most 3%, no more than ten points excess grade 3 or worse harm, and no clinically important quality-of-life loss. These are mission gates, not claims about the sponsor's regulator-agreed test. Falsification experiment. Enrol zero new patients. Spend at most £250,000 from the existing data and statistics line only after rights to participant-level data, independent analysis and publication of negative findings are signed. Reconcile every assignment and compare treatment receipt, later care, survival, hospital time, pain, quality of life, bleeding, fistula, pancreatitis, sepsis, thrombosis, heart rhythm and death. Stop adoption if records cannot be reconciled, the analysis plan is missing, vital status is under 90%, missingness differs by more than five points, the locked primary endpoint fails or any mission gate fails. If the randomized record is too small or irretrievable, publish that failure and use the completed LAP-PIE feasibility result before pricing a replacement trial. Major safety risks are bleeding, vessel injury, pancreatic leak or inflammation, infection, thrombosis, dangerous heart rhythm, anaesthetic harm, biased missing records and false reassurance from a selected registry.", "sourceIds": [ "timmer-2024-crossfire", "clinicaltrials-nct03899636-direct-rct", "clinicaltrials-nct03899649-direct-registry", "martin-2024-direct-registry-safety", "martin-2026-direct-registry-survival-preprint", "rai-2022-lappie-protocol", "isrctn-14986389-lappie", "nice-2017-ire-pancreatic" ], "links": [ { "target": "experiments", "relation": "part-of" }, { "target": "unknown-ire-added-survival-after-modern-chemotherapy", "relation": "tests" }, { "target": "failure-direct-selection-and-randomized-reporting-gap", "relation": "repairs" }, { "target": "trial-pdac-irreversible-electroporation-action-map", "relation": "implemented-by" }, { "target": "programme-500m-integrated-mission", "relation": "constrained-by" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "moderate", "randomised": "strong", "independentReplication": "weak", "sampleSize": "Existing phase 3 target n=528 estimated; zero new patients; independent recovery capped at £250,000", "effect": "Existing assignments can decide adoption or establish the need for a replacement trial", "limits": [ "Actual randomized count unknown", "Thresholds are mission choices", "Original analysis plan is unavailable", "Open-label care can affect non-survival measures", "A negative overall result may hide but cannot prove a subgroup" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The selected local-dominant state and physical ablation create a coherent but unproved benefit model." }, "animalEvidence": { "rating": "moderate", "reason": "Animal ablation supports safety planning, not the clinical effect threshold." }, "inVitroEvidence": { "rating": "strong", "reason": "Cell killing is known; the proposed experiment tests added patient value." }, "reproducibility": { "rating": "weak", "reason": "The decision still depends on one sponsor-controlled phase 3 record." }, "commercialConflicts": { "rating": "high", "reason": "Signed independent access and negative-result rights are required because the sponsor owns NanoKnife." }, "recency": { "rating": "current", "reason": "The gate responds to the phase 3 reporting gap visible in September 2026." } } }, "layout": { "x": 574, "y": 498 }, "updated": "2026-09-15" }, { "id": "hypothesis-dynamic-cyst-interception-rule", "title": "Hypothesis: three locked risk models can find the IPMN interception window", "shortTitle": "Dynamic cyst interception rule", "type": "hypothesis", "status": "hypothesis", "scope": [ "pdac" ], "summary": "Separate models for current advanced disease, future cyst progression and PDAC elsewhere may support safer action than a single cyst score.", "content": "Rationale: one combined label hides three decisions, so build three locked models. Model A asks whether HGD or invasive cancer is already present and may use MRI/EUS, cytology and a molecular classifier. Model B asks whether a currently low-grade IPMN will reach HGD or invasion within three years and must use longitudinal change, not pathology available only after surgery. Model C asks whether PDAC will arise elsewhere and must use whole-pancreas and host risk rather than cyst features alone. Prediction: the three-model rule will improve net benefit over Kyoto-based care at the same missed-invasive-cancer limit. Supporting evidence: large prospective cohorts show separate within-cyst and elsewhere-in-pancreas routes, while multiple worrisome features enrich three-year risk. Evidence against: current molecular evidence detects disease already present, long-term cohorts conflict on when surveillance can stop, and no marker-guided action has reduced death. Assumptions: HGD offers a treatable window, model inputs are measurable before surgery, event rates transport across systems, and surgery or interception can be delivered with acceptable harm. Study: enroll 6,000 consecutive, surgery-fit patients across at least 20 systems before specialist sampling; collect fixed MRI, clinical data and blood, with EUS or cyst fluid only under a prespecified safety rule. Freeze all three models after an internal training third, validate unchanged in the remaining systems, follow every negative for at least five years and adjudicate HGD, invasive cancer, operation harm, interval cancer, pancreatic function and death. Then randomize the action, not merely the assay: model-guided surgery or low-harm interception versus current care. Initial gates: at least 80% of detected advanced lesions are HGD or stage I rather than later invasive disease; fewer than one major procedure complication per additional HGD or stage-I cancer found; at least 25% of positive operations contain HGD or invasive cancer; Model C captures separate PDAC without more than 20 false-positive invasive work-ups per treatable cancer. Falsification: stop if external calibration shifts by more than five percentage points, sensitivity depends on post-resection information, interval cancers exceed the locked limit, or a positive assay does not lead to a safe action. The 6,000 target is a planning number: at a 2% five-year event rate it yields about 120 total events before loss and competing death, enough for bounded validation but not a mortality claim. Estimated cost: £28m–£40m for the five-year cohort and assay validation, followed by a separately gated action trial. Major safety risks include EUS pancreatitis or bleeding, anxiety, false-positive operations, endocrine and exocrine pancreatic loss, and delayed invasive cancer after a false negative.", "contentSections": [ "Supporting evidence: two cancer routes repeat in prospective and health-system cohorts, and risk rises when several worrisome features accumulate.", "Evidence against: no current classifier forecasts future progression in a consecutive low-risk cohort and no randomized marker-guided action has shown survival benefit." ], "sourceIds": [ "ohtsuka-2024-bd-ipmn-surveillance", "mirzaian-2026-low-risk-cysts", "chhoda-2026-in-cyst", "mayo-2026-in-cyst-release", "singhi-2026-pancreaseq-gc", "pea-2026-post-surveillance-resection", "henn-2023-cyst-surgery-registry", "clinicaltrials-nct04239573-ea2185" ], "links": [ { "target": "experiments", "relation": "part-of" }, { "target": "unknown-dangerous-precursor-selection", "relation": "tests" }, { "target": "claim-ipmn-has-two-cancer-routes", "relation": "depends-on" }, { "target": "claim-cyst-classifier-current-disease-not-future-risk", "relation": "depends-on" }, { "target": "claim-ipmn-surgery-timing-tradeoff", "relation": "depends-on" }, { "target": "programme-500m-integrated-mission", "relation": "constrains" } ], "evidence": { "humanProspective": "moderate for risk estimates; none for the proposed action rule", "humanRetrospective": "strong for failure modes and surgical timing", "randomised": "none for this rule", "independentReplication": "none", "sampleSize": "Proposed n=6,000 across at least 20 systems; about 120 events expected at 2% before attrition", "effect": "Unobserved; designed to improve net benefit while holding the missed-invasive-cancer limit fixed", "limits": [ "Five-year follow-up is slow", "Competing death is material", "EUS-FNA is selected and invasive", "HGD adjudication usually requires resection", "Mortality requires a later and larger trial", "The 2% planning rate may not transport" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Current advanced disease, later cyst progression and separate PDAC are biologically and clinically different states." }, "animalEvidence": { "rating": "weak", "reason": "Models can test lesion biology but cannot establish human surveillance or surgery thresholds." }, "inVitroEvidence": { "rating": "moderate", "reason": "Molecular assays can support Model A, while Models B and C need longitudinal human outcomes." }, "reproducibility": { "rating": "none", "reason": "The three-model rule has not been tested." }, "commercialConflicts": { "rating": "high", "reason": "Assay owners may benefit from broader testing, so independent analysis, fixed data rights and publication of negative results are mandatory." }, "recency": { "rating": "current", "reason": "The design responds to 2024–2026 prospective, molecular and trial evidence." } } }, "layout": { "x": 71, "y": 91 }, "updated": "2026-09-14" }, { "id": "hypothesis-electrochemotherapy-randomized-data-gate", "title": "Recover the 90-person electrochemotherapy trial before any new device study", "shortTitle": "ECT randomized-data gate", "type": "hypothesis", "status": "testing", "scope": [ "pdac", "locally-advanced", "trial-recovery", "drug-delivery" ], "summary": "A short independent recovery can determine whether the transitioned phase 2b record contains a useful response, survival and harm answer; recruitment and equipment spending remain zero.", "content": "Biological rationale. Reversible electric pulses can increase tumour-cell exposure to bleomycin while sparing large structural proteins and may improve local control. Supporting evidence. Twenty-five selected human cases show feasible open delivery and scan changes; the 90-person trial fixes systemic treatment and assigns ECT. Contrary evidence. Eleven of 25 had died before six-month imaging, electrode comparisons were chosen and non-significant, PanECT has no control and the randomized trial has no public result. Assumptions. The EU CT transition preserved its assignment key, dated plan, images, device logs, systemic-treatment records and outcomes; IGEA and sites will grant independent access and publication rights. Prediction. Within six months, a custody audit will reconcile at least 80 assignments or prove that recruitment failed, account for at least 95% of assigned people, recover at least 90% of pulse logs and images and reproduce the fixed primary analysis. Falsification experiment. Enroll zero patients, buy no equipment and spend at most £250,000 from existing statistics and device-validation lines. Before analysis, freeze all-assigned flow, imaging rule, death handling and treatment-effect model. Report response with death counted as failure, overall and quality-adjusted survival, chemotherapy delivery and every operation and bleomycin harm. Stop if fewer than 80 assignments exist, vital status is below 95%, device coverage is below 90%, independent image agreement is below 80%, severe harm rises five points, chemotherapy delivery falls ten points or response does not translate into either patient-valued time or a credible survival signal. Do not combine cut-surface PanECT, calcium electroporation or irreversible electroporation into this estimate. A new survival trial receives £0 now and could be priced only after a passing audit and equal portfolio displacement. Major safety risks are bleeding, fistula, arrhythmia, pancreatitis, vessel or bowel injury, bleomycin lung toxicity, treatment delay and false benefit from conditioning imaging on survival.", "sourceIds": [ "izzo-2021-pancreatic-ect-phase12", "izzo-2021-laparoscopic-ect-protocol", "eudract-2018-003925-27-ect", "clinicaltrials-nct04281290-panect-results", "igea-2026-cliniporator-vitae" ], "links": [ { "target": "experiments", "relation": "part-of" }, { "target": "unknown-electrochemotherapy-added-benefit-and-delivery", "relation": "tests" }, { "target": "failure-electrochemotherapy-endpoint-and-trial-drift", "relation": "repairs" }, { "target": "trial-pdac-electrochemotherapy-action-map", "relation": "implemented-by" }, { "target": "company-igea", "relation": "depends-on" }, { "target": "programme-5m-decision-core", "relation": "constrained-by" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "not-applicable", "randomised": "weak — existing trial not reported", "independentReplication": "weak", "sampleSize": "Zero new participants; recover at least 80 of 90 planned assignments; analysis capped at £250,000", "effect": "Convert a hidden randomized record into a go or stop decision", "limits": [ "Actual enrollment is unknown", "The trial may still be active", "Survival is secondary", "Device files may be incomplete", "Independent rights are not negotiated", "Thresholds and cap are mission choices", "Response cannot substitute for survival" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The gate links a known delivery mechanism to measured coverage and patient outcomes." }, "animalEvidence": { "rating": "not-applicable", "reason": "The missing evidence is already in a human randomized programme." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "No further cell experiment is needed before data recovery." }, "reproducibility": { "rating": "weak", "reason": "Independent analysis is required because sponsor-linked studies dominate." }, "commercialConflicts": { "rating": "high", "reason": "Raw device data and publication rights must be secured from involved owners." }, "recency": { "rating": "current", "reason": "The gate addresses the current transitioned trial and March 2026 PanECT reporting." } } }, "layout": { "x": 754, "y": 678 }, "updated": "2026-09-15" }, { "id": "hypothesis-elraglusib-survival-confirmation-gate", "title": "Hypothesis: elraglusib adds repeatable useful survival to first-line chemotherapy", "shortTitle": "Elraglusib confirmation", "type": "hypothesis", "status": "hypothesis", "scope": [ "pdac" ], "summary": "The phase 2 survival gain may be real despite unchanged PFS; a full-data gate and blinded all-randomized phase 3 can falsify it.", "content": "Biological rationale. GSK-3β is a shared control enzyme across tumour survival, state change and immune circuits. Blocking it may alter lethal disease processes that routine scan measurements miss. Supporting evidence. In 269 relevant randomized people, elraglusib plus gemcitabine/nab-paclitaxel improved ITT median OS from 7.2 to 8.9 months, HR 0.68, and one-year survival from 22.3% to 42.1%. Contrary evidence. PFS HR was 0.90, response and disease control were not significant, the treated-population result excluded 20.4% of controls versus 9.4% of the treatment arm, the primary analysis changed after the trial began, severe toxicity rose by 10.2 points and a small second study did not exceed historical standards. Assumptions. Survival follow-up is complete enough, the all-randomized effect is not created by missing outcomes or later treatment, a matched placebo can preserve useful masking, the infusion product can be manufactured consistently and any benefit is not tied to a post-outcome subgroup. Prediction. Independent recovery preserves ITT HR at or below 0.75 with its 95% interval below 1.0; a phase 3 then produces OS HR at or below 0.80, at least a ten-point one-year survival gain and positive quality-adjusted survival without losing more than ten points of chemotherapy dose intensity. Falsification experiment. First rerun the existing data with 99% survival accounting and missingness tipping points. Stop before a new trial if plausible missing outcomes make HR exceed 0.85 or the endpoint change followed unblinded comparative review. If that gate passes, randomize about 600-650 people to blinded elraglusib or placebo beside identical chemotherapy. Falsify benefit if OS HR exceeds 0.80, the interval includes 1.0, quality-adjusted survival is not positive, severe harm rises above ten points without compensating benefit, or the effect fails outside a subgroup discovered after outcomes. Estimated cost. Stage 0 uses existing mission statistics capacity and adds no separate budget. A global phase 3 is not priced until a manufacturing and majority-funding partner is secured; mission capital stays £0 now and any later share must replace lower-value work. Estimated time. Stage 0 takes six to nine months after data access; phase 3 needs roughly four to six years. Major safety risks are neutropenia, infection, fatigue, transient altered colour or contrast vision, treatment unmasking, more clinic time, chemotherapy dose loss and a false survival conclusion caused by unequal follow-up.", "sourceIds": [ "mahalingam-2026-elraglusib-randomized", "clinicaltrials-nct03678883-elraglusib", "pathak-2026-elraglusib-folfirinox", "actuate-2026-q2-10q" ], "links": [ { "target": "experiments", "relation": "part-of" }, { "target": "claim-elraglusib-randomized-survival-signal", "relation": "supported-by" }, { "target": "failure-elraglusib-analysis-and-survival-discordance", "relation": "constrained-by" }, { "target": "unknown-elraglusib-survival-replication", "relation": "tests" }, { "target": "trial-elraglusib-action-map", "relation": "gated-by" }, { "target": "company-actuate-therapeutics", "relation": "requires-access-from" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "moderate", "independentReplication": "none", "sampleSize": "Reanalysis n=269; conditional confirmation planning case n=600-650", "effect": "Generated internal-robustness and external-survival gates from a phase 2 ITT HR of 0.68", "limits": [ "Thresholds are mission decisions", "Phase 3 cost is unpriced", "Drug supply and partner absent", "OS mechanism unresolved", "Masking may fail" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "The target touches several relevant circuits, but a causal route to survival is not established." }, "animalEvidence": { "rating": "moderate", "reason": "Models support testing but do not settle the survival-PFS mismatch." }, "inVitroEvidence": { "rating": "moderate", "reason": "Laboratory target effects are real but are not a surrogate for useful survival." }, "reproducibility": { "rating": "none", "reason": "The proposed central effect awaits independent randomized confirmation." }, "commercialConflicts": { "rating": "high", "reason": "The test requires a company-controlled drug and existing sponsor-controlled data." }, "recency": { "rating": "current", "reason": "The hypothesis uses 2026 clinical and company records." } } }, "layout": { "x": 206, "y": 124 }, "updated": "2026-09-15" }, { "id": "hypothesis-eus-rfa-existing-study-gate", "title": "Audit existing endoscopic ablation studies before funding another probe trial", "shortTitle": "EUS-RFA evidence gate", "type": "hypothesis", "status": "testing", "scope": [ "pdac", "locally-advanced", "trial-recovery", "local-treatment" ], "summary": "A small independent audit can test whether current studies contain a consistent delivery and outcome signal; new enrollment and equipment spending remain zero.", "content": "Biological rationale. Endoscopic ultrasound gives a short access path into pancreatic tumours, and controlled radiofrequency heat can kill local tissue. Supporting evidence. HybridTherm achieved 94.1% first-session feasibility, visible ablation and no severe procedure event. Contrary evidence. Its randomized comparison showed no survival or resection gain, stopped early, excluded three intervention assignments and has a 66-versus-40 enrollment conflict. PANCARDINAL-1 remains single-arm. Assumptions. Sites and manufacturers will provide dated consent, assignment, procedure, imaging, chemotherapy, surgery, quality-of-life and survival records with negative-result publication rights. Prediction. Within nine months, an independent audit will reconcile 100% of HybridTherm's registry count, recover at least 95% vital status across completed EUS-RFA cohorts and show whether delivered ablation volume relates to local control without reducing chemotherapy. Falsification experiment. Enroll zero new people, buy no probe and spend at most £300,000. Freeze cohort definitions and count every assigned patient in their original arm. Map intended and achieved heat volume against viable tumour, vessels and ducts; report local and distant failure separately; use overall and quality-adjusted survival as the decision outputs. Stop the efficacy route if the 26-person registry mismatch cannot be reconciled, vital status is below 95%, procedure logs are below 90%, chemotherapy delivery falls by ten points, severe procedure harm exceeds five percent or no consistent outcome signal remains after accounting for disease burden. Even a passing audit does not justify a new trial automatically: it must first beat the opportunity cost of already-randomized systemic and early-detection programmes. Major risks are pancreatitis, bleeding, infection, duct or bowel injury, thrombosis, biliary obstruction, treatment delay and false benefit from selecting people whose disease remains local.", "sourceIds": [ "testoni-2021-hybridtherm-rct", "clinicaltrials-nct02336672-hybridtherm", "clinicaltrials-nct04990609-pancardinal1" ], "links": [ { "target": "experiments", "relation": "part-of" }, { "target": "unknown-eus-rfa-added-benefit-after-hybridtherm", "relation": "tests" }, { "target": "failure-hybridtherm-product-accrual-and-custody", "relation": "repairs" }, { "target": "trial-pdac-eus-rfa-action-map", "relation": "implemented-by" }, { "target": "company-erbe-elektromedizin", "relation": "depends-on" }, { "target": "programme-5m-decision-core", "relation": "constrained-by" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "moderate", "independentReplication": "weak", "sampleSize": "Zero new participants; reconcile the HybridTherm record and completed EUS-RFA cohorts; cap £300,000", "effect": "Turn scattered device records into a stop or narrowly defined comparison decision", "limits": [ "Raw access is unconfirmed", "Probe geometries differ", "Cohorts may be too small for outcome modeling", "HybridTherm no longer exists", "The thresholds and cap are mission choices", "A positive association would not prove causality" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The audit connects mapped heat delivery to measured tissue and patient outcomes." }, "animalEvidence": { "rating": "not-applicable", "reason": "The next missing evidence already exists in human records." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "No further bench work is needed before resolving the human evidence chain." }, "reproducibility": { "rating": "weak", "reason": "Independent record reconciliation is the experiment." }, "commercialConflicts": { "rating": "high", "reason": "Independent custody and publication rights are required across device owners." }, "recency": { "rating": "current", "reason": "The gate uses completed and recruiting records checked in September 2026." } } }, "layout": { "x": 790, "y": 714 }, "updated": "2026-09-15" }, { "id": "hypothesis-fapi-theranostic-pdac", "title": "Hypothesis: absorbed FAP-radiation dose, not scan brightness, predicts PDAC response", "shortTitle": "FAP dose-response gate", "type": "hypothesis", "status": "hypothesis", "scope": [ "pdac", "pnet", "transfer-unproven" ], "summary": "Use existing LuMIERE data to test whether time-integrated absorbed dose predicts lesion control; do not commission another uncontrolled dose-escalation study.", "content": "Biological rationale. PDAC contains many FAP-positive fibroblasts around malignant cells. A FAP-binding carrier can bring beta radiation into that neighbourhood, but the useful variable is energy absorbed by each lesion, not how bright one scan appears. Supporting evidence: human delivery is no longer hypothetical. The 2022 report treated five pancreatic cancers and LuMIERE treated nine, set a 9.25-GBq phase 2 dose and continues disease-specific expansion. Contrary evidence: no pancreatic partial response was reported, tumour-lesion dose remains unpublished and broad fibroblast removal has harmed pancreatic models. Assumptions: lesion identifiers can be joined across imaging and response, dose calculations transfer across centres, co-treatment can be measured, and normal-organ exposure leaves a useful tumour range. The remaining hypothesis is that time-integrated lesion dose predicts control better than baseline SUV, while normal-organ dose limits usable exposure. Prediction: within imaging-selected PDAC, higher absorbed tumour Gy will associate with later lesion growth after adjustment for lesion size, site, isotope activity, treatment line and concurrent chemotherapy; baseline SUV alone will have weaker out-of-sample prediction. Test without new exposure. Obtain every LuMIERE pancreatic scan, serial SPECT measurement, lesion identifier, administered activity, standard-treatment dose, toxicity and response, including failed screens and early withdrawals. Freeze the calculation in one centre and test it in the other centres. Require at least 80% of target lesions to have usable linked dose and response, cross-centre calibration error no greater than 10%, and an externally tested dose-response area under the receiver-operating curve of at least 0.70 with confidence interval above 0.55. Falsify the routing hypothesis if these gates fail, if baseline SUV performs equally well, if lesion dose has no ordered relation with control, or if safe organ limits prevent the predicted tumour exposure. Clinical gate. A valid lesion model does not prove patient benefit. Let ongoing NCT04939610 establish PDAC response and combination safety. A later randomized comparison of radioligand plus the same standard treatment versus standard treatment alone can be designed only if the dose-response reproduces, the phase 2 PDAC signal crosses its declared bound and standard-treatment dose intensity is preserved. Overall and quality-adjusted survival decide; response alone does not. Current cost is £0 in new treatment capital: data recovery and analysis sit inside assigned imaging and data work. Any later trial is unfunded and must remove equal money elsewhere before entering the £500m ceiling. Major risks are kidney or marrow injury, bowel or healing-tissue irradiation, uneven FAP across lesions, treatment of inflammation and a false causal claim from selected stable disease.", "sourceIds": [ "pang-2022-fapi-pet", "lee-2024-prospective-fapi-resectable-pdac", "evans-2026-fapi-feasibility", "baum-2022-fap2286-firsthuman", "mcconathy-2026-lumiere-phase1", "clinicaltrials-nct04939610-lumiere", "baum-2026-3bp3940", "clinicaltrials-nct05432193-frontier", "clinicaltrials-nct07229768-ctr-fapi", "ozdemir-2014-fibroblast-depletion", "rhim-2014-stroma-restrains" ], "links": [ { "target": "experiments", "relation": "part-of" }, { "target": "technology-fapi-pet-staging-not-screening", "relation": "extends" }, { "target": "failure-bulk-stroma-removal", "relation": "constrained-by" }, { "target": "treatment", "relation": "tests" }, { "target": "claim-fap-radioligand-feasible-pdac-efficacy-unproven", "relation": "extends" }, { "target": "failure-fapi-scan-dose-benefit-collapse", "relation": "repairs" }, { "target": "unknown-pdac-fap-radioligand-treatment-effect", "relation": "tests" }, { "target": "trial-pdac-fap-radioligand-action-map", "relation": "gated-by" } ], "evidence": { "humanProspective": "moderate — multicentre human dose escalation exists", "humanRetrospective": "moderate", "randomised": "none", "independentReplication": "moderate for delivery, none for PDAC dose-response or benefit", "sampleSize": "Existing LuMIERE phase 1: 27 treated, 9 pancreatic; live total plan n=222", "effect": "Analysis gate: ≥80% linked lesions, ≤10% cross-centre calibration error and external dose-response AUC ≥0.70 with lower confidence bound >0.55", "limits": [ "Generated dose-response hypothesis", "LuMIERE tumour dosimetry is not public", "Imaging performance does not prove treatment benefit", "Agent and isotope properties differ", "FAP is present in repair and inflammation", "No randomized survival evidence" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "Human delivery is shown, but FAP often sits on nearby support cells rather than malignant cells." }, "animalEvidence": { "rating": "moderate", "reason": "Pancreatic models support dose delivery and cell killing while broad fibroblast removal gives a safety warning." }, "inVitroEvidence": { "rating": "moderate", "reason": "Carrier binding and radiation effects support the mechanism, not the clinical route." }, "reproducibility": { "rating": "weak", "reason": "Delivery repeats, while lesion dose-response and PDAC patient benefit have not reproduced." }, "commercialConflicts": { "rating": "high", "reason": "The decisive current dataset is sponsor-controlled and the platform has strong product-development incentives." }, "recency": { "rating": "current", "reason": "Updated with June 2026 phase 1 results and July 2026 trial status." } } }, "layout": { "x": 56, "y": 127 }, "updated": "2026-09-15" }, { "id": "hypothesis-gallbladder-two-stage-interception-rule", "title": "Hypothesis: country-specific risk followed by lesion confirmation can make gallbladder prevention safe", "shortTitle": "Gallbladder two-stage rule", "type": "hypothesis", "status": "hypothesis", "scope": [ "gallbladder", "hepatobiliary" ], "summary": "A baseline country and patient-risk model should decide who receives a second test; only the second-stage result should influence optional surgery.", "content": "Rationale: one universal threshold cannot cover common gallstones, ordinary small polyps, PSC, endemic South American or Indian risk and low-incidence European surgery. Stage A estimates future HGD or invasive-cancer risk from country and region, age, sex, ancestry where lawful and useful, gallstone and inflammation history, fixed ultrasound features and competing health risk. It must exclude consequences of existing disease such as open surgery from a prevention score. Stage B is applied only above a locked Stage-A threshold and combines a repeat expert ultrasound or CT with independently measured blood candidates; GBCseeker, GAIA-MIL and miR-4533 are candidates, not the final model. PSC and congenital biliary disease keep separate action rules because their evidence and operation risks differ.\n\nPrediction: compared with current local care at the same missed-invasive-cancer limit, the two-stage rule will reduce low-yield follow-up and operations while increasing the proportion of optional operations containing HGD or stage-I cancer. Study: spend no more than £5m first to align fields, obtain code and data and run locked silent validation inside EULAT, Chile BiLS and P-iGBC, adding high-incidence North Indian sites. If that stage achieves at least 80% complete outcome linkage and calibration error below five percentage points within each country, release £15m–£20m for a 40,000-person prospective validation linked to cancer registries and all surgery outcomes. At 0.2% target prevalence, 40,000 people yield about 80 target events before loss; at 1% they yield about 400. That can validate broad calibration and harm workload, not mortality or every subgroup. Do not pool countries until local intercepts, operation setting and cancer incidence are shown.\n\nAction gate: for each country estimate net benefit as expected cancer deaths prevented minus operative deaths and quality-weighted serious harm. The simple mortality condition is p × b > m, where p is preventable lethal-disease risk, b is the fraction prevented and m is local operative mortality. Require the lower confidence bound for net benefit above zero, at least five expected prevented cancer deaths per operative death, fewer than ten false-positive invasive work-ups per HGD or stage-I cancer, and fewer follow-up scans than current care. Then randomize or use a stepped-wedge rollout of model-guided optional cholecystectomy versus current care, with HGD/invasive cancer, interval cancer, bile-duct injury, 90-day complications, quality of life, cost and cancer death. Falsification: stop if external calibration misses by more than five points, sub-10-mm sensitivity remains below 80%, most positives arise from a single centre, the rule depends on information available only during or after surgery, or serious harm erases expected benefit. Major safety risks are unnecessary surgery, bile-duct injury, operative death, false reassurance, anxiety, and widening access gaps. The test can fail even if its AUC looks good.", "contentSections": [ "Supporting evidence: removal of the organ can prevent future gallbladder cancer; high-incidence cohorts, before-diagnosis serum and externally tested imaging models provide ingredients for a staged rule.", "Evidence against: present polyp rules have very low HGD yield, diagnostic models were trained after lesions were visible, a 100,000-person future-risk cohort did not start, and no risk-guided surgery trial has reduced cancer mortality.", "Assumptions: active cohorts can align core fields, cancer registries can link operated and unoperated outcomes, elective operation harm can be estimated by country, and at least one second-stage signal adds information beyond fixed imaging. Estimated time: two years for field alignment and silent testing, three to five years for prospective validation, and at least a further five years for a mortality-capable action trial." ], "sourceIds": [ "gonzalez-2025-cholecystectomy-chile", "lorenzo-bermejo-2025-gallbladder-specimens", "van-dijk-2025-polyp-study", "clinicaltrials-nct06192719-eulat-gbc", "koshiol-2026-chile-bils", "clinicaltrials-nct06531408-pigbc", "blandino-2025-gallbladder-mirna", "yang-2025-gbcseeker", "gupta-2026-gaia-mil", "wong-2025-ambrose-cholecystectomy" ], "links": [ { "target": "experiments", "relation": "part-of" }, { "target": "unknown-gallbladder-prevention-selector", "relation": "tests" }, { "target": "trial-gallbladder-prevention-detection-action-map", "relation": "depends-on" }, { "target": "claim-gallbladder-polyp-guidelines-still-overtreat", "relation": "learns-from" }, { "target": "claim-gallbladder-diagnostic-models-existing-lesions-only", "relation": "learns-from" }, { "target": "claim-gallbladder-prevention-must-beat-surgery-harm", "relation": "depends-on" }, { "target": "programme-500m-integrated-mission", "relation": "could-enter" } ], "evidence": { "humanProspective": "moderate for ingredients; none for the combined rule", "humanRetrospective": "strong for risk candidates and failure modes", "randomised": "none", "independentReplication": "none for the proposed action rule", "sampleSize": "Proposed first stage uses active cohorts; conditional prospective validation n=40,000 across high- and low-incidence countries", "effect": "Unobserved; designed to increase HGD or stage-I yield while lowering low-value scans and operations at a fixed missed-cancer limit", "limits": [ "The 40,000 target is a planning number", "At low incidence only about 80 target events may occur", "Long latency limits mortality inference", "Candidate blood and imaging markers may not transport", "Local elective surgery harm must be estimated", "PSC and congenital-risk pathways need separate rules" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The organ can be removed before invasion, but only if a dangerous state is selected early enough." }, "animalEvidence": { "rating": "weak", "reason": "Models can study progression but cannot set human surgical thresholds." }, "inVitroEvidence": { "rating": "moderate", "reason": "Blood and tissue markers offer measurable candidates without proven future-risk accuracy." }, "reproducibility": { "rating": "none", "reason": "No independent group has tested this two-stage country-specific action rule." }, "commercialConflicts": { "rating": "high", "reason": "Sequencing and imaging tools could gain from broad testing, so code, data and negative-result rights must be public-interest terms." }, "recency": { "rating": "current", "reason": "The design responds to 2025-2026 cohort, AI, biomarker, surgery and registry evidence." } } }, "layout": { "x": 74, "y": 96 }, "updated": "2026-09-14" }, { "id": "hypothesis-gannon-independent-safety-gate", "title": "Add an independent visibility and coverage gate to GANNON", "shortTitle": "GANNON safety gate", "type": "hypothesis", "status": "testing", "scope": [ "pdac", "locally-advanced", "oligometastatic", "medical-device" ], "summary": "The active 50-person feasibility study can answer a valuable engineering question if every screened patient and every loss of visibility is independently captured; it cannot yet answer survival.", "content": "Biological rationale. Histotripsy can destroy tissue mechanically without placing a probe through the pancreas and may spare stronger structures. Supporting evidence. Six of nine pig targets were reached, treatment zones matched the planned scale and vessels sometimes crossed the zone intact. Contrary evidence. The bubble cloud was visible in only five of nine pigs; three treatments struck bowel or nearby tissue and two animals were killed early. There is no human pancreas result. Assumptions. The GANNON sponsor will export continuous imaging, pulse, motion, treatment-plan, screen-failure and outcome data and allow an independent safety committee to pause enrollment. Prediction. In the first ten treated people, at least nine will maintain target and bubble-cloud visibility through at least 95% of delivered pulses, achieved treatment volume will cover at least 90% of the safe planned volume, and there will be no bowel perforation, grade 3 or worse pancreatitis or treatment-related death by 30 days. Falsification experiment. Fund no device and no added enrollment. Offer at most £400,000 for independent capture, review and public reporting inside the existing trial. Define the safe operating region before the first supported case. Stop energy automatically on lost visibility, motion outside bounds or bubble-cloud excursion. Pause support after any bowel injury, any treatment-related death, two grade 3 pancreas events, target visibility below 90% in two cases, or plan-to-treatment overlap below 80%. Publish the screened denominator and reasons for exclusion. Survival, quality of life and chemotherapy delivery remain descriptive until a later same-care randomized comparison. Major risks are bowel perforation, sepsis, pancreatitis, duct damage, thrombosis, skin injury, incomplete tumour coverage, treatment delay and confusing FDA liver authorization with pancreatic cancer benefit.", "sourceIds": [ "gannon-2026-pancreas-histotripsy-safety", "clinicaltrials-nct06282809-gannon", "fda-den220087-edison" ], "links": [ { "target": "experiments", "relation": "part-of" }, { "target": "unknown-histotripsy-human-coverage-and-benefit", "relation": "tests" }, { "target": "failure-histotripsy-acoustic-window-and-translation", "relation": "repairs" }, { "target": "trial-pdac-histotripsy-action-map", "relation": "implemented-by" }, { "target": "company-histosonics", "relation": "depends-on" }, { "target": "programme-5m-decision-core", "relation": "constrained-by" } ], "evidence": { "humanProspective": "none — embedded gate proposed for active trial", "humanRetrospective": "none", "randomised": "none", "independentReplication": "weak", "sampleSize": "First ten existing GANNON participants; zero added enrollment; cap £400,000", "effect": "Test whether safe closed-loop pancreatic targeting works before any efficacy claim", "limits": [ "Sponsor agreement is unconfirmed", "Ten people cannot establish uncommon harm", "Coverage thresholds need technical validation", "Safe selected cases do not represent all PDAC", "No causal survival estimate", "The cap and stop rules are mission choices" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The test measures whether the mechanical effect stays inside a visible target." }, "animalEvidence": { "rating": "moderate", "reason": "The gate directly addresses the observed pig failure mode." }, "inVitroEvidence": { "rating": "moderate", "reason": "Phantoms can validate tracking thresholds but not living bowel motion." }, "reproducibility": { "rating": "weak", "reason": "Independent monitoring and open data are required to improve it." }, "commercialConflicts": { "rating": "high", "reason": "Funding is conditional on independent pause authority and raw export from the company-sponsored study." }, "recency": { "rating": "current", "reason": "The proposal acts on the recruiting 2026 trial." } } }, "layout": { "x": 826, "y": 750 }, "updated": "2026-09-15" }, { "id": "hypothesis-hcc-liver-reserve-co-control", "title": "Hypothesis: co-treating liver injury and HCC will extend usable control more than tumour treatment alone", "shortTitle": "HCC liver co-control", "type": "hypothesis", "status": "hypothesis", "scope": [ "hcc", "hepatobiliary" ], "summary": "A protocol that actively preserves hepatic reserve alongside cancer treatment may prevent decompensation, keep later options open and improve survival.", "content": "Biological rationale and supporting evidence: HCC arises in an injured organ, and early-stage and systemic-treatment cohorts show large competing risks from liver dysfunction. The intervention is a timed control route: treat viral or alcohol cause where possible; manage portal hypertension, nutrition and sarcopenia; avoid preventable liver toxicity; measure ALBI and decompensation after every cancer treatment; and switch local or systemic treatment before reserve is irreversibly lost. Evidence against or missing: better liver function may simply mark less aggressive disease; several bundle components are already recommended; enforcing them may add no benefit; cancer progression may dominate too quickly; and delayed tumour treatment could harm patients. TRACER and PREMIUM test surveillance routes, not liver-reserve co-control during treatment, so this question does not duplicate them.\n\nAssumptions: enough decompensation is preventable, reserve can change on the timescale of cancer care, the bundle can be delivered without delaying tumour treatment, and its components do not reduce tumour control. First prediction: a £6m, three-year randomized readiness stage in 240 patients at 12 centres can achieve at least 85% delivery of required liver-care actions while losing no more than five percentage points in on-time cancer treatment or delivered dose intensity. With 120 people per arm, this stage is not powered for mortality; it estimates delivery, ALBI change, first decompensation, treatment delay and later-line access. Expansion prediction: if the readiness gates pass, a 24-centre trial in about 600 patients will reduce death or first hepatic decompensation by at least 25% over 18 months while keeping tumour response and dose intensity non-inferior. With roughly 380 joint events, HR 0.75 is a plausible 80%-power planning target, subject to the final clustering calculation.\n\nFalsification experiment: publish the route and analysis before enrolment; randomize within treatment and liver-risk strata; adjudicate tumour death, liver death and decompensation blind to arm; and collect treatment delays, dose intensity, later-line access and quality of life. Stop after the £6m stage for delivery below 85%, more than a five-point loss in on-time treatment or dose intensity, excess 90-day death or decompensation, or no measurable separation in ALBI or required action completion. Do not release the further £12m-£19m for the 600-person stage unless every readiness gate passes. In the expanded trial, stop for conditional power below 20%. A null joint endpoint with adequate delivery falsifies this implementation; tumour-control loss falsifies it even if ALBI improves. Total estimated cost is £18m-£25m and time is four to five years including follow-up. Major safety risks are delayed cancer treatment, hypotension or kidney injury during portal-pressure treatment, drug interactions, infection, harmful weight loss and centre differences that hide the result.", "sourceIds": [ "seif-el-dahan-2026-early-hcc-causes-death", "storandt-2024-atezo-bev-liver-function", "finn-2020-imbrave150" ], "links": [ { "target": "claim-hcc-tumour-and-liver-are-competing-failure-axes", "relation": "depends-on" }, { "target": "claim-hcc-repeat-treatment-sequence", "relation": "extends" }, { "target": "experiments", "relation": "tests" }, { "target": "research-frontier-ranking", "relation": "ranked-by" }, { "target": "programme-25m-prospective-observatory", "relation": "could-enter" }, { "target": "trial-hcc-action-map", "relation": "constrained-by" }, { "target": "programme-500m-integrated-mission", "relation": "could-enter" } ], "evidence": { "humanProspective": "none for the combined protocol", "humanRetrospective": "strong for the two-axis problem", "randomised": "none for the combined protocol", "independentReplication": "moderate for liver reserve as prognosis; none for the proposed intervention", "sampleSize": "Proposed about 600 patients at 24 centres; roughly 380 joint events for a planning target near HR 0.75", "effect": "Predicted at least 25% reduction in death or first decompensation with non-inferior tumour control", "limits": [ "Generated programme hypothesis", "Bundle components may be hard to attribute", "The 240-person stage is not powered for mortality", "The larger trial may be underpowered after centre correlation", "Usual care can improve during the trial", "Prognostic evidence does not prove modifiability" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Preventing avoidable liver decline can directly prevent decompensation and preserve access to later cancer treatment." }, "animalEvidence": { "rating": "weak", "reason": "Animal models cannot test delivery of a multi-part liver-care route across human HCC treatment." }, "inVitroEvidence": { "rating": "weak", "reason": "Cell systems omit organ reserve, portal hypertension and treatment sequencing." }, "reproducibility": { "rating": "weak", "reason": "The two-axis prognosis is repeated, but the proposed route has no randomized efficacy result." }, "commercialConflicts": { "rating": "moderate", "reason": "Most actions are clinical care, but medicines and procedures can carry company or provider interests." }, "recency": { "rating": "current", "reason": "The design uses cause-of-death and treatment-state evidence checked through 2026." } } }, "layout": { "x": 50, "y": 123 }, "updated": "2026-09-14" }, { "id": "hypothesis-heat-existing-randomization-recovery-gate", "title": "Recover the heat-only randomization before exposing another patient", "shortTitle": "Hyperthermia data-recovery gate", "type": "hypothesis", "status": "testing", "scope": [ "pdac", "locally-advanced", "regional-hyperthermia", "data-recovery" ], "summary": "A capped audit of HEATPAC and HEAT can answer whether any interpretable heat signal exists more cheaply and safely than another trial.", "content": "Biological rationale. Heating deep tissue to 40-43 degrees C may improve blood flow and oxygenation and may weaken cancer cells' repair of chemotherapy or radiation damage. Supporting evidence. HEATPAC used the correct component design: identical FOLFIRINOX and gemcitabine chemoradiotherapy in both groups, with random assignment changing only weekly regional heat. It also planned direct temperature measurement, which tests whether the physical treatment reached its target. Contrary evidence. HEAT missed disease-free survival, did not isolate heat and enrolled only 34.8% of plan. HEATPAC has no public result and an unknown registry status years after estimated completion. Assumptions. At least one site's allocation, consent, treatment, temperature, harm and outcome records survive; legal rights allow independent reconstruction; and missingness can be measured rather than hidden. Prediction. An independent audit will reconcile at least 95% of assigned participants and every random assignment, report all deaths and serious harms, and recover time-stamped temperature data for at least 80% of delivered heat sessions. It will produce an all-assigned estimate with uncertainty for one-year survival, progression-free survival, local failure and severe harm. Falsification experiment. Spend no more than £250,000 from the existing data and statistics budget and enroll zero new patients. Search investigator, sponsor, ethics, pharmacy, radiation and device archives. Hash source files, freeze a public analysis plan before opening outcomes, compare registry versions with the final protocol, and publish the full result whether positive, negative or incomplete. Stop the route if fewer than 80% of assignments or deaths can be reconstructed, allocation integrity fails, temperature delivery cannot be checked, severe heat-related harm is materially higher, or an apparent benefit disappears in the all-assigned analysis. Do not use temperature received as a causal dose comparison because survival and treatment tolerance can both cause higher exposure. A new randomized trial is considered only if existing records show a credible heat-specific signal, the design adds current care rather than replacing it, and the expected information is worth more than its patient and money cost. Major safety risks are deep-tissue hot spots, burns, pain, bleeding or perforation from temperature probes, infection, treatment delay and displacement of effective systemic treatment.", "sourceIds": [ "issels-2023-heat-rct", "clinicaltrials-nct01077427-heat", "eudract-2008-004802-14-heat", "datta-2017-heatpac-protocol", "clinicaltrials-nct02439593-heatpac" ], "links": [ { "target": "experiments", "relation": "part-of" }, { "target": "unknown-regional-hyperthermia-added-benefit-modern-care", "relation": "tests" }, { "target": "failure-heat-component-and-registry-drift", "relation": "repairs" }, { "target": "trial-pdac-regional-hyperthermia-action-map", "relation": "implemented-by" }, { "target": "programme-25m-prospective-observatory", "relation": "constrained-by" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "not-applicable", "randomised": "moderate", "independentReplication": "none", "sampleSize": "Existing HEAT n=117; HEATPAC actual enrollment unknown; zero new mission-funded participants; audit capped at £250,000", "effect": "Recover or close an unresolved randomized route before new exposure", "limits": [ "Data access is not secured", "Records may be lost", "Actual HEATPAC enrollment is unknown", "A small phase 2 trial cannot establish cure", "Old treatment backbones limit transfer", "Temperature received is affected by post-randomization behaviour", "Budget and thresholds are mission choices" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "There is a plausible treatment-sensitizing mechanism and a design capable of testing it." }, "animalEvidence": { "rating": "moderate", "reason": "Model evidence supports recovery of the human experiment, not a new efficacy assumption." }, "inVitroEvidence": { "rating": "moderate", "reason": "Laboratory sensitization is credible enough to preserve the question." }, "reproducibility": { "rating": "none", "reason": "No reported independent pancreatic randomization isolates heat." }, "commercialConflicts": { "rating": "low", "reason": "Independent analysis and public negative-result rights address academic ownership and reporting risk." }, "recency": { "rating": "mixed", "reason": "This is a recovery decision about older trials, not a claim that their regimens remain current." } } }, "layout": { "x": 634, "y": 558 }, "updated": "2026-09-15" }, { "id": "hypothesis-hifu-existing-randomized-gate", "title": "Recover the existing HIFU randomizations before exposing another patient", "shortTitle": "Existing HIFU data gate", "type": "hypothesis", "status": "testing", "scope": [ "pancreatic-all", "pdac", "trial-recovery", "patient-reported-outcomes" ], "summary": "The 40-person academic randomization and 90-person industry study can decide whether thermal HIFU deserves a larger pain or survival test; no new cohort is needed now.", "content": "Biological rationale. Focused sound can destroy a local tumour and nearby pain-carrying tissue without an incision. Supporting evidence. Prospective and retrospective human studies repeatedly report pain relief, including 84% in a 50-person prospective cohort. Contrary evidence. No auditable prospective randomized result is public; the strongest old survival contrast comes from a paper whose abstract and methods disagree about allocation; one academic trial has remained recruiting since 2017; one single-arm study is stale; and PULS stopped after an industrial-partner decision. Assumptions. HIFU-PC15 assignments and outcomes still exist; NCT05601323 can release every randomized participant after its 2027 completion; device logs can be exported; and pain, analgesic, chemotherapy and survival data can be aligned. Prediction. A twelve-month recovery and analysis package will account for at least 95% of HIFU-PC15 assignments and at least 99% of NCT05601323 assignments, recover delivered-energy data for at least 90% of treated sessions and report pain together with opioid dose, quality of life, chemotherapy dose and delay, local failure, distant failure, survival and harm. Falsification experiment. Enroll zero patients and commit £0 until either database locks. Then spend at most £250,000 from existing statistics and device-validation lines only after contracts guarantee independent analysis, raw data export and negative-result publication. Stop the treatment-development branch if either assignment denominator cannot be recovered, serious device harm exceeds control by five percentage points, chemotherapy delivery falls by ten points, no pain improvement remains after analgesic adjustment, or survival and quality-adjusted survival do not favour HIFU. A positive pain result may support a symptom trial even if survival is neutral; it must be compared with cheaper pain routes. A larger survival trial remains blocked and receives £0 now. Major safety risks are skin or bowel burn, vascular injury, pancreatitis, anaesthesia, treatment delay, false reassurance from local shrinkage and commercial suppression of a negative result.", "sourceIds": [ "li-2015-hifu-s1", "marinova-2018-hifu-prospective", "marinova-2024-hifu-pc-protocol", "drks-00012367-hifu-pc", "clinicaltrials-nct05262452-hifu-folfirinox", "clinicaltrials-nct05601323-suizenji", "clinicaltrials-nct06211933-puls-hifu", "sonire-2026-nedo-award" ], "links": [ { "target": "experiments", "relation": "part-of" }, { "target": "unknown-hifu-pain-survival-and-system-transfer", "relation": "tests" }, { "target": "failure-hifu-randomization-and-trial-drift", "relation": "repairs" }, { "target": "trial-pdac-hifu-action-map", "relation": "implemented-by" }, { "target": "programme-5m-decision-core", "relation": "constrained-by" }, { "target": "company-sonire-therapeutics", "relation": "depends-on" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "moderate", "randomised": "weak", "independentReplication": "weak", "sampleSize": "Zero new participants; recover target n=40 and target n=90 existing randomizations; analysis capped at £250,000 after lock", "effect": "Turn two existing randomizations into a go or stop decision for thermal HIFU", "limits": [ "HIFU-PC15 actual enrollment may be low", "NCT05601323 will not complete before 2027", "The two studies mix populations and chemotherapy", "Device logs may not share a common format", "Pain assessment is not blinded", "Thresholds and cost cap are mission choices", "No data-access agreement exists" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The gate preserves a direct local symptom mechanism while requiring patient outcomes." }, "animalEvidence": { "rating": "not-applicable", "reason": "Existing human randomizations are the needed evidence." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Laboratory tissue destruction cannot replace controlled pain and survival data." }, "reproducibility": { "rating": "weak", "reason": "The proposed joined analysis is needed because no auditable randomized result is public." }, "commercialConflicts": { "rating": "high", "reason": "Independent access and publication rights are release conditions for industry-controlled data." }, "recency": { "rating": "current", "reason": "The gate waits for the active 90-person study and responds to the September 2026 PULS closure." } } }, "layout": { "x": 724, "y": 648 }, "updated": "2026-09-15" }, { "id": "hypothesis-hipec-existing-trial-measurement-gate", "title": "Measure every eligibility loss in the live HIPEC study before another operation is funded", "shortTitle": "HIPEC live-study gate", "type": "hypothesis", "status": "testing", "scope": [ "pdac", "hipec", "peritoneal-metastases", "trial-design", "capital-allocation" ], "summary": "The recruiting 40-person study can expose feasibility and selection if its denominator starts before chemotherapy response and its data are held independently.", "content": "Biological rationale. Cytoreduction may remove visible peritoneal disease and HIPEC may kill exposed microscopic deposits. Supporting evidence. Selected prospective series report long survival, and a randomized adjuvant study reduced local recurrence. Contrary evidence. The adjuvant trial did not improve survival; the metastatic pilot had 44% grade 3-plus complications and one 30-day death; every metastatic study confounds selection, surgery and HIPEC. Assumptions. Mayo will add an independent registry of every person assessed for NCT04858009, beginning before the six-month response gate, and release de-identified records, procedure maps and analysis code. Prediction. At least 90% of screened candidates and 100% of enrolled participants can be traced through exclusion, progression, operation, heat and drug delivery, serious harm, systemic-treatment restart, recurrence and death. Falsification experiment. Spend at most £350,000 on measurement and independent analysis, £0 on HIPEC procedures, devices or extra participants. Recover the 42-person adjuvant randomization and attach a concurrent, eligibility-matched non-operative cohort to the active study for planning only. Stop if the screened denominator is below 90%, any enrolled participant disappears, more than 10% lack delivered drug-temperature-time records, more than 10% lack systemic-treatment dates, 90-day mortality exceeds 5%, grade 3-plus harm exceeds 35%, or median systemic-treatment restart exceeds eight weeks. Do not claim efficacy from the matched cohort. After the 40-person report, a randomized component test may compete for the existing selected-local-consolidation budget only if at least 70% of eligible assigned people can receive the package and projected quality-adjusted survival gain exceeds the operative harm. Randomize HIPEC after a fixed surgery decision so the operation is identical in both arms. Major risks are death, pancreatic leak, bleeding, infection, kidney injury, marrow injury, bowel injury, readmission and delayed systemic control.", "sourceIds": [ "padilla-valverde-2024-adjuvant-hipec-rct", "grotz-2022-crs-hipec-pdac-pilot", "clinicaltrials-nct04858009-hipec-pdac", "yan-2024-crs-hipec-pdac-retrospective" ], "links": [ { "target": "experiments", "relation": "part-of" }, { "target": "unknown-hipec-added-value-peritoneal-pdac", "relation": "tests" }, { "target": "failure-hipec-local-endpoint-and-package-selection", "relation": "repairs" }, { "target": "trial-pdac-hipec-action-map", "relation": "implemented-by" }, { "target": "hypothesis-selected-local-consolidation", "relation": "constrained-by" }, { "target": "programme-25m-prospective-observatory", "relation": "constrained-by" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "weak", "randomised": "weak — recovered adjuvant data, future metastatic component test conditional", "independentReplication": "none for metastatic contribution", "sampleSize": "Existing randomized n=42, recruiting single-arm target n=40; zero mission-funded participants; measurement cap £350,000", "effect": "Turn a selected one-arm operation study into a bounded feasibility and future-randomization decision", "limits": [ "Sponsor agreement unconfirmed", "Historic randomization may be incomplete", "External comparison is non-causal", "Small safety denominator", "Future randomization may be infeasible", "Thresholds are mission choices" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "The gate measures the whole local-treatment route while withholding causal claims." }, "animalEvidence": { "rating": "not-applicable", "reason": "Existing human treatment makes route accounting more useful than more models." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "The gate is about delivery, harm and patient benefit." }, "reproducibility": { "rating": "moderate", "reason": "Independent custody and common clocks would allow the active study to be checked against older cohorts." }, "commercialConflicts": { "rating": "low", "reason": "Milestone funding goes only to measurement and public analysis, not procedures." }, "recency": { "rating": "current", "reason": "The gate can be inserted while NCT04858009 is recruiting." } } }, "layout": { "x": 1126, "y": 1050 }, "updated": "2026-09-15" }, { "id": "hypothesis-human-dormancy-reawakening-gate", "title": "Hypothesis: a treatment-linked human liver reawakening state identifies preventable PDAC recurrence", "shortTitle": "Dormancy reawakening gate", "type": "hypothesis", "status": "hypothesis", "scope": [ "pdac", "transfer-unproven" ], "summary": "If chemotherapy-driven liver injury awakens a reproducible subset of residual PDAC cells in people, a short low-burden intervention could prevent some liver recurrences.", "content": "Biological rationale: a 2026 PDAC model links chemotherapy-induced hepatocyte senescence to TGF-beta, neutrophil extracellular traps, CCDC25 signalling and renewed disseminated-cell growth. An older reversible PDAC model independently shows that residual cells can survive visible regression and later restart disease. Supporting evidence is causal only in models; human samples show linked host signals, not the complete route. Contrary evidence: chemotherapy also eliminates micrometastases and improves survival after surgery; NET-DNA can reflect inflammation or occult growing cancer; late recurrence may be continuous growth rather than awakening. Assumptions: the host signal occurs before imaging recurrence, covers a useful fraction of patients, and can be blocked without infection, thrombosis, wound-healing or chemotherapy harm. Predicted result: a locked host-injury plus tumour-derived panel rises after chemotherapy and before liver recurrence, is absent in matched non-liver recurrence, and identifies a state in which blocking the route suppresses outgrowth in patient-derived liver cultures. Falsification experiment: embed serial blood before, during and after adjuvant treatment in 250 resected patients; compare liver, local, lung and no recurrence; analyse with outcome labels concealed; reproduce the tissue signal in a second laboratory. Do not give an anti-NET or CCDC25 intervention unless the human timing and perturbation both pass. Stop if adjusted liver-recurrence discrimination is below AUC 0.70, if lead time is under eight weeks, if the signal merely tracks total ctDNA, or if target blockade impairs antimicrobial function at the needed exposure. Planning estimate: £2.5m and 30 months. Major safety risks are infection, thrombosis, liver injury and weakening chemotherapy. Planning rank score uses impact 4/5 × technical probability 0.25 × speed 0.67 per year ÷ cost £2.5m = 0.268; this ranks below the direct RAS and MRD-action experiments but above a drug trial based only on mouse dormancy.", "sourceIds": [ "shi-2026-chemo-dormant-dtc", "lin-2013-reversible-pdac-dormancy", "conroy-2022-prodige24-five-year", "botta-2024-tumor-informed-ctdna", "effenberger-2012-pdac-bone-marrow-dtc", "nordgard-2022-unresectable-dtc", "chalabi-hajkarim-2025-organotropism", "pei-2025-spatial-metastatic-pdac", "tanaka-2012-microscopic-local-recurrence", "mcclellan-2026-quiescent-pdac-car-t" ], "links": [ { "target": "experiments", "relation": "part-of" }, { "target": "claim-pdac-dormancy-model-proven-human-uncertain", "relation": "tests" }, { "target": "unknown-dormant-cell-actionability", "relation": "tests" }, { "target": "hypothesis-human-liver-niche-gate", "relation": "competes-with" }, { "target": "programme-25m-prospective-observatory", "relation": "tests" }, { "target": "claim-human-pdac-dtc-prognostic-not-dormancy", "relation": "bounded-by" }, { "target": "failure-pdac-dormancy-sample-gap", "relation": "bounded-by" }, { "target": "trial-pdac-residual-cell-state-action-map", "relation": "bounded-by" } ], "evidence": { "humanProspective": "none", "humanRetrospective": "weak", "randomised": "none", "independentReplication": "weak", "sampleSize": "Proposed resected-PDAC cohort n=250", "effect": "Target: liver-recurrence AUC at least 0.70, at least eight weeks lead time and added information beyond ctDNA", "limits": [ "Generated hypothesis", "Causal route is model-derived", "New 2026 result is not independently replicated", "Host signal may be a consequence of occult disease", "Anti-NET treatment could cause serious harm", "Rank inputs are explicit programme judgements" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "Model reawakening is coherent with human occult dissemination, but the full liver route is not shown in people." }, "animalEvidence": { "rating": "strong", "reason": "The senescence-NET-CCDC25 chain is causal in a PDAC dormancy model." }, "inVitroEvidence": { "rating": "moderate", "reason": "Patient-derived liver systems can test blockade, with incomplete immune and injury context." }, "reproducibility": { "rating": "none", "reason": "The 2026 mechanism and proposed human state have no independent clinical reproduction." }, "commercialConflicts": { "rating": "moderate", "reason": "Assay and later target access require declared ownership, open methods and independent analysis." }, "recency": { "rating": "current", "reason": "The evidence set includes 2025 human spatial studies and 2026 treatment-specific models." } } }, "layout": { "x": 57, "y": 129 }, "updated": "2026-09-14", "contentSections": [ "The 250-person test is a nested subset of the existing 800-person £25m observatory, not an extra £2.5m programme. Use the existing molecular-assay, biobanking, rapid-biopsy and spatial-pathology lines. Add archived primary tissue, safe recurrence biopsy before new treatment and rapid autopsy; use marrow only as an optional comparison. Require at least 25 pre-imaging recurrence cases with a candidate state, ancestry support in 80% of pairs and 80% two-laboratory agreement before any intervention." ] }, { "id": "hypothesis-human-liver-niche-gate", "title": "Hypothesis: validate a human liver-niche state before attempting metastasis interception", "shortTitle": "Human liver-niche gate", "type": "hypothesis", "status": "hypothesis", "scope": [ "pdac", "colorectal", "transfer-unproven" ], "summary": "A reproducible blood-and-tissue signature preceding liver metastasis could create an interception window, but the mouse exosome result must first survive a human prospective test.", "content": "Biological rationale: PDAC-derived exosomal MIF helped form a liver niche before metastasis in mice, and pancreatic plus colorectal cancers frequently kill through liver disease. Contrary evidence: human association does not prove the niche causes metastasis; occult metastases may already be producing the signal; broad stromal manipulation can accelerate disease. Assumptions: a host response precedes image-detectable metastasis by months, is measurable in serial blood, and has a safely targetable component. Prediction: among resected patients without visible liver disease, a locked plasma-protein, vesicle and myeloid-state panel rises before liver recurrence but not isolated local recurrence, and the signal remains after adjustment for total tumour burden. Falsification experiment: prospective serial sampling in 150 resected PDAC and 150 liver-metastasis-prone colorectal patients with blinded assay and prespecified time-to-liver-recurrence analysis. Do not test a drug unless the signature independently validates and perturbation reverses it in matched human liver models. Planning estimate: £3m, 24–30 months; sampling risk is low, but a later interception drug could impair immunity, repair or liver function.", "contentSections": [ "Major safety risks: serial blood sampling is low risk; any later niche intervention could impair infection control, wound repair, vascular integrity or liver function, so the observational gate must pass before drug exposure." ], "sourceIds": [ "costa-silva-2015-liver-niche", "iacobuzio-donahue-2009-autopsy-patterns", "hishinuma-2006-recurrence-autopsy", "rhim-2014-stroma-restrains" ], "links": [ { "target": "experiments", "relation": "part-of" }, { "target": "claim-liver-niche-before-metastasis", "relation": "tests" }, { "target": "claim-pdac-death-can-be-local-or-metastatic", "relation": "constrained-by" }, { "target": "programme-5m-decision-core", "relation": "tests" } ], "evidence": { "humanProspective": "none", "humanRetrospective": "weak", "randomised": "none", "independentReplication": "weak", "sampleSize": "Proposed prospective cohorts n=300 total", "effect": "Target: independent prediction of liver-specific recurrence beyond total tumour burden", "limits": [ "Generated hypothesis", "Core causal evidence is mouse work", "Reverse causation from occult metastases", "Assay not fixed", "Drug intervention is deliberately outside the first experiment" ] }, "layout": { "x": 51, "y": 125 }, "updated": "2026-09-14" }, { "id": "hypothesis-human-replication-stress-selector", "title": "Hypothesis: a tumour-measured replication-stress state selects WEE1 benefit", "shortTitle": "Replication-stress selector", "type": "hypothesis", "status": "hypothesis", "scope": [ "pdac" ], "summary": "WEE1 treatment may work only when chemotherapy creates measurable tumour DNA stress and the backup pause is actually removed.", "content": "Biological rationale. A cancer cell copying damaged DNA uses WEE1 as an emergency stop. If gemcitabine raises replication stress and a WEE1 inhibitor removes that stop inside the same tumour, the cell may enter division with lethal damage. Supporting evidence. PDAC models show combination activity; a 34-person adavosertib study established tolerable dosing and surrogate-tissue checkpoint change; and the completed azenosertib safety stage includes planned paired tumour biopsies. Contrary evidence. Randomized CHK1 plus gemcitabine did not improve survival after reaching planned exposure; a newer CHK1 combination produced no response even though two participant organoids were sensitive; broad CDK4/6 routes repeatedly failed; and no controlled PDAC WEE1 result exists. Assumptions. At least ten of 12 paired tumour samples are usable, tumour checkpoint change can be distinguished from chemotherapy damage alone, the state is not created only in dying cells, and the assay repeats by two methods. Prediction. At least eight paired tumours show reduced WEE1-checkpoint output plus increased DNA damage, and at least five of 12 participants remain alive without progression at six months. The state should be stronger in controlled than progressing disease and should not be replaced by a hair, blood or organoid result. Falsification experiment. Recover and independently analyse all 12 NCT06015659 records. Falsify the route for PDAC if fewer than ten pairs are usable, fewer than eight show tumour change, fewer than 5/12 reach six-month PFS, a treatment-related death occurs, the marker does not agree across two methods, or marker-positive tumours progress at the same rate. Only if every gate passes should identical chemotherapy be randomized with or without a WEE1 inhibitor. Estimated cost is £0 for new recruitment; data recovery uses the existing decision core. A later randomized test remains unpriced and unfunded. Estimated time is six to nine months after data access. Major safety risks are low blood counts, infection, fatigue, nausea, forcing damaged healthy cells through division, biopsy harm and wasting scarce patients on a model-supported but human-inactive route.", "sourceIds": [ "laquente-2017-chk1-pdac", "huffman-2023-ly2880070-pdac", "cuneo-2019-adavosertib-pdac", "huffman-2024-azenosertib-pdac-design", "clinicaltrials-nct06015659-azenosertib" ], "links": [ { "target": "experiments", "relation": "part-of" }, { "target": "failure-broad-cell-cycle-checkpoint-pdac", "relation": "constrained-by" }, { "target": "unknown-wee1-replication-stress-pdac", "relation": "tests" }, { "target": "trial-cell-cycle-checkpoint-action-map", "relation": "gated-by" } ], "evidence": { "humanProspective": "weak", "humanRetrospective": "not-applicable", "randomised": "none", "independentReplication": "none", "sampleSize": "Existing WEE1 safety stage n=12; no new participants", "effect": "Generated paired-tumour, six-month-control and safety falsification gates", "limits": [ "Thresholds are mission decisions", "Data access unconfirmed", "No comparator in existing stage", "Marker assays may not be available", "Later randomized design unpriced" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "The proposed interaction requires chemotherapy damage and checkpoint removal in the same human tumour." }, "animalEvidence": { "rating": "strong", "reason": "Models support the interaction but past checkpoint translation failed." }, "inVitroEvidence": { "rating": "strong", "reason": "Replication stress and DNA damage are directly measurable, though organoid sensitivity alone failed to predict patients." }, "reproducibility": { "rating": "none", "reason": "No human selector or controlled WEE1 efficacy effect has repeated." }, "commercialConflicts": { "rating": "moderate", "reason": "The drug is company supplied, while the proposed analysis can be independent." }, "recency": { "rating": "current", "reason": "The hypothesis is tied to the current completed 12-person record." } } }, "layout": { "x": 216, "y": 134 }, "updated": "2026-09-15" }, { "id": "hypothesis-k912-component-identification-gate", "title": "Measure the K-912 interaction term before another sonodynamic patient", "shortTitle": "K-912 component gate", "type": "hypothesis", "status": "testing", "scope": [ "pdac", "cholangiocarcinoma", "sonodynamic-therapy", "mechanism" ], "summary": "A capped data-and-tissue experiment can decide whether K-912 is sound-activated or merely another drug added to thermal HIFU.", "content": "Biological rationale. Sound could trigger micelle disruption, cavitation or chemical reactions that increase active epirubicin inside the focal volume. Supporting evidence. The first-human package produced nine coagulation signals at much lower power than earlier thermal HIFU, with little severe harm through 30 days. Contrary evidence. No component was isolated; complete coagulation followed 150 watts more clearly than K-912 dose, and local scan changes did not imply systemic control. Assumptions. The investigators, Kowa and Sonire will release all 12 participant records and the exact K-912 formulation, while four independent laboratories can reproduce acoustic waveforms and matched temperature in patient-derived pancreatic models. Prediction. Compared with K-912 alone and temperature-matched heat alone, ultrasound plus K-912 will at least double intratumour free epirubicin and exceed Bliss-independent expected cell killing by 15 percentage points in at least three of four models, without increasing injury in matched normal pancreatic duct and heart-cell models. Falsification experiment. Enroll nobody and spend at most £450,000 over six months: £150,000 for participant-file recovery and £300,000 for randomized, blinded four-way laboratory work using no treatment, K-912, matched heat and K-912 plus the original acoustic waveform. Measure particle break-up, free drug, temperature, cavitation, cell death and recovery. Stop the sonodynamic label if release rises less than twofold, excess kill is under 15 points in two or more models, the effect disappears when temperature is matched, or normal-cell injury removes the window. If the interaction fails, evaluate the parts only under their existing chemotherapy and thermal-HIFU evidence. Major safety risks are cumulative epirubicin heart and marrow damage, pancreatitis, bowel heating, skin burn, cavitation injury, liver or bile-duct injury and delay of effective systemic therapy.", "sourceIds": [ "muragaki-2026-k912-sonodynamic" ], "links": [ { "target": "experiments", "relation": "part-of" }, { "target": "unknown-k912-ultrasound-drug-interaction-and-benefit", "relation": "tests" }, { "target": "failure-k912-component-and-local-endpoint", "relation": "repairs" }, { "target": "trial-pdac-k912-sonodynamic-action-map", "relation": "implemented-by" }, { "target": "company-sonire-therapeutics", "relation": "depends-on" }, { "target": "programme-5m-decision-core", "relation": "constrained-by" } ], "evidence": { "humanProspective": "weak", "humanRetrospective": "none", "randomised": "none in humans; planned randomized laboratory comparison", "independentReplication": "none", "sampleSize": "Existing n=12 plus four independent patient-derived model systems; zero new people; cap £450,000", "effect": "Identify or reject an ultrasound-specific K-912 interaction before clinical expansion", "limits": [ "Laboratory interaction may not transfer to people", "Exact waveform may be proprietary", "Stored participant samples may be absent", "Formulation access is unconfirmed", "Four models cannot cover full tumour variation", "Thresholds and cap are mission choices" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "The proposed measurements distinguish heat, drug and their claimed acoustic interaction." }, "animalEvidence": { "rating": "moderate", "reason": "Animal evidence supports a gate but should not trigger another human cohort." }, "inVitroEvidence": { "rating": "weak", "reason": "This experiment is designed to create the missing matched-temperature component evidence." }, "reproducibility": { "rating": "moderate", "reason": "Four independent laboratories and blinded allocation test whether the effect travels." }, "commercialConflicts": { "rating": "high", "reason": "Release of formulation and device waveform must not give vendors control of analysis or publication." }, "recency": { "rating": "current", "reason": "The gate responds to the 2026 human report before further sonodynamic expansion." } } }, "layout": { "x": 1060, "y": 984 }, "updated": "2026-09-15" }, { "id": "hypothesis-kras-tcr-human-persistence-gate", "title": "Mutation-HLA-matched TCR cells can produce repeatable PDAC regression when presentation and persistence are retained", "shortTitle": "KRAS TCR persistence gate", "type": "hypothesis", "status": "proposed", "scope": [ "pdac", "pan-cancer", "transfer-unproven" ], "summary": "Use the existing KRAS-G12V trial to test a full presentation-entry-persistence chain before any new cohort or controlled survival trial.", "content": "Biological rationale. KRAS is a trunk driver inside most pancreatic cancer cells. A matched TCR can read a mutant KRAS fragment displayed by the correct HLA molecule, giving greater cancer specificity than a surface protein shared with normal tissue. Supporting evidence. One KRAS-G12D patient had 72% regression lasting at least six months, and one of five KRAS-G12V patients had complete liver-metastasis regression lasting 5.5 months. Contrary evidence. A second G12D patient progressed; four of five G12V patients did not respond; repeat infusions produced no response; mesothelin CAR-T reached tissue without control; and tumours can lose HLA presentation or disable T cells. Assumptions. The mutation is present across lesions, the matching HLA displays enough fragment, conditioning does not dominate harm, the cell product reaches tumours, receptor and function persist, and anti-receptor antibodies do not remove repeat doses. Prediction. In at least 25 infused NCT04146298 participants, at least five have independently confirmed responses, at least four last six months, and a measurement fixed before outcomes separates durable response from failure while paired tumour sampling succeeds in at least 80%. Falsification experiment. Recruit nobody outside the existing trial. Retain every screened, collected and manufactured denominator; assay mutation fraction, HLA and display machinery, tumour entry, cell state, receptor persistence and anti-receptor antibodies. Falsify the branch if fewer than five respond, fewer than four responses last six months, no fixed measurement predicts outcome, collected-to-infused completion is below 70%, treatment causes death or normal-tissue attack, or repeat dosing remains immunogenic. Estimated mission cost is £0; data recovery and protocol repair are requested from the active programme. Estimated time is about 27 months to the registry's December 2028 completion estimate. Major safety risks are conditioning-related infection and bleeding, cytokine-release syndrome, neurologic injury, incorrect normal-tissue recognition, receptor rejection and delayed care during custom manufacture.", "sourceIds": [ "xu-2026-kras-g12v-tcr", "clinicaltrials-nct04146298-kras-g12v-tcr", "leidner-2022-kras-tcr", "aznar-2025-mesothelin-car-t-resistance" ], "links": [ { "target": "experiments", "relation": "part-of" }, { "target": "claim-kras-g12v-tcr-early-signal", "relation": "supported-by" }, { "target": "claim-kras-tcr-one-response-not-platform-proof", "relation": "supported-by" }, { "target": "failure-cell-therapy-route-and-denominator-pdac", "relation": "must-escape" }, { "target": "trial-pdac-engineered-cell-action-map", "relation": "implemented-by" } ], "evidence": { "humanProspective": "weak", "humanRetrospective": "not-applicable", "randomised": "none", "independentReplication": "weak", "sampleSize": "Seven reported PDAC patients across two mutation-HLA routes; existing G12V trial planned n=30", "effect": "Prospective mechanism and replication gate; no assigned survival effect", "limits": [ "Thresholds are programme decisions", "No concurrent control", "Narrow mutation-HLA reach", "Biopsy feasibility unproved", "A positive measurement stage would not establish useful-life benefit" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Two human regressions show the driver fragment can be recognized in PDAC under matching HLA conditions." }, "animalEvidence": { "rating": "strong", "reason": "Matched organoids and xenografts support natural KRAS-G12V presentation and killing." }, "inVitroEvidence": { "rating": "strong", "reason": "Mutation specificity, HLA dependence, cell potency and receptor antibodies can be measured." }, "reproducibility": { "rating": "weak", "reason": "Responses occurred in two routes but remain single patients within tiny cohorts." }, "commercialConflicts": { "rating": "moderate", "reason": "The receptor developers run the trial and control patient-level records." }, "recency": { "rating": "current", "reason": "Uses the 2026 interim result and current registry." } } }, "layout": { "x": 256, "y": 182 }, "updated": "2026-09-15" }, { "id": "hypothesis-kras-wildtype-driver-route-completion", "title": "A DNA-to-RNA reflex route can deliver rare-driver treatment before fitness is lost", "shortTitle": "Rare-driver route completion", "type": "hypothesis", "status": "hypothesis", "scope": [ "pdac", "precision-treatment", "gene-fusion", "care-delivery", "capital-allocation" ], "summary": "Broaden the already funded NRG1 route to every validated KRAS-wild-type driver without adding another cohort or another budget.", "content": "Biological rationale. A pancreatic tumour without activated KRAS may depend on another measurable growth driver, and RNA can expose fusions that DNA misses. Supporting evidence. Three large or repeated profiling series find KRAS-wild-type enrichment for BRAF and receptor fusions, while NTRK, MET, ROS1, RET, BRAF and NRG1 cases show human drug sensitivity. Evidence against. Most treatment records are single cases, tumour-wide approvals pool other cancers, and Know Your Tumor delivered matched treatment to only 46/1,856 referrals. Assumptions. Existing English laboratories can expose their target coverage, adequate tissue can reach RNA testing, a specialist group can make a decision within seven days, and approved or trial access is possible without delaying standard therapy. Prediction. Reusing the existing 5,000-person NRG1 route will produce valid KRAS status in at least 95%, reflex RNA in at least 90% of true wild-type or driver-negative cases, a result within 14 days in at least 85%, and a documented access decision within seven days for at least 90% of strong drivers. At 10.7% KRAS wild type, 5,000 × 0.107 = 535 reflex candidates. If 20% carry a targetable fusion, about 107 could be identified; that 20% is a planning assumption bounded by published 18.6% to 30.8% fusion rates, not a promised treatment yield. Falsification experiment. Join tissue, DNA, RNA, review, access, treatment and outcome for every diagnosis and reject the route if any gate fails, if standard treatment is delayed in more than 5%, or if laboratories disagree on more than 10% of split samples. Estimated cost is no more than the existing £1.5m NRG1 route plus its £150,000 data-recovery line; this pass adds £0 and broadens the deliverable. Estimated time is 24 months. Major safety risks include biopsy harm, false driver calls, treatment delay, off-label toxicity, loss of effective standard treatment and unequal access. No variant of uncertain meaning triggers treatment. Each target's response and survival remain separate.", "sourceIds": [ "philip-2022-kras-wildtype-pdac", "fusco-2021-kras-wildtype-fusions", "singh-2023-kras-wildtype-drivers", "pishvaian-2020-know-your-tumor", "mehdi-2024-kras-wildtype-case-series", "oreilly-2019-ntrk-pdac-response", "fda-2018-larotrectinib-approval", "fda-2024-bizengri-multidisciplinary-review", "nhs-england-2026-cancer-genomic-test-directory" ], "links": [ { "target": "experiments", "relation": "part-of" }, { "target": "hypothesis-nrg1-rna-route-completion", "relation": "extends" }, { "target": "unknown-kras-wildtype-driver-route-completion", "relation": "tests" }, { "target": "failure-kras-wildtype-matching-route-and-selection", "relation": "succeeds-where" }, { "target": "trial-pdac-kras-wildtype-action-map", "relation": "implemented-by" } ], "evidence": { "humanProspective": "weak", "humanRetrospective": "strong", "randomised": "weak — this is a measurement and delivery hypothesis", "independentReplication": "moderate — driver enrichment repeats, but the proposed route does not", "sampleSize": "Existing route n=5,000; about 535 reflex candidates at 10.7% KRAS wild type and about 107 fusion candidates at a 20% planning assumption", "effect": "Falsifiable whole-route completion experiment with £0 added capital", "limits": [ "Fusion-yield assumption", "Rare target groups", "No pooled efficacy claim", "Changing access", "Tissue and RNA failure", "No untreated testing control" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Alternative drivers replace a missing common growth switch and can be directly inhibited." }, "animalEvidence": { "rating": "moderate", "reason": "Several driver models support sensitivity but cannot establish route utility." }, "inVitroEvidence": { "rating": "strong", "reason": "Patient-derived models show direct dependence for selected BRAF and fusion states." }, "reproducibility": { "rating": "weak", "reason": "The exact national DNA-to-RNA-to-treatment route has not been tested." }, "commercialConflicts": { "rating": "moderate", "reason": "Drug and profiling owners benefit, so analysis and publication must remain independent." }, "recency": { "rating": "current", "reason": "The experiment uses July 2026 English panel coverage and current approvals." } } }, "layout": { "x": 2220, "y": 1660 }, "updated": "2026-09-15" }, { "id": "hypothesis-laser-existing-study-recovery-gate", "title": "Recover the failed pancreatic laser programmes before treating anyone else", "shortTitle": "Laser data-recovery gate", "type": "hypothesis", "status": "testing", "scope": [ "pdac", "laser-ablation", "trial-recovery", "thermal-dosimetry" ], "summary": "Up to 47 existing records can test delivery, thermal coverage and failure modes more cheaply and safely than another device cohort.", "content": "Biological rationale. Laser energy can be measured and heat spread can be reconstructed from fibre position, power, time, perfusion and anatomy. Supporting evidence. Nine endoscopic procedures completed without reported harm and created dose-dependent dead tissue. Contrary evidence. Three of 13 open surgical treatments caused pancreatic-fluid leaks, a randomized study stopped at eight, another study stopped at five for device loss and a planning study is stale. Assumptions. Sponsors, sites and Clinical Laserthermia Systems will provide assignment, screening, procedure, calibration, image, temperature, adverse-event and follow-up files for up to 47 records across the n=9, n=15, n=8, n=5 and target-n=10 programmes, with de-duplication and independent publication rights. Prediction. Reanalysis will reconcile every assigned person, explain at least 90% of attempted-but-not-completed procedures and show whether a locked heat model predicts dead-zone volume within 25% while holding temperatures at protected structures below protocol limits. Falsification experiment. Enroll nobody, buy no device or equity and spend at most £350,000. Freeze definitions for procedure completion, treated volume, thermal injury, chemotherapy delay, fistula, admission and death before opening outcome files. Publish the eight-person randomized flow even if effect estimates are unusable. Stop if raw access is refused, participant overlap cannot be resolved, less than 90% of procedures have calibration and imaging, model error exceeds 25%, serious procedure harm reaches 10% or any planned safety boundary was crossed without detection. A pass permits only a design decision, not clinical use. Major risks are pancreatitis, pancreatic-fluid leak, bleeding, bowel or vessel burns, infection, hardware loss and interruption of systemic care.", "sourceIds": [ "dimatteo-2018-eus-laser-ablation", "paiella-2021-imilt-pancreas", "clinicaltrials-nct02702986-imilt-verona", "clinicaltrials-nct02973217-imilt-marseille", "clinicaltrials-nct03187587-imilt-randomized", "clinicaltrials-nct03784417-eus-laser", "clinicaltrials-nct05549960-laser-model", "cls-2026-investment-case" ], "links": [ { "target": "experiments", "relation": "part-of" }, { "target": "unknown-laser-ablation-delivery-and-net-benefit", "relation": "tests" }, { "target": "failure-laser-fistula-accrual-and-device-continuity", "relation": "repairs" }, { "target": "trial-pdac-laser-ablation-action-map", "relation": "implemented-by" }, { "target": "company-clinical-laserthermia-systems", "relation": "depends-on" }, { "target": "programme-5m-decision-core", "relation": "constrained-by" } ], "evidence": { "humanProspective": "weak", "humanRetrospective": "none", "randomised": "weak — existing n=8 record without results", "independentReplication": "weak", "sampleSize": "Up to 47 existing records before de-duplication; zero new participants; cap £350,000", "effect": "Convert fragmented and failed studies into a stop or design decision", "limits": [ "Raw files may no longer exist", "Protocols and devices differ", "Temperature maps may be incomplete", "Small samples limit harm precision", "Company cooperation is unknown", "Thresholds and cap are mission choices" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Power, duration, geometry, perfusion and heat injury form a testable physical chain." }, "animalEvidence": { "rating": "moderate", "reason": "Animal data can constrain heat models but cannot replace human failure records." }, "inVitroEvidence": { "rating": "strong", "reason": "Thermal injury thresholds can support reconstruction." }, "reproducibility": { "rating": "moderate", "reason": "A locked cross-study model directly tests whether the device effect travels." }, "commercialConflicts": { "rating": "high", "reason": "Payment requires full data custody and negative-result publication independent of the vendor." }, "recency": { "rating": "current", "reason": "The gate responds to present registry gaps and current company priorities." } } }, "layout": { "x": 988, "y": 912 }, "updated": "2026-09-15" }, { "id": "hypothesis-localized-losartan-existing-trial-gate", "title": "Losartan adds useful route completion in localized PDAC when it measurably opens tumour flow", "shortTitle": "Localized losartan data gate", "type": "hypothesis", "status": "proposed", "scope": [ "pdac", "transfer-unproven" ], "summary": "Use the completed 168-person allocation and require flow, treatment, surgery, recurrence and survival to move together before any new trial.", "content": "Biological rationale. Losartan may reduce collagen and solid pressure, open compressed vessels and improve chemotherapy or radiation delivery before localized disease escapes. Supporting evidence. One 49-person route reported 34/49 clean-margin resections and 31.4-month median survival. Contrary evidence. That route combined four treatments without control; AFPAC found no response, progression or survival improvement in 88 randomized advanced patients, and plasma TGF-beta did not change differently. Assumptions. NCT03563248 preserved allocation, enough no-losartan controls, treatment exposure, surgery and survival follow-up; any measured flow change precedes outcome. Prediction. In all 168 assigned participants, losartan improves a fixed treatment-route endpoint by at least ten percentage points or gives PFS HR at most 0.75, while OS does not worsen, severe harm rises no more than five points and at least 80% have usable flow or oxygen measurement. Falsification experiment. Recruit nobody. Release the protocol and analysis plan, retain every allocated person, estimate the losartan and nivolumab contributions separately, and join flow to chemotherapy completion, surgery, recurrence, function and survival. Falsify for a confidence interval including no effect on both route completion and PFS, no flow change, R0-only change without recurrence benefit, missing outcomes above 5%, severe-harm excess or an invented subgroup. Estimated mission cost is £0 beyond requesting the existing result; time is up to one year for release while survival follow-up continues. Major risks are hidden interaction between trial factors, blood-pressure harm, incomplete flow measurement, surgery selection and data refusal.", "sourceIds": [ "clinicaltrials-nct03563248-losartan-localized", "murphy-2019-losartan-lapc", "ramaswamy-2025-afpac-losartan" ], "links": [ { "target": "experiments", "relation": "part-of" }, { "target": "failure-losartan-delivery-signal-not-replicated", "relation": "learns-from" }, { "target": "unknown-localized-losartan-contribution-pdac", "relation": "resolves" }, { "target": "trial-pdac-vascular-hypoxia-action-map", "relation": "implemented-by" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "moderate", "independentReplication": "moderate", "sampleSize": "Existing randomized n=168, prior uncontrolled n=49 and contrary randomized n=88", "effect": "Existing-data contribution gate; no assigned treatment benefit", "limits": [ "Thresholds are mission decisions", "Factor interaction may reduce power", "OS follow-up is incomplete", "Flow measurement coverage is unknown" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The physical decompression chain can be joined to treatment delivery and outcome." }, "animalEvidence": { "rating": "strong", "reason": "Models support pressure reduction and perfusion improvement." }, "inVitroEvidence": { "rating": "weak", "reason": "Static culture is a poor test of pressure and flow." }, "reproducibility": { "rating": "weak", "reason": "The localized claim remains unconfirmed despite an advanced negative trial." }, "commercialConflicts": { "rating": "low", "reason": "The generic drug has little proprietary value, but the centre controls the record." }, "recency": { "rating": "current", "reason": "The 168-person record was checked in 2026." } } }, "layout": { "x": 340, "y": 294 }, "updated": "2026-09-15" }, { "id": "hypothesis-low-burden-ras-immune-extinction", "title": "Hypothesis: deep RAS suppression and immune priming work best at minimal burden", "shortTitle": "Low-burden RAS + immune", "type": "hypothesis", "status": "hypothesis", "scope": [ "pdac", "colorectal", "transfer-unproven" ], "summary": "Move proven common-RAS suppression into molecular residual disease, then add immune treatment only after tumour killing and a favourable T-cell-to-myeloid state are shown.", "content": "Biological rationale: daraxonrasib proves that common RAS dependence remains clinically exploitable in treated metastatic PDAC, but bulky heterogeneous disease still evolves resistance. Residual disease contains fewer cells and may have fewer resistant clones. A lymph-node-targeted KRAS vaccine and personalised neoantigen vaccine can induce immunity in some patients, although neither has randomized cure proof. Contrary evidence is now direct in humans. ELI-002 missed its main randomized disease-free-survival endpoint. Plerixafor plus PD-1 blockade moved CD8 T cells into tumours but produced no responses in 21 reported patients while suppressive macrophages rose. A vaccine and dual-checkpoint programme generated tumour-specific T-cell clones but only two responses among 57 treated patients and 68% grade 3 or worse related toxicity. POLAR suggests that a prepared HRD state may permit a long tail, but it missed its prespecified thresholds and had no randomized comparator. Assumptions: a validated MRD state exists; RAS inhibition creates measurable tumour killing; the immune intervention improves antigen-specific function without worsening suppressive myeloid or regulatory states. Prediction: in post-resection MRD-positive patients, sequential RAS suppression then immune priming will clear molecular disease more often and lengthen recurrence-free survival versus RAS suppression alone. Falsification experiment: first run a 20–30-patient paired-tissue pharmacology stage. Stop if tumour killing is absent, antigen-specific function does not rise, the intratumour T-cell-to-suppressive-myeloid ratio does not improve, or grade 3 or worse related toxicity exceeds the fixed safety bound. Only then open a randomized phase 2 of about 120 patients, with blinded MRD clearance, 18-month recurrence-free survival and a futility boundary if clearance improves by less than 10 percentage points. Planning estimate: £18m, 48 months; major risks are overlapping skin, gut, lung and immune toxicity, false-negative MRD and reproducing immune recruitment without tumour control. This is a generated hypothesis, not a treatment recommendation.", "sourceIds": [ "oreilly-2026-daraxonrasib", "fda-2026-rasonque-approval", "elicio-2026-amplify7p-results", "rojass-2025-neoantigen-followup", "oreilly-2019-durvalumab-tremelimumab", "ho-2025-cxcr4-pd1-pdac", "bever-2026-vaccine-checkpoint-pdac", "park-2026-polar-pdac" ], "links": [ { "target": "experiments", "relation": "part-of" }, { "target": "claim-daraxonrasib-common-ras-survival-breakthrough", "relation": "extends" }, { "target": "failure-eli002-phase2-missed-dfs", "relation": "constrained-by" }, { "target": "failure-immune-recruitment-without-response", "relation": "constrained-by" }, { "target": "unknown-ctdna-actionability", "relation": "depends-on" }, { "target": "programme-100m-randomized-platform", "relation": "tests" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "weak", "randomised": "weak", "independentReplication": "weak", "sampleSize": "Proposed paired-tissue gate n=20–30, then randomized phase 2 n≈120", "effect": "Target: at least 10-point absolute MRD-clearance gain and recurrence-free HR ≤0.70", "limits": [ "Generated combination hypothesis", "No human evidence for this sequence", "MRD assay sensitivity uncertain", "T-cell recruitment has failed as a stand-alone success marker", "Immune component may add toxicity without efficacy", "Budget and probability are planning judgments" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "Human RAS dependence and immune recruitment are established separately, but their low-burden sequence has not been tested." }, "animalEvidence": { "rating": "moderate", "reason": "Models support RAS-driven immune change and layered immune barriers, but several model-led combinations failed clinically." }, "inVitroEvidence": { "rating": "moderate", "reason": "Antigen-specific T-cell assays show recognition, while human trials show that this does not ensure tumour control." }, "reproducibility": { "rating": "none", "reason": "No clinical study has reproduced the proposed RAS-suppression then immune-priming sequence in molecular residual disease." }, "commercialConflicts": { "rating": "high", "reason": "RAS drugs, vaccines, checkpoint agents and MRD assays are controlled by different commercial parties with linked trial interests." }, "recency": { "rating": "current", "reason": "The design incorporates the 2026 RAS, vaccine-checkpoint and selected HRD maintenance results." } } }, "layout": { "x": 61, "y": 122 }, "updated": "2026-09-14" }, { "id": "hypothesis-measured-fixed-backbone-sonoporation", "title": "Hypothesis: sonoporation helps only when a fixed drug backbone gains measured tumour delivery", "shortTitle": "Measured sonoporation gate", "type": "hypothesis", "status": "hypothesis", "scope": [ "pdac", "pan-cancer", "transfer-unproven" ], "summary": "A physical delivery effect should appear as more active drug in tumour and a planned treatment interaction on one chemotherapy backbone before any survival trial is repeated.", "content": "Biological rationale. Gas microbubbles expand and contract in an ultrasound pressure field. Near a tumour vessel, that motion may briefly change the transport boundary and allow more chemotherapy to cross. In engineering terms, the intervention changes the channel, not the payload. Its benefit can therefore depend on molecule size, timing, infusion schedule, tumour perfusion and acoustic geometry. Supporting evidence. One 78-person randomized trial reported median survival of 9.1 versus 6.1 months. NCT04821284's highlighted gemcitabine/nab-paclitaxel subgroup reports 407.5 versus 234 days. Contrary evidence. The independent trial abstract reports no overall PFS or OS difference, both P>0.36, and withholds the information needed to judge the subgroup. The exploratory perfusion method agreed with clinical direction in 10/13 people but did not measure chemotherapy concentration or randomized contribution. Assumptions. The subgroup was planned, randomization remains balanced within it, a fixed backbone has physical properties that make ultrasound delivery useful, and a safe measurement can show increased active drug inside human tumour. Prediction. If the device genuinely adds delivery on gemcitabine/nab-paclitaxel, the complete NCT04821284 data will show a treatment-by-backbone interaction with its confidence interval excluding no difference, and device exposure will increase a locked tumour-drug measure that tracks outcome. Falsification. Stage zero costs £0: release and independently reanalyse the completed trial. Stop if the overall result is null and the planned interaction fails, if the subgroup was selected after outcomes were seen, or if missing data can plausibly create the difference. No new participant is exposed from an abstract. Only if that gate passes may a measurement stage use one fixed backbone and identical chemotherapy in both groups. It must capture at least 90% of randomized participants for treatment and survival, deliver at least 90% of planned device sessions in at least 80%, produce an independently validated tumour-delivery measure in at least 80% of its nested sample, and show at least a 25% median increase in the locked delivery measure without worse grade 3 or higher harm or chemotherapy dose intensity. Any failed gate closes the route. A later confirmatory trial must use overall and quality-adjusted survival, public stopping rules and an independent statistics team. Estimated cost: £0 now. Any later measurement or confirmation must displace equal capital from a lower-value branch; no device licence or company investment precedes the interaction and delivery gates. Main risks are vascular injury, cavitation outside tumour, contrast-agent reactions, missed deep lesions, operator dependence, extra treatment burden, chemotherapy delay and a chance subgroup mistaken for a system-specific effect.", "sourceIds": [ "wang-2023-sonochemotherapy-pdac", "clinicaltrials-nct04821284-sonoporation", "ius-2026-sonoporation-nct04821284-abstract", "adler-2026-sonoporation-perfusion" ], "links": [ { "target": "experiments", "relation": "part-of" }, { "target": "claim-sonochemotherapy-randomized-signal", "relation": "constrained-by" }, { "target": "failure-sonoporation-overall-replication-subgroup-rescue", "relation": "repairs" }, { "target": "unknown-pdac-sonoporation-treatment-effect", "relation": "tests" }, { "target": "trial-pdac-sonoporation-action-map", "relation": "gated-by" }, { "target": "hypothesis-selected-local-consolidation", "relation": "shares-physical-treatment-boundary-with" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "none", "randomised": "moderate — sufficient for data recovery, not another trial", "independentReplication": "weak — negative overall result is abstract-only", "sampleSize": "Stage zero uses the completed NCT04821284 dataset; any later sample size is conditional and not funded", "effect": "Generated gates: planned interaction first; ≥90% follow-up; ≥90% sessions in ≥80%; delivery measure in ≥80%; ≥25% median delivery increase; no severe-harm or chemotherapy-dose penalty", "limits": [ "Thresholds are programme decisions, not validated treatment markers", "A delivery increase may not cause survival gain", "Tumour sampling may be unsafe or biased", "Operator and geometry effects may limit reproducibility", "No current capital is assigned" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "A physical transport change is plausible and molecule-dependent, which creates a testable fixed-backbone prediction." }, "animalEvidence": { "rating": "moderate", "reason": "Model delivery effects support measurement but not clinical adoption." }, "inVitroEvidence": { "rating": "weak", "reason": "Bench systems omit vessel flow, clearance, tissue pressure and patient-scale acoustic geometry." }, "reproducibility": { "rating": "weak", "reason": "The first overall survival signal did not reproduce in the second trial's public abstract." }, "commercialConflicts": { "rating": "high", "reason": "Positive subgroup framing could unlock device and contrast-agent use after an overall negative result." }, "recency": { "rating": "current", "reason": "Designed around the 2026 trial abstract and imaging paper." } } }, "layout": { "x": 166, "y": 164 }, "updated": "2026-09-15" }, { "id": "hypothesis-mitazalimab-controlled-survival-gate", "title": "Mitazalimab adds a clinically useful survival effect to fixed mFOLFIRINOX", "shortTitle": "Mitazalimab survival gate", "type": "hypothesis", "status": "proposed", "scope": [ "pdac", "pan-cancer", "transfer-unproven" ], "summary": "Release the full 94-person record, then require an owner, matched capital and a roughly 450-person placebo-controlled survival trial.", "content": "Rationale. Mitazalimab activates CD40, an immune-cell switch that can improve the showing of cancer fragments to T cells. In OPTIMIZE-1, 57 evaluable participants had 42.1% confirmed response, 12.6-month response duration and 14.9-month median survival. Contrary evidence. The study has no control and excluded part of the enrolled route from efficacy. Sotigalimab's two CD40 groups missed their historical survival gates. Cabiralizumab shows that changing tumour-supporting immune cells need not improve outcomes. Alligator stopped its own phase 3 work after cash fell to SEK16.6m. Assumptions. Final all-enrolled data preserve the signal, drug supply and quality are transferable, a solvent partner owns delivery, mFOLFIRINOX exposure stays equal, and the benefit is not selection or later care. Prediction. About 450 untreated metastatic-PDAC patients randomized 1:1 to mitazalimab or placebo plus identical mFOLFIRINOX yield enough deaths to test an OS HR of 0.75; the useful threshold is HR at most 0.75 with its 95% CI below 1, no severe-event rise above ten points, no treatment-related mortality rise above two points and non-worse patient function. Falsification and finance. Phase zero is full release of all 94 participants, manufacturing records and individual data. Mission capital is £0 before that. If it passes and an independent commercial or public partner supplies the drug and at least half of an estimated £50m trial, the mission may fund up to £5m for rights and startup and up to £20m against enrollment and data milestones. Stop at interim futility, safety breach, unequal chemotherapy delivery, supply failure, missing outcome above 5% or refusal of full publication. Estimated duration is four years. Major risks are severe infection and blood-count effects from chemotherapy, immune and infusion reactions, liver injury, financial failure and an underpowered or unavailable trial.", "contentSections": [ "Supporting evidence is the 57-person response, duration and survival signal. It supports one controlled test, not use. If all gates pass, mission capital remains capped at £25m total and at least half of trial cost must come from the solvent partner." ], "sourceIds": [ "van-laethem-2024-optimize1", "van-laethem-2025-optimize1-biomarkers", "clinicaltrials-nct04888312-optimize1", "padron-2022-prince", "clinicaltrials-nct03214250-prince", "clinicaltrials-nct03336216-cabiralizumab", "alligator-2026-q2" ], "links": [ { "target": "experiments", "relation": "part-of" }, { "target": "claim-cd40-signal-without-contribution-proof", "relation": "supported-by" }, { "target": "unknown-cd40-added-survival-pdac", "relation": "resolves" }, { "target": "failure-immune-reprogramming-without-patient-benefit", "relation": "must-escape" }, { "target": "trial-immune-reprogramming-action-map", "relation": "implemented-by" }, { "target": "company-alligator-bioscience", "relation": "depends-on" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "not-applicable", "randomised": "weak", "independentReplication": "weak", "sampleSize": "Current efficacy n=57; registry n=94; proposed randomized n about 450", "effect": "Tests OS HR 0.75 with safety and function gates; current contribution is unknown", "limits": [ "Cost is a planning estimate", "Event assumptions need final statistical design", "Product ownership and supply are unsettled", "Fixed marker score is not independently validated" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The mechanism and human activity justify one controlled test." }, "animalEvidence": { "rating": "strong", "reason": "Multiple PDAC models support combining CD40 activation and chemotherapy." }, "inVitroEvidence": { "rating": "strong", "reason": "CD40 activation and downstream immune effects are directly measurable." }, "reproducibility": { "rating": "weak", "reason": "Mitazalimab has one uncontrolled efficacy set and another CD40 drug did not confirm broad benefit." }, "commercialConflicts": { "rating": "high", "reason": "The product needs a sale or license and the current owner controls its records." }, "recency": { "rating": "current", "reason": "Uses the 2026 development stop and current trial record." } } }, "layout": { "x": 298, "y": 238 }, "updated": "2026-09-15" }, { "id": "hypothesis-mtap-prmt5-ras-deeper-control", "title": "Hypothesis: MTAP-selected PRMT5 inhibition deepens and extends RAS control", "shortTitle": "PRMT5 plus RAS control", "type": "hypothesis", "status": "hypothesis", "scope": [ "pdac" ], "summary": "Vopimetostat may add a second dependency to RAS blockade in MTAP-deleted PDAC, but the interaction requires direct randomized measurement.", "content": "Rationale. RAS inhibition removes a central growth signal. In MTAP-deleted cells, PRMT5 already runs with reduced reserve because MTA has accumulated; an MTA-cooperative inhibitor can reduce that reserve further. The combination may therefore close two independent routes and make tumour escape harder. This is plausible but not yet a measured interaction. Supporting evidence. Vopimetostat alone produced a 15% response rate in the 39-person mature pancreatic set. Its RAS combinations produced 25 responses among 39 currently evaluable patients. AMG 193 independently showed complete tumour PRMT5 inhibition and pan-cancer responses. Contrary evidence. Broad epigenetic combinations failed; there is no randomized combination control; follow-up is short; 15 treated combination patients were not response evaluable; and adding a second drug can reduce RAS dose intensity or cause cumulative harm. Assumptions. Complete MTAP loss creates comparable PRMT5 stress across enrolled tumours, both drugs reach the same lesions, the chosen doses preserve each drug's active exposure, and added response can translate into longer control rather than faster resistance. Prediction. Against the identical RAS inhibitor alone, the combination should increase confirmed response by at least 15 percentage points, reduce progression risk by at least 20%, extend response duration and preserve at least 90% RAS-drug dose intensity. The effect should require complete MTAP loss and tumour PRMT5 inhibition. Falsification. Stop the interaction claim if randomized PFS HR exceeds 0.80, the planned 95% interval includes 1.0, response gain is below 15 points, severe related harm rises by more than 10 points, RAS dose intensity falls by more than 10 points, or benefit appears equally in MTAP-intact disease. Sequence. First mature all 54 existing PDAC records; second lock assays and comparator; third run a sponsor-funded randomized contribution test with independent analysis and public negative results. Timing is roughly one year for mature early data and three to five years for controlled survival. Mission capital is £0 because Tango has sufficient disclosed funding. Main safety risks are anemia, thrombocytopenia, nausea, fatigue, rash, mouth inflammation, interaction-driven dose reduction and delayed marrow toxicity.", "sourceIds": [ "richards-2006-ci994-pdac", "heumann-2022-azacitidine-pdac", "rodon-2024-amg193-prmt5", "tango-2025-vopimetostat-monotherapy", "clinicaltrials-nct06922591-vopimetostat-ras", "tango-2026-vopimetostat-ras-data", "tango-2026-q2-10q" ], "links": [ { "target": "experiments", "relation": "part-of" }, { "target": "claim-mtap-prmt5-selected-human-signal", "relation": "supported-by" }, { "target": "failure-vopimetostat-ras-contribution-unknown", "relation": "constrained-by" }, { "target": "unknown-mtap-prmt5-ras-contribution", "relation": "tests" }, { "target": "trial-mtap-prmt5-action-map", "relation": "gated-by" }, { "target": "company-tango-therapeutics", "relation": "requires-access-from" }, { "target": "claim-daraxonrasib-kras-amplification-resistance", "relation": "must-measure-alongside" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "strong", "randomised": "none", "independentReplication": "none", "sampleSize": "Uses 54 treated combination patients; future contribution trial unpriced and unstarted", "effect": "Generated a randomized added-effect test with response, PFS, dose-intensity and harm thresholds", "limits": [ "Thresholds are mission decisions", "No controlled interaction estimate", "The route may not prevent known RAS escape", "Future drug access and protocol are unnegotiated" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "Two measurable dependencies are present, but their clinical interaction is unproved." }, "animalEvidence": { "rating": "moderate", "reason": "Combination models support the idea but cannot provide the added human effect." }, "inVitroEvidence": { "rating": "strong", "reason": "The MTAP-MTA-PRMT5 dependency and RAS state are measurable." }, "reproducibility": { "rating": "none", "reason": "No independent randomized combination test exists." }, "commercialConflicts": { "rating": "high", "reason": "The hypothesis depends on two company-controlled agents and interim sponsor data." }, "recency": { "rating": "current", "reason": "The gates use June 2026 combination data." } } }, "layout": { "x": 192, "y": 113 }, "updated": "2026-09-15" }, { "id": "hypothesis-multicompartment-kras-sirna-delivery", "title": "Hypothesis: dual-route KRAS siRNA works only with repeated multi-region knockdown", "shortTitle": "Dual-route KRAS RNA gate", "type": "hypothesis", "status": "hypothesis", "scope": [ "pdac", "colorectal", "pan-cancer", "transfer-unproven" ], "summary": "Local and under-skin delivery can add to chemotherapy only if the same product reduces mutant KRAS across separated tumour regions and does not sacrifice systemic treatment.", "content": "Biological rationale. Mutant KRAS continuously drives PDAC through messenger RNA translated into protein. A matching siRNA can destroy that message after it reaches the cytosol and loads the RISC cutting system. Direct injection raises local concentration; under-skin dosing may reach dispersed disease. The combination is intended to cover two compartments. Supporting evidence. Mutant KRAS is a validated human target, SIL204 reduces RNA and growth in sponsor cell and animal models, and local injection was feasible with the first-generation implant. Contrary evidence. The older implant did not improve survival overall, mutation results existed for only 31/59 people, its local product was undetectable in plasma and no tumour-wide human knockdown result validates transfer to SIL204. The new drug, formulation and dual route had no reported human dose by the public cutoff. Assumptions. The new chemistry can survive circulation, enter malignant rather than only stromal cells, escape the endosome, load RISC without unacceptable innate-immune activation, and reach a sufficient fraction of separated tumour regions. Prediction. Clinical benefit requires repeated reduction of mutant KRAS RNA and protein across at least two primary-tumour regions, active systemic exposure and target reduction in an accessible distant lesion when one can be sampled safely. A single injected-region result or plasma concentration alone will not predict benefit. Falsification inside the existing trial. No new cohort or mission money is needed. Before Segment 2, at least 80% of exposed participants must have usable paired primary samples; at least 70% must show 50% or greater mutant-RNA reduction with concordant KRAS-protein reduction; and at least 70% must agree across two or more regions. Systemic exposure must exceed a locked active range, and every safely sampled distant lesion must be reported. At least 75% must receive 80% or more of planned chemotherapy. Stop for severe EUS or injection harm above 5%, spatial disagreement above 30%, wild-type RAS suppression beyond the safety limit, absent systemic exposure, assay disagreement above 20%, or sponsor refusal to publish all participants. These gates only establish mechanism. Segment 3 must then show overall and quality-adjusted survival against identical chemotherapy under an independently locked analysis. Estimated time: the Segment 1 target report is due before Segment 2; the sponsor expects the first segment within one year of actual dosing. Estimated cost: £0 now. The live sponsor trial owns exposure; the mission offers no capital, licence or endorsement. Any later independent confirmation must replace equal portfolio capital. Main safety risks are pancreatitis, bleeding, infection and tissue injury from repeated EUS injection; inflammatory or complement reactions; unintended wild-type RAS suppression; off-target RNA cutting; and loss of effective chemotherapy through added toxicity or visits.", "sourceIds": [ "alagesan-2026-sig12d-loder", "ctis-2026-sil204-phase23", "silexion-2026-sil204-site-initiation", "silexion-2026-halfyear-filing" ], "links": [ { "target": "experiments", "relation": "part-of" }, { "target": "claim-sil204-trial-scales-before-human-knockdown", "relation": "constrained-by" }, { "target": "failure-sig12d-loder-randomized-os-not-shown", "relation": "repairs" }, { "target": "unknown-sil204-human-target-delivery", "relation": "tests" }, { "target": "trial-sil204-action-map", "relation": "gated-by" }, { "target": "hypothesis-prospective-ras-resistance-routing", "relation": "shares-target-with" } ], "evidence": { "humanProspective": "weak", "humanRetrospective": "none", "randomised": "none for SIL204", "independentReplication": "none", "sampleSize": "Uses the planned 15-21-person Segment 1; adds no new participant", "effect": "Generated multi-region target, systemic exposure, chemotherapy and harm gates", "limits": [ "Thresholds are mission decisions", "Safe distant-lesion sampling may be rare", "Knockdown can occur without tumour death", "Adaptive trial changes can weaken the locked test", "No mission capital or data agreement exists" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "The target and RNA-cutting mechanism are plausible, while compartment coverage is the central unknown." }, "animalEvidence": { "rating": "moderate", "reason": "Local and systemic sponsor models support the predicted compartment test." }, "inVitroEvidence": { "rating": "moderate", "reason": "SIL204 reduces mutant RNA and cell growth after access in model systems." }, "reproducibility": { "rating": "weak", "reason": "No human SIL204 result or independent product reproduction exists." }, "commercialConflicts": { "rating": "high", "reason": "All product-specific evidence is sponsor-controlled and the sponsor plans no participant-level data sharing." }, "recency": { "rating": "current", "reason": "Built around the 2026 authorised trial." } } }, "layout": { "x": 198, "y": 164 }, "updated": "2026-09-15" }, { "id": "hypothesis-multimodal-mrd-action", "title": "Hypothesis: a multimodal MRD state can guide action where ctDNA alone cannot", "shortTitle": "Multimodal MRD action", "type": "hypothesis", "status": "hypothesis", "scope": [ "pdac", "colorectal", "transfer-unproven" ], "summary": "Combine tumour-informed DNA, longitudinal CA19-9, imaging and clinical state, then prove that acting on the combined state improves outcome.", "content": "Biological rationale: postoperative ctDNA is strongly prognostic in PDAC and has randomized treatment-allocation value in stage II colon cancer, but low-shedding PDAC and assay false negatives limit a DNA-only rule. Longitudinal CA19-9 and imaging contain partly different information. Two 2026 PDAC cohorts sharpen the problem: one found postoperative sensitivity 35.7% and specificity 88.9%, while another achieved 75.0% positive and 82.4% negative predictive value after chemotherapy in only 35 evaluable patients. Contrary evidence: adding weak signals can create overfitting and more false positives; earlier detection of recurrence is useless if available treatment cannot change its course. Assumptions: each channel adds stable out-of-sample information, the output corresponds to a treatment decision, and intervention has activity at low burden. Prediction: a locked multimodal state improves sensitivity at fixed 99% specificity versus tumour-informed ctDNA alone and identifies a group in whom early treatment changes recurrence-free survival. Falsification experiment: external blinded validation across two cohorts, followed by a randomized action trial only if sensitivity rises by at least 15 points without specificity falling below 99%. Before opening a new trial, compare the rule and endpoints with DYNAMIC-Pancreas, CIRCPAC and MAP-03; fund a new randomization only for an uncovered action or patient state. Planning estimate: £8m and 32 months for assay/data validation; £35m and five years for utility. Risks are false alarms, lead-time bias, repeated imaging or radiation, premature toxic therapy and duplication of active trials.", "sourceIds": [ "botta-2024-tumor-informed-ctdna", "tie-2022-dynamic-colon", "fahrmann-2021-ca19-9-trajectory", "nakamura-2024-galaxy-ctdna", "guo-2026-postsurgical-ctdna", "chen-2026-personalized-ctdna", "clinicaltrials-nct05788744-circpac", "clinicaltrials-nct05802394-map03" ], "links": [ { "target": "experiments", "relation": "part-of" }, { "target": "unknown-ctdna-actionability", "relation": "tests" }, { "target": "biomarker-ca19-9-longitudinal-anchor", "relation": "extends" }, { "target": "claim-colon-ctdna-has-clinical-utility-proof", "relation": "may-transfer-from" }, { "target": "programme-25m-prospective-observatory", "relation": "tests" }, { "target": "trial-pdac-ctdna-action-map", "relation": "must-not-duplicate" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "strong", "randomised": "moderate", "independentReplication": "moderate", "sampleSize": "Multiple existing cohorts; proposed two-cohort validation then randomized utility trial", "effect": "Gate: +15-point sensitivity at fixed ≥99% specificity; utility requires improved recurrence-free survival", "limits": [ "Generated multimodal hypothesis", "Colon utility does not establish PDAC utility", "PDAC intervention at molecular relapse is unproved", "Model overfitting and site shift", "Detection can advance diagnosis without changing death", "Three active PDAC programmes already test different ctDNA actions" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Residual tumour can release DNA and alter longitudinal markers, while different channels can capture different failure routes." }, "animalEvidence": { "rating": "not-applicable", "reason": "The proposed clinical classifier must be established in longitudinal human cohorts." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Cell models cannot establish screening specificity or treatment utility." }, "reproducibility": { "rating": "weak", "reason": "The prognostic direction repeats, but 2026 cohorts show variable operating performance and the combined PDAC action rule has not been reproduced." }, "commercialConflicts": { "rating": "moderate", "reason": "Commercial ctDNA assays and model ownership create incentives that require locked independent validation." }, "recency": { "rating": "current", "reason": "The hypothesis now incorporates two 2026 PDAC cohorts and active action-trial records." } } }, "layout": { "x": 63, "y": 129 }, "updated": "2026-09-14" }, { "id": "hypothesis-mwa-existing-data-gate", "title": "Recover and align existing microwave records before another intervention", "shortTitle": "MWA existing-data gate", "type": "hypothesis", "status": "testing", "scope": [ "pdac", "unresectable", "trial-recovery", "local-treatment" ], "summary": "A capped independent recovery can show whether microwave coverage, pain and harm are consistent after correcting known selection; new equipment and recruitment remain zero.", "content": "Biological rationale. Microwave energy generates a fast, measurable heat field and may reduce local tumour pain. Supporting evidence. Pain fell in 32/33 symptomatic cases and complete ablation was reported in 30/74 tumours; a 150-person centre record contains a same-site non-microwave group. Contrary evidence. Only 40.5% were completely ablated, treated groups differed sharply at baseline, nearly 30% of microwave imaging was missing and one small surgical package caused a fatal fistula pathway. Assumptions. The four hospitals, Frankfurt centre and NCT06231160 sponsor will export every screened case, device log, scan, systemic treatment, pain measure and vital status with independent publication rights. Prediction. Within nine months, the audit will recover at least 95% vital status and 90% imaging for at least 200 treated and comparison cases, explain complete-versus-partial coverage and leave no credible survival advantage after matching on tumour size, metastases, performance, treatment era and prior response unless microwave has a real effect. Falsification experiment. Enroll zero people, buy no equipment and spend at most £300,000. Freeze a target-trial-style analysis before opening outcomes. Report complete coverage, pain and opioid change, chemotherapy delivery, local and distant failure, all grade 3-plus harms and survival. Stop if fewer than 150 usable cases exist, coverage logs are below 80%, vital status below 95%, imaging below 90%, untreated eligibility cannot be reconstructed or adjusted survival changes by more than 30% across reasonable models. Do not use the audit to claim causality; use it to decide whether any narrow palliative comparison can outperform the already-designed cryo trial. Major risks are heat injury, bleeding, pancreatitis, fistula, treatment delay and false precision from modeled observational data.", "sourceIds": [ "zhou-2024-pancreatic-mwa", "vogl-2026-tace-mwa", "topal-2025-mwa-dual-ici", "clinicaltrials-nct06231160-mwa-cohort" ], "links": [ { "target": "experiments", "relation": "part-of" }, { "target": "unknown-mwa-safe-coverage-and-patient-benefit", "relation": "tests" }, { "target": "failure-mwa-selection-bundling-and-missing-imaging", "relation": "repairs" }, { "target": "trial-pdac-mwa-action-map", "relation": "implemented-by" }, { "target": "programme-5m-decision-core", "relation": "constrained-by" } ], "evidence": { "humanProspective": "weak", "humanRetrospective": "moderate", "randomised": "none", "independentReplication": "weak", "sampleSize": "Zero new participants; recover at least 150 usable cases and target 200; cap £300,000", "effect": "Convert biased scattered records into a stop or narrow palliative-trial decision", "limits": [ "Raw access unconfirmed", "No model removes unknown confounding", "Historic care spans many years", "Device parameters vary", "Pain measures may be absent", "The cap and thresholds are mission choices", "Cryo may dominate the palliative opportunity" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The audit connects delivered heat coverage to human outcomes." }, "animalEvidence": { "rating": "not-applicable", "reason": "Existing human records are the next decision asset." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "No further bench result can resolve clinical selection bias." }, "reproducibility": { "rating": "weak", "reason": "Independent cross-site alignment is the proposed test." }, "commercialConflicts": { "rating": "moderate", "reason": "Device records must be exported in supplier-neutral units." }, "recency": { "rating": "current", "reason": "The gate follows the 2026 reports and current registry state." } } }, "layout": { "x": 892, "y": 816 }, "updated": "2026-09-15" }, { "id": "hypothesis-nrg1-rna-route-completion", "title": "Automatic RNA testing after a KRAS-wild-type result will find rare treatable fusions in time", "shortTitle": "NRG1 RNA route", "type": "hypothesis", "status": "testing", "scope": [ "pdac", "nrg1", "rna-testing", "care-delivery", "capital-allocation" ], "summary": "Use the existing NHS panel route, add explicit NRG1 coverage, and measure every handoff from tissue to treatment; fund no new drug trial.", "content": "Biological rationale. KRAS-wild-type PDAC is enriched for alternative fusion drivers, and an approved NRG1-directed treatment already produces responses. Supporting evidence. KRAS was wild type in 10.7% of 2,483 profiled PDACs; targetable fusions appeared in 19% to 31% of KRAS-wild-type cases across three cohorts; RNA finds NRG1 fusions that DNA can miss. Contrary evidence. NRG1 itself occurs in roughly 0.5% to 1% of PDAC, the treatment evidence is single arm, most treated patients had stopped by the FDA cutoff, and no UK product route was found. Assumptions. NHS genomic laboratories can add explicit NRG1 coverage to an existing pancreatic RNA panel without building a new platform, centres can return the result without delaying chemotherapy, and lawful linked outcome data can follow the route. Prediction. Across 5,000 consecutive advanced PDAC diagnoses, about 5,000 × 0.107 = 535 KRAS-wild-type tumours will trigger RNA testing and roughly 50 NRG1-positive cases will be found if prevalence is 1%. At least 90% of eligible tumours will receive interpretable RNA results within 21 days of pathology, at least 90% of positive results will reach a molecular treatment meeting within seven days, and every fit, eligible positive patient will receive an access decision. Falsification experiment. Spend at most £1.5m over 24 months across at least 20 NHS centres. Use the existing M219 route, add explicit NRG1 and broad fusion coverage, lock a reflex order after a KRAS-wild-type or driver-negative result, and publish counts and time at every handoff. Compare each centre with its prior consecutive year and stagger implementation dates where operationally possible. Do not delay first-line treatment while waiting. Stop or repair the assay if fewer than 15 NRG1-positive cases appear after 5,000 diagnoses, less than 90% of reflex samples yield a result, median pathology-to-RNA time exceeds 21 days, any treatment is delayed, more than 10% of positive cases lack an access decision, or a false fusion call causes material harm. The clinical treatment estimate comes from the overdue FDA follow-up, not this route study. Buy £0 of Genmab, Partner Therapeutics or drug inventory. Major risks are inadequate tissue, false negative or false positive fusion calls, treatment delay, loss of patient fitness, unequal centre access, privacy breach, unlicensed UK supply and mistaking more tests for longer life.", "sourceIds": [ "philip-2022-kras-wildtype-pdac", "fusco-2021-kras-wildtype-fusions", "schram-2026-nrg1-fusion-detection", "nhs-england-2026-cancer-genomic-test-directory", "fda-2024-bizengri-approval", "fda-2026-ongoing-bizengri-accelerated-approval" ], "links": [ { "target": "experiments", "relation": "part-of" }, { "target": "failure-nrg1-testing-and-confirmation-route", "relation": "repairs" }, { "target": "unknown-nrg1-identification-and-survival-benefit", "relation": "tests" }, { "target": "trial-pdac-nrg1-action-map", "relation": "implemented-by" }, { "target": "programme-5m-decision-core", "relation": "funded-by" }, { "target": "hypothesis-pdac-germline-family-route-completion", "relation": "shares-route-with" } ], "evidence": { "humanProspective": "moderate for treatment; none for proposed route", "humanRetrospective": "moderate", "randomised": "none — staggered operational comparison proposed", "independentReplication": "moderate for fusion enrichment; none for whole route", "sampleSize": "Proposed 5,000 consecutive advanced PDAC diagnoses, expected 535 reflex RNA tests and about 50 NRG1-positive cases; cap £1.5m", "effect": "Target at least 90% interpretable RNA completion within 21 days and at least 90% positive-result review within seven further days", "limits": [ "Prevalence estimate uncertain", "Historical comparison", "Rare positive count", "No treatment randomization", "UK access unconfirmed", "Existing local panels vary", "Thresholds are mission choices" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "A fusion-led tumour can only enter the approved treatment route after the fusion is measured." }, "animalEvidence": { "rating": "not-applicable", "reason": "Human response and assay data make route completion the binding test." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "No laboratory model can measure national tissue-to-treatment loss." }, "reproducibility": { "rating": "moderate", "reason": "Twenty centres, consecutive denominators and common timing fields test transfer." }, "commercialConflicts": { "rating": "low", "reason": "Public laboratories and drug-owner-independent route statistics are required." }, "recency": { "rating": "current", "reason": "The design uses the July 2026 NHS directory and September 2026 FDA status." } } }, "layout": { "x": 1228, "y": 1152 }, "updated": "2026-09-15" }, { "id": "hypothesis-older-pdac-fast-support-route", "title": "Hypothesis: a 72-hour vulnerability-to-action route will reduce severe PDAC treatment harm", "shortTitle": "Fast support route", "type": "hypothesis", "status": "supported", "scope": [ "pdac" ], "summary": "A rapid team that turns functional, nutrition, medicine, mood and social findings into completed actions may preserve treatment without using age as a cutoff.", "content": "Biological rationale: vulnerable older adults with metastatic PDAC have little time for a slow referral chain. Supporting evidence: GIANT's median survival was about 4.5 months, GrantPax showed that an assessment label alone did not own the decision, and three randomized mixed-cancer studies reproduce lower severe toxicity when findings trigger concrete care. Evidence against easy success: none showed a PDAC survival gain; two missed patient-valued primary outcomes; very rapid cancer progression may dominate any support; and an intervention can delay needed treatment. Assumptions: clinics can act within 72 hours, the baseline severe-toxicity rate is about 52%, a ten-point reduction matters to patients, support does not materially delay treatment, and contamination between teams remains measurable. Prediction: completing nutrition, medicines, mobility, symptom, cognition, mood, transport and caregiver actions within 72 hours will reduce grade 3-5 treatment toxicity from about 52% to 42%, add at least seven days alive outside hospital over 16 weeks and not materially reduce treatment starts. Design: every participating clinic first records all consecutive adults aged at least 70 with newly diagnosed metastatic PDAC before the final treatment choice, including people who choose supportive care. People who choose systemic treatment and have at least one impaired short-assessment domain are individually randomized within site and planned regimen to the assessment report alone or the same report plus a named nurse, pharmacist, dietitian, physiotherapist and symptom-care clinician who complete the required actions within 72 hours. Treatment choice stays with the patient and oncologist; every start, dose, delay, hospital day, symptom, functional change, stop reason and death is recorded. Planning maths: for severe toxicity falling from 52% to 42%, a two-sided 5% test with 80% power needs about 390 treated people per arm, 780 total. Inflating by 10% for incomplete measurement gives 780 / 0.90 = 867, rounded to 900 treated participants. A six-site 150-person run-in checks recruitment, contamination and action time before expansion to about 30 clinics. An independent statistician must simulate death as a competing outcome and adjust the final size if the run-in rate differs. Release £0.5m for the run-in; release the remaining £2.5m only if at least 90% are assessed before the plan, at least 85% of named actions finish within 72 hours, median treatment delay is no more than three days and required outcomes are at least 90% complete. Passing result: severe toxicity down at least ten points, at least seven more days alive outside hospital by week 16, treatment initiation no more than five points lower, no 90-day death increase above three points and no subgroup completion gap above ten points. Falsification experiment: the randomized comparison disproves the hypothesis if it misses the ten-point toxicity and seven-day home-time gates or meets any safety stop. Stop for delay above seven days, emergency care up five points, death up three points, assessment-driven denial without recorded patient choice, no patient-valued gain or a ten-point inequality widening. Estimated cost and time: £3m over 30 months from the final unassigned part of the existing £15m patient-safety, access and health-economics line; £3m / 900 treated participants is about £3,333 per randomized participant before the observed denominator. This leaves £0 in that line and keeps the mission at £500m. Main risks: delaying treatment, medicalising normal ageing, covertly denying treatment, forcing treatment, missing cognitive consent needs, caregiver burden, staff shortage and contamination between arms. This experiment tests safer route delivery, not cure or the best tumour drug. Planning rank: impact 3.5/5 × technical probability 0.55 × speed 0.67 per year / £3m = 0.43.", "sourceIds": [ "dotan-2024-giant", "dotan-2025-giant-ga", "betge-2022-grantpax", "mohile-2021-gap70", "li-2021-gain", "ihorst-2026-geriatric-comanagement" ], "links": [ { "target": "experiments", "relation": "part-of" }, { "target": "claim-age-alone-cannot-route-pdac-treatment", "relation": "derived-from" }, { "target": "failure-grantpax-geriatric-route", "relation": "escapes" }, { "target": "unknown-older-pdac-treatment-route", "relation": "tests" }, { "target": "trial-older-pdac-action-map", "relation": "bounded-by" }, { "target": "programme-500m-integrated-mission", "relation": "funded-by" } ], "evidence": { "humanProspective": "strong rationale", "humanRetrospective": "not-applicable", "randomised": "moderate — generated randomized experiment", "independentReplication": "moderate for toxicity mechanism; absent for PDAC route", "sampleSize": "Run-in n=150; planned randomized treated n=900 across about 30 clinics", "effect": "Designed for severe toxicity 52% to 42% plus a seven-day gain alive outside hospital", "limits": [ "Generated design", "Planning assumptions require simulation", "No direct PDAC support effect", "Cannot select the best drug", "Post-assessment treatment choice", "Care-team contamination" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The intervention acts on measured causes of treatment harm and loss rather than age itself." }, "animalEvidence": { "rating": "not-applicable", "reason": "No animal model can validate this care route." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "No cell system can reproduce function, consent or hospital burden." }, "reproducibility": { "rating": "moderate", "reason": "Several randomized mixed-cancer trials support the toxicity effect; the PDAC-specific size and useful-life effect are untested." }, "commercialConflicts": { "rating": "low", "reason": "The intervention is drug-neutral, uses established care roles and requires independent statistics." }, "recency": { "rating": "current", "reason": "The design incorporates 2026 trial results and current registry checks." } } }, "updated": "2026-09-14" }, { "id": "hypothesis-one-time-risk-triggered-workup", "title": "Hypothesis: scalable warning signs plus one-time work-up may beat repeated population screening", "shortTitle": "Risk-triggered work-up", "type": "hypothesis", "status": "hypothesis", "scope": [ "pdac" ], "summary": "A temporary warning state may make one-time work-up viable, but direct imaging creates incidental burden and the second-stage selector has not been validated.", "content": "Rationale: annual population screening suffers extreme base-rate penalties, while a temporary warning sign can raise current cancer probability before a costly test. The 2026 heuriskance framework requires sufficient cancer prevalence, useful lead time and scalable population identification. Glycaemically defined new-onset diabetes is the worked example. It raises rough three-year risk about 14.9-fold, from 3 × 13.9 = 41.7 to 620 cancers per 100,000, but direct imaging is not a solved action. In a 93-person CT pilot, 52 people had extrapancreatic incidental findings and the only PDAC was stage IV. PANDOME found one stage-I PDAC among 109 MRI participants, in deteriorating rather than new-onset diabetes, after four biopsies. EDI has enrolled 8,869 and will measure stage and incidental findings, but it assigns intervention and observation by site, uses ENDPAC above 0 and has no posted result. Workload model: if the original END-PAC sensitivity of 78% and specificity of 85% transported to 100,000 people at 0.62% risk, about 484 cancer-bearing and 14,907 unaffected people would test positive, PPV 3.14%. A second test with 90% sensitivity and 99% specificity inside that selected group would model to about 435 cancer-bearing and 149 unaffected positives, PPV 74.5%; no current test has shown that conditional performance and correlated errors can defeat it. Revised prediction: a locked health-record trigger plus an independently validated second-stage test, followed by one pancreas-imaging episode, will increase aggressive stage-I/II detection without excess serious procedures versus a randomized or cluster-randomized control. Assumptions: reliable trigger capture, a real lead-time window, conditional test performance in the selected population, imaging of treatable disease and effective treatment. Falsification: no stage shift after site and ascertainment adjustment; interval cancers erase sensitivity; incidental or invasive harm produces negative net benefit; or earlier diagnosis does not reduce pancreatic-cancer death. Spend first on EDI data access, cross-site repeat and the second-stage test. Do not launch a duplicate 40,000–60,000-person utility trial until EDI reports or a distinct action gap is proved. Major safety risks are contrast injury, radiation, EUS or biopsy complications, anxiety, incidental procedures and pancreatic surgery for lesions that would not have killed the patient.", "contentSections": [ "Supporting evidence: new-onset diabetes creates a measurable temporary risk state, and PANDOME shows that early-stage detection can occur.", "Evidence against: the CT pilot found substantial incidental abnormalities and only stage-IV disease; EDI remains unfinished and non-randomized by site." ], "sourceIds": [ "chari-2026-heuriskance", "chari-2026-prospective-new-onset-diabetes", "sharma-2018-endpac", "wu-2022-nod-imaging-pilot", "frank-2025-pandome", "edi-nct04662879", "seer-pancreas-statfacts" ], "links": [ { "target": "claim-new-onset-diabetes-enriches-risk", "relation": "depends-on" }, { "target": "claim-screening-base-rate-maths", "relation": "succeeds-where" }, { "target": "experiments", "relation": "tests" }, { "target": "programme", "relation": "supports" }, { "target": "trial-new-onset-diabetes-workup-map", "relation": "constrained-by" } ], "evidence": { "humanProspective": "weak", "humanRetrospective": "moderate", "randomised": "none", "independentReplication": "weak", "sampleSize": "Framework; END-PAC validation n=1,096 with nine cancers; CT pilot n=93; PANDOME n=109; EDI n=8,869 with results pending", "effect": "Predicted stage shift and net benefit from a two-stage selector; not yet observed", "limits": [ "Generated programme hypothesis", "No completed utility trial", "Modelled workload assumes transported performance", "No second-stage test has shown the stated conditional performance", "Lead-time and overdiagnosis bias", "Risk trigger misses cancers without diabetes", "Incidental imaging findings can cause harm" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Some occult pancreatic cancers cause measurable metabolic change months before diagnosis." }, "animalEvidence": { "rating": "not-applicable", "reason": "The hypothesis concerns a human detection and work-up pathway rather than a mechanism requiring animal proof." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Cell or assay work cannot establish stage shift, work-up burden or survival benefit." }, "reproducibility": { "rating": "weak", "reason": "Risk enrichment repeats, but the full trigger-to-second-test-to-imaging action loop has not been reproduced." }, "commercialConflicts": { "rating": "moderate", "reason": "The current studies are mainly nonprofit-led, while any selected blood or imaging product can create sponsor influence in a large trial." }, "recency": { "rating": "current", "reason": "The hypothesis incorporates 2025 and 2026 prospective reports and live EDI status." } } }, "layout": { "x": 43, "y": 119 }, "updated": "2026-09-14" }, { "id": "hypothesis-optune-pax-independent-value-gate", "title": "Keep the proven narrow use while forcing replication, route measurement and a price ceiling", "shortTitle": "Optune Pax value gate", "type": "hypothesis", "status": "testing", "scope": [ "pdac", "locally-advanced", "post-approval", "value" ], "summary": "Independent analysis can protect a real two-month survival result from both dismissal and commercial overreach, without subsidising another company-run cohort.", "content": "Biological rationale. Alternating electric fields can disturb dividing cells throughout the abdominal field without adding a systemic drug, so a survival effect is possible even when standard scan rules are insensitive to the relevant damage. Supporting evidence. PANOVA-3 met its locked overall-survival test in all 571 assigned people: hazard ratio 0.82, 95% CI 0.68-0.99, and 7.9 percentage points more estimated alive at one year. FDA reviewed the full record and approved the narrow use. Contrary evidence. Progression-free survival missed its test, local progression-free survival and response did not establish benefit, surgery was numerically lower, serious adverse events were 5.6 points higher and there is no independent randomized reproduction. Median use was 11.2 hours per day, skin toxicity affected most users and two cost models put current value far outside their chosen thresholds. Assumptions. Participant-level PANOVA-3 data and dated plans can be independently accessed; patients can report burden through progression; a company with $440.6m in cash and short investments can fund replication; and price can be tied to measured benefit. Prediction. An independent all-randomized analysis will preserve an overall-survival hazard ratio at most 0.85 with its 95% interval below 1.00, at least five points more alive at one year, no more than ten points excess grade 3 or worse harm, and no important quality-of-life loss when follow-up continues after progression. A complete care route will show how many eligible people are offered the device, accept, start, maintain at least 12 hours per day, stop for harm and remain on treatment. A company-funded randomized trial on a current backbone will reproduce survival before any metastatic or platform-wide claim. Falsification experiment. Spend at most £500,000 from existing data, statistics and care-delivery lines, with £0 for product and £0 for new company trial participants. Reconcile every PANOVA-3 assignment, outcome, withdrawal, later treatment and device-use trace. Treat usage as behaviour after randomization, not as proof that more hours caused benefit. Collect quality of life, pain, skin injury, hospital-free days, caregiver time and survival through progression in a prospective all-eligible care route. Stop mission procurement if the independent survival gate fails, route completion is below 60%, important quality of life worsens, severe harm exceeds the gate, or all-in added cost exceeds £50,000 times independently verified added quality-adjusted life-years. Block every use outside the approved locally advanced gemcitabine and nab-paclitaxel setting until a randomized trial with the matching population and backbone passes the same gate. Major risks are skin breakdown and infection, treatment interruption, device burden, caregiver burden, false causal claims from usage, loss after progression, unaffordable care and diversion of money from treatments with larger effects.", "sourceIds": [ "macarulla-2025-panova3", "clinicaltrials-nct03377491-panova3", "fda-2026-optune-pax-ssed", "clinicaltrials-nct06390059-panova4", "novocure-2026-panova4-topline", "novocure-2026-q2-10q", "guzauskas-2026-ttfields-cost", "liu-2026-ttfields-cost" ], "links": [ { "target": "experiments", "relation": "part-of" }, { "target": "unknown-ttfields-survival-replication-and-access", "relation": "tests" }, { "target": "failure-panova4-component-and-historical-control", "relation": "repairs" }, { "target": "trial-pdac-ttfields-action-map", "relation": "implemented-by" }, { "target": "company-novocure", "relation": "requires-from" }, { "target": "programme-500m-integrated-mission", "relation": "constrained-by" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong", "independentReplication": "none", "sampleSize": "Existing PANOVA-3 n=571; zero new mission-funded trial participants; independent work capped at £500,000", "effect": "Preserve evidence-based narrow access while blocking unsupported expansion and excessive price", "limits": [ "Thresholds are mission choices", "Participant-level access is not secured", "Care-route observation cannot reproduce causal efficacy", "A modern-backbone trial may take years", "Quality-adjusted life-year estimates are model-dependent", "Price negotiation power is unknown" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "A physical cell-division effect is plausible, but the survival and progression mismatch needs direct analysis." }, "animalEvidence": { "rating": "moderate", "reason": "Model work supports field selection and harm planning, not the procurement threshold." }, "inVitroEvidence": { "rating": "strong", "reason": "Electric-field effects on dividing cells support the testable route." }, "reproducibility": { "rating": "none", "reason": "The hypothesis exists because the pancreatic survival result has not been independently reproduced." }, "commercialConflicts": { "rating": "high", "reason": "Independent access, analysis and negative-result rights are required because Novocure owns the device and pivotal record." }, "recency": { "rating": "current", "reason": "The design responds to the 2026 approval, commercial launch and two 2026 value models." } } }, "layout": { "x": 604, "y": 528 }, "updated": "2026-09-15" }, { "id": "hypothesis-organoid-guided-early-switch", "title": "Hypothesis: a fast organoid result can improve initial PDAC treatment choice", "shortTitle": "Organoid treatment choice", "type": "hypothesis", "status": "hypothesis", "scope": [ "pdac", "cholangiocarcinoma", "colorectal", "transfer-unproven" ], "summary": "Test the whole route from biopsy to useful life; a high score among successful cultures is not enough.", "content": "Biological rationale and supporting evidence: living tumour cultures can expose response differences between otherwise reasonable drugs, and several prospective PDAC series find correlation with patient response. Contrary evidence: usable yield has ranged from 15.8% of all HOPE enrollees to 62% in ORGANOPREDICT, median or mean reporting time has ranged from 53 to 96 days, and PASS-01 shows that waiting until second line leaves little time. AVATAR delivered matched treatment to only 4.9% of people assigned to its precision route and did not improve survival. Assumptions: a tumour-confirmed culture can retain its signal in a 14-day workflow; two currently accessible regimens have meaningfully different predicted effects; a result from one sampled site represents the lethal disease well enough; and teams will follow the assigned policy. The engineering reliability is multiplicative. Even 0.80 culture success × 0.90 correct identity × 0.90 on-time return × 0.90 drug access × 0.90 delivery = 0.525, so only about half of enrolled people would receive the intended output. Prediction: a locked whole-route assay can exceed every gate and an organoid-guided policy can reduce the death hazard by 25% without worse quality-adjusted survival. Falsification experiment, stage 1: recover patient-level results from NCT04931381, NCT04931394 and NCT05351983; test the treatment-by-assay interaction in PASS-01; then run shadow testing in 200 consecutive advanced-PDAC patients inside the existing observatory, with 50 split samples tested by two laboratories. No care changes in this stage. Go only if at least 80% of all eligible patients receive a tumour-confirmed result within 14 days, the two laboratories give the same top-two regimen order in at least 80%, at least 90% of recommended regimens are actually available, route completion differs by no more than ten percentage points across deprivation and ethnicity groups, and a prespecified treatment interaction predicts response beyond clinical state and tumour subtype. Retire patient selection if timely yield is below 70%, laboratory disagreement exceeds 20%, no treatment interaction appears or failed cultures carry different outcomes. If those gates pass and the two Chinese randomized studies cannot answer the policy, randomize 480 people from biopsy to organoid-guided first-line choice or standard choice; late and failed assays default to standard care and remain in their assigned group. Overall survival from allocation is primary and quality-adjusted survival is joint. Planning maths: HR 0.75 requires 4 × (1.96 + 0.84)^2 / ln(0.75)^2 = 379 deaths; 379 / 0.90 events / 0.90 retained = 468, rounded to 480. The shadow stage uses up to £2.5m already assigned to organoid work in the £25m observatory over 24 months. The four-year utility trial is an unfunded £18m planning candidate and can start only if an explicit review removes at least £18m from the fixed portfolio. Major risks include biopsy harm, wrong-clone growth, normal-cell contamination, assay drift, withholding an effective standard regimen, unavailable drugs, unequal access and a control that changes during the trial.", "sourceIds": [ "tiriac-2018-organoid-profiling", "beutel-2021-pdac-organoids", "grossman-2022-hope-organoids", "boileve-2024-organopredict", "wiessner-2024-organoid-biopsy-randomized", "knox-2025-pass01", "sarno-2025-avatar", "clinicaltrials-nct04931381-organoid-advanced", "clinicaltrials-nct04931394-organoid-adjuvant", "clinicaltrials-nct05351983-organoid-screen", "clinicaltrials-nct05927298-prosper-panc", "clinicaltrials-nct06813079-adopt", "ettrich-2026-unitepanc" ], "links": [ { "target": "experiments", "relation": "part-of" }, { "target": "person-david-tuveson", "relation": "extends" }, { "target": "treatment", "relation": "tests" }, { "target": "programme-5m-decision-core", "relation": "tests" }, { "target": "programme-25m-prospective-observatory", "relation": "tests" }, { "target": "programme-500m-integrated-mission", "relation": "competes-for-capital-in" }, { "target": "claim-pdac-organoid-correlation-not-routing-utility", "relation": "depends-on" }, { "target": "failure-organoid-treatment-chain-attrition", "relation": "repairs" }, { "target": "unknown-pdac-organoid-guided-treatment-utility", "relation": "tests" }, { "target": "trial-pdac-organoid-action-map", "relation": "gated-by" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "moderate", "randomised": "weak — existing randomized PDAC studies do not yet establish organoid-guidance utility", "independentReplication": "moderate", "sampleSize": "Proposed shadow stage n=200 with 50 split samples; conditional randomized utility n=480", "effect": "Target death HR 0.75 with no worse quality-adjusted survival, after ≥80% timely yield and ≥80% inter-laboratory regimen-order agreement", "limits": [ "Generated clinical-utility hypothesis", "Existing correlations condition on successful cultures", "Current turnaround is too slow for first-line choice", "Host and metastatic context are incomplete", "Two stale randomized trials may already contain relevant data", "£18m utility stage is not funded" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "A living tumour sample can reveal drug response that sequence and clinical state do not fully encode." }, "animalEvidence": { "rating": "moderate", "reason": "Matched models support response fidelity but cannot prove whole-route benefit." }, "inVitroEvidence": { "rating": "strong", "reason": "Independent PDAC organoid programmes repeatedly observe drug-response variation." }, "reproducibility": { "rating": "weak", "reason": "The proposed fast, cross-laboratory and treatment-interaction rule has not passed." }, "commercialConflicts": { "rating": "moderate", "reason": "Assay providers and drug-screening laboratories require independent protocol, data and analysis control." }, "recency": { "rating": "current", "reason": "Design incorporates published and registered evidence checked through September 2026." } } }, "layout": { "x": 70, "y": 128 }, "updated": "2026-09-15" }, { "id": "hypothesis-particle-therapy-existing-data-gate", "title": "Make delivered dose and systemic risk pass a gate before a particle trial", "shortTitle": "Particle data and selector gate", "type": "hypothesis", "status": "testing", "scope": [ "pdac", "locally-advanced", "proton", "carbon-ion", "data-recovery" ], "summary": "Complete existing cohorts can test whether particle dose is delivered robustly and whether a local-dominant group exists before a costly randomized trial.", "content": "Biological rationale. Protons and carbon ions can concentrate dose in a deep target while sparing normal tissue, and carbon ions may kill cells with damage that is less dependent on oxygen. Supporting evidence. Carbon-ion cohorts repeatedly report high local control, and proton cohorts report low severe treatment toxicity. Contrary evidence. No comparative randomized participant exists. Distant spread remains common, the proton phase 2 missed its survival target, CIPHER enrolled zero, and completed PACK has no result. Assumptions. Complete planning scans, daily anatomy, delivered-dose estimates, chemotherapy, progression, harm, patient reports and survival can be recovered from PACK, PAN009-18 and participating carbon centres; particle and photon dose can be recalculated under common rules; and a local-dominant risk rule can be locked without using later outcomes. Prediction. A shared existing-data audit will reconstruct at least 95% of enrolled participants, at least 90% of planned fractions and at least 80% of serial anatomy needed to estimate delivered rather than planned dose. A particle advantage worth testing will show at least five percentage points less grade 3 or worse gastrointestinal harm or a clinically material gain in chemotherapy delivery, hospital-free time or quality of life, while a locked systemic-risk rule identifies a group with two-year distant-failure risk below 40%. Falsification experiment. Spend at most £500,000 from existing data and radiation-quality budgets and enroll nobody. Recalculate target and bowel dose under common proton, carbon and photon models; quantify range failure from breathing and gas; reconcile every PACK and proton participant; and publish complete failure patterns. Stop if data recovery is below the gates, dose comparisons do not transfer across centres, a planned-dose advantage disappears under daily anatomy, severe harm is not lower, systemic risk stays high or any apparent survival difference depends on post-treatment selection. Only a passing audit can design a roughly 500-person randomized current-care trial; about 379 deaths are required to detect a survival hazard ratio of 0.75 with two-sided 5% error and 80% power under equal allocation. That future trial receives £0 now and must compare identical systemic treatment, eligibility, follow-up and supportive care. Major safety risks are stomach or bowel ulcer, bleeding or perforation, missed target from range error, treatment delay, travel burden and loss of effective chemotherapy.", "sourceIds": [ "shinoto-2016-carbon-ion-gemcitabine", "kawashiro-2018-jcros-carbon-ion", "lai-2026-carbon-ion-risk-adapted", "rapp-2022-proton-phase2", "eckstein-2023-proton-registry", "clinicaltrials-nct02598349-proton", "clinicaltrials-nct03536182-cipher", "clinicaltrials-nct04194268-pack" ], "links": [ { "target": "experiments", "relation": "part-of" }, { "target": "unknown-particle-therapy-added-patient-benefit", "relation": "tests" }, { "target": "failure-particle-therapy-comparator-and-accrual", "relation": "repairs" }, { "target": "trial-pdac-particle-therapy-action-map", "relation": "implemented-by" }, { "target": "hypothesis-selected-local-consolidation", "relation": "uses" }, { "target": "programme-5m-decision-core", "relation": "constrained-by" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "moderate", "randomised": "none", "independentReplication": "moderate", "sampleSize": "Existing completed or active cohorts; zero new participants; audit capped at £500,000; possible future trial roughly 500 people", "effect": "Reject fragile beam plans and systemic-risk groups before randomized capital", "limits": [ "Data access is not secured", "Delivered-dose reconstruction is model-dependent", "Common cross-centre calibration may fail", "Existing cohorts have treatment and selection differences", "A risk rule cannot prove treatment effect", "Power calculation is approximate", "Budget and thresholds are mission choices" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The experiment measures both the physical beam advantage and the systemic condition needed for that advantage to matter." }, "animalEvidence": { "rating": "moderate", "reason": "Model evidence supports local effect but cannot replace delivered human dose." }, "inVitroEvidence": { "rating": "strong", "reason": "Particle damage mechanisms justify a measured clinical test." }, "reproducibility": { "rating": "weak", "reason": "The proposed cross-centre audit is needed because current dose and outcome records are not comparable." }, "commercialConflicts": { "rating": "moderate", "reason": "Independent recalculation and public negative-result rights are required because centres own the infrastructure." }, "recency": { "rating": "current", "reason": "The design responds to completed 2025-2026 records and current recruitment." } } }, "layout": { "x": 664, "y": 588 }, "updated": "2026-09-15" }, { "id": "hypothesis-pdac-biliary-route-completion", "title": "A 48-hour indication-to-treatment biliary route will cut neoadjuvant interruption without adding procedure harm", "shortTitle": "Biliary route completion", "type": "hypothesis", "status": "hypothesis", "scope": [ "pdac" ], "summary": "Cluster-randomize a complete route because component device efficacy is known and the remaining loss is selection, handoff, rescue and follow-up.", "content": "Biological rationale and supporting evidence: obstruction and infection can block chemotherapy or surgery, while unnecessary ERCP adds its own harm. Randomized evidence rejects routine delayed plastic drainage, supports durable metal stents when drainage must span neoadjuvant treatment, and shows EUS-guided rescue can avoid percutaneous drainage in expert centres. A selected surgical cohort recorded stent-failure treatment interruption in 26/162, 16%. Evidence against: the survival comparison was retrospective; a 2026 metal-stent trial met only a wide 15-point noninferiority margin; EUS trials mix cancers and expert centres; and better patency may not increase treatment delivered. Assumptions: baseline interruption is 16%, the route can lower it to 10%, about 30 centres can enroll consecutive jaundiced PDAC patients, a 1.3 cluster penalty is adequate, and surgery capacity can absorb faster referrals. Prediction and planning maths: for 16% versus 10%, a simple two-proportion calculation at 80% power needs about 492 per arm, 983 total; 983 × 1.3 = 1,278, rounded to 1,300. The route will document indication and plan within 48 hours in at least 90%, use metal stents in at least 90% of patients whose drainage must span neoadjuvant treatment, and reduce cholangitis or obstruction causing treatment interruption by at least six points. Falsification experiment: randomize roughly 30 centres in a stepped sequence to a shared triage, device, rescue and follow-up route versus current care for about 1,300 patients. Keep prompt-surgery, neoadjuvant and palliative states separate. Record bilirubin, symptoms, fever, cultures, route and device, complications, reintervention, every treatment date and dose, surgery and 30-day death. Stop for serious procedure harm up five points, 30-day death imbalance, surgery delayed more than seven additional days, no six-point interruption reduction, or no gain in planned treatment delivered. Estimated cost and time: £4m over three years from the £12m remaining patient-safety, access and health-economics line, leaving £8m. Major safety risks include cholangitis, pancreatitis, bleeding, perforation, stent migration, cholecystitis, anaesthesia, delayed surgery and failed rescue.", "sourceIds": [ "van-der-gaag-2010-pbd-rct", "costamagna-2026-sems-early-surgery-rct", "tamura-2021-fcsems-plastic-rct", "seo-2019-covered-uncovered-sems-rct", "lyu-2023-metal-plastic-nat-meta", "vehvilainen-2024-cholangitis-nat-interruption", "chen-2023-element-biliary-rct", "barbosa-2024-eus-ercp-meta", "esge-2018-biliary-stenting-guideline", "asge-2024-therapeutic-eus-guideline" ], "links": [ { "target": "experiments", "relation": "part-of" }, { "target": "claim-biliary-drainage-is-route-not-cancer-treatment", "relation": "builds-on" }, { "target": "failure-routine-plastic-pbd", "relation": "avoids" }, { "target": "failure-biliary-stent-treatment-interruption", "relation": "repairs" }, { "target": "unknown-pdac-biliary-drainage-route", "relation": "tests" }, { "target": "trial-pdac-biliary-drainage-action-map", "relation": "prioritized-by" }, { "target": "programme-500m-integrated-mission", "relation": "funded-by" } ], "evidence": { "humanProspective": "strong for components; weak for implementation", "humanRetrospective": "moderate", "randomised": "strong for components; none for the whole route", "independentReplication": "none for the whole route", "sampleSize": "Proposed cluster-randomized n about 1,300 across about 30 centres", "effect": "Generated target: treatment interruption 16% to 10% without serious procedure harm up five points", "limits": [ "Planning event rate", "Provisional cluster penalty", "Open-label route", "Mixed indications", "Local expertise", "Rare fatal events" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The route removes avoidable procedures and protects necessary drainage through cancer treatment." }, "animalEvidence": { "rating": "not-applicable", "reason": "This is a human care-delivery hypothesis." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Laboratory work cannot test triage, handoff, rescue or treatment dates." }, "reproducibility": { "rating": "none", "reason": "The complete route has not been randomized or independently repeated." }, "commercialConflicts": { "rating": "moderate", "reason": "The route uses commercial devices but should be product-neutral and independently analysed." }, "recency": { "rating": "current", "reason": "The design responds to randomized evidence through 2026." } } }, "updated": "2026-09-14" }, { "id": "hypothesis-pdac-exercise-existing-data-gate", "title": "Existing exercise trials can identify whether a low-function pancreatic group has patient-level benefit", "shortTitle": "Exercise existing-data gate", "type": "hypothesis", "status": "hypothesis", "scope": [ "pdac", "supportive-care", "surgery", "exercise", "capital-allocation" ], "summary": "Recover and harmonize the randomized records before funding another broad programme, then reuse existing care routes if a selector repeats.", "content": "Biological rationale. Strength and aerobic reserve can support independence, treatment tolerance and recovery, especially when baseline reserve is low but trainable. Supporting evidence. Randomized studies show that steps, strength sessions and some short quality measures can change safely. Evidence against. APACaP and PancFit missed their main between-group outcomes, surgical trials did not reduce major complications, and missingness removed many assigned patients. Assumptions. At least four trial teams can provide participant-level records, common function and care outcomes can be aligned, and baseline state is measured before allocation. Prediction. A locked low-function rule will show at least five extra quality-weighted days alive and at home over 16 weeks or at least a ten-point increase in receiving 80% of planned chemotherapy, without more falls, admissions, operations delayed or early deaths, and the direction will repeat in two independent trials. Falsification experiment. Spend at most £250,000 to recover at least 80% of assignments from APACaP, PancFit and three surgical studies; align baseline function, received activity, nutrition, treatment, complications, admissions, death and questionnaires; analyse every assignment with death as an outcome rather than missing. Estimated time is 12 months. Stop if fewer than 80% of assignments are recoverable, the selector is derived and tested in the same people, interaction direction fails to repeat, absolute gain is below the threshold, or burden and harm erase benefit. If the gate passes, embed the locked route inside the already funded palliative or surgical-delivery programme; added trial capital remains £0. Major safety risks include falls, cardiac events, pain, treatment delay, exhaustion, excluding the sickest patients and replacing nutrition or symptom care with a step target.", "sourceIds": [ "neuzillet-2023-apacap-rct", "ngohuang-2023-pancfit-rct", "ausania-2019-pancreatic-prehab-rct", "steindorf-2019-pancreatic-exercise-rct", "zarate-rodriguez-2023-pancreatectomy-steps-rct", "mizuno-2024-hpb-prehab-rct", "li-2024-pancreatic-early-mobilization-rct", "clinicaltrials-nct07470268-hypanc" ], "links": [ { "target": "experiments", "relation": "part-of" }, { "target": "unknown-pdac-exercise-net-benefit-and-selector", "relation": "tests" }, { "target": "failure-pdac-exercise-endpoint-and-attrition", "relation": "succeeds-where" }, { "target": "trial-pdac-exercise-action-map", "relation": "implemented-by" }, { "target": "hypothesis-pdac-palliative-route-completion", "relation": "coordinates-with" }, { "target": "hypothesis-pdac-surgical-route-completion", "relation": "coordinates-with" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "moderate", "randomised": "strong — the gate reuses randomized records", "independentReplication": "weak — the proposed selector has not been tested across trials", "sampleSize": "At least 80% of assignments from five randomized studies; data-recovery cap £250,000", "effect": "Falsifiable existing-data selector gate with no new participant enrollment", "limits": [ "Data sharing", "Outcome alignment", "Different stages", "Missing baseline fields", "Interaction power", "No new causal contrast" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "A trainable low-reserve state is more specific than an all-patient exercise rule." }, "animalEvidence": { "rating": "not-applicable", "reason": "The hypothesis concerns human function, adherence and useful days." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Patient-level route and outcome data are required." }, "reproducibility": { "rating": "weak", "reason": "No locked low-function interaction has repeated across independent trials." }, "commercialConflicts": { "rating": "low", "reason": "Funding buys independent data recovery rather than an exercise product." }, "recency": { "rating": "current", "reason": "The gate includes recent randomized surgery trials and the live 2026 protocol." } } }, "layout": { "x": 2440, "y": 1660 }, "updated": "2026-09-15" }, { "id": "hypothesis-pdac-germline-family-route-completion", "title": "Automatic testing plus consented family contact will turn more inherited findings into action", "shortTitle": "Germline route completion", "type": "hypothesis", "status": "hypothesis", "scope": [ "pdac" ], "summary": "Test an English diagnosis-to-family route that measures eligibility expansion, result timing and relative uptake in one linked programme.", "content": "Biological rationale and supporting evidence: family-history criteria miss some carriers; point-of-care testing can exceed 80% completion; and remote home testing can reach 90% after a relative has entered a study. The unresolved losses are the all-patient offer, timely interpretation, permission to reach relatives and invitation. Evidence against easy success: England currently commissions a narrower R367 route; US academic completion varies from 38% to 82%; informed-relative testing was 31% in one prospective cohort; GIFT found only 19.2% of reported relatives invited and no clear benefit from a navigator alone. Falsification experiment and design: enroll 3,000 consecutive newly diagnosed PDAC patients across about 20 NHS centres. Offer a public-laboratory multigene test at the oncology visit with translated consent and return results within 21 days. Record whether each patient met R367, each pathogenic finding, every treatment or risk-management action, uncertain findings, cost and every eligible adult first-degree relative. For pathogenic-result families, randomize by family to the standard counselling letter and free test link or, with the patient's explicit permission, direct multilingual contact, a named navigator and a free home kit. Assumptions: pathogenic-variant yield is 8%, each carrier has two eligible adult first-degree relatives on average, 70% consent and can be reached, and a 1.3 family-clustering allowance is adequate. Prediction: the complete route will return at least 85% of patient results within 21 days and raise testing among eligible reached relatives from 31% to at least 50%. Planning maths: using an 8% pathogenic-variant yield, two eligible adult relatives per carrier and 70% consent and reach gives 3,000 × 0.08 × 2 × 0.70 = 336 relatives. Detecting 31% versus 50% testing at two-sided 5% and 80% power needs about 104 relatives per arm, 208 total; 208 × 1.3 for family clustering = 271, leaving a 65-relative recruitment buffer. Passing gates: at least 90% offered, 85% tested and resulted within 21 days, 90% of pathogenic results discussed by genetics within 14 days, and at least 50% of eligible reached relatives tested within six months without a greater than ten-point completion gap by ethnicity, language, deprivation or centre. The policy gate is positive incremental net benefit outside R367 after gene-specific actions and counselling cost are counted. Stop or amend for any material privacy breach, care based on a variant of uncertain significance, median turnaround above 21 days, distress or discrimination harm, a greater than ten-point inequality gap, relative testing below 40%, or no positive net benefit outside R367. Do not claim pancreatic-cancer prevention from testing alone; follow actual surveillance, proven prevention for other syndrome cancers and long-term diagnoses separately. Estimated cost and time: £2.5m over three years from the £8m remaining patient-safety, access and health-economics line, leaving £5.5m. Major safety risks: privacy breach, coercive family contact, distress, insurance or employment discrimination, false reassurance, action on an uncertain variant, delayed cancer treatment and unequal access.", "sourceIds": [ "nhs-england-2026-r367", "asco-2024-germline-panel-guideline", "llach-2024-universal-germline", "chittenden-2021-systematic-germline", "walker-2021-embedded-germline-station", "wang-2022-oncology-germline-cascade", "ramsey-2023-mainstreaming-germline", "rodriguez-2024-generate", "mcdonald-2024-germline-protocol", "sinan-2025-pdac-germline-implementation", "grewal-2026-pancan-germline-survey", "katz-2026-gift-cascade", "aguirre-2026-ignite-tx", "chen-2023-fdr-cancer-risk" ], "links": [ { "target": "experiments", "relation": "part-of" }, { "target": "claim-universal-germline-testing-needs-action", "relation": "builds-on" }, { "target": "claim-germline-risk-subset", "relation": "builds-on" }, { "target": "failure-pdac-germline-route-dropout", "relation": "repairs" }, { "target": "system-england-r367-germline-criteria", "relation": "tests" }, { "target": "unknown-pdac-germline-family-action-route", "relation": "tests" }, { "target": "trial-pdac-germline-cascade-action-map", "relation": "prioritized-by" }, { "target": "claim-high-risk-surveillance", "relation": "routes-to" }, { "target": "programme-500m-integrated-mission", "relation": "funded-by" } ], "evidence": { "humanProspective": "moderate for components; none for whole route", "humanRetrospective": "strong for delivery loss", "randomised": "moderate for family components; none for English whole route", "independentReplication": "none for the proposed route", "sampleSize": "Proposed n=3,000 PDAC patients and about 336 reachable adult relatives across about 20 centres", "effect": "Generated targets: 85% timely patient testing and relative testing 31% to at least 50%", "limits": [ "Planning variant yield", "Planning relatives per carrier", "Consent and reach assumptions", "Family clustering estimate", "Gene-specific action heterogeneity", "Long follow-up for cancer outcomes" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Finding a familial pathogenic variant can change selected treatment and proven risk management for blood relatives." }, "animalEvidence": { "rating": "not-applicable", "reason": "This is a human testing, consent and care-delivery hypothesis." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Laboratory classification is necessary but does not test route completion." }, "reproducibility": { "rating": "none", "reason": "No English programme has tested all linked steps and the R367 policy boundary together." }, "commercialConflicts": { "rating": "low", "reason": "Public laboratories, independent statistics and provider-neutral procurement are built into the design." }, "recency": { "rating": "current", "reason": "The design incorporates evidence and policy through 2026." } } }, "updated": "2026-09-14" }, { "id": "hypothesis-pdac-nerve-ablation-existing-trial-safety-gate", "title": "Make the live early-neurolysis trial answer pain and survival together", "shortTitle": "Early-neurolysis safety gate", "type": "hypothesis", "status": "testing", "scope": [ "pdac", "pain", "palliative-care", "trial-design", "capital-allocation" ], "summary": "Do not fund another pain procedure trial; add exact-target labels, death-aware outcomes and an independent mortality stop to the recruiting 94-person comparison.", "content": "Biological rationale. Early celiac-nerve interruption may prevent pain escalation and reduce opioid toxicity. Supporting evidence. One 96-person randomized trial improved pain at three months, and small heat and radiation studies show procedural signals. Contrary evidence. A modern medication-controlled trial found no pain, opioid or quality-of-life advantage; direct ganglion and splanchnic destruction produced randomized survival warnings. Assumptions. The NCT06160323 team will name every delivered target and dose, release de-identified participant-level data and permit an independent monitor. Prediction. At least 90/94 assignments will have vital status and a joint pain-death outcome at three months; at least 90% will have daily opioid dose, pain interference and procedure classification; a useful early route will improve pain interference by at least two points or cut oral-morphine-equivalent dose by at least 25% without worse quality-adjusted survival. Falsification experiment. Enroll no mission-funded participant. Spend at most £250,000 on independent monitoring, data custody and analysis of the live trial. Stratify randomization or analysis by centre, stage and exact ganglion versus plexus procedure. Count death before three months as the worst outcome in a sensitivity analysis and publish all randomized participants. Pause direct ganglion treatment if the death hazard ratio reaches 1.5 after at least 20 deaths, or for any treatment-related death; retain the independent monitor's stricter safety authority. Stop the early-use branch for less than a one-point pain-interference difference, less than 15% opioid reduction, worse quality-adjusted survival, ten points more grade 3-plus harm or more systemic-treatment interruption. Wait for this result before comparing heat or radiation with standard care. Give the planned 52-person single-arm radiation study £0 because 125 people have already supplied an uncontrolled signal. Major risks are bleeding, infection, diarrhoea, low blood pressure, neurologic injury, procedure pain, impaired function, treatment delay and unrecognized survival harm.", "sourceIds": [ "clinicaltrials-nct06160323-upfront-cgn-cpn", "wyse-2011-early-eus-cpn-rct", "kanno-2020-eus-cpn-modern-opioid-rct", "levy-2019-celiac-ganglia-versus-plexus-rct", "dong-2021-splanchnic-neurolysis-rct", "bang-2019-celiac-rfa-versus-cpn-rct", "lawrence-2024-celiac-radiosurgery-phase2", "clinicaltrials-nct07689955-celiac-radiotherapy" ], "links": [ { "target": "experiments", "relation": "part-of" }, { "target": "unknown-pdac-nerve-ablation-net-benefit", "relation": "tests" }, { "target": "failure-pdac-pain-death-and-evaluable-denominators", "relation": "repairs" }, { "target": "trial-pdac-celiac-pain-action-map", "relation": "implemented-by" }, { "target": "programme-5m-decision-core", "relation": "constrained-by" }, { "target": "hypothesis-cryoablation-pragmatic-pain-trial", "relation": "blocks" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "none", "randomised": "strong", "independentReplication": "moderate — enough to justify a guarded resolving trial", "sampleSize": "Live target n=94; zero mission-funded participants; monitoring cap £250,000", "effect": "Turn an active pain trial into a joint benefit-harm decision with exact intervention identity", "limits": [ "Protocol amendment unconfirmed", "Small death count", "Ganglion visibility is not randomized", "Rescue crossover after four weeks", "Excludes current opioid users", "Thresholds are mission choices" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The gate preserves the pain mechanism while measuring the harms that prior trials exposed." }, "animalEvidence": { "rating": "not-applicable", "reason": "Randomized human evidence makes clinical monitoring the priority." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "The outcomes are pain, function and life." }, "reproducibility": { "rating": "strong", "reason": "A masked randomized study with exact target labels can resolve conflict across prior trials." }, "commercialConflicts": { "rating": "low", "reason": "Funds buy independent safety and data access, not procedures or products." }, "recency": { "rating": "current", "reason": "The trial is recruiting and was updated in May 2026." } } }, "layout": { "x": 1156, "y": 1080 }, "updated": "2026-09-15" }, { "id": "hypothesis-pdac-palliative-route-completion", "title": "A needs-triggered embedded palliative route will increase good days without withholding treatment", "shortTitle": "PDAC good-days route", "type": "hypothesis", "status": "testing", "scope": [ "pdac", "palliative-care", "quality-of-life", "care-delivery", "capital-allocation" ], "summary": "Recover the finished Danish trial, then test a defined route that counts every patient and weights survival days by lived quality.", "content": "Biological rationale. Pain, nausea, maldigestion, cachexia, depression and treatment confusion are parallel failure modes that can remove usable life even if tumour treatment works. Supporting evidence. A pancreatic randomized trial improved quality scores and reduced chemotherapy near death; current guidance recommends early care alongside cancer treatment. Contrary evidence. Survival was unchanged, another large symptom trial missed both main endpoints, a later survival claim used historical controls and the service packages differ. Assumptions. Centres can embed a nurse-led screen into the oncology visit, reserve specialists for declared needs, provide home or remote contact, and link patient reports, treatment, admissions and death without forcing treatment withdrawal. Prediction. The route will add at least ten quality-weighted days alive and at home over 24 weeks, reduce chemotherapy in the final 30 days by at least eight percentage points from an expected 28%, and leave overall survival no worse by more than a hazard-ratio upper confidence bound of 1.20. A quality-weighted day gives each day alive and at home a value from zero to one using the patient's repeated quality score; death, hospital days and unreported days after confirmed incapacity receive declared values rather than disappearing. Falsification experiment. First spend at most £150,000 over six months to obtain and independently analyse all assignments from the 250-person NCT04632303 trial; fund no duplicate participants. Then spend at most £2m over 24 months for a stepped rollout across at least 20 NHS centres and 1,500 consecutive unresectable PDAC patients. Screen pain, nausea, nutrition, function, mood, caregiver load and treatment understanding within 72 hours of the treatment decision. A high-need result triggers a specialist visit within seven days; all others receive an embedded review within 21 days and repeat screening every treatment cycle. Publish the full eligible-to-screen-to-visit denominator, component use, quality-weighted days, hospital-free days, emergency visits, treatment delivery, place of death, caregiver burden and cost. Stop or redesign if less than 90% are screened, less than 80% of high-need patients are seen within seven days, median benefit is below ten quality-weighted days, survival crosses the harm bound, beneficial cancer therapy is withheld because of route assignment, opioid or psychiatric harm increases, or any access gap by language, deprivation or centre exceeds ten points. Major risks are undertreatment, mistaken prognosis, opioid toxicity, missed psychiatric crisis, caregiver coercion, digital exclusion, workforce overload and converting supportive care into a rationing tool.", "sourceIds": [ "maltoni-2016-systematic-palliative-qol", "maltoni-2016-systematic-palliative-end-of-life", "woo-2019-pancreatobiliary-palliative-rct", "dufva-2026-home-palliative-pdac", "clinicaltrials-nct04632303-pdac-palliative", "sanders-2024-asco-palliative-guideline" ], "links": [ { "target": "experiments", "relation": "part-of" }, { "target": "failure-pdac-palliative-timing-and-analysis", "relation": "repairs" }, { "target": "unknown-pdac-palliative-route-net-benefit", "relation": "tests" }, { "target": "trial-pdac-palliative-care-action-map", "relation": "implemented-by" }, { "target": "trial-pdac-cachexia-action-map", "relation": "integrates-with" }, { "target": "trial-pert-pdac-action-map", "relation": "integrates-with" }, { "target": "trial-pdac-celiac-pain-action-map", "relation": "provides-care-backbone-for" }, { "target": "programme-25m-prospective-observatory", "relation": "funded-by" } ], "evidence": { "humanProspective": "strong for components; none for proposed route", "humanRetrospective": "moderate", "randomised": "moderate for early care; none for proposed complete route", "independentReplication": "moderate — mixed results", "sampleSize": "Existing randomized recovery target n=250; proposed consecutive route n=1,500 at at least 20 centres; cap £150,000 plus £2m", "effect": "Target at least ten added quality-weighted days alive and at home over 24 weeks and at least eight-point reduction in late chemotherapy", "limits": [ "Novel combined endpoint", "Open-label service", "Stepped rollout can confound calendar time", "Quality reports may be missing", "Workforce variation", "Survival non-inferiority needs more power", "Thresholds are mission choices" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The route directly detects and treats patient-level failure modes." }, "animalEvidence": { "rating": "not-applicable", "reason": "Human quality and care use are the only valid tests." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "A service cannot be reduced to a laboratory mechanism." }, "reproducibility": { "rating": "moderate", "reason": "Twenty centres and a consecutive denominator test whether the route transfers." }, "commercialConflicts": { "rating": "low", "reason": "Provider-neutral care components and independent analysis are required." }, "recency": { "rating": "current", "reason": "The gate begins with the largest current pancreatic-only randomized record." } } }, "layout": { "x": 1264, "y": 1188 }, "updated": "2026-09-15" }, { "id": "hypothesis-pdac-pharmacogenetic-route-completion", "title": "A measured DPYD and UGT1A1 dosing loop can prevent severe harm without losing treatment", "shortTitle": "Measured dosing loop", "type": "hypothesis", "status": "hypothesis", "scope": [ "pdac", "treatment-safety", "pharmacogenetics", "care-delivery", "capital-allocation" ], "summary": "Join the existing tests to treatment and outcome records, then repair timing and dose errors instead of funding another test-efficacy trial.", "content": "Biological rationale. DPYD and UGT1A1 alter clearance of 5-FU and irinotecan's active metabolite, so a locked result-to-dose action can prevent excess exposure. Supporting evidence. Prospective and real-world cohorts report lower carrier toxicity after guided starts. Evidence against. A pancreatic and biliary genotype-guided study failed its tolerability rule, severe toxicity persists after negative tests, and panel coverage varies by ancestry. Assumptions. Laboratories can return results before treatment, clinicians can follow one rule, and linked records can distinguish drug toxicity from disease. Prediction. Among 2,000 consecutive pancreatic patients planned for a fluoropyrimidine or irinotecan regimen, a named-owner route will return at least 95% of required results before cycle one, keep result-caused delay above three working days below 5%, make at least 95% of required starting-dose changes, and reduce severe early drug-linked toxicity among identified poor metabolizers without reducing planned treatment in normal metabolizers. Design. Use 20 NHS centres for 18 months. DPYD testing remains standard care; no patient is randomized to lose it. Join pharmacy order, laboratory result, chemotherapy administration, admission, toxicity, imaging and death. For UGT1A1, measure current care first and introduce a locked poor-metabolizer rule only where agreed by pharmacology and oncology. Keep standard-panel-negative severe cases for blinded wider sequencing and functional testing; do not use reported race as the dosing variable. Scale. If 4% carry a DPYD actionable result, 2,000 people yield about 80 carriers. If 10% are UGT1A1 poor metabolizers, the same route yields about 200, although real pancreatic frequencies must replace both assumptions. Cost and time. Cap at £800,000 and 18 months: £250,000 data joining, £250,000 laboratory and pharmacology work, £200,000 site repair, £100,000 independent analysis and publication. Stop if fewer than 90% of eligible patients are captured, more than 5% start before a required result, result-caused delay exceeds three days in more than 5%, any complete DPD deficiency receives systemic 5-FU, full-dose errors exceed 2% among locked poor-metabolizer groups, severe toxicity exceeds 35% in either identified high-risk group, or 90-day death points materially upward after risk adjustment. Success does not mean cure. It means fewer preventable treatment failures and a reusable dosing-control system. Falsification experiment. Run the consecutive route, keep every intended patient in the denominator, and reject the hypothesis at any stated timing, action, toxicity or death stop. Major safety risks include delayed cancer treatment, unsafe full dosing, permanent underdosing, false reassurance after a negative panel, and unequal variant coverage.", "sourceIds": [ "mhra-2020-dpd-testing", "nhs-england-dpyd-implementation", "henricks-2018-dpyd-guided-dosing", "glewis-2024-pacific-pgx", "shirasu-2019-ugt1a1-folfirinox", "sharma-2019-ugt1a1-mfolfirinox", "heersche-2026-ugt1a1-guided-dosing", "chan-2024-dpyd-non-european-review" ], "links": [ { "target": "experiments", "relation": "part-of" }, { "target": "unknown-pdac-pharmacogenetic-route-net-benefit", "relation": "tests" }, { "target": "failure-pharmacogenetic-panel-and-route-coverage", "relation": "succeeds-where" }, { "target": "trial-pdac-pharmacogenetic-safety-action-map", "relation": "implemented-by" }, { "target": "hypothesis-older-pdac-fast-support-route", "relation": "coordinates-with" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "strong", "randomised": "weak — this is a route-repair hypothesis around established care", "independentReplication": "moderate — component safety evidence repeats but the proposed joined route has not run", "sampleSize": "Proposed n=2,000 across 20 centres; about 80 DPYD carriers at 4% and 200 UGT1A1 poor metabolizers at 10% planning assumptions", "effect": "Falsifiable £800,000 route-completion and safety experiment", "limits": [ "Carrier-frequency assumptions", "No untreated DPYD control", "Toxicity attribution", "Dose escalation varies", "Survival power limited", "Laboratory methods differ" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The measured genes directly control exposure to two core regimen components." }, "animalEvidence": { "rating": "not-applicable", "reason": "The experiment repairs an established human care route." }, "inVitroEvidence": { "rating": "moderate", "reason": "Functional work helps classify residual variants but cannot prove route benefit." }, "reproducibility": { "rating": "weak", "reason": "The exact joined pancreatic route and gates remain untested." }, "commercialConflicts": { "rating": "low", "reason": "Funding buys public route data and independent analysis, not an assay company." }, "recency": { "rating": "current", "reason": "The design incorporates the June 2026 UGT1A1 outcome report." } } }, "layout": { "x": 2000, "y": 1660 }, "updated": "2026-09-15" }, { "id": "hypothesis-pdac-surgical-route-completion", "title": "Named route ownership plus audited proficiency will increase PDAC treatment completion without increasing surgical harm", "shortTitle": "Complete the surgical route", "type": "hypothesis", "status": "hypothesis", "scope": [ "pdac" ], "summary": "Use England's existing specialist network to test a first-scan-to-last-treatment bundle, with current audit measures, full-course delivery and hard safety stops.", "content": "Supporting evidence and rationale: England has a specialist surgical network and 96.4% 90-day survival after Whipple, yet only 62% of patients receive chemotherapy or chemoradiotherapy within 14 weeks and the national record does not show full-course completion. Complications repeatedly track with omission, but progression, frailty, patient choice and broken handoffs are mixed. Evidence against easy success: centralization is observational, current audit does not report full-course completion, complication studies are confounded, and procedure trials do not test this bundle. Prediction: one named owner from first specialist review through nine months, joined surgery and oncology planning, nutrition and pancreatic-enzyme support, procedure-specific proficiency checks, fixed complication rescue, an oncology appointment booked before discharge and weekly route-loss review will raise timely postoperative treatment from 62% to at least 72% and full planned treatment completion by at least ten points without increasing 90-day death. Assumptions: the 62% audit baseline transports to participating centres, the bundle can change a ten-point gap, centre correlation is near 0.01, and reasons for changing treatment can be recorded consistently. Falsification experiment and design: all 23 English HPB centres first record every patient whose scan might permit curative intent, including people later not operated. One pilot centre tests fields without contributing to the effect estimate. The other 22 centres form 11 matched pairs and cross to the bundle in randomized order. The plan is locked at expert MDT but can change for recorded progression, toxicity, patient choice or new evidence. Open, laparoscopic and robotic operations remain allowed only under procedure-specific proficiency and rescue rules; no centre receives money for buying a robot. Primary outcomes are full planned route completion by nine months and chemotherapy or chemoradiotherapy within 14 weeks after Whipple. Co-primary safety is 90-day death; secondary outcomes include expert-review time, resection and non-resection reasons, major complications, organ failure, readmission, total dose delivered, patient-reported recovery, recurrence, survival, cost and inequalities. Planning maths: detecting 62% versus 72% at two-sided 5% and 80% power needs 346 people per arm, 692 total under individual allocation. At about 55 patients per centre and intracluster correlation 0.01, design effect = 1 + (55 - 1) × 0.01 = 1.54. Then 692 × 1.54 ÷ 0.90 for 10% incomplete data = 1,184, rounded to 1,200. If the first 300 patients imply intracluster correlation near 0.02, design effect becomes 2.08 and the need becomes 692 × 2.08 ÷ 0.90 = 1,599, capped at 1,700. An independent statistician must simulate the stepped rollout before release because 22 clusters, time trends and baseline completion can dominate the simple calculation. Passing gates: at least 95% of eligible scans and 95% of planned doses recorded; timely treatment at least 72%; full completion at least ten points above run-in; 90-day mortality no more than one point above matched control; major complications no more than five points higher; and no more than a ten-point route-completion gap across prespecified groups. Stop or amend if recording falls below 90%, mortality exceeds 5% or rises by more than one point, serious complications rise five points, the operation rate falls five points without more progression or refusal, or an inequality gap widens ten points. Estimated cost and time: £2.5m over three years from the £5.5m remaining patient-safety, access and health-economics line, leaving £3m. The release buys data linkage, route staff, recovery support and independent analysis, not capital equipment. Major safety risks: operative death, leak, organ failure, delayed systemic treatment, false pressure to operate, loss of patient choice, travel burden, unequal access and platform use beyond proficiency. A process result cannot claim cure; recurrence and cause-specific death continue for at least five years.", "sourceIds": [ "gudmundsdottir-2023-staging-laparoscopy", "van-dongen-2023-preopanc-staging", "rompen-2026-preopanc2-staging", "theijse-2024-nontherapeutic-laparotomy", "npaca-2026-state-of-nation", "nhs-england-2024-hpb-pancreatic-service", "nice-2018-ng85-pancreatic", "de-wilde-2012-centralization", "latenstein-2021-centralization", "coupland-2016-england-resection-volume", "labori-2016-adjuvant-completion", "henry-2023-complications-adjuvant", "kollbeck-2025-complications-chemo", "van-hilst-2019-leopard2", "korrel-2023-diploma", "bruna-2025-diploma-long-term", "klotz-2024-europa", "liu-2024-robotic-open-pd", "jin-2026-portal", "labori-2024-norpact1", "bai-2025-cispd1", "tan-2026-neoadjuvant-ipd-meta", "clinicaltrials-nct04340141-a021806" ], "links": [ { "target": "hypothesis-pdac-two-timepoint-staging-route", "relation": "contains" }, { "target": "trial-pdac-occult-metastasis-action-map", "relation": "uses" }, { "target": "experiments", "relation": "part-of" }, { "target": "claim-pdac-surgery-is-a-complete-treatment-route", "relation": "builds-on" }, { "target": "system-england-pancreatic-surgery-network", "relation": "uses" }, { "target": "failure-pdac-postoperative-treatment-loss", "relation": "repairs" }, { "target": "failure-leopard2-laparoscopic-whipple-safety", "relation": "learns-from" }, { "target": "unknown-pdac-surgical-route-completion", "relation": "tests" }, { "target": "trial-pdac-surgical-route-action-map", "relation": "prioritized-by" }, { "target": "claim-neoadjuvant-therapy-improves-selection-not-cure", "relation": "stratifies-by" }, { "target": "programme-500m-integrated-mission", "relation": "funded-by" }, { "target": "trial-pdac-perioperative-sequencing-action-map", "relation": "bounded-by" }, { "target": "unknown-pdac-perioperative-treatment-route", "relation": "bounded-by" } ], "evidence": { "humanProspective": "moderate for components; none for bundle", "humanRetrospective": "strong for route loss", "randomised": "strong for procedure components; none for bundle", "independentReplication": "none for proposed route", "sampleSize": "Proposed n=1,200, adaptive cap 1,700, across 22 effect-estimating centres plus one pilot", "effect": "Generated targets: timely postoperative treatment 62% to at least 72%, full-route completion up at least ten points, no 90-day mortality increase", "limits": [ "Generated intervention", "Cluster correlation uncertain", "Only 22 effect-estimating centres", "National audit baseline is treatment receipt, not completion", "Treatment-before-surgery creates different routes", "Five-year follow-up needed for cancer outcomes" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The bundle targets observed loss points that can block treatment after an otherwise justified resection." }, "animalEvidence": { "rating": "not-applicable", "reason": "This is a human care-delivery hypothesis." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Laboratory evidence cannot test route ownership, proficiency or recovery." }, "reproducibility": { "rating": "none", "reason": "The complete bundle has not been tested independently." }, "commercialConflicts": { "rating": "low", "reason": "The programme buys no robot or proprietary platform and requires independent statistics." }, "recency": { "rating": "current", "reason": "The design is anchored to current 2026 English audit results and current randomized operation evidence." } } }, "updated": "2026-09-14", "contentSections": [ "A 500-person intended-operation staging comparison is nested inside this route rather than funded as another trial. Current-care sites retain NICE-selected staging until crossover; route sites lock early high-risk staging and begin each intended open operation with a short camera inspection. £0.4m of the existing £2.5m pays for protocol, pathology fields and independent analysis, leaving £2.1m for route ownership, recovery and treatment completion. Direct view, biopsy and wash cytology remain separate; no automatic no-surgery rule follows cytology alone.", "Timing is a recorded stratum, not a second randomized intervention. Each centre records upfront surgery or treatment first, exact regimen, every cycle, progression before surgery and postoperative delivery. The cluster route trial adjusts for timing and tests completion only. It does not claim that a better handoff identifies the best treatment order; A021806 owns that comparison until its survival result reports." ] }, { "id": "hypothesis-pdac-two-timepoint-staging-route", "title": "Hypothesis: a locked two-timepoint staging rule will prevent futile PDAC laparotomy without harmful delay", "shortTitle": "Two-timepoint staging route", "type": "hypothesis", "status": "supported", "scope": [ "pdac" ], "summary": "Early staging for high-risk scans plus a short same-session check before incision may preserve treatment while preventing avoidable open exploration.", "content": "Biological rationale: scan resolution misses small liver and abdominal-surface deposits, while a camera can inspect and biopsy those surfaces before an open operation. Supporting evidence: PREOPANC-2 found non-therapeutic laparotomy at 4.5% with same-session staging versus 17.1% without, an absolute 12.6-point association and reported NNT 8; Mayo found an 18% positive yield with 1.2% procedure complications and 0.1% major complications. Evidence against easy success: staging was not randomized, expert centres select patients, a separate procedure may delay treatment, and cytology-only disease has no proven treatment rule. Assumptions: the no-staging non-therapeutic-laparotomy rate is near 17%, the locked route can lower it to 7%, centre clustering inflates sample need by about 1.3, same-session inspection adds no more than 30 minutes, and pathology can return fast enough to avoid route delay. Prediction: a two-timepoint rule will reduce open exploration without resection for hidden spread by at least eight percentage points, with major staging harm no higher than 1.5% and no median treatment delay above three days. Falsification experiment: embed a 500-person intended-operation comparison in the existing 23-centre English surgical-route study. Current-care centres use NICE-selected staging; centres crossing to the route use early separate staging only for a locked high-risk state and start every intended open operation with a standard same-session inspection. Biopsy visible lesions and report direct view, pathology and wash cytology separately. Planning maths: a simple individual comparison of 17.1% versus 7.0%, two-sided 5% and 80% power, needs about 162 people per arm, 324 total. Multiplying by a 1.3 centre factor and dividing by 0.90 for incomplete records gives 324 × 1.3 / 0.90 = 468, rounded to 500. The independent statistician must simulate the actual 22-cluster stepped timing before lock. Pass only if staging fields are at least 95% complete, hidden-spread non-therapeutic laparotomy falls at least eight points, no more than 3% have hidden spread first found after a negative laparoscopy at open exploration, median added anaesthetic time is no more than 30 minutes, major staging harm is no more than 1.5%, and systemic treatment or surgery is not delayed more than three days. Stop for major harm above 2%, treatment delay above seven days in more than 5%, a false-positive denial of surgery, automatic exclusion from surgery based only on cytology, or an inequality gap above ten points. Estimated cost and time: £0.4m and 30 months, carved from the existing £2.5m English surgical-route budget for protocol lock, pathology fields and independent analysis; the other £2.1m retains route ownership, recovery and treatment-completion work. No new capital is added, the patient-safety line remains £15m, and the mission remains £500m. Major safety risks: anaesthetic or port injury, bleeding, infection, delay, false reassurance after a negative result, and loss of a potentially useful operation after an uncertain cytology result. This tests safer selection and route completion, not cure or the best systemic treatment.", "sourceIds": [ "gudmundsdottir-2023-staging-laparoscopy", "van-dongen-2023-preopanc-staging", "rompen-2026-preopanc2-staging", "theijse-2024-nontherapeutic-laparotomy", "stoop-2026-global-staging-survey", "nice-2018-ng85-pancreatic" ], "links": [ { "target": "experiments", "relation": "part-of" }, { "target": "claim-staging-laparoscopy-avoids-incision-not-proven-survival", "relation": "derived-from" }, { "target": "failure-staging-yield-without-patient-utility", "relation": "escapes" }, { "target": "unknown-pdac-staging-laparoscopy-rule", "relation": "tests" }, { "target": "trial-pdac-occult-metastasis-action-map", "relation": "bounded-by" }, { "target": "hypothesis-pdac-surgical-route-completion", "relation": "nested-in" }, { "target": "programme-500m-integrated-mission", "relation": "funded-by" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "strong", "randomised": "moderate — generated cluster-timed comparison", "independentReplication": "moderate", "sampleSize": "Planned intended-operation cohort n=500 inside the existing 1,200-person route study", "effect": "Designed for non-therapeutic laparotomy 17.1% to 7.0%, with an eight-point minimum pass gate", "limits": [ "Generated design", "Cluster simulation required", "Bundle attribution", "Cytology action unresolved", "Low event rate in low-risk group", "No direct survival claim" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The procedure directly observes surfaces that scans cannot reliably resolve before open access." }, "animalEvidence": { "rating": "not-applicable", "reason": "Human operations and route outcomes are the decisive test." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "No laboratory model can estimate avoided incision, delay or patient choice." }, "reproducibility": { "rating": "moderate", "reason": "Several cohorts support detection, while the exact English two-timepoint rule remains untested." }, "commercialConflicts": { "rating": "low", "reason": "The experiment uses standard equipment and independent analysis with no product purchase target." }, "recency": { "rating": "current", "reason": "The design is bounded by 2026 evidence, practice variation and current England guidance." } } }, "updated": "2026-09-15" }, { "id": "hypothesis-pdac-vte-route-completion", "title": "A seven-day clot-risk decision and pharmacy route will prevent more PDAC VTE than education alone", "shortTitle": "Seven-day VTE route", "type": "hypothesis", "status": "hypothesis", "scope": [ "pdac" ], "summary": "Cluster-randomize a complete assessment-to-use route because pharmacologic efficacy is known and the remaining failure occurs before or during delivery.", "content": "Rationale: anticoagulant prophylaxis reduces VTE by about 69% across randomized pancreatic-cancer evidence, yet one current cohort found only 30% receiving it after a median 43 days and half of clots within two months. A small quality project changed little with teaching and reached 56% uptake only after adding an electronic prompt. Hypothesis: a named oncology owner, automatic seven-day eligibility and bleeding check, patient choice, same-day prescription, pharmacy confirmation and follow-up at every cycle will raise suitable use from about 30% to 70% and reduce symptomatic VTE without a material bleeding increase. Planning maths: assume untreated six-month VTE risk 12% and treated relative risk .31, so treated risk is 12% × .31 = 3.72%. At 30% use, expected risk is .30 × 3.72% + .70 × 12% = 9.52%. At 70% use it is .70 × 3.72% + .30 × 12% = 6.20%, a 3.31-point difference. A simple two-proportion calculation needs about 2,070 people for 80% power; multiplying by a provisional 1.2 cluster penalty gives 2,485, rounded to 2,500. The final sample requires observed baseline and cluster correlation. Test: randomize about 30 clinics in a stepped sequence, retaining usual clinician choice in both arms. Record every consecutive systemic-treatment start, reason for eligibility or exclusion, offer, choice, fill, use, interruptions, symptomatic and incidental VTE, hospitalization, VTE death, major and relevant non-major bleeding, treatment delays, quality of life and cost through six months. Co-primary results are documented decision within seven days and symptomatic VTE; major bleeding is a separate safety outcome, not hidden in a composite. Go if decision completion reaches at least 80%, suitable use rises by 30 points and symptomatic VTE falls by at least three points. Stop for an observed major-bleeding excess of two points or more, a fatal-bleeding imbalance, no symptomatic-VTE reduction, benefit confined to scan-detected events, or more than 10% unable to obtain the chosen drug. Budget: £3m from the existing £15m patient-safety, access and health-economics line, leaving £12m. This tests delivery, not whether anticoagulants work. Supporting evidence: randomized trials reproduce the drug-class effect and two care audits show late or incomplete delivery. Evidence against: no completed study shows that this full route improves symptomatic VTE, and bleeding, patient choice and pharmacy access can erase benefit. Assumptions: six-month untreated VTE risk is 12%, treatment RR is .31, cluster penalty is 1.2 and clinics can raise suitable use from 30% to 70%. Predicted result: decision completion at least 80%, suitable use up 30 points and symptomatic VTE down at least three points without major bleeding up two points. Falsification experiment: the stepped cluster trial fails the hypothesis if symptomatic VTE does not fall or any stop rule is met. Estimated time: three years. Major safety risks: gastrointestinal or intracranial bleeding, thrombocytopenia, renal drug accumulation and treatment interruption.", "sourceIds": [ "maraveyas-2012-fragem", "pelzer-2015-conko004", "vadhan-raj-2020-cassini-pdac", "frere-2020-pdac-thromboprophylaxis-meta", "heffley-2024-pdac-vte-uptake", "noor-2026-pdac-vte-implementation", "ash-2023-primary-vte-prophylaxis-guide", "overvad-2022-khorana-validation" ], "links": [ { "target": "experiments", "relation": "part-of" }, { "target": "claim-anticoagulation-prevents-pdac-vte-not-survival", "relation": "builds-on" }, { "target": "failure-pdac-vte-prophylaxis-delivery-gap", "relation": "repairs" }, { "target": "unknown-pdac-vte-prophylaxis-rule", "relation": "tests" }, { "target": "trial-pdac-vte-action-map", "relation": "prioritized-by" }, { "target": "programme-500m-integrated-mission", "relation": "funded-by" } ], "evidence": { "humanProspective": "strong for drug effect; weak for implementation effect", "humanRetrospective": "moderate", "randomised": "strong for drug effect; none for the route", "independentReplication": "none for the complete route", "sampleSize": "Proposed cluster-randomized n about 2,500 across about 30 clinics", "effect": "Generated target: suitable use +30 points and symptomatic VTE -3 points without major bleeding +2 points", "limits": [ "Planning event rate", "Provisional cluster penalty", "Open-label care", "Choice of agent varies", "Rare safety outcomes", "Local pharmacy systems differ" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The route increases timely exposure to treatments with a reproduced clinical effect." }, "animalEvidence": { "rating": "not-applicable", "reason": "The hypothesis concerns human care delivery." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Laboratory evidence cannot test assessment, choice, fill and persistence." }, "reproducibility": { "rating": "none", "reason": "The complete seven-day route has not been tested independently." }, "commercialConflicts": { "rating": "low", "reason": "The design is agent-neutral and should be independently funded and analysed." }, "recency": { "rating": "current", "reason": "The design responds to 2024 and 2026 delivery evidence and current guidance." } } }, "layout": { "x": 168, "y": 92 }, "updated": "2026-09-14" }, { "id": "hypothesis-pdt-measured-spatial-gate", "title": "Reconstruct PDT as a measured spatial system before another efficacy trial", "shortTitle": "PDT spatial measurement gate", "type": "hypothesis", "status": "testing", "scope": [ "pdac", "photodynamic-therapy", "trial-recovery", "dosimetry" ], "summary": "Existing participant data should first test whether planned light, perfusion and anatomy can predict the treated volume and off-target injury; failure means stop, not another small cohort.", "content": "Biological rationale. Photodynamic cell killing requires a light-sensitive drug, photons and oxygen in the same place and time. Supporting evidence. Three human studies repeatedly produced local dead tissue; verteporfin gives a short drug-to-light interval and 24 to 48 hours of light sensitivity. Contrary evidence. Dead-zone volume varies at the same nominal energy, 3/8 recent participants had no visible dead zone, multiple fibres injured tissue outside the pancreas and no controlled patient benefit exists. Assumptions. The four study groups will provide participant-level scans, fibre coordinates, calibrated light output, infusion timing, withdrawals, adverse events, later treatment and outcomes for up to 64 existing starters or treated participants across the 15-, 12-, 13- and 24-person studies. Prediction. A locked model using geometry, tumour perfusion and light energy will estimate dead-zone coverage with median absolute error no greater than 20 percentage points and identify every clinically important off-target injury; at least 70% of illuminated participants will achieve 50% or greater coverage of living local tumour. Falsification experiment. Enroll nobody and spend at most £500,000 on independent data recovery, scan alignment and external testing. Train only on the older 15- and 12-person datasets, lock the calculation, test it on all 13 VERTPAC-02 starters, then apply it without outcome labels to the active 24-person cohort until follow-up matures. Require 100% starter reconciliation, 95% procedure and calibration records and 90% usable scan pairs. Stop the pancreatic PDT branch if access is refused, errors exceed 20 points, any serious off-target injury is missed or less than half of treated participants reach 50% coverage. Passing this gate permits a separately funded randomized design; it does not prove efficacy. Major risks are bleeding, pancreatitis, infection, bowel or vessel injury, light toxicity, treatment delay and false confidence from scan-defined dead tissue.", "sourceIds": [ "huggett-2014-verteporfin-pdt", "dewitt-2019-eus-porfimer-pdt", "chandrasekhara-2021-eus-verteporfin-pdt", "clinicaltrials-nct03033225-vertpac02", "clinicaltrials-nct01770132-porfimer-pdt", "clinicaltrials-nct06381154-pdt-pembrolizumab" ], "links": [ { "target": "experiments", "relation": "part-of" }, { "target": "unknown-pdt-spatial-coverage-and-patient-benefit", "relation": "tests" }, { "target": "failure-pdt-geometry-selection-and-bundling", "relation": "repairs" }, { "target": "trial-pdac-pdt-action-map", "relation": "implemented-by" }, { "target": "programme-5m-decision-core", "relation": "constrained-by" }, { "target": "hypothesis-gannon-independent-safety-gate", "relation": "compares-with" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "none", "randomised": "none", "independentReplication": "weak", "sampleSize": "Up to 64 existing study records; zero new participants; cap £500,000", "effect": "Turn a nominal fibre dose into a tested coverage and injury model before any efficacy claim", "limits": [ "Cross-study drugs and routes differ", "Some source data may be unavailable", "Oxygen may not be measured directly", "Scan-defined dead tissue is an imperfect label", "Current follow-up is immature", "Thresholds and cap are mission choices" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The model follows the physical inputs required for photochemical killing." }, "animalEvidence": { "rating": "moderate", "reason": "Animal studies can inform light spread, but the gate is deliberately tested in existing humans." }, "inVitroEvidence": { "rating": "strong", "reason": "Controlled systems establish drug, photon and oxygen dependence." }, "reproducibility": { "rating": "moderate", "reason": "Locked training and external testing across institutions is designed to expose failure." }, "commercialConflicts": { "rating": "moderate", "reason": "Independent custody, full starter flow and publication rights are required before payment." }, "recency": { "rating": "current", "reason": "The gate uses the active 24-person cohort without changing its treatment." } } }, "layout": { "x": 958, "y": 882 }, "updated": "2026-09-15" }, { "id": "hypothesis-pelican-specimen-salvage-gate", "title": "Salvage mechanism from PELICAN without repeating RFA", "shortTitle": "PELICAN specimen-salvage gate", "type": "hypothesis", "status": "testing", "scope": [ "pdac", "radiofrequency-ablation", "tissue-reuse", "immune-mechanism" ], "summary": "A small stored-sample audit can close or redirect the claimed immune mechanism while the harmful clinical route remains stopped.", "content": "Biological rationale. Thermal tumour destruction may release antigens and danger signals, but surgery can also trigger non-specific inflammation and wound repair. Supporting evidence. PELICAN created a randomized treatment assignment and the investigators identify pre- and post-procedure tissue as a possible mechanism source. Contrary evidence. The clinical treatment produced no survival or progression benefit, more severe serious events and worse quality of life. Its CA19-9 interaction is secondary and points against benefit in marker responders. Assumptions. Consent permits reuse; paired tumour or blood exists before and after treatment; operation-only and chemotherapy records can distinguish procedure injury; and at least two independent laboratories can run the same locked assays. Prediction. Among at least 60 paired participants, RFA will produce a reproducible tumour-specific T-cell or antigen-presentation change beyond generic inflammatory markers, and that change will associate with local tumour control after accounting for occult metastasis and chemotherapy delivered. Falsification experiment. Spend at most £150,000 from the existing immune-mechanism and statistics lines and enroll zero patients. Inventory all samples before analysis, freeze marker and missing-data rules, blind laboratories to outcome, reconcile treatment and surgery times and publish null findings. Stop if fewer than 60 pairs exist, paired material covers less than 80% of the selected set, laboratory agreement is below 80%, only broad inflammation changes, or the signal disappears after chemotherapy and operation adjustment. Do not use the CA19-9 subgroup to treat patients. Even a positive mechanism result releases no RFA trial; it may justify only a safer laboratory perturbation. Major safety risk is false resurrection of a route that caused severe harm and poorer quality of life; data risks include selective sample survival, treatment-time bias and outcome unblinding.", "sourceIds": [ "seelen-2026-pelican", "clinicaltrials-nct03690323-pelican" ], "links": [ { "target": "experiments", "relation": "part-of" }, { "target": "unknown-rfa-human-mechanism-after-negative-trial", "relation": "tests" }, { "target": "failure-rfa-selection-and-route-attrition", "relation": "learns-from" }, { "target": "trial-pdac-rfa-action-map", "relation": "implemented-by" }, { "target": "failure-immune-reprogramming-without-patient-benefit", "relation": "bounded-by" }, { "target": "programme-5m-decision-core", "relation": "constrained-by" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong", "independentReplication": "none", "sampleSize": "Existing PELICAN material only; target at least 60 pairs; zero new participants; capped at £150,000", "effect": "Close or redirect an immune-mechanism claim without clinical RFA exposure", "limits": [ "Specimens may not exist", "Consent may not permit reuse", "Paired sampling may be selected", "Operation injury confounds immune change", "Associations do not prove mechanism", "Thresholds and budget are mission choices", "No result can override the randomized clinical harm finding" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "Thermal destruction can change antigen and inflammatory signals." }, "animalEvidence": { "rating": "moderate", "reason": "Animal immune observations justify low-cost reuse only." }, "inVitroEvidence": { "rating": "moderate", "reason": "Cell-death signals can be tested with preserved human material." }, "reproducibility": { "rating": "none", "reason": "The experiment requires two blinded laboratories because no human mechanism is established." }, "commercialConflicts": { "rating": "low", "reason": "No new device or procedure funding is allowed." }, "recency": { "rating": "current", "reason": "The design responds directly to the 2026 result." } } }, "layout": { "x": 730, "y": 654 }, "updated": "2026-09-15" }, { "id": "hypothesis-pipac-existing-data-and-control-gate", "title": "Recover all-starter PIPAC data, then require a controlled add-on test", "shortTitle": "PIPAC control gate", "type": "hypothesis", "status": "testing", "scope": [ "pdac", "peritoneal-metastases", "pipac", "trial-design", "capital-allocation" ], "summary": "A small data-recovery payment can expose selection now; no new procedure should be funded until systemic care can be held constant.", "content": "Biological rationale. PIPAC may raise chemotherapy exposure at abdominal-surface deposits while systemic treatment covers disease elsewhere. Supporting evidence. Multiple pancreatic series report tissue regression; the 156-person registry shows multicentre feasibility and quantifies attrition. Contrary evidence. No pancreatic comparison isolates PIPAC, only 35.2% completed three procedures, and the current phase II study bundles aerosol nab-paclitaxel with intravenous nab-paclitaxel-gemcitabine. Assumptions. All six registry centres and the NCT05371223 team can export screened, enrolled and procedure-level data with dates, fitness, disease sites, prior and concurrent drugs, reasons for non-access and stopping, biopsies, symptoms, quality of life, hospital days and deaths. Prediction. At least 95% of all starters and 90% of screened candidates can be accounted for; time-varying treatment models will shrink the apparent survival advantage of three-plus procedures or nab-paclitaxel materially if selection drives it. Falsification experiment. Spend at most £300,000 on independent data recovery, code and public results, with £0 for new PIPAC devices, drugs or participants. Stop if less than 95% of starters or less than 90% of screened candidates can be linked, procedure dates are missing in more than 5%, or systemic treatment cannot be reconstructed. If recovery is adequate, design but do not yet launch a pragmatic comparison in fit, peritoneal-dominant PDAC: same systemic regimen and supportive care in both groups, random assignment to add PIPAC or not, stratified by disease burden and centre. Primary output is quality-adjusted survival from assignment; key safety outputs are 90-day mortality, bowel injury, obstruction, hospital-free days and delivered systemic-dose intensity. A pilot must retain at least 80% of planned systemic dose, keep excess grade 3-plus harm below ten points and deliver the first PIPAC in at least 85% of those assigned. Failure retires mission funding. Major risks are bowel perforation, infection, bleeding, anaesthetic harm, kidney or marrow injury, treatment delay, obstruction and rapid distant progression.", "sourceIds": [ "digiorgio-2026-pipac-pdac-registry", "clinicaltrials-nct05371223-nab-pipac", "digiorgio-2024-nab-pipac-protocol", "graversen-2017-pipac-pdac" ], "links": [ { "target": "experiments", "relation": "part-of" }, { "target": "unknown-pipac-added-survival-quality-and-systemic-control", "relation": "tests" }, { "target": "failure-pipac-repeat-procedure-selection-and-bundling", "relation": "repairs" }, { "target": "trial-pdac-pipac-action-map", "relation": "implemented-by" }, { "target": "programme-5m-decision-core", "relation": "constrained-by" }, { "target": "hypothesis-direct-randomized-data-recovery-gate", "relation": "supports" } ], "evidence": { "humanProspective": "weak", "humanRetrospective": "moderate", "randomised": "none — planned only after data recovery", "independentReplication": "none for added benefit", "sampleSize": "Existing n=156 registry plus target n=38 trial; zero new participants; data-recovery cap £300,000", "effect": "Convert selection-heavy procedure series into a go or stop decision for a later controlled comparison", "limits": [ "Screening logs may not exist", "Time-varying adjustment cannot remove all confounding", "Randomization may recruit slowly", "Control preferences may cross over", "Thresholds are mission choices", "Sponsor agreement unconfirmed" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "The gate preserves the local-delivery hypothesis while demanding evidence of net added value." }, "animalEvidence": { "rating": "not-applicable", "reason": "Existing human data make clinical attribution the binding problem." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "No laboratory result can replace a held-constant systemic-care comparison." }, "reproducibility": { "rating": "moderate", "reason": "Common data definitions and public code can test the six-centre result before new enrollment." }, "commercialConflicts": { "rating": "moderate", "reason": "Milestone payment and independent custody reduce dependence on procedure owners." }, "recency": { "rating": "current", "reason": "The gate is designed around the August 2026 registry and stale current phase II record." } } }, "layout": { "x": 1096, "y": 1020 }, "updated": "2026-09-15" }, { "id": "hypothesis-pnet-time-updated-surgery-rule", "title": "Hypothesis: a time-updated pNET rule can avoid surgery without adding metastatic harm", "shortTitle": "Time-updated pNET action rule", "type": "hypothesis", "status": "hypothesis", "scope": [ "pnet" ], "summary": "A locked rule using repeated pre-operation measurements and procedure-specific harm should outperform a fixed two-centimetre threshold for 1-2 cm non-functioning pNET.", "content": "Biological rationale: small well-differentiated pNET often remains indolent for years, while duct change, growth, higher grade, FDG uptake, receptor loss and emerging spread mark different danger states. Surgery can remove localized disease but causes pancreatic fistula, major complications, endocrine or exocrine loss and occasional death. A rational rule must update future tumour harm and treatment harm together. Assumptions: ASPEN and PANDORA can supply complete follow-up; consecutive centres can measure the same MRI or CT, duct diameter, growth, core-biopsy Ki-67, FDG PET and somatostatin-receptor state; operation type can be predicted before surgery; and five-year registry linkage can capture metastasis, pancreatic function and death. Prediction: compared with current centre practice at the same five-year metastatic-progression limit, a locked rule will reduce operations and major treatment harm while preserving quality-adjusted survival. Study and estimated cost: spend £3m over two years to obtain full cohorts, define one endpoint and freeze a simple size-plus-duct baseline and a larger time-updated model. If external calibration error is below five percentage points and at least 85% have complete outcome linkage, spend £12m-£18m over five years on 3,000 consecutive 1-2 cm patients across at least 30 centres, with an embedded randomized comparison where both choices remain acceptable. At a 2% five-year metastatic or invasive-progression rate, 3,000 people yield about 60 events before loss: enough to reject a grossly unsafe rule, not prove mortality benefit in every subgroup. Keep sporadic and MEN1 estimates separate and report results by required operation. Action gate: release a randomized expansion only if the rule reduces expected major procedure harm by at least 25%, misses no more than 1.5 additional metastatic progressions per 100 people at five years, has positive lower-bound quality-adjusted net benefit, and beats size plus duct change alone. Falsification: stop if performance depends on post-resection pathology, centre calibration moves by more than five points, biopsy adequacy is below 80%, outcome linkage is below 85%, crossovers are discarded, or major harm is not reduced. Supporting evidence: ASPEN and PANDORA support short-term observation, duct dilation concentrated the smallest aggressive ASPEN tumours, and current surgery cohorts quantify procedure-specific harm. Evidence against: synchronous metastasis and adverse pathology occur in a minority of small tumours, short follow-up can miss late spread, and the MEN1 mRNA assay failed future-risk prediction. Major safety risks include delayed metastatic disease, biopsy pancreatitis or bleeding, radiation exposure, anxiety, false reassurance, unnecessary surgery, pancreatic fistula, diabetes and exocrine failure.", "sourceIds": [ "partelli-2022-aspen-interim", "clinicaltrials-nct03084770-aspen", "heidsma-2021-pandora-pnet", "maekawa-2026-small-pnet-surgery-morbidity", "nell-2018-men1-surgery-watchful-waiting", "van-vliembergen-2026-men1-mnet-assay" ], "links": [ { "target": "experiments", "relation": "part-of" }, { "target": "trial-small-pnet-action-map", "relation": "depends-on" }, { "target": "unknown-small-pnet-surgery-selector", "relation": "tests" }, { "target": "claim-small-nfpnet-surveillance-surgery-tradeoff", "relation": "extends" }, { "target": "failure-aspen-cannot-estimate-surgery-benefit", "relation": "succeeds-where" }, { "target": "failure-men1-mnet-assay-no-progression-prediction", "relation": "constrained-by" } ], "evidence": { "humanProspective": "moderate — components exist, the combined rule is untested", "humanRetrospective": "strong for candidate risk and operation harm", "randomised": "none", "independentReplication": "none for the proposed rule", "sampleSize": "Proposed n=3,000 prospective cohort with embedded random allocation in the 1-2 cm group", "effect": "Unobserved; designed to lower major treatment harm by at least 25% without more than 1.5 additional metastatic progressions per 100 at five years", "limits": [ "Three thousand may yield only about 60 primary events", "Five years misses some late spread", "Biopsy samples can undergrade", "Patient preference can limit randomization", "MEN1 and sporadic disease need separate estimates", "Costs are planning estimates" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Repeated tumour state and procedure-specific harm are more relevant to action than diameter alone." }, "animalEvidence": { "rating": "weak", "reason": "Models may support markers but cannot validate the clinical decision rule." }, "inVitroEvidence": { "rating": "moderate", "reason": "Tissue and molecular measures can add candidates, though none is yet a proven selector." }, "reproducibility": { "rating": "none", "reason": "No group has tested this locked time-updated net-benefit rule." }, "commercialConflicts": { "rating": "high", "reason": "The rule could expand or contract imaging, biopsy, assay, drug, ablation and surgery use." }, "recency": { "rating": "current", "reason": "The hypothesis responds to clinical and surgical evidence available through September 2026." } } }, "updated": "2026-09-14" }, { "id": "hypothesis-prospective-ras-resistance-routing", "title": "Hypothesis: prospective resistance routing can extend common-RAS control", "shortTitle": "Resistance routing", "type": "hypothesis", "status": "hypothesis", "scope": [ "pdac", "colorectal", "pan-cancer", "transfer-unproven" ], "summary": "Serial plasma plus fresh progression tissue has found a repeated RAS escape class; the next test is independent reproduction and whether it can route a safe combination before clinical collapse.", "content": "Biological rationale: common-RAS inhibition produces a large survival gain but not durable eradication; PDAC metastases can be polyclonal and RAS dosage or state can alter dependency. The first paired human series changes this from an unconstrained search. Among 44 phase 1/2 daraxonrasib patients, 26/44 = 59% acquired RAS-pathway changes and 16/44 = 36% acquired mutant KRAS amplification. The 36% class clears the original frequency gate of at least 25%. Models and one G12R patient suggest that escape differs by allele and cell state: EGFR and wild-type RAS can dominate G12R escape, retained G12D signalling can accompany reduced cyclophilin A, and epithelial versus mesenchymal states can use EGFR versus FGFR. Contrary evidence: 11/44 patients had more than one acquired alteration; plasma may miss spatially diverse clones; the main dataset is sponsor-led; and a single routed patient with about five months of control cannot prove benefit. Assumptions: plasma plus more than one progressing lesion capture the dominant escape; a limited combination menu can be supplied; routing is fast enough; and a detected change predicts drug sensitivity. Prediction: independent phase 3 sampling reproduces mutant KRAS amplification at a decision-relevant frequency, at least 70% of progressing patients can receive a mechanistically matched plan within 14 days, and one or more prospectively defined routes improve 6-month PFS by at least 15 points versus non-matched control. Falsification experiment: first analyse 60 independent paired cases with blinded second-laboratory review and multi-lesion tissue where safe. Advance only if the route repeats, routing coverage reaches 70% and the resistant clone is suppressed at clinical exposure; then run a multi-arm randomized phase 2. Planning estimate: £5m and 24 months for confirmation and routing, £25m and 48 months for the randomized test. Main risks are biopsy harm, delay, false causal calls, coexisting escape routes and unsafe combinations.", "sourceIds": [ "oreilly-2026-daraxonrasib", "aronchik-2026-daraxonrasib-resistance", "dorbin-2026-allele-specific-daraxonrasib-resistance", "lin-2026-vertical-kras-inhibition", "maddipati-2015-polyclonal-metastasis", "varghese-2025-kras-dosage", "chan-seng-yue-2020-evolution-subtypes" ], "links": [ { "target": "experiments", "relation": "part-of" }, { "target": "claim-kras-mutant-dosage-may-shape-resistance", "relation": "tests" }, { "target": "claim-daraxonrasib-kras-amplification-resistance", "relation": "passes-first-gate" }, { "target": "claim-metastases-can-be-polyclonal", "relation": "depends-on" }, { "target": "programme-25m-prospective-observatory", "relation": "tests" }, { "target": "programme-100m-randomized-platform", "relation": "tests" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "moderate", "randomised": "none", "independentReplication": "moderate", "sampleSize": "Observed paired resistance cohort n=44; proposed independent n=60 confirmation, then multi-arm phase 2", "effect": "Frequency gate passed: mutant KRAS amplification 36% exceeds 25%. Remaining gates: ≥70% actionable routing within 14 days and +15 points at 6-month PFS", "limits": [ "Generated routing hypothesis", "Human resistance frequency comes from one sponsor-led cohort", "No prospective matched-treatment proof", "Detected changes may be passengers or coexist", "Tumour heterogeneity may defeat limited sampling", "Combination access is a commercial constraint" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Paired human samples show recurrent mutant KRAS amplification and wider RAS-pathway reactivation at progression." }, "animalEvidence": { "rating": "strong", "reason": "Models reproduce several observed human routes and show that selected combinations can delay or suppress escape." }, "inVitroEvidence": { "rating": "strong", "reason": "Human-derived models reproduce allele- and state-specific escape, though prospective treatment prediction remains unproved." }, "reproducibility": { "rating": "none", "reason": "No prospective PDAC programme has reproduced rapid resistance classification followed by matched treatment." }, "commercialConflicts": { "rating": "high", "reason": "The route depends on sponsor-controlled RAS drugs and access to combination compounds." }, "recency": { "rating": "current", "reason": "The hypothesis now includes the August 2026 paired human resistance report and related 2026 mechanism studies." } } }, "layout": { "x": 66, "y": 124 }, "updated": "2026-09-14" }, { "id": "hypothesis-quemliclustat-existing-phase3-gate", "title": "Quemliclustat adds clinically useful survival to gemcitabine and nab-paclitaxel", "shortTitle": "Quemliclustat phase 3 gate", "type": "hypothesis", "status": "testing", "scope": [ "pdac", "transfer-unproven" ], "summary": "Use the completed PRISM-1 enrollment; require all-randomized survival, aligned secondary outcomes and bounded harm before changing treatment or capital allocation.", "content": "Biological rationale. Chemotherapy releases ATP from damaged cells; CD73 converts its breakdown product AMP into adenosine, which suppresses nearby immune cells. Quemliclustat blocks soluble and cell-bound CD73. Supporting evidence. Human paired tumours show reduced adenosine-linked NR4A expression and more T-cell activation after treatment. A 122-person chosen-dose group had median survival of 15.7 months versus 9.8 months in a matched external control, HR 0.634. Contrary evidence. There was no concurrent randomized quemliclustat-free arm, response was 39% versus 41% and PFS was not significantly different. Oleclumab plus chemotherapy missed its main randomized endpoint; oleclumab alone had OS HR 1.26 and its checkpoint combination HR 0.75 with both intervals including no effect. Prediction. Among every PRISM-1 randomized patient, quemliclustat produces OS HR at most 0.80, at least a two-month median survival gain, a 95% confidence interval excluding no effect, PFS HR at most 0.85, and no more than eight percentage points excess serious adverse events or ten points excess grade 3 or worse events. Falsification experiment. Add no cohort. Release the fixed protocol and analysis plan, final assigned denominator, treatment exposure, all-randomized OS, PFS, response, later treatment and harm by arm. Falsify if the main OS test misses, the clinical effect stays above the threshold, harm crosses either line, or benefit appears only in a marker cut chosen after results. Treat NR4A change as a mechanism check, not a success endpoint. Mission cost is £0 before readout and at most £250,000 for independent analysis after data access; expected result timing is the first half of 2027. Major safety risks are chemotherapy toxicity, infection, blood-count suppression, neuropathy, unknown consequences of systemic adenosine blockade and a false-positive survival result from repeated analyses.", "sourceIds": [ "wainberg-2026-quemliclustat-arc8", "coveler-2024-oleclumab-pdac", "clinicaltrials-nct06608927-prism1", "arcus-2026-q2-10q" ], "links": [ { "target": "experiments", "relation": "part-of" }, { "target": "unknown-quemliclustat-survival-contribution", "relation": "resolves" }, { "target": "failure-adenosine-control-and-marker-pdac", "relation": "learns-from" }, { "target": "trial-pdac-adenosine-action-map", "relation": "implemented-by" }, { "target": "company-arcus-biosciences", "relation": "depends-on" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "moderate", "randomised": "strong", "independentReplication": "weak", "sampleSize": "Existing quadruple-masked randomized phase 3 estimated n=610", "effect": "Mission gate: OS HR at most 0.80, at least two median months, 95% CI below 1 and bounded serious harm", "limits": [ "Thresholds are mission choices", "Final enrollment count is not yet posted as actual", "Analysis plan is not public", "Result timing is sponsor guidance", "Mechanism marker is not validated for selection" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The target chain is measurable from extracellular nucleotide processing to immune state." }, "animalEvidence": { "rating": "strong", "reason": "Models support CD73 blockade, discounted because another human product failed overall." }, "inVitroEvidence": { "rating": "strong", "reason": "Quemliclustat blocks adenosine-linked gene activation in human PDAC cell systems." }, "reproducibility": { "rating": "weak", "reason": "No controlled quemliclustat benefit exists and oleclumab did not validate the class." }, "commercialConflicts": { "rating": "high", "reason": "The sponsor and licensing partners own the drug, study and first analysis." }, "recency": { "rating": "current", "reason": "The existing trial is fully enrolled and readout is expected in 2027." } } }, "contentSections": [ "Assumptions — CD73 is a causal immune brake in enough untreated PDAC, quemliclustat reaches that pathway at the phase 3 dose, and randomization balances later care." ], "layout": { "x": 460, "y": 390 }, "updated": "2026-09-15" }, { "id": "hypothesis-randomized-lapc-resection-after-response", "title": "Hypothesis: resection adds useful life only in a locked locally advanced responder state", "shortTitle": "Randomize responder surgery", "type": "hypothesis", "status": "hypothesis", "scope": [ "pdac" ], "summary": "Randomize only after systemic treatment, independent technical agreement and a patient-level state in which both surgery and continued treatment are acceptable.", "content": "Rationale: some initially locally advanced tumours remain confined after systemic treatment and can be completely removed in expert centres, but the observed survival advantage is entangled with the response filter. Proposed test: enrol at diagnosis, but randomize only after at least four months of modern multiagent treatment when there is no distant progression, physical reserve is adequate, two independent expert panels agree that complete visible removal is technically achievable, and the person considers both routes acceptable. Compare surgery within four weeks plus a locked systemic-treatment plan against continued systemic treatment with a locked local-control and later-surgery policy. Stratify allocation by centre, original arterial or venous anatomy, treatment duration, performance status, CA19-9 level and change, imaging change and circulating tumour DNA when measurable. Keep staging-camera findings, biopsy-confirmed spread and abdominal-wash cytology separate. Primary endpoint: overall survival from randomization. Key joint endpoint: quality-adjusted survival, combining time alive with measured function and symptoms. Also measure days outside hospital, pain, treatment received, recurrence route, endocrine and digestive failure, work, 90-day death, major complications and regret. Planning maths: detecting HR 0.75 with two-sided 5% error and 80% power needs about 379 deaths using 4 × (1.96 + 0.84)^2 / ln(0.75)^2. If 80% have an event and 10% are lost or non-evaluable, 379 / 0.80 / 0.90 = 527; round to 540 people. A 120-person feasibility stage must show at least 70% of allocated operations occur, crossover stays below 20%, required fields exceed 90% and control care is not withheld. Stop for 90-day surgical mortality above 5% or more than three points above control, futility for survival, persistent clinically important quality-of-life loss without survival gain, or inability to sustain equipoise. R0, meaning no cancer at the examined cut edge, cannot release the conclusion. Planning cost is £22m over six years because routine treatment remains health-system care. This is an unfunded candidate: it enters the fixed £500m plan only if an existing arm stops or an explicit portfolio review removes at least £22m elsewhere.", "contentSections": [ "Supporting evidence: selected responders can undergo complete visible removal in expert centres after systemic treatment.", "Contrary evidence: the only direct randomized trial is 42 people old, PELICAN found harm without benefit from a different invasive local treatment, and every modern surgery comparison can still be explained by selection.", "Assumptions: a real state exists in which expert teams judge both routes acceptable, enough people accept allocation, current systemic treatment remains available in both groups and centre skill can be measured.", "Prediction: surgery lowers the death hazard by about 25% without a persistent loss of quality-adjusted survival.", "Falsification experiment: reject the hypothesis if randomized survival is futile, surgery cannot be delivered, crossover destroys the contrast or quality-adjusted survival remains worse.", "Major risks include operative death, complications, delayed systemic treatment, permanent digestive or endocrine loss, centre effects and an obsolete control during a six-year trial." ], "sourceIds": [ "stoop-2026-preopanc4-surgery", "dekker-2026-taps-restaging", "verkolf-2026-taps-resection-prediction", "napoli-2025-surgery-versus-oncology", "imamura-2004-locally-invasive-surgery-rct", "seelen-2026-pelican", "theijse-2024-nontherapeutic-laparotomy", "clinicaltrials-nct06714604-duration" ], "links": [ { "target": "experiments", "relation": "candidate-for" }, { "target": "unknown-lapc-surgery-causal-benefit", "relation": "tests" }, { "target": "failure-conversion-surgery-selection-bias", "relation": "repairs" }, { "target": "claim-lapc-response-warrants-surgical-reassessment", "relation": "implements" }, { "target": "hypothesis-selected-local-consolidation", "relation": "specializes" }, { "target": "programme-500m-integrated-mission", "relation": "competes-for-capital-in" }, { "target": "trial-pdac-occult-metastasis-action-map", "relation": "depends-on" }, { "target": "claim-pdac-surgery-is-a-complete-treatment-route", "relation": "depends-on" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "strong", "randomised": "none — absent for the proposed modern question", "independentReplication": "none — absent", "sampleSize": "Proposed n=540 with a 120-person feasibility stage", "effect": "Target death HR 0.75 with no persistent quality-adjusted-life loss", "limits": [ "Equipoise may be difficult after strong surgical beliefs", "Crossovers likely", "No validated systemic-risk classifier", "ctDNA false negatives", "Expertise changes complication risk", "Control treatment can evolve", "Planning assumptions and cost are generated estimates" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "A local-dominant responder state may exist, but scans and blood cannot exclude every distant clone." }, "animalEvidence": { "rating": "weak", "reason": "Models cannot reproduce the clinical selection, operation and recovery system." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Cell response cannot estimate added useful life from surgery." }, "reproducibility": { "rating": "weak", "reason": "The trial tests a generated rule that has not been randomized." }, "commercialConflicts": { "rating": "moderate", "reason": "Independent allocation, outcome review and centre-level reporting are needed to constrain procedural incentives." }, "recency": { "rating": "current", "reason": "Design responds to evidence and registries through September 2026." } } }, "layout": { "x": 118, "y": 188 }, "updated": "2026-09-15" }, { "id": "hypothesis-regional-route-completion", "title": "Fix the largest measured route loss before importing another treatment trial", "shortTitle": "Regional route completion", "type": "hypothesis", "status": "hypothesis", "scope": [ "pdac", "cholangiocarcinoma", "gallbladder", "hcc", "hepatobiliary", "pancreatic-all", "transfer-unproven" ], "summary": "A site-specific repair of the largest measured delay or non-delivery step may increase completed treatment and survival more than adding another drug protocol to a route most patients never reach.", "content": "Biological rationale and supporting evidence: each HPB cancer has a time-limited route from risk or symptoms to diagnosis, stage, organ-function assessment, referral and treatment. African pancreatic and HCC evidence and Latin American CCA and pancreatic cohorts show large late-stage and treatment losses. If one repairable step removes at least 15 percentage points of otherwise eligible patients, fixing that step should increase completed treatment before any new therapy is invented. Evidence against or missing: every current comparison is observational or selected after referral; stage, liver function, tumour biology, wealth, travel and treatment capacity differ; a faster route may shift delay downstream; and treatment completion may rise without better survival or quality of life. Assumptions: consecutive capture can find at least 80% of eligible cases, non-delivery can be classified reliably, one dominant step can be changed locally, capacity can absorb the patients gained, and the bundle does not delay or displace other care.\n\nPrediction: after a £5m, twelve-month common registry stage, a site-specific repair will produce at least a 10 percentage-point absolute increase in completed recommended treatment when a baseline loss of at least 15 points is reproduced at two sites or in 100 patients. Secondary predictions are earlier stage, shorter time at the targeted step and a survival or quality-of-life direction consistent with the completion gain. The test takes about three years after registry readiness. Falsification experiment: randomize clinics or districts in a stepped sequence to the measured repair, with contemporaneous usual care, locked eligibility and independent review of non-delivery. Allowed parts are navigation, transport support, fixed imaging and pathology slots, automatic multidisciplinary referral, treatment scheduling and stock assurance. The bundle must target the measured loss; it cannot be a generic education campaign. The primary endpoint is the proportion of all consecutive eligible cases completing recommended treatment within the locked time window. Secondary endpoints are stage, time at each step, quality of life, toxicity, twelve- and twenty-four-month survival, household cost and loss shifted elsewhere. Stop before the £15m trial if capture or twelve-month follow-up falls below 80% or required-field missingness exceeds 20%. Stop the trial if the completion gain is below 5 points, delay merely moves downstream or a pooled result hides a failed site or cancer. Major safety risks are harmful treatment of people who would not benefit, diagnostic and surgical complications, treatment toxicity, overwhelmed capacity, loss of other services, financial harm and inequitable access to the new fast route. Total capped cost is £5m + £15m = £20m.", "sourceIds": [ "gismalla-2026-africa-pancreatic-review", "sobnach-2024-ssa-hcc-systematic-review", "sobnach-2025-south-africa-hcc-disparities", "sobnach-2026-global-hcc-disparities", "da-fonseca-2024-latam-cca", "peixoto-2025-gaspar-results" ], "links": [ { "target": "experiments", "relation": "part-of" }, { "target": "unknown-regional-route-loss", "relation": "tests" }, { "target": "trial-regional-route-completion-action-map", "relation": "operationalized-by" }, { "target": "programme-unconstrained-global-system", "relation": "could-be-funded-by" }, { "target": "failure-ssa-hcc-curative-route-not-reaching-patients", "relation": "succeeds-where" } ], "evidence": { "humanProspective": "none for the bundle", "humanRetrospective": "moderate for the route failure", "randomised": "none", "independentReplication": "none for the proposed intervention", "sampleSize": "Proposed multi-site route experiment after a one-year registry gate", "effect": "Generated target: at least 10 percentage-point absolute improvement in completed recommended treatment", "limits": [ "Generated hypothesis", "The dominant step may differ by site and cancer", "Treatment completion is not survival", "Stepped designs can be affected by time trends", "Bundles make component effects hard to separate", "Capacity gains may be temporary without local funding" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Earlier diagnosis and completed treatment preserve access to interventions that depend on stage and organ function." }, "animalEvidence": { "rating": "not-applicable", "reason": "The intervention acts on care delivery." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Laboratory results do not predict route completion." }, "reproducibility": { "rating": "none", "reason": "The exact measured-loss intervention has not been tested in these networks." }, "commercialConflicts": { "rating": "moderate", "reason": "Transport, diagnostics and treatment supply create vendor interests that require open procurement and reporting." }, "recency": { "rating": "current", "reason": "The hypothesis is built from current 2024-2026 regional evidence." } } }, "updated": "2026-09-14" }, { "id": "hypothesis-selected-local-consolidation", "title": "Hypothesis: local consolidation can help only after systemic-risk selection", "shortTitle": "Selected local consolidation", "type": "hypothesis", "status": "hypothesis", "scope": [ "pdac", "colorectal", "transfer-unproven" ], "summary": "Repeated local-therapy failures may reflect uncontrolled occult systemic disease; colorectal trials show that long survival appears only after tight organ-confined selection.", "content": "Biological rationale: autopsy and recurrence studies show some PDAC patients die predominantly from local disease, while randomized radiation and ablation programmes often enrolled clinically defined locally advanced disease without a validated systemic-risk gate. Colorectal evidence supplies a bounded transfer test. CLOCC and TransMet produced long-survival tails in narrow liver-only groups, whereas EPOCH improved liver progression without extending life in broader second-line disease. Contrary evidence in PDAC: CROSSFIRE, SABER, CONKO-007, A021501, LAP07 and PELICAN did not establish useful-life benefit from their tested local additions. PELICAN is especially direct: ablation after non-progression produced survival HR 1.07, raised serious events from 11% to 27% and worsened quality of life. Resection must be tested as its own modality; it cannot inherit an ablation or radiation result. Assumptions: duration of response, serial blood, imaging and physical reserve can define a lower-systemic-risk state, and local treatment does not destroy later systemic treatment. Prediction: among patients with sustained systemic control and anatomically dominant local disease, one defined local action reduces death or quality-adjusted-life loss without an offsetting rise in distant failure or treatment harm. The first clean surgery version is the 540-person response-gated resection comparison. Other modalities require separate randomization. Planning estimate: £22m for one selected local test, not several pooled modalities. Stop if extra-field relapse within 12 months exceeds 35%, treatment death exceeds 3% over control, or local control improves without survival, treatment-free-time or quality-of-life separation. Major risks are procedure death, fistula or bleeding, selection bias and a false-negative dissemination test.", "contentSections": [ "Falsification experiment: retain allocation and all crossovers, then reject the route if local control improves without survival or quality-adjusted-life gain, or if distant failure or treatment harm erases the benefit." ], "sourceIds": [ "timmer-2024-crossfire", "yoo-2026-saber", "fietkau-2025-conko007", "katz-2022-a021501", "hammel-2016-lap07", "seelen-2026-pelican", "stoop-2026-preopanc4-surgery", "dekker-2026-taps-restaging", "winter-2013-smad4-failure-patterns", "ruers-2017-clocc-crc-liver-ablation", "adam-2024-transmet", "mulcahy-2021-epoch-radioembolization" ], "links": [ { "target": "experiments", "relation": "part-of" }, { "target": "failure-local-intensification-pdac-survival-not-shown", "relation": "succeeds-where" }, { "target": "claim-pdac-death-can-be-local-or-metastatic", "relation": "depends-on" }, { "target": "unknown-ctdna-actionability", "relation": "depends-on" }, { "target": "unknown-crc-liver-local-cure-selector", "relation": "informed-by" }, { "target": "failure-crc-radioembolization-pfs-without-os", "relation": "learns-from" }, { "target": "programme-100m-randomized-platform", "relation": "tests" }, { "target": "hypothesis-randomized-lapc-resection-after-response", "relation": "specialized-by" }, { "target": "failure-conversion-surgery-selection-bias", "relation": "constrained-by" }, { "target": "trial-lapc-conversion-surgery-action-map", "relation": "constrained-by" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "moderate", "randomised": "strong", "independentReplication": "strong", "sampleSize": "Six prior controlled local programmes; proposed surgery version n=540", "effect": "Target: survival HR ≤0.75 without persistent quality-adjusted-life loss", "limits": [ "Generated selection hypothesis", "Prior randomized PDAC evidence is negative for unselected or differently selected strategies", "Colorectal biology does not prove PDAC benefit", "Classifier does not exist", "Local modalities are not interchangeable", "Molecular false negatives could erase benefit", "Stop thresholds are programme choices", "No capital is currently released" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "A local-dominant state may exist, but scan and blood negativity cannot exclude occult systemic PDAC." }, "animalEvidence": { "rating": "moderate", "reason": "Models support differing spread patterns but cannot validate the human selector." }, "inVitroEvidence": { "rating": "weak", "reason": "Cell response cannot establish metastatic geography or procedure net benefit." }, "reproducibility": { "rating": "weak", "reason": "Selected colorectal studies support the principle, while six PDAC local-treatment programmes remain negative or bounded." }, "commercialConflicts": { "rating": "moderate", "reason": "Radiation, device and procedure programmes create incentives to overstate local-control signals." }, "recency": { "rating": "current", "reason": "The hypothesis incorporates the 2026 PELICAN and PREOPANC-4 results." } } }, "layout": { "x": 58, "y": 132 }, "updated": "2026-09-15" }, { "id": "hypothesis-smart-vendor-independent-randomized-gate", "title": "Rebuild the SMART survival test around performance, not one machine", "shortTitle": "Vendor-independent SMART gate", "type": "hypothesis", "status": "testing", "scope": [ "pdac", "locally-advanced", "radiation", "trial-recovery" ], "summary": "A short recovery stage can decide whether the dormant phase 3 design can become a durable multi-system trial without repeating selected single-arm evidence.", "content": "Biological rationale. Daily soft-tissue imaging, target tracking and same-day replanning may let high-dose radiation control local disease with less bowel injury. Supporting evidence. A 136-person prospective study delivered five fractions with no definitely attributed acute severe gastrointestinal toxicity, and several centres reproduce plan adaptation. Contrary evidence. Survival is uncontrolled, late severe harm reaches 11.5% under the broad possible-attribution rule, CIPHER found no local-technique survival winner, LAP-ABLATE has not started, MASPAC stopped at nine and the machine sponsor entered bankruptcy. Assumptions. Trial rights and dated plans can be transferred; more than one maintained radiation system can meet a common measurable performance specification; and centres will randomize selected non-progressors to continued systemic care with or without local radiation. Prediction. A six-month recovery stage will reconcile all 136 phase 2 participants and its 133-person registry count, release unfiltered late and postoperative harms, secure at least ten centres across at least two maintained system families, and pass blinded plan tests in at least 95% of cases with cross-system target and bowel-dose differences no larger than 5%. Falsification experiment. Spend at most £500,000 from existing radiation and statistics lines and enroll nobody. Recover the phase 2 record, LAP-ABLATE protocol, analysis code and contracts; define target visibility, gating, plan-adaptation time, dose constraints, data export and service continuity as measurable requirements; and run common phantom plus archived-case tests. Stop if rights fail, only one unsupported platform qualifies, cross-system dose differs by more than 5%, the phase 2 denominator cannot be reconciled, complete harms exceed the stated gate or ten centres will not randomize. If all pass, reprice the 267-person design before recruitment and require at least 99% survival accounting, positive quality-adjusted survival, no more than a five-point increase in grade 3 or worse harm, preserved chemotherapy and no postoperative death excess. The mission commits £0 to a new trial now; any future trial must replace equal capital and cannot be vendor-sponsored. Major safety risks are stomach or bowel ulcer, bleeding or perforation, postoperative vascular or healing complications, treatment delay, imaging exclusion, machine downtime and data lock-in.", "sourceIds": [ "parikh-2023-smart-phase2-safety", "chuong-2024-smart-phase2-survival", "clinicaltrials-nct03621644-smart", "clinicaltrials-nct05585554-lap-ablate", "clinicaltrials-nct05114213-maspac", "viewray-2023-q1-10q", "viewray-2023-chapter11-8k", "timmer-2024-crossfire" ], "links": [ { "target": "experiments", "relation": "part-of" }, { "target": "unknown-smart-added-survival-and-surgery-harm", "relation": "tests" }, { "target": "failure-smart-trial-and-vendor-dependency", "relation": "repairs" }, { "target": "trial-pdac-smart-action-map", "relation": "implemented-by" }, { "target": "company-viewray", "relation": "learns-from" }, { "target": "programme-5m-decision-core", "relation": "constrained-by" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "weak", "independentReplication": "none", "sampleSize": "Existing n=136 phase 2; zero new participants in recovery; possible n=267 trial remains blocked; recovery capped at £500,000", "effect": "Convert a vendor-bound dormant plan into a testable or retired treatment question", "limits": [ "Trial rights may not transfer", "Cross-system performance may not be equivalent", "Ten centres may not commit", "The original sample size may be obsolete", "Archived planning tests cannot establish survival", "Performance thresholds are mission choices", "Capital displacement has not been selected" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The gate preserves the measured imaging-and-dose mechanism while testing patient benefit separately." }, "animalEvidence": { "rating": "not-applicable", "reason": "This is a clinical delivery and governance experiment." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Cell response cannot qualify a clinical machine network." }, "reproducibility": { "rating": "weak", "reason": "The proposed common archived-case test is required before multi-system recruitment." }, "commercialConflicts": { "rating": "high", "reason": "No vendor may sponsor or control analysis because the prior plan failed with its machine maker." }, "recency": { "rating": "current", "reason": "The design follows 2025 trial termination and current dormant phase 3 status." } } }, "layout": { "x": 694, "y": 618 }, "updated": "2026-09-15" }, { "id": "hypothesis-tigerpac-existing-final-gate", "title": "The existing TIGeR-PaC final record can decide package use without another cohort", "shortTitle": "TIGeR-PaC final gate", "type": "hypothesis", "status": "testing", "scope": [ "pdac", "trial-delivery", "transfer-unproven" ], "summary": "Use the enrolled 114-person survival experiment; require a fixed final analysis and count every systemic and procedure harm before adoption or platform transfer.", "content": "Biological rationale. Temporary local artery pressure may increase pancreatic-region exposure while lowering active gemcitabine in blood, but the treatment package also removes nab-paclitaxel and changes schedule. Supporting evidence. The first 45 randomized patients had median post-randomization survival of 16 versus 10 months, and the trial has passed two independent monitoring reviews. Contrary evidence. P=0.051 did not cross the interim stopping rule, no hazard ratio or confidence interval was reported, the plan changed after launch, the second interim estimate is withheld, direct tumour delivery is unmeasured and early procedure safety includes sepsis in 4/23. Assumptions. At least the 114 planned randomized assignments exist, the final 86-death analysis remains valid after two interim looks, follow-up is balanced, and quality-of-life and harm collection are complete enough to compare total patient burden. Prediction. Under the final locked plan, the intra-arterial package will either improve overall survival with hazard ratio at most 0.75 and a 95% confidence interval below 1.00, or keep survival within a mission noninferiority bound whose upper 95% hazard-ratio limit is at most 1.20 while reducing severe treatment harm by at least 15 percentage points and improving a fixed quality-of-life measure. These are mission adoption thresholds, not claims about the trial's regulator-agreed boundaries. Falsification experiment. Enrol zero new patients. Obtain the original and changed protocol and analysis plan, interim alpha-spending rule, regulator correspondence and complete all-randomized record after database lock. Reconcile the final count, then analyse survival, progression, local and distant failure, dose, later treatment, hospital time, quality of life and all systemic and procedure harms. Spend at most £250,000 from the existing data and statistics line only after participant-level access and independent-analysis rights are signed. Stop adoption if the trial misses its locked primary rule, neither mission benefit gate passes, missing vital status exceeds 5% or differs by more than five points between arms, or procedure harm erases the systemic-toxicity gain. Do not buy equity or fund another trial. Do not generalize the platform to another drug or cancer until human tumour or local-vein sampling shows greater active-drug exposure at the target. Major safety risks are sepsis, bleeding, clots, artery or organ ischemia, catheter injury, low blood counts and biased undercounting in the open-label comparison.", "sourceIds": [ "clinicaltrials-nct03257033-tigerpac", "renovorx-2023-tigerpac-interim", "renovorx-2025-10k", "renovorx-2026-full-enrollment", "novelli-2026-tigerpac-pk", "hatoum-2024-tamp-rr1-rr2" ], "links": [ { "target": "experiments", "relation": "part-of" }, { "target": "unknown-tigerpac-package-survival-and-delivery", "relation": "tests" }, { "target": "failure-tigerpac-component-and-interim-inference", "relation": "repairs" }, { "target": "trial-pdac-intraarterial-delivery-action-map", "relation": "implemented-by" }, { "target": "programme-500m-integrated-mission", "relation": "constrained-by" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "weak", "randomised": "strong", "independentReplication": "weak", "sampleSize": "Existing target n=114 randomized and 86 deaths; no new participants; independent analysis capped at £250,000", "effect": "One existing record can reject, adopt or narrow the package while preventing an unjustified platform claim", "limits": [ "Thresholds are mission choices", "Final plan is unavailable", "Open-label care can affect non-survival outcomes", "Package components cannot be separated", "Direct tumour delivery requires separate measurement" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Human blood handling changes in the predicted direction, while target-tissue delivery is still unmeasured." }, "animalEvidence": { "rating": "moderate", "reason": "Animal transport work supports the measurement plan, not patient benefit." }, "inVitroEvidence": { "rating": "strong", "reason": "Gemcitabine action is known; route and package value are the live tests." }, "reproducibility": { "rating": "weak", "reason": "The hypothesis depends on one sponsor-controlled phase 3 record." }, "commercialConflicts": { "rating": "high", "reason": "Independent access and analysis are needed because the sponsor owns the device and controls the readout." }, "recency": { "rating": "current", "reason": "Enrollment closed and the final event-driven result is projected for 2027." } } }, "layout": { "x": 538, "y": 462 }, "updated": "2026-09-15" }, { "id": "hypothesis-tripp-ffx-independent-analysis-gate", "title": "Independently analyze all 88 TRIPP-FFX assignments before any new OncoSil trial", "shortTitle": "TRIPP-FFX analysis gate", "type": "hypothesis", "status": "testing", "scope": [ "pdac", "locally-advanced", "trial-recovery", "brachytherapy" ], "summary": "The randomized human experiment already exists; a capped independent analysis can decide the route without buying a device, company stake or another cohort.", "content": "Biological rationale. Silicon-bound phosphorus-32 can concentrate a 100-gray beta-radiation dose inside a tumour while FOLFIRINOX treats disease throughout the body. Supporting evidence. Pilot implantation localized in 40/42 cases, all-enrolled local control reached 82%, 10/50 reached resection and an adjusted comparison estimates 189 extra days within 30 months. Contrary evidence. All modern efficacy evidence is uncontrolled or selected; an older phosphorus-32 randomization changed tumours but added serious events and no survival. Assumptions. OncoSil and sites will provide the locked assignment key, dated plan, dose calculations, implantation and scan files, chemotherapy delivery, symptoms, quality of life, operations and vital status with independent publication rights. Prediction. Analysis of all 88 will show whether OncoSil improves 16-week local control by at least 15 points and whether that maps to either a hazard ratio at or below 0.75 for progression or at least two additional quality-adjusted months without a five-point rise in severe harm. Falsification experiment. Enroll zero people, buy no device or equity and spend at most £350,000. Freeze the all-assigned analysis, competing local and distant failure, death handling and missing-data rules before receiving arm labels. Require 100% assignment reconciliation, 95% vital status, 90% dose and implantation records and 90% quality-of-life coverage through sixteen weeks. Stop the branch if access is refused, severe harm rises five points, chemotherapy delivery falls ten points, local control improves without patient-valued time, or the survival interval excludes the target effect. Because final follow-up is planned through July 2027, issue a provisional result now and a locked survival update later. Major risks are pancreatitis, bleeding, infection, ulcer or perforation, off-target radiation, chemotherapy delay, surgical harm and selective reporting.", "sourceIds": [ "clinicaltrials-nct05466799-tripp-ffx", "ross-2022-panco-p32", "lim-2026-p32-landmark", "rosemurgy-2008-p32-randomized" ], "links": [ { "target": "experiments", "relation": "part-of" }, { "target": "unknown-oncosil-randomized-survival-and-delivery", "relation": "tests" }, { "target": "failure-p32-liquefaction-survival-and-reporting-gap", "relation": "repairs" }, { "target": "trial-pdac-p32-action-map", "relation": "implemented-by" }, { "target": "company-oncosil-medical", "relation": "depends-on" }, { "target": "programme-5m-decision-core", "relation": "constrained-by" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "moderate", "randomised": "moderate — dataset exists, result pending", "independentReplication": "weak", "sampleSize": "All 88 existing assignments; zero new participants; cap £350,000", "effect": "Convert the completed randomized primary record into a go or stop decision", "limits": [ "Final survival follow-up incomplete", "Power for survival is limited", "Quality-of-life missingness may exceed threshold", "Company access unconfirmed", "Thresholds and cap are mission choices", "Local control may not translate", "A positive phase 2 result may still need replication" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The analysis links measured localized dose to clinical outcomes on fixed chemotherapy." }, "animalEvidence": { "rating": "not-applicable", "reason": "The needed evidence is already in a randomized human study." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "No bench experiment should delay analysis of the human comparison." }, "reproducibility": { "rating": "moderate", "reason": "Blinded independent analysis and later locked update can test robustness." }, "commercialConflicts": { "rating": "high", "reason": "Support is conditional on independent custody and negative-result publication." }, "recency": { "rating": "current", "reason": "The gate acts after January 2026 primary completion." } } }, "layout": { "x": 922, "y": 846 }, "updated": "2026-09-15" }, { "id": "hypothesis-tumour-verified-selective-autophagy-blockade", "title": "Hypothesis: tumour-confirmed selective autophagy blockade can expose RAS-treated PDAC", "shortTitle": "Tumour autophagy gate", "type": "hypothesis", "status": "hypothesis", "scope": [ "pdac", "pan-cancer", "transfer-unproven" ], "summary": "A selective inhibitor may add to RAS treatment only if it blocks the intended recycling process inside tumour, preserves RAS-drug exposure and produces a controlled patient effect.", "content": "Biological rationale. RAS is the dominant growth-control circuit in PDAC. Models indicate that inhibiting its ERK or MEK branch can increase dependence on autophagy, the cell's internal recycling process. HCQ combinations failed clinically, but HCQ acts at the lysosome and did not produce consistent blood-cell evidence of blockade; tumour blockade was not established. A direct inhibitor of ULK1/2, proteins that help start autophagy, is closer to the proposed mechanism. Supporting evidence. Early inlexisertib data showed direct-target change in blood cells, and models support the adaptive-recycling route; neither establishes the required tumour or patient effect. Contrary evidence. Three HCQ programmes found no useful response signal, the randomized combination reduced the ERK-inhibitor dose, trial-derived organoids showed no clear synergy, and the direct inlexisertib programme ended without a final result. Assumptions. A tolerable selective drug can reach tumour, paired biopsies can measure both its direct target and stopped recycling, RAS-drug dose can be preserved, and a patient subset depends materially on this route rather than several parallel escape routes. Prediction. In that subset, a selective autophagy inhibitor will reduce its direct tumour target signal and increase blocked-recycling cargo, while deeper blockade will predict tumour control only when at least 80% of the planned RAS-inhibitor dose is delivered. Falsification experiment in two stages. Stage zero uses no new patient exposure: recover NCT04892017 participant-level data and paired samples, with disease, dose, drug exposure, tumour and blood markers, lesion response and all missing denominators joined. Stop if pancreatic combination data do not exist or tumour measurements are absent; that means the programme generated no test, not a negative mechanism result. A replacement drug may enter a small measurement study only after two independent laboratories reproduce combination killing in patient-derived models at human-achievable exposure. In that study, at least 80% of paired biopsies must be usable; at least 70% of exposed tumours must show both 50% or greater direct-target suppression and the prespecified blocked-cargo change; and at least 75% of patients must receive 80% or more of planned RAS-drug dose through two cycles. Failure of any gate falsifies the actionable hypothesis and stops expansion. A response gate is then at least 4 of the first 20 response-evaluable patients controlled for four months, with everyone treated retained in a second all-patient denominator. This is only a signal gate. Benefit requires a randomized add-on comparison with identical RAS treatment in both groups and overall plus quality-adjusted survival as the decision outcomes. Estimated cost: £0 now; existing data-recovery and resistance-pharmacology teams own stage zero. Any later study must displace equal portfolio capital. Main risks are normal-cell recycling injury, liver or eye toxicity, reduced RAS-drug dose, false target measurement, tumour sampling bias and simultaneous escape routes that make single-partner blockade irrelevant.", "sourceIds": [ "surana-2025-erk-hcq-pdac", "clinicaltrials-nct05221320-ulixertinib-hcq", "manji-2023-mekiauto", "clinicaltrials-nct04214418-mekiauto", "clinicaltrials-nct04892017-inlexisertib", "deciphera-2022-dcc3116-phase1", "alistar-2017-devimistat-phase1", "philip-2024-avenger500" ], "links": [ { "target": "experiments", "relation": "part-of" }, { "target": "claim-hcq-combinations-fail-pdac-not-autophagy", "relation": "constrained-by" }, { "target": "failure-metabolic-repurposing-pdac", "relation": "repairs" }, { "target": "unknown-selective-autophagy-pdac-contribution", "relation": "tests" }, { "target": "trial-pdac-autophagy-action-map", "relation": "gated-by" }, { "target": "hypothesis-prospective-ras-resistance-routing", "relation": "could-supply-partner-to" } ], "evidence": { "humanProspective": "weak — selective-inhibitor data are incomplete and not PDAC-specific", "humanRetrospective": "none", "randomised": "none for the selective inhibitor; the relevant HCQ combination missed its randomized signal gate", "independentReplication": "none", "sampleSize": "Stage-zero recovery from n=91 inlexisertib study; any later measurement stage begins with 20 response-evaluable PDAC patients only after laboratory and tumour-pharmacology gates", "effect": "Generated gate: ≥80% usable paired biopsies; tumour target and cargo change in ≥70%; ≥80% RAS-drug dose delivered in ≥75%; four-month control in ≥4/20 before any comparative expansion", "limits": [ "Thresholds are programme decisions, not validated biomarkers", "Recovery may show that decisive samples were never collected", "ULK1/2 target inhibition may not equal complete autophagy blockade", "A small response gate cannot prove benefit", "No current budget or suitable agent is assigned" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "The hypothesis follows a known adaptive route but requires proof in human tumour rather than blood or models." }, "animalEvidence": { "rating": "moderate", "reason": "Direct ULK inhibition has model support, which has not yet survived a complete human test." }, "inVitroEvidence": { "rating": "weak", "reason": "Sponsor models support combinations; trial-derived organoids did not support the HCQ pair." }, "reproducibility": { "rating": "weak", "reason": "No independent selective-inhibitor human dataset exists." }, "commercialConflicts": { "rating": "high", "reason": "Selective-inhibitor evidence and missing final results originate with the drug sponsor." }, "recency": { "rating": "current", "reason": "Designed after the 2025 randomized result and 2026 inlexisertib termination." } } }, "layout": { "x": 64, "y": 145 }, "updated": "2026-09-15" }, { "id": "hypothesis-vcn01-replication-to-survival-gate", "title": "Hypothesis: tumour-replicating VCN-01 adds useful survival without blocking chemotherapy", "shortTitle": "VCN-01 survival gate", "type": "hypothesis", "status": "hypothesis", "scope": [ "pdac" ], "summary": "VCN-01 may add value through tumour-cell lysis and barrier opening, but only if clean all-randomized evidence links repeatable tumour pharmacology to longer useful life.", "content": "Biological rationale. VCN-01 is a tumour-killing adenovirus with two coupled functions. Broken RB1 control should permit selective viral copying and tumour-cell rupture; virus-made PH20 should cut hyaluronan and lower the transport barrier for chemotherapy and immune cells. Supporting human evidence. Virus appeared in treated tumour biopsies, and the reported VIRAGE FAS favoured VCN-01 for survival and progression. Contrary evidence. Pelareorep changed immune markers without clinical benefit; PEGPH20 removed hyaluronan without extending survival; 26% of phase 1 candidates were excluded for baseline Ad5 antibodies; severe and fatal phase 1 toxicity occurred; and VIRAGE's complete denominator and final analysis rules are unavailable. Assumptions. Viral copying and PH20 occur across primary and metastatic lesions, chemotherapy remains deliverable, baseline immunity does not restrict most future patients, and the phase 2 result is not created by exclusions or imbalance. Prediction. A confirmed benefit should appear in all-randomized survival, not only response, blood virus or people who remain well enough for later doses. It should be strongest where repeated blood viral peaks and PH20 indicate activity, but that relation must be fixed before outcomes are analysed. Falsification sequence. Stage 0 costs £0: recover VIRAGE. Go requires at least 99% randomized outcome accounting, independently reproduced all-randomized OS HR at or below 0.75 with its 95% interval excluding 1.0 or a documented regulator-agreed success rule fixed before outcome access, no missing-outcome scenario that moves HR above 0.85, no severe-toxicity increase above ten points and no chemotherapy dose loss above ten points. Stage 1 is the sponsor-funded six-person VIRAGE2 safety and dosing check: at least five of six must receive three doses and at least 80% planned chemotherapy, with repeat blood viral activity and no treatment-related death; it cannot test efficacy. Stage 2 is a double-blind placebo-controlled phase 3 with all-randomized OS primary, quality-adjusted survival, complete participant flow, baseline liver-metastasis and Ad5-antibody strata, and an independent analysis copy. Stop if the sponsor withholds data, selects efficacy by treatment completion, changes alpha after access, cannot finance or supply the trial, or the phase 3 crosses harm or futility boundaries. Estimated time is months for data recovery, to Q3 2027 for initial VIRAGE2 pharmacology, and at least four to six years for mature phase 3 survival. Mission capital remains £0 until Stage 0 and 1 pass; any later cofunding must displace equal lower-value capital and buy public data rights. Main safety risks are acute inflammatory reactions, fever, liver-enzyme injury, infection, low blood counts, enterocolitis, treatment-related death, added chemotherapy loss and immune exclusion of people with high baseline Ad5 antibodies.", "sourceIds": [ "noonan-2016-pelareorep-pdac", "bazan-peregrino-2021-vcn01-intratumour", "garcia-carbonero-2022-vcn01-intravenous", "clinicaltrials-nct05673811-virage", "theriva-2023-virage-protocol", "theriva-2025-virage-topline", "esmo-2025-virage-2216mo", "ctis-2026-virage2", "theriva-2026-q2-10q" ], "links": [ { "target": "experiments", "relation": "part-of" }, { "target": "claim-vcn01-randomized-signal-not-survival-proof", "relation": "supported-by" }, { "target": "failure-virage-analysis-and-survivor-selection", "relation": "constrained-by" }, { "target": "failure-oncolytic-virus-pelareorep", "relation": "learns-from" }, { "target": "unknown-vcn01-all-enrolled-survival", "relation": "tests" }, { "target": "trial-vcn01-action-map", "relation": "gated-by" }, { "target": "company-theriva-biologics", "relation": "requires-access-from" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "not-applicable", "randomised": "moderate", "independentReplication": "none", "sampleSize": "Uses existing VIRAGE n=112 and VIRAGE2 n=6; future phase 3 unpriced and unstarted", "effect": "Generated staged recovery, dosing and confirmation gates", "limits": [ "Thresholds are mission decisions", "VIRAGE2 cannot test survival", "Phase 3 size and cost are not public", "Product access and public-data rights are unnegotiated" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "The mechanism is directly observed in human tumour, but two separate anti-stroma precedents warn that delivery change need not extend life." }, "animalEvidence": { "rating": "moderate", "reason": "Models support the coupled lysis and transport claim." }, "inVitroEvidence": { "rating": "moderate", "reason": "VCN-01 function can be measured after access to tumour cells." }, "reproducibility": { "rating": "none", "reason": "No independent randomized VCN-01 study exists." }, "commercialConflicts": { "rating": "high", "reason": "The hypothesis depends on a company-controlled product and unpublished data." }, "recency": { "rating": "current", "reason": "The stages follow the live six-person trial and current phase 3 financing state." } } }, "layout": { "x": 178, "y": 98 }, "updated": "2026-09-15" }, { "id": "investment", "title": "Investment and capital efficiency", "shortTitle": "Investment", "type": "investor", "status": "supported", "scope": [ "pdac", "pnet", "cholangiocarcinoma", "gallbladder", "hcc", "colorectal", "pan-cancer" ], "summary": "Capital is reconstructed in bounded ledgers: $2.0067bn of observed US federal pancreatic-cancer funding, EUR 312.720m of EU title-visible target-cancer commitments and £24.106m of UK title-visible pancreatic-cancer commitments remain separate.", "content": "The largest reproducible decade ledger is US public funding. NCI tagged $1.9407bn to pancreatic cancer across fiscal years 2015–2024; DoD's dedicated pancreatic programme appropriated another $66m in FY2020–FY2024. The common-window observation is $2.0067bn = $1.9407bn + $66m. It is not PDAC-only and remains incomplete because NCI categories overlap, some values are estimates and other institutes or broad programmes are omitted. A strict title query of official CORDIS project files adds a separate EUR 312.720m lower-bound commitment across 159 Horizon 2020 and Horizon Europe projects starting from 2015 in the target disease set. A reproducible UKRI title query adds a separate £24.106m lower-bound commitment for explicitly named pancreatic-cancer projects starting from 2015. Neither title ledger can be added to the dollar ledger: they use different disease scopes, currencies and accounting bases, miss broad awards, and include the full value of some multi-cancer projects. Charity accounting remains separate: PanCAN reported $249m cumulative research investment by June 2025, FY2025 Research and Scientific Affairs expense of $13.823m and a narrower $6.494m Research line; Lustgarten reported more than $301m cumulative grants since 1998 and $13.85m of 2025 commitments; Pancreatic Cancer UK reported £17.5m cumulative investment and £2.552m in new 2024/25 grants. These figures mix lifetime totals, annual expense and commitments, so summing them would double-count and fabricate precision.\n\nCommercial capital uses separate ledgers. Selected recent disclosed financing events total $115m: PAQ $77m, Valar $22m, Mainz $6m and Adenocyte $10m. Investor links now show the named PAQ and Valar syndicates. A separate historical record shows Third Rock's $45m Revolution Medicines launch in 2015, about 11 years before PDAC approval, but the original platform differed and the money cannot be assigned to daraxonrasib. Cosomil adds ¥499m of NEDO public startup support while its private round amounts remain undisclosed. Jacobio's $100m AstraZeneca upfront is deal cash; a further $1.915bn is contingent and excluded. GRAIL's $7.1bn acquisition price is ownership consideration, not R&D. Company R&D is another boundary: Natera reported $1.348926bn across 2023–2025 but did not allocate it by product or disease; Revolution Medicines reported $738.9m in the first half of 2026, including $227.8m identified as third-party daraxonrasib cost; GRAIL reported $95.45m in the same half-year. Financing can later fund R&D, so those columns cannot be added without double counting.\n\nCapital efficiency can be ranked only by evidence unit: randomized overall survival, recurrence reduction, a validated treatment rule, decisive negative knowledge or reusable trial infrastructure. It cannot be inferred by dividing public funding by population survival because funding is overlapping, international, lagged and not causally assigned. The observable pattern is still informative: broad detection and stromal programmes attracted large capital before utility or survival proof, while common-RAS treatment has now produced randomized survival evidence. Underfunded relative to decision value are longitudinal resistance measurement, residual-disease action, independent replication and recovery of failed-trial specimens. Full university subsidy, public funding outside the measured ledgers, pharma allocation and most private venture history remain unobservable; the atlas labels that uncertainty rather than inventing a global total.", "sourceIds": [ "nci-2026-research-funding", "nci-2023-budget-factbook", "nci-2022-budget-factbook", "cdmrp-2026-pcarp-funding", "cordis-h2020-target-cancer-title-query-2026", "cordis-horizon-target-cancer-title-query-2026", "ukri-gtr-pancreatic-title-query-2026", "pancan-2025-audited-financials", "lustgarten-2025-grants", "pcuk-2025-annual-report", "revolution-medicines-2026-q2-10q", "ftc-2024-illumina-grail", "sasieni-2026-nhs-galleri", "alagesan-2026-sig12d-loder", "ctis-2026-sil204-phase23", "silexion-2026-sil204-site-initiation", "silexion-2026-halfyear-filing", "bazan-peregrino-2021-vcn01-intratumour", "garcia-carbonero-2022-vcn01-intravenous", "clinicaltrials-nct05673811-virage", "theriva-2023-virage-protocol", "theriva-2025-virage-topline", "esmo-2025-virage-2216mo", "ctis-2026-virage2", "theriva-2026-q2-10q", "richards-2006-ci994-pdac", "heumann-2022-azacitidine-pdac", "sohal-2020-thu-decitabine-pdac", "baretti-2024-entinostat-nivolumab-pdac", "clinicaltrials-nct03250273-entinostat-nivolumab", "safyan-2026-azacitidine-pembrolizumab-pdac", "clinicaltrials-nct03264404-azacitidine-pembrolizumab", "rodon-2024-amg193-prmt5", "matsumoto-2026-mtap-pancreatic", "pavlick-2026-mtap-genomics", "clinicaltrials-nct05732831-vopimetostat", "tango-2025-vopimetostat-monotherapy", "clinicaltrials-nct06922591-vopimetostat-ras", "tango-2026-vopimetostat-ras-data", "tango-2026-q2-10q", "panbela-2024-q3-10q", "panbela-2025-note-default", "sec-2025-panbela-registration-abandoned", "clinicaltrials-nct05254171-aspire", "ctis-2024-514714-12-00-aspire", "clinicaltrials-nct03257033-tigerpac", "novelli-2026-tigerpac-pk", "renovorx-2025-10k", "renovorx-2026-q2-10q", "renovorx-2026-full-enrollment", "igan-2026-adenocyte-financing" ], "links": [ { "target": "companies", "relation": "supports" }, { "target": "funding", "relation": "supports" }, { "target": "programme", "relation": "supports" }, { "target": "claim-observed-ten-year-research-capital", "relation": "contains" }, { "target": "claim-cordis-title-visible-target-cancer-capital", "relation": "contains" }, { "target": "claim-ukri-title-visible-pancreatic-capital", "relation": "contains" }, { "target": "failure-grail-capital-scaled-before-utility", "relation": "contains" }, { "target": "company-revolution-medicines", "relation": "contains" }, { "target": "company-elicio-therapeutics", "relation": "contains" }, { "target": "claim-current-capital-frontier-2026", "relation": "contains" }, { "target": "claim-commercial-capital-ledger-boundaries", "relation": "contains" }, { "target": "investor-third-rock-revolution-medicines", "relation": "contains" }, { "target": "investor-paq-series-b-syndicate", "relation": "contains" }, { "target": "investor-valar-series-a-syndicate", "relation": "contains" }, { "target": "investor-cosomil-public-private-capital", "relation": "contains" }, { "target": "company-silexion-therapeutics", "relation": "does-not-fund" }, { "target": "claim-sil204-trial-scales-before-human-knockdown", "relation": "bounded-by" }, { "target": "claim-vcn01-randomized-signal-not-survival-proof", "relation": "bounded-by" }, { "target": "failure-virage-analysis-and-survivor-selection", "relation": "learns-from" }, { "target": "unknown-vcn01-all-enrolled-survival", "relation": "leaves-open" }, { "target": "trial-vcn01-action-map", "relation": "implements" }, { "target": "hypothesis-vcn01-replication-to-survival-gate", "relation": "tests" }, { "target": "company-theriva-biologics", "relation": "contains" }, { "target": "failure-broad-epigenetic-treatment-pdac", "relation": "learns-from" }, { "target": "claim-mtap-prmt5-selected-human-signal", "relation": "bounded-by" }, { "target": "failure-vopimetostat-ras-contribution-unknown", "relation": "learns-from" }, { "target": "unknown-mtap-prmt5-ras-contribution", "relation": "leaves-open" }, { "target": "trial-mtap-prmt5-action-map", "relation": "implements" }, { "target": "hypothesis-mtap-prmt5-ras-deeper-control", "relation": "tests" }, { "target": "company-tango-therapeutics", "relation": "contains" }, { "target": "company-panbela-therapeutics", "relation": "contains" }, { "target": "trial-pdac-polyamine-action-map", "relation": "constrains" }, { "target": "hypothesis-aspire-data-rescue-gate", "relation": "tests" }, { "target": "company-renovorx", "relation": "contains" }, { "target": "trial-pdac-intraarterial-delivery-action-map", "relation": "constrains" }, { "target": "hypothesis-tigerpac-existing-final-gate", "relation": "tests" }, { "target": "investor-igan-adenocyte", "relation": "contains" } ], "evidence": { "humanProspective": "not-applicable", "humanRetrospective": "not-applicable", "randomised": "not-applicable", "independentReplication": "moderate", "sampleSize": "Ten NCI fiscal years, five DoD years, 159 CORDIS title matches, 74 UKRI title matches, four charity disclosures, four investor or public-startup capital records, one licensing deal, one acquisition and three company R&D disclosures", "effect": "$2.0067bn observed US federal funding in the 2015–2024 window; EUR 312.720m separate EU title-visible commitments; £24.106m separate UKRI title-visible commitments; commercial financing, licensing, acquisition and R&D kept in separate ledgers", "limits": [ "No global total", "Funding categories overlap", "Currencies and accounting bases differ", "University and most company programme costs are unavailable", "Acquisition and valuation are not research spend", "Capital cannot be causally mapped to population survival" ] }, "layout": { "x": 89, "y": 105 }, "updated": "2026-09-15", "contentSections": [ "Panbela receives £0 equity, licence and new-trial capital. $142,000 cash against $12.5m nine-month operating use gives a crude 3.1-day cash ratio, followed by contractor termination and a $12.0m note default. The valuable asset may be the existing masked ASPIRE comparison. Release at most £250,000 from the existing data line for custody diligence; any later rescue must displace the same amount elsewhere inside £500m.", "Silexion is a no-invest decision at the current gate. Its March 2026 cash of $2.413m against $3.646m first-quarter operating cash use models to 1.99 months at unchanged burn before new financing, and the filing states substantial going-concern doubt. More important, SIL204 has no reported human dose or tumour knockdown and its CTIS record says participant-level data will not be shared. Regulatory authorisation and site activation do not meet a target, data or efficacy gate.", "Theriva is a £0 decision at the current gate. VIRAGE has a credible randomized direction and human tumour-replication evidence, but its public record does not reconcile 112 enrolled with 96 analysed or the conflicting alpha statements. Early-August cash was $9.6m and current cash does not fund phase 3. The mission offers independent data recovery, not equity, licence or trial capital. Any later cofunding requires the all-randomized result, six-person dosing safety, a lead commercial partner, public-data rights and equal displacement inside the £500m ceiling.", "Tango is a £0 mission-investment decision. Its $1.0bn June 2026 cash and marketable-securities balance is about 12.0 times the annualized six-month operating cash use: 1000 ÷ (83.226 × 2) = 6.01 years. That simple ratio is not guidance, but shows the contribution test is not short of capital. Require exact MTAP selection, all-treated maturity, a RAS-alone control, independent analysis and public negative results.", "RenovoRx is a £0 equity and £0 new-trial decision. At June 2026, $9.479m cash divided by average six-month operating use of $6.576m / 6 gives 8.65 months of crude unchanged-use coverage. Enrollment then closed with 78 of 86 deaths, so mission subsidy cannot improve recruitment. Reserve at most £250,000 from the existing data and statistics line after database lock for participant-level access, fixed-plan verification, independent analysis and public reporting.", "Adenocyte adds a $10m first close led by iGan in May 2026. It is another example of capital moving toward a test that fits an existing reimbursable workflow before independent clinical utility exists. Keep it separate from scientific achievement: the investor reports four lesions among 90 high-risk people, while both company-sponsored registries remain observational and have no posted results." ] }, { "id": "investor-cosomil-public-private-capital", "title": "Cosomil capital network — public grant plus undisclosed venture rounds", "shortTitle": "Cosomil capital", "type": "investor", "status": "open", "scope": [ "pdac", "pancreatic-all", "colorectal", "pan-cancer" ], "summary": "NEDO awarded ¥499m while ANRI and other venture firms backed undisclosed rounds; none of the amounts proves test accuracy or screening benefit.", "content": "The capital stack combines unlike instruments. NEDO lists a ¥499m grant decision for fiscal years 2024-2025 under its deep-tech startup programme. A pre-Series A named Coral Capital, ANRI and Green Core but did not disclose the amount. A 2026 Series A1 was led by Delight Ventures with Animal Spirits, Gogin Capital, ANRI and Green Core; its amount and valuation were also undisclosed. The public grant planned a roughly 1,000-sample US study, stage 0/I sample work, colorectal work, a finger-prick kit, automation and FDA consultation. The later ENZEAVOUR study targets 10,000 asymptomatic adults in Japan. These are different uses and cannot be added to a disclosed venture total. The decision gate is prospective specificity, complete negative follow-up and treatable-stage positive predictive value—not number of investors or public support. If the 10,000-person study yields too few cancers or verifies mainly positives, the capital has bought feasibility, not screening validation.", "sourceIds": [ "nedo-2024-cosomil-profile", "cosomil-2026-series-a1", "jrct-2026-enzeavour-feasibility" ], "links": [ { "target": "investment", "relation": "part-of" }, { "target": "company-cosomil", "relation": "funds" }, { "target": "trial-enzeavour-japan-screening", "relation": "funds" }, { "target": "claim-screening-base-rate-maths", "relation": "constrained-by" } ], "evidence": { "humanProspective": "none — planned study, no result", "humanRetrospective": "not-applicable — not assessed in this record", "randomised": "none", "independentReplication": "none", "sampleSize": "¥499m public grant; two private rounds with undisclosed amounts; ENZEAVOUR target n=10,000", "effect": "Capital and planned work, no prospective accuracy or utility result", "limits": [ "Private amounts and valuations undisclosed", "Government and venture finance are not additive categories", "Grant covers several products and countries", "Partial verification in screening study", "No randomized utility" ] }, "layout": { "x": 94, "y": 119 }, "updated": "2026-09-14" }, { "id": "investor-igan-adenocyte", "title": "iGan Partners — $10m first close for Adenocyte's detection route", "shortTitle": "iGan / Adenocyte", "type": "investor", "status": "supported", "scope": [ "pdac", "pancreatic-all" ], "summary": "The May 2026 financing confirms investor appetite for a reimbursable diagnostic workflow before independent utility evidence.", "content": "iGan Partners led the first close of a $10m Adenocyte financing in May 2026. The disclosed use is clinical and commercial expansion plus continued development of LINFU for pancreatic, breast and prostate cytology sampling. The announcement emphasises existing billing codes, clinical use and four unsuspected lesions among 90 high-risk participants. That combination reveals the investment thesis: fit a technically distinct test into an existing reimbursable procedure before a long mortality trial finishes. It also creates the central risk. Reimbursement, 150 procedures and a selected yield claim do not establish sensitivity, specificity, false-negative disease, procedure harm or net benefit. The two linked pancreatic registries are observational, invitation-only and have no posted results. Treat the $10m as a capital-flow signal, not scientific validation. A mission partnership should require the full locked 90-person result, all-participant follow-up, central pathology, independent analysis, negative-result publication and explicit pricing before contributing money or sites.", "sourceIds": [ "igan-2026-adenocyte-financing", "clinicaltrials-nct04793633-linfu", "clinicaltrials-nct06276764-linfu-ipmn" ], "links": [ { "target": "investment", "relation": "part-of" }, { "target": "company-adenocyte", "relation": "funds" }, { "target": "claim-current-capital-frontier-2026", "relation": "extends" } ], "evidence": { "humanProspective": "weak — investor-reported early procedures only", "humanRetrospective": "none", "randomised": "none", "independentReplication": "none", "sampleSize": "$10m first close; investor reports >150 procedures and 90-person high-risk result", "effect": "Capital and strategy signal; no patient-benefit effect", "limits": [ "Investor announcement", "Final close not independently verified", "Clinical claim lacks a peer-reviewed report", "No public fund allocation by work package", "No posted registry results" ], "dimensions": { "biologicalPlausibility": { "rating": "not-applicable", "reason": "This record maps capital and investor logic rather than establishing a biological effect." }, "animalEvidence": { "rating": "not-applicable", "reason": "Animal work cannot validate an investment or human utility claim." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Laboratory evidence is separate from the disclosed financing event." }, "reproducibility": { "rating": "weak", "reason": "The financing event is public, but the associated performance claim lacks independent reproduction." }, "commercialConflicts": { "rating": "high", "reason": "The lead investor reports both its investment and the portfolio company's clinical claims." }, "recency": { "rating": "current", "reason": "The financing was announced in May 2026 and checked in September 2026." } } }, "layout": { "x": 158, "y": 105 }, "updated": "2026-09-15" }, { "id": "investor-paq-series-b-syndicate", "title": "PAQ Series B syndicate — $77m committed before human KRAS-degrader efficacy", "shortTitle": "PAQ Series B", "type": "investor", "status": "open", "scope": [ "pdac", "colorectal", "pan-cancer", "transfer-unproven" ], "summary": "A six-investor syndicate financed the first $39m and a later extension brought the Series B total to $77m before any human efficacy result.", "content": "The first $39m Series B was co-led by MRL Ventures Fund and Bayland Capital with JJDC, LAV Fund, BioTrack Capital and Sherpa Health Partners. A January 2026 extension brought the total Series B to $77m, mechanically adding $38m, but the extension announcement did not identify allocations or full terms. The capital supports clinical work on G12D degrader PT0253 and pan-KRAS degrader PT0511. At the evidence cut, both are early clinical programmes and PT0511 has no human response result. This is an explicit example of financing ahead of efficacy. The rational investment case is different pharmacology and broad KRAS reach; the failure case is that degradation may not improve tumour exposure, depth, tolerability or resistance relative to inhibitors. Follow-on capital should require human target degradation at tolerable exposure, PDAC cohort disclosure and a defined route to controlled contribution testing.", "sourceIds": [ "paq-2025-series-b-investors", "paq-2026-seriesb-pt0511", "nct07300150-pt0511" ], "links": [ { "target": "investment", "relation": "part-of" }, { "target": "company-paq-therapeutics", "relation": "funds" }, { "target": "company-revolution-medicines", "relation": "competes-with" }, { "target": "unknown-durable-ras-control", "relation": "depends-on" } ], "evidence": { "humanProspective": "none — planned study, no result", "humanRetrospective": "not-applicable", "randomised": "none", "independentReplication": "none", "sampleSize": "$39m initial Series B; $77m total after extension; one phase 1 programme", "effect": "$77m financing before human efficacy", "limits": [ "Company announcements", "Extension investors and allocations undisclosed", "Terms and valuation undisclosed", "No human efficacy", "Money covers more than one asset and cancer" ] }, "layout": { "x": 92, "y": 117 }, "updated": "2026-09-14" }, { "id": "investor-third-rock-revolution-medicines", "title": "Third Rock Ventures — $45m launch capital before the RAS pivot", "shortTitle": "Third Rock / RevMed", "type": "investor", "status": "supported", "scope": [ "pdac", "colorectal", "pan-cancer", "transfer-unproven" ], "summary": "Third Rock created Revolution Medicines with a $45m Series A in 2015, years before the company produced common-RAS survival evidence.", "content": "The 2015 origin announcement ties investor, founders, university licence and company formation into one record. Third Rock Ventures launched Revolution Medicines with $45m. Mark Goldsmith was founding chief executive; Martin Burke was scientific founder and advisory-board chair; licensed University of Illinois chemistry underpinned the original natural-product platform. The later common-RAS programme cannot be credited mechanically to one founding cheque or the original technical thesis: the company changed focus, raised far more capital and spent across several programmes. The useful capital lesson is time and option value. Roughly 11 years separated launch from the 2026 PDAC approval, so a 15-year vehicle can match drug-development duration better than a short fund. This is historical financing and origin evidence, not proof that venture formation caused the clinical result or that the initial $45m funded daraxonrasib specifically.", "sourceIds": [ "third-rock-2015-revolution-medicines-launch", "revolution-medicines-2026-q2-10q", "oreilly-2026-daraxonrasib" ], "links": [ { "target": "investment", "relation": "part-of" }, { "target": "company-revolution-medicines", "relation": "funds" }, { "target": "claim-daraxonrasib-common-ras-survival-breakthrough", "relation": "preceded" }, { "target": "claim-commercial-capital-ledger-boundaries", "relation": "constrained-by" } ], "evidence": { "humanProspective": "not-applicable", "humanRetrospective": "not-applicable", "randomised": "not-applicable", "independentReplication": "moderate", "sampleSize": "One $45m founding round, one current filing and one pivotal randomized clinical result", "effect": "$45m company launch; about 11 years to PDAC approval", "limits": [ "Company announcement", "Later financing not reconstructed here", "Original platform differed from later RAS focus", "No asset-level tracing of the $45m", "Investment did not guarantee outcome" ] }, "layout": { "x": 91, "y": 116 }, "updated": "2026-09-14" }, { "id": "investor-valar-series-a-syndicate", "title": "DCVC, a16z and Pear VC — $22m for Valar before a randomized utility test", "shortTitle": "Valar Series A", "type": "investor", "status": "open", "scope": [ "pdac", "pan-cancer", "transfer-unproven" ], "summary": "The 2024 Series A financed computational pathology expansion; PDAC external validation arrived later, while a model-directed trial remains missing.", "content": "DCVC and Andreessen Horowitz co-led Valar Labs' $22m Series A, with Pear VC participating. The announcement names founders Anirudh Joshi, Viswesh Krishna and Damir Vrabac and board additions from the lead investors. The stated use was broad precision-oncology expansion, initially centred on bladder cancer, not a ring-fenced PDAC programme. A later PDAC study used 178 development and 299 external-validation cases and found a biomarker-treatment interaction, but treatment was not assigned by the model. Therefore $22m bought platform and company development, not proven pancreatic benefit. The next evidence unit is a randomized treatment-choice trial with a frozen model, an independent analysis and access to failed slides. Until that exists, the investor graph should label this as a plausible low-cost routing platform whose commercial funding moved ahead of causal utility evidence.", "sourceIds": [ "valar-2024-series-a", "hendifar-2026-ai-chemo-pathology" ], "links": [ { "target": "investment", "relation": "part-of" }, { "target": "company-valar-labs", "relation": "funds" }, { "target": "technology-ai-pathology-chemo-selection", "relation": "funds" }, { "target": "programme-25m-prospective-observatory", "relation": "could-support" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "moderate", "randomised": "none", "independentReplication": "weak", "sampleSize": "$22m Series A; PDAC development n=178 and external validation n=299", "effect": "Financing and external validation, no model-directed treatment effect", "limits": [ "Company announcement", "Round not PDAC-specific", "Valuation and terms undisclosed", "Company authors and patents", "No randomized utility" ] }, "layout": { "x": 93, "y": 118 }, "updated": "2026-09-14" }, { "id": "lab-africa-hpb-cancer-consortium", "title": "Africa HepatoPancreatoBiliary Cancer Consortium — registry network under construction", "shortTitle": "Africa HPB consortium", "type": "lab", "status": "open", "scope": [ "pdac", "pnet", "cholangiocarcinoma", "gallbladder", "hcc", "hepatobiliary", "pancreatic-all" ], "summary": "AHPBCC supplies a multi-country route for African HPB registries, but its published record is a roadmap rather than proof of a complete working dataset.", "content": "Mission capability: define a minimum African HPB dataset, connect clinical and population registries, and measure local risk factors, tumour features, treatment and survival across settings missing from most discovery cohorts. The 2026 roadmap includes authors working in Nigeria, Ghana, Ethiopia, Rwanda, Burkina Faso, Egypt and Zambia, with US partners. That breadth is useful for protocol design and site selection. It is not evidence that the named countries already contribute complete, linked or comparable data. The need is now measured: an African pancreatic review found 26,850 cases but nearly 40% lacked stage and fewer than half of studies reported complete treatment paths; a sub-Saharan HCC review found reports from only 15 of 48 countries, with 6% curative treatment and 84% supportive care. First contract gate: five sites must return twelve months of denominator, pathology, stage, recommended and delivered treatment, reason for non-delivery, follow-up and missingness under one dictionary. Stop or redesign if more than 20% of required fields remain missing or less than 80% of diagnosed patients can be followed at twelve months. Main risk: a consortium name can hide uneven staff, registries, pathology and follow-up.", "sourceIds": [ "asombang-2026-africa-hpb-registry", "jonas-2025-sub-saharan-hcc-pathways", "gismalla-2026-africa-pancreatic-review", "sobnach-2024-ssa-hcc-systematic-review" ], "links": [ { "target": "labs", "relation": "part-of" }, { "target": "lab-uct-groote-schuur-hpb", "relation": "could-collaborate-with" }, { "target": "datasets", "relation": "could-create" }, { "target": "programme-unconstrained-global-system", "relation": "could-test" }, { "target": "trial-regional-route-completion-action-map", "relation": "could-run" } ], "evidence": { "humanProspective": "none — no complete prospective registry result reported", "humanRetrospective": "moderate — two reviews quantify the gaps the roadmap proposes to close", "randomised": "not-applicable", "independentReplication": "none", "sampleSize": "Roadmap with authors across at least seven African countries; reviews cover 26,850 pancreatic and 3,989 HCC cases, but no common registry cohort is reported", "effect": "Network and registry-design assessment", "limits": [ "Roadmap, not a completed registry", "Named authors do not prove active site delivery", "Data completeness and follow-up unknown", "Population coverage and sustainable funding unknown", "Existing reviews use incomplete and incompatible source records" ], "dimensions": { "biologicalPlausibility": { "rating": "not-applicable", "reason": "This is a registry-network proposal." }, "animalEvidence": { "rating": "not-applicable", "reason": "Animal evidence does not establish registry function." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Laboratory results do not establish data completeness." }, "reproducibility": { "rating": "none", "reason": "No common multi-country registry output is yet reported." }, "commercialConflicts": { "rating": "unknown", "reason": "Funding and product relationships require current review." }, "recency": { "rating": "current", "reason": "The consortium roadmap was published in 2026." } } }, "layout": { "x": 57, "y": 123 }, "updated": "2026-09-14" }, { "id": "lab-alliance-a021806-network", "title": "Alliance A021806 network — the live modern timing test", "shortTitle": "Alliance A021806", "type": "lab", "status": "supported", "scope": [ "pdac" ], "summary": "Alliance and NCI have completed accrual of 358 people to the direct perioperative-versus-adjuvant mFOLFIRINOX comparison; survival follow-up runs to 2028-2030.", "content": "Current operating evidence: NCT04340141 names Alliance for Clinical Trials in Oncology as sponsor, NCI as collaborator and Cristina Ferrone as chair. It reports 358 actual participants, active but not recruiting, after opening in October 2020. Central imaging review restricts entry to technically resectable PDAC. The arms plan the same twelve mFOLFIRINOX cycles: eight before and four after surgery versus all twelve after upfront surgery. Overall survival from randomization is primary. Disease-free survival, unresectability, R0, distant metastasis, adverse events, dose intensity, cycles received and eight-week quality of life are named outcomes. Mission role: finish follow-up, recover every assigned cycle and surgery event, and supply the modern timing answer before another broad trial. Access gate: confirm the actual accruing centres behind 453 US and Canadian location records, event count, missing data, patient-level access, analysis code, independent statistics and a dated publication plan. Estimated primary completion is December 2028 and final completion November 2030; no result is posted.", "sourceIds": [ "clinicaltrials-nct04340141-a021806" ], "links": [ { "target": "labs", "relation": "part-of" }, { "target": "trial-pdac-perioperative-sequencing-action-map", "relation": "runs" }, { "target": "unknown-pdac-perioperative-treatment-route", "relation": "tests" }, { "target": "person-cristina-ferrone", "relation": "includes" }, { "target": "programme-100m-randomized-platform", "relation": "could-partner-with" } ], "evidence": { "humanProspective": "strong — active multicentre randomized route", "humanRetrospective": "not-applicable", "randomised": "strong design; no result", "independentReplication": "none for the modern timing question", "sampleSize": "Actual n=358; 453 registry location records in the US and Canada", "effect": "Capability and access record, not a clinical result", "limits": [ "No posted result", "Primary completion estimated in 2028", "Actual accruing sites not stated", "Patient-level access unconfirmed", "Fit centrally reviewed population" ], "dimensions": { "biologicalPlausibility": { "rating": "not-applicable", "reason": "This is a cooperative-trial network record." }, "animalEvidence": { "rating": "not-applicable", "reason": "Not relevant to route delivery." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Not relevant to route delivery." }, "reproducibility": { "rating": "moderate", "reason": "The network completed multicentre accrual; the outcome has not reported." }, "commercialConflicts": { "rating": "low", "reason": "The registry names Alliance and NCI rather than a proprietary product sponsor." }, "recency": { "rating": "current", "reason": "Status, accrual and dates were verified in September 2026." } } }, "updated": "2026-09-14" }, { "id": "lab-bidmc-thrombosis-pdi", "title": "Beth Israel Deaconess thrombosis and haemostasis network — PDI mechanism to cancer trial", "shortTitle": "BIDMC PDI network", "type": "lab", "status": "supported", "scope": [ "pdac", "pan-cancer", "transfer-unproven" ], "summary": "This academic network linked PDI assays, a 57-person cancer study and a patent family; its next value is open methods and independent phase-3 analysis.", "content": "Capability: the Beth Israel Deaconess and Harvard thrombosis group joined PDI biology, platelet-dependent thrombin generation, D-dimer and P-selectin assays to a multicentre cancer trial. CATIQ tested 500 mg and 1,000 mg isoquercetin in 57 people and showed dose-linked marker changes, but no controlled clinical endpoint. The group and institution are linked to WO2020112828A1 and later family members. Current route: Beth Israel is listed as a site in the Quercis phase 3 and Jeffrey Zwicker leads the clinical path. Mission use: independent assay transfer, protocol and dataset access, expected-event reconstruction and comparison of marker change with adjudicated clinical events. Required independence: reproduce assays outside the originating lab and keep the phase-3 event committee and statistics free of product ownership. The network is strong for mechanism and trial start-up; clinical prevention remains unproved until the randomized result.", "sourceIds": [ "zwicker-2019-isoquercetin-catiq", "clinicaltrials-nct06861088-catiq-p3", "patent-wo2020112828-isoquercetin-cancer" ], "links": [ { "target": "labs", "relation": "part-of" }, { "target": "person-jeffrey-zwicker", "relation": "includes" }, { "target": "claim-kinisoquin-phase2-marker-not-clinical-proof", "relation": "produced" }, { "target": "company-quercis-pharma", "relation": "works-with" }, { "target": "claim-kinisoquin-access-stack", "relation": "could-unlock" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "not-applicable", "randomised": "none in completed Kinisoquin evidence", "independentReplication": "none", "sampleSize": "One 57-person phase 2, one active phase 3 site and one patent family", "effect": "Strong mechanism-to-trial capability; randomized patient benefit unresolved", "limits": [ "Institutional network rather than one formal laboratory", "Public sources only", "No assay transfer data", "Originator independence limit", "No phase 3 result" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "The network connected a clotting mechanism to dose-linked human marker changes." }, "animalEvidence": { "rating": "moderate", "reason": "PDI and P-selectin thrombosis models support the research route." }, "inVitroEvidence": { "rating": "strong", "reason": "Multiple target and coagulation assays form the lab's main demonstrated capability." }, "reproducibility": { "rating": "none", "reason": "Independent assay transfer and randomized clinical replication are not reported." }, "commercialConflicts": { "rating": "high", "reason": "Academic discovery, patents, supplied product and sponsor development are connected." }, "recency": { "rating": "current", "reason": "The network remains linked to an active 2026 phase 3 study." } } }, "layout": { "x": 172, "y": 96 }, "updated": "2026-09-14" }, { "id": "lab-cnio-experimental-oncology", "title": "CNIO Experimental Oncology Group — multi-node RAS resistance", "shortTitle": "CNIO RAS models", "type": "lab", "status": "contested", "scope": [ "pdac", "pan-cancer", "transfer-unproven" ], "summary": "The Spanish group produced a strong triple-node mouse regression signal, but human-ready drugs, independent reproduction and conflict management are missing.", "content": "Mission capability: genetically engineered PDAC, orthotopic immune-competent models and patient-derived xenografts used to test simultaneous RAS, EGFR/HER2 and STAT3 suppression. The 2026 result is one of the clearest experimental examples of durable resistance prevention across several model types. Main limits are equally clear: small groups, no human efficacy, daraxonrasib exposure above the trial setting, questionable afatinib tolerability, and a STAT3 degrader with unsuitable drug properties. The first publication was retracted for an undisclosed Vega Oncotargets interest and replaced after direct review with declarations. Best role: hypothesis source under independent replication; the group should not control the confirmatory assay, capital decision or first human trial.", "sourceIds": [ "cnio-experimental-oncology-2026", "liaki-2026-triple-combination-republished", "pnas-2026-liaki-retraction" ], "links": [ { "target": "labs", "relation": "part-of" }, { "target": "claim-kras-egfr-stat3-mouse-regression", "relation": "supports" }, { "target": "failure-liaki-undisclosed-conflict-retraction", "relation": "depends-on" } ], "evidence": { "humanProspective": "none", "humanRetrospective": "none", "randomised": "none", "independentReplication": "none", "sampleSize": "Multiple small mouse systems and six patient-derived xenograft models", "effect": "Durable mouse tumour regression under triple blockade", "limits": [ "No human evidence", "Non-clinical drug conditions", "No independent repeat", "Founder financial interests", "Original publication retracted and replaced" ] }, "layout": { "x": 91, "y": 109 }, "updated": "2026-09-14" }, { "id": "lab-cnio-genetic-epidemiology", "title": "CNIO Genetic and Molecular Epidemiology Group — European risk and detection validation", "shortTitle": "CNIO epidemiology", "type": "lab", "status": "supported", "scope": [ "pdac", "pancreatic-all", "pan-cancer", "transfer-unproven" ], "summary": "The group adds European pancreatic-risk cohorts and methods for joining clinical, exposure, inherited and measured biological data.", "content": "Mission capability: distinguish risk enrichment from diagnostic accuracy and test whether signals transport across European populations. The group is led by Núria Malats and participates in PANCAID, a 2023-2027 EU project with €9.845m total cost. PANCAID plans to select a multi-marker blood panel using banked samples and then design a future international prospective study. The correct mission role is to force that transition and prevent a fitted panel from being called screening-ready. Required outputs are a frozen assay, explicit intended population, independent site split, complete disease controls, calibration, false-positive work-up and a prospective decision study. Main weakness: the current grant record describes intended work and funding, not a validated product or improved outcome.", "sourceIds": [ "cnio-malats-profile-2026", "cordis-pancaid-101096309" ], "links": [ { "target": "labs", "relation": "part-of" }, { "target": "person-nuria-malats", "relation": "led-by" }, { "target": "detection-integrated-workflows", "relation": "supports" }, { "target": "claim-screening-base-rate-maths", "relation": "constrained-by" } ], "evidence": { "humanProspective": "none — planned study, no result", "humanRetrospective": "strong", "randomised": "none", "independentReplication": "weak — mixed evidence", "sampleSize": "PANCAID has 19 participating organisations; CORDIS does not fix sample counts", "effect": "Capability and grant record, not clinical validation", "limits": [ "Banked-sample development", "Prospective study is not yet the funded result", "No utility endpoint", "Institutional willingness and data rights untested" ] }, "layout": { "x": 98, "y": 118 }, "updated": "2026-09-14" }, { "id": "lab-cruk-scotland-morton", "title": "Cancer Research UK Scotland Institute — Morton pancreatic group", "shortTitle": "CRUK Scotland Morton", "type": "lab", "status": "supported", "scope": [ "pdac", "pancreatic-all", "transfer-unproven" ], "summary": "A UK pancreatic-model group covering fibroblasts, immune interaction, radiotherapy and homologous-recombination deficiency.", "content": "Mission capability: genetically grounded pancreatic models tied to fibroblast, immune, metastatic, radiotherapy and DNA-repair questions. UKRI lists £2.049m for a current pancreatic microenvironment project, placing a visible funding anchor beside the capability. Main risk: earlier CXCR2 and stromal mouse signals illustrate the translation problem; model response cannot set clinical priority alone. Best role: independent replication of resistance and tumour-environment mechanisms generated in US centres, using shared compounds and locked readouts.", "sourceIds": [ "cruk-scotland-morton-lab-2026", "ukri-gtr-pancreatic-title-query-2026" ], "links": [ { "target": "labs", "relation": "part-of" }, { "target": "claim-immune-exclusion-is-multicellular", "relation": "tests" }, { "target": "claim-ukri-title-visible-pancreatic-capital", "relation": "funds" } ], "evidence": { "humanProspective": "none", "humanRetrospective": "weak", "randomised": "none", "independentReplication": "moderate", "sampleSize": "One current laboratory programme and linked public award", "effect": "Capability assessment", "limits": [ "Main evidence is preclinical", "Institutional profile and award title do not prove translation", "Named project outcomes pending" ] }, "layout": { "x": 89, "y": 109 }, "updated": "2026-09-14" }, { "id": "lab-danafarber-hale-ras", "title": "Dana-Farber Hale Family Center and Center for RAS Therapeutics", "shortTitle": "Dana-Farber RAS", "type": "lab", "status": "supported", "scope": [ "pdac", "colorectal", "cholangiocarcinoma", "pan-cancer", "transfer-unproven" ], "summary": "Andrew Aguirre's programme can compare RAS alleles and escape routes across cancers while retaining a pancreatic clinic.", "content": "Mission capability: functional genetics, patient-derived models and clinical specimens around common-RAS and mutation-specific drugs. Cross-cancer comparison is useful when the same allele escapes differently in pancreatic, colorectal, lung or biliary tissue. Main risk: model sensitivity is not a patient outcome, and access to several company compounds can bias which combinations receive attention. Best role: own the prospective resistance taxonomy and combination rejection screen, with compound-neutral governance and mandatory publication of negative screens.", "sourceIds": [ "danafarber-aguirre-profile-2026" ], "links": [ { "target": "labs", "relation": "part-of" }, { "target": "person-andrew-aguirre", "relation": "employs" }, { "target": "unknown-durable-ras-control", "relation": "tests" } ], "evidence": { "humanProspective": "not-applicable", "humanRetrospective": "not-applicable", "randomised": "not-applicable", "independentReplication": "not-applicable", "sampleSize": "One integrated clinical and laboratory programme", "effect": "Capability assessment", "limits": [ "Institutional profile evidence", "Compound access and conflicts need live review", "Human delivery not proved by models" ] }, "layout": { "x": 73, "y": 109 }, "updated": "2026-09-14" }, { "id": "lab-dpcg-radar-panc-network", "title": "Dutch Pancreatic Cancer Group RADAR-PANC network", "shortTitle": "RADAR-PANC network", "type": "lab", "status": "supported", "scope": [ "pdac" ], "summary": "A Dutch national cohort-and-trial network with a University of Birmingham site, able to test whether scheduled surveillance changes survival and quality of life.", "content": "Current operating evidence: NCT04875325 names UMC Utrecht as sponsor, the Dutch Pancreatic Cancer Group and University of Birmingham as collaborators, ten Dutch hospitals plus Birmingham as sites, and four principal investigators. The trial is nested in the Dutch PACAP and UK PACOPS cohorts, which allows routine outcomes and patient-reported measures to support a randomized care-strategy test. It is unusually valuable because it asks the endpoint that observational imaging reports cannot settle: overall survival, with quality of life, worry, treatment, compliance and cost also measured. Mission role: complete and independently analyse RADAR-PANC, align fields with CIRCPAC and the wider residual-disease programme, and provide procedure-level harms and recurrence-treatment details. Main limits: the registry was last verified in September 2024, still says enrolling by invitation and has no posted result shortly before an estimated October 2026 completion. The protocol's 49% relative survival-improvement assumption may leave the trial underpowered for a smaller effect. Before contracting, confirm randomized count, event maturity, intervention acceptance, control imaging, UK contribution, analysis code, patient-level access and a dated publication plan.", "contentSections": [ "The same Dutch Pancreatic Cancer Group network also produced PREOPANC and PREOPANC-2 staging analyses. PREOPANC-2 enrolled 369 randomized patients across many Dutch centres and recorded 322 surgical explorations; the staging method itself was not randomized. This makes the network a strong partner for protocol definitions and comparison data, not proof that its rule transfers to England." ], "sourceIds": [ "van-dongen-2023-preopanc-staging", "rompen-2026-preopanc2-staging", "stoop-2026-global-staging-survey", "daamen-2024-radar-panc-protocol", "clinicaltrials-nct04875325-radar-panc", "andel-2025-routine-imaging", "van-goor-2025-nationwide-imaging" ], "links": [ { "target": "person-marc-besselink", "relation": "includes" }, { "target": "trial-pdac-occult-metastasis-action-map", "relation": "could-partner-with" }, { "target": "labs", "relation": "part-of" }, { "target": "trial-pdac-postresection-surveillance-action-map", "relation": "runs" }, { "target": "unknown-postresection-surveillance-utility", "relation": "tests" }, { "target": "person-lois-daamen", "relation": "includes" }, { "target": "programme-100m-randomized-platform", "relation": "could-partner-with" }, { "target": "trial-pdac-ctdna-action-map", "relation": "sets-standard-imaging-comparator-for" } ], "evidence": { "humanProspective": "strong — national cohorts and randomized design", "humanRetrospective": "strong — linked registry work", "randomised": "strong design; no result", "independentReplication": "none for utility", "sampleSize": "RADAR-PANC planned n=306 across 11 listed sites; linked observational cohorts n=333 and n=1,311", "effect": "Capability and access record, not a clinical result", "limits": [ "Current accrual unconfirmed", "No posted randomized result", "Optimistic power assumption", "Crossover risk", "Patient-level data access unconfirmed", "One UK site listed" ], "dimensions": { "biologicalPlausibility": { "rating": "not-applicable", "reason": "This is a trial-network capability record." }, "animalEvidence": { "rating": "not-applicable", "reason": "Not relevant to network delivery." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Not relevant to network delivery." }, "reproducibility": { "rating": "moderate", "reason": "The network has delivered linked national, international and randomized work, while the decisive trial remains unfinished." }, "commercialConflicts": { "rating": "low", "reason": "The core question is a non-proprietary follow-up schedule." }, "recency": { "rating": "current", "reason": "The trial's current public deadline is October 2026, though the registry update is stale." } } }, "updated": "2026-09-15" }, { "id": "lab-ecog-acrin-geriatric-oncology-network", "title": "ECOG-ACRIN geriatric-oncology network — a completed older-PDAC trial route", "shortTitle": "ECOG-ACRIN older-PDAC network", "type": "lab", "status": "supported", "scope": [ "pdac" ], "summary": "ECOG-ACRIN recruited 176 vulnerable older adults across 92 US sites and collected treatment, ageing, quality-of-life and imaging measures in one PDAC trial.", "content": "Mission capability: reuse a national route that has already recruited the population routinely excluded from fitter-patient trials. GIANT was designed and conducted by ECOG-ACRIN with National Cancer Institute funding. It enrolled 176 people across 92 US sites, used a short vulnerability screen, repeated a fuller inventory and recorded survival, response, toxicity, falls, emergency care, admissions, treatment changes, quality of life, muscle and inflammatory markers. That joined field set is more valuable than a new age-only registry. Proposed mission role: obtain the locked data dictionary, treatment-start and discontinuation reasons, complete patient-level follow-up, site performance and missingness; map the smallest usable assessment; and adapt the action layer with UK and European sites. Main limits: the trial's planned survival difference failed; the current registry has no result tables; 92 enrolling sites do not guarantee equal data quality; and a US network may not transport to NHS staffing or access. Data access, patient consent, independent analysis, missing-data recovery and site-level equity must pass before money releases. The network supplies trial infrastructure, not proof that one regimen or assessment rule is best.", "sourceIds": [ "dotan-2024-giant", "dotan-2025-giant-ga", "clinicaltrials-nct04233866-giant" ], "links": [ { "target": "labs", "relation": "part-of" }, { "target": "person-efrat-dotan", "relation": "includes" }, { "target": "trial-older-pdac-action-map", "relation": "ran" }, { "target": "hypothesis-older-pdac-fast-support-route", "relation": "could-test" } ], "evidence": { "humanProspective": "strong — completed multicentre randomized recruitment", "humanRetrospective": "not-applicable", "randomised": "strong infrastructure; negative regimen comparison", "independentReplication": "not-applicable", "sampleSize": "GIANT actual n=176 at 92 US sites", "effect": "Capability and access record, not an efficacy estimate", "limits": [ "Patient-level access unconfirmed", "Registry results absent", "Site contribution uneven", "US-to-NHS transfer unknown", "Full trial report pending", "Support intervention not tested" ], "dimensions": { "biologicalPlausibility": { "rating": "not-applicable", "reason": "This record evaluates network capability." }, "animalEvidence": { "rating": "not-applicable", "reason": "Not relevant to clinical network delivery." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Not relevant to clinical network delivery." }, "reproducibility": { "rating": "moderate", "reason": "Multisite recruitment is demonstrated; transfer and action delivery remain untested." }, "commercialConflicts": { "rating": "low", "reason": "The National Cancer Institute funded the cooperative trial network." }, "recency": { "rating": "current", "reason": "Trial follow-up remains active with final completion estimated in December 2026." } } }, "updated": "2026-09-14" }, { "id": "lab-hopkins-early-detection", "title": "Johns Hopkins Pancreatic Cancer Early Detection Laboratory", "shortTitle": "Hopkins detection lab", "type": "lab", "status": "supported", "scope": [ "pdac", "pancreatic-all" ], "summary": "Longitudinal inherited-risk surveillance and biomarker continuity make this a strong source of samples collected before diagnosis.", "content": "Mission capability: serial samples, inherited-risk cohorts, MRI and endoscopic-ultrasound surveillance, and genetic, epigenetic and protein markers under Michael Goggins. These samples can test whether a marker rises before an actionable lesion rather than merely separating diagnosed cases from controls. Main risk: expert-centre families are a small, selected fraction of future PDAC and results may not transport to new-onset diabetes or ordinary health systems. Best role: reference cohort and assay comparator, not sole recruitment system.", "sourceIds": [ "hopkins-goggins-profile-2026", "dbouk-2022-caps5", "blackford-2024-high-risk-surveillance" ], "links": [ { "target": "labs", "relation": "part-of" }, { "target": "person-michael-goggins", "relation": "employs" }, { "target": "unknown-actionable-interception-window", "relation": "tests" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "moderate", "randomised": "none", "independentReplication": "moderate", "sampleSize": "Prospective high-risk surveillance programmes and longitudinal biospecimens", "effect": "Capability assessment; linked surveillance cohorts show earlier stage and survival association", "limits": [ "Selected inherited-risk population", "Expert-centre care", "No randomized mortality proof" ] }, "layout": { "x": 75, "y": 109 }, "updated": "2026-09-14" }, { "id": "lab-imperial-vapor2", "title": "Imperial VAPOR2 network — UK prospective breath-test validation", "shortTitle": "VAPOR2 network", "type": "lab", "status": "open", "scope": [ "pdac", "pancreatic-all", "pan-cancer", "transfer-unproven" ], "summary": "A prospective 6,079-person study spanning UK health systems can test a breath pipeline beyond retrospective case-control accuracy.", "content": "Mission capability: recruit the symptomatic and referred population in which pancreatic detection would operate, freeze the breath collection and analysis pipeline, and count diagnostic consequences. ClinicalTrials.gov lists Imperial as sponsor, George Hanna as principal investigator and Pancreatic Cancer UK plus 22 health-system collaborators. The estimated enrolment is 6,079. That scale can produce a useful accuracy estimate if case capture and negative follow-up are complete. It cannot prove clinical utility without a comparison of decisions and outcomes. The mission contract should require a public analysis plan, site-stratified calibration, independent chemical analysis, all failed tests, work-up procedures, time to diagnosis, health-economic results and disclosure of patents or commercial interests.", "sourceIds": [ "nct07243262-vapor2", "imperial-hanna-profile-2026" ], "links": [ { "target": "labs", "relation": "part-of" }, { "target": "person-george-hanna", "relation": "led-by" }, { "target": "detection-rna-protein-metabolite-methods", "relation": "tests" }, { "target": "claim-case-control-blood-tests-overstate-screening-readiness", "relation": "advances-beyond" } ], "evidence": { "humanProspective": "none — planned study, no result", "humanRetrospective": "weak", "randomised": "none", "independentReplication": "none", "sampleSize": "Target n=6,079; 22 listed health-system collaborators plus charity and Imperial clinical partner", "effect": "No prospective result posted", "limits": [ "Observational", "No decision or mortality comparison", "One listed location despite many collaborators", "Assay ownership and conflicts need current review" ] }, "layout": { "x": 98, "y": 117 }, "updated": "2026-09-14" }, { "id": "lab-lacog-gaspar", "title": "LACOG GASPAR — Latin American pancreatic data and site network", "shortTitle": "LACOG GASPAR", "type": "lab", "status": "supported", "scope": [ "pdac", "pancreatic-all", "pan-cancer" ], "summary": "GASPAR demonstrates a working three-country route for pancreatic real-world data, but its sample and private-care concentration are too narrow to stand for Latin America.", "content": "Mission capability: recruit through LACOG, GAICO and cancer centres in Brazil, Argentina and Colombia; collect treatment and outcome data; and identify where care is lost before designing an intervention. The 2025 abstract reports 79 advanced pancreatic cases within 201 total participants. Of the pancreatic group, 67 received first-line treatment, 27 second-line and 10 third-line, so only 27/79 = 34% reached second line and 10/79 = 13% reached third line. This is useful delivery evidence, not a comparison proving why attrition occurred. A mission contract should expand public-sector and underrepresented sites, reconcile the registry's 120-person total with the abstract's 201, publish the data dictionary and missingness, and test follow-up completeness. Main risk: scaling a convenient private-network sample could give precise answers about the wrong population.", "sourceIds": [ "clinicaltrials-nct05924789-gaspar", "peixoto-2025-gaspar-results", "oncoclinicas-2026-peixoto-profile" ], "links": [ { "target": "labs", "relation": "part-of" }, { "target": "person-renata-dalpino-peixoto", "relation": "led-by" }, { "target": "trial-gaspar-lacog-0222", "relation": "runs" }, { "target": "datasets", "relation": "could-extend" }, { "target": "programme-unconstrained-global-system", "relation": "could-test" } ], "evidence": { "humanProspective": "moderate — bidirectional observational cohort", "humanRetrospective": "moderate", "randomised": "none", "independentReplication": "none", "sampleSize": "Abstract n=201 overall, n=79 pancreatic, seven sites and three countries; registry n=120", "effect": "Working regional data route with large selection and representation limits", "limits": [ "Observational cohort", "Only 79 pancreatic cases", "Private-care and racial concentration", "Registry and abstract counts conflict", "No intervention or mortality comparison" ], "dimensions": { "biologicalPlausibility": { "rating": "not-applicable", "reason": "This is a data and delivery network." }, "animalEvidence": { "rating": "not-applicable", "reason": "Animal work does not validate regional representativeness." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Laboratory evidence is not the basis for this recommendation." }, "reproducibility": { "rating": "weak", "reason": "The network has one reported cohort and an unresolved enrolment discrepancy." }, "commercialConflicts": { "rating": "moderate", "reason": "Commercial oncology sites and an industry collaborator require governance and data-access terms." }, "recency": { "rating": "current", "reason": "The result was reported in 2025 and registry status was checked in 2026." } } }, "layout": { "x": 48, "y": 121 }, "updated": "2026-09-14" }, { "id": "lab-latam-cca-registry", "title": "LATAM-CCA Registry — five-country cholangiocarcinoma route", "shortTitle": "LATAM-CCA Registry", "type": "lab", "status": "supported", "scope": [ "cholangiocarcinoma", "hepatobiliary", "transfer-unproven" ], "summary": "The registry proves that five Latin American expert centres can harmonize CCA records, but its referral selection and missing population denominator limit it as a route-completion platform.", "content": "The LATAM-CCA Registry links referral centres in Argentina, Brazil, Chile, Ecuador and Peru through the ESCALON and international CCA networks. Its first report harmonized 309 eligible cases across intrahepatic, perihilar and distal CCA, with treatment and survival detail unavailable in many regional datasets. Mission capability: extend a working five-country dictionary into a prospective route from population catchment to expert treatment, and test whether a measured referral or delivery repair changes outcomes. Contract gates: include consecutive eligible cases rather than complete records only; publish screened, excluded and missing counts; link a catchment denominator; time-stamp first contact through treatment; record why a recommended action was not delivered; allow qualified independent access to de-identified data; and keep negative results public. Main risks: expert-centre selection, undercounting remote and undiagnosed patients, internal rather than central review, self-reported ethnicity used as a weak stand-in for structural factors, and industry funding or investigator relationships. It should collaborate with GASPAR but not merge disease-specific fields or claim that five centres represent Latin America.", "sourceIds": [ "da-fonseca-2024-latam-cca" ], "links": [ { "target": "labs", "relation": "part-of" }, { "target": "datasets", "relation": "maintains" }, { "target": "lab-lacog-gaspar", "relation": "could-collaborate-with" }, { "target": "trial-regional-route-completion-action-map", "relation": "could-run" }, { "target": "programme-unconstrained-global-system", "relation": "could-test" } ], "evidence": { "humanProspective": "none — the published analysis is retrospective", "humanRetrospective": "moderate — harmonized five-centre cohort", "randomised": "none", "independentReplication": "none", "sampleSize": "309 eligible CCA patients across five countries; 276 in treatment analysis", "effect": "Working regional registry and potential trial route, not population coverage or intervention proof", "limits": [ "Five expert referral centres", "Not population based", "Mandatory-data exclusions", "No external central review", "Remote and undiagnosed patients absent", "Data not openly available", "Industry funding and author relationships" ], "dimensions": { "biologicalPlausibility": { "rating": "not-applicable", "reason": "This is a registry-network assessment." }, "animalEvidence": { "rating": "not-applicable", "reason": "Animal studies cannot establish data or delivery coverage." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Laboratory work is not the basis for the route recommendation." }, "reproducibility": { "rating": "weak", "reason": "One harmonized cohort is reported; population linkage and external repeat are absent." }, "commercialConflicts": { "rating": "moderate", "reason": "Incyte supported the registry and a senior investigator reported several product-company relationships." }, "recency": { "rating": "current", "reason": "The registry result was published in late 2024." } } }, "updated": "2026-09-14" }, { "id": "lab-mit-jameel-clinical-ai", "title": "MIT Jameel Clinic and Barzilay group — clinical AI methods", "shortTitle": "MIT clinical AI", "type": "lab", "status": "supported", "scope": [ "pdac", "pan-cancer", "transfer-unproven" ], "summary": "The group can connect risk, pathology, records and molecule models to locked external tests, but most demonstrated clinical work is outside PDAC.", "content": "Mission capability: define the decision a model changes, freeze its inputs and outputs, test calibration across hospitals and connect model uncertainty to a prospective experiment. The same group spans clinical prediction and drug design, which helps prevent a hand-off between modelling and laboratory tests. Main risk: an accurate model can learn care patterns rather than cancer biology, and strong work in breast or lung cancer does not prove PDAC value. Best role: methods and independent trial design around existing CT, pathology, organoid and residual-disease datasets; no deployment claim without external and prospective validation.", "sourceIds": [ "mit-barzilay-profile-2026", "mit-2026-barzilay-aacr" ], "links": [ { "target": "labs", "relation": "part-of" }, { "target": "person-regina-barzilay", "relation": "employs" }, { "target": "ai-mission-work-packages", "relation": "supports" } ], "evidence": { "humanProspective": "weak", "humanRetrospective": "strong", "randomised": "none", "independentReplication": "moderate", "sampleSize": "Cross-cancer AI programme; PDAC-specific work not pooled", "effect": "Methods capability assessment", "limits": [ "Much evidence outside PDAC", "Retrospective accuracy is not utility", "Data access and transport remain limiting" ] }, "layout": { "x": 85, "y": 109 }, "updated": "2026-09-14" }, { "id": "lab-msk-olayan-vaccines", "title": "MSK Olayan Center for Cancer Vaccines", "shortTitle": "Olayan vaccine center", "type": "lab", "status": "supported", "scope": [ "pdac", "colorectal", "pan-cancer", "transfer-unproven" ], "summary": "The centre joins individual vaccine design, manufacture, treatment and immune measurement, reducing hand-off delay in residual disease.", "content": "Mission capability: work backwards from rare long-term PDAC survivors, select mutation-derived targets, manufacture an individual RNA vaccine and track induced T cells under one clinical programme. That integration directly attacks the time constraint after surgery. Main risk: phase 1 survival association compared immune responders with nonresponders after treatment and cannot prove a vaccine effect. Best role: own manufacturing-time and immune-mechanism data inside the randomized phase 2, while an independent group evaluates all assigned patients, non-manufactured products and recurrence.", "sourceIds": [ "msk-balachandran-profile-2026", "rojass-2025-neoantigen-followup", "nct05968326-imcode003" ], "links": [ { "target": "labs", "relation": "part-of" }, { "target": "person-vinod-balachandran", "relation": "employs" }, { "target": "claim-neoantigen-vaccine-frontier", "relation": "supports" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "moderate", "randomised": "none", "independentReplication": "weak", "sampleSize": "Phase 1 vaccinated cohort n=16; randomized phase 2 recruiting", "effect": "Integrated vaccine capability; efficacy unproved", "limits": [ "Nonrandomized efficacy signal", "Selected immune responders", "Sponsor and institutional concentration", "Manufacturing failures must remain in denominator" ] }, "layout": { "x": 79, "y": 109 }, "updated": "2026-09-14" }, { "id": "lab-msk-rubenstein", "title": "MSK David M. Rubenstein Center for Pancreatic Cancer Research", "shortTitle": "MSK Rubenstein Center", "type": "lab", "status": "supported", "scope": [ "pdac", "pnet", "cholangiocarcinoma", "gallbladder", "hepatobiliary" ], "summary": "A rare concentration of human evolution, pancreatic trial execution and hepatopancreatobiliary clinical work, with matching concentration and sponsor risks.", "content": "Mission capability: link resection, biopsy and rapid-autopsy tissue to common-RAS, vaccine, imaging and hepatobiliary trials. Christine Iacobuzio-Donahue supplies the human evolutionary reconstruction; Eileen O'Reilly supplies multi-centre clinical execution. This directly serves the top bottleneck because resistant clones and recurrence routes can be measured around treatment rather than inferred from mice. Main risk: one institution can control patients, specimens, analysis and sponsor relationships, creating correlated evidence and limits on negative publication. Best role: lead clinical accrual and human tissue while an outside laboratory repeats assays and an independent statistician controls go decisions.", "sourceIds": [ "msk-oreilly-profile-2026", "msk-iacobuzio-donahue-profile-2026" ], "links": [ { "target": "labs", "relation": "part-of" }, { "target": "person-eileen-oreilly", "relation": "employs" }, { "target": "person-christine-iacobuzio-donahue", "relation": "employs" }, { "target": "recommended-strategy", "relation": "supports" } ], "evidence": { "humanProspective": "not-applicable", "humanRetrospective": "not-applicable", "randomised": "not-applicable", "independentReplication": "not-applicable", "sampleSize": "Two linked programme leaders and their institutional portfolios", "effect": "Capability assessment", "limits": [ "Institutional profile evidence", "Current compound and data contracts not reviewed", "Sponsor concentration" ] }, "layout": { "x": 71, "y": 109 }, "updated": "2026-09-14" }, { "id": "lab-nccs-planet-liver", "title": "National Cancer Centre Singapore PLANet — HCC evolution and trial network", "shortTitle": "NCCS PLANet", "type": "lab", "status": "supported", "scope": [ "hcc", "hepatobiliary", "pan-cancer" ], "summary": "The NCCS-led Asia-Pacific network joins multiregion HCC evolution, recurrence tissue and prospective clinical studies across countries with high disease burden.", "content": "Mission capability: distinguish seeded recurrence, polyclonal recurrence and new field cancers using multiregion tissue, then test route-specific adjuvant strategies across an Asia-Pacific clinical network. PLANet's current evidence includes 408 whole-genome and 406 RNA-sequenced primary-tumour samples from 106 surgical patients plus 24 paired recurrences. The official programme spans translational platforms and prospective multicentre studies under Pierce Chow. Best role: lead the HCC recurrence-classification lane and provide a non-Western external test for mission assays. Main risks: only 24 paired recurrences underpin the latest route map, surgical cases exclude much advanced disease, and treatment-sensitivity claims are observational. The group should contribute locked classifiers and samples to an independently analysed randomized trial rather than treating a recurrence score as clinical utility.", "sourceIds": [ "nccs-liver-research-programme-2026", "zhang-2026-planet-hcc-recurrence" ], "links": [ { "target": "labs", "relation": "part-of" }, { "target": "claim-hcc-recurrence-distinct-clonal-routes", "relation": "supports" }, { "target": "model-hcc-causal-chain", "relation": "supports" }, { "target": "failure-imbrave050-interim-benefit-disappeared", "relation": "could-explain" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "strong", "randomised": "none", "independentReplication": "moderate", "sampleSize": "106 multiregion primary tumours and 24 paired recurrences in the latest PLANet analysis", "effect": "Capability assessment; recurrence-route observations require intervention validation", "limits": [ "Small paired-recurrence subset", "Surgical selection", "No validated treatment selector", "Institutional programme claims require delivery testing" ] }, "layout": { "x": 99, "y": 113 }, "updated": "2026-09-14" }, { "id": "lab-nyu-edrn-detection", "title": "NYU Perlmutter pancreatic early-detection and EDRN programme", "shortTitle": "NYU EDRN detection", "type": "lab", "status": "supported", "scope": [ "pdac", "pancreatic-all" ], "summary": "Anirban Maitra's programme links precursor pathology to blinded multi-institution assay validation in inherited risk, cysts and new-onset diabetes.", "content": "Mission capability: move candidate biomarkers from tissue and discovery cohorts into blinded testing across intended-use populations. The NCI early-detection centre spans inherited risk, pancreatic cysts and new-onset diabetes, which allows the same assay to face several base rates. Main risk: a broad cancer-centre leadership role can dilute direct ownership, and biomarker panels repeatedly lose accuracy outside case-control studies. Best role: common validation core under locked protocols and independent statistics, with false positives, work-up harm and negative panels published.", "sourceIds": [ "nyulangone-maitra-profile-2026", "nci-maitra-edrn-2025" ], "links": [ { "target": "labs", "relation": "part-of" }, { "target": "person-anirban-maitra", "relation": "employs" }, { "target": "detection-modality-ranking", "relation": "supports" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "strong", "randomised": "none", "independentReplication": "moderate", "sampleSize": "Multi-institution biomarker validation centre across three risk settings", "effect": "Capability assessment", "limits": [ "Programme description rather than outcome study", "Assay performance not pooled here", "Current grant transfer and contracts need confirmation" ] }, "layout": { "x": 81, "y": 109 }, "updated": "2026-09-14" }, { "id": "lab-onomichi-pancreas-biliary-centre", "title": "JA Onomichi Pancreatic and Biliary Disease Center — community detection route", "shortTitle": "JA Onomichi", "type": "lab", "status": "supported", "scope": [ "pdac", "pancreatic-all", "hepatobiliary" ], "summary": "This Japanese centre links primary-care risk recognition to EUS, selective duct sampling and longitudinal clinical follow-up.", "content": "Mission capability: test whether a complete regional referral route can move diagnosis early enough to change treatment and survival. The centre's advantage is not a unique machine; it is the 2007-onward relationship with community physicians, specialist triage and long-term case capture. The reported stage shift makes it a strong delivery partner and the ENZEAVOUR study adds a new assay workflow. Contracted work should compare regions or introduce the route in steps, record every eligible person and negative follow-up, and separate the value of risk recognition, imaging and invasive testing. Main risks are single-region dependence, specialist skill requirements, invasive pancreatic-juice sampling and lack of randomized mortality evidence.", "sourceIds": [ "ja-onomichi-hanada-profile-2026", "hanada-2026-onomichi-method", "jrct-2026-enzeavour-feasibility" ], "links": [ { "target": "labs", "relation": "part-of" }, { "target": "person-keiji-hanada", "relation": "led-by" }, { "target": "claim-onomichi-community-pathway-stage-shift", "relation": "supports" }, { "target": "trial-enzeavour-japan-screening", "relation": "runs" }, { "target": "detection-integrated-workflows", "relation": "could-test" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "strong", "randomised": "none", "independentReplication": "none", "sampleSize": "Regional pathway 29,642 assessed and 925 cancers", "effect": "Delivery capability with an observed early-stage shift", "limits": [ "Single-region evidence", "No concurrent control", "Skill-intensive work-up", "Institutional willingness and data rights untested" ] }, "layout": { "x": 98, "y": 116 }, "updated": "2026-09-14" }, { "id": "lab-penn-vonderheide-immunity", "title": "Penn Vonderheide laboratory — pancreatic immune mechanism", "shortTitle": "Penn immune lab", "type": "lab", "status": "supported", "scope": [ "pdac", "pan-cancer", "transfer-unproven" ], "summary": "The group joins CD40, vaccines, checkpoint treatment and radiation in mouse-to-human work but faces PDAC's large immune translation gap.", "content": "Mission capability: identify whether antigen presentation, T-cell entry, activation or persistence is the limiting step, then connect measurements to clinical combinations. Main risk: PDAC mouse models often overpredict immune efficacy, while unselected checkpoint trials have near-zero response. Best role: build a patient-specific immune failure assay and reject combinations unless human samples show the intended step changing at tolerable exposure.", "sourceIds": [ "penn-vonderheide-profile-2026", "royal-2010-ipilimumab", "oreilly-2019-durvalumab-tremelimumab" ], "links": [ { "target": "labs", "relation": "part-of" }, { "target": "person-robert-vonderheide", "relation": "employs" }, { "target": "claim-immune-exclusion-is-multicellular", "relation": "tests" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "moderate", "randomised": "moderate", "independentReplication": "moderate", "sampleSize": "Laboratory and clinical programme; no pooled effect", "effect": "Capability assessment against repeated human checkpoint failure", "limits": [ "Model-to-human gap", "No validated immune selector", "Multi-agent trials can obscure contribution" ] }, "layout": { "x": 83, "y": 109 }, "updated": "2026-09-14" }, { "id": "lab-princess-margaret-pancurx", "title": "Princess Margaret and PanCuRx pancreatic precision-treatment network", "shortTitle": "Princess Margaret PanCuRx", "type": "lab", "status": "supported", "scope": [ "pdac", "pan-cancer" ], "summary": "A Canadian clinical and translational network with randomized chemotherapy, whole-genome and RNA analysis, organoid generation and model-directed treatment capability.", "content": "Demonstrated capability: PASS-01 randomized 160 previously untreated metastatic-PDAC patients and generated genome, RNA and organoid correlates across Canadian and US centres. PROSPER-PANC is an active Ontario prospective cohort targeting 200 resectable and advanced cases under one tissue, blood, genome, RNA and organoid route. ADOPT is a recruiting single-site phase 2 that assigns one of ten approved drugs from a model screen in about 25 progressive or maintenance patients. This network can expose the missing route fields and run a treatment policy. It has not yet shown organoid-guidance benefit: PASS-01 did not randomize the correlate, PROSPER-PANC is observational and ADOPT is single-arm. Mission access requires patient-level data, failed-culture records, timing, exact recommendations, treatment delivery and permission for independent analysis. Access unconfirmed.", "sourceIds": [ "knox-2025-pass01", "clinicaltrials-nct05927298-prosper-panc", "clinicaltrials-nct06813079-adopt" ], "links": [ { "target": "labs", "relation": "part-of" }, { "target": "person-jennifer-knox", "relation": "works-with" }, { "target": "person-david-tuveson", "relation": "collaborated-with" }, { "target": "trial-pdac-organoid-action-map", "relation": "could-test" }, { "target": "hypothesis-organoid-guided-early-switch", "relation": "could-test" } ], "evidence": { "humanProspective": "strong — one randomized trial and two current prospective programmes", "humanRetrospective": "moderate", "randomised": "strong — PASS-01 randomized first-line chemotherapy", "independentReplication": "moderate", "sampleSize": "PASS-01 n=160; PROSPER-PANC estimated n=200; ADOPT estimated n=25", "effect": "Execution capability, not proof that organoid guidance improves outcome", "limits": [ "Patient-level access unconfirmed", "Organoid policy was not randomized in PASS-01", "ADOPT is single-site and single-arm", "Treatment and assay ownership need review", "Availability unconfirmed" ], "dimensions": { "biologicalPlausibility": { "rating": "not-applicable", "reason": "This record describes verified programme capability rather than a biological claim." }, "animalEvidence": { "rating": "not-applicable", "reason": "Animal evidence is not needed to establish the public clinical network." }, "inVitroEvidence": { "rating": "strong", "reason": "The programmes explicitly generate and test patient-derived organoids." }, "reproducibility": { "rating": "moderate", "reason": "The capability spans a completed multicentre trial and current observational and interventional work." }, "commercialConflicts": { "rating": "unknown", "reason": "Platform, drug and institutional interests require fresh review before contracting." }, "recency": { "rating": "current", "reason": "Registry and publication evidence was current through June 2026." } } }, "layout": { "x": 139, "y": 195 }, "updated": "2026-09-15" }, { "id": "lab-renji-pancreatic-center", "title": "Renji Hospital Comprehensive Pancreatic Cancer Center", "shortTitle": "Renji pancreatic center", "type": "lab", "status": "supported", "scope": [ "pdac", "pancreatic-all", "pan-cancer" ], "summary": "Liwei Wang's centre supplies a broad Chinese recruitment network and an independent G12D programme moving from phase 1 to blinded phase 3.", "content": "Mission capability: recruit mutation-selected pancreatic patients across China and test whether efficacy and toxicity generalise beyond US and European sponsor networks. The centre led the HRS-4642 plus chemotherapy study and its placebo-controlled first-line phase 3. Main risk: the current response signal has only 30 untreated patients, no controlled drug contribution and 90.3% grade 3 or worse treatment-related adverse events. Best role: independent international replication and common-specimen trial partner, with raw toxicity and dose-intensity data available outside the sponsor.", "sourceIds": [ "shsmu-2026-liwei-wang-profile", "cui-2026-hrs4642-pdac", "nct07232875-hrs4642-phase3" ], "links": [ { "target": "labs", "relation": "part-of" }, { "target": "person-liwei-wang", "relation": "employs" }, { "target": "claim-hrs4642-g12d-chemo-signal", "relation": "supports" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "none", "randomised": "none", "independentReplication": "none", "sampleSize": "Phase 1b/2 PDAC cohort n=31; first-line treated n=30; phase 3 recruiting", "effect": "Capability and international-replication assessment", "limits": [ "Sponsor-linked uncontrolled signal", "High combined toxicity", "Phase 3 outcome unknown", "Institutional profile supplies role, not performance" ] }, "layout": { "x": 87, "y": 109 }, "updated": "2026-09-14" }, { "id": "lab-tata-varanasi-gallbladder", "title": "Tata Memorial Centre Varanasi — endemic gallbladder pathway", "shortTitle": "Tata Varanasi", "type": "lab", "status": "supported", "scope": [ "gallbladder", "hepatobiliary" ], "summary": "A high-volume North Indian centre can test prevention, referral and treatment-delivery rules where gallbladder cancer is common and four in five new patients present metastatic.", "content": "Mission capability: build a complete high-incidence gallbladder pathway from gallstone risk through diagnostic referral, surgery, systemic treatment and survival. The Varanasi units operate under Tata Memorial Centre and the prospective 1,500-patient cohort demonstrates enough volume to test operational changes quickly. Best role: lead an Indian multi-centre risk and earlier-presentation programme, with population denominators and rural referral data rather than another resection-only series. Main risk: the current evidence is one treatment-centre cohort. It cannot estimate population incidence, referral loss or causal treatment benefit, and a high-volume endemic centre may not represent low-incidence health systems. A mission contract should require consecutive capture before tertiary referral, reasons planned treatment was not received, and external centres across North and Northeast India.", "sourceIds": [ "mpmmcc-varanasi-profile-2026", "vineet-2026-north-india-gallbladder-cohort" ], "links": [ { "target": "labs", "relation": "part-of" }, { "target": "claim-gallbladder-endemic-late-presentation", "relation": "supports" }, { "target": "claim-gallbladder-prevention-selection-bottleneck", "relation": "could-test" }, { "target": "model-gallbladder-causal-chain", "relation": "supports" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "none", "independentReplication": "none", "sampleSize": "Prospective gallbladder cohort n=1,500 at the centre", "effect": "Capability assessment; 80.2% metastatic presentation establishes the pathway gap", "limits": [ "Single-centre evidence", "No population denominator", "No intervention comparison", "Institutional willingness and data rights untested" ] }, "layout": { "x": 96, "y": 113 }, "updated": "2026-09-14" }, { "id": "lab-tuveson-organoid-models", "title": "Tuveson laboratory — organoids and engineered pancreatic models", "shortTitle": "Tuveson models", "type": "lab", "status": "supported", "scope": [ "pdac", "pancreatic-all", "pan-cancer" ], "summary": "A major pancreatic organoid and mouse-model capability is useful for rejecting combinations quickly, not declaring clinical success.", "content": "Mission capability: organoids, engineered mice, RAS biology, redox metabolism and tumour–fibroblast interaction using a large human-derived collection. Its highest value is throughput: weak mechanistic combinations can be killed before a trial and promising ones can be tested across diverse states. Main risk: organoids and mice omit human immune history, treatment exposure, pharmacology and metastatic delivery. Organisational note: David Tuveson was still at Cold Spring Harbor Laboratory on 14 September 2026, with a move to lead Duke Cancer Institute announced for 31 October 2026. Contract the specimen and model capability, not an institution name that is about to change.", "contentSections": [ "PASS-01 supplies the clinical throughput limit: organoids were generated in 50%, while the second-line window was short. The laboratory remains valuable for drug rejection and cross-laboratory testing; clinical guidance requires the separate all-patient route and randomized utility gate." ], "sourceIds": [ "cshl-tuveson-profile-2026", "cshl-tuveson-duke-transition-2026", "tiriac-2018-organoid-profiling", "knox-2025-pass01" ], "links": [ { "target": "labs", "relation": "part-of" }, { "target": "person-david-tuveson", "relation": "employs" }, { "target": "technology-ai-organoid-active-learning", "relation": "supports" }, { "target": "trial-pdac-organoid-action-map", "relation": "could-test" } ], "evidence": { "humanProspective": "none", "humanRetrospective": "moderate", "randomised": "none", "independentReplication": "strong", "sampleSize": "Large model and organoid programme; no single pooled cohort", "effect": "Capability assessment", "limits": [ "Model-to-human gap", "Institutional transition in progress", "Human response prediction remains unproved" ] }, "layout": { "x": 77, "y": 109 }, "updated": "2026-09-15" }, { "id": "lab-uct-groote-schuur-hpb", "title": "UCT and Groote Schuur Surgical Gastroenterology — African HPB delivery route", "shortTitle": "UCT–Groote Schuur HPB", "type": "lab", "status": "supported", "scope": [ "pdac", "pnet", "cholangiocarcinoma", "hcc", "hepatobiliary", "pancreatic-all" ], "summary": "The Cape Town unit is an evidenced route into pancreatic, biliary and liver care under the capacity and access constraints the mission must solve in Africa.", "content": "Mission capability: join pancreatic surgery, biliary drainage, liver disease, palliative care, biomarker work and prospective patient follow-up in one South African academic hospital route. The UCT directory names pancreatic adenocarcinoma, GEP-NET and hilar-obstruction projects; the IHPBA paper makes access, prevention, screening, curative treatment, palliation, infrastructure and policy part of the same HCC system. This is now backed by two patient comparisons. In Cape Town, 551 public and 51 private HCC patients had median survival of 68 versus 703 days, with much later public-sector disease; the small selected resection and ablation groups had no significant survival difference. In a 959-person South Africa-Sweden comparison, curative-intent treatment was 9.2% versus 42.5% and survival remained lower within treatment groups. Together these show both an operating clinical unit and losses before and after treatment selection. They do not identify one causal repair. A contract should measure consecutive eligibility, diagnosis stage, liver function, pathology completion, referral loss, treatment delivery, travel cost, time to care and survival before testing the largest measured loss. The unit is not a proxy for Africa: it needs linked sites, a population denominator and an independent data team.", "sourceIds": [ "uct-2026-cancer-projects", "jonas-2025-sub-saharan-hcc-pathways", "asombang-2026-africa-hpb-registry", "sobnach-2024-ssa-hcc-systematic-review", "sobnach-2025-south-africa-hcc-disparities", "sobnach-2026-global-hcc-disparities" ], "links": [ { "target": "labs", "relation": "part-of" }, { "target": "person-eduard-jonas", "relation": "led-by" }, { "target": "lab-africa-hpb-cancer-consortium", "relation": "could-collaborate-with" }, { "target": "programme-unconstrained-global-system", "relation": "could-test" }, { "target": "person-sanju-sobnach", "relation": "contains" }, { "target": "trial-regional-route-completion-action-map", "relation": "could-run" } ], "evidence": { "humanProspective": "none for a tested delivery repair", "humanRetrospective": "strong for outcome and access description", "randomised": "none", "independentReplication": "moderate for the existence of a route gap; none for a repair", "sampleSize": "Cape Town public/private HCC cohort n=602; South Africa/Sweden comparison n=959; pancreatic and biliary project denominators not reported", "effect": "Observed stage, treatment and survival gaps plus a delivery-site assessment; no causal intervention effect", "limits": [ "Single South African academic route", "No completed multi-country registry shown", "No randomized delivery repair or mortality result", "Public/private and international comparisons have major baseline differences", "Selected treatment-group comparisons do not remove confounding", "Capacity, contracting and data access untested" ], "dimensions": { "biologicalPlausibility": { "rating": "not-applicable", "reason": "This is a delivery-centre assessment." }, "animalEvidence": { "rating": "not-applicable", "reason": "Animal work cannot validate a care route." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Laboratory evidence is not the basis for the recommendation." }, "reproducibility": { "rating": "moderate", "reason": "Two HCC comparisons reproduce a large route gap, but the repair has not been tested across African health systems." }, "commercialConflicts": { "rating": "unknown", "reason": "Current sponsor and product ties have not been audited." }, "recency": { "rating": "current", "reason": "The route record includes a 2026 international HCC comparison and current UCT projects." } } }, "layout": { "x": 63, "y": 121 }, "updated": "2026-09-14" }, { "id": "lab-zhejiang-pancreatic-centre", "title": "Zhejiang University pancreatic centre — positive sequence with an external-validity debt", "shortTitle": "Zhejiang pancreas centre", "type": "lab", "status": "supported", "scope": [ "pdac" ], "summary": "The Hangzhou centre delivered a 324-person randomized phase 3 timing trial, but one-centre execution and its tumour-location mix limit transfer.", "content": "Current operating evidence: the Second Affiliated Hospital of Zhejiang University sponsored and solely hosted NCT03750669. It completed 324 randomized patients and reported a positive primary event-free-survival result after sequential gemcitabine/nab-paclitaxel and modified FOLFIRINOX before surgery. Median event-free survival was 15.3 versus 10.9 months, HR .71; median overall survival was 35.4 versus 27.2 months, HR .73 with 95% CI .53-1.00. Half of patients had body or tail tumours, unlike many Whipple-heavy western trials. Severe events were 47.6% versus 30.7%. Mission role: expose the exact delivery system that made the complex regimen feasible, provide patient-level data and code, and reproduce it across outside centres before adoption. Required access: regimen deviations, all allocated-patient flow, resection and full-course completion, pathology review, tumour-location interaction, serious harm, recurrence and survival, plus independent analysis and negative-result publication. The centre proves execution in one setting, not general transfer.", "sourceIds": [ "bai-2025-cispd1" ], "links": [ { "target": "labs", "relation": "part-of" }, { "target": "trial-pdac-perioperative-sequencing-action-map", "relation": "ran" }, { "target": "person-tingbo-liang", "relation": "includes" }, { "target": "unknown-pdac-perioperative-treatment-route", "relation": "tests" }, { "target": "lab-alliance-a021806-network", "relation": "needs-independent-replication-by" } ], "evidence": { "humanProspective": "strong — completed randomized phase 3", "humanRetrospective": "not-applicable", "randomised": "strong single-centre result", "independentReplication": "none", "sampleSize": "CISPD-1 n=324", "effect": "Capability and access record; EFS HR .71 and OS HR .73, 95% CI .53-1.00", "limits": [ "Single centre", "Complex sequence", "Half body or tail tumours", "Higher severe-event rate", "Patient-level access unconfirmed", "External reproduction absent" ], "dimensions": { "biologicalPlausibility": { "rating": "not-applicable", "reason": "This record assesses trial delivery and transfer." }, "animalEvidence": { "rating": "not-applicable", "reason": "Not relevant to institutional capability." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Not relevant to institutional capability." }, "reproducibility": { "rating": "weak", "reason": "A completed phase 3 trial proves internal delivery, while outside centres have not reproduced the regimen." }, "commercialConflicts": { "rating": "unknown", "reason": "Complete current funding, product-supply and data-control terms require direct review." }, "recency": { "rating": "current", "reason": "The trial completed and reported in 2025." } } }, "updated": "2026-09-14" }, { "id": "labs", "title": "Laboratories and institutions ranked by mission capability", "shortTitle": "Labs", "type": "lab", "status": "supported", "scope": [ "pdac", "pnet", "cholangiocarcinoma", "gallbladder", "hcc", "colorectal", "pan-cancer", "transfer-unproven" ], "summary": "No single centre spans human evolution, risk cohorts, rapid models, RAS resistance, immune manufacture, AI and independent global trials; the mission needs a contracted network.", "content": "The strongest unit is a network, not one famous laboratory. Memorial Sloan Kettering's Rubenstein Center covers human evolution and multi-centre treatment but concentrates sponsor relationships. Dana-Farber covers RAS resistance; Johns Hopkins inherited-risk surveillance; Tuveson and Cancer Research UK Scotland rapid model rejection; MSK Olayan vaccine manufacture; NYU and EDRN blinded detection validation; Penn immune mechanism; and MIT locked external AI testing. The Beth Israel Deaconess thrombosis and haemostasis network adds PDI assays, cancer-coagulation trial design and direct CATIQ access, but it needs outside assay transfer and independent phase-3 analysis. The international map includes Renji's Chinese trial network, CNIO's Spanish RAS models and epidemiology group, JA Onomichi's Japanese community detection route, Imperial's prospective UK VAPOR2 network, Tata Varanasi's high-volume gallbladder pathway and NCCS PLANet's Asia-Pacific HCC work. Regional delivery now has four distinct routes: UCT and Groote Schuur with patient-level HCC comparisons; the Africa HPB Cancer Consortium roadmap; LACOG GASPAR across Brazil, Argentina and Colombia; and the five-country LATAM-CCA Registry. The names no longer hide the evidence limits. African pancreatic records lack stage for nearly 40% of 26,850 cases, and a sub-Saharan HCC review found reports from only 15 of 48 countries. GASPAR has 79 pancreatic cases, 76% private coverage across its combined cohort and a 201-versus-120 enrolment conflict. LATAM-CCA has 309 eligible referral-centre cases but no population denominator and excludes missing mandatory records. Contracts must separate discovery, assay, trial, statistics and replication control, and require consecutive capture, route dates, reasons for non-delivery, common fields, complete failed-result release, patient-level data access and declared product interests.", "sourceIds": [ "msk-iacobuzio-donahue-profile-2026", "danafarber-aguirre-profile-2026", "hopkins-goggins-profile-2026", "cshl-tuveson-profile-2026", "msk-balachandran-profile-2026", "nyulangone-maitra-profile-2026", "penn-vonderheide-profile-2026", "mit-barzilay-profile-2026", "shsmu-2026-liwei-wang-profile", "cnio-experimental-oncology-2026", "cnio-malats-profile-2026", "cordis-pancaid-101096309", "ja-onomichi-hanada-profile-2026", "hanada-2026-onomichi-method", "nct07243262-vapor2", "imperial-hanna-profile-2026", "cruk-scotland-morton-lab-2026", "mpmmcc-varanasi-profile-2026", "nccs-liver-research-programme-2026", "uct-2026-cancer-projects", "jonas-2025-sub-saharan-hcc-pathways", "asombang-2026-africa-hpb-registry", "clinicaltrials-nct05924789-gaspar", "peixoto-2025-gaspar-results", "gismalla-2026-africa-pancreatic-review", "sobnach-2024-ssa-hcc-systematic-review", "sobnach-2025-south-africa-hcc-disparities", "sobnach-2026-global-hcc-disparities", "da-fonseca-2024-latam-cca", "zwicker-2019-isoquercetin-catiq", "clinicaltrials-nct06861088-catiq-p3", "patent-wo2020112828-isoquercetin-cancer", "daamen-2024-radar-panc-protocol", "clinicaltrials-nct04875325-radar-panc", "clinicaltrials-nct04340141-a021806", "bai-2025-cispd1", "dotan-2024-giant", "dotan-2025-giant-ga", "clinicaltrials-nct04233866-giant", "betge-2022-grantpax", "clinicaltrials-nct02143219-pamela70", "mohile-2021-gap70", "li-2021-gain", "ihorst-2026-geriatric-comanagement", "clinicaltrials-nct06040801-frail-gi" ], "links": [ { "target": "lab-princess-margaret-pancurx", "relation": "contains" }, { "target": "people", "relation": "depends-on" }, { "target": "lab-msk-rubenstein", "relation": "contains" }, { "target": "lab-danafarber-hale-ras", "relation": "contains" }, { "target": "lab-hopkins-early-detection", "relation": "contains" }, { "target": "lab-tuveson-organoid-models", "relation": "contains" }, { "target": "lab-msk-olayan-vaccines", "relation": "contains" }, { "target": "lab-nyu-edrn-detection", "relation": "contains" }, { "target": "lab-penn-vonderheide-immunity", "relation": "contains" }, { "target": "lab-mit-jameel-clinical-ai", "relation": "contains" }, { "target": "lab-renji-pancreatic-center", "relation": "contains" }, { "target": "lab-cruk-scotland-morton", "relation": "contains" }, { "target": "lab-cnio-experimental-oncology", "relation": "contains" }, { "target": "lab-tata-varanasi-gallbladder", "relation": "contains" }, { "target": "lab-nccs-planet-liver", "relation": "contains" }, { "target": "lab-onomichi-pancreas-biliary-centre", "relation": "contains" }, { "target": "lab-imperial-vapor2", "relation": "contains" }, { "target": "lab-cnio-genetic-epidemiology", "relation": "contains" }, { "target": "lab-uct-groote-schuur-hpb", "relation": "contains" }, { "target": "lab-africa-hpb-cancer-consortium", "relation": "contains" }, { "target": "lab-lacog-gaspar", "relation": "contains" }, { "target": "lab-latam-cca-registry", "relation": "contains" }, { "target": "lab-bidmc-thrombosis-pdi", "relation": "contains" }, { "target": "lab-dpcg-radar-panc-network", "relation": "contains" }, { "target": "lab-alliance-a021806-network", "relation": "contains" }, { "target": "lab-zhejiang-pancreatic-centre", "relation": "contains" }, { "target": "lab-ecog-acrin-geriatric-oncology-network", "relation": "contains" } ], "evidence": { "humanProspective": "not-applicable", "humanRetrospective": "not-applicable", "randomised": "not-applicable", "independentReplication": "not-applicable", "sampleSize": "Twenty-four laboratory, centre or network records across the US, Canada, UK, Netherlands, Spain, Japan, China, India, Singapore, South Africa and wider African and Latin American networks", "effect": "Capability map and network design, not a clinical-effect claim", "limits": [ "Selection is judgement", "Institutional profiles describe capability but not delivery quality", "Contracts and willingness are untested", "Africa has a named South African route and consortium roadmap, not proven continent-wide delivery", "Latin America has two selected referral networks, not representative population coverage", "Several detection programmes remain observational or use stored samples" ] }, "layout": { "x": 69, "y": 106 }, "updated": "2026-09-15", "contentSections": [ "The Dutch Pancreatic Cancer Group RADAR-PANC network adds an existing Netherlands-UK randomized route for post-resection follow-up. Its value is the enrolled cohort, survival and quality-of-life design, and ten Dutch sites plus Birmingham—not an assumption that its optimistic effect size or September 2024 registry status remains current. Access requires direct confirmation of accrual, crossover, data maturity, analysis rights and a result timetable.", "Two perioperative networks are separated by what they can prove. Alliance and NCI completed 358-person multicentre A021806 accrual and now owe mature survival and full-route reporting. Zhejiang's centre completed the positive 324-person CISPD-1 trial but owes patient-level access and outside-centre reproduction because the regimen, tumour-location mix and delivery system are unusual. Neither network alone settles universal timing.", "The ECOG-ACRIN geriatric-oncology network adds a demonstrated route across 92 enrolling US sites. It can supply the GIANT field set and patient-level pathway only if consent, data access, missingness and independent analysis pass. Trial reach is capability, not proof that the network can transfer unchanged to the NHS.", "The Princess Margaret and PanCuRx network adds a current organoid-route capability: PASS-01 randomized first-line treatment, PROSPER-PANC follows an estimated 200 people and ADOPT tests model-directed therapy in about 25. Patient-level access, failed-culture records and independent analysis remain unconfirmed." ] }, { "id": "mission-coverage", "title": "Mission coverage: what the atlas answers and what remains unknown", "shortTitle": "Coverage map", "type": "system", "status": "supported", "scope": [ "pdac", "pnet", "cholangiocarcinoma", "gallbladder", "hcc", "colorectal", "pan-cancer", "transfer-unproven" ], "summary": "All twenty requested dossier sections resolve to one linked evidence system. The current release contains 682 connected research pages, 864 unique used source records and 488 fully scored decision records after 79 directed passes. Search has stopped changing the five-gap ranking or programme allocation; scientific unknowns remain explicit.", "content": "Start with the mission and reading guide, then open the five largest gaps and recommended strategy. Those pages state the decision. Disease systems, detection, treatment, trials and failures let the reader challenge it. Evidence quality shows why a claim is strong or weak. Unknowns, experiments and risks show what could reverse it. Companies, investment, patents, people, laboratories, datasets, AI and funding show who or what could execute it. The programme pages turn the same records into £5m, £25m, £100m, £500m and unconstrained plans.\n\nCoverage achieved. The atlas separates PDAC, pNET, cholangiocarcinoma, gallbladder cancer, HCC and colorectal cancer rather than pooling them. It contains all twenty requested dossier sections plus evidence, trial, patent, dataset, company, investor, person, laboratory, timeline and search-audit pages. All 682 pages are reachable from the mission through the graph. All 864 source records are attached to at least one page. The 488 decision records that can change a programme, treatment ranking, experiment or capital gate expose the full evidence breakdown. Duplicate URLs, DOIs, publication-record PubMed IDs and repeated source IDs within a page are absent. Research build and audit scripts reproduce those counts and reject missing links, disconnected pages, unused sources, uncited claims and unscored programme records.\n\nWhy searching stopped. Seventy-nine directed passes covered supporting and opposing results, failures, recent trials, non-US and non-UK work, company and capital claims, patents, adjacent fields, source access and new regulator actions. The last seven passes changed useful details: four hepatobiliary trial boundaries, one pNET sequencing question, one cheap nutrition-replication candidate, one negative complementary-care result, one family-testing delivery signal, one AI reader limit, one pathology-AI selection signal and two treatment-access labels. The final sweeps found TORCH, a large state-specific HCC survival result, and PANCprAId, a retrospective treatment-interaction signal from randomized-trial slides; they corrected the HCC and AI pages without changing the five main bottlenecks, releasing programme money or changing the company ranking. Further search now reopens only for a registered trial result, regulator action, independently reproduced patient outcome, failed core assumption or evidence crossing a stated stop or release threshold. This is a decision rule, not a claim that no future discovery matters.\n\nPermanent limits. Public records cannot establish complete freedom to operate: exact claim mapping, live national patent status, combination rights and legal risk still require patent counsel. Global representation remains uneven because many regions lack common stage, treatment-route and population-denominator data. Some recent results remain sponsor-controlled, some trial records are stale, and future follow-up can reverse current estimates. The atlas therefore distinguishes a completed research and publication task from a solved cancer problem. It does not claim that these cancers are now reliably preventable, detectable or curable. It identifies the smallest measured failures that must change, the tests that could show change, and the conditions under which capital should stop.\n\nPublication gate. The final source check tested all 864 URLs: 798 were directly accessible, 64 returned access-control responses, two failed at the network or server layer, and none returned a persistent not-found response. Search-engine retrieval independently confirmed both network failures; access status does not establish claim truth. Completion also requires the research build, mission audit, application lint and production build to pass, then the exact public route must serve this release. The release remains a research and capital-allocation tool, not personal medical advice.", "sourceIds": [ "beaufils-2026-pancpraid", "lyu-2026-torch-hcc", "oreilly-2026-daraxonrasib", "conroy-2022-prodige24-five-year", "blackford-2024-high-risk-surveillance", "sasieni-2026-nhs-galleri", "tie-2022-dynamic-colon", "qin-2025-cares310-final", "subbiah-2023-rly4008", "patent-wo2026015825-daraxonrasib-pdac", "patent-wo2022268051-hengrui-kras-g12d", "patent-wo2024022507-hengrui-liposome", "asombang-2026-africa-hpb-registry", "peixoto-2025-gaspar-results", "gismalla-2026-africa-pancreatic-review", "sobnach-2024-ssa-hcc-systematic-review", "sobnach-2025-south-africa-hcc-disparities", "sobnach-2026-global-hcc-disparities", "da-fonseca-2024-latam-cca", "fda-2026-bizengri-cholangiocarcinoma", "ema-2026-daraxonrasib-phased-review" ], "links": [ { "target": "reading-guide", "relation": "depends-on" }, { "target": "mission", "relation": "audits" }, { "target": "evidence-quality", "relation": "depends-on" }, { "target": "disease-system", "relation": "covers" }, { "target": "detection", "relation": "covers" }, { "target": "treatment", "relation": "covers" }, { "target": "trials", "relation": "covers" }, { "target": "failures", "relation": "covers" }, { "target": "ranking-provisional-pdac-bottlenecks", "relation": "audits" }, { "target": "research-frontier-ranking", "relation": "covers" }, { "target": "companies", "relation": "covers" }, { "target": "investment", "relation": "covers" }, { "target": "ai-programme", "relation": "covers" }, { "target": "people", "relation": "covers" }, { "target": "transferability-matrix", "relation": "covers" }, { "target": "experiments", "relation": "covers" }, { "target": "programme", "relation": "covers" }, { "target": "funding", "relation": "covers" }, { "target": "risks", "relation": "covers" }, { "target": "unknowns", "relation": "covers" }, { "target": "recommended-strategy", "relation": "audits" }, { "target": "patent-landscape", "relation": "open-gap" }, { "target": "trial-regional-route-completion-action-map", "relation": "covers" }, { "target": "research-stopping-audit", "relation": "depends-on" } ], "evidence": { "humanProspective": "strong — linked decision records", "humanRetrospective": "strong — linked disease and detection records", "randomised": "strong — linked treatment and failure records", "independentReplication": "moderate — uneven across individual claims and stated on each record", "sampleSize": "682 connected research pages, 864 unique used source records, 488 fully scored decision records and 79 directed research passes. Final URL audit: 798 accessible, 64 access-controlled, 2 network or server failures and 0 persistent not-found targets", "effect": "Coverage and release audit; it tests whether every requested question resolves to checked records, not whether cancer control has been achieved", "limits": [ "Structural and publication completion do not mean that the scientific problem is solved", "Recent sponsor-controlled evidence and future follow-up may change", "Patent legal analysis and representative global delivery data remain incomplete", "The stopping rule is tied to decisions and can reopen when new evidence crosses a stated threshold" ], "dimensions": { "biologicalPlausibility": { "rating": "not-applicable", "reason": "This record audits research coverage rather than proposing a biological mechanism." }, "animalEvidence": { "rating": "not-applicable", "reason": "Animal results belong to linked claims and do not establish coverage." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Laboratory results belong to linked claims and do not establish coverage." }, "reproducibility": { "rating": "strong", "reason": "Automated checks reproduce record counts, source use, graph reach and required decision fields; scientific judgement remains inspectable in each record." }, "commercialConflicts": { "rating": "moderate", "reason": "Several recent treatment, diagnostic and company records depend partly on sponsor-controlled reports, and each is labelled." }, "recency": { "rating": "current", "reason": "The audit describes the atlas as checked on 15 September 2026." } } }, "layout": { "x": 53, "y": 18 }, "updated": "2026-09-15" }, { "id": "mission", "title": "Make lethal gastrointestinal cancers preventable, detectable and controllable", "shortTitle": "Mission", "type": "mission", "status": "supported", "scope": [ "pdac", "pnet", "cholangiocarcinoma", "gallbladder", "hcc", "colorectal" ], "summary": "PDAC will not become reliably curable through detection alone. The first missing capability is safe eradication of occult systemic disease, built around durable common-RAS control and a treatment-linked residual-disease measurement loop.", "content": "Shortest defensible answer: make systemic treatment work durably at low tumour burden, then connect detection to that treatment. PDAC is often systemic before it is clinically visible, and current local treatment does not remove the main risk: in selected resected patients receiving mFOLFIRINOX, five-year disease-free survival was 26.1%, so 73.9% had recurrence or death. Moving every pancreatic cancer to today's localised-stage outcome would be a large gain—43.6% rather than 13.7% five-year relative survival—but would still leave a 56.4-point deficit from 100% five-year relative survival. Detection alone is therefore not the endpoint. The strongest new treatment fact is that broad RAS inhibition works in humans: daraxonrasib increased median overall survival from 6.6 to 13.2 months after prior treatment in RAS-G12 PDAC. Median progression at 7.3 months shows that targetability is proved but durable control is not. The first programme should measure residual disease and resistance prospectively around common-RAS therapy, reproduce the resulting treatment rules in two laboratories, and randomise only rules that can be delivered safely and fast enough to change care. Release £25m for the four-year observatory; release the next £75m only after one reproducible resistance or residual-disease rule survives the stated gates. Prevention remains important but presently covers only fractions of risk: avoidable exposures can reduce incidence, inherited surveillance reaches a minority, and no common removable cause or safe dangerous-precursor rule exists. General-population screening is unsafe at today's base rate unless specificity is extraordinary: at 13.9 cases per 100,000 people per year, even a 99%-specific, 100%-sensitive test would produce about 1,000 false positives for 13.9 true cancers, or roughly 72 false positives per true positive. pNET, cholangiocarcinoma, gallbladder cancer, HCC and colorectal cancer require separate causal models. Their transferable lessons are methods—viral prevention, precursor removal, molecular selection, residual-disease trials and radioligand delivery—not an assumption that the same biology applies. The ten-year target is not a universal cure claim. It is at least one randomised, practice-changing result that reduces recurrence or extends survival materially in a defined group, with every negative result and stopping decision retained. Current planning probability is 30–50% for at least one practice-changing result by 2036, 10–20% for a material population mortality reduction in a defined group, and below 10–15% that most newly treated PDAC becomes durably controlled. The largest avoidable error would be spending as if a biomarker, model, response rate or company valuation were already a patient outcome.", "sourceIds": [ "gudmundsdottir-2023-staging-laparoscopy", "rompen-2026-preopanc2-staging", "theijse-2024-nontherapeutic-laparotomy", "conroy-2022-prodige24-five-year", "seer-pancreas-statfacts", "oreilly-2026-daraxonrasib", "hu-2018-germline-risk", "uspstf-2019-pancreatic-screening", "botta-2024-tumor-informed-ctdna", "tie-2022-dynamic-colon", "chang-1997-hbv-vaccine-hcc", "nhs-england-2026-r367", "wang-2022-oncology-germline-cascade", "katz-2026-gift-cascade", "npaca-2026-state-of-nation", "nhs-england-2024-hpb-pancreatic-service", "nice-2018-ng85-pancreatic", "de-wilde-2012-centralization", "lemmens-2011-centralization", "latenstein-2021-centralization", "coupland-2016-england-resection-volume", "labori-2016-adjuvant-completion", "henry-2023-complications-adjuvant", "kollbeck-2025-complications-chemo", "van-hilst-2019-leopard2", "korrel-2023-diploma", "bruna-2025-diploma-long-term", "klotz-2024-europa", "liu-2024-robotic-open-pd", "jin-2026-portal", "andel-2025-routine-imaging", "van-goor-2025-nationwide-imaging", "daamen-2024-radar-panc-protocol", "clinicaltrials-nct04875325-radar-panc", "tai-2025-pan-asian-esmo-pancreatic", "labori-2024-norpact1", "bai-2025-cispd1", "tan-2026-neoadjuvant-ipd-meta", "ghaneh-2023-espac5", "clinicaltrials-nct04340141-a021806", "effenberger-2012-pdac-bone-marrow-dtc", "nordgard-2022-unresectable-dtc", "chalabi-hajkarim-2025-organotropism", "pei-2025-spatial-metastatic-pdac", "mcclellan-2026-quiescent-pdac-car-t", "tanaka-2012-microscopic-local-recurrence", "dotan-2024-giant", "dotan-2025-giant-ga", "clinicaltrials-nct04233866-giant", "betge-2022-grantpax", "clinicaltrials-nct02143219-pamela70", "mohile-2021-gap70", "li-2021-gain", "ihorst-2026-geriatric-comanagement", "clinicaltrials-nct06040801-frail-gi", "stoop-2026-preopanc4-surgery", "dekker-2026-taps-restaging", "seelen-2026-pelican", "kunzmann-2021-neolap", "napoli-2025-surgery-versus-oncology" ], "links": [ { "target": "reading-guide", "relation": "explained-by" }, { "target": "claim-pdac-organoid-correlation-not-routing-utility", "relation": "depends-on" }, { "target": "unknown-pdac-organoid-guided-treatment-utility", "relation": "leaves-open" }, { "target": "claim-staging-laparoscopy-avoids-incision-not-proven-survival", "relation": "includes" }, { "target": "hypothesis-pdac-two-timepoint-staging-route", "relation": "tests" }, { "target": "pdac", "relation": "part-of" }, { "target": "pnet", "relation": "part-of" }, { "target": "hepatobiliary", "relation": "part-of" }, { "target": "colorectal", "relation": "part-of" }, { "target": "bottlenecks", "relation": "ranks" }, { "target": "mission-coverage", "relation": "audited-by" }, { "target": "people", "relation": "depends-on" }, { "target": "companies", "relation": "depends-on" }, { "target": "investment", "relation": "depends-on" }, { "target": "ai-programme", "relation": "depends-on" }, { "target": "funding", "relation": "depends-on" }, { "target": "risks", "relation": "depends-on" }, { "target": "timeline", "relation": "depends-on" }, { "target": "claim-universal-germline-testing-needs-action", "relation": "depends-on" }, { "target": "unknown-pdac-germline-family-action-route", "relation": "depends-on" }, { "target": "hypothesis-pdac-germline-family-route-completion", "relation": "tests" }, { "target": "claim-pdac-surgery-is-a-complete-treatment-route", "relation": "links-to" }, { "target": "system-england-pancreatic-surgery-network", "relation": "links-to" }, { "target": "failure-pdac-postoperative-treatment-loss", "relation": "links-to" }, { "target": "failure-leopard2-laparoscopic-whipple-safety", "relation": "links-to" }, { "target": "unknown-pdac-surgical-route-completion", "relation": "links-to" }, { "target": "trial-pdac-surgical-route-action-map", "relation": "links-to" }, { "target": "hypothesis-pdac-surgical-route-completion", "relation": "links-to" }, { "target": "claim-routine-imaging-after-pdac-resection-not-yet-proven", "relation": "bounded-by" }, { "target": "unknown-postresection-surveillance-utility", "relation": "bounded-by" }, { "target": "trial-pdac-postresection-surveillance-action-map", "relation": "bounded-by" }, { "target": "claim-neoadjuvant-therapy-improves-selection-not-cure", "relation": "bounded-by" }, { "target": "failure-norpact1-neoadjuvant-delivery", "relation": "bounded-by" }, { "target": "unknown-pdac-perioperative-treatment-route", "relation": "bounded-by" }, { "target": "trial-pdac-perioperative-sequencing-action-map", "relation": "bounded-by" }, { "target": "claim-human-pdac-dtc-prognostic-not-dormancy", "relation": "bounded-by" }, { "target": "failure-pdac-dormancy-sample-gap", "relation": "bounded-by" }, { "target": "trial-pdac-residual-cell-state-action-map", "relation": "bounded-by" }, { "target": "hypothesis-human-dormancy-reawakening-gate", "relation": "bounded-by" }, { "target": "claim-age-alone-cannot-route-pdac-treatment", "relation": "bounded-by" }, { "target": "failure-grantpax-geriatric-route", "relation": "bounded-by" }, { "target": "unknown-older-pdac-treatment-route", "relation": "bounded-by" }, { "target": "trial-older-pdac-action-map", "relation": "bounded-by" }, { "target": "hypothesis-older-pdac-fast-support-route", "relation": "tests" }, { "target": "claim-lapc-response-warrants-surgical-reassessment", "relation": "bounded-by" }, { "target": "failure-conversion-surgery-selection-bias", "relation": "bounded-by" }, { "target": "unknown-lapc-surgery-causal-benefit", "relation": "tests" }, { "target": "trial-lapc-conversion-surgery-action-map", "relation": "bounded-by" }, { "target": "hypothesis-randomized-lapc-resection-after-response", "relation": "could-fund-after-review" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "strong", "randomised": "strong", "independentReplication": "strong", "sampleSize": "Synthesis across population registry data, randomized treatment trials, prospective surveillance cohorts and residual-disease studies", "effect": "Decision synthesis; the cited treatment and survival values are direct human outcomes, while programme probabilities are planning judgements", "limits": [ "No single study tests the full strategy", "SEER pancreatic data mix histologies", "Detection-only calculation is an upper-bound thought experiment", "Programme costs and probabilities are judgement", "New first-line and postoperative RAS trials can change the ranking" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The conclusion follows human recurrence, metastatic progression and common-RAS response rather than a model alone." }, "animalEvidence": { "rating": "not-applicable", "reason": "Animal evidence is not required for the mission-level conclusion drawn from human outcomes." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Laboratory activity is not used as proof that the proposed clinical system works." }, "reproducibility": { "rating": "moderate", "reason": "The component failures repeat, while the full treatment-linked measurement strategy remains untested." }, "commercialConflicts": { "rating": "moderate", "reason": "The conclusion includes manufacturer-funded RAS evidence alongside public trials and registries." }, "recency": { "rating": "current", "reason": "The synthesis includes current 2026 treatment, detection and trial evidence." } } }, "layout": { "x": 50, "y": 48 }, "updated": "2026-09-15", "contentSections": [ "The surgical route now starts before incision. A standardized camera check can detect scan-hidden surface spread and avoid an open exploration, but positive yield alone is not survival. The £500m mission nests a 500-person timing-and-action check inside the existing £2.5m English surgical route, separates biopsy-confirmed spread from wash cytology and stops for procedure harm, treatment delay or false denial of surgery. No capital is added.", "The surgical-route pass prevents a major counting error: an operation is not a completed curative route. The atlas now links current English commissioning and audit to centralization cohorts, randomized open, laparoscopic and robotic comparisons, postoperative complications and treatment completion. The mission test is full planned care without greater 90-day harm, not more procedures, shorter stay or faster recovery alone.", "Post-resection follow-up is an evidence-recovery task before it is a new programme. Two observational cohorts associate scheduled imaging with more recurrence treatment and longer survival, but neither randomizes the schedule. RADAR-PANC already tests three-monthly CA19-9 plus CT against mainly symptom-led follow-up with overall survival primary; CIRCPAC separately tests ctDNA-directed scan intensity. Finish and align those studies before funding another surveillance trial.", "Treatment timing is now a bounded open route. Nine randomized resectable-disease trials support event-free control but not a conclusive overall-survival gain, and PREOPANC, NORPACT-1 and CISPD-1 conflict. A021806 has completed 358-person accrual to the direct modern mFOLFIRINOX timing comparison. Finish it before funding another broad sequence trial; measure survival and complete treatment delivery, not R0 or node status alone.", "Human PDAC dormancy is not yet a treatment target. Bone-marrow tumour-cell signals predict recurrence and death, while primary and end-stage tissue show route diversity; none observes a low-cycling residual cell during the disease-free interval and links it to later recurrence. The existing £25m observatory must establish ancestry, state, time and added prediction before any anti-NET, EREG or other dormancy intervention.", "The mission cannot define eligibility by age alone. For older or otherwise vulnerable metastatic-PDAC patients, the decision unit is tumour route plus measured host reserve plus patient priorities plus completed support. The final £3m safety-line experiment tests whether action within 72 hours reduces severe toxicity and hospital time without delaying or covertly denying treatment. It is a safety-and-delivery result, not cure.", "Initially locally advanced PDAC must be reconsidered after systemic treatment, but reconsideration is not proof that surgery helps. PREOPANC-4 shows that expert centres can resect 155 of 180 selected explorations with 3% 90-day mortality, while 44% had a major complication. TAPS and matched cohorts show much longer survival after surgery but cannot separate the response-and-fitness filter from the operation. The missing causal test randomizes 540 agreed responders between resection and continued treatment and counts survival, quality-adjusted life and harm from allocation. Its £22m is not added to the fixed plan; it requires an explicit equal portfolio displacement.", "Organoid-guided treatment is now bounded by the whole route. A miniature tumour can reproduce some response differences, but published accuracy conditions on successful cultures. Real programmes reached drug testing or complete evaluation in 15.8% to 39.1% of all enrolled people and often took 6.8 to 13.7 weeks. The mission therefore funds shadow testing and data recovery inside existing model work, not clinical guidance or a new trial." ] }, { "id": "model-cholangiocarcinoma-causal-chain", "title": "Cholangiocarcinoma causal chain: current working model", "shortTitle": "CCA causal chain", "type": "system", "status": "open", "scope": [ "cholangiocarcinoma" ], "summary": "Anatomical site and cause split cholangiocarcinoma into diseases with different drivers, detectability, surgery and drug targets.", "content": "1. Injury and susceptibility: primary sclerosing cholangitis, bile-duct cysts, stones, chronic liver disease and the proven carcinogenic liver flukes Opisthorchis viverrini and Clonorchis sinensis create different risk routes; many patients have no recognised syndrome. 2. Prevention: infection control can remove one direct cause. A Thai One Health programme reported a fall from 6.0% infection to 0.3%, but no controlled cancer outcome; eCHEC randomizes behaviour and infection, not cancer. 3. Initiation and site: intrahepatic, perihilar and distal tumours arise in different duct environments and cannot share one assumed precursor. 4. Genomic routes: international sequencing links FGFR2 fusions and IDH1 mutations mainly to intrahepatic disease, while other sites more often carry different driver patterns. 5. Detection: duct obstruction can reveal advanced perihilar or distal disease, while small intrahepatic tumours can remain silent. Annual MRI/MRCP plus CA19-9 failed to create early curable detection in a five-year prospective Swedish PSC cohort; 23.8% developed severe or progressive duct change, creating roughly nine non-malignant findings per malignancy. Retrospective Australian PSC data associated surveillance with lower all-cause mortality but not better survival after cancer diagnosis. In endemic northeastern Thailand, ultrasound shifted stage among diagnosed cancers without proving population mortality benefit. 6. Local treatment: surgery offers the only established route to long disease-free survival, but an international cohort of 1,221 curative-intent intrahepatic resections recorded 59.4% recurrence and median survival of 36.8 months. BILCAP is compatible with adjuvant benefit but its primary intention-to-treat survival test was not significant; ACTICCA-1 has no posted result after its estimated completion. 7. Residual disease: tumour-informed ctDNA repeatedly identifies high recurrence risk and can lead imaging by months, yet no biliary study has randomized an action from the result. 8. Advanced treatment: chemotherapy plus durvalumab improves the survival tail across mixed biliary sites. FGFR2 and IDH1 drugs validate molecular selection in their subgroups, yet current control is measured in months and first-line FGFR2 targeting has not shown longer overall survival. The mission requires cause-specific prevention, site-specific risk finding, molecular classification before treatment, residual-disease action after surgery and resistance control after targeted response.", "sourceIds": [ "jusakul-2017-cholangiocarcinoma", "nakamura-2015-biliary-genomics", "iarc-2012-liver-flukes", "charoensuk-2024-one-health-opisthorchis", "clinicaltrials-nct05321992-echec", "villard-2023-psc-surveillance", "tan-2024-psc-mrcp-surveillance", "khuntikeo-2020-thailand-cca-screening", "hu-2020-icc-resection-recurrence", "primrose-2019-bilcap", "clinicaltrials-nct02170090-acticca1", "yu-2025-btc-ctdna-recurrence", "yoo-2025-extrahepatic-cca-ctdna", "oh-2022-topaz1", "abou-alfa-2020-claridhy", "bekaii-saab-2026-fight302" ], "links": [ { "target": "cholangiocarcinoma", "relation": "part-of" }, { "target": "claim-biliary-cancers-not-one-disease", "relation": "depends-on" }, { "target": "claim-bilcap-primary-endpoint-uncertainty", "relation": "depends-on" }, { "target": "claim-biliary-immunochemotherapy-tail", "relation": "depends-on" }, { "target": "claim-fgfr2-targeting-valid-but-not-durable", "relation": "depends-on" }, { "target": "claim-cca-surveillance-depends-on-risk-population", "relation": "depends-on" }, { "target": "claim-icc-resection-often-leaves-recurrence-risk", "relation": "depends-on" }, { "target": "disease-system", "relation": "part-of" }, { "target": "claim-fluke-control-prevents-exposure-not-yet-cancer", "relation": "depends-on" }, { "target": "claim-cca-psc-surveillance-workload", "relation": "depends-on" }, { "target": "claim-cca-ctdna-prognostic-not-action-guiding", "relation": "depends-on" }, { "target": "trial-cca-action-map", "relation": "depends-on" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "strong", "randomised": "strong", "independentReplication": "strong", "sampleSize": "Synthesis across international genomics, prospective and retrospective surveillance, 1,221 resections, and phase 2 and 3 trials", "effect": "Working causal model; linked records retain site-specific effects", "limits": [ "No surveillance route has randomized cancer-mortality evidence", "Anatomical subtypes remain under-separated in mixed trials", "Resistance mechanisms need direct paired human sampling", "Postoperative residual-disease action is not mapped" ] }, "layout": { "x": 40, "y": 12 }, "updated": "2026-09-14" }, { "id": "model-colorectal-causal-chain", "title": "Colorectal and liver-metastasis causal chain: current working model", "shortTitle": "CRC causal chain", "type": "system", "status": "open", "scope": [ "colorectal" ], "summary": "Accessible precursor removal can prevent primary cancer; after invasion, early seeding, liver geography and residual disease determine whether local cure remains possible.", "content": "1. Initiation and precursor: adenomatous and serrated routes produce mucosal lesions accessible to stool tests and endoscopy. 2. Prevention: repeated stool testing and one-time sigmoidoscopy reduced incidence and disease-specific mortality. One colonoscopy invitation in NordICC reduced incidence but not mortality at 13 years; 42.0% attendance makes delivery part of the mechanism. 3. Invasion and classification: tumour location, mismatch repair, RAS, BRAF and HER2 create different treatment routes. MSI-high or mismatch-repair-deficient metastatic disease has a durable first-line checkpoint-sensitive state; BRAF V600E requires combined BRAF and EGFR blockade; HER2-positive disease is rarer and has smaller proof-of-concept evidence. The drugs do not transfer to unselected PDAC, but the assay-to-treatment method does. 4. Anatomy: selected locally advanced rectal cancer is a pelvic local-control problem. PROSPECT showed that neoadjuvant FOLFOX with selective radiation was noninferior to routine chemoradiotherapy, so rectal and colon pathways cannot be silently merged. 5. Dissemination: human phylogenetic work indicates that some metastases seed early. Portal blood makes the liver a common first major organ, but biology, not anatomy alone, determines confinement. 6. Organ-limited disease: local treatment can change survival only after strong selection. CLOCC randomized 119 people with fewer than ten liver metastases and no extrahepatic disease; ablation with or without resection improved the long-survival tail, with eight-year survival 35.9% versus 8.9%. TransMet produced a larger effect in an even narrower transplant-eligible group. In contrast, phase III radioembolization for broader second-line liver involvement improved liver progression by 1.9 months but not survival and increased severe events. 7. Residual disease: postoperative ctDNA strongly marks recurrence risk, but the action determines utility. DYNAMIC stage II validated one treatment-omission rule. DYNAMIC-III did not validate stage III omission or escalation, and ALTAIR did not validate trifluridine/tipiracil at molecular relapse. FIND increased earlier curative-intent treatment, but survival is pending. 8. Advanced resistance: successive targeted and immune subgroups make colorectal cancer more classified than PDAC, yet heterogeneous multi-organ disease remains lethal. The useful transfer is the closed chain from measurable systemic risk to a validated local or systemic action, not an assumption that colorectal biology matches pancreatic biology.", "sourceIds": [ "mandel-2000-fobt-incidence", "atkin-2010-flexible-sigmoidoscopy", "bretthauer-2022-nordicc-10y", "kaminski-2026-nordicc-13y", "andre-2025-keynote177-five-year", "tabernero-2021-beacon-crc", "sartore-bianchi-2016-heracles", "schrag-2023-prospect-rectal", "hu-2019-crc-early-seeding", "adam-2024-transmet", "ruers-2017-clocc-crc-liver-ablation", "mulcahy-2021-epoch-radioembolization", "tie-2022-dynamic-colon", "tie-2025-dynamic3", "bando-2026-altair", "mo-2026-find-crc-surveillance", "nakamura-2024-galaxy-ctdna" ], "links": [ { "target": "colorectal", "relation": "part-of" }, { "target": "colorectal-liver-metastasis", "relation": "part-of" }, { "target": "claim-colorectal-screening-prevents-cancer-and-death", "relation": "depends-on" }, { "target": "claim-mcrc-early-seeding", "relation": "depends-on" }, { "target": "claim-transmet-selected-liver-only-crc", "relation": "depends-on" }, { "target": "claim-crc-liver-ablation-selected-survival", "relation": "depends-on" }, { "target": "failure-crc-radioembolization-pfs-without-os", "relation": "depends-on" }, { "target": "claim-colon-ctdna-has-clinical-utility-proof", "relation": "depends-on" }, { "target": "claim-crc-ctdna-utility-is-action-specific", "relation": "depends-on" }, { "target": "claim-crc-classification-produces-different-treatment-routes", "relation": "depends-on" }, { "target": "claim-rectal-anatomy-changes-local-treatment", "relation": "depends-on" }, { "target": "disease-system", "relation": "part-of" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "strong", "randomised": "strong", "independentReplication": "strong", "sampleSize": "Synthesis across population screening, randomized molecular treatment, rectal local treatment, tumour phylogeny, residual-disease and organ-directed treatment evidence", "effect": "Working causal model; linked records contain endpoint effects", "limits": [ "Rare molecular subgroup evidence does not cover most colorectal cancer", "Early-seeding estimates come from selected sampled tumours", "Transplant applies to a tiny group", "Treatment sequencing and resistance remain incomplete", "FIND survival is immature", "Population invitation effects depend on attendance" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The chain links accessible precursors, tumour state, residual disease and metastatic geography to distinct decisions." }, "animalEvidence": { "rating": "moderate", "reason": "Models support mechanisms, but the chain is anchored by human randomized outcomes." }, "inVitroEvidence": { "rating": "moderate", "reason": "Cell and organoid work helps define dependencies but cannot prove care-route benefit." }, "reproducibility": { "rating": "strong", "reason": "Several links have multiple human studies, while the liver selector and positive-MRD action remain open." }, "commercialConflicts": { "rating": "moderate", "reason": "Drug and assay company interests affect several nodes and are kept beside each result." }, "recency": { "rating": "current", "reason": "The working chain includes randomized evidence through July 2026." } } }, "layout": { "x": 86, "y": 36 }, "updated": "2026-09-14" }, { "id": "model-gallbladder-causal-chain", "title": "Gallbladder cancer causal chain: open working model", "shortTitle": "Gallbladder chain", "type": "system", "status": "open", "scope": [ "gallbladder" ], "summary": "Gallbladder cancer needs a geography- and inflammation-aware model; current mixed biliary trials do not supply a complete one.", "content": "1. Susceptibility and injury: gallstones, chronic mucosal inflammation, congenital biliary conditions, infection and geographically patterned exposures appear to create risk, but common gallstones make absolute-risk selection difficult. 2. Development: dysplasia and carcinoma can be found incidentally after cholecystectomy, while symptomatic disease can already invade liver or nodes. 3. Biology: genomic studies show site-linked differences from cholangiocarcinoma, including different frequencies of targetable receptor changes. HER2 amplification plus strong IHC 3+ expression identifies one actionable group; amplification with IHC 2+ did not behave the same. 4. Prevention and detection: Chilean aggregate data are compatible with prevention by cholecystectomy, but do not prove individual benefit. A 10,561-patient South American surgical series provides candidate risk factors, not a prospectively validated selection rule. In a US ultrasound cohort, overall cancer occurrence was almost identical with and without a recorded polyp; large initial polyps were riskier, but small polyp growth was common without cancer. 5. Late presentation in an endemic region: at a high-volume North Indian centre, 80.2% of 1,500 prospectively recorded new patients were already metastatic, only 92 received definitive surgery and many assigned palliative treatment never received one cycle. 6. Local treatment: extent and timing of surgery depend on invasion discovered before or after cholecystectomy. A Dutch registry found residual disease in 35% of re-resected incidental cancers, but selection makes the apparent survival benefit uncertain. A 163-patient randomized trial found no outcome advantage for segment 4b/5 liver resection over a wedge; more tissue removal is not a substitute for margin control and staging. Residual nodal or liver disease drives recurrence. 7. Systemic treatment: gallbladder patients enter mixed biliary chemotherapy and immunotherapy trials, but subgroup precision is lower than the overall result. The main prevention bottleneck is selecting future cancer from common gallstones and polyps; the main treatment bottleneck is reaching patients before metastatic spread, then applying site- and dependency-specific classification followed by durable control.", "contentSections": [ "The prevention chain has four observable states: common susceptibility without a visible dangerous lesion; visible benign or uncertain disease; HGD or early invasive cancer; and metastatic presentation. Current polyp rules blur the middle two states, while diagnostic AI mostly studies the last two. A useful model must predict movement between states before surgery and separately measure the after-removal pathology decision.", "Geography changes both sides of the decision. EULAT and Chile BiLS can discover high-incidence candidates, P-iGBC can define low-incidence routine-surgery pathology, and North Indian centres expose the late-presentation load. None can be pooled into a universal action threshold without local incidence, calibration and elective-operation harm." ], "sourceIds": [ "nakamura-2015-biliary-genomics", "jusakul-2017-cholangiocarcinoma", "gonzalez-2025-cholecystectomy-chile", "lorenzo-bermejo-2025-gallbladder-specimens", "szpakowski-2020-gallbladder-polyps", "van-dijk-2025-polyp-study", "yang-2025-gbcseeker", "gupta-2026-gaia-mil", "clinicaltrials-nct06192719-eulat-gbc", "clinicaltrials-nct06531408-pigbc", "wong-2025-ambrose-cholecystectomy", "vineet-2026-north-india-gallbladder-cohort", "de-savornin-lohman-2020-incidental-gallbladder", "singh-2026-gallbladder-liver-resection-rct", "pant-2026-zanidatamab-final", "primrose-2019-bilcap", "edeline-2019-prodige12", "oh-2022-topaz1" ], "links": [ { "target": "gallbladder", "relation": "part-of" }, { "target": "claim-biliary-cancers-not-one-disease", "relation": "depends-on" }, { "target": "claim-gallbladder-prevention-selection-bottleneck", "relation": "depends-on" }, { "target": "claim-gallbladder-endemic-late-presentation", "relation": "depends-on" }, { "target": "claim-gallbladder-polyp-selector", "relation": "depends-on" }, { "target": "claim-gallbladder-polyp-guidelines-still-overtreat", "relation": "depends-on" }, { "target": "claim-gallbladder-diagnostic-models-existing-lesions-only", "relation": "depends-on" }, { "target": "claim-gallbladder-prevention-must-beat-surgery-harm", "relation": "depends-on" }, { "target": "trial-gallbladder-prevention-detection-action-map", "relation": "depends-on" }, { "target": "hypothesis-gallbladder-two-stage-interception-rule", "relation": "tested-by" }, { "target": "claim-gallbladder-incidental-residual-disease", "relation": "depends-on" }, { "target": "claim-gallbladder-liver-resection-extent", "relation": "depends-on" }, { "target": "claim-zanidatamab-her2-biliary-selection", "relation": "depends-on" }, { "target": "claim-bilcap-primary-endpoint-uncertainty", "relation": "depends-on" }, { "target": "failure-adjuvant-gemox-biliary", "relation": "depends-on" }, { "target": "claim-biliary-immunochemotherapy-tail", "relation": "depends-on" }, { "target": "unknowns", "relation": "supports" }, { "target": "disease-system", "relation": "part-of" } ], "evidence": { "humanProspective": "moderate for separate states; none for a complete prevention chain", "humanRetrospective": "moderate", "randomised": "moderate", "independentReplication": "moderate", "sampleSize": "Synthesis across 10,561 South American surgical specimens, 21,706 international operations, 622,227 ultrasound patients, active cohort plans, genomic studies and mixed biliary trials", "effect": "No combined effect estimate", "limits": [ "Prevention rule lacks prospective validation", "The North Indian treatment cohort is single-centre and does not supply a population screening rule", "Existing-lesion models do not forecast future cancer", "Country-level baseline risk and elective harm are not joined", "Gallbladder-specific systemic-treatment effects often unavailable", "The surgical RCT is single-centre with 27-month median follow-up", "Re-resection benefit remains vulnerable to selection bias" ] }, "layout": { "x": 55, "y": 11 }, "updated": "2026-09-14" }, { "id": "model-hcc-causal-chain", "title": "HCC causal chain: current working model", "shortTitle": "HCC causal chain", "type": "system", "status": "open", "scope": [ "hcc" ], "summary": "Chronic liver injury creates both malignant clones and a failing organ, so prevention, cancer control and liver function cannot be separated.", "content": "1. Susceptibility and injury: hepatitis B or C, alcohol and metabolic liver disease produce long-lived inflammation, cell death and repair; inherited and environmental factors change exposure and response. 2. Initiation and field risk: cancer can arise in a cirrhotic or chronically injured liver containing many altered clones, so removing one tumour does not remove the field. 3. Prevention: hepatitis B vaccination and viral suppression prove that removing an upstream driver can lower HCC incidence, but they do not cover non-viral causes or erase established cirrhosis. Prospective alcohol-cirrhosis data suggest that abstinence may reduce HCC before decompensation, not after it, but confounding prevents a causal effect estimate. 4. Detection: a known high-risk organ population permits repeated ultrasound-based surveillance; one randomized Chinese trial reduced HCC mortality, although modern replication is limited. 5. Local treatment: resection, ablation and transplant can remove visible disease, but choice is constrained by tumour distribution, liver reserve and donor organs. In the 302-patient Japanese SURF trial, resection did not beat ablation: five-year survival was about 70% in both groups while fewer than half remained recurrence-free. Transplant can also remove the damaged field; the original Milan result reached 92% four-year recurrence-free survival only in 35 patients inside strict pathological limits. 6. Residual or new disease: recurrence is not one route. PLANet multiregion sequencing found polyclonal and monoclonal patterns; more than half of intrahepatic recurrences were polyclonal, while most distant metastases traced to one dominant clone. Those observations need a clinical classifier. In a randomized trial of 240 highly selected early recurrences, repeat surgery and ablation did not differ significantly in survival; surgery reduced local repeat recurrence but increased complications. This supports a repeatable treatment sequence rather than one universally best procedure. Adjuvant sorafenib and atezolizumab-bevacizumab failed durable low-burden benefit, showing that activity in advanced disease does not transfer automatically. 7. Advanced disease: combined vascular and immune treatment extends survival, but progression and liver failure still dominate. A 322-person treated cohort separated median survival at 21.6 months for Child-Pugh A, 9.1 for B7 and 4.7 for B8-C12; this is prognosis under selected care, not a randomized treatment interaction. 8. Coupled control: among 1,336 people with early HCC, 114 deaths were assigned to non-cancer liver disease and 220 to HCC. The mission therefore requires continued cause control, early detection, classification of recurrence origin, repeatable local control and systemic treatment that preserves liver reserve. Every trial should report tumour progression, decompensation and cause-specific death rather than hide the organ failure inside overall survival.", "contentSections": [ "9. Surveillance utility: the six-month programme mortality signal does not validate every newer test. MRI improved stage and false-referral measures without a mortality result; shorter ultrasound intervals and extra serum markers failed at the next clinical endpoint. TRACER and PREMIUM now test late-stage disease and HCC death. Until they report, test sensitivity cannot stand in for route utility.", "10. Action design: the untested link is whether active liver-reserve care during tumour treatment changes decompensation, later-line access and survival. The £6m first stage tests delivery and safety in 240 randomized patients. It does not claim mortality power and cannot expand unless tumour treatment is preserved." ], "sourceIds": [ "tcga-2017-hcc", "chang-1997-hbv-vaccine-hcc", "liaw-2004-lamivudine-hbv", "rodriguez-2021-alcohol-abstinence-hcc", "zhang-2004-hcc-screening", "bruix-2015-storm-sorafenib", "finn-2020-imbrave150", "yopp-2026-imbrave050-update", "kawaguchi-2025-surf-hcc", "mazzaferro-1996-hcc-transplant", "xia-2020-repeat-hcc-treatment", "zhang-2026-planet-hcc-recurrence", "seif-el-dahan-2026-early-hcc-causes-death", "storandt-2024-atezo-bev-liver-function", "lim-2025-attention-hbv", "rhee-2025-miracle-hcc", "hirode-2026-hcc-biomarker-rct", "trinchet-2011-hcc-surveillance-interval", "singal-2024-tracer-protocol", "clinicaltrials-nct06084234-tracer", "ioannou-2026-premium-protocol", "clinicaltrials-nct05486572-premium" ], "links": [ { "target": "hcc", "relation": "part-of" }, { "target": "claim-hcc-prevention-is-proven-for-viral-cause", "relation": "depends-on" }, { "target": "claim-hcc-surveillance-mortality-signal", "relation": "depends-on" }, { "target": "failure-sorafenib-adjuvant-hcc", "relation": "depends-on" }, { "target": "failure-imbrave050-interim-benefit-disappeared", "relation": "depends-on" }, { "target": "claim-hcc-systemic-combination-improves-not-cures", "relation": "depends-on" }, { "target": "claim-hcc-local-control-long-survival-not-eradication", "relation": "depends-on" }, { "target": "claim-hcc-recurrence-distinct-clonal-routes", "relation": "depends-on" }, { "target": "claim-hcc-transplant-selection", "relation": "depends-on" }, { "target": "claim-hcc-repeat-treatment-sequence", "relation": "depends-on" }, { "target": "claim-hcc-alcohol-interception-window", "relation": "depends-on" }, { "target": "claim-hcc-tumour-and-liver-are-competing-failure-axes", "relation": "depends-on" }, { "target": "hypothesis-hcc-liver-reserve-co-control", "relation": "motivates" }, { "target": "claim-hcc-early-hbv-treatment-signal-immature", "relation": "depends-on" }, { "target": "claim-hcc-surveillance-test-improvement-not-mortality", "relation": "depends-on" }, { "target": "trial-hcc-action-map", "relation": "depends-on" }, { "target": "disease-system", "relation": "part-of" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "strong", "randomised": "strong", "independentReplication": "strong", "sampleSize": "Synthesis across population prevention, screening, genomic and treatment evidence", "effect": "Working causal model; linked records contain intervention effects", "limits": [ "Metabolic liver-disease prevention still needs direct cancer-endpoint evidence", "Alcohol abstinence evidence is observational and stage-dependent", "Early antiviral expansion remains interim", "Modern comparative surveillance mortality results are pending", "Recurrence-origin classification is not yet ready for treatment selection", "Transplant results depend on strict selection and scarce organs", "Repeat-local-treatment evidence covers only fit patients with small recurrence", "Liver reserve predicts outcome but modifiability during HCC treatment is unproved" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Each step preserves the dual causal route through malignant burden and organ failure." }, "animalEvidence": { "rating": "moderate", "reason": "Models support individual mechanisms but cannot prove the human prevention, surveillance or care-route effects." }, "inVitroEvidence": { "rating": "moderate", "reason": "Laboratory systems explain parts of tumour and viral biology but omit liver reserve and care delivery." }, "reproducibility": { "rating": "moderate", "reason": "Several treatment and prevention effects repeat, while surveillance utility and co-control remain unsettled." }, "commercialConflicts": { "rating": "moderate", "reason": "The synthesis includes public cohorts and company-funded medicine trials." }, "recency": { "rating": "current", "reason": "The model includes completed and recruiting evidence checked through September 2026." } } }, "layout": { "x": 70, "y": 9 }, "updated": "2026-09-14" }, { "id": "model-pdac-causal-chain", "title": "PDAC causal chain: current working model", "shortTitle": "PDAC causal chain", "type": "system", "status": "open", "scope": [ "pdac" ], "summary": "A linked route from susceptibility through precursor mutation, changing cell state, immune, vascular, stromal and neural support, dissemination, treatment selection, residual disease, host wasting and death.", "content": "1. Susceptibility: a small inherited subset and chronic inflammatory states raise risk, but most cases lack a known high-risk mutation. 2. Initiation: common drivers appear in microscopic PanIN; IPMN provides a second visible precursor route. Mutation presence is common enough that progression, not initiation alone, is the diagnostic target. 3. Evolution and cell state: KRAS-driven epithelium accumulates tumour-suppressor loss and copy-number change while classical, basal-like and stress states coexist and change. Treatment-specific model work shows that relapse can arise through gene-control changes without a new resistance mutation. A fixed rare cancer-stem-cell hierarchy is not established; tumour-initiating capacity can move between cells under stromal and treatment pressure. 4. Local ecosystem and delivery: fibroblast, matrix, irregular vessels, myeloid cells, metabolism and nerves can support or restrain different parts of growth. Drug transport varies between human tumours. Opening a single matrix barrier increased response but not survival in phase 3, so delivery must be measured and paired with an effective cell-killing treatment. Human tissue and model work make Schwann cells active participants in nerve invasion, not a proven treatment target. 5. Dissemination and persistence: timing varies, metastases may contain several founding clones, and the liver can be conditioned before visible disease. Dormant residual cells and treatment-triggered reawakening are causal in PDAC models, but a human dormant-cell state has not been validated. 6. Treatment selection: surgery and combination chemotherapy remove or suppress much disease but leave occult cells and select changing states. Multi-RAS inhibition doubled median survival after prior therapy in a 2026 randomized trial, yet median progression at 7.3 months shows that initial dependence is not durable control. 7. Residual disease: ctDNA predicts recurrence strongly in retrospective cohorts, but no completed PDAC trial proves that acting on it improves survival. The missing chain is measurement to a timely, safe treatment that clears the detected state. 8. Host failure and death: most autopsied patients have widespread metastasis, while a substantial minority die from locally destructive disease. Muscle loss independently tracks short survival and can reduce treatment capacity, but IL-6 blockade preserving muscle did not improve the main survival endpoint. The highest-value intervention point remains before residual cells expand into heterogeneous visible metastases, provided treatment can eradicate them and the host remains fit enough to receive it.", "sourceIds": [ "hu-2018-germline-risk", "kanda-2012-early-panin-mutations", "chan-seng-yue-2020-evolution-subtypes", "singhal-2024-classical-kras-resistance", "lodestijn-2021-environment-clonogenicity", "koay-2014-human-pdac-drug-transport", "hingorani-2020-halo301", "fujii-nishimura-2018-schwann-met", "stupakov-2026-schwann-mechanics", "notta-2016-punctuated-evolution", "shi-2026-chemo-dormant-dtc", "conroy-2022-prodige24-five-year", "oreilly-2026-daraxonrasib", "botta-2024-tumor-informed-ctdna", "babic-2019-muscle-loss-survival", "chen-2025-tocilizumab-cachexia", "iacobuzio-donahue-2009-autopsy-patterns" ], "links": [ { "target": "pdac", "relation": "part-of" }, { "target": "claim-germline-risk-subset", "relation": "depends-on" }, { "target": "claim-early-panin-driver-mutations-not-progression-markers", "relation": "depends-on" }, { "target": "claim-pdac-subtypes-are-mixed-evolving-states", "relation": "depends-on" }, { "target": "claim-pdac-resistance-uses-reversible-cell-states", "relation": "depends-on" }, { "target": "claim-pdac-stemness-is-dynamic", "relation": "depends-on" }, { "target": "claim-immune-exclusion-is-multicellular", "relation": "depends-on" }, { "target": "claim-pdac-delivery-is-variable-not-single-barrier", "relation": "depends-on" }, { "target": "claim-fibrosis-nerve-mechanosensing", "relation": "depends-on" }, { "target": "claim-pdac-progression-timing-uncertain", "relation": "depends-on" }, { "target": "claim-pdac-dormancy-model-proven-human-uncertain", "relation": "depends-on" }, { "target": "claim-daraxonrasib-common-ras-survival-breakthrough", "relation": "depends-on" }, { "target": "unknown-ctdna-actionability", "relation": "depends-on" }, { "target": "claim-cachexia-is-parallel-lethal-system", "relation": "depends-on" }, { "target": "claim-pdac-death-can-be-local-or-metastatic", "relation": "depends-on" }, { "target": "ranking-provisional-pdac-bottlenecks", "relation": "supports" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "strong", "randomised": "moderate", "independentReplication": "moderate", "sampleSize": "Synthesis across tissue, genomic, autopsy, cohort and randomised treatment evidence", "effect": "Working causal model; no single combined effect estimate", "limits": [ "Several causal transitions still rely on animal models", "Dissemination timing remains contested", "Human dormancy is not directly mapped", "Microbiome evidence does not yet define an intervention", "Neural and epigenetic treatment routes remain model-derived", "Each linked claim retains its own limits" ] }, "layout": { "x": 28, "y": 45 }, "updated": "2026-09-14" }, { "id": "model-pnet-causal-chain", "title": "Pancreatic neuroendocrine tumour causal chain: current working model", "shortTitle": "pNET causal chain", "type": "system", "status": "open", "scope": [ "pnet" ], "summary": "Cell lineage, differentiation, grade, hormone production, inherited syndromes and receptor expression make pNET a separate system from PDAC.", "content": "1. Susceptibility: a minority arises in inherited syndromes such as MEN1 or VHL, while most are sporadic. 2. Initiation and evolution: human genomic work identifies recurrent MEN1, DAXX or ATRX and mTOR-pathway alterations, with different patterns from KRAS-dominated PDAC. 3. Clinical state: hormone-producing tumours can be found through functional syndromes; non-functioning tumours may grow silently. Grade, differentiation and growth rate change the disease more than anatomical label alone. 4. Classification boundary: well-differentiated G3 pNET and poorly differentiated pancreatic neuroendocrine carcinoma cannot be merged. Expert morphology alone agreed in only 38% of 70 high-grade samples; RB, DAXX, ATRX and mutation evidence improved classification. In 49 retrospective PanNEC-G3 records, RB loss and/or KRAS mutation marked higher platinum response, but the survival difference was not significant. 5. Observability: somatostatin-receptor expression enables functional imaging and receptor-directed treatment in selected differentiated tumours. 6. Local control: two resection cohorts of 487 and 1,020 patients found recurrence in 12.3% and 15.1%. Grade, size, nodes and vascular invasion separated risk; the liver was the leading recurrence site and very late recurrence occurred. These selected cohorts support durable control and tailored follow-up, not a universal surgical cure rate. 7. Evolution under treatment: a prospective 21-person progression cohort found higher Ki-67 in 13 of 16 adequate repeat biopsies and grade 2 to 3 change in eight; combined biopsy and dual PET changed treatment in seven. The result shows possible state drift, not benefit from reclassification. 8. Chronic control: in pooled low-grade enteropancreatic NETs, lanreotide halved progression or death risk but did not improve survival; randomized everolimus, sunitinib and cabozantinib trials delayed progression for months. SEQTOR found no significant PFS or OS winner between everolimus-first and streptozotocin-plus-5-FU-first sequences, while chemotherapy produced more first-line response. 9. Directed radiation: OCLURANDOM gives pNET-specific randomized evidence that lutetium-177 DOTATATE can produce substantially longer control than sunitinib, while late marrow injury and one treatment-related leukaemia prevent a simple cure claim. Receptor loss, liver burden and cumulative toxicity narrow later options. The mission for well-differentiated pNET is closer to chronic control than for PDAC, but it requires time-updated grade, receptor and molecular classification inside a treatment-strategy trial.", "contentSections": [ "10. Small localized state: short prospective follow-up supports observation for many asymptomatic non-functioning tumours no larger than 2 cm, but synchronous spread and adverse pathology occur in a minority. Duct dilation, measured growth and grade are warning features; operation type and pancreatic function are competing harms. Diameter is an input, not the decision.", "11. Inherited early-detection state: MEN1 can produce multifocal tumours in adolescence. Imaging detects them, but a prospective circulating RNA assay did not predict development, growth or metastasis. Detection therefore opens a longitudinal decision; it does not identify the lethal lesion." ], "sourceIds": [ "scarpa-2017-pnet", "sigel-2018-panenen-differentiation", "tanaka-2020-panec-g3", "marchegiani-2019-pnet-recurrence", "dong-2020-pnet-recurrence", "raymond-2011-sunitinib-pnet", "yao-2011-everolimus-pnet", "chan-2025-cabinet", "caplin-2014-clarinet-lanreotide", "baudin-2026-oclurandom", "mollazadegan-2025-pnet-recharacterization", "capdevila-2025-seqtor-pnet", "partelli-2022-aspen-interim", "maekawa-2026-small-pnet-surgery-morbidity", "klein-haneveld-2021-men1-young-screening", "van-vliembergen-2026-men1-mnet-assay" ], "links": [ { "target": "pnet", "relation": "part-of" }, { "target": "claim-pnet-distinct-from-pdac", "relation": "depends-on" }, { "target": "claim-pnet-targeted-control-differs-from-pdac", "relation": "depends-on" }, { "target": "claim-cabozantinib-pnet-pfs-not-cure", "relation": "depends-on" }, { "target": "claim-lanreotide-pnet-control-not-survival", "relation": "depends-on" }, { "target": "claim-prrt-pnet-randomized-control", "relation": "depends-on" }, { "target": "claim-pnet-surgical-control-is-risk-stratified", "relation": "depends-on" }, { "target": "claim-panec-needs-separate-classification", "relation": "depends-on" }, { "target": "claim-pnet-state-changes-at-progression", "relation": "depends-on" }, { "target": "claim-pnet-seqtor-no-sequence-winner", "relation": "depends-on" }, { "target": "unknown-pnet-reclassification-actionability", "relation": "depends-on" }, { "target": "unknown-small-pnet-surgery-selector", "relation": "depends-on" }, { "target": "hypothesis-pnet-time-updated-surgery-rule", "relation": "tested-by" }, { "target": "disease-system", "relation": "part-of" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "strong", "randomised": "strong", "independentReplication": "strong", "sampleSize": "Synthesis across whole-genome sequencing, randomized treatment programmes, two resection cohorts and two high-grade classification cohorts", "effect": "Working causal model; linked records retain treatment effects", "limits": [ "Resection evidence is retrospective and selected", "PanNEC treatment classification lacks prospective validation", "Treatment sequence evidence is incomplete", "Overall-survival effects remain less certain than progression effects" ] }, "layout": { "x": 10, "y": 11 }, "updated": "2026-09-14" }, { "id": "patent-landscape", "title": "Patent landscape: ownership can block combinations, but a patent is not scientific proof", "shortTitle": "Patents", "type": "system", "status": "supported", "scope": [ "pdac", "colorectal", "pan-cancer", "transfer-unproven" ], "summary": "Fourteen programme-critical families plus their operating dependencies show where compound, assay, data, supply and manufacturing rights may constrain the mission; none proves benefit.", "content": "Daraxonrasib: WO2026015825A1 claims use of a RAS(ON) multi-selective inhibitor including daraxonrasib in pancreatic cancer. This makes compound and data access a real constraint for independent residual-disease combinations. HRS-4642: two Hengrui-assigned families form the probable compound-and-liposome control stack. WO2022268051A1 claims fused tetracyclic KRAS-G12D inhibitors. WO2024022507A1 claims a formula-I compound from that family in lipid and liposome compositions, loading methods, freeze-dried and reconstituted products, and cancer uses including pancreas, liver, gallbladder, bile duct and colorectal cancer. Hengrui's prospectus describes HRS-4642 as its proprietary liposomal injectable KRAS-G12D inhibitor. No checked primary public document prints the HRS-4642 code beside either patent publication, so exact structure-to-claim mapping and live national claim scope still require counsel. ELI-002: US20250009888A1 covers lipid-linked mutant-KRAS sequences and related immune use. The cited application is flagged as abandoned while later related applications exist, so legal status and claim continuity require a live family search by counsel. FAPI: US20240042066A1 covers FAP inhibitor compounds and includes lutetium-177 treatment in pancreatic xenograft mice; it does not prove human dosimetry or benefit. A patent record answers who may control an implementation, not whether the mechanism works. The programme needs freedom-to-operate work only after a human target, assay and experiment pass their scientific gate. Licence terms should include research use, combination rights, trial data access, negative-result publication, manufacturing transfer and step-in rights. The map is deliberately not a legal opinion and is incomplete across national families.", "contentSections": [ "Core RAS access: US11690915B2 is a granted Revolution Medicines member of the broad RAS(ON) tri-complex chemistry family. The later WO2026015825A1 pancreatic-use application is only one layer. A mission combination therefore needs compound supply, combination rights, biomarker and resistance data, not merely permission to cite the clinical result.", "Setidegrasib: WO2022173032A1 covers Astellas bifunctional KRAS-G12D degraders and pancreatic uses. The PCT record is marked ceased, but national family members include grants, so reading the PCT status as abandonment of the invention would be wrong. HRS-4642: WO2022268051A1 and WO2024022507A1 are the probable Hengrui compound and liposome families. Both list national applications beyond their PCT entries. The product-code match is triangulated from the later family's reference to the earlier formula-I compound and Hengrui's description of HRS-4642 as a proprietary liposomal injectable, but remains an inference until exact structure and claims are matched by counsel.", "Residual disease: WO2019200228A1 is a Natera patient-specific ctDNA monitoring family. It shows why an MRD-action trial needs assay and data rights even when the biological question is public. The PCT entry is marked ceased while national members include grants, and its examples are not pancreatic utility proof.", "Early detection: WO2019195268A2 is a GRAIL targeted-methylation panel family that includes pancreatic and hepatobiliary classification. US12258636B2 is a narrower active grant on ZFP30 and related pancreatic methylation markers. The latter's small case-control examples show the distinction between a patentable assay and a clinically useful screen. Neither family removes the base-rate, lead-time or mortality gates.", "Practical access order: first reproduce the biological or analytical result; second identify the exact product, family members and live jurisdictions; third map composition, method, use, software, data and manufacturing claims; fourth negotiate only the rights needed for the falsifying experiment; fifth expand rights after the gate passes. Patent counts are not a measure of scientific progress.", "Natera shows why a family list is not an access plan. Its filing describes more than 650 patents across the company, Signatera processing in company laboratories, no general backup for tests run at only one site, one validated sequencing platform, sole-sourced inputs and pending companion-diagnostic submissions. An MRD contract must therefore cover assay specification, algorithm version, raw and derived data, turnaround, supply continuity, failure handling and negative-result publication as well as patent rights.", "Ponsegromab adds two cachexia families. WO2020039321A2 is a Pfizer anti-GDF15 antibody family with 2018 priority; its PCT entry is marked ceased while listed national members include grants. WO2025245200A1 is a pending use-and-dose PCT filing with 2024 priority covering ponsegromab with systemic chemotherapy in cancer cachexia including pancreatic cancer. A follow-on experiment also needs drug, GDF-15 assay, chemotherapy-delivery, function, missing-data and publication access.", "Kinisoquin adds two public families and an unresolved operating stack. US10391096B2 covers an isoquercetin, vitamin B3, vitamin C and folic-acid composition for venous thrombosis, including cancer-associated thrombosis, and names Quercis in its ownership history. WO2020112828A1 and US11872241B2 cover isoquercetin or quercetin routes for cancer thrombotic events and begin with Beth Israel Deaconess. The checked records do not prove the licence chain, exact phase-3 formulation or manufacturing controls. Independent evaluation needs those items plus live claims, protocol and dose rules, patient-level clot and bleed data and comparison with generic LMWH or factor-Xa prophylaxis.", "Adenocyte adds US20240131366A1 for controlled cavitation and tissue sampling. The publication is enough to identify a device route and assignee, but not enough to establish the full international family, live claim scope, validity, product match or operating freedom." ], "sourceIds": [ "patent-wo2026015825-daraxonrasib-pdac", "patent-us11690915-rason", "patent-wo2022173032-setidegrasib", "patent-wo2022268051-hengrui-kras-g12d", "patent-wo2024022507-hengrui-liposome", "hengrui-2025-hk-prospectus", "patent-us20250009888-elicio-kras-vaccine", "patent-us20240042066-fapi", "patent-wo2019200228-natera-mrd", "patent-wo2019195268-grail-methylation", "patent-us12258636-zfp30-pdac", "patent-wo2020039321-anti-gdf15", "patent-wo2025245200-ponsegromab-cachexia", "oreilly-2026-daraxonrasib", "elicio-2026-amplify7p-results", "natera-2025-10k", "patent-us10391096-kinisoquin", "patent-wo2020112828-isoquercetin-cancer", "clinicaltrials-nct06861088-catiq-p3", "quercis-2026-kinisoquin-profile", "zwicker-2019-isoquercetin-catiq", "patent-us20240131366-linfu" ], "links": [ { "target": "company-revolution-medicines", "relation": "protects" }, { "target": "company-astellas", "relation": "protects" }, { "target": "claim-setidegrasib-g12d-degradation", "relation": "constrains" }, { "target": "company-hengrui", "relation": "protects" }, { "target": "claim-hrs4642-g12d-chemo-signal", "relation": "constrains" }, { "target": "company-elicio-therapeutics", "relation": "protects" }, { "target": "company-grail", "relation": "protects" }, { "target": "unknown-ctdna-actionability", "relation": "constrains" }, { "target": "detection-integrated-workflows", "relation": "constrains" }, { "target": "hypothesis-fapi-theranostic-pdac", "relation": "may-constrain" }, { "target": "recommended-strategy", "relation": "requires-licence-work-for" }, { "target": "investment", "relation": "part-of" }, { "target": "claim-mrd-platform-access-is-more-than-patent-license", "relation": "contains" }, { "target": "claim-ponsegromab-access-stack", "relation": "contains" }, { "target": "company-quercis-pharma", "relation": "protects" }, { "target": "claim-kinisoquin-access-stack", "relation": "contains" }, { "target": "company-adenocyte", "relation": "protects" } ], "evidence": { "humanProspective": "not-applicable", "humanRetrospective": "not-applicable", "randomised": "not-applicable", "independentReplication": "not-applicable", "sampleSize": "Fourteen priority patent families plus audited operating stacks spanning RAS inhibition, KRAS degradation, liposomal delivery, vaccination, radioligands, residual disease, early detection cachexia and thrombosis", "effect": "Ownership and licence map; no efficacy effect", "limits": [ "Priority families only; not an exhaustive family search", "HRS-4642 code-to-claim mapping remains an inference", "Exact national claims and legal status require counsel", "Legal and supply status may change", "Google does not provide a legal opinion", "Claims differ by jurisdiction", "No freedom-to-operate opinion", "No contract terms or laboratory contingency test" ], "dimensions": { "biologicalPlausibility": { "rating": "not-applicable", "reason": "The record maps rights and access; patents do not establish that a biological mechanism works." }, "animalEvidence": { "rating": "not-applicable", "reason": "Animal examples inside a patent may support a filing but do not establish ownership scope or human benefit." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Laboratory examples do not determine freedom to operate or prove clinical efficacy." }, "reproducibility": { "rating": "moderate", "reason": "Primary patent records and the Hengrui prospectus independently support the probable ownership stack, but the exact product-code and claim match is not public." }, "commercialConflicts": { "rating": "high", "reason": "The assignees and product sponsor control compound, formulation, trial and access information relevant to negotiation." }, "recency": { "rating": "current", "reason": "The priority families, national applications and 2025 company filing are current enough for programme triage, but legal status must be refreshed before action." } } }, "layout": { "x": 92, "y": 101 }, "updated": "2026-09-14" }, { "id": "pdac", "title": "Pancreatic ductal adenocarcinoma", "shortTitle": "PDAC", "type": "cancer", "status": "supported", "scope": [ "pdac" ], "summary": "PDAC is usually a systemic RAS-driven disease by the time it is diagnosed; local removal can cure a minority, while recurrence, resistance and host decline remain the central routes to death.", "content": "PDAC is kept separate from other pancreatic cancers. Human genomic data place KRAS activation early, followed by loss of CDKN2A, TP53 and SMAD4 routes in changing combinations. Human phylogenetic work shows that metastatic capacity develops over years in some tumours, but model and recurrence evidence also shows that dissemination can precede clinical diagnosis. Dense fibroblast, immune, nerve and vascular states alter drug exposure and cell state without forming one uniformly removable barrier.\n\nCurrent curative intent requires complete surgery plus systemic treatment. Even in selected resected patients receiving modified FOLFIRINOX, five-year disease-free survival was 26.1%, so apparent local disease often contains occult systemic disease. In advanced PDAC, chemotherapy extends survival by months. Daraxonrasib has now produced a large randomized survival gain after prior treatment in RAS-G12 disease, but median progression at 7.3 months shows that common-RAS inhibition is control rather than eradication.\n\nThe mission chain is risk enrichment → actionable early detection → complete local and systemic treatment → residual-disease measurement → resistance-directed intervention → maintained nutrition and organ function. The largest current break is after initial systemic response: the disease is measurable enough to show dependence, but not classified and intercepted quickly enough to prevent resistant or dormant populations from returning. Claims from colorectal, lung, melanoma or blood cancers enter only through explicit transfer links and require a PDAC falsification test.", "sourceIds": [ "tcga-2017-pdac", "yachida-2010-metastatic-timing", "rhim-2012-early-dissemination", "conroy-2022-prodige24-five-year", "wainberg-2023-napoli3", "oreilly-2026-daraxonrasib", "singhal-2024-classical-kras-resistance", "koay-2014-human-pdac-drug-transport", "uemura-2026-elemental-diet-pdac", "wode-2026-mistral" ], "links": [ { "target": "disease-system", "relation": "part-of" }, { "target": "model-pdac-causal-chain", "relation": "contains" }, { "target": "detection", "relation": "part-of" }, { "target": "treatment", "relation": "part-of" }, { "target": "residual-disease", "relation": "part-of" }, { "target": "claim-elemental-diet-survival-signal-needs-replication", "relation": "contains" }, { "target": "failure-mistletoe-pdac-no-quality-benefit", "relation": "contains" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "strong", "randomised": "strong", "independentReplication": "strong for poor systemic control; mixed for specific mechanisms", "sampleSize": "Integrated human genomic, resection, metastatic treatment, resistance and tissue-delivery studies", "effect": "Selected resected five-year disease-free survival 26.1%; later-line daraxonrasib median PFS 7.3 months despite a large survival gain", "limits": [ "PDAC states vary across patients and treatment", "Early-dissemination evidence includes models", "No current route reliably eradicates systemic disease", "Recent RAS results need longer resistance and low-burden follow-up" ] }, "layout": { "x": 20, "y": 23 }, "updated": "2026-09-14" }, { "id": "people", "title": "People and laboratories", "shortTitle": "People", "type": "person", "status": "supported", "scope": [ "pdac", "pnet", "cholangiocarcinoma", "gallbladder", "hcc", "colorectal", "pan-cancer" ], "summary": "Researchers and clinicians are ranked by demonstrated capability and complementary mission role, not fame or publication count.", "content": "Rank people by the capability the mission lacks, not by paper count. Human trials and delivery: Eileen O'Reilly is the central clinical operator; Wungki Park adds early drug development and Liwei Wang adds Chinese common-RAS execution. Human disease evolution: Christine Iacobuzio-Donahue connects autopsy, metastatic route and resistant clones. RAS biology: Andrew Aguirre joins functional genetics to clinical specimens. Detection and risk: Suresh Chari covers new-onset diabetes, Michael Goggins inherited-risk surveillance, Anirban Maitra biomarker validation, Keiji Hanada Japanese community delivery, George Hanna prospective NHS breath validation and Núria Malats European risk epidemiology. Models: David Tuveson supplies organoids and engineered systems, with human evidence kept above model results. Low-burden immunity: Vinod Balachandran joins individual vaccine manufacture to deep immune tracking; Robert Vonderheide tests immune exclusion. Regina Barzilay covers locked external AI tests. Bert Vogelstein supplies colorectal prevention and liquid-biopsy transfer checks. Christina Yap is the proposed trial-statistics chair, conditional on a conflict screen. Ali Stunt is the proposed paid patient co-chair, backed by a wider panel. Eduard Jonas adds South African pancreatic and HPB clinical delivery. Sanju Sobnach adds a connected HCC evidence chain: a sub-Saharan review, a Cape Town public/private cohort and a South Africa/Sweden comparison. Renata D'Alpino Peixoto adds a Latin American pancreatic route through LACOG GASPAR. The ten-seat working table still has ten voting seats and now keeps twelve further contributors linked. Jeffrey Zwicker adds direct CATIQ phase-2, patent and current phase-3 access for the PDI clot-prevention route; those connected roles make independent statistics and event review mandatory. No regional researcher closes representation alone: UCT supplies deep South African evidence rather than continent-wide delivery, while GASPAR has 79 pancreatic cases, mostly private coverage and a registry-count conflict. Local patient, primary-care, pathology and public-health leaders still need appointment at each site. Availability, conflicts and willingness have not been checked.", "sourceIds": [ "rompen-2026-preopanc2-staging", "theijse-2024-nontherapeutic-laparotomy", "stoop-2026-global-staging-survey", "msk-oreilly-profile-2026", "msk-iacobuzio-donahue-profile-2026", "danafarber-aguirre-profile-2026", "mdanderson-chari-profile-2026", "ja-onomichi-hanada-profile-2026", "imperial-hanna-profile-2026", "cnio-malats-profile-2026", "cshl-tuveson-profile-2026", "msk-balachandran-profile-2026", "mit-barzilay-profile-2026", "icr-christina-yap-profile-2026", "pancreatic-cancer-action-ali-stunt-profile-2026", "shsmu-2026-liwei-wang-profile", "uct-2026-cancer-projects", "jonas-2025-sub-saharan-hcc-pathways", "clinicaltrials-nct05924789-gaspar", "peixoto-2025-gaspar-results", "sobnach-2024-ssa-hcc-systematic-review", "sobnach-2025-south-africa-hcc-disparities", "sobnach-2026-global-hcc-disparities", "zwicker-2019-isoquercetin-catiq", "clinicaltrials-nct06861088-catiq-p3", "patent-wo2020112828-isoquercetin-cancer", "daamen-2024-radar-panc-protocol", "clinicaltrials-nct04875325-radar-panc", "clinicaltrials-nct04340141-a021806", "bai-2025-cispd1", "dotan-2024-giant", "dotan-2025-giant-ga", "clinicaltrials-nct04233866-giant", "betge-2022-grantpax", "clinicaltrials-nct02143219-pamela70", "mohile-2021-gap70", "li-2021-gain", "ihorst-2026-geriatric-comanagement", "clinicaltrials-nct06040801-frail-gi", "stoop-2026-preopanc4-surgery", "dekker-2026-taps-restaging", "verkolf-2026-taps-resection-prediction" ], "links": [ { "target": "person-jennifer-knox", "relation": "contains" }, { "target": "person-marc-besselink", "relation": "contains" }, { "target": "programme", "relation": "supports" }, { "target": "funding", "relation": "depends-on" }, { "target": "companies", "relation": "supports" }, { "target": "labs", "relation": "depends-on" }, { "target": "programme-mission-table-people", "relation": "contains" }, { "target": "person-eileen-oreilly", "relation": "contains" }, { "target": "person-christine-iacobuzio-donahue", "relation": "contains" }, { "target": "person-andrew-aguirre", "relation": "contains" }, { "target": "person-suresh-chari", "relation": "contains" }, { "target": "person-david-tuveson", "relation": "contains" }, { "target": "person-vinod-balachandran", "relation": "contains" }, { "target": "person-regina-barzilay", "relation": "contains" }, { "target": "person-michael-goggins", "relation": "contains" }, { "target": "person-anirban-maitra", "relation": "contains" }, { "target": "person-wungki-park", "relation": "contains" }, { "target": "person-robert-vonderheide", "relation": "contains" }, { "target": "person-bert-vogelstein", "relation": "contains" }, { "target": "person-liwei-wang", "relation": "contains" }, { "target": "person-keiji-hanada", "relation": "contains" }, { "target": "person-george-hanna", "relation": "contains" }, { "target": "person-nuria-malats", "relation": "contains" }, { "target": "person-christina-yap", "relation": "contains" }, { "target": "person-ali-stunt", "relation": "contains" }, { "target": "person-eduard-jonas", "relation": "contains" }, { "target": "person-renata-dalpino-peixoto", "relation": "contains" }, { "target": "person-sanju-sobnach", "relation": "contains" }, { "target": "person-jeffrey-zwicker", "relation": "contains" }, { "target": "person-lois-daamen", "relation": "contains" }, { "target": "person-cristina-ferrone", "relation": "contains" }, { "target": "person-tingbo-liang", "relation": "contains" }, { "target": "person-efrat-dotan", "relation": "contains" } ], "evidence": { "humanProspective": "not-applicable", "humanRetrospective": "not-applicable", "randomised": "not-applicable", "independentReplication": "not-applicable", "sampleSize": "25 named people across trials, evolution, RAS, detection, models, immunity, AI, statistics, patient oversight, transfer and regional delivery", "effect": "Capability ranking, not a clinical-effect claim", "limits": [ "Selection is judgement", "Profiles and trial roles do not prove future execution", "Availability and conflicts require current review", "Statistics and patient seats are proposed, not appointed", "The African outcome route is anchored in one South African centre and an unproved consortium roadmap", "The Latin American route has selected referral cohorts rather than population coverage" ] }, "layout": { "x": 62, "y": 104 }, "updated": "2026-09-15", "contentSections": [ "The hidden-spread pass adds Marc Besselink because his current work joins PREOPANC-2 trial infrastructure, the five-centre TAPS surgery cohort and a 76-country staging survey. The proposed role is protocol and definition advice for the English route. Availability, conflicts, patient-level access and English authority remain unconfirmed.", "Lois Daamen is linked as RADAR-PANC's coordinating investigator because the mission needs direct access to the randomized surveillance protocol, actual accrual, crossover, outcome definitions and eventual data. That role is evidence of study leadership, not proof of availability, data access or independence; all require current confirmation.", "Cristina Ferrone is linked as chair of A021806, the current 358-person multicentre timing test. Tingbo Liang is linked as principal investigator of the positive 324-person CISPD-1 trial. The first role gives a route to complete the North American survival answer; the second gives a route to test whether one-centre Chinese delivery transfers. Availability, conflicts and patient-level access remain unconfirmed for both.", "The older-and-vulnerable treatment route adds Efrat Dotan because she led GIANT, the first national randomized metastatic-PDAC treatment study built for this population. The proposed role is complete data recovery and route design. Availability, data rights and independence remain unconfirmed.", "The conversion-surgery pass expands Marc Besselink's evidence-linked role. PREOPANC-4, the TAPS response models and the earlier staging work connect nationwide implementation, expert-centre selection, operative harm and the explicit limits of non-randomized survival comparison. The proposed role is protocol and data advice, not authority over allocation or analysis; availability, conflicts and data rights remain unconfirmed.", "Organoid-routing capability now includes Jennifer Knox. PASS-01 links randomized first-line treatment to genome, RNA and organoid measurements; PROSPER-PANC and ADOPT extend that route in Ontario. This supports a role in data recovery and policy testing, not an assumption of availability or assay benefit." ] }, { "id": "person-ali-stunt", "title": "Ali Stunt — patient oversight and earlier-diagnosis delivery", "shortTitle": "Ali Stunt", "type": "person", "status": "open", "scope": [ "pdac", "pancreatic-all" ], "summary": "A long-term survivor and charity founder can force the mission to count delay, burden, false alarms, access and quality of life as patient outcomes.", "content": "Stunt founded Pancreatic Cancer Action in 2010 after her own diagnosis and stepped down as chief executive in 2023 to focus on advocacy. Proposed mission role: paid patient and public co-chair with equal voting rights on detection thresholds, consent, false-positive work-up, trial burden, access and quality-of-life measures. Her lived experience and work on early diagnosis are directly relevant to a mission built around health-system pathways. They are not scientific evidence and do not establish that she is available or representative of every patient. Appointment would require a transparent selection process, declared organisational interests, compensation and a wider patient panel spanning unresectable disease, hereditary risk, different ethnic groups and rural access.", "sourceIds": [ "pancreatic-cancer-action-ali-stunt-profile-2026" ], "links": [ { "target": "people", "relation": "part-of" }, { "target": "programme-mission-table-people", "relation": "patient-seat-on" }, { "target": "detection", "relation": "oversees-burden-of" }, { "target": "risks", "relation": "reviews" } ], "evidence": { "humanProspective": "not-applicable", "humanRetrospective": "not-applicable", "randomised": "not-applicable", "independentReplication": "not-applicable", "sampleSize": "One proposed named co-chair plus a required wider patient panel", "effect": "Governance capability, not a clinical-effect claim", "limits": [ "Official charity profile", "Availability not checked", "One survivor cannot represent the full disease population", "Scientific decisions still require independent evidence review" ] }, "layout": { "x": 73, "y": 116 }, "updated": "2026-09-14" }, { "id": "person-andrew-aguirre", "title": "Andrew Aguirre — RAS biology, resistance and pancreatic translation", "shortTitle": "Andrew Aguirre", "type": "person", "status": "supported", "scope": [ "pdac", "colorectal", "hepatobiliary", "pan-cancer" ], "summary": "A laboratory-to-clinic lead who combines pancreatic models, functional genetics and a cross-cancer RAS therapeutics centre.", "content": "Current roles: medical oncologist at Dana-Farber, associate professor at Harvard Medical School, Broad Institute member, associate director of the Hale Family Center for Pancreatic Cancer Research and co-director of Dana-Farber's Center for RAS Therapeutics. Mission role: build the resistance map around common-RAS and G12D drugs, test combinations in patient-linked models, and reject combinations that do not suppress the escape route they claim to target. Unique capability: one programme can compare RAS alleles and resistance across pancreatic, colorectal, lung and biliary cancers while retaining a PDAC-specific clinic. Main weakness: model sensitivity and target engagement do not establish survival benefit; a centre built partly around company compounds also needs independent compound selection, full disclosures and publication of negative screens.", "sourceIds": [ "danafarber-aguirre-profile-2026", "oreilly-2026-daraxonrasib", "park-2026-setidegrasib" ], "links": [ { "target": "people", "relation": "part-of" }, { "target": "claim-daraxonrasib-common-ras-survival-breakthrough", "relation": "works-on" }, { "target": "claim-setidegrasib-g12d-degradation", "relation": "works-on" }, { "target": "hypothesis-prospective-ras-resistance-routing", "relation": "tests" }, { "target": "programme-5m-decision-core", "relation": "supports" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "moderate", "randomised": "not-applicable", "independentReplication": "moderate", "sampleSize": null, "effect": "Capability record, not an efficacy claim", "limits": [ "Institutional profile", "Model-to-human translation gap", "Company compounds and collaborations require disclosure", "Negative combination screens must be published" ] }, "layout": { "x": 89, "y": 102 }, "updated": "2026-09-14" }, { "id": "person-anirban-maitra", "title": "Anirban Maitra — pancreatic interception, pathology and biomarker validation", "shortTitle": "Anirban Maitra", "type": "person", "status": "supported", "scope": [ "pdac", "pancreatic-all" ], "summary": "A pathology and programme lead joining precursor biology, liquid biopsies, cysts, new-onset diabetes and blinded multi-centre validation.", "content": "Current role, corrected from older MD Anderson pages: director of NYU Langone's Perlmutter Cancer Center, associate dean for cancer research, Perlmutter Professor of Pathology and professor of medicine. His active NCI early-detection centre moved with him to NYU and runs blinded multi-institution biomarker validation across inherited risk, pancreatic cysts and new-onset diabetes. Mission role: build the bridge from precursor tissue and candidate markers to locked assays in intended populations, with common specimens and an explicit clinical action. Unique capability: pathology, biomarker consortia and cancer-centre authority sit in one programme. Main weakness: leadership breadth can dilute direct experimental ownership; candidate panels and liquid biopsies have repeatedly looked stronger in case-control samples than in screening. The mission should fund a common validation core with independent statistics, assay blinding and public negative results rather than another local marker panel.", "sourceIds": [ "nyulangone-maitra-profile-2026", "nci-maitra-edrn-2025", "wu-2011-gnas-ipmn" ], "links": [ { "target": "people", "relation": "part-of" }, { "target": "claim-ipmn-mutations-identify-lineage-not-progression", "relation": "works-on" }, { "target": "claim-new-onset-diabetes-enriches-risk", "relation": "works-on" }, { "target": "claim-case-control-blood-tests-overstate-screening-readiness", "relation": "must-overcome" }, { "target": "programme-25m-prospective-observatory", "relation": "supports" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "strong", "randomised": "none", "independentReplication": "moderate", "sampleSize": null, "effect": "Capability record, not an efficacy claim", "limits": [ "Institutional and grant records", "Detection utility remains unproved", "Leadership role is broader than PDAC", "Biomarker patents and consulting interests require current disclosure", "Independent blinded validation required" ] }, "layout": { "x": 95, "y": 103 }, "updated": "2026-09-14" }, { "id": "person-bert-vogelstein", "title": "Bert Vogelstein — colorectal evolution, prevention and blood detection transfer", "shortTitle": "Bert Vogelstein", "type": "person", "status": "supported", "scope": [ "colorectal", "pan-cancer", "transfer-unproven", "pdac" ], "summary": "A transfer lead for the genetic model of colorectal progression and blood-based cancer detection, with strict disease-property checks before PDAC use.", "content": "Current role: Clayton Professor of Oncology and co-director of the Ludwig Center at Johns Hopkins. His group established the stepwise genetic model of colorectal tumour evolution and helped develop mutation-and-protein blood tests such as CancerSEEK. Mission role: separate what made colorectal prevention work—an accessible precursor, long dwell time, removable lesions and a scalable test—from what can and cannot transfer to PDAC. He can also challenge liquid-biopsy teams to move from case-control separation to prospective utility. Main weakness: PDAC often has a hidden microscopic precursor, low shed at small burden and early systemic risk. A pan-cancer assay or colon-derived progression model cannot be assumed to solve those properties, and commercial intellectual-property interests around blood detection must be disclosed.", "sourceIds": [ "hopkins-vogelstein-profile-2026", "cohen-2018-cancerseek", "mandel-2000-fobt-incidence" ], "links": [ { "target": "people", "relation": "part-of" }, { "target": "model-colorectal-causal-chain", "relation": "works-on" }, { "target": "claim-colorectal-screening-prevents-cancer-and-death", "relation": "supports" }, { "target": "claim-case-control-blood-tests-overstate-screening-readiness", "relation": "must-overcome" }, { "target": "detection", "relation": "may-transfer-to" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "strong", "randomised": "strong", "independentReplication": "strong", "sampleSize": "Capability spans large colorectal prevention trials; CancerSEEK case-control n=1,817", "effect": "Proven colorectal prevention capability; PDAC transfer unproved", "limits": [ "Institutional profile", "Colorectal and PDAC precursor biology differ", "CancerSEEK was case-control", "Blood-test intellectual-property and company interests require disclosure" ] }, "layout": { "x": 92, "y": 101 }, "updated": "2026-09-14" }, { "id": "person-christina-yap", "title": "Christina Yap — independent trial rules and stopping decisions", "shortTitle": "Christina Yap", "type": "person", "status": "open", "scope": [ "pdac", "pancreatic-all", "colorectal", "hepatobiliary", "pan-cancer" ], "summary": "Yap brings clinical-trial statistics and adaptive platform design to a programme that must stop weak ideas early without corrupting error rates.", "content": "Current role: professor in clinical-trial biostatistics at the Institute of Cancer Research. Her listed experience includes early-phase, adaptive, stratified, basket, umbrella and platform trials. Proposed mission role: statistics chair for estimands, sample-size rules, multiplicity, missing data, model locking, interim looks and stop/go decisions across cancers. She should be called independent only after a formal conflict and sponsor-dependence screen; an institutional profile cannot prove that condition. Her profile also does not establish a direct pancreatic programme or availability. The mission needs this seat because the programme's value comes partly from stopping: if ten £2m branches are opened and six are stopped after a £0.5m test, £9m is preserved for the branches that clear their gates.", "sourceIds": [ "icr-christina-yap-profile-2026" ], "links": [ { "target": "people", "relation": "part-of" }, { "target": "programme-mission-table-people", "relation": "statistics-seat-on" }, { "target": "programme-100m-randomized-platform", "relation": "could-design" }, { "target": "risks", "relation": "controls" } ], "evidence": { "humanProspective": "not-applicable", "humanRetrospective": "not-applicable", "randomised": "not-applicable", "independentReplication": "not-applicable", "sampleSize": "One proposed statistics chair; worked example preserves £9m if six of ten £2m branches stop after £0.5m", "effect": "Trial-design and capital-control capability, not a clinical-effect claim", "limits": [ "Institutional profile", "Direct pancreatic programme not established", "Independence requires live conflict review", "Availability not checked" ] }, "layout": { "x": 73, "y": 117 }, "updated": "2026-09-14" }, { "id": "person-christine-iacobuzio-donahue", "title": "Christine Iacobuzio-Donahue — human PDAC evolution and metastasis", "shortTitle": "C. Iacobuzio-Donahue", "type": "person", "status": "supported", "scope": [ "pdac" ], "summary": "A pathology and evolution lead for reconstructing how primary tumours seed, diversify and kill across patients.", "content": "Current role: David M. Rubenstein Center chair for pancreatic cancer research and director of the Center for Pancreatic Cancer Research at Memorial Sloan Kettering. Her laboratory uses human pancreatic cancer to study the evolutionary steps from a mixed primary tumour to distant metastases. Mission role: own the human tissue, rapid-autopsy and spatial reconstruction needed to decide when dissemination occurs, whether resistant clones pre-exist, and which patients die from local versus distant failure. This capability directly addresses the programme's first bottleneck. Main weakness: end-stage autopsy and resection material can reconstruct history but cannot by itself prove which intervention would have changed it. Prospective serial sampling and treatment-linked outcomes are required.", "sourceIds": [ "msk-iacobuzio-donahue-profile-2026", "yachida-2010-metastatic-timing" ], "links": [ { "target": "people", "relation": "part-of" }, { "target": "claim-pdac-progression-timing-uncertain", "relation": "works-on" }, { "target": "claim-metastases-can-be-polyclonal", "relation": "works-on" }, { "target": "claim-pdac-death-can-be-local-or-metastatic", "relation": "works-on" }, { "target": "residual-disease", "relation": "supports" } ], "evidence": { "humanProspective": "weak", "humanRetrospective": "strong", "randomised": "not-applicable", "independentReplication": "moderate", "sampleSize": null, "effect": "Capability record, not an efficacy claim", "limits": [ "Institutional profile", "Much evidence is retrospective or end-stage", "Evolutionary reconstruction does not identify an effective treatment alone" ] }, "layout": { "x": 68, "y": 101 }, "updated": "2026-09-14" }, { "id": "person-cristina-ferrone", "title": "Cristina Ferrone — current multicentre perioperative-timing trial chair", "shortTitle": "Cristina Ferrone", "type": "person", "status": "supported", "scope": [ "pdac" ], "summary": "Ferrone chairs A021806, the 358-person North American randomized comparison that directly tests modern mFOLFIRINOX timing.", "content": "Evidence-linked role: the September 2026 NCT04340141 registry names Cristina R. Ferrone of Massachusetts General Hospital as study chair. A021806 has 358 actual participants and compares eight preoperative plus four postoperative mFOLFIRINOX cycles against upfront surgery plus twelve postoperative cycles, with overall survival primary. That role gives access to the exact multicentre route the mission needs to finish before another broad timing trial. Proposed mission role: obtain the locked protocol, complete intention-to-treat route and dose data, confirm event maturity, protect survival follow-up, and align timing fields with the surgical-route programme. Main limits: chairing a trial does not prove availability, independence, data rights or result timing; the registry lists 453 US and Canadian location records but does not state which actually enrolled patients. Estimated primary completion is December 2028. Any contract is for timely completion, independent analysis and durable access, not another study.", "sourceIds": [ "clinicaltrials-nct04340141-a021806" ], "links": [ { "target": "people", "relation": "part-of" }, { "target": "trial-pdac-perioperative-sequencing-action-map", "relation": "leads" }, { "target": "lab-alliance-a021806-network", "relation": "works-with" }, { "target": "unknown-pdac-perioperative-treatment-route", "relation": "studies" } ], "evidence": { "humanProspective": "strong — active randomized trial leadership", "humanRetrospective": "not-applicable", "randomised": "strong design; result pending", "independentReplication": "not-applicable", "sampleSize": "A021806 actual n=358", "effect": "Capability record, not an efficacy estimate", "limits": [ "Availability unconfirmed", "Trial result pending", "Data rights unknown", "Location list does not prove accrual", "Collaborative work cannot be assigned to one chair" ], "dimensions": { "biologicalPlausibility": { "rating": "not-applicable", "reason": "This record assesses trial leadership, not a biological claim." }, "animalEvidence": { "rating": "not-applicable", "reason": "Not relevant to trial leadership." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Not relevant to trial leadership." }, "reproducibility": { "rating": "moderate", "reason": "The role is verified in a current multicentre randomized registry; delivery results remain pending." }, "commercialConflicts": { "rating": "low", "reason": "Alliance sponsors and NCI collaborates on a comparison of established chemotherapy timing." }, "recency": { "rating": "current", "reason": "Registry role and status were verified in September 2026." } } }, "updated": "2026-09-14" }, { "id": "person-david-tuveson", "title": "David Tuveson — pancreatic models, organoids and treatment testing", "shortTitle": "David Tuveson", "type": "person", "status": "supported", "scope": [ "pdac", "pancreatic-all" ], "summary": "A model-building and biology lead whose lab links engineered mice and patient-derived organoids to diagnostics and treatment experiments.", "content": "Current role on 14 September 2026: professor and Cancer Center director at Cold Spring Harbor Laboratory, with an announced move to become executive director of Duke Cancer Institute on 31 October 2026. Demonstrated fit: his group built widely used mouse and organoid systems, studies RAS, redox metabolism, biomarkers and cancer-cell–fibroblast communication, and keeps a large human organoid collection. Mission role: run fast mechanism and combination screens against models that preserve different tumour states, then set a hard gate for human confirmation. Main weakness: even advanced organoids and engineered mice cannot reproduce human immune history, treatment exposure or metastatic spread. The programme should use this group to reject weak mechanisms cheaply, not to declare clinical success.", "contentSections": [ "Current human boundary: PASS-01 generated organoids in 50% of its randomized first-line cohort and found that second-line treatment lasted a median 2.1 months. The programme role is therefore to help lock fast shadow testing and treatment-interaction analysis, not to treat organoid growth as patient benefit." ], "sourceIds": [ "cshl-tuveson-profile-2026", "cshl-tuveson-duke-transition-2026", "knox-2025-pass01" ], "links": [ { "target": "people", "relation": "part-of" }, { "target": "claim-immune-exclusion-is-multicellular", "relation": "works-on" }, { "target": "claim-pdac-metabolic-adaptation-is-redundant", "relation": "works-on" }, { "target": "experiments", "relation": "supports" }, { "target": "funding", "relation": "depends-on" }, { "target": "trial-pdac-organoid-action-map", "relation": "could-test" } ], "evidence": { "humanProspective": "none", "humanRetrospective": "not-applicable", "randomised": "not-applicable", "independentReplication": "moderate", "sampleSize": null, "effect": "Capability record, not an efficacy claim", "limits": [ "Institutional profile", "Laboratory models do not prove patient benefit", "Institutional transition is pending" ] }, "layout": { "x": 65, "y": 100 }, "updated": "2026-09-15" }, { "id": "person-eduard-jonas", "title": "Eduard Jonas — South African pancreatic surgery and HPB delivery", "shortTitle": "Eduard Jonas", "type": "person", "status": "supported", "scope": [ "pdac", "pnet", "cholangiocarcinoma", "hcc", "hepatobiliary", "pancreatic-all" ], "summary": "Jonas connects pancreatic and liver clinical work at UCT and Groote Schuur to the practical access, registry and capacity constraints that determine whether an intervention can work in sub-Saharan Africa.", "content": "Current evidence places Jonas in the Surgical Gastroenterology Unit at the University of Cape Town and Groote Schuur Hospital. UCT names him as investigator or contact for pancreatic adenocarcinoma biomarker work, prospective quality-of-life follow-up, locally advanced PDAC evidence work, malignant hilar obstruction and GEP-NET natural history. He led the 2025 IHPBA pathway paper and co-authored a registered review of 3,989 HCC patients, a 602-person Cape Town public/private comparison and a 959-person South Africa-Sweden comparison. This moves his record beyond a project directory: the linked work measures stage, treatment and survival gaps while showing that observed comparisons cannot isolate one repair. Mission capability: design an African delivery route whose denominators, referral loss, pathology, surgery, liver disease, palliation and cost are measured rather than borrowed from high-income centres. Best role: African HPB delivery partner beside an outcomes lead, local public-health and patient leaders, not the sole continent-wide lead. Main limit: the evidence is observational and does not prove that a route repair improves survival.", "sourceIds": [ "uct-2026-cancer-projects", "jonas-2025-sub-saharan-hcc-pathways", "asombang-2026-africa-hpb-registry", "sobnach-2024-ssa-hcc-systematic-review", "sobnach-2025-south-africa-hcc-disparities", "sobnach-2026-global-hcc-disparities" ], "links": [ { "target": "people", "relation": "part-of" }, { "target": "lab-uct-groote-schuur-hpb", "relation": "leads" }, { "target": "lab-africa-hpb-cancer-consortium", "relation": "could-collaborate-with" }, { "target": "programme-mission-table-people", "relation": "international-contributor-to" }, { "target": "person-sanju-sobnach", "relation": "collaborates-with" }, { "target": "trial-regional-route-completion-action-map", "relation": "could-lead" } ], "evidence": { "humanProspective": "none for an intervention", "humanRetrospective": "strong — pancreatic and biliary work plus three linked HCC evidence sets", "randomised": "none", "independentReplication": "moderate for route measurement; none for a repair", "sampleSize": "HCC evidence includes n=3,989 systematic review, n=602 Cape Town comparison and n=959 South Africa/Sweden comparison", "effect": "People-map assessment, not a clinical-effect claim", "limits": [ "Project directory does not establish completion", "The HCC outcome papers are observational, not intervention results", "One South African centre cannot represent a continent", "Availability, conflicts, data rights and delivery capacity require current checks" ], "dimensions": { "biologicalPlausibility": { "rating": "not-applicable", "reason": "This record ranks delivery capability rather than a biological mechanism." }, "animalEvidence": { "rating": "not-applicable", "reason": "Animal work does not establish health-system execution." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Laboratory evidence is not the basis for this delivery role." }, "reproducibility": { "rating": "moderate", "reason": "The group has repeated route measurement across a review and two cohorts, but no delivery repair has been reproduced." }, "commercialConflicts": { "rating": "unknown", "reason": "A current conflict and sponsor review has not been completed." }, "recency": { "rating": "current", "reason": "The evidence includes a 2026 international cohort and current UCT project records." } } }, "layout": { "x": 61, "y": 118 }, "updated": "2026-09-14" }, { "id": "person-efrat-dotan", "title": "Efrat Dotan — older-adult PDAC trial lead", "shortTitle": "Efrat Dotan", "type": "person", "status": "supported", "scope": [ "pdac" ], "summary": "Dotan chaired GIANT, the first national randomized metastatic-PDAC treatment study built specifically for vulnerable older adults.", "content": "Evidence-linked role: the GIANT publications and registry name Efrat Dotan as the trial's lead investigator. The study enrolled 176 vulnerable adults aged 70-90 at 92 US sites and joined drug comparison to repeated measures of complex daily activity, nutrition, cognition, mood, quality of life, muscle and hospital use. Proposed mission role: recover the complete GIANT patient route, advise a rapid assessment-to-action field set, preserve patient-valued endpoints and help distinguish a prognostic risk domain from a correctable treatment barrier. Unique capability: a completed multicentre PDAC-specific network and a structured assessment made for vulnerable adults. Main limits: GIANT stopped for futility, the full trial result is not yet a peer-reviewed article, its registry has no posted results, and leadership does not prove availability, data rights or independence. Any agreement must cover complete treatment-start and discontinuation data, negative-result publication and independent analysis; it must not turn one assessment score into an automatic treatment denial.", "sourceIds": [ "dotan-2024-giant", "dotan-2025-giant-ga", "clinicaltrials-nct04233866-giant" ], "links": [ { "target": "people", "relation": "part-of" }, { "target": "trial-older-pdac-action-map", "relation": "leads" }, { "target": "lab-ecog-acrin-geriatric-oncology-network", "relation": "works-with" }, { "target": "unknown-older-pdac-treatment-route", "relation": "studies" } ], "evidence": { "humanProspective": "strong — randomized PDAC trial leadership", "humanRetrospective": "not-applicable", "randomised": "strong design; full result incomplete", "independentReplication": "not-applicable", "sampleSize": "GIANT actual n=176 at 92 enrolling sites", "effect": "Capability record, not a treatment-effect claim", "limits": [ "Availability unconfirmed", "Data rights unknown", "Full result not peer-reviewed", "Trial stopped for futility", "Registry results absent", "Network work is collaborative" ], "dimensions": { "biologicalPlausibility": { "rating": "not-applicable", "reason": "This record assesses trial leadership and data capability." }, "animalEvidence": { "rating": "not-applicable", "reason": "Not relevant to leadership." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Not relevant to leadership." }, "reproducibility": { "rating": "moderate", "reason": "The role is verified across trial publications and the registry; access remains unconfirmed." }, "commercialConflicts": { "rating": "moderate", "reason": "GIANT compares established drugs in a publicly funded network; investigator disclosures still require contract review." }, "recency": { "rating": "current", "reason": "The latest linked trial analysis is from 2025 and follow-up remains active in 2026." } } }, "updated": "2026-09-14" }, { "id": "person-eileen-oreilly", "title": "Eileen O'Reilly — PDAC and hepatobiliary trial execution", "shortTitle": "Eileen O'Reilly", "type": "person", "status": "supported", "scope": [ "pdac", "pnet", "cholangiocarcinoma", "gallbladder", "hcc", "hepatobiliary" ], "summary": "A clinical-trial and network lead spanning pancreatic, liver and biliary cancers, with direct involvement in major targeted and immune studies.", "content": "Current roles at Memorial Sloan Kettering include medical co-director of the Rubenstein Center, head of hepatopancreatobiliary and neuroendocrine cancers, pancreatic SPORE co-lead and NCI Alliance GI committee co-chair. Her active portfolio spans common-RAS drugs, vaccines, molecularly selected treatment, FAPI imaging and hepatobiliary trials. Mission role: assemble multi-centre trials, force common endpoints and specimens, and move the strongest biology into earlier disease and residual disease. Main weakness: breadth and access to industry can create correlated evidence and financial conflicts. Each programme decision needs an independent statistical lead, complete disclosure and replication outside the sponsor network.", "sourceIds": [ "msk-oreilly-profile-2026", "oreilly-2026-daraxonrasib", "wainberg-2025-eli002-final", "alagesan-2026-sig12d-loder" ], "links": [ { "target": "people", "relation": "part-of" }, { "target": "claim-daraxonrasib-common-ras-survival-breakthrough", "relation": "led-trial-for" }, { "target": "claim-eli002-mrd-frontier", "relation": "led-trial-for" }, { "target": "trial-daraxonrasib-expansion-programme", "relation": "supports" }, { "target": "companies", "relation": "requires-independent-check-from" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "moderate", "randomised": "strong", "independentReplication": "moderate", "sampleSize": null, "effect": "Capability record, not an efficacy claim", "limits": [ "Institutional profile", "Many industry relationships", "Trial leadership does not eliminate sponsor bias", "Requires independent statistics and replication" ] }, "layout": { "x": 77, "y": 100 }, "updated": "2026-09-14" }, { "id": "person-george-hanna", "title": "George Hanna — prospective NHS breath-test validation", "shortTitle": "George Hanna", "type": "person", "status": "open", "scope": [ "pdac", "pancreatic-all", "pan-cancer", "transfer-unproven" ], "summary": "Hanna leads VAPOR2, a planned 6,079-person prospective UK cohort that can test breath volatile compounds in the clinical population where they would be used.", "content": "Current roles listed by Imperial include surgical and NHS clinical leadership. His group joins breath chemistry, analytical hardware, model building and health economics. Mission role: UK health-system validation lead, using VAPOR2 to establish fixed-threshold sensitivity, specificity, positive predictive value, failed samples and route-to-diagnosis in symptomatic or referred patients. The study is recruiting, observational and has no posted result; it will not by itself prove mortality benefit. The earlier pancreatic signal came from a small case-control cohort and should not be treated as clinical validation. Because the institution reports patents and product-development work, assay ownership and financial interests must be disclosed and independent laboratories must reproduce the locked pipeline.", "sourceIds": [ "imperial-hanna-profile-2026", "nct07243262-vapor2" ], "links": [ { "target": "people", "relation": "part-of" }, { "target": "lab-imperial-vapor2", "relation": "leads" }, { "target": "detection-rna-protein-metabolite-methods", "relation": "tests" }, { "target": "programme-25m-prospective-observatory", "relation": "could-support" } ], "evidence": { "humanProspective": "none — planned study, no result", "humanRetrospective": "weak", "randomised": "none", "independentReplication": "none", "sampleSize": "VAPOR2 target n=6,079", "effect": "No VAPOR2 result posted", "limits": [ "Observational cohort", "No utility or mortality endpoint", "Earlier result was small and case-control", "Institutional intellectual-property interests require separation", "Availability not checked" ] }, "layout": { "x": 73, "y": 118 }, "updated": "2026-09-14" }, { "id": "person-jeffrey-zwicker", "title": "Jeffrey Zwicker — clinical thrombosis lead for isoquercetin and current Kinisoquin testing", "shortTitle": "Jeffrey Zwicker", "type": "person", "status": "supported", "scope": [ "pdac", "pan-cancer", "transfer-unproven" ], "summary": "Zwicker led the CATIQ phase 2, is named on the cancer-thrombosis patent family and is a current phase-3 investigator; he is useful for data access but not independent confirmation.", "content": "Role: thrombosis and haemostasis clinician-researcher linked to Beth Israel Deaconess and Harvard in the CATIQ paper, and named by Memorial Sloan Kettering as investigator for the current Kinisoquin phase 3. Contribution: led the 57-person sequential-dose isoquercetin study that established human marker activity and helped create the VTE endpoint and PDI route. Ownership link: named inventor on WO2020112828A1 and later family members covering isoquercetin or quercetin for cancer thrombotic events. Mission role: obtain the complete phase 2 dataset, expected-event assumptions, phase 3 protocol, event definitions, dose-selection plan and licence chain. Independence limit: the same investigator route connects discovery, patent, phase 2 and phase 3, so an independent statistician and event committee must verify any decisive result. This is a capability and access map, not an endorsement or allegation.", "sourceIds": [ "zwicker-2019-isoquercetin-catiq", "clinicaltrials-nct06861088-catiq-p3", "patent-wo2020112828-isoquercetin-cancer" ], "links": [ { "target": "people", "relation": "part-of" }, { "target": "lab-bidmc-thrombosis-pdi", "relation": "works-in" }, { "target": "company-quercis-pharma", "relation": "clinical-investigator-for" }, { "target": "claim-kinisoquin-phase2-marker-not-clinical-proof", "relation": "led" }, { "target": "claim-kinisoquin-access-stack", "relation": "could-unlock" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "not-applicable", "randomised": "none in completed Kinisoquin evidence", "independentReplication": "none", "sampleSize": "One 57-person phase 2, one active 480-person phase 3 and one patent family", "effect": "High direct capability and data access; independence is limited by repeated programme roles", "limits": [ "Public-record profile only", "Current institutional title not separately checked", "No interview", "No performance prediction", "Commercial and patent links require disclosure" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "His programme generated target-consistent human coagulation changes." }, "animalEvidence": { "rating": "moderate", "reason": "The programme links mechanistic models to clinical testing." }, "inVitroEvidence": { "rating": "strong", "reason": "PDI and coagulation assay work is central to the reported contribution." }, "reproducibility": { "rating": "none", "reason": "Independent randomized clinical reproduction has not reported." }, "commercialConflicts": { "rating": "high", "reason": "Discovery, patent and current trial roles are connected to the same product route." }, "recency": { "rating": "current", "reason": "The current phase 3 role and registry were checked in September 2026." } } }, "layout": { "x": 170, "y": 94 }, "updated": "2026-09-14" }, { "id": "person-jennifer-knox", "title": "Jennifer Knox", "shortTitle": "Jennifer Knox", "type": "person", "status": "supported", "scope": [ "pdac", "pan-cancer" ], "summary": "Clinical trial lead connecting randomized first-line PDAC treatment, molecular and organoid correlates, and the Canadian organoid-guidance route.", "content": "Public evidence role: Jennifer Knox is first author of PASS-01, the 160-person randomized comparison that integrated genome, RNA and organoid measurements, and is an author on the ADOPT organoid-guidance protocol. The University Health Network records connect her group to PROSPER-PANC, which targets 200 prospective patients, and ADOPT, which tests model-chosen treatment in about 25. Mission role: obtain the complete PASS-01 organoid and treatment data, define the all-patient route denominator, and help decide whether any assay-by-treatment interaction is strong enough for a randomized policy. Constraint: authorship proves relevant experience, not availability, authority over the data, willingness to collaborate or independence from the platform. Availability unconfirmed; data-access and conflict review are required before appointment.", "sourceIds": [ "knox-2025-pass01", "clinicaltrials-nct05927298-prosper-panc", "clinicaltrials-nct06813079-adopt" ], "links": [ { "target": "people", "relation": "part-of" }, { "target": "lab-princess-margaret-pancurx", "relation": "works-with" }, { "target": "trial-pdac-organoid-action-map", "relation": "could-test" }, { "target": "hypothesis-organoid-guided-early-switch", "relation": "could-test" } ], "evidence": { "humanProspective": "strong — randomized and prospective PDAC trial leadership", "humanRetrospective": "moderate", "randomised": "strong — first author of PASS-01", "independentReplication": "moderate", "sampleSize": "PASS-01 n=160 randomized; PROSPER-PANC estimated n=200; ADOPT estimated n=25", "effect": "Evidence-linked execution capability; not a treatment-effect record", "limits": [ "Availability unconfirmed", "Authorship is not appointment", "Data authority unconfirmed", "ADOPT is single-arm", "Current conflicts require review" ], "dimensions": { "biologicalPlausibility": { "rating": "not-applicable", "reason": "This record establishes documented clinical and translational capability." }, "animalEvidence": { "rating": "not-applicable", "reason": "Animal evidence is not needed to verify public study roles." }, "inVitroEvidence": { "rating": "moderate", "reason": "The linked programme includes patient-derived organoid development and screening." }, "reproducibility": { "rating": "moderate", "reason": "Relevant roles span a completed randomized trial and two current Canadian programmes." }, "commercialConflicts": { "rating": "unknown", "reason": "A fresh institutional and individual conflict review would be required." }, "recency": { "rating": "current", "reason": "The linked records were published or verified in 2025 and 2026." } } }, "layout": { "x": 131, "y": 187 }, "updated": "2026-09-15" }, { "id": "person-keiji-hanada", "title": "Keiji Hanada — community pancreatic detection in Japan", "shortTitle": "Keiji Hanada", "type": "person", "status": "supported", "scope": [ "pdac", "pancreatic-all" ], "summary": "Hanada joins community risk assessment, referral, advanced endoscopy and long follow-up in one operating Japanese pathway.", "content": "Current roles listed by JA Onomichi General Hospital include vice president and head of its pancreatic and biliary disease centre. Mission capability: reproduce and test a community-to-specialist route rather than export a stand-alone biomarker. The Onomichi programme began with local physicians in 2007, later expanded across Hiroshima as Hi-PEACE, and reported a higher share of stage 0/I diagnoses across eras. Hanada also leads the 10,000-person ENZEAVOUR feasibility study. Best role: Japanese health-system detection lead, accountable for denominators, referral completion, invasive work-up, interval cancers and transport beyond Onomichi. Main weakness: the Onomichi comparison is observational and ENZEAVOUR is uncontrolled with mainly positive-test verification; neither proves mortality benefit.", "sourceIds": [ "ja-onomichi-hanada-profile-2026", "hanada-2026-onomichi-method", "jrct-2026-enzeavour-feasibility" ], "links": [ { "target": "people", "relation": "part-of" }, { "target": "lab-onomichi-pancreas-biliary-centre", "relation": "leads" }, { "target": "claim-onomichi-community-pathway-stage-shift", "relation": "led" }, { "target": "trial-enzeavour-japan-screening", "relation": "leads" }, { "target": "programme-mission-table-people", "relation": "health-system-seat-on" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "strong", "randomised": "none", "independentReplication": "none", "sampleSize": "Onomichi pathway 29,642 assessed and 925 cancers; ENZEAVOUR target 10,000", "effect": "Capability record; Onomichi stage 0/I share rose 8.2 percentage points across eras", "limits": [ "No randomized utility result", "Institutional profile", "Availability and conflicts untested", "Japanese pathway may not transport to other systems" ] }, "layout": { "x": 73, "y": 115 }, "updated": "2026-09-14" }, { "id": "person-liwei-wang", "title": "Liwei Wang — Chinese pancreatic trial network and KRAS-G12D translation", "shortTitle": "Liwei Wang", "type": "person", "status": "supported", "scope": [ "pdac", "pan-cancer" ], "summary": "Wang leads the Renji pancreatic oncology centre and the HRS-4642 clinical programme, adding a major Chinese trial capability missing from a US-heavy table.", "content": "Current role: professor, chief physician and chair of Oncology and the Comprehensive Pancreatic Cancer Center at Renji Hospital, Shanghai Jiao Tong University School of Medicine. Mission capability: access to a broad Chinese pancreatic trial network, mutation-selected drug development and execution from phase 1 through phase 3. Evidence: senior clinical leadership on the multicentre HRS-4642 plus chemotherapy study and principal-investigator role for its blinded first-line phase 3. Why important: he brings recruitment scale, an independent RAS molecule and a non-US patient population to test whether efficacy and toxicity generalize. Why not an automatic mission-table seat: published controlled benefit is not yet available, several authors are sponsor employees, and discovery, patent and programme-finance details need deeper verification. Best role now: international trial and replication partner for RAS and precision-treatment work.", "sourceIds": [ "shsmu-2026-liwei-wang-profile", "cui-2026-hrs4642-pdac", "nct07232875-hrs4642-phase3" ], "links": [ { "target": "people", "relation": "part-of" }, { "target": "claim-hrs4642-g12d-chemo-signal", "relation": "led" }, { "target": "company-hengrui", "relation": "collaborates-with" }, { "target": "programme-mission-table-people", "relation": "international-contributor-to" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "none — ongoing trial, no result", "independentReplication": "none", "sampleSize": "Multicentre phase 1b/2 n=31; phase 3 ongoing", "effect": "People-map assessment, not an efficacy claim", "limits": [ "Institution profile is self-reported", "Grant and patent map incomplete", "Controlled phase 3 result unavailable", "Sponsor relationship requires conflict management" ] }, "layout": { "x": 71, "y": 113 }, "updated": "2026-09-14" }, { "id": "person-lois-daamen", "title": "Lois A Daamen", "shortTitle": "Lois Daamen", "type": "person", "status": "supported", "scope": [ "pdac" ], "summary": "A surgical-oncology investigator connecting national PDAC recurrence cohorts to the randomized RADAR-PANC surveillance test.", "content": "Evidence-linked role: Daamen is first author of the RADAR-PANC protocol, a listed principal investigator in the current NCT04875325 registry and a senior author on both the 333-person international and 1,311-person Dutch routine-imaging studies. That chain matters because the same group moved from a large but confounded survival association to a randomized overall-survival and quality-of-life test. Proposed mission role: post-resection surveillance and recurrence-route lead, responsible for actual randomization, crossover, scan burden, recurrence treatment, quality of life, cost and timely release of negative as well as positive results. Main limits: the registry has not been verified since September 2024, its estimated completion moved to October 2026 and no result is posted; investigator availability, current role, conflicts, data rights and UK recruitment must be confirmed directly. The mission should not ask this team to duplicate its trial. It should pay only for rapid data completion, independent analysis, common fields with CIRCPAC and durable access.", "sourceIds": [ "daamen-2024-radar-panc-protocol", "clinicaltrials-nct04875325-radar-panc", "andel-2025-routine-imaging", "van-goor-2025-nationwide-imaging" ], "links": [ { "target": "people", "relation": "part-of" }, { "target": "trial-pdac-postresection-surveillance-action-map", "relation": "leads" }, { "target": "lab-dpcg-radar-panc-network", "relation": "works-with" }, { "target": "unknown-postresection-surveillance-utility", "relation": "studies" }, { "target": "claim-routine-imaging-after-pdac-resection-not-yet-proven", "relation": "authored-evidence-for" } ], "evidence": { "humanProspective": "strong — trial leadership and prospective cohort work", "humanRetrospective": "strong — nationwide outcome analysis", "randomised": "strong design; result pending", "independentReplication": "not-applicable", "sampleSize": "Linked studies n=306 planned randomized, n=333 international and n=1,311 nationwide", "effect": "Capability record, not an efficacy estimate", "limits": [ "Availability unconfirmed", "Registry update stale", "Trial result pending", "Data-access terms unknown", "Collaborative work cannot be assigned to one author" ], "dimensions": { "biologicalPlausibility": { "rating": "not-applicable", "reason": "This record assesses human trial capability rather than a biological claim." }, "animalEvidence": { "rating": "not-applicable", "reason": "Not relevant to the role." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Not relevant to the role." }, "reproducibility": { "rating": "moderate", "reason": "The investigator is linked to national, international and randomized surveillance work." }, "commercialConflicts": { "rating": "low", "reason": "The linked programme is academic and does not require a proprietary imaging product." }, "recency": { "rating": "current", "reason": "The live trial remains due for a 2026 result, subject to direct confirmation." } } }, "updated": "2026-09-14" }, { "id": "person-marc-besselink", "title": "Marc Besselink", "shortTitle": "Marc Besselink", "type": "person", "status": "supported", "scope": [ "pdac", "pan-cancer" ], "summary": "Pancreatic surgeon and trial-network lead linking PREOPANC staging, TAPS selection models, PREOPANC-4 nationwide implementation and international practice measurement.", "content": "Public evidence role: Marc Besselink is the senior author on the 2026 PREOPANC-2 staging analysis, the TAPS non-therapeutic-laparotomy study, the 617-surgeon global staging survey, the two 2026 TAPS prediction papers and the PREOPANC-4 nationwide surgery report. Those records connect randomized-treatment infrastructure, cross-centre response selection, surgical harm and country-level implementation. PREOPANC-4 explored 180 selected locally advanced cases and performed 155 resections; major morbidity was 44% and 90-day mortality was 3%. The TAPS models cover 1,835 and 2,338 patients and explicitly separate prognostic prediction from causal treatment effect. Mission role: advise the English route on staging timing, response fields, non-therapeutic-laparotomy definitions and a possible surgery-versus-continued-treatment randomization. Constraint: authorship proves relevant work, not availability, willingness, independence from local practice or authority over English sites. Availability unconfirmed; conflict and data-access review are required before appointment.", "contentSections": [ "Earlier connected work includes 322 surgical explorations in PREOPANC-2, 663 selected locally advanced PDAC cases in TAPS and a staging survey spanning surgeons in 76 countries." ], "sourceIds": [ "rompen-2026-preopanc2-staging", "theijse-2024-nontherapeutic-laparotomy", "stoop-2026-global-staging-survey", "stoop-2026-preopanc4-surgery", "dekker-2026-taps-restaging", "verkolf-2026-taps-resection-prediction" ], "links": [ { "target": "people", "relation": "part-of" }, { "target": "lab-dpcg-radar-panc-network", "relation": "works-with" }, { "target": "trial-pdac-occult-metastasis-action-map", "relation": "leads" }, { "target": "hypothesis-pdac-two-timepoint-staging-route", "relation": "could-test" }, { "target": "trial-lapc-conversion-surgery-action-map", "relation": "could-test" }, { "target": "hypothesis-randomized-lapc-resection-after-response", "relation": "could-test" } ], "evidence": { "humanProspective": "strong — multicentre randomized parent-trial leadership", "humanRetrospective": "strong — international surgical cohort leadership", "randomised": "moderate — treatment randomized but staging method was not", "independentReplication": "moderate", "sampleSize": "PREOPANC-2 n=369; TAPS n=663, n=1,835 and n=2,338; PREOPANC-4 explored n=180; survey n=617", "effect": "Evidence-linked investigator capability; not a treatment-effect record", "limits": [ "Availability unconfirmed", "Authorship is not appointment", "Staging use not randomized", "Data access unconfirmed", "English delivery role unconfirmed" ], "dimensions": { "biologicalPlausibility": { "rating": "not-applicable", "reason": "This record establishes documented human-study capability rather than a biological claim." }, "animalEvidence": { "rating": "not-applicable", "reason": "No animal evidence is needed to verify clinical network roles." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "No laboratory evidence is needed to verify authorship and trial leadership." }, "reproducibility": { "rating": "strong", "reason": "Relevant leadership appears across staging, prediction, cohort and nationwide implementation records." }, "commercialConflicts": { "rating": "unknown", "reason": "A current appointment would require fresh conflict and institutional review." }, "recency": { "rating": "current", "reason": "The linked staging, TAPS and PREOPANC-4 reports were published in 2026." } } }, "updated": "2026-09-15" }, { "id": "person-michael-goggins", "title": "Michael Goggins — inherited-risk surveillance and pancreatic biomarkers", "shortTitle": "Michael Goggins", "type": "person", "status": "supported", "scope": [ "pdac", "pancreatic-all" ], "summary": "A surveillance and biomarker lead with direct experience following inherited high-risk cohorts over time.", "content": "Current role: Sol Goldman Professor at Johns Hopkins and director of its Pancreatic Cancer Early Detection Laboratory. His work covers inherited susceptibility, surveillance outcomes and genetic, epigenetic and protein markers. Mission role: define the highest-risk surveillance cohorts, provide serial samples before diagnosis, and set the clinical rules for moving from marker to MRI or endoscopic ultrasound. Main weakness: expert-centre inherited-risk surveillance covers a small, selected fraction of future PDAC and does not automatically transfer to new-onset diabetes or the general population. The programme needs his longitudinal samples and clinical discipline while separately testing scale and equity.", "sourceIds": [ "hopkins-goggins-profile-2026", "dbouk-2022-caps5", "blackford-2024-high-risk-surveillance" ], "links": [ { "target": "people", "relation": "part-of" }, { "target": "claim-high-risk-surveillance", "relation": "works-on" }, { "target": "claim-germline-risk-subset", "relation": "works-on" }, { "target": "biomarker-ca19-9-longitudinal-anchor", "relation": "connects-to" }, { "target": "detection", "relation": "supports" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "moderate", "randomised": "none", "independentReplication": "moderate", "sampleSize": null, "effect": "Capability record, not an efficacy claim", "limits": [ "Institutional profile", "High-risk cohorts are selected", "Surveillance evidence is nonrandomized", "Does not solve population-scale screening" ] }, "layout": { "x": 74, "y": 103 }, "updated": "2026-09-14" }, { "id": "person-nuria-malats", "title": "Núria Malats — European pancreatic risk epidemiology", "shortTitle": "Núria Malats", "type": "person", "status": "supported", "scope": [ "pdac", "pancreatic-all", "pan-cancer", "transfer-unproven" ], "summary": "Malats links pancreatic risk epidemiology, international cohorts and multi-source data to clinical and public-health questions.", "content": "Current role: leader of CNIO's Genetic and Molecular Epidemiology Group, which focuses on pancreatic and bladder cancers. Mission role: European risk and transportability lead, separating causal risk, correlated signals and population-specific calibration before a biomarker enters a screening route. CNIO participates in the €9.845m PANCAID project, which is developing a multi-marker liquid-biopsy panel and plans banked-sample validation before designing a later prospective study. That is useful discovery infrastructure, not clinical proof. Best mission contribution: make the future study prospective, population-based and decision-linked, with common negative controls and external sites. Main risks are multi-marker overfitting, dependence on stored case-control samples and no current clinical-utility result.", "sourceIds": [ "cnio-malats-profile-2026", "cordis-pancaid-101096309" ], "links": [ { "target": "people", "relation": "part-of" }, { "target": "lab-cnio-genetic-epidemiology", "relation": "leads" }, { "target": "detection-integrated-workflows", "relation": "supports" }, { "target": "programme-25m-prospective-observatory", "relation": "could-support" }, { "target": "funding", "relation": "connected-to" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "strong", "randomised": "none", "independentReplication": "weak — mixed evidence", "sampleSize": "PANCAID is a 19-participant EU project with total cost €9.845m; sample counts are not fixed in the CORDIS record", "effect": "Risk-epidemiology capability, not a test-performance claim", "limits": [ "Institutional profile", "PANCAID has no utility result", "Prospective validation is future work", "Availability and conflicts untested" ] }, "layout": { "x": 73, "y": 119 }, "updated": "2026-09-14" }, { "id": "person-regina-barzilay", "title": "Regina Barzilay — clinical AI and model-to-experiment design", "shortTitle": "Regina Barzilay", "type": "person", "status": "supported", "scope": [ "pan-cancer", "pdac", "transfer-unproven" ], "summary": "An external AI lead who can turn imaging, pathology and molecular data into locked predictions and design the trial that can disprove them.", "content": "Current role: MIT School of Engineering Distinguished Professor for AI and Health and AI faculty lead at the MIT Jameel Clinic. Her programmes span cancer-risk prediction, pathology models, clinical records, molecule design and protein binders. Mission role: define which decision a model changes, lock inputs and outputs before validation, quantify calibration and transport across hospitals, and make the model face a prospective experiment. Unique capability: she can connect clinical prediction with molecule and protein design without treating either as a substitute for biology or trials. Main weakness: her strongest public clinical examples are outside PDAC, and an accurate model can still encode care patterns or fail to improve outcome. The first PDAC work should be a methods and trial-design role around existing CT, pathology, organoid and molecular-residual-disease data, with no deployment claim until external and prospective validation.", "sourceIds": [ "mit-barzilay-profile-2026", "mit-2026-barzilay-aacr", "hendifar-2026-ai-chemo-pathology" ], "links": [ { "target": "people", "relation": "part-of" }, { "target": "ai-programme", "relation": "supports" }, { "target": "technology-ai-pathology-chemo-selection", "relation": "connects-to" }, { "target": "technology-opportunistic-ct-ai-pancreas", "relation": "connects-to" }, { "target": "hypothesis-multimodal-mrd-action", "relation": "tests" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "strong", "randomised": "none", "independentReplication": "moderate", "sampleSize": null, "effect": "Cross-cancer capability record; PDAC mission effect unproved", "limits": [ "Institutional sources", "Most demonstrated clinical work is outside PDAC", "Prediction does not prove clinical utility", "Models can inherit hospital and treatment-selection bias", "Prospective external tests required" ] }, "layout": { "x": 98, "y": 102 }, "updated": "2026-09-14" }, { "id": "person-renata-dalpino-peixoto", "title": "Renata D'Alpino Peixoto — Latin American pancreatic data network", "shortTitle": "Renata D'Alpino Peixoto", "type": "person", "status": "supported", "scope": [ "pdac", "pnet", "hepatobiliary", "pancreatic-all", "pan-cancer" ], "summary": "Peixoto led the GASPAR multi-country cohort, giving the mission a named route into Latin American pancreatic data and participating centres.", "content": "Oncoclinicas identifies Peixoto as a gastrointestinal and neuroendocrine clinical oncologist in Sao Paulo. ClinicalTrials.gov names her as principal investigator of LACOG 0222 GASPAR, and the 2025 results abstract lists her first. GASPAR reports 201 people across seven sites in Brazil, Argentina and Colombia, including 79 with advanced pancreatic cancer. Mission capability: convene Latin American centres, standardize real-world pancreatic data and expose treatment attrition and access limits before a trial is designed. Best role now: Latin American data and delivery contributor, not a sole regional representative or treatment-efficacy lead. Main limits: the cohort is observational and small; 76% of the combined cohort had private coverage; 90% of the pancreatic cohort was recorded as white and none as Black; and the registry's 120-person actual enrolment conflicts with the 201-person abstract without explanation.", "sourceIds": [ "oncoclinicas-2026-peixoto-profile", "clinicaltrials-nct05924789-gaspar", "peixoto-2025-gaspar-results" ], "links": [ { "target": "people", "relation": "part-of" }, { "target": "lab-lacog-gaspar", "relation": "leads" }, { "target": "trial-gaspar-lacog-0222", "relation": "led" }, { "target": "programme-mission-table-people", "relation": "international-contributor-to" } ], "evidence": { "humanProspective": "moderate — bidirectional observational cohort", "humanRetrospective": "moderate", "randomised": "none", "independentReplication": "none", "sampleSize": "GASPAR n=201 overall and n=79 pancreatic in the abstract; registry reports n=120 actual", "effect": "People-map assessment, not a treatment-effect claim", "limits": [ "Small selected advanced-disease cohort", "Mostly private coverage", "Poor racial representation", "Registry and abstract enrolment totals conflict", "Availability, conflicts and data rights require current checks" ], "dimensions": { "biologicalPlausibility": { "rating": "not-applicable", "reason": "This record ranks data-network capability rather than biology." }, "animalEvidence": { "rating": "not-applicable", "reason": "Animal work does not establish regional cohort execution." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Laboratory evidence is not the basis for this role." }, "reproducibility": { "rating": "weak", "reason": "One multi-country cohort is reported and its enrolment count conflicts with the registry." }, "commercialConflicts": { "rating": "moderate", "reason": "The employer is a commercial oncology group and the registry names Astellas as collaborator; the effect on the study requires review." }, "recency": { "rating": "current", "reason": "The cohort result is from 2025 and role and registry status were checked in 2026." } } }, "layout": { "x": 51, "y": 118 }, "updated": "2026-09-14" }, { "id": "person-robert-vonderheide", "title": "Robert Vonderheide — pancreatic immune activation and CD40", "shortTitle": "Robert Vonderheide", "type": "person", "status": "supported", "scope": [ "pdac", "pan-cancer" ], "summary": "An immune-mechanism and translation lead focused on antigen presentation, CD40 activation and combinations that may convert immune-cold PDAC.", "content": "Current role: director of the Abramson Cancer Center at the University of Pennsylvania and leader of a laboratory spanning mouse models and clinical investigation. His group focuses on CD40, vaccines, checkpoint blockade, radiation and the suppressive pancreatic tumour environment. Mission role: define which missing immune step blocks each patient, then test combinations in human tumour samples rather than adding more agents by intuition. Main weakness: PDAC immunotherapy has a large mouse-to-human translation gap, and unselected checkpoint trials have produced almost no responses. The programme should require measurable human antigen presentation, T-cell entry, expansion and persistence before treating response rate as the next gate.", "sourceIds": [ "penn-vonderheide-profile-2026", "royal-2010-ipilimumab", "oreilly-2019-durvalumab-tremelimumab", "aznar-2025-mesothelin-car-t-resistance" ], "links": [ { "target": "people", "relation": "part-of" }, { "target": "claim-immune-exclusion-is-multicellular", "relation": "works-on" }, { "target": "failure-checkpoint-blockade-unselected-pdac", "relation": "must-overcome" }, { "target": "failure-mesothelin-car-t-pdac", "relation": "must-overcome" }, { "target": "claim-neoantigen-vaccine-frontier", "relation": "connects-to" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "moderate", "randomised": "weak", "independentReplication": "moderate", "sampleSize": null, "effect": "Capability record, not an efficacy claim", "limits": [ "Institutional profile", "Large mouse-to-human gap", "Unselected PDAC immunotherapy has repeatedly failed", "Combination complexity can obscure the active component" ] }, "layout": { "x": 83, "y": 102 }, "updated": "2026-09-14" }, { "id": "person-sanju-sobnach", "title": "Sanju Sobnach — African HCC outcomes and route measurement", "shortTitle": "Sanju Sobnach", "type": "person", "status": "supported", "scope": [ "hcc", "hepatobiliary", "transfer-unproven" ], "summary": "Sobnach connects a sub-Saharan HCC systematic review to public/private and South Africa/Sweden patient comparisons at UCT, making him a strong route-measurement lead rather than proof of an intervention.", "content": "Current evidence places Sobnach in the University of Cape Town surgery group and as first author of three linked HCC analyses: a registered review of 39 reports and 3,989 patients across sub-Saharan Africa; a 602-patient Cape Town public/private comparison; and a 959-patient South Africa/Sweden comparison. This is a useful chain from continent-wide reporting gaps to patient-level stage, treatment and survival differences. Mission capability: define route-loss fields, distinguish selection from treatment effect, run outcome comparisons and connect them to an African clinical site. Best role: African HCC outcomes and registry methods lead alongside local site, patient, primary-care and public-health leaders. Main limits: all reported effect evidence is observational; the group has not yet shown that a delivery intervention changes treatment completion or survival; UCT cannot represent other countries; availability, conflicts, data rights and operational capacity require direct checks.", "sourceIds": [ "sobnach-2024-ssa-hcc-systematic-review", "sobnach-2025-south-africa-hcc-disparities", "sobnach-2026-global-hcc-disparities" ], "links": [ { "target": "people", "relation": "part-of" }, { "target": "lab-uct-groote-schuur-hpb", "relation": "works-at" }, { "target": "trial-regional-route-completion-action-map", "relation": "could-lead" }, { "target": "person-eduard-jonas", "relation": "collaborates-with" }, { "target": "programme-mission-table-people", "relation": "international-contributor-to" } ], "evidence": { "humanProspective": "none for an intervention", "humanRetrospective": "strong for relevant outcome analysis", "randomised": "none", "independentReplication": "moderate for route failure across three linked evidence sets", "sampleSize": "Systematic review n=3,989; Cape Town cohort n=602; South Africa/Sweden cohort n=959", "effect": "People-map assessment, not a clinical-effect claim", "limits": [ "Observational programme", "Single home institution", "Overlapping research group and patient sources", "No tested route repair", "Future availability and authority unverified" ], "dimensions": { "biologicalPlausibility": { "rating": "not-applicable", "reason": "This record ranks research and delivery capability." }, "animalEvidence": { "rating": "not-applicable", "reason": "Animal work does not establish outcome-research execution." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Laboratory evidence is not the basis for this role." }, "reproducibility": { "rating": "moderate", "reason": "The team has produced a connected review and two patient-level comparisons, but no intervention replication." }, "commercialConflicts": { "rating": "low", "reason": "The comparative cohorts report foundation funding rather than a tested product sponsor." }, "recency": { "rating": "current", "reason": "The latest multicentre result was published in 2026." } } }, "updated": "2026-09-14" }, { "id": "person-suresh-chari", "title": "Suresh Chari — new-onset diabetes and pancreatic early detection", "shortTitle": "Suresh Chari", "type": "person", "status": "supported", "scope": [ "pdac", "pancreatic-all" ], "summary": "A clinical risk-enrichment lead for turning pancreatic-cancer-associated diabetes into a testable care pathway.", "content": "Current role: professor of gastroenterology, hepatology and nutrition at MD Anderson. His listed work spans the END-PAC score, prospective new-onset-diabetes cohorts, biomarker resources, electronic-record signals and imaging AI. Mission role: design a health-system pathway that identifies older adults whose new or rapidly changing diabetes is caused by occult pancreatic cancer, then tests them with a second-stage assay and pancreas imaging. Main weakness: the signal is diluted by ordinary type 2 diabetes and varies across racial and ethnic groups. The person record is therefore tied to the transportability and clinical-utility experiment, not to a claim that diabetes alone is an adequate screen.", "sourceIds": [ "mdanderson-chari-profile-2026", "sharma-2018-endpac", "chari-2026-prospective-new-onset-diabetes" ], "links": [ { "target": "people", "relation": "part-of" }, { "target": "claim-new-onset-diabetes-enriches-risk", "relation": "works-on" }, { "target": "hypothesis-one-time-risk-triggered-workup", "relation": "tests" }, { "target": "technology-opportunistic-ct-ai-pancreas", "relation": "connects-to" }, { "target": "experiments", "relation": "supports" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "moderate", "randomised": "none", "independentReplication": "moderate", "sampleSize": null, "effect": "Capability record, not an efficacy claim", "limits": [ "Institutional profile", "Risk enrichment has not shown mortality benefit", "Population transportability remains unresolved" ] }, "layout": { "x": 71, "y": 102 }, "updated": "2026-09-14" }, { "id": "person-tingbo-liang", "title": "Tingbo Liang — single-centre randomized treatment-first capability", "shortTitle": "Tingbo Liang", "type": "person", "status": "supported", "scope": [ "pdac" ], "summary": "Liang is the registered principal investigator for CISPD-1, the positive 324-person Chinese phase 3 timing trial whose transfer must be tested.", "content": "Evidence-linked role: the NCT03750669 registry names Tingbo Liang as responsible principal investigator at the Second Affiliated Hospital of Zhejiang University, and the 2025 report lists him as senior author. CISPD-1 randomized 324 resectable patients and reported event-free survival HR .71 and overall survival HR .73, with the 95% confidence interval reaching 1.00. Proposed mission role: supply the full protocol, route data, tumour-location strata, treatment delivery, severe-event detail, analysis code and an outside-centre reproduction plan. Main limits: the trial ran at one expert Hangzhou centre, used a complex sequential regimen and enrolled an unusual 50% body-or-tail tumour mix. Investigator availability, data rights, conflicts and ability to transfer delivery are unconfirmed. The positive result supports reproduction, not immediate universal adoption.", "sourceIds": [ "bai-2025-cispd1" ], "links": [ { "target": "people", "relation": "part-of" }, { "target": "trial-pdac-perioperative-sequencing-action-map", "relation": "contributed-to" }, { "target": "lab-zhejiang-pancreatic-centre", "relation": "works-with" }, { "target": "unknown-pdac-perioperative-treatment-route", "relation": "studies" } ], "evidence": { "humanProspective": "strong — phase 3 trial leadership", "humanRetrospective": "not-applicable", "randomised": "strong single-centre result", "independentReplication": "none", "sampleSize": "CISPD-1 n=324", "effect": "Capability record, not an independent efficacy estimate", "limits": [ "Availability unconfirmed", "Single-centre result", "Data access unknown", "Complex regimen", "External delivery unproved" ], "dimensions": { "biologicalPlausibility": { "rating": "not-applicable", "reason": "This record assesses trial capability." }, "animalEvidence": { "rating": "not-applicable", "reason": "Not relevant to the role." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Not relevant to the role." }, "reproducibility": { "rating": "weak", "reason": "One randomized centre supports the capability, but the positive sequence has no outside reproduction." }, "commercialConflicts": { "rating": "unknown", "reason": "The public abstract and registry do not supply a complete current conflict and data-rights audit." }, "recency": { "rating": "current", "reason": "The phase 3 report appeared in late 2025." } } }, "updated": "2026-09-14" }, { "id": "person-vinod-balachandran", "title": "Vinod Balachandran — exceptional survivors and personalised cancer vaccines", "shortTitle": "Vinod Balachandran", "type": "person", "status": "supported", "scope": [ "pdac", "hepatobiliary", "pan-cancer" ], "summary": "A surgeon-scientist who can connect rare durable PDAC immunity to vaccine design, in-house manufacture and prospective trials.", "content": "Current role: hepatopancreatobiliary surgeon at Memorial Sloan Kettering, member of its Immuno-Oncology Program and director of the Olayan Center for Cancer Vaccines. His group worked backwards from rare long-term PDAC survivors to mutation-derived neoantigen recognition, then ran a 16-person personalised RNA-vaccine study with deep immune tracking. Mission role: own rapid patient-specific vaccine manufacture and the experiment that asks whether vaccine-induced T cells actually remove residual disease. Unique capability: the Olayan Center joins design, manufacture, dosing and repeated immune measurement in one system. Main weakness: the current clinical signal compares immune responders with nonresponders after everyone was offered the vaccine; it is neither randomized nor proof that vaccination caused longer survival. The phase 2 trial must report all assigned patients, manufacturing failures, time to product, toxicity and recurrence, not only immune responders.", "sourceIds": [ "msk-balachandran-profile-2026", "rojass-2025-neoantigen-followup" ], "links": [ { "target": "people", "relation": "part-of" }, { "target": "claim-neoantigen-vaccine-frontier", "relation": "led-trial-for" }, { "target": "hypothesis-low-burden-ras-immune-extinction", "relation": "tests" }, { "target": "claim-immune-exclusion-is-multicellular", "relation": "works-on" }, { "target": "programme-100m-randomized-platform", "relation": "supports" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "moderate", "randomised": "none", "independentReplication": "weak", "sampleSize": "16 vaccinated patients in the published PDAC follow-up", "effect": "Capability record; immune engagement shown, treatment effect unproved", "limits": [ "Institutional profile", "Single-centre phase 1 cohort", "Responder analysis is post-treatment selection", "Personalised manufacturing may miss the treatment window", "Randomized phase 2 pending" ] }, "layout": { "x": 86, "y": 100 }, "updated": "2026-09-14" }, { "id": "person-wungki-park", "title": "Wungki Park — early clinical tests of KRAS and immune combinations", "shortTitle": "Wungki Park", "type": "person", "status": "supported", "scope": [ "pdac", "cholangiocarcinoma", "gallbladder", "hepatobiliary" ], "summary": "An early-phase clinical lead working on KRAS-targeted drugs, degradation and immune combinations across pancreas and biliary cancers.", "content": "Current role: gastrointestinal medical oncologist and assistant attending physician at Memorial Sloan Kettering. His listed trials include daraxonrasib combinations, setidegrasib, other RAS drugs, myeloid-targeted immunotherapy and CLDN18.2 therapies. Mission role: build early trials that measure target engagement and resistance before launching large studies, then carry useful agents into first-line and residual disease. Main weakness: a wide early-phase portfolio creates multiple comparisons and sponsor exposure. Response signals must be judged against trial design, denominators, mutation selection and later randomized results; the POLAR study and siG12D-LODER trial show why technically interesting studies can still miss their endpoints.", "sourceIds": [ "msk-park-profile-2026", "park-2026-setidegrasib", "park-2026-polar-pdac", "alagesan-2026-sig12d-loder" ], "links": [ { "target": "people", "relation": "part-of" }, { "target": "claim-setidegrasib-g12d-degradation", "relation": "led-trial-for" }, { "target": "failure-polar-missed-primary-thresholds", "relation": "led-trial-for" }, { "target": "failure-sig12d-loder-randomized-os-not-shown", "relation": "coauthored" }, { "target": "experiments", "relation": "supports" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "moderate", "randomised": "moderate", "independentReplication": "weak", "sampleSize": null, "effect": "Capability record, not an efficacy claim", "limits": [ "Institutional profile", "Early-phase focus", "Multiple sponsor relationships", "Many programmes lack randomized results" ] }, "layout": { "x": 80, "y": 101 }, "updated": "2026-09-14" }, { "id": "pnet", "title": "Pancreatic neuroendocrine tumours", "shortTitle": "pNET", "type": "cancer", "status": "supported", "scope": [ "pnet" ], "summary": "pNET is a separate pancreatic disease system in which grade, differentiation, receptor expression and distribution determine routes to surgical or chronic control.", "content": "Pancreatic neuroendocrine tumours share an organ with PDAC but not one disease model. Well-differentiated pNET can remain controlled for years after surgery, somatostatin analogues, kinase inhibitors or receptor-directed radiation. That claim does not transfer to poorly differentiated pancreatic neuroendocrine carcinoma, which needs separate tissue and molecular classification and often platinum-based treatment. Classification also changes with time. In a prospective 21-person progressive metastatic cohort, 13 of 16 adequate repeat biopsies showed higher Ki-67 and eight changed from grade 2 to 3; combined biopsy and dual PET changed treatment in seven people, without proving that the changes improved outcome. SEQTOR randomized 141 people to everolimus and streptozotocin plus 5-FU in opposite orders and found no significant PFS or survival winner, although chemotherapy shrank more tumours. The route to durable control is therefore repeated grade, differentiation and receptor assessment tied to a sequence trial—not a one-time diagnosis or a list of active drugs.", "contentSections": [ "Early localized disease creates a different trade-off. ASPEN observed 406 of 500 small sporadic non-functioning tumours and PANDORA observed 76; neither reported a new distant metastasis during median 25 or 17 months, but neither establishes lifetime safety. In ASPEN, patient or centre preference explained 79 of 94 operations and 13% of operated patients had severe complications. A 2026 cohort of 460 small-tumour resections reported 14.8% major morbidity and 0.9% 90-day mortality. The rule cannot be diameter alone: size above 10 mm and duct dilation concentrated ASPEN's adverse findings, while the actual required procedure changes harm.", "MEN1 makes early detection rational but does not solve action selection. Among 350 DutchMEN patients, five developed a clinically relevant tumour before 18 and two later had nodal metastasis. A 110-person prospective circulating mRNA study then failed to predict new tumour, growth or metastasis. Surveillance should therefore find and follow lesions, while a separate study tests which change should trigger treatment." ], "sourceIds": [ "scarpa-2017-pnet", "marchegiani-2019-pnet-recurrence", "dong-2020-pnet-recurrence", "sigel-2018-panenen-differentiation", "tanaka-2020-panec-g3", "baudin-2026-oclurandom", "chan-2026-control-nets", "mollazadegan-2025-pnet-recharacterization", "capdevila-2025-seqtor-pnet", "partelli-2022-aspen-interim", "heidsma-2021-pandora-pnet", "maekawa-2026-small-pnet-surgery-morbidity", "klein-haneveld-2021-men1-young-screening", "van-vliembergen-2026-men1-mnet-assay" ], "links": [ { "target": "disease-system", "relation": "part-of" }, { "target": "model-pnet-causal-chain", "relation": "contains" }, { "target": "claim-pnet-surgical-control-is-risk-stratified", "relation": "contains" }, { "target": "claim-panec-needs-separate-classification", "relation": "contains" }, { "target": "claim-pnet-state-changes-at-progression", "relation": "contains" }, { "target": "claim-pnet-seqtor-no-sequence-winner", "relation": "contains" }, { "target": "unknown-pnet-reclassification-actionability", "relation": "contains" }, { "target": "claim-small-nfpnet-surveillance-surgery-tradeoff", "relation": "contains" }, { "target": "claim-men1-surveillance-finds-lesions-not-danger", "relation": "contains" }, { "target": "unknown-small-pnet-surgery-selector", "relation": "contains" }, { "target": "trial-small-pnet-action-map", "relation": "contains" }, { "target": "treatment", "relation": "part-of" }, { "target": "transfer", "relation": "may-transfer-to" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "strong", "randomised": "strong for chronic control; weak for curative sequencing", "independentReplication": "moderate", "sampleSize": "Large resection cohorts, high-grade classification series and randomized drug or radioligand trials", "effect": "Some well-differentiated pNETs can be controlled for years; grade, differentiation and receptor state determine route", "limits": [ "Poorly differentiated carcinoma is a separate disease", "Retrospective surgery cohorts are selected", "Treatment sequence is not established", "Overall-survival proof is weaker than progression proof" ] }, "layout": { "x": 6, "y": 7 }, "updated": "2026-09-14" }, { "id": "prevention-ranking", "title": "Prevention map: PDAC risk can be reduced and enriched, but not yet reliably intercepted", "shortTitle": "Prevention ranking", "type": "system", "status": "supported", "scope": [ "pdac", "hcc", "colorectal", "pan-cancer", "transfer-unproven" ], "summary": "Smoking, weight, glucose control and selected surveillance matter, but PDAC lacks a removable cause or safely selectable precursor covering most cases.", "content": "Proven prevention in related cancers sets the target: hepatitis-B vaccination and antiviral suppression remove an upstream HCC cause; colorectal screening removes premalignant polyps and reduces deaths. PDAC has no equivalent covering most cases. Modifiable risk: global burden modelling attributed 21.1% of pancreatic deaths to smoking, 8.9% to high fasting glucose and 6.2% to high BMI in 2017; these fractions overlap and cannot be added. A UK Biobank cohort of 340,631 people and 1,129 cancers associated a favourable lifestyle with HR 0.48 versus unfavourable lifestyle, but the absolute risk reduction in people over 70 was 0.19% and observational confounding remains. A 182,011-person Multiethnic Cohort found 9% lower excess risk per year since smoking cessation, while bariatric-surgery and GLP-1 studies reported favourable pancreatic-cancer associations. None randomized a pancreatic-cancer prevention action, so no surgery or diabetes drug should be prescribed solely for that purpose. Polygenic risk scores change relative risk and may help combine age and clinical triggers; they do not lift most people to an invasive-screening threshold. Pancreatitis: Korean data show recurrent acute pancreatitis at 12.69 cancers per 10,000 person-years, about 0.127% yearly, HR 5.00, and chronic pancreatitis with acute episodes at 12.12 per 10,000, HR 5.74. A 2026 Danish cohort showed that five-year pancreatic-cancer incidence fell from 3.4% to 0.5% after excluding the first two years, so a new diagnosis is partly an occult-cancer warning rather than a clean prevention group. Risk remains too low for pancreatectomy. International consensus supports surveillance for inherited PRSS1 pancreatitis in specialist centres, not SPINK1 N34S alone. Inherited and familial risk: MRI/EUS surveillance shifts stage in expert cohorts, but covers a minority, requires lifelong specialist capacity and lacks randomized mortality proof. New-onset diabetes is a temporary warning state: 18,838-person prospective surveillance found 0.62% three-year risk, but a 93-scan pilot found 56% extrapancreatic incidental findings and only one stage-IV PDAC; PANDOME found one stage-I PDAC among 109 participants. EDI's 8,869-person non-randomized site comparison is unfinished. Precursors: PanIN is usually invisible; IPMN is visible but common; KRAS and GNAS do not identify lethal progression. Immune interception: mutant-KRAS vaccination is safe and immunogenic in 20 high-risk people but has not prevented cancer. Microbiome: prediagnostic oral Porphyromonas gingivalis carriage was associated with OR 1.60 and Aggregatibacter with OR 2.20 in 361 cases and 371 controls, but association is not a causal, reversible intervention. Mendelian-randomization claims remain instrument- and ancestry-sensitive. Alcohol matters mainly through heavy exposure and pancreatitis in current evidence; no pancreatic-cancer prevention trial supports a test or drug. Broader inflammatory states may mark risk but do not name a safe intervention. Practical programme: universal smoking cessation, healthy weight, metabolic care and alcohol-harm reduction for broad health benefit; germline testing and specialist surveillance where risk criteria are met; prompt diagnostic exclusion after new unexplained pancreatitis; complete EDI before duplicating a large diabetes work-up trial; long-term precursor classification; no microbiome treatment, PRS-only screening, metabolic medicine or prevention vaccine for pancreatic prevention outside trials. The missing capability is a marker of dangerous progression paired with a safe interceptive action. Germline delivery is also a family-prevention route, not just a risk label. Family-history criteria miss some carriers, while relatives who enter a low-friction remote programme usually complete testing. The larger loss happens before entry: one prospective pancreatic cohort found only 31% of informed first-degree relatives tested, and a pan-cancer randomized study found only 19.2% of reported relatives invited. England's current R367 rule remains age- and history-limited. A prospective English comparison should measure pathogenic findings and actual gene-specific actions inside and outside that rule before national expansion; variants of uncertain significance are never an action.", "contentSections": [ "Gallbladder cancer is a separate high-variance prevention case. Removing the organ can remove future risk, but common stones and polyps make a universal operation rule unsafe. The current priority is not a larger diagnostic AUC study. It is a country-specific future-risk rule followed by independent lesion confirmation, tested against local operation death, bile-duct injury, other complications, HGD or stage-I yield and interval cancer.", "HCC retains the clearest cause-control proof. HBV vaccination and antiviral treatment reduce cancer or liver-disease events, while alcohol abstinence has broad health value and an observational HCC signal before decompensation. ATTENTION suggests that treatment may help earlier in noncirrhotic HBV: HCC occurred in 2 of 369 treated people and 7 of 365 observed people. Only nine HCCs occurred, the 97.5% confidence interval crossed no effect and the trial did not cross its early-stop boundary. Keep the result as a live prevention boundary, not a booked population effect.", "Colorectal prevention demonstrates both possibility and delivery loss. Repeated stool testing and one-time sigmoidoscopy reduced incidence and death. In NordICC, one colonoscopy invitation reduced incidence from 1.80% to 1.46% over 13 years but did not significantly reduce colorectal-cancer death; only 42.0% underwent screening. The intervention rule therefore includes invitation, uptake, procedure quality and follow-up, not colonoscopy alone." ], "sourceIds": [ "chang-1997-hbv-vaccine-hcc", "mandel-2000-fobt-incidence", "ando-2019-smoking-japan", "bogumil-2024-smoking-cessation-multiethnic", "zhang-2022-uk-biobank-risk-fractions", "schauer-2019-bariatric-pancreatic-risk", "wang-2025-glp1-pancreatic-risk", "klein-2024-ukb-prs-lifestyle", "ding-2026-prs-modifiable-risk", "kim-2023-pancreatitis-risk-korea", "davidsen-2026-chronic-pancreatitis-mortality", "greenhalf-2020-prss1-surveillance", "blackford-2024-high-risk-surveillance", "chari-2026-prospective-new-onset-diabetes", "wu-2022-nod-imaging-pilot", "frank-2025-pandome", "edi-nct04662879", "haldar-2026-kras-interception", "fan-2018-oral-microbiome-pdac-risk", "van-dijk-2025-polyp-study", "wong-2025-ambrose-cholecystectomy", "clinicaltrials-nct06192719-eulat-gbc", "koshiol-2026-chile-bils", "lim-2025-attention-hbv", "bretthauer-2022-nordicc-10y", "kaminski-2026-nordicc-13y", "nhs-england-2026-r367", "llach-2024-universal-germline", "wang-2022-oncology-germline-cascade", "rodriguez-2024-generate", "katz-2026-gift-cascade" ], "links": [ { "target": "detection", "relation": "part-of" }, { "target": "claim-modifiable-risk-reduces-but-cannot-eliminate-pdac", "relation": "extends" }, { "target": "claim-germline-risk-subset", "relation": "depends-on" }, { "target": "claim-pancreatitis-high-relative-low-absolute-risk", "relation": "depends-on" }, { "target": "unknown-dangerous-precursor-selection", "relation": "blocked-by" }, { "target": "claim-kras-interception-vaccine-immune-response-not-prevention", "relation": "ranks" }, { "target": "transfer-hpv-screening-action-loop", "relation": "contrasts-with" }, { "target": "claim-hcc-prevention-is-proven-for-viral-cause", "relation": "contrasts-with" }, { "target": "claim-hcc-early-hbv-treatment-signal-immature", "relation": "ranks" }, { "target": "prevention-risk-interception-audit", "relation": "contains" }, { "target": "claim-pdac-primary-prevention-action-gap", "relation": "contains" }, { "target": "trial-new-onset-diabetes-workup-map", "relation": "contains" }, { "target": "trial-gallbladder-prevention-detection-action-map", "relation": "contains" }, { "target": "hypothesis-gallbladder-two-stage-interception-rule", "relation": "ranks" }, { "target": "claim-crc-colonoscopy-invitation-benefit-limited-by-participation", "relation": "ranks" }, { "target": "claim-universal-germline-testing-needs-action", "relation": "ranks" }, { "target": "failure-pdac-germline-route-dropout", "relation": "learns-from" }, { "target": "hypothesis-pdac-germline-family-route-completion", "relation": "ranks" } ], "evidence": { "humanProspective": "strong for risk; weak for PDAC prevention action", "humanRetrospective": "strong", "randomised": "strong — related cancers only; none for PDAC prevention", "independentReplication": "strong for major risk factors; weak for named preventive actions", "sampleSize": "Large population cohorts, hereditary surveillance, pancreatitis registries, germline-delivery cohorts and trials, two small diabetes-imaging studies and unfinished EDI n=8,869", "effect": "Risk reduction and enrichment; no intervention prevents most PDAC and no diabetes work-up has a completed utility result", "limits": [ "Observational lifestyle and intervention estimates", "Overlapping attributable fractions", "UK Biobank selection and ancestry", "Surveillance cohorts are specialist and nonrandomized", "Diabetes imaging studies have one cancer each", "Microbiome causality unproved", "Prevention and mortality endpoints require long follow-up", "Germline testing is not itself prevention" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Inherited risk, tobacco, metabolic injury, pancreatitis and precursor evolution each have coherent links to pancreatic cancer." }, "animalEvidence": { "rating": "moderate", "reason": "Models support several causes and interception ideas but do not establish human population benefit." }, "inVitroEvidence": { "rating": "weak", "reason": "Laboratory systems cannot resolve absolute population risk, screening harm or long-term prevention." }, "reproducibility": { "rating": "moderate", "reason": "Major risk directions repeat, while action effects and dangerous-progression selectors do not." }, "commercialConflicts": { "rating": "moderate", "reason": "Most risk evidence is academic, but drugs, tests and vaccines create incentives to overstate prevention readiness." }, "recency": { "rating": "current", "reason": "The ranking includes prospective risk, imaging and registry evidence current through September 2026." } } }, "layout": { "x": 110, "y": 49 }, "updated": "2026-09-14" }, { "id": "prevention-risk-interception-audit", "title": "Prevention audit: risk only matters when it changes a safe action", "shortTitle": "Prevention action map", "type": "system", "status": "supported", "scope": [ "pdac", "pnet", "cholangiocarcinoma", "gallbladder", "hcc", "colorectal", "pan-cancer", "transfer-unproven" ], "summary": "Most PDAC risk markers alter probability without naming a preventable lesion; the useful system must separate ordinary risk, lifetime risk, active premalignancy and probable occult cancer.", "content": "Inherited risk: hereditary syndromes, family history and germline sequencing identify a real minority with BRCA2, ATM, PALB2, CDKN2A, mismatch-repair or pancreatitis-gene routes. Their value is an entry rule for counselling, cascade testing and expert MRI/MRCP plus EUS surveillance, not proof that surveillance reduces mortality. Polygenic risk scores spread small effects across the population and may improve a combined age and clinical model, but do not by themselves raise most people to an invasive-work-up threshold. A rational lifetime-risk group is therefore defined by a validated syndrome, family pattern or combined model with an action attached.\n\nAcquired risk: smoking, obesity, high glucose, diabetes and new-onset diabetes, chronic pancreatitis, heavy alcohol exposure and broader inflammatory states all change risk. Smoking cessation, healthy weight, metabolic care and alcohol-harm reduction are justified by broad health benefit, but no pancreatic-specific randomized trial shows that a named intervention prevents cancer. Cohorts suggesting lower incidence after bariatric surgery or GLP-1 medicines are not prescriptions for pancreatic prevention. Pancreatitis raises relative risk while leaving long-term absolute risk too low for prophylactic pancreatectomy; a Danish lag analysis also shows that much of the high early rate is cancer already present. New-onset diabetes is more useful as a temporary occult-cancer warning than as a lifetime label. It raises three-year risk to 0.62%, yet the 93-person CT pilot found incidental extrapancreatic findings in 56% and its only PDAC was stage IV. PANDOME found one stage-I PDAC among 109 participants. EDI is the first large operating test and remains unfinished. Microbiome associations may mark smoking, inflammation, oral health or tumour effects; without a reproduced causal intervention they do not support antibiotics, probiotics or microbiome transplantation.\n\nPrecursor lesions: IPMN is visible but common and heterogeneous; PanIN is usually microscopic and inaccessible. Cysts can enter surveillance or surgery pathways, but size and ordinary driver mutations do not reliably distinguish the minority that will become lethal. The central prevention experiment is not finding more precursor lesions. It is validating a dangerous-progression state, its time window and a treatment safer than pancreatic surgery. Immune interception with a mutant-KRAS vaccine has shown immune response in 20 high-risk people, not cancer prevention.\n\nCirculating biomarkers include ctDNA and broader cfDNA, methylation, exosomes, proteomics, metabolic signatures and immune signatures. Imaging and organ-specific biological markers can localise or confirm disease. None becomes prevention merely because it detects a signal: it must identify active premalignant change early enough for a safe interceptive action. Current studies often use diagnosed cases, healthy controls or small stored-sample cohorts, so stage-specific sensitivity, specificity, lead time and intervention benefit remain missing.\n\nWorking four-state rule. Ordinary risk: broad prevention only; no pancreatic test. Elevated lifetime risk: inherited, familial or strongly combined risk; expert surveillance and research enrolment. Active premalignant change: a visible lesion or reproducible longitudinal molecular state with defined progression probability; intensified sampling or a trial, not automatic surgery. Probable occult malignancy: a temporary risk trigger plus concordant blood or imaging evidence; rapid pancreas-protocol imaging and specialist confirmation. Movement between states must use absolute risk, not a relative-risk label. At 13.9 cancers per 100,000 person-years, even a perfect-sensitivity 99%-specific test yields about 72 false positives per true cancer. At 0.62% three-year risk, an END-PAC-like rule with 78% sensitivity and 85% specificity would model to about 31 unaffected positives per cancer-bearing positive if its performance transported. Prevention therefore needs both enrichment, a second-stage selector and a low-harm action loop.", "contentSections": [ "Gallbladder prevention has a removable organ but still lacks the future-risk rule that would make optional removal safer than observation. POLYP found one high-grade lesion and no cancer among 68 operations supported by the revised European polyp rule. AMBROSE measured 30-day complications in 8.0%, death in 0.4%, bile leak in 1.3% and severe bile-duct injury in 0.2% across 21,706 operations, so diagnostic yield alone is not enough. GBCseeker and GAIA-MIL classify selected visible lesions; EULAT and Chile BiLS discover candidates; P-iGBC triages pathology after removal. None yet shows that a risk-guided operation prevents more deaths than it causes. A country-specific two-stage rule should release capital only after silent external calibration, complete outcome linkage and explicit operation-harm maths.", "Liver-fluke-related cholangiocarcinoma is the clearest direct prevention route in the related cancers. IARC concludes that chronic Opisthorchis viverrini and Clonorchis sinensis infection causes cholangiocarcinoma. A Thai One Health programme reported infection falling from 6.0% to 0.3%, a 95% relative change, but without randomization or cancer follow-up. eCHEC randomizes 480 people and can test behaviour and stool infection at 12 months. The missing work is long-term linkage from exposure reduction to cancer incidence, stage, surgery and death; behaviour alone cannot stand in for that result." ], "sourceIds": [ "hu-2018-germline-risk", "blackford-2024-high-risk-surveillance", "klein-2024-ukb-prs-lifestyle", "ding-2026-prs-modifiable-risk", "ando-2019-smoking-japan", "bogumil-2024-smoking-cessation-multiethnic", "zhang-2022-uk-biobank-risk-fractions", "schauer-2019-bariatric-pancreatic-risk", "wang-2025-glp1-pancreatic-risk", "kim-2023-pancreatitis-risk-korea", "davidsen-2026-chronic-pancreatitis-mortality", "greenhalf-2020-prss1-surveillance", "chari-2026-prospective-new-onset-diabetes", "sharma-2018-endpac", "wu-2022-nod-imaging-pilot", "frank-2025-pandome", "edi-nct04662879", "fan-2018-oral-microbiome-pdac-risk", "kanda-2012-early-panin-mutations", "haldar-2026-kras-interception", "fahrmann-2021-ca19-9-trajectory", "seer-pancreas-statfacts", "van-dijk-2025-polyp-study", "wong-2025-ambrose-cholecystectomy", "yang-2025-gbcseeker", "gupta-2026-gaia-mil", "clinicaltrials-nct06192719-eulat-gbc", "koshiol-2026-chile-bils", "clinicaltrials-nct06531408-pigbc", "iarc-2012-liver-flukes", "charoensuk-2024-one-health-opisthorchis", "clinicaltrials-nct05321992-echec" ], "links": [ { "target": "prevention-ranking", "relation": "extends" }, { "target": "detection", "relation": "part-of" }, { "target": "claim-germline-risk-subset", "relation": "depends-on" }, { "target": "claim-pancreatitis-high-relative-low-absolute-risk", "relation": "depends-on" }, { "target": "claim-new-onset-diabetes-enriches-risk", "relation": "depends-on" }, { "target": "unknown-dangerous-precursor-selection", "relation": "blocked-by" }, { "target": "claim-kras-interception-vaccine-immune-response-not-prevention", "relation": "constrained-by" }, { "target": "claim-screening-base-rate-maths", "relation": "constrained-by" }, { "target": "claim-pdac-primary-prevention-action-gap", "relation": "contains" }, { "target": "trial-new-onset-diabetes-workup-map", "relation": "contains" }, { "target": "trial-gallbladder-prevention-detection-action-map", "relation": "contains" }, { "target": "hypothesis-gallbladder-two-stage-interception-rule", "relation": "contains" }, { "target": "unknown-gallbladder-prevention-selector", "relation": "blocked-by" }, { "target": "claim-fluke-control-prevents-exposure-not-yet-cancer", "relation": "contains" }, { "target": "failure-echec-cancer-outcome-mismatch", "relation": "constrained-by" } ], "evidence": { "humanProspective": "strong for risk association; weak for interception", "humanRetrospective": "strong", "randomised": "none showing prevention of most PDAC", "independentReplication": "strong for major risk factors; mixed for molecular states", "sampleSize": "Population cohorts, inherited-risk surveillance, pancreatitis registries, two small diabetes-imaging studies, EDI n=8,869, stored-sample biomarkers and a 20-person vaccine interception study", "effect": "Risk can be enriched, but no marker-action pair prevents most PDAC", "limits": [ "Risk factors overlap and cannot be added", "Relative risk can hide low absolute risk", "High-risk surveillance is nonrandomized", "Diabetes work-up utility is unfinished", "Precursor progression is poorly predicted", "Biomarker detection is not an interceptive treatment" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The four states match distinct inherited, inflammatory, precursor and occult-cancer routes observed in people." }, "animalEvidence": { "rating": "moderate", "reason": "Models support precursor and inflammatory biology but cannot validate the proposed human action thresholds." }, "inVitroEvidence": { "rating": "weak", "reason": "Laboratory markers do not establish population selection, work-up harm or prevention." }, "reproducibility": { "rating": "moderate", "reason": "Major risk states repeat, while movement rules and safe interceptive actions remain unproved." }, "commercialConflicts": { "rating": "moderate", "reason": "Genetic, blood and imaging products can benefit from broad risk labels, so each action threshold needs independent control." }, "recency": { "rating": "current", "reason": "The audit includes chronic-pancreatitis and diabetes-work-up evidence current through September 2026." } } }, "layout": { "x": 105, "y": 84 }, "updated": "2026-09-14" }, { "id": "programme-100m-randomized-platform", "title": "£100 million programme — two randomized PDAC decision trials around systemic control", "shortTitle": "£100m trial platform", "type": "programme", "status": "hypothesis", "scope": [ "pdac" ], "summary": "Use existing RAS drugs rather than inventing another, and fund a new residual-disease action trial only if active PDAC studies leave a material decision gap.", "content": "Objective: turn common-RAS suppression into durable systemic control and establish whether measured residual disease should change treatment. Trial A: a 300–450-patient randomized phase 2 platform after initial RAS-inhibitor response. Its first candidate route is now mutant KRAS amplification, reported at progression in 16/44 = 36% of a paired sponsor cohort. Do not assign a combination from that result alone. Release the first £3m of Trial A for independent plasma plus multi-lesion reproduction, early-detection timing and clinical-exposure tests. Randomize a fixed deeper-RAS or other matched combination before radiographic progression only if the class repeats, routing completes within 14 days and coexisting escape is measured. Trial B is conditional, not automatic. DYNAMIC-Pancreas, RADAR-PANC, CIRCPAC and MAP-03 already test adjuvant choice, intensified surveillance and early treatment switching. First spend up to £3m to obtain protocols, live recruitment and event counts, assay definitions, data rights and a common field set across those studies. Open a new 500–700-patient post-resection strategy trial only if none can test the locked mission action with a current comparator, adequate power, overall survival, quality of life and false-positive harm. Otherwise redirect Trial B money into assay cross-calibration, shared follow-up and the missing action within an existing network. Exact drugs depend on safety and access; do not duplicate Revolution Medicines' adjuvant or dual-RAS trials or an active ctDNA action trial. Leadership: O'Reilly for cooperative trials, Park for RAS pharmacology, Iacobuzio-Donahue for resistance, Goggins for assays and an independent chair for allocation and analysis. Budget envelope: £28m Trial A; up to £30m conditional Trial B or existing-trial extension; £9m drug and assay access; £8m serial biopsies, blood and spatial pathology; £6m safety pharmacology and manufacturing support; £6m data, randomization and statistics; £4m organoid and perturbation loop; £3m regulatory and health economics; £2m independent replication; £4m reserve. Total ceiling = £100m, with Trial A expansion and unspent Trial B capital held until their gates pass. Go from safety stage: target engagement in at least 70% of evaluable tumours and no treatment-related mortality signal above a fixed bound. Go to expansion: a prespecified early endpoint with at least 90% probability of a clinically worthwhile effect under the trial model. Stop an arm for failure to reproduce its route, futility, intolerable overlapping toxicity or resistance that the combination does not suppress. Time: 18 months to safety and biology, four to six years to clinical results. Planning outcomes: 15–30% at least one phase-3-ready strategy; 40–55% decisive negative arms; 15–25% inconclusive or obsolete comparator; 5–10% operational failure. Practice-changing efficacy is possible; reliable cure for most patients is not a plausible £100m deliverable.", "contentSections": [ "Hidden-spread staging changes who enters a localized or residual-disease treatment state, but it creates no £100m treatment arm. Record early and same-session camera findings, biopsy, wash cytology and later recurrence as separate fields. A negative result cannot prove absence of systemic disease, and cytology alone cannot assign a treatment without the locked clinical review. The staging experiment is paid from the existing surgical-route budget.", "Partners and institutions: a cooperative PDAC trial network anchored by MSK, Dana-Farber, Johns Hopkins, Penn and at least four NHS, European and Asian centres; an independent statistics centre; patient representatives; trial-grade assay laboratories; and drug partners selected only after pharmacology and data-access terms pass review.", "First experiments: before Trial A opens, run a 40–60-patient pharmacology stage that tests tumour exposure, target engagement, serial blood recovery, biopsy safety and ex-vivo suppression of the dominant escape route. Trial A opens only if at least 70% of evaluable tumours meet the exposure and engagement gate. Before Trial B receives more than £3m, complete the DYNAMIC-Pancreas, RADAR-PANC, CIRCPAC and MAP-03 gap audit and a locked analytical comparison across two laboratories. A new residual-disease randomization opens only for an action, patient state or endpoint that the active trials cannot answer.", "Precursor exclusion gate: this £100m platform does not absorb a new cyst trial. EA2185 no longer answers its original randomized question, but 3P-C results and EA2185 randomized-subset data must be obtained before defining any replacement. A cyst action programme belongs in the separately held detection capital only after the three risk models validate and a recruitment pilot passes.", "RADAR-PANC now joins the £3m post-resection gap audit without adding capital. Before any ctDNA trial claims an imaging benefit, obtain RADAR-PANC's actual 306-person randomization, intervention acceptance, control imaging, survival events, recurrence treatments, quality of life, false-positive procedures, radiation and cost. Its large assumed survival effect may leave a smaller real benefit underpowered. A new scan-schedule trial is prohibited unless RADAR-PANC fails operationally or cannot answer after data recovery.", "Perioperative timing creates no new £100m trial. NCT04340141 A021806 already has 358 actual participants randomized to eight preoperative plus four postoperative mFOLFIRINOX cycles or surgery followed by twelve cycles, with overall survival primary. Obtain route and outcome data through the existing observatory and trial-audit work. Do not launch another broad timing study before its estimated December 2028 primary result. Any later study must name a patient state, regimen or action that A021806 cannot answer.", "No dormancy drug enters the £100m platform from a mouse mechanism or prognostic marrow marker. The £25m observatory must first show a human residual state with ancestry, at least eight weeks lead time, added prediction beyond stage plus ctDNA and a target that can be perturbed without impairing infection control or tissue repair. Until then, anti-NET, CCDC25 and EREG routes remain model hypotheses.", "Every £100m systemic-treatment arm must retain older and vulnerable patients through a common short assessment, exact starting plan, dose, delay, functional change, hospital use and stop reason. The separate 72-hour support experiment is paid from the £15m safety line, so no new £100m branch or capital is added. Until its result, age alone cannot exclude, dose-reduce or assign supportive care, and a favourable toxicity result cannot substitute for tumour control or survival.", "Locally advanced surgery is not silently absorbed into the £100m systemic platform. Current data justify expert reassessment but do not isolate surgery's useful-life effect. The proposed 540-person, £22m randomization can enter only after a platform arm stops or a portfolio review removes at least £22m elsewhere. No current capital is assigned. Until then, use the observatory to retain the response fields needed for eligibility and do not let R0 rate or selected survival release capital.", "No £100m treatment arm may use an organoid recommendation until the existing randomized studies are recovered and the £25m shadow route passes timing, identity, laboratory agreement, drug access and a treatment interaction. The existing £4m organoid loop remains a drug and combination rejection tool. A later 480-person, £18m policy trial is not assigned current capital and must replace at least £18m of stopped or lower-value work.", "FAP-targeted radiation receives no current £100m treatment arm. Human feasibility and a phase 2 dose are known; NCT04939610 already owns non-randomized PDAC response and combination-safety work. A later radioligand-plus-standard-care comparison may enter only after an externally tested lesion dose-response, a prespecified PDAC phase 2 signal, preserved standard-treatment dose intensity and acceptable organ exposure. It must replace equal cost from a stopped or lower-value arm before opening.", "Autophagy receives no current £100m arm. HCQ combinations have failed, and direct-inhibitor results are incomplete. A selective add-on may compete only after complete NCT04892017 recovery, independent model reproduction, paired tumour blockade, preserved RAS-drug dose and a declared four-month-control gate. Any randomized contribution test must replace equal cost from a stopped or lower-value arm.", "Sonoporation receives no current £100m arm. The independent randomized abstract is negative overall and its highlighted chemotherapy subgroup is incomplete. A fixed-backbone test may compete only after full NCT04821284 release, a planned treatment-by-backbone interaction, a validated tumour-delivery measure, preserved chemotherapy dose intensity and acceptable harm. It must replace equal cost from a stopped or lower-value arm.", "SIL204 receives no £100m arm or company investment. Its own 403-person adaptive trial already plans survival testing. Mission participation stops unless Segment 1 publicly shows multi-region tumour knockdown, systemic exposure, acceptable EUS harm and preserved chemotherapy, and unless mutation and sample denominators are locked. Any independent confirmation must replace equal cost from a stopped or lower-value arm.", "No VCN-01 phase 3 enters the £100m platform now. First recover and independently reproduce all-randomized VIRAGE survival and let the sponsor-funded six-person study answer repeat-dose safety. A later blinded trial can enter only after explicit portfolio review, equal capital displacement, commercial product and manufacturing support, public participant-level data and an all-randomized OS endpoint.", "Do not fund a separate MTAP-PRMT5 arm from the £100m platform. Tango and Revolution Medicines own and can finance the products. Demand the missing contribution design: the identical RAS inhibitor alone versus RAS inhibitor plus vopimetostat, with exact MTAP loss, all-randomized PFS and OS, response duration, quality-adjusted survival, toxicity and dose intensity. A package-versus-chemotherapy trial does not answer the component question.", "Reserve no £100m-platform arm for elraglusib now. First audit all NCT03678883 randomized records with £0 added capital. If ITT HR remains at most 0.75 under missingness, site, region and later-treatment tests, a blinded 600-to-650-person phase 3 may compete for the platform only with a solvent majority-funding partner and equal displacement from a lower-ranked branch.", "Do not allocate a platform arm to CDK4/6, CHK1 or WEE1 now. The first two routes failed human comparisons. WEE1 must first clear the existing 12-person tumour and outcome gate for £0. A later same-chemotherapy randomized contribution test can compete only after that result and equal displacement from a lower-value branch." ], "sourceIds": [ "gudmundsdottir-2023-staging-laparoscopy", "rompen-2026-preopanc2-staging", "oreilly-2026-daraxonrasib", "aronchik-2026-daraxonrasib-resistance", "dorbin-2026-allele-specific-daraxonrasib-resistance", "lin-2026-vertical-kras-inhibition", "nct07252232-daraxonrasib-resected", "tie-2022-dynamic-colon", "botta-2024-tumor-informed-ctdna", "varghese-2025-kras-dosage", "dynamic-pancreas-registry", "lee-2024-dynamic-pancreas-feasibility", "clinicaltrials-nct05788744-circpac", "clinicaltrials-nct05802394-map03", "guo-2026-postsurgical-ctdna", "clinicaltrials-nct04239573-ea2185", "clinicaltrials-nct04207944-3pc", "andel-2025-routine-imaging", "van-goor-2025-nationwide-imaging", "daamen-2024-radar-panc-protocol", "clinicaltrials-nct04875325-radar-panc", "tai-2025-pan-asian-esmo-pancreatic", "labori-2024-norpact1", "bai-2025-cispd1", "tan-2026-neoadjuvant-ipd-meta", "ghaneh-2023-espac5", "clinicaltrials-nct04340141-a021806", "effenberger-2012-pdac-bone-marrow-dtc", "nordgard-2022-unresectable-dtc", "chalabi-hajkarim-2025-organotropism", "pei-2025-spatial-metastatic-pdac", "mcclellan-2026-quiescent-pdac-car-t", "tanaka-2012-microscopic-local-recurrence", "dotan-2024-giant", "dotan-2025-giant-ga", "clinicaltrials-nct04233866-giant", "betge-2022-grantpax", "clinicaltrials-nct02143219-pamela70", "mohile-2021-gap70", "li-2021-gain", "ihorst-2026-geriatric-comanagement", "clinicaltrials-nct06040801-frail-gi", "stoop-2026-preopanc4-surgery", "dekker-2026-taps-restaging", "seelen-2026-pelican", "clinicaltrials-nct06714604-duration", "baum-2022-fap2286-firsthuman", "mcconathy-2026-lumiere-phase1", "clinicaltrials-nct04939610-lumiere", "baum-2026-3bp3940", "clinicaltrials-nct05432193-frontier", "clinicaltrials-nct07229768-ctr-fapi", "alistar-2017-devimistat-phase1", "philip-2024-avenger500", "surana-2025-erk-hcq-pdac", "clinicaltrials-nct05221320-ulixertinib-hcq", "manji-2023-mekiauto", "clinicaltrials-nct04214418-mekiauto", "clinicaltrials-nct04892017-inlexisertib", "deciphera-2022-dcc3116-phase1", "clinicaltrials-nct04524702-paricalcitol-hcq", "dimcevski-2016-sonoporation-phase1", "wang-2023-sonochemotherapy-pdac", "clinicaltrials-nct04821284-sonoporation", "ius-2026-sonoporation-nct04821284-abstract", "adler-2026-sonoporation-perfusion", "alagesan-2026-sig12d-loder", "ctis-2026-sil204-phase23", "silexion-2026-sil204-site-initiation", "silexion-2026-halfyear-filing", "richards-2006-ci994-pdac", "heumann-2022-azacitidine-pdac", "sohal-2020-thu-decitabine-pdac", "baretti-2024-entinostat-nivolumab-pdac", "clinicaltrials-nct03250273-entinostat-nivolumab", "safyan-2026-azacitidine-pembrolizumab-pdac", "clinicaltrials-nct03264404-azacitidine-pembrolizumab", "rodon-2024-amg193-prmt5", "matsumoto-2026-mtap-pancreatic", "pavlick-2026-mtap-genomics", "clinicaltrials-nct05732831-vopimetostat", "tango-2025-vopimetostat-monotherapy", "clinicaltrials-nct06922591-vopimetostat-ras", "tango-2026-vopimetostat-ras-data", "tango-2026-q2-10q", "mahalingam-2026-elraglusib-randomized", "clinicaltrials-nct03678883-elraglusib", "pathak-2026-elraglusib-folfirinox", "clinicaltrials-nct05077800-elraglusib-folfirinox", "actuate-2026-q2-10q", "sec-2026-actuate-submissions", "chiorean-2023-abemaciclib-pdac", "baghdadi-2019-palbociclib-cdkn2a", "hidalgo-2022-palbociclib-nabpaclitaxel", "laquente-2017-chk1-pdac", "huffman-2023-ly2880070-pdac", "cuneo-2019-adavosertib-pdac", "huffman-2024-azenosertib-pdac-design", "clinicaltrials-nct06015659-azenosertib" ], "links": [ { "target": "hypothesis-organoid-guided-early-switch", "relation": "gates" }, { "target": "trial-pdac-occult-metastasis-action-map", "relation": "bounded-by" }, { "target": "programme", "relation": "part-of" }, { "target": "programme-25m-prospective-observatory", "relation": "depends-on" }, { "target": "claim-daraxonrasib-common-ras-survival-breakthrough", "relation": "extends" }, { "target": "claim-daraxonrasib-kras-amplification-resistance", "relation": "must-reproduce" }, { "target": "unknown-ctdna-actionability", "relation": "tests" }, { "target": "trial-pdac-ctdna-action-map", "relation": "must-not-duplicate" }, { "target": "trial-pancreatic-precursor-action-map", "relation": "must-not-duplicate" }, { "target": "person-wungki-park", "relation": "led-by" }, { "target": "company-revolution-medicines", "relation": "requires-compound-from" }, { "target": "claim-routine-imaging-after-pdac-resection-not-yet-proven", "relation": "audits" }, { "target": "unknown-postresection-surveillance-utility", "relation": "audits" }, { "target": "trial-pdac-postresection-surveillance-action-map", "relation": "audits" }, { "target": "claim-neoadjuvant-therapy-improves-selection-not-cure", "relation": "audits" }, { "target": "failure-norpact1-neoadjuvant-delivery", "relation": "audits" }, { "target": "unknown-pdac-perioperative-treatment-route", "relation": "audits" }, { "target": "trial-pdac-perioperative-sequencing-action-map", "relation": "audits" }, { "target": "claim-human-pdac-dtc-prognostic-not-dormancy", "relation": "bounded-by" }, { "target": "failure-pdac-dormancy-sample-gap", "relation": "bounded-by" }, { "target": "trial-pdac-residual-cell-state-action-map", "relation": "bounded-by" }, { "target": "hypothesis-human-dormancy-reawakening-gate", "relation": "bounded-by" }, { "target": "claim-age-alone-cannot-route-pdac-treatment", "relation": "bounded-by" }, { "target": "trial-older-pdac-action-map", "relation": "must-not-duplicate" }, { "target": "hypothesis-older-pdac-fast-support-route", "relation": "depends-on" }, { "target": "trial-lapc-conversion-surgery-action-map", "relation": "bounded-by" }, { "target": "unknown-lapc-surgery-causal-benefit", "relation": "leaves-open" }, { "target": "hypothesis-randomized-lapc-resection-after-response", "relation": "could-fund-after-review" }, { "target": "trial-pdac-fap-radioligand-action-map", "relation": "must-not-duplicate" }, { "target": "unknown-pdac-fap-radioligand-treatment-effect", "relation": "leaves-open" }, { "target": "hypothesis-fapi-theranostic-pdac", "relation": "could-fund-after-review" }, { "target": "trial-pdac-autophagy-action-map", "relation": "must-not-duplicate" }, { "target": "unknown-selective-autophagy-pdac-contribution", "relation": "leaves-open" }, { "target": "hypothesis-tumour-verified-selective-autophagy-blockade", "relation": "could-fund-after-review" }, { "target": "trial-pdac-sonoporation-action-map", "relation": "must-not-duplicate" }, { "target": "unknown-pdac-sonoporation-treatment-effect", "relation": "leaves-open" }, { "target": "hypothesis-measured-fixed-backbone-sonoporation", "relation": "could-fund-after-review" }, { "target": "trial-sil204-action-map", "relation": "must-not-duplicate" }, { "target": "unknown-sil204-human-target-delivery", "relation": "leaves-open" }, { "target": "hypothesis-multicompartment-kras-sirna-delivery", "relation": "could-fund-after-review" }, { "target": "failure-broad-epigenetic-treatment-pdac", "relation": "learns-from" }, { "target": "claim-mtap-prmt5-selected-human-signal", "relation": "bounded-by" }, { "target": "failure-vopimetostat-ras-contribution-unknown", "relation": "learns-from" }, { "target": "unknown-mtap-prmt5-ras-contribution", "relation": "leaves-open" }, { "target": "trial-mtap-prmt5-action-map", "relation": "implements" }, { "target": "hypothesis-mtap-prmt5-ras-deeper-control", "relation": "tests" }, { "target": "company-tango-therapeutics", "relation": "contains" }, { "target": "claim-elraglusib-randomized-survival-signal", "relation": "bounded-by" }, { "target": "failure-elraglusib-analysis-and-survival-discordance", "relation": "learns-from" }, { "target": "unknown-elraglusib-survival-replication", "relation": "leaves-open" }, { "target": "trial-elraglusib-action-map", "relation": "implements" }, { "target": "hypothesis-elraglusib-survival-confirmation-gate", "relation": "tests" }, { "target": "company-actuate-therapeutics", "relation": "contains" }, { "target": "failure-broad-cell-cycle-checkpoint-pdac", "relation": "learns-from" }, { "target": "unknown-wee1-replication-stress-pdac", "relation": "leaves-open" }, { "target": "trial-cell-cycle-checkpoint-action-map", "relation": "implements" }, { "target": "hypothesis-human-replication-stress-selector", "relation": "tests" } ], "evidence": { "humanProspective": "not-applicable", "humanRetrospective": "not-applicable", "randomised": "not-applicable", "independentReplication": "none", "sampleSize": "Proposed Trial A n=300-450; Trial B n=500-700", "effect": "Designed to select phase 3 strategies and test a treatment rule", "limits": [ "Generated programme design", "Drug access not negotiated", "Trial designs require current event-rate modelling", "Comparator may change", "Costs and probabilities are planning estimates", "No guarantee of enough eligible patients", "Active ctDNA trials may make Trial B unnecessary or require a different action" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "Human RAS benefit and residual-disease risk support both trials, but neither treatment rule is validated." }, "animalEvidence": { "rating": "weak", "reason": "Model evidence informs candidate combinations but does not prove benefit from early resistance action." }, "inVitroEvidence": { "rating": "moderate", "reason": "Perturbation models can reject combinations, but they have not prospectively selected these trial arms." }, "reproducibility": { "rating": "none", "reason": "The proposed resistance-directed and residual-disease action strategies have not been reproduced." }, "commercialConflicts": { "rating": "high", "reason": "Compound, assay and combination rights are held by commercial partners whose incentives may shape access." }, "recency": { "rating": "current", "reason": "The platform is designed around 2026 RAS evidence and active postoperative trials." } } }, "layout": { "x": 54, "y": 118 }, "updated": "2026-09-15" }, { "id": "programme-25m-prospective-observatory", "title": "£25 million programme — a prospective PDAC residual-disease and resistance observatory", "shortTitle": "£25m observatory", "type": "programme", "status": "hypothesis", "scope": [ "pdac" ], "summary": "A four-year, multi-centre cohort would measure disease before, during and after treatment and produce selectors for intervention trials rather than another retrospective signature.", "content": "Objective: follow 800 people across resectable, locally advanced and metastatic PDAC with fixed blood, imaging and tissue times so residual disease and RAS resistance can be seen rather than inferred. The resistance lane now has a specific independent test: reproduce the sponsor-reported treatment-emergent RAS-pathway rate of 26/44 = 59% and mutant KRAS-amplification rate of 16/44 = 36%, determine how early amplification appears, and test whether plasma agrees with more than one progressing lesion. Leadership: Iacobuzio-Donahue for evolution, O'Reilly for the trial network, Chari and Goggins for detection and samples, Tuveson for functional models, plus independent statistics and health economics. Partners: MSK, Johns Hopkins, MD Anderson, CSHL, Penn and at least three NHS or European centres. Budget: £5.0m recruitment and clinical sampling; £4.0m molecular assays and biobanking; £3.0m imaging and opportunistic-CT analysis; £3.0m rapid biopsy and spatial pathology; £2.5m organoid and perturbation work; £2.0m data platform and external validation; £1.5m independent statistics and trial design; £1.5m access to compounds, assays and intellectual property; £1.0m patient safety, ethics and health economics; £1.5m reserve. Total = £25.0m. The first 200 patients form a locked feasibility stage. Go: over 75% completion of required serial samples, under 10% serious biopsy complications, independent reproduction of at least one decision-relevant resistance class, externally reproducible marker performance and at least one result that changes a prespecified treatment or trial choice. Stop or narrow: missingness destroys longitudinal inference, the 36% amplification class does not reproduce within a prespecified bound, multi-lesion sampling contradicts the plasma route, the selector fails external validation, or proposed biopsies do not alter a decision. Time: two years to the first locked dataset, four years to recurrence and resistance results. Planning outcomes: 25–40% one validated trial selector; 35–50% decisive negative or narrowing result; 15–25% ambiguous; 5–10% operational failure. It cannot establish population screening benefit or phase 3 efficacy.", "sourceIds": [ "oreilly-2026-daraxonrasib", "aronchik-2026-daraxonrasib-resistance", "dorbin-2026-allele-specific-daraxonrasib-resistance", "botta-2024-tumor-informed-ctdna", "blackford-2024-high-risk-surveillance", "aznar-2025-mesothelin-car-t-resistance", "effenberger-2012-pdac-bone-marrow-dtc", "nordgard-2022-unresectable-dtc", "chalabi-hajkarim-2025-organotropism", "pei-2025-spatial-metastatic-pdac", "mcclellan-2026-quiescent-pdac-car-t", "tanaka-2012-microscopic-local-recurrence", "alistar-2017-devimistat-phase1", "philip-2024-avenger500", "surana-2025-erk-hcq-pdac", "clinicaltrials-nct05221320-ulixertinib-hcq", "manji-2023-mekiauto", "clinicaltrials-nct04214418-mekiauto", "clinicaltrials-nct04892017-inlexisertib", "deciphera-2022-dcc3116-phase1", "clinicaltrials-nct04524702-paricalcitol-hcq", "dimcevski-2016-sonoporation-phase1", "wang-2023-sonochemotherapy-pdac", "clinicaltrials-nct04821284-sonoporation", "ius-2026-sonoporation-nct04821284-abstract", "adler-2026-sonoporation-perfusion", "alagesan-2026-sig12d-loder", "ctis-2026-sil204-phase23", "silexion-2026-sil204-site-initiation", "silexion-2026-halfyear-filing", "richards-2006-ci994-pdac", "heumann-2022-azacitidine-pdac", "sohal-2020-thu-decitabine-pdac", "baretti-2024-entinostat-nivolumab-pdac", "clinicaltrials-nct03250273-entinostat-nivolumab", "safyan-2026-azacitidine-pembrolizumab-pdac", "clinicaltrials-nct03264404-azacitidine-pembrolizumab", "rodon-2024-amg193-prmt5", "matsumoto-2026-mtap-pancreatic", "pavlick-2026-mtap-genomics", "clinicaltrials-nct05732831-vopimetostat", "tango-2025-vopimetostat-monotherapy", "clinicaltrials-nct06922591-vopimetostat-ras", "tango-2026-vopimetostat-ras-data", "tango-2026-q2-10q", "mahalingam-2026-elraglusib-randomized", "clinicaltrials-nct03678883-elraglusib", "pathak-2026-elraglusib-folfirinox", "clinicaltrials-nct05077800-elraglusib-folfirinox", "actuate-2026-q2-10q", "sec-2026-actuate-submissions" ], "links": [ { "target": "hypothesis-organoid-guided-early-switch", "relation": "tests" }, { "target": "trial-pdac-organoid-action-map", "relation": "uses" }, { "target": "programme", "relation": "part-of" }, { "target": "programme-5m-decision-core", "relation": "extends" }, { "target": "residual-disease", "relation": "targets" }, { "target": "unknown-ctdna-actionability", "relation": "resolves" }, { "target": "claim-daraxonrasib-kras-amplification-resistance", "relation": "must-reproduce" }, { "target": "person-eileen-oreilly", "relation": "led-by" }, { "target": "person-michael-goggins", "relation": "led-by" }, { "target": "claim-human-pdac-dtc-prognostic-not-dormancy", "relation": "tests" }, { "target": "failure-pdac-dormancy-sample-gap", "relation": "tests" }, { "target": "trial-pdac-residual-cell-state-action-map", "relation": "tests" }, { "target": "hypothesis-human-dormancy-reawakening-gate", "relation": "tests" }, { "target": "unknown-selective-autophagy-pdac-contribution", "relation": "measures" }, { "target": "hypothesis-tumour-verified-selective-autophagy-blockade", "relation": "gates" }, { "target": "trial-pdac-autophagy-action-map", "relation": "constrained-by" }, { "target": "unknown-pdac-sonoporation-treatment-effect", "relation": "measures" }, { "target": "hypothesis-measured-fixed-backbone-sonoporation", "relation": "gates" }, { "target": "trial-pdac-sonoporation-action-map", "relation": "constrained-by" }, { "target": "unknown-sil204-human-target-delivery", "relation": "measures" }, { "target": "hypothesis-multicompartment-kras-sirna-delivery", "relation": "gates" }, { "target": "trial-sil204-action-map", "relation": "constrained-by" }, { "target": "failure-broad-epigenetic-treatment-pdac", "relation": "learns-from" }, { "target": "claim-mtap-prmt5-selected-human-signal", "relation": "bounded-by" }, { "target": "failure-vopimetostat-ras-contribution-unknown", "relation": "learns-from" }, { "target": "unknown-mtap-prmt5-ras-contribution", "relation": "leaves-open" }, { "target": "trial-mtap-prmt5-action-map", "relation": "implements" }, { "target": "hypothesis-mtap-prmt5-ras-deeper-control", "relation": "tests" }, { "target": "company-tango-therapeutics", "relation": "contains" }, { "target": "claim-elraglusib-randomized-survival-signal", "relation": "bounded-by" }, { "target": "failure-elraglusib-analysis-and-survival-discordance", "relation": "learns-from" }, { "target": "unknown-elraglusib-survival-replication", "relation": "leaves-open" }, { "target": "trial-elraglusib-action-map", "relation": "implements" }, { "target": "hypothesis-elraglusib-survival-confirmation-gate", "relation": "tests" }, { "target": "company-actuate-therapeutics", "relation": "contains" } ], "evidence": { "humanProspective": "not-applicable", "humanRetrospective": "not-applicable", "randomised": "not-applicable", "independentReplication": "none", "sampleSize": "Proposed n=800, with a locked n=200 feasibility stage", "effect": "Decision-enabling cohort; no direct efficacy claim", "limits": [ "Generated programme design", "Costs are planning estimates", "Observational cohort cannot establish treatment benefit", "Multi-centre data agreements are untested", "Probability ranges are judgement" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "Recurrence and RAS escape are observed in people, but a stable decision rule has not been shown." }, "animalEvidence": { "rating": "weak", "reason": "Animal systems support some escape mechanisms but cannot establish a human longitudinal selector." }, "inVitroEvidence": { "rating": "moderate", "reason": "Human-derived models can test sampled resistance states, with uncertain clinical prediction." }, "reproducibility": { "rating": "none", "reason": "No network has reproduced the proposed fixed-timepoint PDAC observatory and its treatment rule." }, "commercialConflicts": { "rating": "high", "reason": "Drug, assay and trial-data access can create sponsor control unless contracts protect independent analysis." }, "recency": { "rating": "current", "reason": "The design is tied to current common-RAS, ctDNA and surveillance results." } } }, "layout": { "x": 51, "y": 117 }, "updated": "2026-09-15", "contentSections": [ "A 250-person resected residual-cell subset now uses the existing £4m assay and biobank, £3m rapid-biopsy and spatial-pathology, and shared recruitment lines; it adds no capital to the £25m total. Bank primary tissue, keep fixed blood and imaging, obtain safe recurrence tissue before new treatment and connect rapid autopsy. Marrow is optional. Go only with at least 25 pre-imaging cases, ancestry support in 80% of pairs, two-laboratory agreement in 80%, eight weeks lead time and improvement beyond stage plus ctDNA.", "The existing £2.5m organoid and perturbation line now owns a 200-person shadow route, including 50 split samples. It records tissue adequacy, tumour identity, culture failure, result time, top-two regimen order, current drug access and later outcome. At least 80% must return a tumour-confirmed result within 14 days and two laboratories must agree in at least 80%; failure retires patient selection while preserving organoids for drug rejection. No new capital is added.", "Selective-autophagy work adds no cohort or budget to the £25m observatory. Existing biopsy, organoid and resistance-pharmacology lines may define a paired tumour assay and independently test combination killing at human-achievable exposure. They do not expose patients until NCT04892017 data are recovered and two laboratories reproduce the predicted effect. No new capital is added.", "Sonoporation adds no cohort or budget to the £25m observatory. Existing imaging and pharmacology teams may lock a safe direct or validated tumour-drug delivery measure only after the NCT04821284 planned interaction passes. Perfusion-image direction alone does not qualify. No new participant exposure or capital is added.", "SIL204 adds no cohort or budget to the £25m observatory. Existing biopsy and resistance-pharmacology teams may validate the multi-region assays under a public-data contract, but the sponsor trial supplies all patient exposure. Plasma concentration, an injected-region result or first-generation subgroup survival cannot pass the human target gate.", "VCN-01 is outside the £25m observatory and adds no cohort, biopsy or budget line. Its mechanism samples and survival data already exist under sponsor control. Recover those first; do not use observatory participants to rebuild a company dataset.", "MTAP-selected PRMT5 work adds no cohort or budget. The route is biologically and commercially credible, but Tango and its partners already fund a 225-person monotherapy programme and a 183-person combination programme. Require all 54 treated PDAC records to mature, exact MTAP-loss assays and a randomized RAS-inhibitor-alone contribution arm. Existing mission teams may review assays and analysis only under public-data terms.", "Do not put elraglusib into the prospective observatory or recruit another cohort. Its immediate question is already contained in 269 randomized records. The observatory may supply assay comparisons only after the survival result survives independent analysis; it does not absorb Actuate's financing need." ] }, { "id": "programme-500m-integrated-mission", "title": "£500 million programme — integrate systemic control, residual disease and risk-triggered detection", "shortTitle": "£500m integrated mission", "type": "programme", "status": "hypothesis", "scope": [ "pdac", "pnet", "cholangiocarcinoma", "gallbladder", "hcc", "colorectal", "transfer-unproven" ], "summary": "Run linked systemic-control and residual-disease programmes now; use colorectal evidence as an action check, and reserve large trials until a distinct gap is proved.", "content": "Objective: create at least one intervention that reduces PDAC death, not merely a new marker or response rate. Budget: £160m for two global randomized systemic-control trials built around RAS inhibition and resistance; £85m for a residual-disease action platform across resected PDAC and a colorectal reference cohort; £105m reserved for risk-triggered detection, of which only £5m releases first for EDI and PANDOME data access, cross-site analysis, assay comparison and second-stage selector validation while the other £100m is held; £45m for longitudinal samples, spatial pathology, organoids and perturbation; £25m for diagnostic and trial-grade manufacturing; £20m for data, independent statistics and regulatory work; £20m for pNET, biliary, gallbladder and HCC transfer tests that must meet disease-specific gates; £15m for patient safety, access and health economics; £25m reserve. Total ceiling = £500m. EDI already enrolled 8,869 in a non-randomized site comparison and is expected to complete in 2028. A CT pilot found extrapancreatic incidental findings in 52 of 93 participants and its only PDAC was stage IV; PANDOME found one stage-I PDAC among 109 participants after four biopsies. Do not pay for a 40,000–60,000-person duplicate until complete operating data show the uncovered decision. Release the held £100m only if a locked second-stage selector validates outside its development systems, the new trial uses individual or cluster randomization, both arms have complete negative follow-up, and the action can measure aggressive stage-I/II yield, interval cancers, incidental procedures, quality of life, cost and pancreatic-cancer death. Programme chair: an independent operator without a product stake. Scientific table: Iacobuzio-Donahue, Chari, Goggins, Tuveson, O'Reilly, Park and Vonderheide, joined by radiology AI, statistics, manufacturing and patient leaders. Company partners may include Revolution Medicines, Astellas, GRAIL and assay providers, but data rights, negative-result publication and independent analysis are mandatory. Three-year gates: reproduce resistance classes; validate an MRD action threshold; obtain complete EDI fields and reproduce a second-stage detection rule. Five-year gates: at least one randomized phase 2 effect large enough for phase 3 and no negative net-benefit signal in detection. Stop the detection lane if site adjustment removes stage shift, interval cancers erase sensitivity, correlated test errors destroy the modelled workload or serious work-up harm per additional treatable cancer crosses the locked limit. Eight-to-ten-year gate: improvement in cancer-specific survival or a validated surrogate backed by later mortality. Planning outcomes: 30–50% at least one practice-changing result; 30–45% only decisive negative knowledge; 10–20% mixed or obsolete results; 5–10% operational failure. Chance of making PDAC reliably controllable for most patients within ten years remains below 10–15% because biology and time, not only money, constrain the result. The £15m patient-safety, access and health-economics line now has named internal gates: £3m for the PDAC clot route, £4m for biliary route completion, £2.5m for the English germline-to-family study, £2.5m for surgical-route completion and £3m held for later safety and access gaps. £3m + £4m + £2.5m + £2.5m + £3m = £15m. Germline money pays for 3,000 consecutive patient tests, R367-stratified action and cost analysis, and a family-cluster randomized contact test; it does not pay for another selected-volunteer uptake report or treat a test as prevention.", "contentSections": [ "ASPIRE data diligence creates no new budget line. At most £250,000 comes from the existing £20m data and independent-statistics line to verify legal custody, assignment keys, raw records and follow-up. No rescue invoices, ivospemin licence, equity or new cohort are funded before at least 400 assignments reconcile and 90% 12-month vital status is recoverable without a greater than ten-point arm missingness gap. Any later rescue needs itemized costs and equal displacement elsewhere. £500m + £0 expansion = £500m.", "The hidden-spread staging test is nested inside the existing £2.5m English surgical-route allocation. Ring-fence £0.4m for a locked protocol, pathology fields and independent analysis; £2.1m remains for route ownership, recovery and treatment completion. The first 500 intended operations compare current NICE-selected practice with the two-timepoint route during the stepped rollout. This is a reallocation within £2.5m, not a new line: the patient-safety total remains £15m, its unassigned balance remains £0, and the mission remains £500m.", "The £5m first-release audit now covers both risk-triggered work-up and pancreatic precursors. It must obtain complete EDI and PANDOME operating fields, the EA2185 amendment and randomized-subset data, 3P-C component outcomes, SIPP-T3 accrual reasons, and an independent feasibility set for current-disease cyst classifiers. This does not increase the release: it reduces duplicate spending by resolving existing evidence first.", "The held £100m can fund at most one detection action branch initially. A precursor branch competes with the new-onset-diabetes branch on expected preventable deaths per pound, serious procedures per additional HGD or stage-I cancer, and time to a mortality answer. The precursor branch cannot release more than £40m for its 6,000-person validation cohort until the three models—current advanced disease, future cyst progression and separate PDAC—validate outside their development systems. No intervention expansion follows unless at least 25% of positive operations contain HGD or invasive cancer and major procedure harm stays below one per additional target lesion found.", "Gallbladder prevention uses the existing £20m related-cancer line and reserve, not the held PDAC detection tranche. Release £5m to align EULAT, Chile BiLS, P-iGBC and North Indian fields and test existing imaging and blood models silently. Up to £15m more can leave the reserve for a 40,000-person country-stratified validation only if outcome linkage exceeds 80%, local calibration error stays below five points and the modelled operation harm leaves positive net benefit. Do not fund another stand-alone cohort, universal polyp cutoff or enriched diagnostic AUC. A later mortality trial requires a separate decision after the validation stage.", "Small-pNET action selection also uses the related-cancer line. Release £3m to obtain complete ASPEN and PANDORA data, resolve ASPEN's unknown registry status, define a common five-year progression endpoint and compare a size-plus-duct rule with a time-updated model that includes procedure-specific harm. This leaves £12m of the £20m line for cholangiocarcinoma and HCC work after the £5m gallbladder and £3m pNET releases. Conditional gallbladder and pNET expansions compete for no more than £15m total from the £25m reserve; they cannot both receive their full modelled amount. pNET expansion requires calibration error below five points, outcome linkage above 85% and a modelled reduction of at least 25% in major treatment harm without more than 1.5 additional metastatic progressions per 100 at five years.", "Cholangiocarcinoma receives £6m of the remaining related-cancer line, leaving £6m for HCC. The first stage joins 6,000 PSC records, 10,000 endemic-area prevention participants and 600 resected biliary cancers under one field set but three separate analyses. It does not start three new trials. It measures whether existing networks can complete the route from exposure or risk through action, harm and survival. PSC expansion stops above ten benign invasive procedures per additional resectable target or below 80% action completion. Liver-fluke expansion stops below 50% infection reduction at 24 months or without long-term cancer linkage. Residual-disease expansion stops below 90% panel success, below 85% result return before action or without a safe named treatment. At most one passing CCA route may receive up to £6m from the £10m reserve left after the capped gallbladder and pNET competition; this preserves the £500m ceiling.", "HCC receives the final £6m in the related-cancer line. It does not fund another GALAD or MRI superiority trial while TRACER and PREMIUM remain viable. It funds a three-year randomized readiness stage in 240 patients at 12 centres for active liver-reserve care during local or systemic tumour treatment. Required actions must be delivered in at least 85%; on-time cancer treatment and dose intensity may fall by no more than five percentage points; excess 90-day death or decompensation stops the stage; and there must be measurable separation in ALBI or action completion. The stage is not powered for mortality. If it passes, a 600-person outcome trial needs a further £12m-£19m. No such expansion is assigned inside the current ceiling: it can occur only after an explicit portfolio review stops or reduces another branch by the same amount. £500m + £0 reallocation = £500m; adding an unfunded expansion would break the budget and is prohibited.", "Within the existing £85m residual-disease line, cap the colorectal reference work at £5m. It may obtain patient-level DYNAMIC, DYNAMIC-III, ALTAIR and FIND data, compare assay states, mature FIND survival follow-up and align fields with CIRCULATE-US. It may not start another general prognostic cohort, repeat trifluridine/tipiracil, or duplicate ctDNA-negative omission and ctDNA-positive mFOLFIRINOX while NCT05174169 remains viable. Any commercial assay contract must freeze the assay and algorithm version, return raw and derived data, guarantee turnaround and supply continuity, permit negative-result publication and define a validated backup. The other £80m remains for the PDAC action platform and is still held behind its own gates. £5m + £80m = £85m; no new programme money is created.", "Cachexia creates no new budget line and no duplicate pivotal trial. NCT06989437 already tests ponsegromab versus placebo with first-line chemotherapy in an estimated 982 metastatic-PDAC patients. Use the existing £20m data and independent-statistics line to seek arm-level data, lock the endpoint hierarchy and audit GDF-15, function, missing wearables, chemotherapy dose changes, quality-adjusted survival and OS. Weight alone cannot release new capital. A follow-on study can compete for reserve only if the sponsor trial fails operationally, denies usable data or leaves a named frail, adjuvant or low-resource group uncovered; it must then show at least a ten-point gain in treatment delivery or an equally important patient outcome without excess severe harm. This preserves £500m: the cachexia audit is inside the existing data line, not £500m plus another programme.", "Pancreatic enzyme work also creates no new budget line and no duplicate efficacy trial. PERTseverance already randomizes an estimated 100 selected patients to immediate or delayed Creon; NCT05642962 compares two doses in 17 patients; PANCAX-3 provides single-arm feasibility data. Use the same existing £20m data and independent-statistics line to obtain patient-level fields and join symptoms, enzyme-function evidence, dose, adherence, pharmacy supply, chemotherapy delivered, quality of life and survival. Weight alone cannot release capital. A new trial can compete for reserve only if the existing studies cannot estimate the selected treatment effect and routine-care fields show a repairable gap. The first randomized stage must then show at least a ten-point gain in patients receiving 80% of planned chemotherapy or a gain in quality-adjusted survival, without serious treatment harm. The total remains £500m.", "PDAC clot prevention uses £3m of the existing £15m patient-safety, access and health-economics line, leaving £12m. Do not fund another conventional anticoagulant-versus-no-treatment efficacy trial: five randomized studies already estimate an eight-point absolute VTE reduction and CONKO-004 found no survival gain. First recover imPaCT-PRO data and refresh CATIQ, PREPOSTEROUS Pancreas and PINPOINT. The funded experiment is an agent-neutral seven-day assessment-to-use route across about 30 clinics and 2,500 people. Release requires at least 80% documented decisions, prophylaxis use up 30 points and symptomatic VTE down three points. Stop for major bleeding up two points, fatal-bleeding imbalance, no symptomatic reduction, incidental-only benefit or drug access failure above 10%. No Quercis investment or licence follows marker data; any Kinisoquin decision requires the exact product, licence, manufacture, randomized event data and fair generic comparator.", "Biliary-route work uses £4m of the £12m left in the existing patient-safety, access and health-economics line after clot prevention, leaving £8m. Do not fund another broad metal-versus-plastic or EUS-versus-ERCP trial. The funded roughly 1,300-person, 30-centre stepped study tests the whole route: drainage indication within 48 hours, prompt surgery when drainage is avoidable, durable metal stenting through neoadjuvant treatment, expert rescue after failed ERCP, infection response, reintervention and every treatment date. Release requires at least 90% documented plans, at least 90% metal use when drainage must span treatment and interruption down from about 16% to 10%. Stop for procedure harm up five points, death imbalance, surgery delayed over seven additional days, no interruption reduction or no gain in treatment delivered. Device contracts require independent statistics, adjudication, patient-level data, traceability, supply continuity and negative-result publication. After this step, £8m remained.", "Germline-to-family work uses £2.5m of the £8m left in the patient-safety, access and health-economics line, leaving £5.5m. Enrol 3,000 consecutive newly diagnosed PDAC patients at about 20 NHS centres, offer point-of-care public-laboratory testing, and report pathogenic yield, time, cost and documented action inside versus outside R367. Families with a pathogenic result enter a family-cluster randomized test of the standard letter and free link versus consented direct multilingual contact, a named navigator and a free home kit. Release requires at least 85% patient results within 21 days, at least 50% of eligible reached relatives tested and positive net benefit outside R367. Stop for privacy breach, action on a variant of uncertain significance, inequality above ten points, slow results or no policy value. £3m clot route + £4m biliary route + £2.5m germline route + £5.5m remainder = £15m.", "Surgical-route work uses £2.5m of the £5.5m left in the patient-safety, access and health-economics line, leaving £3m. All 23 English HPB centres record every potentially curable scan and planned route; one pilot centre fixes the fields and the other 22 cross to a named-owner, joined surgery-oncology, recovery and proficiency bundle in 11 matched pairs. Planning starts at 1,200 people and can rise to 1,700 after a masked 300-person cluster check. Release requires at least 95% route and dose capture, postoperative treatment within 14 weeks rising from 62% to at least 72%, full-course completion up ten points, no more than a one-point 90-day mortality rise and no more than a five-point serious-complication rise. Stop above 5% mortality, for unexplained operation-rate loss or a ten-point inequality widening. No money buys a robot. The final safety-line arithmetic is £3m clot + £4m biliary + £2.5m germline + £2.5m surgical route + £3m remainder = £15m.", "Post-resection surveillance creates no new budget line. NCT04875325 RADAR-PANC already tests three-monthly CA19-9 plus CT against mainly symptom-led care, with overall survival primary, and NCT05788744 CIRCPAC tests ctDNA-directed scan intensity. Use the existing £3m audit inside the £100m randomized-platform plan to obtain actual accrual, crossover, survival, quality of life, recurrence treatment, false-positive procedures, radiation and cost. No new broad imaging trial releases while either existing study can answer the chosen action. The total remains £500m.", "Perioperative sequencing creates no new budget line. Use the existing £2.5m surgical-route programme and the £25m observatory's data work to align A021806, PREOPANC, NORPACT-1 and CISPD-1 fields: anatomy, assigned first treatment, every dose, severe harm, progression, operation, pathology, postoperative treatment, recurrence, quality of life and survival. A021806 is already externally funded and accrued; the atlas pays no duplicate trial. The £15m safety-line arithmetic and £500m total do not change.", "The human residual-cell gate creates no new budget line. Its 250-person subset is inside the existing £25m observatory, using already allocated assay, biobank, biopsy and spatial-pathology funds. No anti-dormancy intervention draws from the £100m platform until ancestry, timing, cross-laboratory agreement, added prediction and safety all pass. The total remains £500m.", "The older-and-vulnerable PDAC route uses the final £3m in the existing £15m patient-safety, access and health-economics line, leaving £0 unassigned there. Release £0.5m for a six-site, 150-person run-in; hold £2.5m until assessment-before-plan, 72-hour action, delay and data-completion gates pass. Expansion targets 900 treated participants at about 30 clinics and tests severe toxicity plus days alive outside hospital. Do not fund another age-defined drug comparison. Final arithmetic: £3m clot + £4m biliary + £2.5m germline + £2.5m surgical route + £3m older-and-vulnerable route = £15m. The total remains £500m.", "Conversion surgery creates a high-value but unfunded causal question. Use the existing observatory and English route to capture initial anatomy, treatment, response, expert reassessment, staging and outcomes now. A 540-person surgery-versus-continued-treatment trial has a planning cost of £22m and cannot be paid from the fully assigned £15m safety line or silently added to the £500m ceiling. It may start only after an explicit review removes at least £22m from a stopped, failed or lower-value branch. Until then, the mission offers reassessment and records the decision without claiming surgery caused selected survival. £500m + £0 current release = £500m.", "Organoid route work creates no new budget line. Data recovery uses the £5m decision core, and the 200-person shadow test uses the existing £2.5m organoid line inside the £25m observatory. The £4m organoid loop inside the £100m platform remains for drug rejection. A 480-person utility trial has a planning cost of £18m but receives £0 now; it can start only after an explicit review removes at least £18m elsewhere. £500m + £0 current release = £500m.", "FAP-radioligand work creates no new budget line. Existing data-contract and imaging work requests LuMIERE lesion dose-response data and FRONTIER results; the active LuMIERE programme supplies its own treatment exposure. Scan brightness, stable disease and mixed-cancer response cannot release capital. Any later randomized PDAC treatment trial receives £0 now and can enter the £500m ceiling only after an explicit review removes at least the same amount elsewhere. £500m + £0 current release = £500m.", "Autophagy creates no new budget line. Data recovery uses existing contracts and pharmacology work; no HCQ combination or replacement treatment cohort receives money now. A future selective-inhibitor measurement study and randomized add-on trial may enter only after the tumour, dose-preservation and early-control gates pass and an explicit review removes at least the same cost elsewhere. £500m + £0 current release = £500m.", "Sonoporation creates no new budget line. Full NCT04821284 release and interaction analysis use existing data-contract work; no device cohort, licence or company investment receives money now. A future fixed-backbone measurement study may enter only after the planned interaction and tumour-delivery gates pass and an explicit review removes at least the same cost elsewhere. £500m + £0 current release = £500m.", "SIL204 creates no new budget line, trial or equity position. The live sponsor programme supplies exposure; existing data and pharmacology teams can help only under a public contract and before Segment 2. Human multi-region knockdown, systemic coverage, chemotherapy preservation and safety must all pass. Any later independent test requires equal portfolio removal. £500m + £0 current release = £500m.", "The VCN-01 review adds £0, so the £500m ceiling and every existing programme total remain unchanged. Data recovery is the current action. A future phase 3 would require a separate displacement decision and a commercial partner funding most product, manufacturing and trial cost.", "The MTAP-PRMT5 review adds £0, leaving the £500m ceiling unchanged. Tango's $1.0bn disclosed balance and active programmes fund the next proof. Mission leverage is limited to exact selection, complete data and a randomized component design; any later mission trial requires equal portfolio displacement.", "The elraglusib review adds £0 and leaves the £500m ceiling unchanged. The phase 2 ITT survival signal merits immediate independent data recovery, but Actuate's 2.82-month simple June cash-to-burn ratio does not justify rescuing the company. A later blinded phase 3 can enter only after the full-data gate, a manufacturing and majority-funding partner, public negative-result rights and equal portfolio displacement.", "The TIGeR-PaC review adds £0 in trial and company capital and leaves the £500m ceiling unchanged. Enrollment is closed and the event-driven result is expected in 2027. At most £250,000 from the existing data and statistics line may support independent all-randomized analysis after lock and access. Package adoption, catheter mechanism and transfer to another drug or cancer remain three separate decisions." ], "sourceIds": [ "gudmundsdottir-2023-staging-laparoscopy", "rompen-2026-preopanc2-staging", "nice-2018-ng85-pancreatic", "oreilly-2026-daraxonrasib", "chari-2026-prospective-new-onset-diabetes", "sharma-2018-endpac", "wu-2022-nod-imaging-pilot", "frank-2025-pandome", "edi-nct04662879", "ohtsuka-2024-bd-ipmn-surveillance", "singhi-2026-pancreaseq-gc", "henn-2023-cyst-surgery-registry", "weinberg-2020-ea2185-protocol", "clinicaltrials-nct04239573-ea2185", "clinicaltrials-nct04207944-3pc", "clinicaltrials-nct06519097-sipp-t3", "clinicaltrials-nct06192719-eulat-gbc", "clinicaltrials-nct06531408-pigbc", "koshiol-2026-chile-bils", "yang-2025-gbcseeker", "gupta-2026-gaia-mil", "wong-2025-ambrose-cholecystectomy", "tie-2022-dynamic-colon", "sasieni-2026-nhs-galleri", "baudin-2026-oclurandom", "partelli-2022-aspen-interim", "clinicaltrials-nct03084770-aspen", "heidsma-2021-pandora-pnet", "maekawa-2026-small-pnet-surgery-morbidity", "villard-2023-psc-surveillance", "charoensuk-2024-one-health-opisthorchis", "clinicaltrials-nct05321992-echec", "yu-2025-btc-ctdna-recurrence", "yoo-2025-extrahepatic-cca-ctdna", "clinicaltrials-nct02170090-acticca1", "lim-2025-attention-hbv", "trinchet-2011-hcc-surveillance-interval", "hirode-2026-hcc-biomarker-rct", "rhee-2025-miracle-hcc", "clinicaltrials-nct06084234-tracer", "clinicaltrials-nct05486572-premium", "tie-2025-dynamic3", "bando-2026-altair", "mo-2026-find-crc-surveillance", "clinicaltrials-nct05174169-circulate-us", "ruers-2017-clocc-crc-liver-ablation", "adam-2024-transmet", "mulcahy-2021-epoch-radioembolization", "natera-2025-10k", "patent-wo2019200228-natera-mrd", "woo-2016-pert-rct", "saito-2018-pert-rct", "picozzi-2025-pert-advanced-pdac", "dugic-2026-pert-mortality", "clinicaltrials-nct04098237-pancax3", "clinicaltrials-nct05642962-pertzye-dose", "clinicaltrials-nct06099119-pertseverance", "lemanska-2023-opensafely-pert-audit", "dhsc-2024-pert-shortage-alert", "maraveyas-2012-fragem", "pelzer-2015-conko004", "vadhan-raj-2020-cassini-pdac", "frere-2020-pdac-thromboprophylaxis-meta", "heffley-2024-pdac-vte-uptake", "noor-2026-pdac-vte-implementation", "ash-2023-primary-vte-prophylaxis-guide", "overvad-2022-khorana-validation", "clinicaltrials-nct05178628-impact-pro", "clinicaltrials-nct06861088-catiq-p3", "clinicaltrials-nct05245877-preposterous-pancreas", "clinicaltrials-nct07196020-pinpoint-vte", "zwicker-2019-isoquercetin-catiq", "quercis-2026-kinisoquin-profile", "patent-us10391096-kinisoquin", "patent-wo2020112828-isoquercetin-cancer", "van-der-gaag-2010-pbd-rct", "costamagna-2026-sems-early-surgery-rct", "tamura-2021-fcsems-plastic-rct", "seo-2019-covered-uncovered-sems-rct", "lyu-2023-metal-plastic-nat-meta", "vehvilainen-2024-cholangitis-nat-interruption", "chen-2023-element-biliary-rct", "paik-2018-eus-ercp-biliary-rct", "barbosa-2024-eus-ercp-meta", "esge-2018-biliary-stenting-guideline", "asge-2024-therapeutic-eus-guideline", "nhs-england-2026-r367", "asco-2024-germline-panel-guideline", "llach-2024-universal-germline", "chittenden-2021-systematic-germline", "walker-2021-embedded-germline-station", "wang-2022-oncology-germline-cascade", "ramsey-2023-mainstreaming-germline", "rodriguez-2024-generate", "mcdonald-2024-germline-protocol", "sinan-2025-pdac-germline-implementation", "grewal-2026-pancan-germline-survey", "katz-2026-gift-cascade", "chen-2023-fdr-cancer-risk", "npaca-2026-state-of-nation", "nhs-england-2024-hpb-pancreatic-service", "de-wilde-2012-centralization", "lemmens-2011-centralization", "latenstein-2021-centralization", "coupland-2016-england-resection-volume", "labori-2016-adjuvant-completion", "henry-2023-complications-adjuvant", "kollbeck-2025-complications-chemo", "van-hilst-2019-leopard2", "korrel-2023-diploma", "bruna-2025-diploma-long-term", "klotz-2024-europa", "liu-2024-robotic-open-pd", "jin-2026-portal", "andel-2025-routine-imaging", "van-goor-2025-nationwide-imaging", "daamen-2024-radar-panc-protocol", "clinicaltrials-nct04875325-radar-panc", "tai-2025-pan-asian-esmo-pancreatic", "labori-2024-norpact1", "bai-2025-cispd1", "tan-2026-neoadjuvant-ipd-meta", "ghaneh-2023-espac5", "clinicaltrials-nct04340141-a021806", "effenberger-2012-pdac-bone-marrow-dtc", "nordgard-2022-unresectable-dtc", "chalabi-hajkarim-2025-organotropism", "pei-2025-spatial-metastatic-pdac", "mcclellan-2026-quiescent-pdac-car-t", "tanaka-2012-microscopic-local-recurrence", "dotan-2024-giant", "dotan-2025-giant-ga", "clinicaltrials-nct04233866-giant", "betge-2022-grantpax", "clinicaltrials-nct02143219-pamela70", "mohile-2021-gap70", "li-2021-gain", "ihorst-2026-geriatric-comanagement", "clinicaltrials-nct06040801-frail-gi", "stoop-2026-preopanc4-surgery", "dekker-2026-taps-restaging", "verkolf-2026-taps-resection-prediction", "seelen-2026-pelican", "clinicaltrials-nct06714604-duration", "baum-2022-fap2286-firsthuman", "mcconathy-2026-lumiere-phase1", "clinicaltrials-nct04939610-lumiere", "baum-2026-3bp3940", "clinicaltrials-nct05432193-frontier", "clinicaltrials-nct07229768-ctr-fapi", "alistar-2017-devimistat-phase1", "philip-2024-avenger500", "surana-2025-erk-hcq-pdac", "clinicaltrials-nct05221320-ulixertinib-hcq", "manji-2023-mekiauto", "clinicaltrials-nct04214418-mekiauto", "clinicaltrials-nct04892017-inlexisertib", "deciphera-2022-dcc3116-phase1", "clinicaltrials-nct04524702-paricalcitol-hcq", "dimcevski-2016-sonoporation-phase1", "wang-2023-sonochemotherapy-pdac", "clinicaltrials-nct04821284-sonoporation", "ius-2026-sonoporation-nct04821284-abstract", "adler-2026-sonoporation-perfusion", "alagesan-2026-sig12d-loder", "ctis-2026-sil204-phase23", "silexion-2026-sil204-site-initiation", "silexion-2026-halfyear-filing", "richards-2006-ci994-pdac", "heumann-2022-azacitidine-pdac", "sohal-2020-thu-decitabine-pdac", "baretti-2024-entinostat-nivolumab-pdac", "clinicaltrials-nct03250273-entinostat-nivolumab", "safyan-2026-azacitidine-pembrolizumab-pdac", "clinicaltrials-nct03264404-azacitidine-pembrolizumab", "rodon-2024-amg193-prmt5", "matsumoto-2026-mtap-pancreatic", "pavlick-2026-mtap-genomics", "clinicaltrials-nct05732831-vopimetostat", "tango-2025-vopimetostat-monotherapy", "clinicaltrials-nct06922591-vopimetostat-ras", "tango-2026-vopimetostat-ras-data", "tango-2026-q2-10q", "mahalingam-2026-elraglusib-randomized", "clinicaltrials-nct03678883-elraglusib", "pathak-2026-elraglusib-folfirinox", "clinicaltrials-nct05077800-elraglusib-folfirinox", "actuate-2026-q2-10q", "sec-2026-actuate-submissions", "clinicaltrials-nct05254171-aspire", "ctis-2024-514714-12-00-aspire", "panbela-2024-q3-10q", "panbela-2025-note-default", "clinicaltrials-nct03257033-tigerpac", "hatoum-2024-tamp-rr1-rr2", "novelli-2026-tigerpac-pk", "renovorx-2023-tigerpac-interim", "renovorx-2025-10k", "renovorx-2026-q2-10q", "renovorx-2026-full-enrollment" ], "links": [ { "target": "hypothesis-organoid-guided-early-switch", "relation": "holds-capital-for" }, { "target": "trial-pdac-organoid-action-map", "relation": "audits" }, { "target": "hypothesis-pdac-two-timepoint-staging-route", "relation": "funds" }, { "target": "unknown-pdac-staging-laparoscopy-rule", "relation": "must-resolve" }, { "target": "programme", "relation": "part-of" }, { "target": "programme-100m-randomized-platform", "relation": "extends" }, { "target": "hypothesis-one-time-risk-triggered-workup", "relation": "tests" }, { "target": "technology-opportunistic-ct-ai-pancreas", "relation": "tests" }, { "target": "person-christine-iacobuzio-donahue", "relation": "designed-by" }, { "target": "person-suresh-chari", "relation": "designed-by" }, { "target": "trial-new-onset-diabetes-workup-map", "relation": "must-not-duplicate" }, { "target": "trial-pancreatic-precursor-action-map", "relation": "must-not-duplicate" }, { "target": "hypothesis-dynamic-cyst-interception-rule", "relation": "tests" }, { "target": "trial-gallbladder-prevention-detection-action-map", "relation": "audits" }, { "target": "hypothesis-gallbladder-two-stage-interception-rule", "relation": "tests" }, { "target": "trial-small-pnet-action-map", "relation": "audits" }, { "target": "hypothesis-pnet-time-updated-surgery-rule", "relation": "tests" }, { "target": "trial-cca-action-map", "relation": "audits" }, { "target": "hypothesis-cca-route-specific-action-platform", "relation": "tests" }, { "target": "trial-hcc-action-map", "relation": "audits" }, { "target": "hypothesis-hcc-liver-reserve-co-control", "relation": "tests" }, { "target": "trial-crc-action-map", "relation": "audits" }, { "target": "unknown-crc-ctdna-positive-action", "relation": "constrains" }, { "target": "unknown-crc-liver-local-cure-selector", "relation": "constrains" }, { "target": "trial-pdac-cachexia-action-map", "relation": "must-not-duplicate" }, { "target": "unknown-cachexia-treatment-utility", "relation": "constrains" }, { "target": "trial-pert-pdac-action-map", "relation": "must-not-duplicate" }, { "target": "unknown-pert-pdac-utility", "relation": "constrains" }, { "target": "trial-pdac-vte-action-map", "relation": "must-not-duplicate" }, { "target": "unknown-pdac-vte-prophylaxis-rule", "relation": "constrains" }, { "target": "hypothesis-pdac-vte-route-completion", "relation": "funds" }, { "target": "claim-kinisoquin-access-stack", "relation": "constrained-by" }, { "target": "trial-pdac-biliary-drainage-action-map", "relation": "must-not-duplicate" }, { "target": "unknown-pdac-biliary-drainage-route", "relation": "constrains" }, { "target": "hypothesis-pdac-biliary-route-completion", "relation": "funds" }, { "target": "company-boston-scientific-biliary", "relation": "may-contract-with" }, { "target": "trial-pdac-germline-cascade-action-map", "relation": "must-not-duplicate" }, { "target": "unknown-pdac-germline-family-action-route", "relation": "constrains" }, { "target": "hypothesis-pdac-germline-family-route-completion", "relation": "funds" }, { "target": "system-england-r367-germline-criteria", "relation": "could-change" }, { "target": "claim-pdac-surgery-is-a-complete-treatment-route", "relation": "funds" }, { "target": "system-england-pancreatic-surgery-network", "relation": "funds" }, { "target": "failure-pdac-postoperative-treatment-loss", "relation": "funds" }, { "target": "failure-leopard2-laparoscopic-whipple-safety", "relation": "funds" }, { "target": "unknown-pdac-surgical-route-completion", "relation": "funds" }, { "target": "trial-pdac-surgical-route-action-map", "relation": "funds" }, { "target": "hypothesis-pdac-surgical-route-completion", "relation": "funds" }, { "target": "claim-routine-imaging-after-pdac-resection-not-yet-proven", "relation": "funds" }, { "target": "unknown-postresection-surveillance-utility", "relation": "funds" }, { "target": "trial-pdac-postresection-surveillance-action-map", "relation": "funds" }, { "target": "claim-neoadjuvant-therapy-improves-selection-not-cure", "relation": "funds" }, { "target": "failure-norpact1-neoadjuvant-delivery", "relation": "funds" }, { "target": "unknown-pdac-perioperative-treatment-route", "relation": "funds" }, { "target": "trial-pdac-perioperative-sequencing-action-map", "relation": "funds" }, { "target": "claim-human-pdac-dtc-prognostic-not-dormancy", "relation": "funds" }, { "target": "failure-pdac-dormancy-sample-gap", "relation": "funds" }, { "target": "trial-pdac-residual-cell-state-action-map", "relation": "funds" }, { "target": "hypothesis-human-dormancy-reawakening-gate", "relation": "funds" }, { "target": "claim-age-alone-cannot-route-pdac-treatment", "relation": "funds" }, { "target": "failure-grantpax-geriatric-route", "relation": "funds" }, { "target": "unknown-older-pdac-treatment-route", "relation": "funds" }, { "target": "trial-older-pdac-action-map", "relation": "funds" }, { "target": "hypothesis-older-pdac-fast-support-route", "relation": "funds" }, { "target": "person-efrat-dotan", "relation": "could-contract" }, { "target": "lab-ecog-acrin-geriatric-oncology-network", "relation": "could-contract" }, { "target": "claim-lapc-response-warrants-surgical-reassessment", "relation": "records" }, { "target": "unknown-lapc-surgery-causal-benefit", "relation": "leaves-open" }, { "target": "trial-lapc-conversion-surgery-action-map", "relation": "bounded-by" }, { "target": "hypothesis-randomized-lapc-resection-after-response", "relation": "could-fund-after-review" }, { "target": "trial-pdac-fap-radioligand-action-map", "relation": "audits" }, { "target": "unknown-pdac-fap-radioligand-treatment-effect", "relation": "leaves-open" }, { "target": "hypothesis-fapi-theranostic-pdac", "relation": "could-fund-after-review" }, { "target": "company-novartis-fap-radioligand", "relation": "could-contract" }, { "target": "trial-pdac-autophagy-action-map", "relation": "audits" }, { "target": "unknown-selective-autophagy-pdac-contribution", "relation": "leaves-open" }, { "target": "hypothesis-tumour-verified-selective-autophagy-blockade", "relation": "could-fund-after-review" }, { "target": "trial-pdac-sonoporation-action-map", "relation": "audits" }, { "target": "unknown-pdac-sonoporation-treatment-effect", "relation": "leaves-open" }, { "target": "hypothesis-measured-fixed-backbone-sonoporation", "relation": "could-fund-after-review" }, { "target": "trial-sil204-action-map", "relation": "audits" }, { "target": "unknown-sil204-human-target-delivery", "relation": "leaves-open" }, { "target": "hypothesis-multicompartment-kras-sirna-delivery", "relation": "could-fund-after-review" }, { "target": "failure-broad-epigenetic-treatment-pdac", "relation": "learns-from" }, { "target": "claim-mtap-prmt5-selected-human-signal", "relation": "bounded-by" }, { "target": "failure-vopimetostat-ras-contribution-unknown", "relation": "learns-from" }, { "target": "unknown-mtap-prmt5-ras-contribution", "relation": "leaves-open" }, { "target": "trial-mtap-prmt5-action-map", "relation": "implements" }, { "target": "hypothesis-mtap-prmt5-ras-deeper-control", "relation": "tests" }, { "target": "company-tango-therapeutics", "relation": "contains" }, { "target": "claim-elraglusib-randomized-survival-signal", "relation": "bounded-by" }, { "target": "failure-elraglusib-analysis-and-survival-discordance", "relation": "learns-from" }, { "target": "unknown-elraglusib-survival-replication", "relation": "leaves-open" }, { "target": "trial-elraglusib-action-map", "relation": "implements" }, { "target": "hypothesis-elraglusib-survival-confirmation-gate", "relation": "tests" }, { "target": "company-actuate-therapeutics", "relation": "contains" }, { "target": "failure-aspire-trial-operational-collapse", "relation": "learns-from" }, { "target": "unknown-aspire-assigned-denominator-and-results", "relation": "leaves-open" }, { "target": "trial-pdac-polyamine-action-map", "relation": "implements" }, { "target": "hypothesis-aspire-data-rescue-gate", "relation": "tests" }, { "target": "company-panbela-therapeutics", "relation": "contains" }, { "target": "failure-tigerpac-component-and-interim-inference", "relation": "learns-from" }, { "target": "unknown-tigerpac-package-survival-and-delivery", "relation": "leaves-open" }, { "target": "trial-pdac-intraarterial-delivery-action-map", "relation": "implements" }, { "target": "hypothesis-tigerpac-existing-final-gate", "relation": "tests" }, { "target": "company-renovorx", "relation": "contains" } ], "evidence": { "humanProspective": "not-applicable", "humanRetrospective": "not-applicable", "randomised": "not-applicable", "independentReplication": "none", "sampleSize": "Proposed multiple trials; detection starts with EDI n=8,869 and holds a possible 40,000-60,000-person randomized programme behind evidence gates", "effect": "Mission portfolio; outcome probabilities are decision priors", "limits": [ "Generated programme design", "Costs are planning estimates", "Several drug and data agreements unnegotiated", "Probability ranges are subjective", "Global recruitment and regulation may extend timelines", "Related cancers receive limited budgets rather than equal coverage" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "Each lane has human evidence, but joining treatment, residual disease and detection into one mortality programme is untested." }, "animalEvidence": { "rating": "weak", "reason": "Animal evidence supports selected mechanisms but cannot validate this human portfolio." }, "inVitroEvidence": { "rating": "moderate", "reason": "Laboratory systems can screen resistance routes, not establish programme-wide clinical benefit." }, "reproducibility": { "rating": "none", "reason": "No independent group has reproduced the integrated programme or its stated outcome distribution." }, "commercialConflicts": { "rating": "high", "reason": "Several lanes rely on company-owned medicines, assays and data, requiring independent contracts and analysis." }, "recency": { "rating": "current", "reason": "The portfolio uses current RAS and residual-disease evidence plus EDI status and diabetes-imaging results checked through September 2026." } } }, "layout": { "x": 57, "y": 119 }, "updated": "2026-09-15" }, { "id": "programme-5m-decision-core", "title": "£5 million programme — build the PDAC decision core and kill weak ideas quickly", "shortTitle": "£5m decision core", "type": "programme", "status": "hypothesis", "scope": [ "pdac" ], "summary": "A two-year programme cannot develop a new drug; it can create the shared samples, measurements and independent analysis needed to choose one expensive intervention well.", "content": "Objective: determine which residual-disease and common-RAS resistance measurements are reproducible enough to control a later trial. Leadership: Christine Iacobuzio-Donahue for human evolution and tissue, Suresh Chari for clinical cohorts, David Tuveson for model testing, and an independent trial statistician with no product equity. Partners: MSK, MD Anderson, CSHL, Johns Hopkins and one NHS pathology network; data access from Revolution Medicines should be negotiated, not assumed. Budget: £1.20m for data and specimen agreements; £1.25m for serial blood and matched tissue from 150–200 patients; £0.90m for locked ctDNA, CA19-9 and immune assays; £0.75m for organoid and resistance perturbation work; £0.45m for independent statistics, data engineering and audit; £0.25m for patient input, ethics and health economics; £0.20m reserve. Total = £5.00m. First tests: cross-lab assay concordance, serial RAS-clone dynamics under daraxonrasib or chemotherapy, and whether blood changes precede imaging recurrence. Go: at least two labs reproduce the same resistance class; sample completeness exceeds 80%; a marker predicts recurrence or resistance with a confidence interval narrow enough to change trial stratification. Stop: specimen access under 60% after six months, assay discordance over 20%, or no reproducible signal after the fixed analysis. Time: 24 months. Planning outcome distribution: 15–25% trial-ready selector; 45–60% useful negative result that retires a route; 15–25% ambiguous result; 5–10% operational failure. It does not fund a mortality trial, population screening or a new medicine.", "contentSections": [ "A bounded precursor audit fits inside the £1.20m data-and-specimen-agreement line without opening another trial. Obtain the EA2185 amendment, randomized-subset accrual and events; obtain 3P-C component outcomes and harms; document SIPP-T3 screen failures and participant disposition; and negotiate an independent PancreaSeq GC dataset. Go only if the audit identifies a decision that existing data cannot recover and a feasible recruitment route. Stop a proposed precursor trial if it repeats EA2185's original question, uses cyst growth alone, or cannot recruit 60 eligible participants across sites in 12 months.", "Organoid work inside the existing £0.75m line is limited to data recovery, protocol comparison and fast failure tests. Obtain NCT04931381, NCT04931394 and NCT05351983 results; align PASS-01, PROSPER-PANC and UNITEPANC fields; and define the all-patient route denominator. It does not fund a new guidance cohort or change treatment.", "FAP-radioligand work is a data-recovery task inside the existing data-and-specimen agreement, not a new treatment line. Request LuMIERE patient- and lesion-level scan, dose, toxicity and response data and FRONTIER results; build the all-denominator table. Go to external dose-response testing only with at least 80% linked lesions and cross-centre calibration error no greater than 10%. No patient receives treatment and no new capital is added.", "Autophagy work is data recovery inside the existing agreement and pharmacology audit. Request the complete 91-person NCT04892017 dataset, registry history, pancreatic-cohort disposition, tumour and blood measurements, combination exposure, response, toxicity and missing denominators. Do not fund another HCQ cohort or infer biology from a business termination. No patient receives treatment and no new capital is added.", "Sonoporation work is a data-release request inside the existing trial and data agreement. Obtain NCT04821284 protocol versions, analysis plan, participant flow, backbone and stage counts, dose, harm, quality of life and survival events; reconcile 97 reported with 120 estimated. Do not fund another device cohort or treat the exploratory 13-person perfusion series as drug delivery. No new capital is added.", "SIL204 adds no trial or budget. Offer assay alignment and independent analysis inside existing pharmacology and data agreements only if the sponsor changes the no-participant-data plan. Use the live 15-to-21-person Segment 1 to test paired multi-region mutant-KRAS RNA and protein reduction, systemic exposure, EUS harm and chemotherapy delivery. Stop support before Segment 2 if any gate or public-data contract fails.", "VCN-01 uses no £5m experiment capital. The decision core may reproduce the public VIRAGE statistics and specify the data request, but Theriva must release all 112 records and fund access. The branch stops if denominators or the final analysis plan remain unavailable.", "MTAP-selected PRMT5 work adds no cohort or budget. The route is biologically and commercially credible, but Tango and its partners already fund a 225-person monotherapy programme and a 183-person combination programme. Require all 54 treated PDAC records to mature, exact MTAP-loss assays and a randomized RAS-inhibitor-alone contribution arm. Existing mission teams may review assays and analysis only under public-data terms.", "Elraglusib uses the existing independent statistics and data-access function, not a new budget line. Request the 269-person weekly-versus-control randomized record, protocol versions and dated analysis plans. Reproduce ITT survival and missingness limits before discussing phase 3. The gate is HR at most 0.75, 95% CI excluding 1.0 and no plausible missing-data case above 0.85.", "Cell-cycle work adds no cohort and no budget. Use the existing data-access and statistics function to recover the 12-person azenosertib safety stage. Publish paired tumour change, six-month PFS, harm and the recorded expansion decision. Failure at the 10-pair, 8-target-change, 5/12 control or no-treatment-death gates closes PDAC WEE1 work." ], "sourceIds": [ "oreilly-2026-daraxonrasib", "varghese-2025-kras-dosage", "botta-2024-tumor-informed-ctdna", "yachida-2010-metastatic-timing", "weinberg-2020-ea2185-protocol", "clinicaltrials-nct04239573-ea2185", "clinicaltrials-nct04207944-3pc", "clinicaltrials-nct06519097-sipp-t3", "singhi-2026-pancreaseq-gc", "baum-2022-fap2286-firsthuman", "mcconathy-2026-lumiere-phase1", "clinicaltrials-nct04939610-lumiere", "baum-2026-3bp3940", "clinicaltrials-nct05432193-frontier", "clinicaltrials-nct07229768-ctr-fapi", "alistar-2017-devimistat-phase1", "philip-2024-avenger500", "surana-2025-erk-hcq-pdac", "clinicaltrials-nct05221320-ulixertinib-hcq", "manji-2023-mekiauto", "clinicaltrials-nct04214418-mekiauto", "clinicaltrials-nct04892017-inlexisertib", "deciphera-2022-dcc3116-phase1", "clinicaltrials-nct04524702-paricalcitol-hcq", "dimcevski-2016-sonoporation-phase1", "wang-2023-sonochemotherapy-pdac", "clinicaltrials-nct04821284-sonoporation", "ius-2026-sonoporation-nct04821284-abstract", "adler-2026-sonoporation-perfusion", "alagesan-2026-sig12d-loder", "ctis-2026-sil204-phase23", "silexion-2026-sil204-site-initiation", "silexion-2026-halfyear-filing", "richards-2006-ci994-pdac", "heumann-2022-azacitidine-pdac", "sohal-2020-thu-decitabine-pdac", "baretti-2024-entinostat-nivolumab-pdac", "clinicaltrials-nct03250273-entinostat-nivolumab", "safyan-2026-azacitidine-pembrolizumab-pdac", "clinicaltrials-nct03264404-azacitidine-pembrolizumab", "rodon-2024-amg193-prmt5", "matsumoto-2026-mtap-pancreatic", "pavlick-2026-mtap-genomics", "clinicaltrials-nct05732831-vopimetostat", "tango-2025-vopimetostat-monotherapy", "clinicaltrials-nct06922591-vopimetostat-ras", "tango-2026-vopimetostat-ras-data", "tango-2026-q2-10q", "mahalingam-2026-elraglusib-randomized", "clinicaltrials-nct03678883-elraglusib", "pathak-2026-elraglusib-folfirinox", "clinicaltrials-nct05077800-elraglusib-folfirinox", "actuate-2026-q2-10q", "sec-2026-actuate-submissions", "chiorean-2023-abemaciclib-pdac", "baghdadi-2019-palbociclib-cdkn2a", "hidalgo-2022-palbociclib-nabpaclitaxel", "laquente-2017-chk1-pdac", "huffman-2023-ly2880070-pdac", "cuneo-2019-adavosertib-pdac", "huffman-2024-azenosertib-pdac-design", "clinicaltrials-nct06015659-azenosertib" ], "links": [ { "target": "trial-pdac-organoid-action-map", "relation": "audits" }, { "target": "programme", "relation": "part-of" }, { "target": "ranking-provisional-pdac-bottlenecks", "relation": "targets" }, { "target": "person-christine-iacobuzio-donahue", "relation": "led-by" }, { "target": "person-suresh-chari", "relation": "led-by" }, { "target": "person-david-tuveson", "relation": "led-by" }, { "target": "company-revolution-medicines", "relation": "requires-data-from" }, { "target": "trial-pancreatic-precursor-action-map", "relation": "audits" }, { "target": "failure-ea2185-randomized-question-closed", "relation": "learns-from" }, { "target": "failure-sipp-t3-accrual-collapse", "relation": "learns-from" }, { "target": "trial-pdac-fap-radioligand-action-map", "relation": "audits" }, { "target": "hypothesis-fapi-theranostic-pdac", "relation": "tests" }, { "target": "failure-fapi-scan-dose-benefit-collapse", "relation": "learns-from" }, { "target": "trial-pdac-autophagy-action-map", "relation": "audits" }, { "target": "failure-metabolic-repurposing-pdac", "relation": "learns-from" }, { "target": "hypothesis-tumour-verified-selective-autophagy-blockade", "relation": "tests" }, { "target": "trial-pdac-sonoporation-action-map", "relation": "audits" }, { "target": "failure-sonoporation-overall-replication-subgroup-rescue", "relation": "learns-from" }, { "target": "hypothesis-measured-fixed-backbone-sonoporation", "relation": "tests" }, { "target": "trial-sil204-action-map", "relation": "audits" }, { "target": "failure-sig12d-loder-randomized-os-not-shown", "relation": "learns-from" }, { "target": "hypothesis-multicompartment-kras-sirna-delivery", "relation": "tests" }, { "target": "failure-broad-epigenetic-treatment-pdac", "relation": "learns-from" }, { "target": "claim-mtap-prmt5-selected-human-signal", "relation": "bounded-by" }, { "target": "failure-vopimetostat-ras-contribution-unknown", "relation": "learns-from" }, { "target": "unknown-mtap-prmt5-ras-contribution", "relation": "leaves-open" }, { "target": "trial-mtap-prmt5-action-map", "relation": "implements" }, { "target": "hypothesis-mtap-prmt5-ras-deeper-control", "relation": "tests" }, { "target": "company-tango-therapeutics", "relation": "contains" }, { "target": "claim-elraglusib-randomized-survival-signal", "relation": "bounded-by" }, { "target": "failure-elraglusib-analysis-and-survival-discordance", "relation": "learns-from" }, { "target": "unknown-elraglusib-survival-replication", "relation": "leaves-open" }, { "target": "trial-elraglusib-action-map", "relation": "implements" }, { "target": "hypothesis-elraglusib-survival-confirmation-gate", "relation": "tests" }, { "target": "company-actuate-therapeutics", "relation": "contains" }, { "target": "failure-broad-cell-cycle-checkpoint-pdac", "relation": "learns-from" }, { "target": "unknown-wee1-replication-stress-pdac", "relation": "leaves-open" }, { "target": "trial-cell-cycle-checkpoint-action-map", "relation": "implements" }, { "target": "hypothesis-human-replication-stress-selector", "relation": "tests" } ], "evidence": { "humanProspective": "not-applicable", "humanRetrospective": "not-applicable", "randomised": "not-applicable", "independentReplication": "none", "sampleSize": "Proposed 150-200 patients plus existing trial specimens", "effect": "Decision-enabling programme; no direct patient-outcome claim", "limits": [ "Generated programme design", "Costs are 2026 planning estimates", "Data licenses not negotiated", "Probability ranges are judgement", "Cannot prove clinical utility or survival" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "Human RAS response, recurrence and ctDNA association support measuring escape, but not the proposed selector." }, "animalEvidence": { "rating": "weak", "reason": "Models show individual resistance routes but do not validate the proposed human decision system." }, "inVitroEvidence": { "rating": "moderate", "reason": "Organoids can test combinations, but prospective prediction of patient response remains unproved." }, "reproducibility": { "rating": "none", "reason": "The planned cross-laboratory resistance and residual-disease rule has not yet been reproduced." }, "commercialConflicts": { "rating": "high", "reason": "Useful execution may require company-controlled drug, specimen and trial-data access." }, "recency": { "rating": "current", "reason": "The programme follows the 2026 common-RAS result and current residual-disease evidence." } } }, "layout": { "x": 48, "y": 116 }, "updated": "2026-09-15" }, { "id": "programme-mission-table-people", "title": "Mission table: ten complementary seats, with conflicts kept visible", "shortTitle": "Mission team", "type": "programme", "status": "open", "scope": [ "pdac", "pancreatic-all", "colorectal", "hepatobiliary", "pan-cancer", "transfer-unproven" ], "summary": "The working table is selected by missing capability: trials, human evolution, RAS resistance, detection delivery, models, immunity, AI, statistics and patient oversight.", "content": "Working shortlist, not a prize list or an appointment. Seat 1, multi-centre trial execution: Eileen O'Reilly. Seat 2, human evolution: Christine Iacobuzio-Donahue. Seat 3, RAS resistance: Andrew Aguirre. Seat 4, risk-enriched detection: Suresh Chari. Seat 5, Japanese health-system detection: Keiji Hanada, because the Onomichi route joins community risk assessment, referral and specialist diagnosis and reported stage 0/I rising from 13.9% to 22.1% across eras. Seat 6, models and fast rejection: David Tuveson. Seat 7, residual-disease immunity and manufacture: Vinod Balachandran. Seat 8, external AI testing: Regina Barzilay. Seat 9, trial statistics and stopping rules: Christina Yap, conditional on a formal conflict screen; her role is to lock estimands, multiplicity and interim rules. Seat 10, paid patient and public co-chair: Ali Stunt, subject to a transparent selection process and backed by a wider patient panel. Michael Goggins, Anirban Maitra, Wungki Park, Robert Vonderheide, Bert Vogelstein, Liwei Wang, George Hanna, Núria Malats, Eduard Jonas and Renata D'Alpino Peixoto remain linked contributors. Jonas adds South African pancreatic and HPB pathway work; Peixoto adds the LACOG Latin American data network. Neither gets a voting seat until site capacity, representation, conflicts, willingness and data rights are checked. Governance: no investigator votes on a product in which they, their institution or close collaborators hold equity, patents or sponsor dependence; the statistics chair's independence must be checked before appointment; each go decision requires the patient co-chair; negative results enter the failure database within 90 days. Geographic names now span the US, Japan, UK, China, Europe, Africa and Latin America, but a name is not proof of representative delivery.", "sourceIds": [ "msk-oreilly-profile-2026", "msk-iacobuzio-donahue-profile-2026", "danafarber-aguirre-profile-2026", "mdanderson-chari-profile-2026", "ja-onomichi-hanada-profile-2026", "cshl-tuveson-profile-2026", "msk-balachandran-profile-2026", "mit-barzilay-profile-2026", "icr-christina-yap-profile-2026", "pancreatic-cancer-action-ali-stunt-profile-2026", "uct-2026-cancer-projects", "jonas-2025-sub-saharan-hcc-pathways", "clinicaltrials-nct05924789-gaspar", "peixoto-2025-gaspar-results", "clinicaltrials-nct04340141-a021806", "bai-2025-cispd1" ], "links": [ { "target": "people", "relation": "part-of" }, { "target": "person-eileen-oreilly", "relation": "employs" }, { "target": "person-christine-iacobuzio-donahue", "relation": "employs" }, { "target": "person-andrew-aguirre", "relation": "employs" }, { "target": "person-suresh-chari", "relation": "employs" }, { "target": "person-david-tuveson", "relation": "employs" }, { "target": "person-vinod-balachandran", "relation": "employs" }, { "target": "person-regina-barzilay", "relation": "employs" }, { "target": "person-keiji-hanada", "relation": "employs" }, { "target": "person-christina-yap", "relation": "employs" }, { "target": "person-ali-stunt", "relation": "employs" }, { "target": "person-eduard-jonas", "relation": "international-contributor" }, { "target": "person-renata-dalpino-peixoto", "relation": "international-contributor" }, { "target": "programme-25m-prospective-observatory", "relation": "supports" }, { "target": "programme-100m-randomized-platform", "relation": "supports" }, { "target": "risks", "relation": "depends-on" }, { "target": "person-cristina-ferrone", "relation": "consults" }, { "target": "person-tingbo-liang", "relation": "consults" } ], "evidence": { "humanProspective": "not-applicable", "humanRetrospective": "not-applicable", "randomised": "not-applicable", "independentReplication": "not-applicable", "sampleSize": "10 named working seats plus 10 linked contributors", "effect": "Capability and governance design, not a clinical-effect claim", "limits": [ "Selection is judgment", "Availability and willingness not checked", "Conflicts require live disclosure review", "Named statistics and patient seats are proposals, not appointments", "African and Latin American contributors are named but their network reach is not yet representative", "The LACOG registry and abstract enrolment totals conflict", "One patient co-chair requires a wider representative panel" ], "dimensions": { "biologicalPlausibility": { "rating": "not-applicable", "reason": "This is a capability and governance selection rather than a biological claim." }, "animalEvidence": { "rating": "not-applicable", "reason": "Animal evidence does not establish a person's availability or programme fit." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Laboratory evidence belongs to linked research records rather than this team design." }, "reproducibility": { "rating": "weak", "reason": "Roles and work histories are documented, but the proposed table has not operated together." }, "commercialConflicts": { "rating": "high", "reason": "Many candidates have institutional, patent, product or sponsor ties that require current review." }, "recency": { "rating": "current", "reason": "Profiles and active roles were checked in September 2026, but availability was not checked." } } }, "layout": { "x": 83, "y": 112 }, "updated": "2026-09-14", "contentSections": [ "Two non-voting perioperative data leads are added without enlarging the ten-seat decision table. Cristina Ferrone supplies A021806 protocol, route and mature outcome access. Tingbo Liang supplies CISPD-1 delivery and outside-centre reproduction access. Both remain conditional on availability, conflicts, patient-level rights and independent analysis." ] }, { "id": "programme-unconstrained-global-system", "title": "Effectively unconstrained programme — a £5 billion staged global cancer-control system", "shortTitle": "£5bn global system", "type": "programme", "status": "hypothesis", "scope": [ "pdac", "pnet", "cholangiocarcinoma", "gallbladder", "hcc", "colorectal", "hepatobiliary", "pan-cancer", "transfer-unproven" ], "summary": "Unconstrained capital should buy scale, comparison and manufacturing, not remove stop rules; no more than the first £500m is exposed until the key biological gates are met.", "content": "Use a ten-year ceiling of £5.0bn as a planning case, released in stages. Allocation: £0.50bn for the first three years of cohorts, samples, resistance maps and assay validation; £1.50bn for multi-arm phase 2 and phase 3 systemic and residual-disease trials; £1.00bn for randomized risk-triggered and multicancer detection with mortality follow-up; £0.65bn for trial-grade diagnostics, radiotracers, cell or vaccine manufacturing and distributed biobanks; £0.45bn for disease-specific pNET, biliary, gallbladder, HCC and colorectal programmes; £0.35bn for global data, statistics, regulation and health economics; £0.25bn for independent replication, negative-result recovery and technology replacement; £0.30bn reserve. Total = £5.00bn. The £0.35bn global line includes a capped £20m regional route test: £5m for one year of common consecutive registration, then at most £15m for a randomized repair only if a loss of at least 15 percentage points is found at two sites or in 100 patients. £5m + £15m = £20m; the total remains £5.00bn. The first £500m follows the integrated programme. Later tranches release only after prespecified human gates; failure in one lane reallocates money rather than lowering its endpoint. Build public-interest data rights, common assays, complete trial reporting and manufacturing access into every contract. Acquire a company only when ownership is cheaper than licensing and its platform has human target engagement; do not buy a valuation as a substitute for evidence. Ten-year planning outcomes: 60–80% chance of at least one material reduction in death or recurrence in one scoped cancer; 25–45% chance of two or more practice-changing capabilities; 5–15% chance that most newly treated PDAC patients reach durable control sufficient to make another cause of death more likely. These are subjective ranges, not model outputs. Even £5bn cannot force tumour biology to cooperate, shorten survival follow-up, create a missing causal target or make an unsafe therapy acceptable.", "contentSections": [ "Objective: build a global system that can find, test, manufacture and deploy at least one intervention that reduces death or recurrence, while closing routes that fail. Leadership: an independent mission chief and statistician without product equity, a patient co-chair, disease leads for PDAC, pNET, biliary cancer, HCC and colorectal cancer, and accountable heads of trials, diagnostics, manufacturing, data and access.", "Partners and institutions: multi-continent cancer centres and health systems; national trial groups; public biobanks; regulators and health-technology assessors; trial-grade assay, drug, vaccine, cell and radiotracer manufacturers; and industry partners under contracts that preserve negative-result publication, external audit and public-interest data rights. Current named routes include UCT and Groote Schuur, the Africa HPB Cancer Consortium, LACOG GASPAR and the five-country LATAM-CCA Registry. Geographic release requires proof: at least five African and five Latin American sites must each return one year of population or catchment denominator, first contact, imaging, pathology, stage and organ reserve, recommendation, delivered treatment or reason for loss, follow-up and missingness under one dictionary. Capture and twelve-month follow-up must each reach 80%, and required-field missingness must stay at or below 20%. No intervention money releases until one repairable step accounts for at least 15 percentage points of route loss at two sites or 100 patients. The matched repair must increase completed recommended treatment by at least 10 points; it stops below five points, for excess harm or if delay only moves downstream." ], "sourceIds": [ "oreilly-2026-daraxonrasib", "sasieni-2026-nhs-galleri", "tie-2022-dynamic-colon", "chang-1997-hbv-vaccine-hcc", "baudin-2026-oclurandom", "asombang-2026-africa-hpb-registry", "clinicaltrials-nct05924789-gaspar", "peixoto-2025-gaspar-results", "gismalla-2026-africa-pancreatic-review", "sobnach-2024-ssa-hcc-systematic-review", "sobnach-2025-south-africa-hcc-disparities", "sobnach-2026-global-hcc-disparities", "da-fonseca-2024-latam-cca" ], "links": [ { "target": "programme", "relation": "part-of" }, { "target": "programme-500m-integrated-mission", "relation": "depends-on" }, { "target": "funding", "relation": "depends-on" }, { "target": "investment", "relation": "contrasts-with" }, { "target": "risks", "relation": "constrained-by" }, { "target": "timeline", "relation": "depends-on" }, { "target": "lab-uct-groote-schuur-hpb", "relation": "could-contract" }, { "target": "lab-africa-hpb-cancer-consortium", "relation": "could-contract" }, { "target": "lab-lacog-gaspar", "relation": "could-contract" }, { "target": "lab-latam-cca-registry", "relation": "could-contract" }, { "target": "trial-regional-route-completion-action-map", "relation": "could-fund" }, { "target": "hypothesis-regional-route-completion", "relation": "could-test" } ], "evidence": { "humanProspective": "not-applicable", "humanRetrospective": "not-applicable", "randomised": "not-applicable", "independentReplication": "none", "sampleSize": "Proposed global portfolio; exact trials depend on early gates", "effect": "Ten-year planning case, not a forecast", "limits": [ "Generated programme design", "£5bn is a planning ceiling", "Probabilities are subjective", "Future standards and costs will change", "Governance at this scale is a major failure risk", "Unconstrained money does not remove biological or time constraints" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "The component routes have human support, but money cannot establish that the combined system will control several cancers." }, "animalEvidence": { "rating": "weak", "reason": "Animal evidence can guide mechanisms but does not support a £5bn outcome forecast." }, "inVitroEvidence": { "rating": "moderate", "reason": "Laboratory systems can narrow choices without proving population mortality effects." }, "reproducibility": { "rating": "none", "reason": "No comparable global programme has reproduced the proposed allocation or outcome probabilities." }, "commercialConflicts": { "rating": "high", "reason": "A portfolio of this size would depend on many owners of drugs, diagnostics, data and manufacturing." }, "recency": { "rating": "current", "reason": "The planning case uses current clinical, regulatory and programme evidence." } } }, "layout": { "x": 60, "y": 120 }, "updated": "2026-09-14" }, { "id": "programme", "title": "Programme designs", "shortTitle": "Programme", "type": "programme", "status": "hypothesis", "scope": [ "pdac", "pnet", "cholangiocarcinoma", "gallbladder", "hcc", "colorectal", "pan-cancer" ], "summary": "Concrete programmes at five capital levels, with people, partners, milestones, stop rules and probability ranges.", "content": "The programmes buy different levels of evidence; they are not smaller and larger versions of one plan. £5m over two years builds a decision core around 150–200 patients, cross-laboratory assays and RAS or residual-disease measurements. It cannot develop a drug or prove survival. Planning probability: 15–25% for a trial-ready selector, 45–60% for a useful negative result. £25m over four years builds an 800-person prospective observatory with fixed blood, imaging and tissue times, rapid pathology and organoid tests, external validation and health economics. Probability: 25–40% for one validated trial selector, 35–50% for decisive narrowing. £100m funds a resistance-directed randomized action platform and holds any new post-resection residual-disease trial behind an audit of DYNAMIC-Pancreas, CIRCPAC and MAP-03. Probability: 15–30% for at least one phase-3-ready strategy; 40–55% for decisive negative arms. £500m links global systemic-control trials and residual-disease action, but its £105m risk-triggered detection allocation releases only £5m for EDI and PANDOME data access, cross-site analysis and second-stage selector validation. The other £100m remains held until EDI reports and a distinct randomized action gap is proved; disease-specific lanes remain for pNET, biliary, gallbladder, HCC and colorectal work. Probability: 30–50% for at least one practice-changing result, but below 10–15% that most PDAC becomes reliably controllable within ten years. An unconstrained planning case is capped at £5bn over ten years and still uses tranches: £0.50bn for measurement and assay gates, £1.50bn for treatment trials, £1.00bn for detection with mortality follow-up, £0.65bn for manufacturing and biobanks, £0.45bn for related cancers, £0.35bn for data and regulation, £0.25bn for independent replication and failed-result recovery, and £0.30bn reserve. Total = £5.00bn. Its probability of at least one material reduction in death or recurrence in one scoped cancer is estimated at 60–80%; probability that most newly treated PDAC patients reach durable control remains only 5–15%. The nonlinearity is the point: capital can buy samples, comparison, manufacturing and statistical power, but cannot force a missing target to exist or compress ten-year survival follow-up. Every level has named leaders, partners, exact allocations, milestones, safety gates, stop rules and an outcome distribution. No later tranche releases because an earlier team produced papers, patents or response rates; it releases only after a reproduced human measurement supports a timely, tolerable action.", "sourceIds": [ "conroy-2022-prodige24-five-year", "oreilly-2026-daraxonrasib", "botta-2024-tumor-informed-ctdna", "sarno-2025-avatar", "tie-2022-dynamic-colon", "sasieni-2026-nhs-galleri", "edi-nct04662879", "frank-2025-pandome" ], "links": [ { "target": "mission", "relation": "part-of" }, { "target": "programme-5m-decision-core", "relation": "contains" }, { "target": "programme-25m-prospective-observatory", "relation": "contains" }, { "target": "programme-100m-randomized-platform", "relation": "contains" }, { "target": "programme-500m-integrated-mission", "relation": "contains" }, { "target": "programme-unconstrained-global-system", "relation": "contains" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "strong", "randomised": "strong", "independentReplication": "moderate", "sampleSize": "Five costed programme designs from £5m to £5bn", "effect": "Generated programme and probability ranges; no direct efficacy claim", "limits": [ "Costs are planning estimates rather than quotes", "Probabilities are judgement", "Compound and data access not negotiated", "Later comparator standards will change", "Clinical follow-up cannot be accelerated by spending alone" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "The programmes target human recurrence and RAS escape, but their complete decision systems are untested." }, "animalEvidence": { "rating": "weak", "reason": "Model evidence can narrow mechanisms but does not validate any budget-level outcome distribution." }, "inVitroEvidence": { "rating": "moderate", "reason": "Human-derived models can reject combinations but have not prospectively selected these programmes." }, "reproducibility": { "rating": "none", "reason": "No independent programme has reproduced the proposed staged allocations and outcomes." }, "commercialConflicts": { "rating": "high", "reason": "Several programme lanes require commercial drug, assay, data and manufacturing rights." }, "recency": { "rating": "current", "reason": "All five budgets use clinical and programme evidence current to September 2026." } } }, "layout": { "x": 50, "y": 110 }, "updated": "2026-09-14" }, { "id": "ranking-provisional-pdac-bottlenecks", "title": "Five largest PDAC gaps: mission impact and first-capital priority", "shortTitle": "Five largest gaps", "type": "bottleneck", "status": "open", "scope": [ "pdac", "pancreatic-all" ], "summary": "Durable systemic control ranks first by mission impact; resistance and residual-disease measurement rank first for the initial £25m because they can produce a clear decision sooner.", "content": "The primary ranking asks how much of the mission each solved gap could unlock. Mission-impact index = patient reach from 1 to 5 × outcome gain if solved from 1 to 5. It is a transparent planning scale, not a clinical effect estimate. Rank 1 — safely eliminate occult and established systemic PDAC: reach 5 × outcome 5 = 25. In selected resected patients given mFOLFIRINOX, five-year disease-free survival was 26.1%, so 73.9% had recurrence or death; in metastatic disease daraxonrasib doubled median survival in RAS-G12 disease but median progression still occurred at 7.3 months. The gap covers disease after surgery and established metastasis. Durable common-RAS treatment is the leading current route into this gap, not a separate gap. Rank 2 — detect biologically dangerous disease while treatment can still eliminate it: reach 4 × outcome 4 = 16. Only 15.4% of all pancreatic cancers in SEER are localised. An intentionally unrealistic stage-shift ceiling moves every case to current localised five-year survival: 43.6% instead of 13.7%, a 29.9-point gain and 3.18-fold improvement, yet still leaves a 56.4-point deficit from 100% five-year relative survival. Detection needs treatment capability and cannot be valued alone. Rank 3 — measure residual disease and prove that action changes recurrence: reach 3 × outcome 4 = 12. This mainly reaches people who complete curative-intent local treatment. Postoperative ctDNA predicts recurrence, while PDAC lacks a randomized action result; stage II colon cancer supplies a transferable trial design, not proof of PDAC benefit. Rank 4 — classify and reverse each patient's immune, stromal and delivery barrier without intolerable harm: reach 4 × outcome 2 = 8. Unselected checkpoint therapy is nearly inactive, vaccines show immune response without randomized recurrence benefit, and bulk stroma removal failed. The likely solution is several selectors rather than one universal drug. Rank 5 — prevent disease beyond small inherited-risk groups: reach 2 × outcome 3 = 6. Known high-risk genes covered 5.5% of cases in one large study, and PDAC lacks a removable cause with hepatitis-B-like population reach. Prevention remains high value but has a narrower visible starting population and longer proof time.\n\nCapital priority differs from mission impact. For a fixed first £25m, information index = decision impact from 1 to 5 × chance of a clear result ÷ years. Resistance measurement scores 5 × 0.35 ÷ 4 = 0.44; residual-disease action 4 × 0.35 ÷ 4 = 0.35; risk-enriched detection 4 × 0.25 ÷ 5 = 0.20; immune or delivery classification 3 × 0.25 ÷ 4 = 0.19; broad prevention 3 × 0.20 ÷ 5 = 0.12. These are programme judgements with ranges, not observed probabilities. They explain why the recommended first programme joins RAS resistance to residual disease even though detection ranks second by eventual mission impact. What would overturn the order: lead-time-corrected evidence that most screen-found stage I PDAC is cured by current care; a prevention target covering a large population fraction; failure to find any repeated RAS escape class; or a randomized PDAC result showing that residual-disease action does not help.", "sourceIds": [ "conroy-2022-prodige24-five-year", "wainberg-2023-napoli3", "seer-pancreas-statfacts", "strickler-2022-sotorasib-pdac", "oreilly-2026-daraxonrasib", "botta-2024-tumor-informed-ctdna", "tie-2022-dynamic-colon", "oreilly-2019-durvalumab-tremelimumab", "hu-2018-germline-risk", "chang-1997-hbv-vaccine-hcc", "nhs-england-2026-r367", "wang-2022-oncology-germline-cascade", "katz-2026-gift-cascade", "npaca-2026-state-of-nation", "nhs-england-2024-hpb-pancreatic-service", "nice-2018-ng85-pancreatic", "de-wilde-2012-centralization", "lemmens-2011-centralization", "latenstein-2021-centralization", "coupland-2016-england-resection-volume", "labori-2016-adjuvant-completion", "henry-2023-complications-adjuvant", "kollbeck-2025-complications-chemo", "van-hilst-2019-leopard2", "korrel-2023-diploma", "bruna-2025-diploma-long-term", "klotz-2024-europa", "liu-2024-robotic-open-pd", "jin-2026-portal" ], "links": [ { "target": "bottlenecks", "relation": "part-of" }, { "target": "claim-adjuvant-mfolfirinox-improves-but-recurrence-remains", "relation": "depends-on" }, { "target": "claim-stage-shift-insufficient", "relation": "depends-on" }, { "target": "claim-kras-g12c-inhibition-proves-targetability-not-general-control", "relation": "depends-on" }, { "target": "claim-daraxonrasib-common-ras-survival-breakthrough", "relation": "depends-on" }, { "target": "unknown-ctdna-actionability", "relation": "depends-on" }, { "target": "failure-checkpoint-blockade-unselected-pdac", "relation": "depends-on" }, { "target": "claim-germline-risk-subset", "relation": "depends-on" }, { "target": "programme", "relation": "ranks" }, { "target": "claim-universal-germline-testing-needs-action", "relation": "depends-on" }, { "target": "failure-pdac-germline-route-dropout", "relation": "depends-on" }, { "target": "hypothesis-pdac-germline-family-route-completion", "relation": "ranks" }, { "target": "claim-pdac-surgery-is-a-complete-treatment-route", "relation": "links-to" }, { "target": "system-england-pancreatic-surgery-network", "relation": "links-to" }, { "target": "failure-pdac-postoperative-treatment-loss", "relation": "links-to" }, { "target": "failure-leopard2-laparoscopic-whipple-safety", "relation": "links-to" }, { "target": "unknown-pdac-surgical-route-completion", "relation": "links-to" }, { "target": "trial-pdac-surgical-route-action-map", "relation": "links-to" }, { "target": "hypothesis-pdac-surgical-route-completion", "relation": "links-to" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "strong", "randomised": "strong", "independentReplication": "strong", "sampleSize": "Synthesis across population registry, randomised treatment trials, prospective surveillance and biomarker cohorts", "effect": "Mission-impact indices: systemic control 25, detection 16, MRD action 12, immune or delivery classification 8, prevention 6. First-£25m information indices: resistance 0.44, MRD 0.35, detection 0.20, immune or delivery 0.19, prevention 0.12", "limits": [ "Planning ranks rather than clinical effect estimates", "SEER pancreatic data mix histologies", "Upper-bound stage-shift calculation is not causal", "Reach, outcome and probability inputs are judgement", "Capabilities overlap", "Scores should be recomputed when current trials report" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The ranking follows observed recurrence, metastatic failure and common-RAS treatment response in people." }, "animalEvidence": { "rating": "not-applicable", "reason": "The ranking rests on human outcomes rather than requiring an animal result." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Laboratory activity is not used to set the five mission-impact ranks." }, "reproducibility": { "rating": "moderate", "reason": "The component findings repeat, but their ordering is a programme judgement rather than a reproduced experiment." }, "commercialConflicts": { "rating": "moderate", "reason": "The synthesis includes manufacturer-funded RAS evidence alongside public trials and registries." }, "recency": { "rating": "current", "reason": "The ranking includes the 2026 common-RAS phase 3 result and current mortality evidence." } } }, "layout": { "x": 50, "y": 57 }, "updated": "2026-09-14", "contentSections": [ "Surgical-route completion enters the delivery tier, below durable systemic control and early interception but above another marginal procedure comparison. England already has a 23-centre specialist base and 96.4% 90-day survival after Whipple; the repairable signal is that only 62% receive postoperative treatment within 14 weeks and full-course completion is unreported. A £2.5m route experiment has higher expected information value than buying robots because either result changes staffing, audit and referral rules across the existing network." ] }, { "id": "reading-guide", "title": "How to read the atlas without learning medicine first", "shortTitle": "Reading guide", "type": "system", "status": "supported", "scope": [ "pdac", "pnet", "cholangiocarcinoma", "gallbladder", "hcc", "colorectal", "pan-cancer" ], "summary": "Treat each cancer route as a measured control system: the disease has hidden state, tests are imperfect sensors, treatments are actuators, trials estimate what the actuator changes, and delivery determines how much of that result reaches the population.", "content": "Start with the unit of analysis. A cancer is not one object and a treatment is not one command. The useful model is a distributed system with changing clones, different organs, a supporting tissue environment and a patient whose reserve constrains every action. A scan sees macroscopic structure. A blood marker samples a noisy output. A biopsy samples one place at one time. Each sensor has a detection limit and a false-alarm rate. No sensor result is useful until it is connected to a decision that improves survival, function or suffering.\n\nRead treatment results as a chain, not a headline. Biological efficacy asks whether the intervention can change the target. Eligibility asks how many patients have that target and are fit enough to receive it. Route completion asks how many actually receive the intended test, referral, drug, operation and follow-up on time. Population effect is approximately efficacy × eligibility × completion. For example, 60% efficacy × 40% eligibility × 50% completion = 12% of the starting population affected. This is why a strong drug result in a narrow, selected group can coexist with little population change.\n\nRead endpoints as different outputs. Overall survival is time until death from any cause and is the hardest patient outcome. Progression-free survival is time until measured growth or death; it can improve without extending life. Response rate is the fraction whose measured tumour shrinks by a stated rule; shrinkage can be brief and is not cure. Disease-free survival after local treatment is time without detected recurrence or death. Patient-reported outcomes measure how people feel or function. The atlas does not exchange an earlier output for a later one without direct evidence that the substitution holds in that disease and setting.\n\nRead study design like system identification. A prospective study declares what it will collect and follows forward. A retrospective study queries records already produced by care, so missing fields and selection can be part of the result. Random assignment is an A/B allocation intended to balance known and unknown causes. It does not repair loss after assignment, treatment switching or selective reporting. Intention-to-treat analysis keeps people in their assigned group and protects the comparison created by randomization. A single-arm study can show that something happened after treatment but usually cannot identify what would have happened without it.\n\nRead numbers in absolute as well as relative units. A hazard ratio compares event rates through follow-up; it is not the probability that one person benefits. A 95% confidence interval describes the range of effects compatible with the data and model at that confidence rule; width is the precision signal. A p value does not measure importance and does not give the probability that the claim is true. Median survival is the time by which half the group has had the event; it is not an individual forecast or the mean lifespan. Always ask for the denominator, the absolute difference, the follow-up time and who was missing.\n\nRead diagnosis through base-rate maths. Sensitivity is the true-positive catch rate among people who have the condition. Specificity is the true-negative rejection rate among people who do not. Positive predictive value asks what fraction of positive results are real, and it changes sharply with prevalence. With prevalence p, sensitivity s and specificity c: PPV = s×p / (s×p + (1-c)×(1-p)). At low prevalence, a small false-positive fraction can overwhelm the true cases. A diagnostic claim is therefore incomplete without the intended population and the action caused by a positive result.\n\nRead common clinical route words literally. Resectable means surgeons judge that visible disease can be removed; it does not mean systemic disease is absent. Neoadjuvant treatment occurs before an operation. Adjuvant treatment occurs after it. Metastatic disease has established new tumours away from the original site. Minimal residual disease means tumour material is inferred below ordinary imaging, often through circulating tumour DNA, or ctDNA: small DNA fragments shed into blood. An organoid is a small laboratory-grown model made from a patient's tumour. It is a sandbox instance, not the live system; build success, time, sampling and proof that its recommendation helps patients all matter.\n\nRead the investment pages as staged capital allocation. A release gate states what evidence must exist before more money moves. A stop rule states what failure ends or changes the work. A programme that cannot name its next decision, measurement and stopping condition is activity, not a decision system. Costs and probabilities in the atlas are planning judgements unless a source states them; clinical effects and trial counts remain tied to their source records.", "sourceIds": [ "nci-cancer-terms-dictionary", "fda-2018-oncology-endpoints", "hopewell-2025-consort" ], "links": [ { "target": "mission", "relation": "explains" }, { "target": "evidence-quality", "relation": "explains" }, { "target": "disease-system", "relation": "explains" }, { "target": "programme", "relation": "explains" } ], "evidence": { "humanProspective": "not-applicable — this is a reading aid", "humanRetrospective": "not-applicable — this is a reading aid", "randomised": "not-applicable — this is a reading aid", "independentReplication": "strong — definitions follow established public standards", "sampleSize": "Not applicable", "effect": "No patient-benefit claim; the engineering and business comparisons are teaching aids", "limits": [ "Comparisons simplify biology", "The term box does not replace clinical advice", "Definitions do not prove that any test or treatment works" ], "dimensions": { "biologicalPlausibility": { "rating": "not-applicable", "reason": "The page explains how to read evidence rather than proposing a biological mechanism." }, "animalEvidence": { "rating": "not-applicable", "reason": "No animal claim is made." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "No laboratory-effect claim is made." }, "reproducibility": { "rating": "strong", "reason": "The clinical terms and trial-reporting distinctions come from maintained NCI, FDA and CONSORT sources." }, "commercialConflicts": { "rating": "none", "reason": "The definition anchors are public or academic standards, not product marketing." }, "recency": { "rating": "current", "reason": "The page combines the current NCI dictionary, FDA endpoint guidance and CONSORT 2025 reporting standard." } } }, "updated": "2026-09-15" }, { "id": "recommended-strategy", "title": "Recommended strategy: build the residual-disease and resistance control loop around common-RAS treatment", "shortTitle": "What to do next", "type": "programme", "status": "hypothesis", "scope": [ "pdac", "colorectal", "pan-cancer", "transfer-unproven" ], "summary": "Spend the first £25m measuring when and how PDAC escapes; release the next £75m only when a reproducible treatment rule is ready for randomized testing.", "content": "Problem to attack: occult systemic PDAC and acquired resistance cause death after apparently successful surgery and after initially active systemic treatment. Why this problem: 73.9% of selected resected patients had recurrence or death within five years despite mFOLFIRINOX, while daraxonrasib proved broad RAS dependence in humans but median progression still occurred at 7.3 months. Scientific thesis: deep common-RAS suppression will have its largest curative reach at minimal burden, but only if serial molecular and tissue measurements expose the escape route early enough to apply a matched, tolerable intervention. First experiments: a 200-patient locked feasibility cohort with baseline, week-four, progression and postoperative serial blood; matched tissue where safe; two-laboratory resistance classification; ctDNA, CA19-9, imaging and immune-state concordance; and clinically achievable combination testing in human-derived models. Required technology: trial-grade common-RAS inhibitors, duplex or tumour-informed DNA measurement, spatial pathology, rapid organoid/co-culture perturbation, and auditable longitudinal data. Required data: complete treatment exposure, toxicity, imaging, recurrence, survival and sample timing; no retrospective convenience labels. People: an independent chair and statistician; O'Reilly for trials, Iacobuzio-Donahue for evolution, Aguirre and Park for RAS resistance, Chari and Goggins for risk and assays, Tuveson for models, Balachandran for low-burden immunity, and Barzilay for model-to-experiment work. Likely partners: MSK, Johns Hopkins, Dana-Farber, CSHL, Penn, NHS and European centres; Revolution Medicines or Astellas only under enforceable compound, data and negative-publication terms. Capital: £25m for the four-year observatory; release up to £75m more for randomized action trials only after gates. Milestones: six months—contracts and assay lock; 18 months—80% serial-sample completeness and cross-lab concordance; 30 months—one reproduced resistance or MRD treatment rule; four to six years—randomized phase 2 result; eight to ten years—mature survival. Failure criteria: sample completeness below 60%; assay disagreement above 20%; no resistance class covering 25% of progression; no target engagement at safe exposure; MRD action futility or excess serious harm; or partner control that prevents independent analysis. Planning probability: 25–40% of one validated trial selector from the first £25m; 15–30% of a phase-3-ready strategy after £100m; below 10–15% that this alone makes most PDAC durably controllable within ten years. Success unlocks rational first-line and post-resection combinations, a treatment-linked MRD system and a stronger reason to invest in detection, because finding disease earlier would then lead to an intervention capable of eradicating it.", "sourceIds": [ "gudmundsdottir-2023-staging-laparoscopy", "rompen-2026-preopanc2-staging", "conroy-2022-prodige24-five-year", "oreilly-2026-daraxonrasib", "botta-2024-tumor-informed-ctdna", "sarno-2025-avatar", "tie-2022-dynamic-colon", "varghese-2025-kras-dosage", "andel-2025-routine-imaging", "van-goor-2025-nationwide-imaging", "daamen-2024-radar-panc-protocol", "clinicaltrials-nct04875325-radar-panc", "tai-2025-pan-asian-esmo-pancreatic", "labori-2024-norpact1", "bai-2025-cispd1", "tan-2026-neoadjuvant-ipd-meta", "ghaneh-2023-espac5", "clinicaltrials-nct04340141-a021806", "effenberger-2012-pdac-bone-marrow-dtc", "nordgard-2022-unresectable-dtc", "chalabi-hajkarim-2025-organotropism", "pei-2025-spatial-metastatic-pdac", "mcclellan-2026-quiescent-pdac-car-t", "tanaka-2012-microscopic-local-recurrence", "dotan-2024-giant", "dotan-2025-giant-ga", "clinicaltrials-nct04233866-giant", "betge-2022-grantpax", "clinicaltrials-nct02143219-pamela70", "mohile-2021-gap70", "li-2021-gain", "ihorst-2026-geriatric-comanagement", "clinicaltrials-nct06040801-frail-gi", "stoop-2026-preopanc4-surgery", "dekker-2026-taps-restaging", "guggenberger-2023-neolap-ct", "seelen-2026-pelican", "baum-2022-fap2286-firsthuman", "mcconathy-2026-lumiere-phase1", "clinicaltrials-nct04939610-lumiere", "baum-2026-3bp3940", "clinicaltrials-nct05432193-frontier", "clinicaltrials-nct07229768-ctr-fapi", "alistar-2017-devimistat-phase1", "philip-2024-avenger500", "surana-2025-erk-hcq-pdac", "clinicaltrials-nct05221320-ulixertinib-hcq", "manji-2023-mekiauto", "clinicaltrials-nct04214418-mekiauto", "clinicaltrials-nct04892017-inlexisertib", "deciphera-2022-dcc3116-phase1", "clinicaltrials-nct04524702-paricalcitol-hcq", "dimcevski-2016-sonoporation-phase1", "wang-2023-sonochemotherapy-pdac", "clinicaltrials-nct04821284-sonoporation", "ius-2026-sonoporation-nct04821284-abstract", "adler-2026-sonoporation-perfusion", "alagesan-2026-sig12d-loder", "ctis-2026-sil204-phase23", "silexion-2026-sil204-site-initiation", "silexion-2026-halfyear-filing", "bazan-peregrino-2021-vcn01-intratumour", "garcia-carbonero-2022-vcn01-intravenous", "clinicaltrials-nct05673811-virage", "theriva-2023-virage-protocol", "theriva-2025-virage-topline", "esmo-2025-virage-2216mo", "ctis-2026-virage2", "theriva-2026-q2-10q", "richards-2006-ci994-pdac", "heumann-2022-azacitidine-pdac", "sohal-2020-thu-decitabine-pdac", "baretti-2024-entinostat-nivolumab-pdac", "clinicaltrials-nct03250273-entinostat-nivolumab", "safyan-2026-azacitidine-pembrolizumab-pdac", "clinicaltrials-nct03264404-azacitidine-pembrolizumab", "rodon-2024-amg193-prmt5", "matsumoto-2026-mtap-pancreatic", "pavlick-2026-mtap-genomics", "clinicaltrials-nct05732831-vopimetostat", "tango-2025-vopimetostat-monotherapy", "clinicaltrials-nct06922591-vopimetostat-ras", "tango-2026-vopimetostat-ras-data", "tango-2026-q2-10q", "mahalingam-2026-elraglusib-randomized", "clinicaltrials-nct03678883-elraglusib", "pathak-2026-elraglusib-folfirinox", "clinicaltrials-nct05077800-elraglusib-folfirinox", "actuate-2026-q2-10q", "sec-2026-actuate-submissions", "chiorean-2023-abemaciclib-pdac", "baghdadi-2019-palbociclib-cdkn2a", "hidalgo-2022-palbociclib-nabpaclitaxel", "laquente-2017-chk1-pdac", "huffman-2023-ly2880070-pdac", "cuneo-2019-adavosertib-pdac", "huffman-2024-azenosertib-pdac-design", "clinicaltrials-nct06015659-azenosertib" ], "links": [ { "target": "hypothesis-organoid-guided-early-switch", "relation": "gates" }, { "target": "trial-pdac-organoid-action-map", "relation": "depends-on" }, { "target": "hypothesis-pdac-two-timepoint-staging-route", "relation": "includes" }, { "target": "mission", "relation": "answers" }, { "target": "ranking-provisional-pdac-bottlenecks", "relation": "implements" }, { "target": "programme-25m-prospective-observatory", "relation": "starts-with" }, { "target": "programme-100m-randomized-platform", "relation": "advances-to" }, { "target": "unknown-durable-ras-control", "relation": "tests" }, { "target": "unknown-ctdna-actionability", "relation": "tests" }, { "target": "risks", "relation": "bounded-by" }, { "target": "timeline", "relation": "scheduled-by" }, { "target": "programme-mission-table-people", "relation": "staffed-by" }, { "target": "claim-routine-imaging-after-pdac-resection-not-yet-proven", "relation": "bounded-by" }, { "target": "unknown-postresection-surveillance-utility", "relation": "bounded-by" }, { "target": "trial-pdac-postresection-surveillance-action-map", "relation": "bounded-by" }, { "target": "claim-neoadjuvant-therapy-improves-selection-not-cure", "relation": "bounded-by" }, { "target": "failure-norpact1-neoadjuvant-delivery", "relation": "bounded-by" }, { "target": "unknown-pdac-perioperative-treatment-route", "relation": "bounded-by" }, { "target": "trial-pdac-perioperative-sequencing-action-map", "relation": "bounded-by" }, { "target": "claim-human-pdac-dtc-prognostic-not-dormancy", "relation": "bounded-by" }, { "target": "failure-pdac-dormancy-sample-gap", "relation": "bounded-by" }, { "target": "trial-pdac-residual-cell-state-action-map", "relation": "bounded-by" }, { "target": "hypothesis-human-dormancy-reawakening-gate", "relation": "bounded-by" }, { "target": "claim-age-alone-cannot-route-pdac-treatment", "relation": "bounded-by" }, { "target": "unknown-older-pdac-treatment-route", "relation": "tests" }, { "target": "trial-older-pdac-action-map", "relation": "bounded-by" }, { "target": "hypothesis-older-pdac-fast-support-route", "relation": "implements" }, { "target": "claim-lapc-response-warrants-surgical-reassessment", "relation": "implements" }, { "target": "unknown-lapc-surgery-causal-benefit", "relation": "leaves-open" }, { "target": "trial-lapc-conversion-surgery-action-map", "relation": "bounded-by" }, { "target": "hypothesis-randomized-lapc-resection-after-response", "relation": "could-advance-to" }, { "target": "claim-fap-radioligand-feasible-pdac-efficacy-unproven", "relation": "bounded-by" }, { "target": "unknown-pdac-fap-radioligand-treatment-effect", "relation": "leaves-open" }, { "target": "trial-pdac-fap-radioligand-action-map", "relation": "implements" }, { "target": "hypothesis-fapi-theranostic-pdac", "relation": "tests" }, { "target": "claim-hcq-combinations-fail-pdac-not-autophagy", "relation": "bounded-by" }, { "target": "unknown-selective-autophagy-pdac-contribution", "relation": "leaves-open" }, { "target": "trial-pdac-autophagy-action-map", "relation": "implements" }, { "target": "hypothesis-tumour-verified-selective-autophagy-blockade", "relation": "tests" }, { "target": "claim-sonochemotherapy-randomized-signal", "relation": "bounded-by" }, { "target": "unknown-pdac-sonoporation-treatment-effect", "relation": "leaves-open" }, { "target": "trial-pdac-sonoporation-action-map", "relation": "implements" }, { "target": "hypothesis-measured-fixed-backbone-sonoporation", "relation": "tests" }, { "target": "claim-sil204-trial-scales-before-human-knockdown", "relation": "bounded-by" }, { "target": "unknown-sil204-human-target-delivery", "relation": "leaves-open" }, { "target": "trial-sil204-action-map", "relation": "implements" }, { "target": "hypothesis-multicompartment-kras-sirna-delivery", "relation": "tests" }, { "target": "claim-vcn01-randomized-signal-not-survival-proof", "relation": "bounded-by" }, { "target": "failure-virage-analysis-and-survivor-selection", "relation": "learns-from" }, { "target": "unknown-vcn01-all-enrolled-survival", "relation": "leaves-open" }, { "target": "trial-vcn01-action-map", "relation": "implements" }, { "target": "hypothesis-vcn01-replication-to-survival-gate", "relation": "tests" }, { "target": "company-theriva-biologics", "relation": "contains" }, { "target": "failure-broad-epigenetic-treatment-pdac", "relation": "learns-from" }, { "target": "claim-mtap-prmt5-selected-human-signal", "relation": "bounded-by" }, { "target": "failure-vopimetostat-ras-contribution-unknown", "relation": "learns-from" }, { "target": "unknown-mtap-prmt5-ras-contribution", "relation": "leaves-open" }, { "target": "trial-mtap-prmt5-action-map", "relation": "implements" }, { "target": "hypothesis-mtap-prmt5-ras-deeper-control", "relation": "tests" }, { "target": "company-tango-therapeutics", "relation": "contains" }, { "target": "claim-elraglusib-randomized-survival-signal", "relation": "bounded-by" }, { "target": "failure-elraglusib-analysis-and-survival-discordance", "relation": "learns-from" }, { "target": "unknown-elraglusib-survival-replication", "relation": "leaves-open" }, { "target": "trial-elraglusib-action-map", "relation": "implements" }, { "target": "hypothesis-elraglusib-survival-confirmation-gate", "relation": "tests" }, { "target": "company-actuate-therapeutics", "relation": "contains" }, { "target": "failure-broad-cell-cycle-checkpoint-pdac", "relation": "learns-from" }, { "target": "unknown-wee1-replication-stress-pdac", "relation": "leaves-open" }, { "target": "trial-cell-cycle-checkpoint-action-map", "relation": "implements" }, { "target": "hypothesis-human-replication-stress-selector", "relation": "tests" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "strong", "randomised": "strong", "independentReplication": "moderate", "sampleSize": "Strategy synthesized from randomized systemic, adjuvant and MRD-transfer evidence; proposed feasibility n=200", "effect": "Generated strategy; no direct efficacy claim", "limits": [ "Planning costs and probabilities are judgement", "Compound and data access not negotiated", "Resistance may be too diverse", "MRD may remain non-actionable", "New trial results can change the priority" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "Human RAS response and recurrent disease support the thesis, but the full resistance-and-MRD loop is untested." }, "animalEvidence": { "rating": "weak", "reason": "Models support individual resistance routes but not the proposed clinical system as a whole." }, "inVitroEvidence": { "rating": "moderate", "reason": "Human-derived models can test combinations, but model response has not yet predicted this treatment route." }, "reproducibility": { "rating": "none", "reason": "No group has reproduced the complete measurement, routing and intervention programme." }, "commercialConflicts": { "rating": "high", "reason": "Execution depends on company-controlled compounds, data and combination access." }, "recency": { "rating": "current", "reason": "The recommendation responds directly to the 2026 common-RAS result and current MRD limits." } } }, "layout": { "x": 60, "y": 121 }, "updated": "2026-09-15", "contentSections": [ "Start the surgical route with a complete hidden-spread state. Record who receives early staging before treatment and who receives a same-session check before incision; keep visible spread, biopsy and wash cytology separate. The immediate operational target is fewer open explorations that cannot remove the cancer without delaying effective treatment. Survival remains a later outcome, not an inference from detection.", "Surveillance is subordinate to treatment utility. RADAR-PANC already tests whether scheduled CA19-9 plus CT improves survival and quality of life; CIRCPAC tests whether ctDNA should intensify scans. Recover both results and align their outcomes before a new study. An earlier recurrence date without longer or better life is not a mission success.", "Do not let treatment timing become a substitute for systemic control. Across nine randomized resectable-disease trials, event-free survival improves but overall survival remains inconclusive. Finish A021806, recover full-route data, and use its modern same-regimen comparison to decide timing. A new trial is justified only for a patient state or treatment absent from that study.", "Treat dormancy as a sample-and-action problem, not a label for every late recurrence. Human marrow assays predict risk but do not identify the lethal resting cell. Use the observatory to link primary, serial blood, recurrence and autopsy by ancestry, and require added value beyond stage plus ctDNA. Do not buy an anti-dormancy drug programme before that chain exists.", "Make the older-and-vulnerable route an immediate safety dependency. Record every person before the final treatment decision, explain the exact functional or support risk, complete named actions within 72 hours and keep untreated patients in the denominator. The aim is to convert treatment activity into useful life with less severe harm; it cannot rescue an ineffective tumour drug.", "For initially locally advanced disease, make reassessment universal and surgery conditional. After at least four months without distant progression, send the full response state to an independent expert panel; persistent vessel contact on CT is not a permanent veto. Explain that current evidence predicts technical removal and prognosis but not the operation's added useful life. Preserve the 540-person randomized test as a portfolio option, but fund it only by stopping or reducing at least £22m of lower-value work.", "Use organoids first as a drug-rejection tool and a shadow treatment selector. Recover the missing randomized Chinese studies, join PASS-01 and active Canadian and German routes, and count every patient from biopsy. No organoid result changes care until timing, identity, two-laboratory agreement, drug access and treatment interaction pass. A later 480-person policy trial remains unfunded and must displace £18m elsewhere.", "Treat FAP-targeted radiation as a data-access option, not a new mission programme. Request LuMIERE lesion dose and response data and FRONTIER results; count every failed screen. NCT04939610 already runs the PDAC single-arm response and combination-safety work. No new cohort, licence or investment follows scan brightness or stable disease. A later controlled trial can compete for capital only after lesion dose-response reproduces, standard chemotherapy dose intensity is preserved and equal money leaves a lower-value branch.", "Treat autophagy as a failed-agent branch with one recoverable information asset. Do not run another HCQ combination. Obtain all 91 NCT04892017 records and determine whether pancreatic combination exposure and paired tumour measurements exist. A selective replacement must show direct tumour target suppression, blocked-recycling cargo and preserved RAS-drug dose before any response expansion. A later randomized add-on trial competes for, rather than expands, the fixed portfolio.", "Treat sonoporation as a completed-data obligation, not a device programme. Obtain the full NCT04821284 intended-to-treat result, reconcile 97 reported with 120 estimated, and test the planned treatment-by-backbone interaction. Keep its 13-person perfusion signal separate from drug delivery. No new cohort, licence or investment follows the incomplete gemcitabine subgroup. A fixed-backbone measurement study can compete only after the interaction passes and must displace equal capital.", "Treat SIL204 as a live data-access option, not an investment. The first-generation product missed overall survival and the second-generation product has no reported human target result. Offer independent multi-region tumour pharmacology and statistics only if participant flow, mutation, target, dose, harm and negative results become public. Use the existing 15-to-21-person safety segment, reconcile the 166 and 403 descriptions, and stop before expansion if knockdown, spatial coverage, systemic exposure or chemotherapy delivery fails.", "VCN-01 does not displace the central RAS, resistance and residual-disease strategy. It enters as a £0 data-recovery branch: obtain the complete 112-person VIRAGE record and use the six-person repeat-dose study only for safety and pharmacology. A clean all-randomized estimate and blinded phase 3 design may later compete for capital, but only through equal displacement and public-data rights.", "MTAP-selected PRMT5 work adds no cohort or budget. The route is biologically and commercially credible, but Tango and its partners already fund a 225-person monotherapy programme and a 183-person combination programme. Require all 54 treated PDAC records to mature, exact MTAP-loss assays and a randomized RAS-inhibitor-alone contribution arm. Existing mission teams may review assays and analysis only under public-data terms.", "Elraglusib enters as a £0 data-recovery branch above other unconfirmed treatment signals. Request every NCT03678883 randomized record and the dated endpoint-change history under the paper's access promise. Reanalyse all randomized survival, missing outcomes, site, region, later treatment and quality-adjusted life. A blinded phase 3 may compete for capital only if HR stays at most 0.75 under the audit and a solvent partner funds manufacturing and most trial cost.", "Close broad cell-cycle work. CDK4/6 and CHK1 results are negative despite common pathway faults, selected patients, adequate exposure and model sensitivity. Keep only a £0 WEE1 data-recovery task: publish all 12 completed azenosertib records and paired tumours. No new trial or company investment follows unless tumour checkpoint change, six-month control, safety and a reproducible selector all pass." ] }, { "id": "regulatory-bizengri-nrg1-cholangiocarcinoma", "title": "FDA approved zenocutuzumab for NRG1-fusion cholangiocarcinoma on a 19-person response set", "shortTitle": "NRG1 CCA approval", "type": "regulatory", "status": "supported", "scope": [ "cholangiocarcinoma", "nrg1", "rare-driver", "treatment" ], "summary": "The May 2026 US approval creates an available route for a very rare molecular group; 7 of 19 people responded, without a randomized survival comparison.", "content": "On 8 May 2026 the FDA approved zenocutuzumab-zbco, branded Bizengri, for adults with advanced unresectable or metastatic cholangiocarcinoma containing an NRG1 gene fusion after earlier systemic treatment. NRG1 is a gene-fusion driver that forces growth signalling through the HER3-HER2 receptor pair; the two-target antibody blocks that pair. The evidence set contained 19 people in a single-arm study. Seven responded, 36.8%, and response duration ranged from 2.8 to 12.9 months. The approval changes access and makes RNA-capable fusion detection an immediate care-route requirement. It does not prove an overall-survival gain, does not show that most treated people obtain durable control, and does not transfer to cholangiocarcinoma without the fusion. The same drug already had a pancreatic NRG1-fusion indication, so the reusable asset is the testing-to-referral network rather than a broad tumour label.", "sourceIds": [ "fda-2026-bizengri-cholangiocarcinoma", "fda-2024-bizengri-multidisciplinary-review", "schram-2025-zenocutuzumab-nrg1" ], "links": [ { "target": "cholangiocarcinoma", "relation": "part-of" }, { "target": "claim-nrg1-approved-response-not-durable-control", "relation": "extends" }, { "target": "company-genmab-partner-bizengri", "relation": "involves" }, { "target": "hypothesis-cca-route-specific-action-platform", "relation": "informs" }, { "target": "trial-pdac-nrg1-action-map", "relation": "shares-testing-route-with" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "not-applicable", "randomised": "none", "independentReplication": "weak", "sampleSize": "n=19 single-arm cholangiocarcinoma efficacy set", "effect": "Objective response 36.8%; response duration 2.8-12.9 months", "limits": [ "Single arm", "Nineteen people", "Response endpoint", "No comparative survival", "Ultra-rare selected fusion", "Sponsor-linked development", "Does not apply without NRG1 fusion" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The fusion, receptor pair and blocking antibody form a direct target chain." }, "animalEvidence": { "rating": "moderate", "reason": "Models support target action but do not set the clinical claim." }, "inVitroEvidence": { "rating": "strong", "reason": "Target blocking is established; patient durability remains the limit." }, "reproducibility": { "rating": "weak", "reason": "The cholangiocarcinoma result comes from one small single-arm programme." }, "commercialConflicts": { "rating": "high", "reason": "The approval evidence comes from the drug-development programme." }, "recency": { "rating": "current", "reason": "FDA status checked 15 September 2026." } } }, "updated": "2026-09-15" }, { "id": "regulatory-daraxonrasib-first-common-ras-approval", "title": "FDA approved daraxonrasib for metastatic pancreatic adenocarcinoma after prior therapy", "shortTitle": "Daraxonrasib approval", "type": "regulatory", "status": "supported", "scope": [ "pdac" ], "summary": "The 26 August 2026 U.S. approval turns broad RAS inhibition from a research route into clinical practice, but only for metastatic disease after prior treatment or when multiagent therapy is unsuitable.", "content": "The FDA approved once-daily daraxonrasib, branded Rasonque, on 26 August 2026 for adults with metastatic pancreatic adenocarcinoma who had received at least one prior systemic therapy or were not candidates for multiagent systemic therapy. The label rests on RASolute 302. It does not require a RAS-G12-only population because the overall randomized population also showed the survival effect. Practical limits are substantial: dermatologic toxicity occurred in 86% and grade 3 dermatologic toxicity in 10%; stomatitis occurred in 57%, diarrhoea in 63%, and rare fatal gastrointestinal perforation and pneumonitis occurred. Approval confirms availability, not cure, long-term safety, first-line benefit or benefit after surgery.", "contentSections": [ "European access remains pending. EMA granted orphan designation on 20 April 2026 and began a phased review in July. Both actions can speed regulatory work; neither is a European marketing authorization, and EMA gives no fixed decision date. The US indication must not be described as EU availability." ], "sourceIds": [ "fda-2026-rasonque-approval", "ema-2026-daraxonrasib-phased-review", "ema-2026-daraxonrasib-orphan-designation", "oreilly-2026-daraxonrasib" ], "links": [ { "target": "claim-daraxonrasib-common-ras-survival-breakthrough", "relation": "extends" }, { "target": "treatment", "relation": "part-of" }, { "target": "timeline", "relation": "part-of" }, { "target": "companies", "relation": "part-of" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong", "independentReplication": "none", "sampleSize": "500-patient registration trial", "effect": "Full U.S. approval after prior therapy or when multiagent therapy is unsuitable", "limits": [ "Metastatic indication", "No curative claim", "No first-line or post-resection efficacy established", "Manufacturer-funded pivotal trial", "Common and sometimes severe toxicity", "Rare fatal gastrointestinal perforation and pneumonitis", "No EU marketing authorization as of 15 September 2026" ] }, "layout": { "x": 89, "y": 55 }, "updated": "2026-09-14" }, { "id": "regulatory-fda-optune-pax-2026", "title": "FDA approved Optune Pax for one locally advanced PDAC treatment pair", "shortTitle": "Optune Pax approval", "type": "regulatory", "status": "supported", "scope": [ "pdac", "locally-advanced", "device" ], "summary": "The February 2026 approval is narrow: adult locally advanced pancreatic cancer, with gemcitabine and nab-paclitaxel.", "content": "FDA approved premarket application P250034 on 11 February 2026. Optune Pax is a class III prescription device for adults with locally advanced pancreatic cancer used together with gemcitabine and nab-paclitaxel. It sends preset 150 kHz alternating electric fields across the abdominal cavity through insulated skin arrays. The physician and patient cannot change the field parameters. U.S. instructions say average use should be at least 12 hours per day and arrays should be replaced every three to four days. People with an electrical implant in the torso or a known sensitivity to electrode gels should not use it. Approval rests on PANOVA-3's all-randomized survival result, not the selected group who managed 28 days of device treatment. The legal and evidence boundary is exact: the approval does not prove use with FOLFIRINOX or NALIRIFOX, after surgery, in metastatic disease, or as a cure. FDA found a favourable benefit-risk balance in the approved setting despite no significant progression-free-survival gain and common skin injury. Regulatory approval answers whether marketing is allowed for that use. It does not answer independent replication, comparative affordability, who can sustain the device, or whether another treatment backbone preserves the effect.", "sourceIds": [ "fda-2026-optune-pax-ssed", "clinicaltrials-nct03377491-panova3", "macarulla-2025-panova3" ], "links": [ { "target": "claim-panova3-ttfields-small-os-benefit", "relation": "implements" }, { "target": "treatment", "relation": "part-of" }, { "target": "timeline", "relation": "part-of" }, { "target": "company-novocure", "relation": "held-by" }, { "target": "trial-pdac-ttfields-action-map", "relation": "constrains" }, { "target": "regulatory-daraxonrasib-first-common-ras-approval", "relation": "compares-with" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong", "independentReplication": "none", "sampleSize": "571-person pivotal trial; 547-person safety population", "effect": "U.S. approval for one adult locally advanced disease and chemotherapy setting", "limits": [ "No metastatic indication", "No use with another backbone established", "No curative claim", "One sponsor-run pivotal trial", "Common skin toxicity", "Daily wear and maintenance burden", "Approval does not establish cost-effectiveness" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "The device has measurable physical output and cell evidence, while the clinical pathway remains partly unresolved." }, "animalEvidence": { "rating": "moderate", "reason": "Animal evidence contributed to the safety case, not the clinical effect estimate." }, "inVitroEvidence": { "rating": "strong", "reason": "Cell work supports the selected 150 kHz field and division effect." }, "reproducibility": { "rating": "none", "reason": "No independent pancreatic randomized trial has reproduced the survival result." }, "commercialConflicts": { "rating": "high", "reason": "The manufacturer submitted and sponsored the evidence package." }, "recency": { "rating": "current", "reason": "The approval and current instructions were checked in September 2026." } } }, "layout": { "x": 622, "y": 546 }, "updated": "2026-09-15" }, { "id": "research-frontier-ranking", "title": "Research frontier: 23 ideas ranked by mission impact and testability", "shortTitle": "Research frontier", "type": "bottleneck", "status": "open", "scope": [ "pdac", "pnet", "cholangiocarcinoma", "gallbladder", "hcc", "colorectal", "pan-cancer", "transfer-unproven" ], "summary": "The top frontier is not one magic drug: it is a linked system for RAS control, resistance measurement, MRD action and risk-enriched detection.", "content": "Ranked for the next programme decision, not by publication volume. 1: prospective common-RAS resistance routing; high impact and a £5m human atlas can fail fast. 2: multimodal MRD action after PDAC surgery; central to recurrence but requires both a better state estimate and an active treatment. 3: move common-RAS suppression into first-line and residual disease; largest current treatment effect, with toxicity and resistance as hard limits. 4: recover and independently audit the elraglusib survival result, then run a blinded all-randomized confirmation only if the phase 2 record survives; peer-reviewed ITT OS favours treatment but PFS and response do not. 5: isolate the added effect of MTAP-selected PRMT5 inhibition beside RAS blockade; the route reaches about 28% of PDAC and has a large early response signal, but no RAS-drug-alone control. 6: blinded fast organoid rejection of ineffective regimens; useful only if results arrive before cycle two and change care. 7: patient-specific immune-exclusion measurement tied to a small intervention menu; generic checkpoint combinations should not be repeated. 8: risk-triggered one-time pancreas work-up for new-onset diabetes or other temporary warning states. 9: HCC tumour and liver-reserve co-control, because early HCC still causes roughly one non-cancer liver death for every two cancer deaths and advanced-treatment survival collapses with impaired reserve. 10: human validation of a liver pre-metastatic niche before any interception drug. 11: test whether a treatment-linked human reawakening state predicts liver recurrence beyond ctDNA before giving an anti-dormancy drug. 12: recover and independently reproduce VCN-01 VIRAGE, then run a blinded all-randomized survival confirmation only if the 112-person record survives. 13: FAP-radioligand dose-response from existing human data, not another scan or uncontrolled treatment cohort. 14: local consolidation only after a locked systemic-risk classifier. 15: rapid personalised neoantigen manufacturing at residual burden, conditional on randomized benefit. 16: opportunistic analysis of existing abdominal CT, validated across scanners, health systems and diagnosis windows. 17: quantitative precursor progression rules for IPMN/PanIN that separate lineage from danger. 18: G12D degrader plus rational resistance partner, after the phase 3 first-line result. 19: radiopharmaceutical expansion in selected pNET, where randomized control already exists but cure does not. 20: prevention by smoking, obesity, diabetes and pancreatitis control; population value is real but it cannot remove most PDAC. 21: tumour-confirmed selective autophagy blockade after RAS treatment; current HCQ combinations failed and the direct-inhibitor dataset is missing. 22: measured fixed-backbone sonoporation; the overall replication is negative in an abstract, so first recover the completed trial and require a planned interaction plus direct tumour delivery. 23: multi-compartment KRAS siRNA delivery; use the live SIL204 safety segment to prove repeated human tumour knockdown and systemic coverage before its adaptive programme scales. The top five converge on the same missing capability: observe and extinguish the evolving systemic clone before clinical relapse. Ranking will change when first-line RAS, EDI, MRD-action, organoid-guidance and HCC co-control trials report.", "sourceIds": [ "rompen-2026-preopanc2-staging", "gudmundsdottir-2023-staging-laparoscopy", "oreilly-2026-daraxonrasib", "botta-2024-tumor-informed-ctdna", "beutel-2021-pdac-organoids", "oreilly-2019-durvalumab-tremelimumab", "chari-2026-heuriskance", "seif-el-dahan-2026-early-hcc-causes-death", "storandt-2024-atezo-bev-liver-function", "costa-silva-2015-liver-niche", "shi-2026-chemo-dormant-dtc", "lee-2024-prospective-fapi-resectable-pdac", "yoo-2026-saber", "rojass-2025-neoantigen-followup", "cao-2023-panda-ct-ai", "dbouk-2022-caps5", "park-2026-setidegrasib", "baudin-2026-oclurandom", "zhang-2022-uk-biobank-risk-fractions", "effenberger-2012-pdac-bone-marrow-dtc", "nordgard-2022-unresectable-dtc", "chalabi-hajkarim-2025-organotropism", "pei-2025-spatial-metastatic-pdac", "mcclellan-2026-quiescent-pdac-car-t", "tanaka-2012-microscopic-local-recurrence", "dotan-2024-giant", "dotan-2025-giant-ga", "clinicaltrials-nct04233866-giant", "betge-2022-grantpax", "clinicaltrials-nct02143219-pamela70", "mohile-2021-gap70", "li-2021-gain", "ihorst-2026-geriatric-comanagement", "clinicaltrials-nct06040801-frail-gi", "stoop-2026-preopanc4-surgery", "dekker-2026-taps-restaging", "seelen-2026-pelican", "clinicaltrials-nct06714604-duration", "baum-2022-fap2286-firsthuman", "mcconathy-2026-lumiere-phase1", "clinicaltrials-nct04939610-lumiere", "baum-2026-3bp3940", "clinicaltrials-nct05432193-frontier", "clinicaltrials-nct07229768-ctr-fapi", "alistar-2017-devimistat-phase1", "philip-2024-avenger500", "surana-2025-erk-hcq-pdac", "clinicaltrials-nct05221320-ulixertinib-hcq", "manji-2023-mekiauto", "clinicaltrials-nct04214418-mekiauto", "clinicaltrials-nct04892017-inlexisertib", "deciphera-2022-dcc3116-phase1", "clinicaltrials-nct04524702-paricalcitol-hcq", "dimcevski-2016-sonoporation-phase1", "wang-2023-sonochemotherapy-pdac", "clinicaltrials-nct04821284-sonoporation", "ius-2026-sonoporation-nct04821284-abstract", "adler-2026-sonoporation-perfusion", "alagesan-2026-sig12d-loder", "ctis-2026-sil204-phase23", "silexion-2026-sil204-site-initiation", "silexion-2026-halfyear-filing", "bazan-peregrino-2021-vcn01-intratumour", "garcia-carbonero-2022-vcn01-intravenous", "clinicaltrials-nct05673811-virage", "theriva-2023-virage-protocol", "theriva-2025-virage-topline", "esmo-2025-virage-2216mo", "ctis-2026-virage2", "theriva-2026-q2-10q", "richards-2006-ci994-pdac", "heumann-2022-azacitidine-pdac", "sohal-2020-thu-decitabine-pdac", "baretti-2024-entinostat-nivolumab-pdac", "clinicaltrials-nct03250273-entinostat-nivolumab", "safyan-2026-azacitidine-pembrolizumab-pdac", "clinicaltrials-nct03264404-azacitidine-pembrolizumab", "rodon-2024-amg193-prmt5", "matsumoto-2026-mtap-pancreatic", "pavlick-2026-mtap-genomics", "clinicaltrials-nct05732831-vopimetostat", "tango-2025-vopimetostat-monotherapy", "clinicaltrials-nct06922591-vopimetostat-ras", "tango-2026-vopimetostat-ras-data", "tango-2026-q2-10q", "mahalingam-2026-elraglusib-randomized", "clinicaltrials-nct03678883-elraglusib", "pathak-2026-elraglusib-folfirinox", "clinicaltrials-nct05077800-elraglusib-folfirinox", "actuate-2026-q2-10q", "sec-2026-actuate-submissions" ], "links": [ { "target": "trial-pdac-organoid-action-map", "relation": "ranks" }, { "target": "failure-organoid-treatment-chain-attrition", "relation": "constrained-by" }, { "target": "hypothesis-pdac-two-timepoint-staging-route", "relation": "ranks" }, { "target": "claim-kras-egfr-stat3-mouse-regression", "relation": "contains" }, { "target": "bottlenecks", "relation": "part-of" }, { "target": "hypothesis-prospective-ras-resistance-routing", "relation": "ranks" }, { "target": "hypothesis-multimodal-mrd-action", "relation": "ranks" }, { "target": "hypothesis-low-burden-ras-immune-extinction", "relation": "ranks" }, { "target": "hypothesis-organoid-guided-early-switch", "relation": "ranks" }, { "target": "claim-immune-exclusion-is-multicellular", "relation": "ranks" }, { "target": "hypothesis-one-time-risk-triggered-workup", "relation": "ranks" }, { "target": "hypothesis-hcc-liver-reserve-co-control", "relation": "ranks" }, { "target": "hypothesis-human-liver-niche-gate", "relation": "ranks" }, { "target": "hypothesis-human-dormancy-reawakening-gate", "relation": "ranks" }, { "target": "hypothesis-fapi-theranostic-pdac", "relation": "ranks" }, { "target": "hypothesis-selected-local-consolidation", "relation": "ranks" }, { "target": "claim-neoantigen-vaccine-frontier", "relation": "ranks" }, { "target": "technology-opportunistic-ct-ai-pancreas", "relation": "ranks" }, { "target": "claim-ipmn-mutations-identify-lineage-not-progression", "relation": "ranks" }, { "target": "claim-setidegrasib-g12d-degradation", "relation": "ranks" }, { "target": "claim-prrt-pnet-randomized-control", "relation": "ranks" }, { "target": "claim-modifiable-risk-reduces-but-cannot-eliminate-pdac", "relation": "ranks" }, { "target": "claim-human-pdac-dtc-prognostic-not-dormancy", "relation": "ranks" }, { "target": "failure-pdac-dormancy-sample-gap", "relation": "ranks" }, { "target": "trial-pdac-residual-cell-state-action-map", "relation": "ranks" }, { "target": "hypothesis-older-pdac-fast-support-route", "relation": "ranks" }, { "target": "claim-age-alone-cannot-route-pdac-treatment", "relation": "bounded-by" }, { "target": "trial-older-pdac-action-map", "relation": "bounded-by" }, { "target": "claim-lapc-response-warrants-surgical-reassessment", "relation": "bounded-by" }, { "target": "unknown-lapc-surgery-causal-benefit", "relation": "ranks" }, { "target": "hypothesis-randomized-lapc-resection-after-response", "relation": "ranks" }, { "target": "claim-fap-radioligand-feasible-pdac-efficacy-unproven", "relation": "ranks" }, { "target": "unknown-pdac-fap-radioligand-treatment-effect", "relation": "ranks" }, { "target": "trial-pdac-fap-radioligand-action-map", "relation": "bounded-by" }, { "target": "claim-hcq-combinations-fail-pdac-not-autophagy", "relation": "ranks" }, { "target": "unknown-selective-autophagy-pdac-contribution", "relation": "ranks" }, { "target": "hypothesis-tumour-verified-selective-autophagy-blockade", "relation": "ranks" }, { "target": "claim-sonochemotherapy-randomized-signal", "relation": "ranks" }, { "target": "unknown-pdac-sonoporation-treatment-effect", "relation": "ranks" }, { "target": "hypothesis-measured-fixed-backbone-sonoporation", "relation": "ranks" }, { "target": "claim-sil204-trial-scales-before-human-knockdown", "relation": "ranks" }, { "target": "unknown-sil204-human-target-delivery", "relation": "ranks" }, { "target": "hypothesis-multicompartment-kras-sirna-delivery", "relation": "ranks" }, { "target": "claim-vcn01-randomized-signal-not-survival-proof", "relation": "bounded-by" }, { "target": "failure-virage-analysis-and-survivor-selection", "relation": "learns-from" }, { "target": "unknown-vcn01-all-enrolled-survival", "relation": "leaves-open" }, { "target": "trial-vcn01-action-map", "relation": "implements" }, { "target": "hypothesis-vcn01-replication-to-survival-gate", "relation": "tests" }, { "target": "company-theriva-biologics", "relation": "contains" }, { "target": "failure-broad-epigenetic-treatment-pdac", "relation": "learns-from" }, { "target": "claim-mtap-prmt5-selected-human-signal", "relation": "bounded-by" }, { "target": "failure-vopimetostat-ras-contribution-unknown", "relation": "learns-from" }, { "target": "unknown-mtap-prmt5-ras-contribution", "relation": "leaves-open" }, { "target": "trial-mtap-prmt5-action-map", "relation": "implements" }, { "target": "hypothesis-mtap-prmt5-ras-deeper-control", "relation": "tests" }, { "target": "company-tango-therapeutics", "relation": "contains" }, { "target": "claim-elraglusib-randomized-survival-signal", "relation": "bounded-by" }, { "target": "failure-elraglusib-analysis-and-survival-discordance", "relation": "learns-from" }, { "target": "unknown-elraglusib-survival-replication", "relation": "leaves-open" }, { "target": "trial-elraglusib-action-map", "relation": "implements" }, { "target": "hypothesis-elraglusib-survival-confirmation-gate", "relation": "tests" }, { "target": "company-actuate-therapeutics", "relation": "contains" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "strong", "randomised": "moderate", "independentReplication": "moderate", "sampleSize": "23 ranked programme ideas linked to their underlying evidence", "effect": "Decision ranking; no pooled clinical effect", "limits": [ "Ranking contains explicit judgment", "Ideas are at different maturity", "Cost and speed affect experiment priority separately from ultimate impact", "Several transfers remain unproved", "Major 2026–2028 trial readouts can reorder the list" ] }, "layout": { "x": 74, "y": 120 }, "updated": "2026-09-15", "contentSections": [ "The hidden-spread staging route is an immediate surgical-safety dependency, not a new biological frontier. Its strongest result is an avoided-incision signal with NNT 8 after treatment; its main uncertainty is timing and cytology-only action. It uses £0.4m already inside the surgical-route budget and must not displace systemic-control or residual-disease work.", "Residual-cell state capture ranks above an anti-dormancy treatment. Human marrow studies establish prognosis, and recent organ-linked and spatial studies show route diversity, but the disease-free cell remains unsampled. The next value is in joining ancestry, state and time inside the existing observatory. Anti-NET, CCDC25 and EREG interventions stay below human pharmacology until that gate passes.", "The 72-hour older-and-vulnerable route sits beside, rather than inside, the biological frontier ranking. It uses the already assigned patient-safety line because randomized evidence supports lower toxicity now. Its scientific uncertainty is whether that transfer holds in PDAC and adds days alive outside hospital without slowing treatment. It must not displace common-RAS, residual-disease or resistance work.", "Locally advanced responder surgery sharpens frontier item 14 rather than adding a new rank. It is the cleanest test of selected local consolidation, but not the highest current capital priority: expert feasibility is established, treatment-effect causality is not, and 44% major morbidity raises the benefit threshold. Keep the 540-person test ready; release £22m only after equal capital leaves a lower-value branch.", "Organoid-guided choice remains frontier item 6 but its next action has narrowed. Do not build another correlation cohort. Recover the two stale randomized Chinese trials and completed Swiss feasibility data, test treatment interaction in PASS-01, and run a 200-person shadow route inside existing organoid funds. The route advances only if at least 80% receive tumour-confirmed results within 14 days and two laboratories agree on regimen order in at least 80%.", "FAP-targeted radiation remains frontier item 13, but its next experiment has changed. Human delivery and a phase 2 dose are established. The missing information is the map from scan uptake to time-integrated lesion Gy to PDAC control, followed by comparative useful life. Recover LuMIERE and FRONTIER data, and let NCT04939610 answer its already-running response question. The route moves upward only after dose-response reproduces and a controlled treatment design is justified; it receives no new treatment capital now.", "Autophagy enters at rank 21, not as a revival of HCQ. The human signal is negative for tested HCQ combinations and incomplete for direct ULK1/2 inhibition. Recover NCT04892017 and require paired tumour target plus blocked-cargo measurements before considering an agent. It can move upward only if tumour blockade repeats, RAS-drug dose is preserved and a randomized contribution test becomes feasible; current capital is £0.", "Sonoporation enters at rank 22 as data release, not device development. The first randomized trial's survival signal did not reproduce overall in the available NCT04821284 abstract. The highlighted gemcitabine subgroup moves only if the complete dataset shows it was planned, balanced and supported by the treatment-by-backbone interaction. A later experiment must measure actual tumour delivery and receives £0 now.", "KRAS siRNA enters at rank 23 as a proof-order correction. SIL204's 403-person plan already exists, so the mission adds no trial. Require paired multi-region mutant-RNA and protein reduction, systemic exposure, preserved chemotherapy and complete public data before dose-selection expansion. The branch gets £0 and can rise only after human target delivery reproduces.", "VCN-01 enters at rank 12 because it combines direct human tumour replication with a randomized phase 2 direction, evidence stronger than the delivery-only or uncontrolled routes below it. It remains a data-recovery rank, not a phase 3 cheque: the registry says 112 enrolled, the sponsor reports 96 in the FAS, the public alpha rules conflict and the later-dose subgroup selects survivors. Recover and independently reproduce the complete result, then let a sponsor-funded six-person dosing study answer safety and pharmacology only.", "MTAP-selected PRMT5 plus RAS treatment enters frontier rank 5. It clears reach, mechanism and early-response gates: about 28% of PDAC has the deletion, human biopsies show PRMT5 inhibition, and 25 of 39 currently evaluable combination patients responded. It does not clear contribution or survival. The existing sponsor programme should publish all 54 treated PDAC records and compare each RAS drug against itself plus vopimetostat. This adds £0 to the mission.", "Elraglusib enters at rank 4 because a peer-reviewed randomized phase 2 gives an all-randomized OS result of 8.9 versus 7.2 months, HR 0.68. It stays below proven treatment because the larger headline excludes 36 people before treatment, PFS and response do not improve significantly, severe events rise by 10.2 points and no independent randomized trial repeats the result. Recover the full record for £0 first. Only a robust audit and a solvent majority-funding partner can release blinded phase 3 confirmation." ] }, { "id": "research-stopping-audit", "title": "Deliberate stopping-rule searches changed the atlas rather than merely confirming it", "shortTitle": "Stopping audit", "type": "system", "status": "supported", "scope": [ "pdac", "pnet", "cholangiocarcinoma", "gallbladder", "hcc", "colorectal", "pan-cancer", "transfer-unproven" ], "summary": "79 directed passes cover the highest-value treatment, detection, delivery, failure, company, capital, interpretation and source-access branches; the last seven changed treatment boundaries or access but not the five-gap ranking or programme allocation.", "content": "Contradiction to the main conclusion: PANOVA-3 showed a 2.0-month randomized survival gain from tumour-treating fields, so the atlas no longer states that local or physical additions never extend survival. Omitted researchers and laboratories: Liwei Wang and the Renji network were added as an international RAS trial capability; a worldwide lab pass added centres in China, Spain, Scotland, India and Singapore while retaining African and Latin American gaps. Non-US/non-UK work: Chinese HRS-4642 and new chemotherapy–dormancy work, Japanese ENZEAVOUR, CT-AI and HCC local-treatment trials, Spanish AVATAR and CNIO combination work, Israeli Alpha DaRT, a 1,500-patient North Indian gallbladder cohort and the Asia-Pacific PLANet HCC network now sit in the graph. Failed approaches: AVATAR showed that only 4 of 81 patients assigned to a complex precision system received matched treatment. A separate integrity search found that a striking CNIO RAS–EGFR–STAT3 mouse study was retracted for undisclosed interests and republished with declarations; the atlas preserves both records and does not treat retraction alone as scientific disproof. Last 12 months: daraxonrasib phase 3 and approval, its paired human resistance series, setidegrasib, HRS-4642, PANOVA-3, KRAS interception, NHS-Galleri, chemotherapy-linked dormant-cell reawakening, new HCC recurrence work, two 2026 postoperative ctDNA cohorts and three recent immune-combination reports were checked. The RAS-resistance pass crossed a gate: 26/44 progressing patients acquired RAS-pathway changes and 16/44 = 36% acquired mutant KRAS amplification, above the mission's 25% frequency threshold. The £25m programme now tests independent phase 3 reproduction and multi-lesion agreement; the £100m programme cannot release expansion capital until that succeeds and a combination works at clinical exposure. The ctDNA pass changed the programme: DYNAMIC-Pancreas, Denmark's CIRCPAC and China's MAP-03 already test three different actions, while the new cohorts show materially variable sensitivity and predictive values. The £100m design now releases only £3m for a gap and assay audit before any new residual-disease trial; the remaining trial capital is held unless an uncovered action, patient state or endpoint is demonstrated. The immune pass also changed a gate: CXCR4 plus PD-1 treatment and vaccine plus dual-checkpoint treatment moved T cells into tumours but produced zero responses among 21 and two responses among 57, respectively, while suppressive myeloid states increased. The low-burden RAS-plus-immune experiment now stops if it changes infiltration without tumour killing or a better T-cell-to-myeloid state. POLAR keeps a selected HRD route open but missed its prespecified response and six-month progression thresholds. Adjacent disciplines: mechanobiology and nerve repair produced a narrower stroma–Schwann hypothesis. Recently funded startups: PAQ's $77m Series B, Cosomil's undisclosed Series A1, Mainz's $6m runway financing and Jacobio's $100m upfront licence were separated from contingent value. Newly initiated trials: PT0511, HRS-4642 phase 3, ENZEAVOUR and JAB-23E73 were added with no-results warnings. New preprints: a 3D multi-omic chemotherapy-resistance preprint was added as an unreplicated discovery result. These searches lowered confidence in a blanket local-treatment dismissal, strengthened the systemic-resistance priority, exposed delivery, host and base-rate constraints, prevented likely trial duplication and increased Asian and European representation. The germline-delivery pass changed the programme rather than adding prevalence. It found a live policy split: NHS England's May 2026 R367 rule remains age- and history-limited, while ASCO treats exocrine pancreatic cancer as an all-patient indication. It also separated high completion among relatives already enrolled in GENERATE from population reach: GIFT found only 19.2% of reported relatives invited, and a pancreatic cohort found only 31% of informed first-degree relatives tested. The £2.5m route now measures action inside and outside R367 and randomizes consented family contact; it stops for privacy, inequality or uncertain-variant harm. Later passes established diminishing decision value: they kept adding boundary conditions but did not change the five-gap ranking or programme allocation. The sonoporation pass corrected an outdated replication claim. A second randomized trial now exists: its accepted 2026 abstract reports no overall PFS or OS difference, both P>0.36, after a first 78-person trial reported longer survival. The attractive gemcitabine subgroup lacks counts, uncertainty, balance and an interaction test. The portfolio adds no money or trial; it requests the complete result and direct drug-delivery evidence. The KRAS-siRNA pass found an authorised 403-person successor trial after a failed first-generation overall result. It adds no cohort or money. The useful action is to make multi-region human knockdown and public data a gate inside the planned 15-to-21-person safety segment before expansion.", "contentSections": [ "Pass 49 — electrochemotherapy. The 25-person selected cohort, 90-person randomized protocol and transitioned registry, PanECT paper and posted results, IREC protocol and IGEA device record were joined. Delivery is feasible, but 11/25 lacked six-month imaging through death, chosen electrode groups do not prove superiority and PanECT has no control. The randomized trial has no public result after registry transition. Search stops at £0 equity, equipment or recruitment and at most £250,000 for independent recovery of at least 80 assignments with pulse logs, death-aware response, survival and harm.", "Pass 48 — high-intensity focused ultrasound. Prospective symptom evidence, the internally inconsistent 120-person comparison, the HIFU-PC15 protocol and registry, three current thermal-device trials and the distinct SonoPANIII non-thermal route were joined. Pain relief is plausible; survival benefit is not established. HIFU-PC15 has remained recruiting since 2017 without a result, NCT05262452 is stale, and PULS stopped after an industrial-partner decision. Search stops at £0 for devices or new cohorts and at most £250,000 after database lock to recover the 40- and 90-person randomizations with patient outcomes and device logs. Non-thermal drug-delivery ultrasound remains a separate claim.", "Pass 47 — radiofrequency ablation and PELICAN. The 188-person analysed randomized trial and its stale registry record were joined. Median survival was 12.1 versus 11.6 months, HR 1.07, and progression-free survival also did not improve. Grade 3 or worse serious events rose from 11% to 27%, an absolute 16-point increase and about one additional severe serious event per 6.25 treated patients. Only 70 of 95 assigned to ablation received it; surgery exposed hidden metastases in 21. Search stops at £0 for equipment or a new treatment cohort and at most £150,000 for two-laboratory analysis of existing paired specimens, which cannot reopen the failed efficacy route.", "Pass 46 — MRI-guided adaptive ablative radiation. The 136-person phase 2, current registries, CROSSFIRE and ViewRay's SEC record were joined. SMART can deliver 50 Gy in five fractions after induction selection, but the single arm cannot assign survival or surgery benefit. Late severe toxicity changes from 0% definitely related to 11.5% possibly related. LAP-ABLATE remains not yet recruiting on a 2022 record after its machine sponsor filed Chapter 11; MASPAC stopped at nine for low accrual. Search stops at a vendor-independent recovery gate: £0 for equity, equipment or recruitment and at most £500,000 to reconcile the phase 2 and qualify at least two maintained systems and ten centres before repricing the randomized test.", "Pass 45 — proton and carbon-ion radiation. Japanese, Taiwanese, German and U.S. papers and current registries were joined. Particle physics and high local control are credible, but none of the pancreatic evidence is randomized. The stark recent boundary is 94.7% one-year local control versus 44.5% freedom from distant spread. CIPHER enrolled zero; PACK completed 25 without posted results; PAN009-18 targets 60 with completion now estimated in 2039. Search stops at delivered-dose and systemic-risk recovery: £0 for another cohort or facility and at most £500,000 for common dose recalculation, complete outcomes and a locked local-dominant selector before any roughly 500-person randomized proposal.", "Pass 44 — regional hyperthermia. HEAT's paper, U.S. registry and European regulator record were joined with the HEATPAC protocol and registry. HEAT randomized only 117 of 336 planned people, missed disease-free survival and cannot isolate heat because cisplatin, gemcitabine timing and dose, capecitabine and treatment interactions also changed. HEATPAC has the cleaner heat-only comparison and direct temperature measurement, but its status has been unknown since 2017 and no result is public after estimated 2021 completion. Search stops at existing-data recovery: £0 for another cohort and at most £250,000 to reconcile assignments, survival, harm and temperature traces before any current-care trial.", "Pass 43 — tumour-treating fields and Optune Pax. PANOVA-3's paper, posted trial results, FDA review, PANOVA-4, two cost models and Novocure's latest filing were joined. The pass keeps the real narrow result: median survival rose 2.0 months and one-year survival 7.9 points with gemcitabine and nab-paclitaxel. It also exposes the cost and boundary: no significant PFS gain, median use 11.2 hours per day, skin toxicity in 76.3%, no independent reproduction and no evidence with another backbone. PANOVA-4 adds both fields and atezolizumab against an old control, so one contrast cannot estimate at least five causal terms. Search stops at £0 equity, product subsidy or company cohort and at most £500,000 for independent PANOVA-3 analysis and all-eligible route measurement after access and publication rights.", "Pass 42 — irreversible electroporation and NanoKnife. CROSSFIRE, two distinct DIRECT registrations, the registry safety paper and survival preprint, UK LAP-PIE, NICE guidance and AngioDynamics finance were joined. The device kills local tissue, but CROSSFIRE stopped for futility against radiotherapy and lacked a chemotherapy-only arm. The registry's 18-versus-10-month association follows an eight-versus-four-month diagnosis-to-enrollment filter, with only ten prospectively enrolled controls. Separate phase 3 NCT03899636 already tests the same chemotherapy with or without IRE; its 528 count and April 2025 completion remain estimated and no result is posted. Search stops at £0 new cohort or equity and at most £250,000 for complete independent randomized-data recovery after access and publication rights.", "Pass 41 — local artery drug delivery and TIGeR-PaC. The early 43-person pooled study, 16-person randomized blood study, 45-person first interim set, current registry, changed analysis plan, company finance and August 2026 enrollment close were joined. The pass separated three claims: the replacement package may help patients, the catheter may increase active drug in tumour cells, and the platform may transfer to other medicines or cancers. TIGeR-PaC can answer only the first. Route, removal of nab-paclitaxel and schedule change together; peripheral blood does not measure tumour drug; the first interim P=0.051 lacked a hazard ratio and interval. It adds no cohort or equity. Search stops at the existing 114-person final gate and at most £250,000 for independent analysis after lock and access.", "Pass 40 — ivospemin and ASPIRE trial integrity. The 50-person phase 1 signal, severe retinal and liver harms, two trial registries, the sponsor's last full quarter, note default and SEC order were joined. The pass found a potentially high-value randomized record trapped behind contractor termination, company distress and incompatible public trial states: U.S. planned n=600 and recruiting versus European planned n=547 and ended. It adds no cohort, licence or equity. Search stops at a £250,000 custody audit requiring at least 400 reconciled assignments, a usable blind and key and 90% 12-month vital status before any rescue is priced.", "Pass 22 — hidden metastasis before PDAC surgery. Mayo, PREOPANC, PREOPANC-2, TAPS, a 2026 Japanese cohort, the current global survey, NICE and a live Japanese study were compared. The pass found a strong avoided-incision signal but no randomized survival result, and exposed the field's habit of combining visible spread with wash cytology. It adds no budget: a 500-person action check is nested inside the existing English surgical route. Search stops at separate timing, pathology, delay and patient outcomes; reopen for randomized staging utility, a validated low-risk omission rule or a cytology-only treatment result.", "The gallbladder pass changed the related-cancer allocation. A 26-centre prospective cohort showed that the revised polyp rule produced one high-grade lesion and no cancer among 68 indicated operations. Two externally tested imaging and blood models diagnose selected visible lesions but do not validate future risk, and transporting one model to 2% prevalence reduces modelled PPV to 7.9%. AMBROSE supplies the missing harm bound: 0.4% 30-day mortality and 0.2% severe bile-duct injury across 21,706 mixed-setting operations. The planned 100,000-person incidence cohort still says not yet recruiting and has no results. The £500m design now releases £5m to join active EULAT, Chile BiLS, P-iGBC and North Indian fields and holds further funding behind country calibration, outcome linkage and net-benefit gates. No complete risk-to-action-to-mortality path was found, so this branch has not reached diminishing information value.", "The pNET pass separated three decisions that the earlier disease page had compressed. ASPEN and PANDORA support short-term observation of selected small non-functioning tumours, but ASPEN's choice was driven by centre and patient preference and its registry now has unknown status after a passed 2024 completion estimate. A current 460-operation cohort found 14.8% major morbidity and 0.9% 90-day mortality. MEN1 imaging finds lesions in adolescence, while a prospective 110-person mRNA assay failed to predict development, growth or metastasis. The £500m design now releases £3m for complete follow-up and rule definition; conditional pNET and gallbladder expansions compete for one capped reserve rather than silently exceeding £500m. No randomized small-pNET action path was found.", "The cholangiocarcinoma pass separated primary prevention, risk surveillance and postoperative recurrence. Liver-fluke infection is a proven cause; a 3,600-person One Health programme reported infection falling from 6.0% to 0.3%, but no cancer effect, and the randomized 480-person eCHEC trial ends at 12-month behaviour and infection. In PSC, annual MRI/MRCP produced severe or progressive duct findings in 23.8% of 512 people and roughly nine non-malignant findings per malignancy without durable curable detection. Postoperative ctDNA detected 15 of 16 recurrences in one cohort and repeated strong risk separation in STAMP, but no study randomized an action. The £500m design now releases £6m to rank the three routes, leaves £6m for HCC and permits only one CCA route to seek reserve expansion. No complete cholangiocarcinoma prevention or residual-disease action path was found.", "The HCC pass rejected three shortcuts and changed the final related-cancer allocation. Three-month ultrasound found more tiny lesions but no more small HCC. Three additional serum markers did not improve early-stage detection. Noncontrast MRI improved stage and reduced false referrals but did not show fewer deaths. TRACER and PREMIUM already test the missing late-stage and mortality questions, so a duplicate surveillance trial is excluded. ATTENTION keeps earlier HBV treatment open but its interim result did not cross the stopping boundary. The remaining £6m now funds a 240-person randomized liver-reserve readiness stage; a 600-person outcome trial cannot start unless delivery, safety, ALBI and tumour-treatment gates pass. HCC surveillance and co-control have not reached complete answers, but the next non-duplicate actions are now explicit.", "The colorectal pass rejected the shortcut that a validated ctDNA risk marker has general clinical utility. DYNAMIC stage II supports one omission rule. DYNAMIC-III did not validate stage III omission or escalation. ALTAIR did not validate trifluridine/tipiracil at molecular relapse and added 69.7 percentage points of severe hematologic toxicity. FIND increased curative-intent treatment after earlier imaging, but survival is immature. CIRCULATE-US already covers the live chemotherapy questions. The remaining useful searches are treatment sensitivity in positive residual disease, mature FIND survival and a locked organ-confined selector; another broad prognostic cohort has low information value.", "The regional delivery pass replaced a name-only world map with measured limits. An African pancreatic review contains 26,850 patients but nearly 40% lack stage and fewer than half of studies report complete treatment paths. A sub-Saharan HCC review found curative treatment in 6% and supportive care in 84% of 3,989 patients. Cape Town public/private and South Africa/Sweden comparisons show large stage, treatment and survival gaps without isolating one cause. LATAM-CCA adds 309 cases from five countries, while its referral selection and missing population denominator remain explicit. This changed the £5bn programme: only £5m releases for route capture, and £15m remains held until a repairable loss of at least 15 points is reproduced. More generic regional cohorts now have low value; a common dated route or randomized repair would have high value.", "The cachexia pass rejected weight or muscle as a stand-alone success condition. Tocilizumab reduced muscle loss but missed six-month and median survival and raised severe treatment toxicity by 24.7 points. Ponsegromab produced 1.22 to 2.81 kg placebo-adjusted weight gains in a randomized mixed-cancer trial, but survival and chemotherapy delivery were not its main test. A cheap olanzapine trial improved appetite and weight, while only 13 participants had HPB cancer and survival was absent. NCT06989437 already randomizes an estimated 982 PDAC patients across 220 locations. The £500m design now funds only data and endpoint access inside its existing statistics line; another broad trial is excluded unless the live study fails or leaves a named patient state uncovered.", "The pancreatic-enzyme pass split an inexpensive care treatment from an unproved survival claim. Randomized trials of 67 and 88 patients missed short weight or BMI endpoints and did not establish survival. Cohorts of 501 and 1,247 reported large survival associations, but the chemotherapy-only subgroup in the larger analysis was not significant and selection can explain the gap. PERTseverance, NCT05642962 and PANCAX-3 already cover delayed treatment, dose and feasibility, although none uses survival as its main endpoint and the first remains recruiting after its estimated June 2026 primary date. The £500m design now funds data joining inside the existing statistics line, rejects a duplicate broad trial and requires symptoms or absorption plus treatment delivery or quality-adjusted survival before capital release.", "The PDAC clot-prevention pass found a strong class effect and rejected a survival shortcut. Five randomized studies and 1,003 patients estimate VTE RR .31, an eight-point absolute reduction and NNT 11.9; CONKO-004 found PFS HR 1.06 and OS HR 1.01. A 243-person cohort found 24% VTE, 30% prophylaxis uptake, median start at day 43 and half of events within two months. The 450-person imPaCT-PRO study has unknown status after a passed completion estimate. Kinisoquin has a 57-person uncontrolled marker study and an active 480-person phase 3, not clinical proof. This changed the £500m design: £3m moves inside the existing £15m patient-safety line to a seven-day route-completion trial, leaving £12m. Another conventional efficacy trial and any product investment are excluded.", "The biliary-obstruction pass reconciled an apparent trial contradiction. Routine delayed plastic drainage raised serious complications from 39% to 74% in 2010; metal-stent drainage produced 29.0% versus 26.5% in 2026 and met a wide 15-point noninferiority margin. The later result supports selected drainage during real delay, not universal use. During neoadjuvant treatment, metal devices reduce obstruction, infection, reintervention and delay, while a 162-person surgical cohort links cholangitis and interruption to worse outcomes without proving causation. EUS drainage is a credible expert-centre route, especially after failed ERCP, but mixed-cancer trials do not establish a PDAC survival effect. This changes the £500m design: £4m inside the remaining safety line funds whole-route completion, leaving £8m; broad device trials and manufacturer-controlled analysis are excluded.", "Pass 17 — PDAC surgical route. Current English audit, commissioning, centralization cohorts, postoperative-treatment cohorts and randomized operation trials were connected. The search changed the programme: the next useful experiment is not another broad open-versus-robotic comparison. It is a first-scan-to-last-treatment route study with procedure-specific safety gates, full-course delivery, current national comparators and an adaptive cluster calculation. Stop searching this branch when the linked audit obtains the missing full-course and cause-of-loss fields, unless a new randomized trial changes 90-day safety or long-term cancer control.", "Pass 18 — post-resection surveillance. Two favourable routine-imaging associations, current Asian guidance, the RADAR-PANC protocol and live registry, and CIRCPAC were compared. This changed the portfolio by adding RADAR-PANC to the existing £3m residual-disease audit and prohibiting another broad imaging trial. Search stops at the live randomized gate: obtain actual accrual, crossover, overall survival, quality of life, recurrence treatment, false-positive work-up, radiation and cost. Reopen when a result posts, the registry misses its October 2026 completion, or a clearly different action appears.", "Pass 19 — perioperative treatment timing. PREOPANC, NORPACT-1, CISPD-1, ESPAC5, PREOPANC-2, A021501, the latest nine-trial analysis and live A021806 registry were compared. The search changed the claim: treatment first improves event-free control and can avoid noncurative exploration, but overall survival remains inconclusive and delivery can fail. It adds no trial or capital. Search stops at A021806's already accrued 358-person randomized gate. Reopen when its survival result posts, its December 2028 primary date moves materially, CISPD-1 is independently reproduced, or a distinct patient state or regimen remains unanswered.", "Pass 20 — human residual cells and dormancy. Two human bone-marrow studies, a 744-person recurrence cohort, two autopsy routes and two 2026 model mechanisms were compared. Human tumour-cell signals reproduce prognosis but not viability, low cycling, recurrence ancestry or treatment utility. Primary and end-stage tissue miss the disease-free interval. This changed the £25m observatory by locking a 250-person ancestry-state-time subset and explicit go gates without adding money. Stop broad searching at the sample boundary; reopen for a human disease-free tissue series, a lineage-linked recurrence set, independent reproduction of a candidate state or a safe human intervention.", "Pass 21 — older and vulnerable PDAC. GIANT, GrantPax, newly posted PAMELA-70, three randomized mixed-cancer care trials and the current Taiwan registry were compared. The search rejected age as a treatment selector and another reduced-regimen comparison. It changed the final £3m safety-line remainder into a gated 72-hour assessment-to-action experiment with a 150-person run-in and 900-person treated comparison. Search stops at complete GIANT data, separate PAMELA outcomes and this action gate; reopen for a randomized PDAC support result, a validated treatment-benefit rule or material harm.", "Pass 23 — locally advanced PDAC conversion surgery. NEOLAP treatment, CT and CA19-9 analyses, TAPS selection and restaging models, PREOPANC-4 nationwide surgery, a 2025 matched comparison, the old 42-person direct trial, randomized PELICAN, CONKO-007 and two live registries were compared. The search changed the route from permanent inoperability to mandatory expert reassessment, while rejecting selected survival as proof of surgery benefit. It adds no current capital: a 540-person, £22m surgery-versus-continued-treatment candidate requires equal portfolio displacement. Search stops at the three-function boundary of technical removal, prognosis and treatment effect; reopen for a modern randomized surgery result, NCT06714604 outcome, external validation outside expert centres or a treatment-effect selector.", "Pass 24 — organoid-guided PDAC treatment. HOPE, Beutel, ORGANOPREDICT, the paired biopsy trial, PASS-01, AVATAR, two stale randomized Chinese studies, the completed Swiss study, PROSPER-PANC, ADOPT and UNITEPANC were compared. The pass retained organoids as a drug-rejection tool but rejected conditional accuracy as clinical utility. It adds no budget: recover existing data and run a 200-person shadow route inside assigned funds. Search stops at the whole-route boundary; reopen for a posted randomized result, a replicated 14-day assay or a treatment-by-assay interaction.", "Pass 25 — FAP-targeted radiation in PDAC. The 2022 FAP-2286 compassionate-use report, 2026 LuMIERE phase 1, active NCT04939610 expansion, terminated FRONTIER programme, four-person CTR-FAPI study and mixed-isotope 3BP-3940 series were compared. Human delivery and a phase 2 dose are established, overturning the old statement that no therapeutic human evidence existed. PDAC benefit is not: LuMIERE treated nine pancreatic cancers, reported four stable cases and no pancreatic partial response, while tumour-lesion dose was not published. The pass cancels a proposed duplicate 20–30-person, £4m dose-escalation study. It adds no capital: recover existing scan-dose-response data and let the live programme answer its declared response question. Search stops at the scan-to-dose-to-benefit boundary; reopen for LuMIERE PDAC dose-response, phase 2 results, FRONTIER data, an independent PDAC response signal or a controlled useful-life comparison.", "Pass 26 — autophagy and metabolic targeting in PDAC. The 528-person AVENGER 500 result was traced back to its uncontrolled 11/18 phase 1 response, and three HCQ combination routes were compared with the terminated 91-person direct ULK1/2 programme. The pass rejects both early-response escalation and another HCQ trial. It does not claim that all autophagy biology failed: the randomized study could not show consistent blood-cell blockade, tumour blockade was not established, and final inlexisertib results are missing. No capital is added. Recover NCT04892017, require paired tumour target and blocked-cargo measurements, preserve at least 80% of the RAS-drug dose, and stop before expansion if any gate fails. Search stops at the human tumour-target boundary; reopen for final inlexisertib data, a repeated tumour assay or a controlled selective-inhibitor result.", "Pass 27 — ultrasound and microbubbles in PDAC. The ten-person historical-control study, 78-person Chinese randomized trial, NCT04821284 registry, 13-person perfusion analysis and accepted 2026 trial abstract were joined. The first randomized survival signal did not reproduce overall: second-trial PFS and OS were both P>0.36. Its gemcitabine/nab-paclitaxel subgroup reports 407.5 versus 234 days but omits the denominator, confidence interval, balance, prespecification and interaction. No capital or duplicate trial is added. Release the complete intended-to-treat dataset, reconcile 97 enrolled with the registry estimate of 120 and separate perfusion imaging from actual tumour drug delivery. Search stops at the overall-to-interaction-to-delivery boundary; reopen for the full report, posted results, a planned interaction with uncertainty, or direct human tumour-delivery evidence.", "Pass 28 — local and systemic KRAS siRNA in PDAC. The 59-person siG12D-LODER study, authorised SIL204 CTIS protocol, July site-activation filing and August corporate filing were joined. The old product missed overall randomized survival; its G12D/V estimate came from 11 treated and five controls. SIL204 changes chemistry, formulation and route and had no reported human dose at the cutoff. CTIS plans 403, while the site release describes about 166 randomized; the records require reconciliation. CTIS also says participant-level data will not be shared. No mission trial, licence, equity or new capital is added. Use the live 15-to-21-person safety segment for paired multi-region tumour knockdown, systemic exposure, chemotherapy and harm gates under a public-data contract. Search stops at the first-human target boundary; reopen for treated-participant pharmacology, a revised protocol, first safety results or independent human tumour data.", "Pass 29 — engineered and tumour-killing viruses in PDAC. Pelareorep's randomized failure, two VCN-01 phase 1 papers, the 112-person VIRAGE registry, public protocol, 96-person sponsor result, ESMO report, six-person VIRAGE2 registry and current Theriva filing were joined. VCN-01 clears the human-delivery boundary: virus appeared in five of six sampled tumours, and its randomized direction favours survival and progression. It does not clear confirmation: 112 enrolled becomes a 96-person FAS, the protocol conflicts on alpha and analysis sets, the strongest repeat-dose subgroup selects survivors, and no full report or independent trial exists. No capital or duplicate cohort is added. Recover the complete allocation and outcome record, use VIRAGE2 only for dosing safety and pharmacology, and require a blinded all-randomized survival trial. Search stops at the all-enrolled-analysis boundary; reopen for posted results, a peer-reviewed full report, the final analysis plan, mature quality-of-life data or a funded phase 3 protocol.", "Pass 30 — broad epigenetic treatment versus MTAP-selected PRMT5 inhibition. Two randomized broad-drug studies, three small uncontrolled state-changing studies, two PRMT5 molecules, two large prevalence series, current registries and Tango's finance and combination disclosures were joined. Broad drug treatment remains a failure: no randomized survival gain and only three responses in each modern checkpoint combination. MTAP deletion creates a different route in about 28% of PDAC, with direct tumour PRMT5 inhibition and a 25-of-39 evaluable combination response signal. The signal does not prove contribution because it joins two RAS drugs and several doses with no RAS-alone control. No money or duplicate cohort is added. Mature all 54 treated PDAC records and randomize the identical RAS inhibitor with or without vopimetostat. Search stops at the component-attribution boundary; reopen for mature all-treated data, a peer-reviewed combination report, a fixed randomized protocol or independent survival evidence.", "Pass 31 — GSK-3β inhibition with elraglusib. A peer-reviewed randomized phase 2, current registry, independent 49-person combination abstract and current SEC filings were joined. The signal is real enough to rank: all-randomized OS was 8.9 versus 7.2 months, HR 0.68. The 2.9-month headline uses a treated population after 9.4% of the drug arm and 20.4% of controls did not start treatment. PFS, response, disease control and response duration did not significantly improve; severe events rose by 10.2 points. Actuate had $4.426m cash against $9.422m six-month operating use and no later company financing filing by 15 September. Search stops at the all-randomized-data boundary. Spend £0 on phase 3 now; recover the full record, then require blinded survival confirmation and a solvent majority-funding partner.", "Pass 32 — cell-cycle and replication-stress checkpoints. Selected palbociclib, a 76-person combination, randomized abemaciclib, randomized CHK1, an 11-person CHK1 replacement, an uncontrolled adavosertib study and the current azenosertib registry were joined. Broad CDK4/6 and CHK1 treatment is a repeated human failure despite model activity, target exposure and a common pathway fault. WEE1 stays open only as data recovery: NCT06015659 planned 12 safety participants before expansion and now reports actual n=12 after primary completion, with no public result. No cohort or money is added. Search stops at the completed-safety-stage boundary; reopen when all 12 outcomes, paired tumour measurements and the expand-or-stop decision are public.", "Pass 33 — antibody-delivered treatment. Randomized GLEAM, IBI343 and IBI389 reports and registries, EBC-129, anetumab, SHR-3821, free exatecan, company finance and licensing were joined. The class does not transfer: zolbetuximab found CLDN18.2 but OS HR was 0.999; anetumab produced 0/14 responses and SHR-3821 0/11 confirmed responses at its main doses. IBI343 remains product-specific Tier C evidence because 10/44 high-marker tumours responded versus 0/12 lower-marker tumours. NCT07066098 already randomizes 201 people and Innovent plus Takeda can fund it. Search stops at the existing-randomized-answer boundary; reopen when the protocol, exact marker rule, full denominators and survival are public.", "Pass 34 — engineered cell treatment. Mesothelin CAR-T, satri-cel CLDN18.2 CAR-T, KRAS-G12D and KRAS-G12V TCR-T, active registries, adjuvant evidence and CARsgen finance were joined. Surface-reading satri-cel has 4/24 uncontrolled responses; mutation-HLA TCR-T has two clear responses across seven reported patients; neither has controlled survival. The negative control is stronger: local mesothelin delivery produced zero responses and NCT03323944 now counts 54 enrolled people while the paper reports nine infused. No new cohort or capital is added. Search stops at the route-accounting-and-response-replication boundary; reopen when full screen-to-outcome denominators, fixed human mechanism and durable response report from existing cohorts.", "Pass 35 — tumour-immune reprogramming. TGF-beta, CD40 and macrophage CSF1R routes were joined across phase 2 and phase 3 trials, tissue work, current registries and developer finance. NIS793 changed its intended tissue system but lost 2.0 median survival months and added 14.6 points of serious harm in a 511-person blinded trial. Cabiralizumab did not add controlled benefit. Mitazalimab retains a 57-person uncontrolled signal while its owner has stopped development. Search stops with two high-information actions: a capped £1m fixed angiogenin test in stored randomized samples, and a £25m matched ceiling for a controlled CD40 survival trial only after complete 94-person data, product transfer and a solvent co-funder.", "Pass 36 — tumour blood flow and hypoxia. Four VEGF-starvation phase 3 trials totaling 2,387 people, MAESTRO's 693-person hypoxia-activated drug test, the 49-person losartan route, randomized AFPAC and current localized registries were joined. Starvation repeatedly failed survival. Evofosfamide improved progression but missed survival. Losartan failed broad advanced disease after an uncontrolled localized signal, while NCT03563248 already holds 168 randomized localized records. Search stops at existing-data recovery: £0 for new cohorts, factorial release of the losartan trial and marker-linked MAESTRO reuse only if fixed samples exist.", "Pass 37 — microbiome intervention versus association. Human bacterial, fungal and survivor signals were separated from mouse treatment effects and from the 40-person randomized kestose pilot, which found no significant response difference. The clinical pipeline lost zero, 13 and an estimated 10 patients to withdrawal, insolvency and suspension, while two current single arms cannot isolate their microbial components. Randomized FMTPanc and PANORAMIX already own the next questions. Search stops at £0 new cohorts and at most £1m for a fixed tumour-enzyme assay plus independent analysis inside PANORAMIX after its 142-versus-206 registry denominator is reconciled.", "Pass 38 — PDAC vaccines. Shared peptides, whole-cell products, bacterial vectors, dendritic cells, KRAS peptides, personalized RNA and preventive vaccination were separated. Four later-stage programmes totaling 1,812 randomized participants missed survival or disease-free survival. Autogene cevumeran generates durable T cells, but its 8-versus-8 responder comparison is not efficacy. IMCODE003 already estimates 260 assigned people and is active but no longer recruiting. Search stops at its all-assigned outcome and complete tissue-to-dose route: £0 for a new cohort and at most £250,000 for independent analysis after data lock and access.", "Pass 39 — CD73 and tumour adenosine. ARC-8's 122-person chosen-dose survival signal was separated from its 90-person randomization, which tested checkpoint addition because every arm received quemliclustat. The matched external OS result did not align with response or PFS; baseline NR4A failed external validation; and oleclumab missed its direct randomized endpoint. PRISM-1 already completed enrollment for a masked 610-person survival test, with a first-half 2027 result expected. Search stops at £0 new trial or equity and at most £250,000 for independent analysis after data access.", "Pass 50 — endoscopic ultrasound radiofrequency ablation. The randomized HybridTherm study found no patient benefit, stopped short of target accrual and left product and study-custody questions. A separate live study can answer safety and route performance but cannot erase the negative comparison. Search stops at £0 for a new device cohort and at most £300,000 to recover existing assignments, treatment delivery, survival, pain and harm.", "Pass 51 — histotripsy. GANNON shows that focused sound can be aimed at some pancreatic tumours, but the ribs, stomach, bowel and gas block a reliable acoustic window and human tumour coverage is not yet established. The FDA-cleared Edison liver indication does not transfer to pancreas. Search stops at the first 30 independently reviewed pancreatic procedures and at most £300,000 for image, coverage, tissue and harm analysis before any efficacy claim.", "Pass 52 — cryoablation. Small selected cohorts report pain relief and local effects, but no randomized patient-benefit result survives selection, attrition and procedural-risk checks. A live cryo-immunotherapy study bundles two interventions. Search stops at £0 for equipment or a broad cohort and at most £1.5m for a pragmatic pain-and-useful-life trial only after a reproducible safe route exists.", "Pass 53 — microwave ablation. Human reports support technical heating and possible pain relief, not added survival or complete safe tumour coverage. Combination studies mix artery treatment, chemotherapy and microwave exposure. Search stops at £0 for a new cohort and at most £250,000 to recover complete imaging, pain, death and harm records from existing studies.", "Pass 54 — phosphorus-32 implantation. OncoSil has a local-control signal, but liquefaction, variable delivered dose, selected surgery and incomplete reporting prevent a survival claim. TRIPP-FFX already owns the randomized question. Search stops at £0 for product, equity or another cohort and at most £350,000 for independent all-assigned analysis after the trial locks.", "Pass 55 — photodynamic treatment. Verteporfin and porfimer can destroy a measured volume after light activation, but tiny studies cannot show useful-life benefit and treatment geometry excludes vessels and ducts. Search stops at £0 for a new cohort and at most £300,000 for spatial dose, untreated-volume, death-aware response and harm recovery from existing work.", "Pass 56 — laser ablation. Endoscopic laser heating is technically possible, but fistula risk, weak accrual, stopped studies and device continuity block a benefit claim. Search stops at £0 for equipment or recruitment and at most £250,000 to recover the existing treated denominator, delivered heat, imaging, pain and harms.", "Pass 57 — ASP-1929 photoimmunotherapy. EGFR is common in pancreatic tumours, but expression is not proof that antibody-light treatment will cover a tumour or improve life; older EGFR treatment did not improve survival. The pancreatic trial has not yet produced human data. Search stops after its first ten evaluable patients unless target coverage, delivered light, viable-tumour loss and acceptable harm all pass; no company or product capital is released.", "Pass 58 — K-912 sonodynamic treatment. The first human study reports a local signal, but ultrasound, K-912 and chemotherapy changed together, so the active component and systemic benefit are unknown. Two live studies already own the next measurements. Search stops at £0 equity or new cohort and at most £350,000 for component, tissue-exposure and all-lesion analysis.", "Pass 59 — pressurised abdominal chemotherapy. PIPAC reaches the abdominal cavity, but 101 of 156 registered procedures stopped because of progression or poor condition and repeat-treatment survivors are selected. Search stops at £0 for new procedures and at most £300,000 to recover complete assignments, systemic treatment, quality of life, survival and stop reasons.", "Pass 60 — heated abdominal chemotherapy. A 42-person randomized adjuvant study reduced local recurrence from 11 of 21 to 2 of 21 without improving median survival, 18.0 versus 17.1 months. Small metastatic reports remain selected and a live Mayo study already tests feasibility. Search stops at £0 for procedures and at most £350,000 for existing-trial measurement and complete peritoneal-disease accounting.", "Pass 61 — cancer-pain nerve ablation. Coeliac plexus, direct ganglion and splanchnic routes produce inconsistent pain effects, while two comparisons raise possible survival harm. Search stops at £0 for a new procedure cohort and at most £250,000 for death-aware, opioid-adjusted analysis of existing randomized records before any routine route change.", "Pass 62 — ferroptosis. Iron-driven membrane failure is a real tumour-cell vulnerability, but pancreas-wide ferroptosis can also accelerate KRAS tumours in mice. The first CNSI-Fe study treated 19 people, only two with pancreatic cancer, reported no direct human ferroptosis measure and had 14 grade 3 or worse events. Search stops at £0 product or equity and at most £400,000 for cell-type, tissue-exposure and component measurement inside the live study.", "Pass 63 — NRG1 fusion treatment. Zenocutuzumab produced 12 responses among 30 pancreatic patients in the FDA review, but 20 stopped for progression or clinical progression and durable control is unproved. The rare route can fail before treatment because RNA fusion testing is inconsistent. Search stops at £0 product or equity, at most £150,000 for full trial recovery, and a capped £1.5m test of consecutive RNA-route completion only if NRG1 coverage is explicit.", "Pass 64 — early palliative care. A 207-person randomized pancreatic trial improved a quality-of-life score by 3.83 points and reduced chemotherapy in the final 30 days by 9.1 points without extending survival; another 288-person pancreatobiliary trial missed both main endpoints. Search stops first at £150,000 to release and independently analyse the completed Danish trial, then at most £2m for a needs-triggered route whose main outcome is quality-weighted days alive and at home.", "Pass 65 — DPYD and UGT1A1 treatment safety. Regulators already require DPD-deficiency testing before systemic fluoropyrimidines, and prospective mixed-cancer cohorts show lower severe toxicity among guided carriers against historical controls. The boundary is the pancreatic route: UGT1A1 risk changes with the regimen, one 19-person pancreatic expansion missed its tolerability rule, standard DPYD panels leave ancestry and negative-test gaps, and national pancreatic completion is not published. Search stops at £0 for a drug, assay company or new efficacy trial and at most £800,000 for 2,000 consecutive patients across 20 centres, with result-time, dose, escalation, toxicity, tumour-control and death gates.", "Pass 66 — KRAS-wild-type rare drivers. Cohorts of 2,483, 795 and 100 pancreatic cancers reproduce enrichment for alternative BRAF and receptor-fusion drivers, and NTRK, MET, ROS1, RET, BRAF and NRG1 cases show real drug sensitivity. Know Your Tumor supplies the systems warning: 46/1,856 referrals, 2.5%, reached its matched-treatment group, so the HR 0.42 selected comparison is not a profiling effect. Search stops by broadening the existing 5,000-person NRG1 DNA-to-RNA route to every supported alternative driver, with £0 added capital and no pooled rare-driver efficacy claim.", "Pass 67 — exercise and rehabilitation. APACaP randomized 313 advanced-PDAC patients but retained 172 for its week-16 primary analysis and missed all three main quality endpoints; PancFit randomized 151 and changed strength sessions without a significant quality or clinical difference. Surgical trials changed steps or selected recovery measures without a repeated major-complication effect. Search stops at £0 for a new broad programme and at most £250,000 to recover at least 80% of five randomized records, test a baseline low-function selector in independent data and require useful days or chemotherapy delivery without delay or harm.", "Pass 68 — interpretation and medical-language audit. NCI's reviewed cancer dictionary, FDA's oncology-endpoint guidance and CONSORT 2025 were joined to a systems, statistics and capital-allocation reading guide. The interface now detects 65 common medical and trial terms on each selected page and explains both the clinical meaning and a software, physics or business comparison. The comparisons are marked as teaching aids, not biological evidence. Search stops because the same definitions now cover the repeated high-load vocabulary; add a term only when a reader encounters an unexplained term that changes a decision.", "Pass 69 — graph and source-access audit. At that checkpoint, all 673 research pages were reachable from the mission and all 839 source records were used; 785 URLs were directly accessible, 52 returned access-control responses and two failed at the network or server layer, with no persistent not-found target. Later passes expanded the final release to the counts reported in the coverage page. Access control is not evidence failure, so the atlas keeps those citations and records the check separately. Search stops at the link layer because no dead target remains; scientific verification stays claim-specific.", "Pass 70 — recent blood-test challenge. Three 2026 primary studies were checked rather than accepting their AUC values. The strongest prospective result found 6 of 8 PDAC cases in 1,926 people with diabetes or obesity, but also returned 40 false positives: 75% sensitivity, 97.9% specificity and 13.0% PPV, with no clinical-utility endpoint. At ordinary annual incidence, unchanged performance would imply about 2,100 false positives for 10.4 true positives per 100,000, or roughly 0.49% PPV; this is a workload illustration, not a transport claim. Two larger blinded studies still began with selected known cancers. The detection ranking therefore does not change. Search stops at new assay examples and continues only for intended-use prospective results, fixed work-up harms or randomized patient outcomes.", "Pass 71 — outside-field transfer. Primary work in model-guided adaptive therapy, programmable T-cell logic, microfluidic organoids, engineered matrices and paired nuclear imaging-treatment was reduced to five required properties. Each property has a direct PDAC failure gate: a dense state sensor and competing cell populations; a safe tumour-wide antigen pair plus entry; all-patient on-time culture and delivered treatment; human mechanics, drug concentration and state measured together; or pancreatic tumour dose followed by patient benefit. None supplies that full chain today, so no programme allocation changes. Search stops at attractive analogies and reopens only when the missing property is measured in human PDAC.", "Pass 72 — newly funded detection startup. Adenocyte's May 2026 $10m first close, company description, two U.S. registries and a device patent were joined. The investor reports four unsuspected lesions among 90 high-risk people, but there is no peer-reviewed complete result, both 500-person registries are observational and invitation-only, and neither posts results. The capital-frontier lower bound rises from $183m to $193m; the detection ranking and programme allocation do not change. Search stops at financing and product claims and reopens for the all-participant table, negative follow-up, central pathology, procedure harm or an independent action result.", "Pass 73 — 2026 randomized hepatobiliary outcomes. TALENTOP improved time to treatment failure after selecting HCC induction responders for resection, but added severe harm and two treatment-related deaths without a mature overall-survival claim. EMERALD-3 improved progression-free survival but not reported overall survival and raised serious events. PROOF 301 screened 1,127 people at 120 sites but randomized 48 of about 300 planned, 4.3% of those screened. These results strengthen response-gated local treatment, endpoint separation and rare-driver network design; they do not change the five-gap ranking or release capital.", "Pass 74 — pancreatic neuroendocrine radioligand combination. CONTROL NETS included only 27 pNET participants in a non-comparative randomized phase 2. The combination response and late progression signal are interesting, while PFS HR 0.41 has a 95% CI of 0.15-1.12 and two radioligand-exposed participants developed treatment-related myeloid cancers. It supports a phase 3 sequencing question, not superiority or cure, and does not change the pNET ranking.", "Pass 75 — 2026 PDAC treatment and supportive-care results. RTOG 0848 did not add survival after resection and chemotherapy and doubled grade 3 toxicity. A 36-person paricalcitol run-in changed the predicted human tissue signals but was not an efficacy test and caused frequent high blood calcium by the oral route. A 60-person elemental-diet trial reported a large survival signal that needs independent replication; a blinded 290-person mistletoe trial found no quality-of-life or weight benefit. These records add one cheap replication candidate and two stop rules, not a new core allocation.", "Pass 76 — delivery, AI and diet stress test. IGNITE-TX supports proactive bilingual navigation plus no-cost family testing in a pan-cancer pilot, but does not establish a PDAC prevention effect. An AI overlay improved novice endoscopic-ultrasound readers on selected image sets, but did not test live procedures or patient outcomes. A 32-person ketogenic-diet screening trial had wide intervals crossing no effect. The final 2026 randomized search added useful boundary conditions without changing the five bottlenecks, funded experiments or company ranking.", "Pass 77 — regulator and access check. FDA's May 2026 approval creates a US treatment route for previously treated NRG1-fusion cholangiocarcinoma, but its evidence set was 19 people, seven responses and no randomized survival comparison. EMA orphan designation and phased review may accelerate daraxonrasib assessment but do not authorize European marketing. These records change access labels, not the treatment ranking.", "Pass 78 — final current-results sweep. TORCH was missed by the earlier HCC searches despite being a July 2026 phase 3 publication. It randomized 241 people with liver-confined BCLC stage B HCC at two Chinese centres. Selective radiofrequency ablation after arterial chemoembolisation increased median overall survival from 35.1 to 88.6 months, HR 0.50, and progression-free survival from 7.3 to 17.7 months, HR 0.47. This materially strengthens a state-specific HCC local-treatment route and corrects the earlier three-result summary. It does not change the PDAC-led five-gap ranking, transfer to other cancers or release new trial capital: the action is to use TORCH as the treatment benchmark, specify tumour geometry and liver reserve, and seek independent replication.", "Pass 79 — final AI validation sweep. PANCprAId trained on 231 retrospective resected-PDAC cases and then found treatment-interaction signals in 313 archived slides from randomized PRODIGE-24 participants. This is stronger than an ordinary held-out classifier but still does not test model-guided care, because treatment was not assigned from the model. The AI programme now treats it as the direct reason to run a locked prospective selection test, not as a clinical product. Further search now reopens only for a registered trial result, regulator action, reproduced patient-outcome effect, failed core assumption or evidence that crosses an existing stop or release threshold." ], "sourceIds": [ "beaufils-2026-pancpraid", "lyu-2026-torch-hcc", "fda-2026-bizengri-cholangiocarcinoma", "ema-2026-daraxonrasib-phased-review", "ema-2026-daraxonrasib-orphan-designation", "aguirre-2026-ignite-tx", "tanaka-2026-ai-eus-reader", "jameson-2026-ketogenic-diet-pdac", "abrams-2026-rtog0848", "perez-2026-paricalcitol-pdac", "uemura-2026-elemental-diet-pdac", "wode-2026-mistral", "chan-2026-control-nets", "sun-2026-talentop-hcc", "abou-alfa-2026-proof301", "kudo-2026-emerald3", "adenocyte-company-linfu", "igan-2026-adenocyte-financing", "clinicaltrials-nct04793633-linfu", "clinicaltrials-nct06276764-linfu-ipmn", "patent-us20240131366-linfu", "zhang-2017-evolutionary-adaptive-therapy", "roybal-2016-combinatorial-t-cell-circuit", "choi-2024-microfluidic-pdac-organoids", "lesavage-2024-engineered-matrix-chemoresistance", "wang-2026-cfdna-prospective-detection", "oberg-2026-pcdc-biomarker-bakeoff", "wang-2026-itih3-ceacam1-panel", "gudmundsdottir-2023-staging-laparoscopy", "van-dongen-2023-preopanc-staging", "rompen-2026-preopanc2-staging", "theijse-2024-nontherapeutic-laparotomy", "stoop-2026-global-staging-survey", "umin-000061741-staging-laparoscopy", "nice-2018-ng85-pancreatic", "macarulla-2025-panova3", "babic-2019-muscle-loss-survival", "chen-2025-tocilizumab-cachexia", "groarke-2024-ponsegromab-cachexia", "clinicaltrials-nct06989437-ponsegromab-pdac", "sandhya-2023-olanzapine-anorexia", "sarno-2025-avatar", "cui-2026-hrs4642-pdac", "jrct-2026-enzeavour-feasibility", "nct06698458-alpha-dart-pdac", "paq-2026-seriesb-pt0511", "jacobio-2026-annual-results", "forjaz-2026-3d-multiomic-preprint", "stupakov-2026-schwann-mechanics", "shi-2026-chemo-dormant-dtc", "liaki-2026-triple-combination-republished", "pnas-2026-liaki-retraction", "vineet-2026-north-india-gallbladder-cohort", "kawaguchi-2025-surf-hcc", "zhang-2026-planet-hcc-recurrence", "aronchik-2026-daraxonrasib-resistance", "dorbin-2026-allele-specific-daraxonrasib-resistance", "lin-2026-vertical-kras-inhibition", "lee-2024-dynamic-pancreas-feasibility", "clinicaltrials-nct05788744-circpac", "clinicaltrials-nct05802394-map03", "guo-2026-postsurgical-ctdna", "chen-2026-personalized-ctdna", "ho-2025-cxcr4-pd1-pdac", "bever-2026-vaccine-checkpoint-pdac", "park-2026-polar-pdac", "van-dijk-2025-polyp-study", "yang-2025-gbcseeker", "gupta-2026-gaia-mil", "wong-2025-ambrose-cholecystectomy", "clinicaltrials-nct03762837-gbc-risk-cohort", "clinicaltrials-nct06192719-eulat-gbc", "koshiol-2026-chile-bils", "clinicaltrials-nct06531408-pigbc", "partelli-2022-aspen-interim", "clinicaltrials-nct03084770-aspen", "heidsma-2021-pandora-pnet", "maekawa-2026-small-pnet-surgery-morbidity", "klein-haneveld-2021-men1-young-screening", "van-vliembergen-2026-men1-mnet-assay", "iarc-2012-liver-flukes", "charoensuk-2024-one-health-opisthorchis", "clinicaltrials-nct05321992-echec", "villard-2023-psc-surveillance", "tan-2024-psc-mrcp-surveillance", "yu-2025-btc-ctdna-recurrence", "yoo-2025-extrahepatic-cca-ctdna", "clinicaltrials-nct05743959-btc-mrd", "lim-2025-attention-hbv", "trinchet-2011-hcc-surveillance-interval", "hirode-2026-hcc-biomarker-rct", "rhee-2025-miracle-hcc", "singal-2024-tracer-protocol", "clinicaltrials-nct06084234-tracer", "ioannou-2026-premium-protocol", "clinicaltrials-nct05486572-premium", "bretthauer-2022-nordicc-10y", "kaminski-2026-nordicc-13y", "tie-2025-dynamic3", "bando-2026-altair", "mo-2026-find-crc-surveillance", "clinicaltrials-nct05174169-circulate-us", "gismalla-2026-africa-pancreatic-review", "sobnach-2024-ssa-hcc-systematic-review", "sobnach-2025-south-africa-hcc-disparities", "sobnach-2026-global-hcc-disparities", "da-fonseca-2024-latam-cca", "woo-2016-pert-rct", "saito-2018-pert-rct", "picozzi-2025-pert-advanced-pdac", "dugic-2026-pert-mortality", "clinicaltrials-nct04098237-pancax3", "clinicaltrials-nct05642962-pertzye-dose", "clinicaltrials-nct06099119-pertseverance", "lemanska-2023-opensafely-pert-audit", "dhsc-2024-pert-shortage-alert", "maraveyas-2012-fragem", "pelzer-2015-conko004", "vadhan-raj-2020-cassini-pdac", "frere-2020-pdac-thromboprophylaxis-meta", "heffley-2024-pdac-vte-uptake", "noor-2026-pdac-vte-implementation", "ash-2023-primary-vte-prophylaxis-guide", "overvad-2022-khorana-validation", "clinicaltrials-nct05178628-impact-pro", "clinicaltrials-nct06861088-catiq-p3", "clinicaltrials-nct05245877-preposterous-pancreas", "clinicaltrials-nct07196020-pinpoint-vte", "zwicker-2019-isoquercetin-catiq", "quercis-2026-kinisoquin-profile", "patent-us10391096-kinisoquin", "patent-wo2020112828-isoquercetin-cancer", "van-der-gaag-2010-pbd-rct", "costamagna-2026-sems-early-surgery-rct", "tamura-2021-fcsems-plastic-rct", "seo-2019-covered-uncovered-sems-rct", "lyu-2023-metal-plastic-nat-meta", "vehvilainen-2024-cholangitis-nat-interruption", "chen-2023-element-biliary-rct", "paik-2018-eus-ercp-biliary-rct", "barbosa-2024-eus-ercp-meta", "esge-2018-biliary-stenting-guideline", "asge-2024-therapeutic-eus-guideline", "nhs-england-2026-r367", "asco-2024-germline-panel-guideline", "llach-2024-universal-germline", "chittenden-2021-systematic-germline", "walker-2021-embedded-germline-station", "wang-2022-oncology-germline-cascade", "ramsey-2023-mainstreaming-germline", "rodriguez-2024-generate", "mcdonald-2024-germline-protocol", "sinan-2025-pdac-germline-implementation", "grewal-2026-pancan-germline-survey", "katz-2026-gift-cascade", "chen-2023-fdr-cancer-risk", "npaca-2026-state-of-nation", "nhs-england-2024-hpb-pancreatic-service", "de-wilde-2012-centralization", "lemmens-2011-centralization", "latenstein-2021-centralization", "coupland-2016-england-resection-volume", "labori-2016-adjuvant-completion", "henry-2023-complications-adjuvant", "kollbeck-2025-complications-chemo", "van-hilst-2019-leopard2", "korrel-2023-diploma", "bruna-2025-diploma-long-term", "klotz-2024-europa", "liu-2024-robotic-open-pd", "jin-2026-portal", "andel-2025-routine-imaging", "van-goor-2025-nationwide-imaging", "daamen-2024-radar-panc-protocol", "clinicaltrials-nct04875325-radar-panc", "tai-2025-pan-asian-esmo-pancreatic", "labori-2024-norpact1", "bai-2025-cispd1", "tan-2026-neoadjuvant-ipd-meta", "ghaneh-2023-espac5", "clinicaltrials-nct04340141-a021806", "effenberger-2012-pdac-bone-marrow-dtc", "nordgard-2022-unresectable-dtc", "chalabi-hajkarim-2025-organotropism", "pei-2025-spatial-metastatic-pdac", "mcclellan-2026-quiescent-pdac-car-t", "tanaka-2012-microscopic-local-recurrence", "dotan-2024-giant", "dotan-2025-giant-ga", "clinicaltrials-nct04233866-giant", "betge-2022-grantpax", "clinicaltrials-nct02143219-pamela70", "mohile-2021-gap70", "li-2021-gain", "ihorst-2026-geriatric-comanagement", "clinicaltrials-nct06040801-frail-gi", "stoop-2026-preopanc4-surgery", "dekker-2026-taps-restaging", "verkolf-2026-taps-resection-prediction", "seelen-2026-pelican", "kunzmann-2021-neolap", "hartlapp-2022-neolap-ca199", "guggenberger-2023-neolap-ct", "napoli-2025-surgery-versus-oncology", "imamura-2004-locally-invasive-surgery-rct", "clinicaltrials-nct06714604-duration", "clinicaltrials-nct06132087-proactive", "baum-2022-fap2286-firsthuman", "mcconathy-2026-lumiere-phase1", "clinicaltrials-nct04939610-lumiere", "baum-2026-3bp3940", "clinicaltrials-nct05432193-frontier", "clinicaltrials-nct07229768-ctr-fapi", "alistar-2017-devimistat-phase1", "philip-2024-avenger500", "surana-2025-erk-hcq-pdac", "clinicaltrials-nct05221320-ulixertinib-hcq", "manji-2023-mekiauto", "clinicaltrials-nct04214418-mekiauto", "clinicaltrials-nct04892017-inlexisertib", "deciphera-2022-dcc3116-phase1", "clinicaltrials-nct04524702-paricalcitol-hcq", "dimcevski-2016-sonoporation-phase1", "wang-2023-sonochemotherapy-pdac", "clinicaltrials-nct04821284-sonoporation", "ius-2026-sonoporation-nct04821284-abstract", "adler-2026-sonoporation-perfusion", "alagesan-2026-sig12d-loder", "ctis-2026-sil204-phase23", "silexion-2026-sil204-site-initiation", "silexion-2026-halfyear-filing", "richards-2006-ci994-pdac", "heumann-2022-azacitidine-pdac", "sohal-2020-thu-decitabine-pdac", "baretti-2024-entinostat-nivolumab-pdac", "clinicaltrials-nct03250273-entinostat-nivolumab", "safyan-2026-azacitidine-pembrolizumab-pdac", "clinicaltrials-nct03264404-azacitidine-pembrolizumab", "rodon-2024-amg193-prmt5", "matsumoto-2026-mtap-pancreatic", "pavlick-2026-mtap-genomics", "clinicaltrials-nct05732831-vopimetostat", "tango-2025-vopimetostat-monotherapy", "clinicaltrials-nct06922591-vopimetostat-ras", "tango-2026-vopimetostat-ras-data", "tango-2026-q2-10q", "mahalingam-2026-elraglusib-randomized", "clinicaltrials-nct03678883-elraglusib", "pathak-2026-elraglusib-folfirinox", "clinicaltrials-nct05077800-elraglusib-folfirinox", "actuate-2026-q2-10q", "sec-2026-actuate-submissions", "chiorean-2023-abemaciclib-pdac", "baghdadi-2019-palbociclib-cdkn2a", "hidalgo-2022-palbociclib-nabpaclitaxel", "laquente-2017-chk1-pdac", "huffman-2023-ly2880070-pdac", "cuneo-2019-adavosertib-pdac", "huffman-2024-azenosertib-pdac-design", "clinicaltrials-nct06015659-azenosertib", "park-2026-gleam-zolbetuximab", "clinicaltrials-nct03816163-gleam", "yu-2025-ibi343-pdac", "clinicaltrials-nct05458219-ibi343", "clinicaltrials-nct07066098-ibi343-phase3", "hao-2024-ibi389-pdac", "xu-2025-ibi389-phase1", "lentz-2025-ebc129-pdac", "clinicaltrials-nct03023722-anetumab", "gou-2026-shr3821", "abou-alfa-2006-exatecan-pdac", "innovent-2026-interim-results", "takeda-2025-ibi343-license", "clinicaltrials-nct03323944-meso-cart", "aznar-2025-mesothelin-car-t-resistance", "qi-2024-ct041-pdac", "clinicaltrials-nct03874897-ct041", "clinicaltrials-nct04581473-ct041", "clinicaltrials-nct05911217-satricel-adjuvant", "carsgen-2025-satricel-pdac-adjuvant", "xu-2026-kras-g12v-tcr", "clinicaltrials-nct04146298-kras-g12v-tcr", "leidner-2022-kras-tcr", "carsgen-2026-interim-results", "clinicaltrials-nct04935359-danis2", "pelletier-2026-nis793-biomarkers", "pietrobono-2026-angiogenin-alk5", "clinicaltrials-nct03336216-cabiralizumab", "padron-2022-prince", "van-laethem-2025-optimize1-biomarkers", "alligator-2026-q2", "kindler-2010-bevacizumab-calgb80303", "van-cutsem-2009-bevacizumab-erlotinib", "rougier-2013-aflibercept-vanilla", "kindler-2011-axitinib-pdac", "clinicaltrials-nct01746979-maestro", "van-cutsem-2016-maestro", "murphy-2019-losartan-lapc", "ramaswamy-2025-afpac-losartan", "clinicaltrials-nct03563248-losartan-localized", "geller-2017-bacterial-gemcitabine", "nakaoka-2024-kestose-pdac", "clinicaltrials-nct04193904-mrx0518-pdac", "sharma-2026-fmtpanc", "ctis-panoramix-2024-517766-41", "middleton-2014-telovac", "le-2015-gvax-crs207", "le-2019-eclipse", "elicio-2026-amplify7p-results", "nct05968326-imcode003", "wainberg-2026-quemliclustat-arc8", "coveler-2024-oleclumab-pdac", "clinicaltrials-nct06608927-prism1", "arcus-2026-q2-10q", "clinicaltrials-nct05254171-aspire", "ctis-2024-514714-12-00-aspire", "singhal-2021-sbp101-phase1", "carla-2024-ivospemin-retinal-atrophy", "panbela-2024-q3-10q", "panbela-2025-note-default", "sec-2025-panbela-registration-abandoned", "clinicaltrials-nct03257033-tigerpac", "hatoum-2024-tamp-rr1-rr2", "novelli-2026-tigerpac-pk", "renovorx-2023-tigerpac-interim", "renovorx-2025-10k", "renovorx-2026-q2-10q", "renovorx-2026-full-enrollment", "timmer-2024-crossfire", "clinicaltrials-nct03899636-direct-rct", "clinicaltrials-nct03899649-direct-registry", "martin-2024-direct-registry-safety", "martin-2026-direct-registry-survival-preprint", "rai-2022-lappie-protocol", "isrctn-14986389-lappie", "nice-2017-ire-pancreatic", "angiodynamics-2026-10k", "clinicaltrials-nct03377491-panova3", "fda-2026-optune-pax-ssed", "clinicaltrials-nct06390059-panova4", "novocure-2026-panova4-topline", "novocure-2026-q2-10q", "guzauskas-2026-ttfields-cost", "liu-2026-ttfields-cost", "issels-2023-heat-rct", "clinicaltrials-nct01077427-heat", "eudract-2008-004802-14-heat", "datta-2017-heatpac-protocol", "clinicaltrials-nct02439593-heatpac", "shinoto-2016-carbon-ion-gemcitabine", "kawashiro-2018-jcros-carbon-ion", "lai-2026-carbon-ion-risk-adapted", "rapp-2022-proton-phase2", "eckstein-2023-proton-registry", "clinicaltrials-nct02598349-proton", "clinicaltrials-nct03536182-cipher", "clinicaltrials-nct04194268-pack", "parikh-2023-smart-phase2-safety", "chuong-2024-smart-phase2-survival", "clinicaltrials-nct03621644-smart", "clinicaltrials-nct05585554-lap-ablate", "clinicaltrials-nct05114213-maspac", "viewray-2023-q1-10q", "viewray-2023-chapter11-8k", "clinicaltrials-nct03690323-pelican", "li-2015-hifu-s1", "marinova-2018-hifu-prospective", "marinova-2024-hifu-pc-protocol", "drks-00012367-hifu-pc", "clinicaltrials-nct05262452-hifu-folfirinox", "clinicaltrials-nct05601323-suizenji", "clinicaltrials-nct06211933-puls-hifu", "clinicaltrials-nct07325214-sonopan3", "izzo-2021-pancreatic-ect-phase12", "izzo-2021-laparoscopic-ect-protocol", "eudract-2018-003925-27-ect", "cebron-2023-panect-preliminary", "clinicaltrials-nct04281290-panect-results", "rudno-rudzinska-2021-irec-protocol", "igea-2026-cliniporator-vitae", "testoni-2021-hybridtherm-rct", "gannon-2026-pancreas-histotripsy-safety", "pusceddu-2026-percutaneous-cryoablation", "zhou-2024-pancreatic-mwa", "ross-2022-panco-p32", "huggett-2014-verteporfin-pdt", "dimatteo-2018-eus-laser-ablation", "clinicaltrials-nct07698613-asp1929-pancreas", "muragaki-2026-k912-sonodynamic", "clinicaltrials-nct05371223-nab-pipac", "padilla-valverde-2024-adjuvant-hipec-rct", "levy-2019-celiac-ganglia-versus-plexus-rct", "xie-2026-cnsi-fe-first-human", "fda-2024-bizengri-multidisciplinary-review", "maltoni-2016-systematic-palliative-qol", "mhra-2020-dpd-testing", "pishvaian-2020-know-your-tumor", "neuzillet-2023-apacap-rct", "nci-cancer-terms-dictionary", "fda-2018-oncology-endpoints", "hopewell-2025-consort" ], "links": [ { "target": "claim-pdac-organoid-correlation-not-routing-utility", "relation": "links-to" }, { "target": "failure-organoid-treatment-chain-attrition", "relation": "links-to" }, { "target": "unknown-pdac-organoid-guided-treatment-utility", "relation": "links-to" }, { "target": "trial-pdac-organoid-action-map", "relation": "links-to" }, { "target": "hypothesis-organoid-guided-early-switch", "relation": "links-to" }, { "target": "claim-staging-laparoscopy-avoids-incision-not-proven-survival", "relation": "found" }, { "target": "failure-staging-yield-without-patient-utility", "relation": "found" }, { "target": "unknown-pdac-staging-laparoscopy-rule", "relation": "found" }, { "target": "trial-pdac-occult-metastasis-action-map", "relation": "found" }, { "target": "hypothesis-pdac-two-timepoint-staging-route", "relation": "changed" }, { "target": "claim-panova3-ttfields-small-os-benefit", "relation": "found" }, { "target": "failure-avatar-precision-medicine-delivery", "relation": "found" }, { "target": "person-liwei-wang", "relation": "found" }, { "target": "trial-enzeavour-japan-screening", "relation": "found" }, { "target": "trial-pdac-ctdna-action-map", "relation": "found" }, { "target": "failure-immune-recruitment-without-response", "relation": "found" }, { "target": "claim-daraxonrasib-kras-amplification-resistance", "relation": "found" }, { "target": "claim-fibrosis-nerve-mechanosensing", "relation": "found" }, { "target": "claim-current-capital-frontier-2026", "relation": "found" }, { "target": "claim-3d-spatial-chemo-resistance-preprint", "relation": "found" }, { "target": "claim-pdac-dormancy-model-proven-human-uncertain", "relation": "found" }, { "target": "claim-kras-egfr-stat3-mouse-regression", "relation": "found" }, { "target": "failure-liaki-undisclosed-conflict-retraction", "relation": "found" }, { "target": "lab-tata-varanasi-gallbladder", "relation": "found" }, { "target": "lab-nccs-planet-liver", "relation": "found" }, { "target": "trial-gallbladder-prevention-detection-action-map", "relation": "found" }, { "target": "failure-nct03762837-risk-cohort-never-started", "relation": "found" }, { "target": "hypothesis-gallbladder-two-stage-interception-rule", "relation": "updates" }, { "target": "trial-small-pnet-action-map", "relation": "found" }, { "target": "failure-aspen-cannot-estimate-surgery-benefit", "relation": "found" }, { "target": "hypothesis-pnet-time-updated-surgery-rule", "relation": "updates" }, { "target": "trial-cca-action-map", "relation": "found" }, { "target": "failure-psc-surveillance-no-curable-detection", "relation": "found" }, { "target": "failure-echec-cancer-outcome-mismatch", "relation": "found" }, { "target": "failure-cca-mrd-prediction-without-action", "relation": "found" }, { "target": "hypothesis-cca-route-specific-action-platform", "relation": "updates" }, { "target": "trial-hcc-action-map", "relation": "found" }, { "target": "failure-hcc-three-month-ultrasound-no-small-cancer-gain", "relation": "found" }, { "target": "failure-hcc-biomarker-addition-no-early-stage-gain", "relation": "found" }, { "target": "hypothesis-hcc-liver-reserve-co-control", "relation": "updates" }, { "target": "ranking-provisional-pdac-bottlenecks", "relation": "updates" }, { "target": "trial-crc-action-map", "relation": "found" }, { "target": "failure-dynamic3-ctdna-guided-chemo-no-rfs-gain", "relation": "found" }, { "target": "failure-altair-mrd-treatment-no-dfs-gain", "relation": "found" }, { "target": "trial-regional-route-completion-action-map", "relation": "found" }, { "target": "failure-ssa-hcc-curative-route-not-reaching-patients", "relation": "found" }, { "target": "hypothesis-regional-route-completion", "relation": "updates" }, { "target": "trial-pdac-cachexia-action-map", "relation": "found" }, { "target": "failure-tocilizumab-cachexia-muscle-without-survival", "relation": "found" }, { "target": "unknown-cachexia-treatment-utility", "relation": "updates" }, { "target": "trial-pert-pdac-action-map", "relation": "found" }, { "target": "failure-pert-short-randomized-trials", "relation": "found" }, { "target": "failure-pert-prescribing-and-supply-gap", "relation": "found" }, { "target": "unknown-pert-pdac-utility", "relation": "updates" }, { "target": "claim-anticoagulation-prevents-pdac-vte-not-survival", "relation": "found" }, { "target": "failure-pdac-vte-prophylaxis-delivery-gap", "relation": "found" }, { "target": "failure-impact-pro-status-unknown", "relation": "found" }, { "target": "trial-pdac-vte-action-map", "relation": "found" }, { "target": "claim-kinisoquin-phase2-marker-not-clinical-proof", "relation": "found" }, { "target": "company-quercis-pharma", "relation": "found" }, { "target": "claim-kinisoquin-access-stack", "relation": "found" }, { "target": "unknown-pdac-vte-prophylaxis-rule", "relation": "updates" }, { "target": "hypothesis-pdac-vte-route-completion", "relation": "updates" }, { "target": "claim-biliary-drainage-is-route-not-cancer-treatment", "relation": "found" }, { "target": "failure-routine-plastic-pbd", "relation": "found" }, { "target": "failure-biliary-stent-treatment-interruption", "relation": "found" }, { "target": "unknown-pdac-biliary-drainage-route", "relation": "updates" }, { "target": "trial-pdac-biliary-drainage-action-map", "relation": "found" }, { "target": "hypothesis-pdac-biliary-route-completion", "relation": "updates" }, { "target": "company-boston-scientific-biliary", "relation": "found" }, { "target": "claim-universal-germline-testing-needs-action", "relation": "found" }, { "target": "failure-pdac-germline-route-dropout", "relation": "found" }, { "target": "system-england-r367-germline-criteria", "relation": "found" }, { "target": "unknown-pdac-germline-family-action-route", "relation": "updates" }, { "target": "trial-pdac-germline-cascade-action-map", "relation": "found" }, { "target": "hypothesis-pdac-germline-family-route-completion", "relation": "updates" }, { "target": "claim-pdac-surgery-is-a-complete-treatment-route", "relation": "links-to" }, { "target": "system-england-pancreatic-surgery-network", "relation": "links-to" }, { "target": "failure-pdac-postoperative-treatment-loss", "relation": "links-to" }, { "target": "failure-leopard2-laparoscopic-whipple-safety", "relation": "links-to" }, { "target": "unknown-pdac-surgical-route-completion", "relation": "links-to" }, { "target": "trial-pdac-surgical-route-action-map", "relation": "links-to" }, { "target": "hypothesis-pdac-surgical-route-completion", "relation": "links-to" }, { "target": "claim-routine-imaging-after-pdac-resection-not-yet-proven", "relation": "links-to" }, { "target": "unknown-postresection-surveillance-utility", "relation": "links-to" }, { "target": "trial-pdac-postresection-surveillance-action-map", "relation": "links-to" }, { "target": "claim-neoadjuvant-therapy-improves-selection-not-cure", "relation": "links-to" }, { "target": "failure-norpact1-neoadjuvant-delivery", "relation": "links-to" }, { "target": "unknown-pdac-perioperative-treatment-route", "relation": "links-to" }, { "target": "trial-pdac-perioperative-sequencing-action-map", "relation": "links-to" }, { "target": "claim-human-pdac-dtc-prognostic-not-dormancy", "relation": "links-to" }, { "target": "failure-pdac-dormancy-sample-gap", "relation": "links-to" }, { "target": "trial-pdac-residual-cell-state-action-map", "relation": "links-to" }, { "target": "hypothesis-human-dormancy-reawakening-gate", "relation": "links-to" }, { "target": "claim-age-alone-cannot-route-pdac-treatment", "relation": "links-to" }, { "target": "failure-grantpax-geriatric-route", "relation": "links-to" }, { "target": "unknown-older-pdac-treatment-route", "relation": "links-to" }, { "target": "trial-older-pdac-action-map", "relation": "links-to" }, { "target": "hypothesis-older-pdac-fast-support-route", "relation": "links-to" }, { "target": "claim-lapc-response-warrants-surgical-reassessment", "relation": "links-to" }, { "target": "failure-conversion-surgery-selection-bias", "relation": "links-to" }, { "target": "unknown-lapc-surgery-causal-benefit", "relation": "links-to" }, { "target": "trial-lapc-conversion-surgery-action-map", "relation": "links-to" }, { "target": "hypothesis-randomized-lapc-resection-after-response", "relation": "links-to" }, { "target": "claim-fap-radioligand-feasible-pdac-efficacy-unproven", "relation": "links-to" }, { "target": "failure-fapi-scan-dose-benefit-collapse", "relation": "links-to" }, { "target": "unknown-pdac-fap-radioligand-treatment-effect", "relation": "links-to" }, { "target": "trial-pdac-fap-radioligand-action-map", "relation": "links-to" }, { "target": "hypothesis-fapi-theranostic-pdac", "relation": "links-to" }, { "target": "claim-hcq-combinations-fail-pdac-not-autophagy", "relation": "found" }, { "target": "failure-metabolic-repurposing-pdac", "relation": "found" }, { "target": "unknown-selective-autophagy-pdac-contribution", "relation": "found" }, { "target": "trial-pdac-autophagy-action-map", "relation": "found" }, { "target": "hypothesis-tumour-verified-selective-autophagy-blockade", "relation": "found" }, { "target": "claim-sonochemotherapy-randomized-signal", "relation": "corrects" }, { "target": "failure-sonoporation-overall-replication-subgroup-rescue", "relation": "records" }, { "target": "unknown-pdac-sonoporation-treatment-effect", "relation": "leaves-open" }, { "target": "trial-pdac-sonoporation-action-map", "relation": "maps" }, { "target": "hypothesis-measured-fixed-backbone-sonoporation", "relation": "gates" }, { "target": "claim-sil204-trial-scales-before-human-knockdown", "relation": "records" }, { "target": "failure-sig12d-loder-randomized-os-not-shown", "relation": "corrects" }, { "target": "unknown-sil204-human-target-delivery", "relation": "leaves-open" }, { "target": "trial-sil204-action-map", "relation": "maps" }, { "target": "hypothesis-multicompartment-kras-sirna-delivery", "relation": "gates" }, { "target": "failure-broad-epigenetic-treatment-pdac", "relation": "learns-from" }, { "target": "claim-mtap-prmt5-selected-human-signal", "relation": "bounded-by" }, { "target": "failure-vopimetostat-ras-contribution-unknown", "relation": "learns-from" }, { "target": "unknown-mtap-prmt5-ras-contribution", "relation": "leaves-open" }, { "target": "trial-mtap-prmt5-action-map", "relation": "implements" }, { "target": "hypothesis-mtap-prmt5-ras-deeper-control", "relation": "tests" }, { "target": "company-tango-therapeutics", "relation": "contains" }, { "target": "claim-elraglusib-randomized-survival-signal", "relation": "bounded-by" }, { "target": "failure-elraglusib-analysis-and-survival-discordance", "relation": "learns-from" }, { "target": "unknown-elraglusib-survival-replication", "relation": "leaves-open" }, { "target": "trial-elraglusib-action-map", "relation": "implements" }, { "target": "hypothesis-elraglusib-survival-confirmation-gate", "relation": "tests" }, { "target": "company-actuate-therapeutics", "relation": "contains" }, { "target": "failure-broad-cell-cycle-checkpoint-pdac", "relation": "learns-from" }, { "target": "unknown-wee1-replication-stress-pdac", "relation": "leaves-open" }, { "target": "trial-cell-cycle-checkpoint-action-map", "relation": "implements" }, { "target": "hypothesis-human-replication-stress-selector", "relation": "tests" }, { "target": "claim-cldn182-adc-selected-response", "relation": "contains" }, { "target": "failure-antibody-target-and-payload-transfer-pdac", "relation": "contains" }, { "target": "unknown-cldn182-adc-survival-effect", "relation": "contains" }, { "target": "trial-cldn182-antibody-action-map", "relation": "mapped-by" }, { "target": "hypothesis-cldn182-adc-survival-test", "relation": "tests" }, { "target": "company-innovent-biologics", "relation": "contains" }, { "target": "claim-satricel-pdac-early-signal", "relation": "contains" }, { "target": "claim-kras-g12v-tcr-early-signal", "relation": "contains" }, { "target": "failure-cell-therapy-route-and-denominator-pdac", "relation": "contains" }, { "target": "unknown-engineered-cell-reproducible-benefit-pdac", "relation": "contains" }, { "target": "trial-pdac-engineered-cell-action-map", "relation": "mapped-by" }, { "target": "hypothesis-kras-tcr-human-persistence-gate", "relation": "tests" }, { "target": "company-carsgen-therapeutics", "relation": "contains" }, { "target": "claim-nis793-target-engagement-without-benefit", "relation": "contains" }, { "target": "claim-cd40-signal-without-contribution-proof", "relation": "contains" }, { "target": "failure-immune-reprogramming-without-patient-benefit", "relation": "contains" }, { "target": "unknown-cd40-added-survival-pdac", "relation": "contains" }, { "target": "trial-immune-reprogramming-action-map", "relation": "mapped-by" }, { "target": "hypothesis-angiogenin-tgfb-existing-data-test", "relation": "tests" }, { "target": "hypothesis-mitazalimab-controlled-survival-gate", "relation": "tests" }, { "target": "company-alligator-bioscience", "relation": "contains" }, { "target": "failure-vascular-starvation-pdac", "relation": "contains" }, { "target": "claim-hypoxia-control-without-survival-pdac", "relation": "contains" }, { "target": "failure-losartan-delivery-signal-not-replicated", "relation": "contains" }, { "target": "unknown-localized-losartan-contribution-pdac", "relation": "contains" }, { "target": "trial-pdac-vascular-hypoxia-action-map", "relation": "mapped-by" }, { "target": "hypothesis-localized-losartan-existing-trial-gate", "relation": "tests" }, { "target": "claim-pdac-microbiome-mechanism-without-human-benefit", "relation": "contains" }, { "target": "failure-microbiome-proxy-without-patient-benefit", "relation": "contains" }, { "target": "unknown-microbiome-treatment-contribution-pdac", "relation": "contains" }, { "target": "trial-pdac-microbiome-action-map", "relation": "mapped-by" }, { "target": "hypothesis-bacterial-gemcitabine-deactivation-gate", "relation": "tests" }, { "target": "claim-pdac-vaccine-immunity-without-proven-benefit", "relation": "contains" }, { "target": "failure-pdac-vaccine-escalation-reversals", "relation": "contains" }, { "target": "unknown-pdac-vaccine-regimen-contribution", "relation": "contains" }, { "target": "trial-pdac-vaccine-action-map", "relation": "mapped-by" }, { "target": "hypothesis-autogene-cevumeran-existing-trial-gate", "relation": "tests" }, { "target": "claim-cd73-blockade-signal-without-concurrent-control", "relation": "contains" }, { "target": "failure-adenosine-control-and-marker-pdac", "relation": "contains" }, { "target": "unknown-quemliclustat-survival-contribution", "relation": "contains" }, { "target": "trial-pdac-adenosine-action-map", "relation": "mapped-by" }, { "target": "hypothesis-quemliclustat-existing-phase3-gate", "relation": "tests" }, { "target": "company-arcus-biosciences", "relation": "contains" }, { "target": "claim-ivospemin-signal-without-controlled-benefit", "relation": "contains" }, { "target": "failure-aspire-trial-operational-collapse", "relation": "contains" }, { "target": "unknown-aspire-assigned-denominator-and-results", "relation": "contains" }, { "target": "trial-pdac-polyamine-action-map", "relation": "mapped-by" }, { "target": "hypothesis-aspire-data-rescue-gate", "relation": "tests" }, { "target": "company-panbela-therapeutics", "relation": "contains" }, { "target": "claim-tamp-blood-exposure-not-tumour-delivery", "relation": "contains" }, { "target": "failure-tigerpac-component-and-interim-inference", "relation": "contains" }, { "target": "unknown-tigerpac-package-survival-and-delivery", "relation": "contains" }, { "target": "trial-pdac-intraarterial-delivery-action-map", "relation": "mapped-by" }, { "target": "hypothesis-tigerpac-existing-final-gate", "relation": "tests" }, { "target": "company-renovorx", "relation": "contains" }, { "target": "claim-ire-ablation-without-proven-added-survival", "relation": "contains" }, { "target": "failure-direct-selection-and-randomized-reporting-gap", "relation": "contains" }, { "target": "unknown-ire-added-survival-after-modern-chemotherapy", "relation": "contains" }, { "target": "trial-pdac-irreversible-electroporation-action-map", "relation": "mapped-by" }, { "target": "hypothesis-direct-randomized-data-recovery-gate", "relation": "tests" }, { "target": "company-angiodynamics", "relation": "contains" }, { "target": "claim-panova3-ttfields-small-os-benefit", "relation": "contains" }, { "target": "failure-panova4-component-and-historical-control", "relation": "contains" }, { "target": "unknown-ttfields-survival-replication-and-access", "relation": "contains" }, { "target": "trial-pdac-ttfields-action-map", "relation": "mapped-by" }, { "target": "hypothesis-optune-pax-independent-value-gate", "relation": "tests" }, { "target": "company-novocure", "relation": "contains" }, { "target": "regulatory-fda-optune-pax-2026", "relation": "contains" }, { "target": "claim-heat-regional-hyperthermia-primary-endpoint-negative", "relation": "contains" }, { "target": "failure-heat-component-and-registry-drift", "relation": "contains" }, { "target": "unknown-regional-hyperthermia-added-benefit-modern-care", "relation": "contains" }, { "target": "trial-pdac-regional-hyperthermia-action-map", "relation": "mapped-by" }, { "target": "hypothesis-heat-existing-randomization-recovery-gate", "relation": "tests" }, { "target": "claim-particle-therapy-local-control-without-survival-proof", "relation": "contains" }, { "target": "failure-particle-therapy-comparator-and-accrual", "relation": "contains" }, { "target": "unknown-particle-therapy-added-patient-benefit", "relation": "contains" }, { "target": "trial-pdac-particle-therapy-action-map", "relation": "mapped-by" }, { "target": "hypothesis-particle-therapy-existing-data-gate", "relation": "tests" }, { "target": "claim-smart-delivery-and-safety-without-added-survival-proof", "relation": "contains" }, { "target": "failure-smart-trial-and-vendor-dependency", "relation": "contains" }, { "target": "unknown-smart-added-survival-and-surgery-harm", "relation": "contains" }, { "target": "trial-pdac-smart-action-map", "relation": "mapped-by" }, { "target": "hypothesis-smart-vendor-independent-randomized-gate", "relation": "tests" }, { "target": "company-viewray", "relation": "contains" }, { "target": "claim-pelican-rfa-no-survival-and-more-harm", "relation": "contains" }, { "target": "failure-rfa-selection-and-route-attrition", "relation": "contains" }, { "target": "unknown-rfa-human-mechanism-after-negative-trial", "relation": "contains" }, { "target": "trial-pdac-rfa-action-map", "relation": "mapped-by" }, { "target": "hypothesis-pelican-specimen-salvage-gate", "relation": "tests" }, { "target": "claim-hifu-pain-signal-without-survival-proof", "relation": "contains" }, { "target": "failure-hifu-randomization-and-trial-drift", "relation": "contains" }, { "target": "unknown-hifu-pain-survival-and-system-transfer", "relation": "contains" }, { "target": "trial-pdac-hifu-action-map", "relation": "mapped-by" }, { "target": "hypothesis-hifu-existing-randomized-gate", "relation": "tests" }, { "target": "claim-electrochemotherapy-local-effect-without-patient-benefit", "relation": "contains" }, { "target": "failure-electrochemotherapy-endpoint-and-trial-drift", "relation": "contains" }, { "target": "unknown-electrochemotherapy-added-benefit-and-delivery", "relation": "contains" }, { "target": "trial-pdac-electrochemotherapy-action-map", "relation": "mapped-by" }, { "target": "hypothesis-electrochemotherapy-randomized-data-gate", "relation": "tests" }, { "target": "company-igea", "relation": "contains" }, { "target": "trial-pdac-eus-rfa-action-map", "relation": "mapped-by" }, { "target": "trial-pdac-histotripsy-action-map", "relation": "mapped-by" }, { "target": "trial-pdac-cryoablation-action-map", "relation": "mapped-by" }, { "target": "trial-pdac-mwa-action-map", "relation": "mapped-by" }, { "target": "trial-pdac-p32-action-map", "relation": "mapped-by" }, { "target": "trial-pdac-pdt-action-map", "relation": "mapped-by" }, { "target": "trial-pdac-laser-ablation-action-map", "relation": "mapped-by" }, { "target": "trial-pdac-asp1929-action-map", "relation": "mapped-by" }, { "target": "trial-pdac-k912-sonodynamic-action-map", "relation": "mapped-by" }, { "target": "trial-pdac-pipac-action-map", "relation": "mapped-by" }, { "target": "trial-pdac-hipec-action-map", "relation": "mapped-by" }, { "target": "trial-pdac-celiac-pain-action-map", "relation": "mapped-by" }, { "target": "trial-pdac-ferroptosis-action-map", "relation": "mapped-by" }, { "target": "trial-pdac-nrg1-action-map", "relation": "mapped-by" }, { "target": "trial-pdac-palliative-care-action-map", "relation": "mapped-by" }, { "target": "trial-pdac-pharmacogenetic-safety-action-map", "relation": "mapped-by" }, { "target": "trial-pdac-kras-wildtype-action-map", "relation": "mapped-by" }, { "target": "trial-pdac-exercise-action-map", "relation": "mapped-by" }, { "target": "trial-2026-randomised-result-stress-test", "relation": "contains" }, { "target": "claim-elemental-diet-survival-signal-needs-replication", "relation": "contains" }, { "target": "failure-mistletoe-pdac-no-quality-benefit", "relation": "contains" }, { "target": "ai-programme", "relation": "audits" }, { "target": "trial-pdac-germline-cascade-action-map", "relation": "audits" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "strong", "randomised": "strong", "independentReplication": "weak — mixed evidence", "sampleSize": "79 directed search passes spanning trials, papers, registries, patents, companies, guidance and preprints, including PDAC treatment, detection, delivery, supportive care, outside-field transfer, interpretation, source access and related cancers", "effect": "Audit record; effects remain in linked evidence", "limits": [ "Search engines are not exhaustive", "Some company and preprint results lack independent confirmation", "Patent families remain incomplete", "Multilingual coverage is still selective", "The stopping rule is decision-specific, not a claim that future research has no value" ] }, "layout": { "x": 88, "y": 122 }, "updated": "2026-09-15" }, { "id": "residual-disease", "title": "Residual disease and recurrence", "shortTitle": "Residual disease", "type": "system", "status": "supported", "scope": [ "pdac", "pnet", "cholangiocarcinoma", "gallbladder", "hcc", "colorectal" ], "summary": "Residual disease is measurable as recurrence risk, but its cell states, locations and treatment rules remain the central missing clinical chain.", "content": "Residual disease is not one object. It can include continuously growing micrometastases, treatment-tolerant changing cell states, dormant disseminated cells or locally persistent tumour below imaging resolution. In selected resected PDAC treated with modified FOLFIRINOX, five-year disease-free survival was 26.1%, so 73.9% had recurrence or death. Tumour-informed ctDNA strongly separates recurrence risk, but blood-negative disease can remain in poorly shedding sites and no completed PDAC trial proves that changing treatment from the result improves survival. The size of that blind spot is material: a 2026 series reported 35.7% postoperative sensitivity and 53.2% negative predictive value, improving later but still missing recurrence. Another small cohort reported a median 4.59-month lead time, which creates an opportunity only if an effective action exists. DYNAMIC-Pancreas, CIRCPAC and MAP-03 now test three different actions; none has supplied the missing randomized utility result. Dormancy is causally demonstrated in PDAC models: residual cells can survive visible regression, and a new model links chemotherapy-induced liver injury to reawakening through neutrophil extracellular traps. The human state and route are not established. The practical programme must preserve four things in one timeline: exact treatment exposure, repeated blood and imaging, tissue at baseline and progression when safe, and complete recurrence and survival. It must distinguish detection from action. A marker earns use only if a fixed intervention improves survival, treatment-free time or quality of life against a concealed-result control. The first target is not every possible residual cell; it is a reproduced state covering enough recurrence to justify a safe low-burden trial.", "contentSections": [ "Cholangiocarcinoma supplies an independent residual-disease test of the same rule. A 56-person biliary cohort detected 15 of 16 recurrences by ctDNA with an average 3.7-month lead, and the 89-person STAMP extrahepatic cohort repeated strong prognostic separation during adjuvant therapy. Neither changed treatment from ctDNA. A separate 18-person prospective feasibility study analysed only 14 and made 12 postoperative comparisons. This strengthens recurrence prediction but leaves treatment utility, assay completion and site-specific action open.", "Standard imaging utility also remains open. Two non-randomized resection cohorts associate routine scans with more recurrence treatment and longer survival, but NCT04875325 RADAR-PANC already tests three-monthly CA19-9 plus CT against mainly symptom-led follow-up with overall survival primary. That trial must report before a duplicate scan schedule is funded; CIRCPAC remains a separate ctDNA-directed intensity test.", "Human occult-cell evidence is prognostic, not a dormant-state proof. Bone-marrow assays found signals in 13.7% of 175 surgical patients and 31% of 48 inoperable patients, with roughly doubled death hazards, but did not prove viability or ancestry. Primary-tumour and rapid-autopsy maps show organ-linked and changing states at the wrong times. The missing sample is the residual cell during the disease-free interval, linked by private alterations to later recurrence." ], "sourceIds": [ "conroy-2022-prodige24-five-year", "botta-2024-tumor-informed-ctdna", "guo-2026-postsurgical-ctdna", "chen-2026-personalized-ctdna", "dynamic-pancreas-registry", "clinicaltrials-nct05788744-circpac", "clinicaltrials-nct05802394-map03", "lin-2013-reversible-pdac-dormancy", "shi-2026-chemo-dormant-dtc", "yu-2025-btc-ctdna-recurrence", "yoo-2025-extrahepatic-cca-ctdna", "park-2026-btc-ctdna-feasibility", "andel-2025-routine-imaging", "van-goor-2025-nationwide-imaging", "daamen-2024-radar-panc-protocol", "clinicaltrials-nct04875325-radar-panc", "tai-2025-pan-asian-esmo-pancreatic", "effenberger-2012-pdac-bone-marrow-dtc", "nordgard-2022-unresectable-dtc", "chalabi-hajkarim-2025-organotropism", "pei-2025-spatial-metastatic-pdac", "mcclellan-2026-quiescent-pdac-car-t", "tanaka-2012-microscopic-local-recurrence" ], "links": [ { "target": "bottlenecks", "relation": "supports" }, { "target": "experiments", "relation": "tests" }, { "target": "claim-pdac-dormancy-model-proven-human-uncertain", "relation": "contains" }, { "target": "unknown-dormant-cell-actionability", "relation": "contains" }, { "target": "unknown-ctdna-actionability", "relation": "contains" }, { "target": "trial-pdac-ctdna-action-map", "relation": "tested-by" }, { "target": "programme-25m-prospective-observatory", "relation": "measured-by" }, { "target": "claim-cca-ctdna-prognostic-not-action-guiding", "relation": "contains" }, { "target": "unknown-cca-postoperative-mrd-action", "relation": "contains" }, { "target": "claim-routine-imaging-after-pdac-resection-not-yet-proven", "relation": "links-to" }, { "target": "unknown-postresection-surveillance-utility", "relation": "links-to" }, { "target": "trial-pdac-postresection-surveillance-action-map", "relation": "links-to" }, { "target": "claim-human-pdac-dtc-prognostic-not-dormancy", "relation": "links-to" }, { "target": "failure-pdac-dormancy-sample-gap", "relation": "links-to" }, { "target": "trial-pdac-residual-cell-state-action-map", "relation": "links-to" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "strong", "randomised": "moderate", "independentReplication": "moderate", "sampleSize": "Five-year randomized adjuvant follow-up, a 298-patient commercial ctDNA cohort, two 2026 performance cohorts, three active action programmes and two PDAC dormancy model programmes", "effect": "Five-year disease-free survival 26.1% after resection and modified FOLFIRINOX; ctDNA is prognostic with variable sensitivity, while action remains unproved", "limits": [ "Residual tissue is rarely sampled", "ctDNA sensitivity varies by burden and site", "Dormancy causality is model-derived", "No completed randomized PDAC ctDNA-action trial", "Recurrence can arise through more than one route" ] }, "layout": { "x": 79, "y": 52 }, "updated": "2026-09-14" }, { "id": "risks", "title": "Mission risks", "shortTitle": "Risks", "type": "regulatory", "status": "supported", "scope": [ "pdac", "pnet", "cholangiocarcinoma", "gallbladder", "hcc", "colorectal", "pan-cancer" ], "summary": "The programme has ten named failure routes with measurable warning signs, mitigations, owners and stop rules.", "content": "Scientific risk — RAS treatment remains temporary because several bypass routes emerge at once. Warning: no resistance class covers at least 25% of progression. Mitigation: collect progression tissue before choosing combinations. Owner: resistance lead. Stop: no reproducible class across two laboratories. Clinical risk — an assay predicts recurrence but changing treatment does not help. Warning: high prognostic separation with no treatment interaction. Mitigation: randomized reveal-versus-hide design. Owner: independent trial chair. Stop: action arm crosses futility or harm boundary. Safety risk — deeper RAS, immune or local combinations injure normal gut, lung, marrow or liver. Warning: exposure-linked grade 3–5 events or treatment-related deaths. Mitigation: staged dosing, organ-specific monitoring and independent safety board. Stop: fixed posterior probability that serious harm exceeds the protocol bound. Detection risk — false positives cause invasive imaging, EUS, biopsy or pancreatic surgery without finding lethal disease. Warning: positive predictive value below the locked minimum or procedure harm exceeds cancers found. Mitigation: risk enrichment and confirmatory steps. Stop: net-benefit model is negative using observed harms. Manufacturing risk — personalised vaccines, assays or cell products arrive after the treatment window. Warning: fewer than 80% released on time. Mitigation: parallel manufacture, fixed specifications and backup sites. Stop: timely yield below 70%. Regulatory risk — a changing assay or AI model breaks trial validity. Warning: version drift or site calibration failure. Mitigation: locked versions, traceable updates and regulator meetings before pivotal use. Owner: regulatory lead. Stop: primary endpoint cannot be mapped to the approved product. Organisational and financial risk — partners control compounds or data, comparators become obsolete, or £500m is committed before gates. Warning: missing data rights, publication vetoes, concentration in one company, or schedule slippage above 25%. Mitigation: staged capital, independent analysis, negative-result publication and replaceable vendors. Owner: programme board. Stop: no enforceable access or more than £500m exposed before randomized signal. These risks are derived from observed failures: AVATAR failed delivery, GRAIL scaled before utility, vaccines produced immune signals without efficacy, and local or stromal interventions changed intermediate measures without durable survival. Clot-prevention risk — a proven class effect is lost through late assessment, non-fill or early discontinuation, or bleeding rises when a broad high-risk label replaces an individual check. Warning: fewer than 80% have a decision within seven days, more than 10% cannot obtain the chosen drug, or major bleeding rises by two points. Mitigation: name the owner, record bleeding and interaction checks, patient choice, fill, use and stop reason, and separate symptomatic from incidental VTE. Owner: supportive-care and thrombosis lead. Stop: fatal-bleeding imbalance, no symptomatic-VTE reduction or benefit confined to scan-detected events. Biliary-route risk — unnecessary drainage adds infection and delay, while necessary drainage fails before chemotherapy or surgery. Warning: no documented indication within 48 hours, plastic use when drainage must span neoadjuvant treatment, cholangitis or obstruction interruption above 10%, or missing rescue ownership. Mitigation: triage prompt surgery versus drainage first, use durable internal drainage when needed, specify expert EUS or percutaneous rescue by anatomy and stability, and track every treatment date. Owner: pancreatic surgeon and therapeutic-endoscopy lead together. Stop: serious procedure harm up five points, 30-day death imbalance, surgery delayed over seven additional days or no treatment-completion gain. Germline-route risk — testing is offered late, results are acted on incorrectly, or relatives are never reached. Warning: fewer than 85% of patients receive a result within 21 days, fewer than 40% of eligible reached relatives test, any care is based on a variant of uncertain significance, or completion differs by more than ten points across ethnicity, language, deprivation or centre. Mitigation: automatic point-of-care offer, translated consent, genetics review of pathogenic results, explicit permission before direct family contact, free home testing and complete denominators. Owner: oncology and clinical-genetics leads together. Stop or amend for a material privacy breach, uncertain-result action, median turnaround above 21 days, discrimination harm, a greater than ten-point inequality gap or no positive net benefit outside R367.", "sourceIds": [ "oreilly-2026-daraxonrasib", "sarno-2025-avatar", "sasieni-2026-nhs-galleri", "elicio-2026-amplify7p-results", "hammel-2016-lap07", "hingorani-2020-halo301", "frere-2020-pdac-thromboprophylaxis-meta", "heffley-2024-pdac-vte-uptake", "noor-2026-pdac-vte-implementation", "ash-2023-primary-vte-prophylaxis-guide", "van-der-gaag-2010-pbd-rct", "costamagna-2026-sems-early-surgery-rct", "vehvilainen-2024-cholangitis-nat-interruption", "lyu-2023-metal-plastic-nat-meta", "asge-2024-therapeutic-eus-guideline", "nhs-england-2026-r367", "wang-2022-oncology-germline-cascade", "rodriguez-2024-generate", "katz-2026-gift-cascade", "npaca-2026-state-of-nation", "nhs-england-2024-hpb-pancreatic-service", "nice-2018-ng85-pancreatic", "de-wilde-2012-centralization", "lemmens-2011-centralization", "latenstein-2021-centralization", "coupland-2016-england-resection-volume", "labori-2016-adjuvant-completion", "henry-2023-complications-adjuvant", "kollbeck-2025-complications-chemo", "van-hilst-2019-leopard2", "korrel-2023-diploma", "bruna-2025-diploma-long-term", "klotz-2024-europa", "liu-2024-robotic-open-pd", "jin-2026-portal" ], "links": [ { "target": "failures", "relation": "depends-on" }, { "target": "programme", "relation": "supports" }, { "target": "unknowns", "relation": "supports" }, { "target": "failure-pdac-vte-prophylaxis-delivery-gap", "relation": "learns-from" }, { "target": "unknown-pdac-vte-prophylaxis-rule", "relation": "constrained-by" }, { "target": "hypothesis-pdac-vte-route-completion", "relation": "controlled-by" }, { "target": "failure-routine-plastic-pbd", "relation": "learns-from" }, { "target": "failure-biliary-stent-treatment-interruption", "relation": "learns-from" }, { "target": "unknown-pdac-biliary-drainage-route", "relation": "constrained-by" }, { "target": "hypothesis-pdac-biliary-route-completion", "relation": "controlled-by" }, { "target": "failure-pdac-germline-route-dropout", "relation": "learns-from" }, { "target": "unknown-pdac-germline-family-action-route", "relation": "constrained-by" }, { "target": "hypothesis-pdac-germline-family-route-completion", "relation": "controlled-by" }, { "target": "claim-pdac-surgery-is-a-complete-treatment-route", "relation": "links-to" }, { "target": "system-england-pancreatic-surgery-network", "relation": "links-to" }, { "target": "failure-pdac-postoperative-treatment-loss", "relation": "links-to" }, { "target": "failure-leopard2-laparoscopic-whipple-safety", "relation": "links-to" }, { "target": "unknown-pdac-surgical-route-completion", "relation": "links-to" }, { "target": "trial-pdac-surgical-route-action-map", "relation": "links-to" }, { "target": "hypothesis-pdac-surgical-route-completion", "relation": "links-to" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "moderate", "randomised": "strong", "independentReplication": "moderate", "sampleSize": "Ten risk classes derived from treatment, detection, delivery and programme failures", "effect": "Risk-control design; thresholds are proposed programme gates", "limits": [ "Several warning thresholds are judgement", "Safety bounds require protocol-specific statistical design", "Commercial and regulatory conditions can change", "Observed failures may not predict every new implementation" ] }, "layout": { "x": 66, "y": 115 }, "updated": "2026-09-14", "contentSections": [ "Surgical-route risks are operative death, leak, organ failure, delayed or omitted systemic treatment, false pressure to operate, loss of patient choice, travel burden, age or deprivation gaps, and platform adoption ahead of proficiency. Hard stops are 90-day mortality above 5% or more than one point higher than matched control, serious complications five points higher, an unexplained five-point operation-rate fall, or inequality widening by ten points. The programme funds route staff and independent analysis, not robots." ] }, { "id": "system-england-pancreatic-surgery-network", "title": "England commissions pancreatic surgery as a specialist network, not a stand-alone operation", "shortTitle": "England surgery network", "type": "system", "status": "supported", "scope": [ "pdac", "pancreatic-all" ], "summary": "The service rule requires at least 80 neoplastic pancreatic procedures per team each year and a connected referral, operation, recovery and oncology route.", "content": "NHS England's September 2024 service specification requires pancreatic and periampullary cancer services to cover catchments of two to four million people and complete at least 80 pancreatic surgical procedures for neoplastic or suspected neoplastic disease per team per year. It also requires specialist MDT review, prompt network referral, clear responsibility across providers, 24/7 interventional and postoperative cover, prehabilitation, enhanced recovery, specialist dietetics and return of outcomes to the specialist MDT. The 2026 national audit reports surgical indicators across 23 English HPB specialist centres. This is a useful operating base, but a service specification is a rule, not evidence of compliance or benefit. The audit shows 77% of all English pancreatic diagnoses had a recorded MDT discussion, treatment waits had a median of 74 days, and only 62% of Whipple patients received chemotherapy or chemoradiotherapy within 14 weeks. The network must therefore be judged from the first eligible scan to the last planned treatment, with each loss and its cause recorded. Volume remains a capability screen, not a guaranteed outcome.", "sourceIds": [ "nhs-england-2024-hpb-pancreatic-service", "npaca-2026-state-of-nation", "nice-2018-ng85-pancreatic", "de-wilde-2012-centralization", "lemmens-2011-centralization", "latenstein-2021-centralization", "coupland-2016-england-resection-volume" ], "links": [ { "target": "disease-system", "relation": "part-of" }, { "target": "claim-pdac-surgery-is-a-complete-treatment-route", "relation": "delivers" }, { "target": "failure-pdac-postoperative-treatment-loss", "relation": "must-repair" }, { "target": "unknown-pdac-surgical-route-completion", "relation": "must-measure" }, { "target": "hypothesis-pdac-surgical-route-completion", "relation": "hosts" } ], "evidence": { "humanProspective": "moderate for routine audit", "humanRetrospective": "strong for volume and regional associations", "randomised": "none for centralization", "independentReplication": "moderate", "sampleSize": "English audit n=17,672 diagnoses and 23 centres; Dutch centralization cohorts n=2,155, n=2,129 and n=10,079; English historical cohort n=31,973", "effect": "Commissioned minimum 80 neoplastic or suspected-neoplastic pancreatic procedures per team per year; observed route gaps remain", "limits": [ "Commissioned standard is not measured compliance", "Centralization studies are observational", "Mixed pancreatic and periampullary indications", "Era changes coincide with reorganization", "Volume does not identify the active care component" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Complex resection, rescue and systemic treatment require coordinated specialist capability." }, "animalEvidence": { "rating": "not-applicable", "reason": "This is a human service design." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "No laboratory model can test network delivery." }, "reproducibility": { "rating": "moderate", "reason": "Several populations associate concentration with safer care, while English volume-survival precision is weaker." }, "commercialConflicts": { "rating": "low", "reason": "The key sources are public specifications, national audit and registry studies." }, "recency": { "rating": "current", "reason": "The current operating boundary comes from 2024 commissioning and the September 2026 audit." } } }, "updated": "2026-09-14" }, { "id": "system-england-r367-germline-criteria", "title": "England's 2026 inherited-pancreatic-cancer test rule remains age- and history-limited", "shortTitle": "England R367 rule", "type": "system", "status": "supported", "scope": [ "pdac", "pancreatic-all" ], "summary": "NHS England commissions R367 for defined age and family-history groups, leaving a measurable difference from all-patient exocrine testing guidance.", "content": "The live NHS England rare and inherited disease directory, version 9.1 dated 20 May 2026, commissions R367 for pancreatic adenocarcinoma diagnosed before age 60, or before 70 when specified personal or family cancers are present. Testing outside the criteria can be considered after specialist genetics multidisciplinary review. This is narrower than ASCO's 2024 all-patient indication for exocrine pancreatic cancer. It also creates a direct test against the Spanish result in which 7/14 pathogenic or likely pathogenic carriers would not have met clinical criteria. The difference does not prove NHS policy is wrong: the Spanish cohort was small, young and routed through high-risk clinics; US guidance does not supply UK cost, capacity or downstream-action estimates. The needed evidence is the variant and action yield among English patients inside and outside R367, time to result, family uptake, inequalities, cost and the harms of uncertain findings. Until those data exist, describe R367 as a current commissioning boundary, not a biological boundary.", "sourceIds": [ "nhs-england-2026-r367", "asco-2024-germline-panel-guideline", "llach-2024-universal-germline", "hu-2018-germline-risk" ], "links": [ { "target": "disease-system", "relation": "part-of" }, { "target": "claim-universal-germline-testing-needs-action", "relation": "constrains" }, { "target": "unknown-pdac-germline-family-action-route", "relation": "creates" }, { "target": "hypothesis-pdac-germline-family-route-completion", "relation": "tested-by" } ], "evidence": { "humanProspective": "not-applicable", "humanRetrospective": "moderate for missed-carrier comparison", "randomised": "none", "independentReplication": "none for England policy expansion", "sampleSize": "Current national rule; comparison cohorts n=179 and n=3,030", "effect": "R367 limits routine testing by age and history; clinical-criteria series missed 7 of 14 carriers", "limits": [ "Commissioning rule is not an outcome study", "Foreign cohorts may not transfer", "No English all-comer cost and action dataset", "Criteria and panels can change" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Pathogenic variants can occur without a typical family history, so criteria can miss carriers." }, "animalEvidence": { "rating": "not-applicable", "reason": "This is a national eligibility and delivery question." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Variant detection technology does not settle who should be offered testing." }, "reproducibility": { "rating": "none", "reason": "No English prospective all-comer comparison yet tests the policy boundary." }, "commercialConflicts": { "rating": "low", "reason": "The central record is a public commissioning rule and the proposed comparison can use public laboratories." }, "recency": { "rating": "current", "reason": "The NHS rule was checked in version 9.1 dated May 2026." } } }, "updated": "2026-09-14" }, { "id": "technology-ai-multimodal-mrd-state", "title": "AI work package: a calibrated residual-disease state estimate, not a recurrence score", "shortTitle": "AI MRD state", "type": "technology", "status": "hypothesis", "scope": [ "pdac", "colorectal", "transfer-unproven" ], "summary": "Combine serial tumour-informed DNA, CA19-9, imaging and treatment history only if the output controls a defined action and is tested prospectively.", "content": "Input: dates and quantitative values from tumour-informed ctDNA, CA19-9, liver function, cross-sectional imaging, pathology margins, nodal state, treatment dose and missingness. Output: calibrated probability of visible recurrence within 3 and 6 months, plus the likely compartment and an abstain flag. Current workflow: clinicians read each stream separately; the cited 298-patient commercial series establishes prognosis but not action. Current error: sensitivity at low burden and false-negative rate are not known well enough across centres; treatment changes contaminate observed outcomes. Model: a time-to-event model with explicit missing-data and competing-risk terms, compared against simple fixed rules before any larger model. Validation: lock the model in about 800 prospective post-resection patients, require calibration slope 0.8–1.2 and at least a 15-point sensitivity gain at 99% or greater specificity, and then randomize the action. Economic value comes only from avoided late relapse or futile scans; £8m over 32 months for the model gate, followed by about £35m for the action trial. A result that changes treatment is a medical-device claim and requires version control, traceable inputs and prospective utility. Safety risks are false reassurance and earlier toxic treatment. AI is useful for time-series fusion; assay shedding and an effective MRD treatment are the limiting factors.", "contentSections": [ "Work-package audit. Input data: serial ctDNA, CA19-9, imaging, pathology, treatment exposure, liver function and explicit missingness. Target output: calibrated three- and six-month recurrence risk, likely compartment and abstain. Current human workflow: clinicians inspect each stream separately. Current error: low-burden false negatives and cross-centre calibration are not established. Available training data: a 298-patient retrospective PDAC series plus marker-trajectory cohorts. Missing data: uniform sampling, negative assays, changed treatment, outside hospitals and a proven PDAC action. Model architecture: competing-risk time-to-event model before any larger multimodal model. Experimental validation loop: lock in about 800 patients, test calibration and sensitivity at at least 99% specificity, then randomise the action. Regulatory implications: an output that changes treatment is medical-device software. Expected economic value: avoided futile scans or late relapse minus earlier toxic therapy. Limiting factor: AI can fuse time series; assay shedding and an effective action are harder." ], "sourceIds": [ "botta-2024-tumor-informed-ctdna", "fahrmann-2021-ca19-9-trajectory", "tie-2022-dynamic-colon" ], "links": [ { "target": "ai-programme", "relation": "part-of" }, { "target": "hypothesis-multimodal-mrd-action", "relation": "supports" }, { "target": "unknown-ctdna-actionability", "relation": "must-resolve" }, { "target": "programme-25m-prospective-observatory", "relation": "tests" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "strong", "randomised": "none", "independentReplication": "moderate", "sampleSize": "Retrospective PDAC n=298; proposed prospective n≈800", "effect": "Generated work package; target ≥15-point sensitivity gain at ≥99% specificity", "limits": [ "Generated hypothesis", "Low-shedding disease may remain invisible", "Treatment labels are confounded", "No proven PDAC MRD action", "High specificity required", "Medical-device validation needed" ] }, "layout": { "x": 132, "y": 113 }, "updated": "2026-09-14" }, { "id": "technology-ai-organoid-active-learning", "title": "AI work package: active-learning organoid screens that return before cycle two", "shortTitle": "AI organoid loop", "type": "technology", "status": "hypothesis", "scope": [ "pdac", "cholangiocarcinoma", "colorectal", "transfer-unproven" ], "summary": "AI can choose the next drug combinations to test, but culture success, turnaround and patient fidelity—not model size—decide clinical value.", "content": "Input: patient organoid images, growth curves, dose-response surfaces, tumour genotype, transcript state and prior drug exposures. Output: the next small batch of compounds or combinations that maximizes information, followed by a ranked reject/retain list with uncertainty. Current workflow: broad fixed panels are slow and consume scarce cells; one PDAC feasibility series produced profiles for 28 patients, 63.6%, with median turnaround 53 days and range 21–126, often too late. A separate automated screen tested 1,172 compounds across 31 lines but did not show patient benefit. Model: Bayesian optimisation or ensemble active learning, benchmarked against random and expert selection. Falsification: in a locked 150-patient workflow, at least 80% of eligible patients must receive a result before cycle two, negative predictive value for clinical resistance must reach 80% or more, and the AI arm must find the same useful hits with no more than 25% of wells. Gate budget £4m over 24 months; only then a £15m randomized treatment-choice study. Safety risk is choosing a regimen that kills a culture but not the patient's metastatic clone. AI reduces experimental count; tissue acquisition, assay fidelity and treatment access are the limiting factors.", "contentSections": [ "Work-package audit. Input data: organoid images, growth, dose response, tumour state, drug exposure, culture failures and patient outcome. Target output: the next small experiment batch and a reject or retain list with uncertainty. Current human workflow: broad fixed panels consume scarce cells and often finish too late. Current error: only 28 of 44 patients produced profiles and median turnaround was 53 days; patient-treatment prediction is unproved. Available training data: 31 screened lines and small feasibility cohorts. Missing data: failed cultures, metastatic-clone fidelity, immune context and prospective treatment outcomes. Model architecture: Bayesian optimisation or ensemble active learning against random and expert choice. Experimental validation loop: locked 150-patient comparison of wells, time and clinical resistance. Regulatory implications: research selection is low risk until it recommends treatment, when assay and device controls apply. Expected economic value: fewer wells and faster rejection. Limiting factor: AI can reduce search; tissue, fidelity, turnaround and drug access dominate.", "Current evidence changes the AI boundary. ORGANOPREDICT still took 6.8 weeks, PASS-01 generated organoids in 50%, and only 4/81 AVATAR participants assigned to a larger precision route received matched treatment. An algorithm may cut wells, but it cannot create tumour tissue, prove culture identity, guarantee drug access or estimate treatment benefit. Train and test on all failures, not only growing cultures. No model-guided care begins before the 200-person shadow route passes timing, cross-laboratory and treatment-interaction gates." ], "sourceIds": [ "beutel-2021-pdac-organoids", "tiriac-2018-organoid-profiling", "kong-2022-pdac-organoid-drug-screen" ], "links": [ { "target": "claim-pdac-organoid-correlation-not-routing-utility", "relation": "bounded-by" }, { "target": "failure-organoid-treatment-chain-attrition", "relation": "bounded-by" }, { "target": "trial-pdac-organoid-action-map", "relation": "gated-by" }, { "target": "ai-programme", "relation": "part-of" }, { "target": "hypothesis-organoid-guided-early-switch", "relation": "supports" }, { "target": "person-david-tuveson", "relation": "could-be-led-by" }, { "target": "programme-5m-decision-core", "relation": "tests" } ], "evidence": { "humanProspective": "weak", "humanRetrospective": "moderate", "randomised": "none", "independentReplication": "moderate", "sampleSize": "PDAC feasibility n=44 with 28 profiles; organoid screen 31 lines; proposed n=150", "effect": "Generated work package; target same hits with ≤25% of wells and result before cycle two", "limits": [ "Generated hypothesis", "Culture failure selects patients", "Median historical turnaround too slow", "Organoids omit immune and metastatic context", "Drug availability constrains action", "No randomized utility proof" ] }, "layout": { "x": 135, "y": 115 }, "updated": "2026-09-15" }, { "id": "technology-ai-pathology-chemo-selection", "title": "Pathology AI may choose between two PDAC chemotherapy backbones, but needs a randomized decision test", "shortTitle": "Pathology chemo AI", "type": "technology", "status": "open", "scope": [ "pdac", "pan-cancer", "transfer-unproven" ], "summary": "A locked slide-based biomarker found different outcomes by chemotherapy family in 299 external patients; observational allocation leaves the causal treatment-choice claim unproved.", "content": "Input: the routine H&E whole-slide image from a diagnostic PDAC biopsy. Output: a binary preference for a fluoropyrimidine-based or gemcitabine-based first-line regimen. Current workflow: outside germline DNA-repair subgroups, clinicians choose between active regimens without a validated predictive test. Data: 178 multi-institutional development patients and 299 independent validation patients from COMPASS and Know Your Tumor; 97.6% of validation patients received mFOLFIRINOX or gemcitabine/nab-paclitaxel. Result: among 173 F-pref patients, median OS was 14.4 months on F-based versus 11.7 on G-based therapy; among 126 G-pref patients, time to next treatment/death favoured G-based therapy but OS did not differ. The treatment-by-biomarker interaction was significant for OS and time to next treatment/death. Current error rate cannot be stated as classification accuracy because the counterfactual—how the same patient would fare on the other regimen—is never observed. Main missing data are randomized assignment, current NALIRIFOX-era performance, scanner/stain drift, turnaround, failure rate and effect by race/site. Model: a proprietary computational-histology feature system; opacity is a reproducibility and regulatory risk. Validation loop: randomize biomarker-stratified patients between the two backbones, lock the model before enrolment, test interaction and net quality-adjusted survival, then measure whether clinicians can receive and use the result before treatment. Economic value could be high because the slide already exists and the decision affects nearly every advanced patient, but only if the prospective interaction is real. AI is plausibly limiting here because it extracts treatment-linked morphology at scale; the final limit is randomized causal evidence, not compute.", "contentSections": [ "Work-package audit. Input data: routine H&E slides, scanner and stain data, treatment, dose, progression, survival and quality of life. Target output: one of two treatment preferences or abstain. Current human workflow: clinicians choose between active chemotherapy families without a validated general selector. Current error cannot be measured as ordinary accuracy because each person receives only one treatment. Available training data: 178 development and 299 external validation patients. Missing data: random assignment, current-regimen performance, failure rate, turnaround and subgroup drift. Model architecture: locked slide-feature model compared with clinical and pathology baselines. Experimental validation loop: stratify then randomise between backbones and test the treatment interaction. Regulatory implications: a treatment recommendation requires device-grade version control, traceability and prospective utility. Expected economic value: a cheap existing slide could affect nearly every advanced patient if the interaction is causal. Limiting factor: image modelling may extract useful morphology; randomised causal evidence is the final gate." ], "sourceIds": [ "hendifar-2026-ai-chemo-pathology", "conroy-2011-folfirinox-metastatic", "wainberg-2023-napoli3" ], "links": [ { "target": "ai-programme", "relation": "part-of" }, { "target": "company-valar-labs", "relation": "developed-by" }, { "target": "treatment", "relation": "supports" }, { "target": "claim-pdac-molecular-subtypes", "relation": "contrasts-with" }, { "target": "experiments", "relation": "tests" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "moderate", "randomised": "none", "independentReplication": "weak", "sampleSize": "Development n=178; locked validation n=299", "effect": "F-pref median OS 14.4 vs 11.7 months; biomarker-treatment OS interaction P=.015 before weighting and P=.005 after weighting", "limits": [ "Nonrandomized treatment choice", "Residual confounding despite weighting", "Company-developed proprietary model", "Several author financial conflicts", "G-pref OS difference not significant", "Prospective clinical utility not tested" ] }, "layout": { "x": 97, "y": 91 }, "updated": "2026-09-14" }, { "id": "technology-ai-protein-design", "title": "AI work package: design binders only after human target validation", "shortTitle": "AI binder design", "type": "technology", "status": "open", "scope": [ "pdac", "hepatobiliary", "colorectal", "pan-cancer", "transfer-unproven" ], "summary": "Structure and binder models can compress the design cycle; they cannot tell whether a target controls human disease, reaches every lesion or is safe to hit.", "content": "Input: a structurally defined extracellular or intracellular target, required binding site, modality and developability constraints. Output: a diverse set of ranked protein, peptide or small-molecule designs plus predicted complex and affinity. AlphaFold 3 improves interaction prediction; BoltzGen reports wet-lab campaigns across 26 targets and nanomolar binders for several held-out targets, while also showing target-dependent diversity collapse and benchmark leakage risk. Current workflow takes repeated structural, display and medicinal-chemistry cycles. Validation loop: surface plasmon resonance, cell binding, functional perturbation, selectivity across human tissue, pharmacokinetics and patient-derived models before animals or trials. A £3m, 24-month mission unit could compare open models on two human-validated PDAC targets with a blinded experimental team; stop if fewer than two orthogonal assays replicate binding or if healthy-tissue activity exceeds the locked limit. Expected value is faster failure and cheaper tool molecules, not automatic drugs. Regulation attaches to the resulting molecule, manufacture and evidence, not the model alone. AI is limiting only after the biological target and delivery route are real; today those are usually the harder problems.", "contentSections": [ "Work-package audit. Input data: target structure, binding site, modality, healthy-tissue expression, assay and manufacture constraints. Target output: diverse candidate binders or molecules with predicted interaction and uncertainty. Current human workflow: repeated structure, display, medicinal-chemistry and assay cycles. Current error is target-dependent and hidden by benchmark overlap; predicted binding also fails to measure function. Available training data: public structures and interaction data plus reported wet-lab campaigns across 26 targets. Missing data: negative assays, novel folds, PDAC delivery, human selectivity and exposure. Model architecture: interaction and generative models compared with conventional design on held-out targets. Experimental validation loop: blinded binding, cell function, selectivity, pharmacology and patient-derived models. Regulatory implications: the resulting product, manufacture and evidence are regulated, not the design model alone. Expected economic value: shorter design cycles and earlier failure. Limiting factor: AI helps design; human target validity, delivery and safety dominate." ], "sourceIds": [ "abramson-2024-alphafold3", "stark-2025-boltzgen", "ozdemir-2014-fibroblast-depletion" ], "links": [ { "target": "ai-programme", "relation": "part-of" }, { "target": "claim-immune-exclusion-is-multicellular", "relation": "depends-on" }, { "target": "failure-bulk-stroma-removal", "relation": "warned-by" }, { "target": "person-regina-barzilay", "relation": "could-be-led-by" } ], "evidence": { "humanProspective": "none", "humanRetrospective": "none", "randomised": "none", "independentReplication": "weak", "sampleSize": "BoltzGen wet-lab campaigns across 26 targets; no PDAC efficacy study", "effect": "Binder generation demonstrated; disease and clinical value unproved", "limits": [ "Preprint", "No PDAC-specific clinical result", "Binding is not functional effect", "Target validity and delivery dominate", "Training-data leakage can inflate benchmarks", "Off-target and manufacturing risk remain" ] }, "layout": { "x": 141, "y": 116 }, "updated": "2026-09-14" }, { "id": "technology-ai-resistance-routing", "title": "AI work package: route observed RAS resistance into a fixed experiment menu", "shortTitle": "AI resistance routing", "type": "technology", "status": "hypothesis", "scope": [ "pdac", "colorectal", "pan-cancer", "transfer-unproven" ], "summary": "A model should classify a measured escape state and abstain when unsupported; it should not invent a combination from sparse retrospective data.", "content": "Input: serial plasma RAS alleles and copy number, paired tissue where safe, pathology state, drug exposure and outcome under daraxonrasib, G12D inhibitors or degraders. Output: one of a fixed set—on-target RAS escape, bypass signalling, lineage or state shift, pharmacologic failure, mixed or unknown—with calibrated confidence and a proposed confirmatory assay. Current workflow is late, sample-poor and case-by-case. Current error is unknown because resistance specimens are not yet collected uniformly; that is the first problem. Model: interpretable multi-view classifier or Bayesian state model, not a free-form treatment generator. Falsification: 60-patient discovery atlas across at least three drugs and two labs; 70% or more of progressing cases routed within 14 days, inter-lab agreement 80% or more, and prospective route prediction must beat mutation-only rules. Cost £5m over 24 months; a later route-matched randomized study about £25m over 48 months. Economic value is avoiding large all-comer combination trials. Safety risk is false confidence that steers patients away from an effective standard. AI can join mixed measurements; specimens, causal perturbation and available escape drugs remain limiting.", "contentSections": [ "Work-package audit. Input data: serial plasma, copy number, progression tissue, pathology state, drug exposure, pharmacology and outcomes. Target output: one fixed escape class, confidence, abstain and confirmatory assay. Current human workflow: late case-by-case review after progression. Current error is unknown because uniform specimens and causal labels are absent. Available training data: treatment-response cohorts and model studies, not a prospective multi-drug resistance atlas. Missing data: fixed sample times, failed biopsies, mixed clones, negative perturbations and route-matched drug response. Model architecture: interpretable multi-view classifier or Bayesian state model against mutation-only rules. Experimental validation loop: sixty patients, at least three drugs, two laboratories, then route-matched randomisation. Regulatory implications: classification used to choose therapy requires traceable device and drug-trial evidence. Expected economic value: avoid broad combination trials. Limiting factor: AI can join measurements; specimens, causal tests and usable partner drugs dominate." ], "sourceIds": [ "oreilly-2026-daraxonrasib", "park-2026-setidegrasib", "varghese-2025-kras-dosage" ], "links": [ { "target": "ai-programme", "relation": "part-of" }, { "target": "hypothesis-prospective-ras-resistance-routing", "relation": "supports" }, { "target": "person-andrew-aguirre", "relation": "could-be-led-by" }, { "target": "programme-5m-decision-core", "relation": "tests" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "moderate", "randomised": "strong", "independentReplication": "weak", "sampleSize": "Daraxonrasib pivotal n=500; generated resistance atlas n=60", "effect": "Generated work package; target ≥70% routing within 14 days and ≥80% inter-lab agreement", "limits": [ "Generated hypothesis", "Uniform resistance samples do not yet exist", "Labels may be mixed or wrong", "Few route-matched drugs have efficacy proof", "Model must abstain", "Prospective treatment utility unproved" ] }, "layout": { "x": 138, "y": 112 }, "updated": "2026-09-14" }, { "id": "technology-ai-trial-operations", "title": "AI work package: improve trial operations, never replace the missing control", "shortTitle": "AI trial operations", "type": "technology", "status": "supported", "scope": [ "pdac", "pnet", "hepatobiliary", "colorectal", "pan-cancer" ], "summary": "Eligibility extraction, site forecasting and data checks can reduce delay; synthetic controls cannot repair treatment-selection bias when survival is the question.", "content": "Inputs: versioned protocol criteria, structured records, scans, pathology reports, site capacity and recruitment history. Outputs: an auditable eligibility list with quoted evidence, missing-data queries, predicted site shortfall and protocol-deviation alerts. Current workflow is manual and slow; current error should be measured locally rather than borrowed from a vendor. The model should retrieve and classify with a human sign-off, never silently exclude a patient. Validation: shadow mode across three trials, sensitivity at least 98% for eligible patients, zero unreviewed exclusions, at least 20% reduction in coordinator time and fixed subgroup error bounds. Synthetic-control models may help plan sample size or interpret a single-arm safety signal, but they cannot estimate the missing counterfactual reliably when treatment choice, supportive care and calendar time differ. ELI-002 subgroup rescue, Valar's observational chemotherapy choice and repeated stromal failures show why the randomized control remains necessary. A £2m, 18-month operations layer may pay for itself through faster accrual. Privacy and regulation attach to patient selection and exported data. AI is useful for throughput; trial question, endpoint and randomization remain human programme design.", "contentSections": [ "Work-package audit. Input data: versioned eligibility rules, patient records, reports, scans, site capacity and accrual history. Target output: quoted eligibility evidence, missing-data queries, site forecasts and deviation alerts. Current human workflow: coordinators search records and reconcile forms manually. Current error must be measured locally; vendor accuracy cannot be borrowed. Available training data: completed screening logs and protocol records, if institutions can release them. Missing data: undocumented exclusions, outside care, race and language errors, protocol changes and true coordinator time. Model architecture: retrieval and classification with mandatory human sign-off, compared with rules and keyword search. Experimental validation loop: shadow mode across three trials at at least 98% eligibility sensitivity and no unreviewed exclusion. Regulatory implications: privacy, audit and trial-record rules apply; adaptive designs require fixed simulations and reporting. Expected economic value: at least 20% coordinator-time saving. Limiting factor: AI can improve throughput; trial question, intervention, control and sites dominate." ], "sourceIds": [ "elicio-2026-amplify7p-results", "hendifar-2026-ai-chemo-pathology", "picozzi-2026-lapis" ], "links": [ { "target": "ai-programme", "relation": "part-of" }, { "target": "programme-100m-randomized-platform", "relation": "supports" }, { "target": "failure-eli002-phase2-missed-dfs", "relation": "warned-by" }, { "target": "technology-ai-pathology-chemo-selection", "relation": "must-validate" } ], "evidence": { "humanProspective": "weak", "humanRetrospective": "moderate", "randomised": "not-applicable", "independentReplication": "weak", "sampleSize": "Proposed shadow validation across three trials", "effect": "Operational target ≥20% coordinator-time reduction at ≥98% eligibility sensitivity", "limits": [ "Local workflow error not yet measured", "Automation can amplify exclusion bias", "Synthetic controls remain confounded", "Privacy and audit requirements", "Does not improve a weak intervention" ] }, "layout": { "x": 147, "y": 117 }, "updated": "2026-09-14" }, { "id": "technology-ai-virtual-cell-loop", "title": "AI work package: a closed perturbation loop before any 'virtual patient' claim", "shortTitle": "AI perturbation loop", "type": "technology", "status": "open", "scope": [ "pdac", "pan-cancer", "transfer-unproven" ], "summary": "Single-cell models become useful when failed predictions improve the next experimental round; current proof is outside PDAC and far from a patient twin.", "content": "Input: baseline single-cell state, explicit genetic or drug perturbation, dose, time and measured post-perturbation state. Output: a distribution of cell-state changes and the next perturbations chosen for information, not a simulated patient's survival. A 2025 closed-loop preprint combined model predictions with CRISPR and drug experiments in blood and immune cells and reported a threefold positive-predictive-value gain in one T-cell activation task. That demonstrates the loop, not a universal cell. PDAC work should begin with matched tumour, fibroblast and myeloid co-cultures from at least 20 patients, hold out entire patients and laboratories, and require predictions to beat nearest-neighbour and pathway baselines in live experiments. Budget £4m over 30 months; stop if the second loop does not improve prospective perturbation ranking by at least 20%. Economic value is retiring combinations before animal or clinical work. Risks are batch identity, missing spatial history and false confidence from transcript-only endpoints. The limiting factor is standardized perturbation data from human PDAC, not compute.", "contentSections": [ "Work-package audit. Input data: baseline cell state, explicit perturbation, dose, time, measured later state, donor and laboratory. Target output: a distribution of cell changes and the next experiments, not survival. Current human workflow: experts choose perturbations and inspect one round at a time. Current error is not established in PDAC; a threefold positive-predictive-value gain came from one outside-PDAC task. Available training data: blood and immune-cell loops plus public single-cell atlases. Missing data: matched PDAC tumour, fibroblast and myeloid perturbations, spatial history, protein function and failed experiments. Model architecture: perturbation model plus active selection against nearest-neighbour and pathway baselines. Experimental validation loop: hold out patients and laboratories, run two live rounds and require at least 20% better ranking in round two. Regulatory implications: research use remains separate from a patient claim. Expected economic value: fewer failed combinations. Limiting factor: standardized human PDAC perturbation data, not compute." ], "sourceIds": [ "boyd-2025-closed-loop-single-cell", "ozdemir-2014-fibroblast-depletion" ], "links": [ { "target": "ai-programme", "relation": "part-of" }, { "target": "claim-immune-exclusion-is-multicellular", "relation": "depends-on" }, { "target": "person-david-tuveson", "relation": "could-be-led-by" }, { "target": "programme-5m-decision-core", "relation": "tests" } ], "evidence": { "humanProspective": "none", "humanRetrospective": "none", "randomised": "none", "independentReplication": "none", "sampleSize": "Published loop outside PDAC; proposed ≥20 PDAC donors", "effect": "Threefold PPV gain in one T-cell task; no PDAC effect", "limits": [ "Preprint", "No PDAC data", "Cell culture omits whole-body context", "Transcript change may not equal tumour killing", "Batch effects", "Virtual-patient language would overstate evidence" ] }, "layout": { "x": 144, "y": 114 }, "updated": "2026-09-14" }, { "id": "technology-fapi-pet-staging-not-screening", "title": "FAPI PET may improve pancreatic-cancer staging; it is not yet a screening test", "shortTitle": "FAPI PET staging", "type": "technology", "status": "open", "scope": [ "pdac", "pancreatic-all", "cholangiocarcinoma", "gastric" ], "summary": "Small studies show stronger lesion uptake and some upstaging versus FDG PET, but sample sizes are 10–36, inflammation also lights up, and no study shows better outcomes.", "content": "FAPI PET images activated fibroblasts rather than tumour cells directly. This matters in PDAC because its dense fibroblast-rich tissue can be more visible than it is with glucose-based FDG PET. In a retrospective 36-person study, FAPI sensitivity was 100% versus 73.1% for primary tumours, 81.8% versus 59.1% for involved nodes and 91.5% versus 44.0% for metastases. Against FDG it upstaged 6 of 23 patients and changed management for 2; against contrast CT it changed management for only 1 of 23. A prospective 20-person potentially resectable study reported lymph-node sensitivity of 50% versus 0% for FDG. A later ten-person feasibility study upstaged three people. These are signals for better staging, not proof of better survival, screening performance or earlier diagnosis. Fibroblast activation also occurs in pancreatitis and other inflammation, creating a biological false-positive route. The decisive study should enroll consecutive people with apparently localized disease, compare FAPI-guided and standard staging, verify lesions pathologically where possible, and measure futile surgery, treatment changes, recurrence and survival. A separate low-prevalence study would be required before any screening claim: specificity, radiation, tracer capacity and downstream workups cannot be inferred from selected cancer cohorts.", "sourceIds": [ "pang-2022-fapi-pet", "lee-2024-prospective-fapi-resectable-pdac", "evans-2026-fapi-feasibility" ], "links": [ { "target": "detection", "relation": "part-of" }, { "target": "claim-screening-base-rate-maths", "relation": "depends-on" }, { "target": "claim-high-risk-surveillance", "relation": "adjacent-to" }, { "target": "claim-localised-outcomes-differ-by-cancer", "relation": "could-improve-selection-for" }, { "target": "experiments", "relation": "tests" } ], "evidence": { "humanProspective": "weak", "humanRetrospective": "moderate", "randomised": "none", "independentReplication": "weak", "sampleSize": "Retrospective n=36; prospective n=20; feasibility n=10", "effect": "Retrospective primary sensitivity 100% vs 73.1% for FDG; management changed in 1/23 versus contrast CT; later studies showed higher node sensitivity or upstaging", "limits": [ "Very small studies", "Selected known or suspected cancers", "No randomised management trial", "No survival evidence", "No screening cohort", "Inflammatory uptake", "Different tracers and protocols", "Limited pathology confirmation" ] }, "layout": { "x": 15, "y": 116 }, "updated": "2026-09-14" }, { "id": "technology-opportunistic-ct-ai-pancreas", "title": "Pancreas AI can re-read existing CT scans, but prospective patient benefit and true false-positive cost remain unproved", "shortTitle": "Opportunistic CT AI", "type": "technology", "status": "open", "scope": [ "pdac", "pancreatic-all" ], "summary": "PANDA reached 92.9% sensitivity and 99.9% adjusted specificity in 20,530 consecutive clinical scans; other prediagnosis models reached about 69–73% sensitivity, but all evidence remains retrospective.", "content": "Input: an abdominal or chest CT already acquired for another clinical reason. Target output: a calibrated alert for a pancreatic lesion or an image pattern associated with later cancer, plus a location that a radiologist can inspect. Current workflow: many scans do not focus on the pancreas, and subtle duct change, atrophy or texture can be missed. PANDA was trained on 3,208 people and tested across centres. In 20,530 consecutive clinical scans, lesion sensitivity was 92.9% and adjusted specificity 99.9%; it found cancers and other lesions missed by initial care. For PDAC identification in the first real-world cohort, positive predictive value was 56.0% before adjustment and 68.9% after 80 of 156 AI false positives judged to be other diseases needing attention were removed from the error count. That relabelling may be clinically reasonable, but it is not the same as a correct cancer result. PANCANAI found 68.7% of 198 scans taken a median seven months before diagnosis, but its cancer-only cohort could not measure specificity. REDMOD found 73.0% of 63 cancers at a median 475-day lead, while external-control specificity was 81.3% to 87.5%; its test set contained roughly one case per seven people, far above population prevalence. The missing experiment is a locked, silent, prospective run across consecutive health-system scans with every alert adjudicated, followed by a randomised workflow trial. It must measure stage shift, unnecessary MRI or endoscopy, biopsies, time to diagnosis, interval cancers, costs and treatment outcomes. AI is the limiting factor only for image recognition and workload. It cannot create scans for people who never receive one, determine whether a found lesion is lethal, or supply an effective treatment. Its best near-term use is opportunistic re-reading of scans that already exist, where added radiation cost is zero.", "contentSections": [ "Work-package audit. Input data: full CT volumes, scanner settings, scan indication, later diagnoses and benign-pancreas controls. Target output: a calibrated, located alert with an abstain state. Current human workflow: a radiologist reads the scan for its original indication. Current error: retrospective sensitivity ranges from 68.7% to 92.9%, while specificity cannot be transported from enriched sets to routine care. Available training data: PANDA used 3,208 people and three studies supply external tests. Missing data: consecutive prospective alerts, invasive follow-up, interval cancers, subgroup error and mortality. Model architecture: a locked 3D detector plus calibration, compared with radiologist and simple imaging rules. Experimental validation loop: silent multi-hospital use, full adjudication, then randomised workflow testing. Regulatory implications: a patient-level detection alert is medical-device software and needs version, input and performance control. Expected economic value: earlier useful diagnosis minus added MRI, endoscopy, biopsy and staff time. Limiting factor: AI helps recognition; scan availability, lesion danger and effective treatment limit patient benefit." ], "sourceIds": [ "cao-2023-panda-ct-ai", "mukherjee-2026-redmod", "ygberg-2025-pancanai-validation" ], "links": [ { "target": "ai-programme", "relation": "part-of" }, { "target": "detection", "relation": "part-of" }, { "target": "claim-screening-base-rate-maths", "relation": "depends-on" }, { "target": "experiments", "relation": "tests" }, { "target": "unknowns", "relation": "supports" } ], "evidence": { "humanProspective": "none", "humanRetrospective": "strong", "randomised": "none", "independentReplication": "moderate", "sampleSize": "PANDA: 20,530 consecutive clinical scans; PANCANAI: 1,083 cancer patients; REDMOD independent test: 493", "effect": "Retrospective prediagnosis sensitivity about 69% to 73%; PANDA real-world clinical-scan sensitivity 92.9% and adjusted specificity 99.9%", "limits": [ "Retrospective", "No randomised patient-benefit trial", "Prevalence differs from screening", "Some error labels were adjusted after clinical review", "Existing CT availability selects a clinical population", "Radiation prevents using CT AI alone as general-population screening" ] }, "layout": { "x": 15, "y": 111 }, "updated": "2026-09-14" }, { "id": "timeline-technology-history", "title": "Technology history: seven paths from promising idea to current decision", "shortTitle": "Technology history", "type": "system", "status": "supported", "scope": [ "pdac", "pancreatic-all", "colorectal", "pan-cancer", "transfer-unproven" ], "summary": "Seven histories show a repeated pattern: discovery can be fast, but reliable patient benefit requires decades of measurement, failed trials and narrower claims.", "content": "1. Cytotoxic systemic treatment. The mapped modern clinical path begins with FOLFIRINOX in 2011: median survival rose from 6.8 to 11.1 months in selected metastatic PDAC. Mature adjuvant results in 2022 showed a five-year overall-survival gain after resection, but 73.9% still had recurrence or death by five years. NALIRIFOX added another first-line option in 2023. Public and cooperative-group trials established benefit; no financing number can be assigned cleanly to the regimens. Current position: chemotherapy remains the backbone and comparator, not durable systemic control.\n\n2. RAS targeting. A 1988 tumour study found codon-12 c-K-ras mutations in 21 of 22 exocrine pancreatic cancers. More than three decades later, sotorasib produced a narrow single-arm signal in the roughly 1–2% G12C subgroup. In 2026, a 500-patient randomized trial of daraxonrasib doubled median survival in RAS-G12 disease, moving the question from targetability to resistance and earlier use. Revolution Medicines reported $738.9m company-wide research spending in the first half of 2026, including $227.8m identified as third-party daraxonrasib cost; those figures are not the full programme cost. First-line and post-resection phase 3 trials are underway. Current position: clinically validated, not durable, sponsor-dependent and unreplicated independently.\n\n3. Blood markers and early detection. By 1986, CA19-9 was elevated in 78% of 91 pancreatic cancers but also in 22% of 111 benign pancreatic, biliary or liver cases, exposing the sensitivity-specificity problem early. A 2019 evidence review found no basis for general-population pancreatic screening. Later work used longitudinal CA19-9, inherited-risk MRI and endoscopic ultrasound, tumour DNA and multi-cancer methylation. Expert surveillance cohorts reported more early-stage diagnoses; the roughly 142,250-person NHS-Galleri trial missed its main stage III/IV endpoint in 2026. The $7.1bn GRAIL acquisition was ownership consideration rather than research spend, but it shows how far capital ran ahead of utility. Current position: selective high-risk surveillance is defensible; general screening still lacks a low-harm, mortality-proven route.\n\n4. Stroma and drug delivery. A 2012 mouse study linked hyaluronan-rich matrix to pressure, collapsed vessels and weak gemcitabine delivery. The mechanistic result encouraged depletion strategies. In 2020, the 494-patient HALO-301 phase 3 trial selected hyaluronan-high metastatic PDAC but PEGPH20 did not improve survival. FibroGen ended pamrevlumab after two randomized failures in 2024. Current position: the stroma is a set of protective and harmful cell states, not bulk material to remove; local delivery, imaging selection and state-specific intervention require direct human tests.\n\n5. Organoids and precision systems. The 2015 Boj study established durable mouse and human pancreatic duct organoids. Tiriac's 2018 platform linked molecular and drug-response profiles to patient examples; a 2021 prospective study tested feasibility; a 2022 bank screened 1,172 approved compounds across 31 lines. Translation failed at the delivery layer in the 2025 AVATAR trial: only 4 of 81 patients assigned to the complex precision system received matched treatment. Current position: organoids are useful for rejecting combinations and studying resistance; treatment selection must meet culture-yield, turnaround and randomized-utility gates.\n\n6. Immunity and vaccines. Unselected checkpoint treatment produced almost no response in a 2019 PDAC trial. Small individual neoantigen-vaccine studies in 2024–2025 showed feasible manufacture and long-lived T cells, but not randomized treatment benefit. A separate off-the-shelf KRAS vaccine missed its randomized disease-free-survival endpoint in 2026. A phase 3 individual-vaccine combination after resection remains active. Current position: measurable immunity is possible, but the field still lacks proof that it prevents recurrence and must expose manufacturing failures as well as infused-patient results.\n\n7. AI tied to a care decision. The PANDA-CT study in 2023 showed that models could identify pancreatic cancer on non-contrast CT retrospectively. External pathology and CT studies in 2025–2026 narrowed the route to images already collected, fixed treatment choices and held-out hospitals. Prospective Japanese CT and UK breath workflows are still studies, not utility proof. Current position: model size is not the main constraint; representative data, stable inputs, abstention, a safe clinical action and a randomized or prospective patient-benefit test are.\n\nAcross all seven paths, the time from discovery to dependable outcome is dominated by human sample access, comparison groups, manufacturing, safety and follow-up. Capital shortens engineering work and buys statistical power, but it cannot turn a marker into a useful screen or an initial response into cure. The future milestone page therefore releases money only after a reproduced measurement controls a real action.", "sourceIds": [ "almoguera-1988-kras-pdac", "boj-2015-pdac-organoids", "haglund-1986-ca19-9", "conroy-2011-folfirinox-metastatic", "conroy-2022-prodige24-five-year", "wainberg-2023-napoli3", "strickler-2022-sotorasib-pdac", "oreilly-2026-daraxonrasib", "revolution-medicines-2026-q2-10q", "uspstf-2019-pancreatic-screening", "fahrmann-2021-ca19-9-trajectory", "blackford-2024-high-risk-surveillance", "sasieni-2026-nhs-galleri", "ftc-2024-illumina-grail", "provenzano-2012-stromal-delivery-barrier", "hingorani-2020-halo301", "fibrogen-2024-pamrevlumab-winddown", "tiriac-2018-organoid-profiling", "beutel-2021-pdac-organoids", "kong-2022-pdac-organoid-drug-screen", "sarno-2025-avatar", "oreilly-2019-durvalumab-tremelimumab", "lau-2024-dendritic-resected-pdac", "rojass-2025-neoantigen-followup", "elicio-2026-amplify7p-results", "nct05968326-imcode003", "cao-2023-panda-ct-ai", "ygberg-2025-pancanai-validation", "umin-2025-kobe-ai-early-pdac" ], "links": [ { "target": "timeline", "relation": "part-of" }, { "target": "treatment", "relation": "explains" }, { "target": "detection", "relation": "explains" }, { "target": "failures", "relation": "depends-on" }, { "target": "investment", "relation": "depends-on" }, { "target": "ai-programme", "relation": "explains" }, { "target": "recommended-strategy", "relation": "supports" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "strong", "randomised": "strong", "independentReplication": "moderate", "sampleSize": "Seven technology paths from 1986–2026 with 29 primary papers, trials, registries, filings and regulatory sources", "effect": "Historical decision map; numerical effects remain attached to the linked claim and source records", "limits": [ "Earliest mapped milestone is not always the first use of the idea", "Commercial spending is incomplete", "Several current trials have no result", "The seven paths are selected rather than exhaustive" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Each path is anchored to a measured disease mechanism or human care decision." }, "animalEvidence": { "rating": "moderate", "reason": "Animal work explains parts of the stroma, RAS and immune paths but repeatedly overpredicted human effect." }, "inVitroEvidence": { "rating": "moderate", "reason": "Organoid and molecular work supplies mechanisms while clinical claims remain separated." }, "reproducibility": { "rating": "moderate", "reason": "Several treatment effects repeat, while detection, organoid routing and vaccine utility remain unsettled." }, "commercialConflicts": { "rating": "high", "reason": "RAS, detection, vaccine and model programmes include sponsor-owned products and data." }, "recency": { "rating": "mixed", "reason": "The record deliberately joins historical origins to current 2026 trials and failures." } } }, "updated": "2026-09-14" }, { "id": "timeline", "title": "One-, two-, four- and ten-year milestones", "shortTitle": "Timeline", "type": "programme", "status": "hypothesis", "scope": [ "pdac", "pnet", "cholangiocarcinoma", "gallbladder", "hcc", "colorectal", "pan-cancer" ], "summary": "Technology histories explain how the field reached 2026; future dates are then tied to explicit prerequisites and probabilities.", "content": "The linked technology-history page reconstructs seven paths from origin through scientific milestone, investment, clinical trial, failure and current position: cytotoxic treatment, RAS targeting, blood detection, stroma, organoids, immunity and AI. The future schedule begins only after those lessons are applied. By September 2027: finish independent reanalysis of key detection, MRD and RAS datasets; open common specimen and data contracts; lock one MRD state, one RAS-resistance taxonomy and one risk-triggered detection workflow. Planning probability 60–75% if access agreements close. Failure is assay discordance above 20% or fewer than 60% usable serial specimens. By September 2028: reproduce the resistance taxonomy in a second laboratory; complete silent prospective CT-AI and MRD validation; return fast organoid results before cycle two in at least 80% or retire that route. Probability 40–60% that at least two selectors pass. By September 2030: report randomized phase 2 evidence for one resistance-directed combination or MRD action rule; report prospective harms and treatable-stage yield from the enriched detection workflow; require all negative results to be public. Probability 25–45% of one phase-3-ready strategy and 35–55% of only decisive negative knowledge. By September 2036: broad RAS first-line and post-resection trials should have mature outcomes; at least one MRD-action trial and one risk-enriched detection utility trial should report cancer-specific survival or a later-validated surrogate. Probability 30–50% of at least one practice-changing result, 10–20% of a material population mortality reduction in a defined group, and below 10–15% that most newly treated PDAC is durably controlled. Prerequisite chain: reproducible human measurement → timely treatment or detection decision → tolerable intervention → randomized benefit → manufacturable and affordable delivery. If a link fails, money cannot legitimately buy the next date. pNET, biliary, HCC and colorectal milestones remain disease-specific: transfer only the trial or selection method unless target biology is shared.", "sourceIds": [ "oreilly-2026-daraxonrasib", "nct07491445-daraxonrasib-first-line", "nct07252232-daraxonrasib-resected", "botta-2024-tumor-informed-ctdna", "dynamic-pancreas-registry", "umin-2025-kobe-ai-early-pdac", "almoguera-1988-kras-pdac", "boj-2015-pdac-organoids", "haglund-1986-ca19-9" ], "links": [ { "target": "programme", "relation": "part-of" }, { "target": "experiments", "relation": "depends-on" }, { "target": "risks", "relation": "depends-on" }, { "target": "timeline-technology-history", "relation": "contains" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "moderate", "randomised": "moderate", "independentReplication": "moderate", "sampleSize": "Four milestone horizons tied to current trials, registries and proposed programmes", "effect": "Planning schedule and probability ranges; not a forecast model output", "limits": [ "Dates depend on contracts, recruitment and comparator stability", "Probabilities are judgement", "Current trials can fail or report late", "Mortality follow-up cannot be compressed", "Related-cancer schedules differ" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "The prerequisite chain follows human evidence, while its dates and success ranges remain planning judgements." }, "animalEvidence": { "rating": "not-applicable", "reason": "Animal results do not establish whether the proposed programme can meet these dates." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Laboratory results cannot validate the schedule or its outcome probabilities." }, "reproducibility": { "rating": "none", "reason": "The future milestones have not occurred and therefore cannot have been reproduced." }, "commercialConflicts": { "rating": "moderate", "reason": "Some dates depend on sponsor trials, compound access and company-controlled data." }, "recency": { "rating": "current", "reason": "The schedule uses active and recently completed programmes checked in September 2026." } } }, "layout": { "x": 55, "y": 118 }, "updated": "2026-09-14" }, { "id": "transfer-car-t-lineage-access", "title": "CAR-T transfer: manufacture transfers; blood-cancer antigen and access do not", "shortTitle": "CAR-T → solid tumour", "type": "claim", "status": "supported", "scope": [ "pan-cancer", "pdac", "transfer-unproven" ], "summary": "CAR-T proves living drugs can expand and persist, but PDAC lacks a safe uniform lineage antigen and places physical and immune barriers between cells and tumour.", "content": "Blood-cancer CAR-T combined a near-uniform surface lineage antigen, direct circulatory access, massive in-patient expansion and an antigen-loss state clinicians could measure. Those engineering lessons—closed manufacture, release testing, persistence monitoring and toxicity control—transfer. The decisive biology does not. PDAC candidate antigens are patchy or shared with essential tissue; infused cells must leave blood, enter fibrotic lesions, survive nutrient and oxygen stress and resist multiple suppressive cell types. The mesothelin PDAC trial produced no objective responses. A credible transfer therefore needs a logic-gated or multi-receptor cell, human-tissue off-tumour safety, measured lesion entry and a kill switch before efficacy scaling. Copying the blood-cancer product with a new antigen is insufficient.", "sourceIds": [ "maude-2018-tisagenlecleucel", "aznar-2025-mesothelin-car-t-resistance" ], "links": [ { "target": "transfer", "relation": "part-of" }, { "target": "failure-mesothelin-car-t-pdac", "relation": "contradicts" }, { "target": "claim-immune-exclusion-is-multicellular", "relation": "depends-on" }, { "target": "treatment", "relation": "may-transfer-to" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "moderate", "randomised": "none", "independentReplication": "strong", "sampleSize": "Pivotal B-ALL CAR-T n=75 infused; PDAC mesothelin CAR-T n=16", "effect": "Blood-cancer efficacy established; PDAC objective responses absent in cited trial", "limits": [ "Cross-disease transfer", "Blood-cancer pivotal study single-arm", "Different antigens and tissue access", "Cell toxicity and manufacturing burden", "No PDAC efficacy platform" ] }, "layout": { "x": 56, "y": 138 }, "updated": "2026-09-14" }, { "id": "transfer-cml-target-addiction", "title": "CML transfer: one dominant, measurable driver can turn lethal cancer into chronic disease", "shortTitle": "CML → target control", "type": "claim", "status": "supported", "scope": [ "pan-cancer", "pdac", "transfer-unproven" ], "summary": "Imatinib worked because BCR-ABL is an unusually central driver and blood disease is repeatedly measurable; PDAC RAS is central but the rest of that property bundle is missing.", "content": "What made CML work: a near-universal fusion driver, a druggable kinase, wide tumour exposure, a quantitative blood marker and serial switching when molecular response is inadequate. In IRIS, the imatinib arm reached estimated ten-year overall survival 83.3% and complete cytogenetic response 82.8%. What transfers: insist on deep target suppression, longitudinal molecular response and resistance-directed switching rather than one scan-defined response. What does not: PDAC is a spatially segregated solid tumour with stromal delivery limits, branching clones and no equally sensitive universal blood marker. Common-RAS inhibition supplies only the driver-drug component. Technology must recreate the missing parts: whole-body exposure, low-burden measurement and a menu of escape-route treatments. This is the clearest model for durable control, but not evidence that PDAC will become CML-like.", "sourceIds": [ "hochhaus-2017-imatinib-iris", "oreilly-2026-daraxonrasib", "botta-2024-tumor-informed-ctdna" ], "links": [ { "target": "transfer", "relation": "part-of" }, { "target": "claim-daraxonrasib-common-ras-survival-breakthrough", "relation": "may-transfer-to" }, { "target": "hypothesis-prospective-ras-resistance-routing", "relation": "supports" }, { "target": "unknown-ctdna-actionability", "relation": "depends-on" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "strong", "randomised": "strong", "independentReplication": "strong", "sampleSize": "IRIS imatinib randomized arm n=553; PDAC transfer evidence separate", "effect": "CML ten-year OS 83.3%; no corresponding PDAC transfer effect", "limits": [ "Cross-disease analogy", "IRIS crossover limits late randomized comparison", "BCR-ABL and KRAS biology differ", "Solid-tumour access and heterogeneity remain", "PDAC molecular monitoring is less sensitive" ] }, "layout": { "x": 44, "y": 136 }, "updated": "2026-09-14" }, { "id": "transfer-colorectal-mrd-action", "title": "Colorectal MRD transfer: a marker becomes useful only when it safely changes treatment", "shortTitle": "Colon MRD → PDAC", "type": "claim", "status": "supported", "scope": [ "colorectal", "pdac", "transfer-unproven" ], "summary": "Stage II de-escalation succeeded, while DYNAMIC-III and ALTAIR show that risk prediction does not validate treatment escalation; PDAC still lacks a proven positive-result action.", "content": "DYNAMIC stage II is valuable because the test controlled a decision: chemotherapy use fell from 28% to 15%, an absolute reduction of 13 points, while two-year recurrence-free survival remained noninferior. Later randomized results make the boundary clearer. DYNAMIC-III reduced treatment in ctDNA-negative stage III disease but failed its noninferiority test, and escalation in ctDNA-positive disease did not improve recurrence-free survival. ALTAIR gave trifluridine/tipiracil at molecular relapse, missed its disease-free-survival endpoint and caused severe hematologic toxicity in 73.0% versus 3.3%. FIND used a positive result to trigger CT and increased curative-intent treatment, but survival remains immature. What transfers: randomize the exact action rule, count treatment burden and false reassurance, and measure outcome. What does not: stage II colon cancer has a much lower recurrence risk than resected PDAC, and neither a successful de-escalation rule nor an earlier-surgery signal proves that escalation helps PDAC. A stronger PDAC assay can improve prognosis while still worsening patients' lives if it triggers ineffective toxic treatment. The transfer is the trial architecture, not the treatment rule.", "sourceIds": [ "tie-2022-dynamic-colon", "tie-2025-dynamic3", "bando-2026-altair", "mo-2026-find-crc-surveillance", "botta-2024-tumor-informed-ctdna", "dynamic-pancreas-registry" ], "links": [ { "target": "transfer", "relation": "part-of" }, { "target": "claim-colon-ctdna-has-clinical-utility-proof", "relation": "extends" }, { "target": "unknown-ctdna-actionability", "relation": "may-transfer-to" }, { "target": "hypothesis-multimodal-mrd-action", "relation": "supports" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "strong", "randomised": "strong", "independentReplication": "moderate", "sampleSize": "DYNAMIC n=455; PDAC retrospective n=298", "effect": "Colon chemotherapy use -13 points with noninferior two-year recurrence-free survival; PDAC action effect unknown", "limits": [ "De-escalation does not prove escalation", "Different baseline recurrence risk", "PDAC ctDNA shedding lower", "No active PDAC MRD treatment proven", "FIND survival is immature", "Assays differ", "Short follow-up versus a cure horizon" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Residual tumour DNA can mark hidden disease, while benefit depends on whether the selected action can control it." }, "animalEvidence": { "rating": "weak", "reason": "Models can support residual-disease biology but not transfer a human care rule between cancers." }, "inVitroEvidence": { "rating": "weak", "reason": "Assay detection and drug response cannot establish clinical utility." }, "reproducibility": { "rating": "strong", "reason": "Several randomized colorectal trials now demonstrate both successful and failed action rules." }, "commercialConflicts": { "rating": "high", "reason": "Commercial assay involvement is frequent and requires public rule and outcome reporting." }, "recency": { "rating": "current", "reason": "The transfer assessment includes randomized evidence through 2026." } } }, "layout": { "x": 68, "y": 138 }, "updated": "2026-09-14" }, { "id": "transfer-egfr-adjuvant-targeting", "title": "EGFR-lung transfer: move a proven driver drug into residual disease, then demand overall survival", "shortTitle": "EGFR → adjuvant RAS", "type": "claim", "status": "supported", "scope": [ "pan-cancer", "pdac", "transfer-unproven" ], "summary": "ADAURA shows a potent driver inhibitor can prevent lethal recurrence after surgery; PDAC can test the same sequence only after common-RAS activity and MRD state are established.", "content": "ADAURA randomized 682 patients with resected EGFR-mutant lung cancer. Five-year overall survival in stage II-IIIA disease was 85% with osimertinib versus 73% with placebo, HR 0.49. The transferable design is powerful: identify a central driver, prove a selective drug in advanced disease, move it to microscopic residual disease, continue long enough to suppress escape, and follow overall survival rather than stopping at scan-free time. PDAC now has the first ingredient through common-RAS survival benefit, but not the rest. RAS variants, rapid resistance, drug toxicity and uncertain MRD measurement may erase the advantage. The recruiting post-resection daraxonrasib phase 3 is the direct test. A mission programme should surround it with resistance and MRD measurement, not launch a redundant adjuvant trial.", "sourceIds": [ "tsuboi-2023-adaura-os", "oreilly-2026-daraxonrasib", "nct07252232-daraxonrasib-resected" ], "links": [ { "target": "transfer", "relation": "part-of" }, { "target": "trial-daraxonrasib-expansion-programme", "relation": "may-transfer-to" }, { "target": "residual-disease", "relation": "supports" }, { "target": "hypothesis-prospective-ras-resistance-routing", "relation": "depends-on" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "moderate", "randomised": "strong", "independentReplication": "moderate", "sampleSize": "ADAURA n=682; PDAC post-resection trial planned n=500", "effect": "Lung stage II-IIIA five-year OS +12 percentage points, HR 0.49; PDAC transfer pending", "limits": [ "EGFR and RAS resistance differ", "Sponsor-funded trials", "PDAC drug toxicity higher", "PDAC MRD measurement less mature", "Adjuvant PDAC result pending" ] }, "layout": { "x": 60, "y": 137 }, "updated": "2026-09-14" }, { "id": "transfer-her2-selection", "title": "HER2 transfer: a small molecular subgroup can justify a dedicated treatment path", "shortTitle": "HER2 → biliary path", "type": "claim", "status": "supported", "scope": [ "cholangiocarcinoma", "gallbladder", "pdac", "pan-cancer", "transfer-unproven" ], "summary": "Breast cancer proves that biomarker-locked targeted therapy can change long-term survival; HER2-positive biliary cancer already supports the selection principle, while most PDAC does not.", "content": "What made HER2 targeting work: a reproducible assay, a biologically important surface target, large randomized trials and treatment continued in the exact positive subgroup. HERA's 5,099 patients showed one year of trastuzumab reduced death risk, HR 0.74; two years added no DFS benefit and increased cardiac burden. What transfers directly: HER2-positive biliary cancers now have a distinct antibody path, including zanidatamab approval. What transfers as method: lock the assay and subgroup before trial entry, then stop extending duration or combinations once marginal benefit disappears. What does not: HER2 is uncommon in PDAC, and an antibody target must be present, accessible and necessary. The lesson is not 'use HER2 drugs in pancreas'; it is 'build small biomarker-defined routes rather than average unlike tumours together.'", "sourceIds": [ "cameron-2017-hera-trastuzumab", "fda-2024-zanidatamab-approval" ], "links": [ { "target": "transfer", "relation": "part-of" }, { "target": "claim-zanidatamab-her2-biliary-selection", "relation": "succeeds-where" }, { "target": "claim-biliary-cancers-not-one-disease", "relation": "supports" }, { "target": "disease-system", "relation": "supports" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "strong", "randomised": "strong", "independentReplication": "strong", "sampleSize": "HERA n=5,099; biliary approval evidence separate", "effect": "HERA death HR 0.74; transfer is subgroup-specific", "limits": [ "Breast and biliary tumour biology differ", "Biliary approval is not cure", "HER2 subgroup is small", "PDAC relevance is mainly trial method", "Sponsor-funded landmark trial" ] }, "layout": { "x": 48, "y": 137 }, "updated": "2026-09-14" }, { "id": "transfer-hpv-screening-action-loop", "title": "Cervical transfer: screening saves lives when the marker names a long, treatable precursor", "shortTitle": "HPV → action loop", "type": "claim", "status": "supported", "scope": [ "pan-cancer", "pdac", "transfer-unproven" ], "summary": "A single HPV-testing round cut cervical-cancer death by about half because a causal infection and accessible precursor had an effective treatment path; PDAC lacks that full chain.", "content": "In rural India, 131,746 women were cluster-randomized. One HPV-test round reduced advanced cervical cancer, HR 0.47, and cervical-cancer death, HR 0.52. The test worked as a system: a causal marker, long precursor dwell, visible cervix, colposcopy, biopsy and removal. What transfers: screening must be judged by completed work-up and mortality, not sensitivity alone; one risk-triggered round can be valuable when it enters a clear action loop. What does not: PDAC has no single common causal infection, PanIN is microscopic, pancreatic sampling is invasive and many positive molecular signals have no safe lesion-removal route. Risk-enriched imaging can imitate only part of the cervical chain. The missing capability is a marker that names dangerous progression and an intervention safer than waiting or pancreatic surgery.", "sourceIds": [ "sankaranarayanan-2009-hpv-screening-india", "dbouk-2022-caps5", "blackford-2024-high-risk-surveillance" ], "links": [ { "target": "transfer", "relation": "part-of" }, { "target": "hypothesis-one-time-risk-triggered-workup", "relation": "may-transfer-to" }, { "target": "claim-stage-shift-insufficient", "relation": "supports" }, { "target": "claim-high-risk-surveillance", "relation": "connects-to" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "moderate", "randomised": "strong", "independentReplication": "strong", "sampleSize": "Cervical screening n=131,746; PDAC surveillance evidence separate", "effect": "Cervical-cancer death HR 0.52; PDAC transfer effect unknown", "limits": [ "Different organ access", "Different precursor visibility", "No single PDAC causal marker", "PDAC surgery carries high burden", "Risk-surveillance evidence is nonrandomized" ] }, "layout": { "x": 64, "y": 139 }, "updated": "2026-09-14" }, { "id": "transfer-melanoma-immunity", "title": "Melanoma transfer: durable immune memory works only when the tumour is visible and reachable", "shortTitle": "Melanoma → immune state", "type": "claim", "status": "supported", "scope": [ "pan-cancer", "pdac", "transfer-unproven" ], "summary": "Checkpoint blockade can produce decade-long metastatic control in melanoma, but PDAC lacks the antigen, entry and activation conditions that made it possible.", "content": "CheckMate 067 shows the destination: median overall survival 71.9 months with nivolumab plus ipilimumab, and 37% of patients alive at ten years. The transferable component is durable self-renewing immune surveillance after finite treatment. The missing PDAC properties are the point: few mutation-derived targets, weak antigen presentation, poor T-cell entry, suppressive myeloid cells, hostile stroma and exhausted or absent clones. Unselected checkpoint trials in PDAC produced almost no response, so adding a checkpoint to every combination is not a transfer strategy. Technology must first measure and repair the patient's missing immune step—antigen, presentation, entry, expansion or persistence—then use checkpoint release only when the preceding chain exists. The melanoma result is proof that immune cure is possible in a solid tumour, not proof that PDAC is one drug away.", "sourceIds": [ "wolchok-2024-checkmate067-ten-year", "oreilly-2019-durvalumab-tremelimumab", "royal-2010-ipilimumab" ], "links": [ { "target": "transfer", "relation": "part-of" }, { "target": "failure-checkpoint-blockade-unselected-pdac", "relation": "contradicts" }, { "target": "claim-immune-exclusion-is-multicellular", "relation": "supports" }, { "target": "hypothesis-low-burden-ras-immune-extinction", "relation": "may-transfer-to" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "moderate", "randomised": "strong", "independentReplication": "strong", "sampleSize": "CheckMate 067 n=945; PDAC checkpoint trials much smaller", "effect": "Melanoma combination median OS 71.9 months; PDAC unselected response near zero", "limits": [ "Large cross-disease property mismatch", "Checkpoint toxicity", "Sponsor-funded melanoma trial", "No validated PDAC immune selector", "Combination contribution versus nivolumab descriptive" ] }, "layout": { "x": 52, "y": 136 }, "updated": "2026-09-14" }, { "id": "transfer-radioligand-selection", "title": "Neuroendocrine transfer: image the target, measure dosimetry, then deliver the radionuclide", "shortTitle": "PRRT → FAPI", "type": "claim", "status": "supported", "scope": [ "pnet", "pdac", "pan-cancer", "transfer-unproven" ], "summary": "PRRT works in selected neuroendocrine disease because the diagnostic scan and therapeutic target match; FAPI-PDAC must first prove the same tumour-to-normal dose window.", "content": "The transferable property is a matched diagnostic-and-treatment pair. Somatostatin-receptor imaging identifies patients whose tumours can concentrate a radioligand; randomized pNET and neuroendocrine evidence shows disease control, while dosimetry and kidney or marrow limits are measurable. PDAC often shows FAP-related tracer uptake, but inflammation also lights up and bowel, marrow, wound healing and normal stroma may receive dose. What transfers: require target-positive imaging, lesion coverage, patient-specific dosimetry and an early stop if tumour-to-normal exposure is poor. What does not: a staging scan's sensitivity is not therapeutic uptake, and FAP is neither tumour-exclusive nor necessarily on the cancer cell. The first PDAC trial should be a 20–30-person dosimetry and safety experiment, not an efficacy claim.", "sourceIds": [ "baudin-2026-oclurandom", "lee-2024-prospective-fapi-resectable-pdac" ], "links": [ { "target": "transfer", "relation": "part-of" }, { "target": "claim-prrt-pnet-randomized-control", "relation": "supports" }, { "target": "hypothesis-fapi-theranostic-pdac", "relation": "may-transfer-to" }, { "target": "claim-pnet-distinct-from-pdac", "relation": "depends-on" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "moderate", "randomised": "strong", "independentReplication": "moderate", "sampleSize": "OCLURANDOM n=84; prospective FAPI PDAC n=20", "effect": "Randomized pNET disease control established; PDAC therapeutic effect untested", "limits": [ "Different target and cancer", "FAPI inflammation uptake", "Stromal target may harm normal repair", "Diagnostic uptake does not prove therapeutic dose", "Small PDAC imaging studies" ] }, "layout": { "x": 72, "y": 140 }, "updated": "2026-09-14" }, { "id": "transfer", "title": "Transfer from other diseases: copy the working property, not the headline", "shortTitle": "Transfer", "type": "system", "status": "supported", "scope": [ "pan-cancer", "pdac", "pnet", "cholangiocarcinoma", "gallbladder", "hcc", "colorectal", "transfer-unproven" ], "summary": "A result transfers only when the target cancer has the same required mechanism, patient state, delivery route and measurable action; eight oncology examples and five outside-field tools are mapped to the property PDAC still lacks.", "content": "Use this page as a router, not as evidence that cancers are interchangeable. The transferability matrix compares eight oncology successes. The outside-field map compares control theory, programmable cell circuits, microfluidics, material mechanics and nuclear medicine. Every route is reduced to a property bundle: what target is present, which patient state is eligible, how the intervention reaches every relevant site, what measurement selects or stops treatment, and which patient outcome proves benefit. A familiar pathway name or similar diagram is insufficient.\n\nThe closest current transfer is trial design. Stage II colon cancer shows that circulating tumour DNA can control an adjuvant-treatment decision without increasing recurrence, but it does not identify the right pancreatic treatment for a positive result. That action must be tested in PDAC. EGFR-mutant lung cancer shows the sequence driver drug → advanced-disease efficacy → postoperative trial → survival follow-up. PDAC now has common-RAS efficacy and an adjuvant trial, while fast escape and the correct residual-disease measurement remain open. Neuroendocrine radioligand work shows how matched imaging, dose measurement and treatment can form one route; FAP targeting in PDAC still lacks tumour-wide specificity, safe absorbed dose and patient benefit.\n\nSome related-cancer lessons transfer more directly. Hepatitis-B prevention reduces HCC by removing a common cause. Colonoscopy can find and remove colorectal precursors. Those routes work because a large causal exposure or accessible precursor exists. PDAC does not inherit either property. Melanoma immune control does not transfer while pancreatic tumours lack sufficient antigen visibility, T-cell entry and a favourable immune-cell state. CAR-T manufacturing transfers, but safe tumour-specific recognition and access do not.\n\nThe capital rule is strict. Copy a protocol, measurement method, manufacturing route or control design only after naming the missing target-cancer property. Fund the smallest human experiment that measures that property. Do not assign a survival probability from the source disease, and stop if the required property is absent. Follow the linked matrix for the ranked cancer comparisons and the outside-field map for the engineering entry tests.", "sourceIds": [ "tsuboi-2023-adaura-os", "hochhaus-2017-imatinib-iris", "tie-2022-dynamic-colon", "cameron-2017-hera-trastuzumab", "baudin-2026-oclurandom", "wolchok-2024-checkmate067-ten-year", "maude-2018-tisagenlecleucel", "sankaranarayanan-2009-hpv-screening-india", "chang-1997-hbv-vaccine-hcc" ], "links": [ { "target": "transferability-matrix", "relation": "maps" }, { "target": "cross-disciplinary-transfer-map", "relation": "maps" }, { "target": "disease-system", "relation": "depends-on" }, { "target": "experiments", "relation": "tests" }, { "target": "research-frontier-ranking", "relation": "informs" } ], "evidence": { "humanProspective": "strong — several source-disease routes have prospective human results", "humanRetrospective": "moderate — some property mapping uses observational target-disease evidence", "randomised": "strong in source diseases; incomplete in PDAC", "independentReplication": "moderate — varies by transfer route", "sampleSize": "Eight ranked cancer transfers plus five outside-field property tests", "effect": "Navigation and decision rule; it makes no pooled treatment-effect claim across diseases", "limits": [ "The same target name can have a different role in another cancer", "Source-disease effect sizes do not transfer", "Manufacturing or measurement transfer does not establish patient benefit", "Each route still needs its stated PDAC or related-cancer human test" ] }, "layout": { "x": 93, "y": 76 }, "updated": "2026-09-15" }, { "id": "transferability-matrix", "title": "Transferability matrix: copy the enabling property, not the cancer label", "shortTitle": "Transfer matrix", "type": "system", "status": "supported", "scope": [ "pdac", "pnet", "cholangiocarcinoma", "gallbladder", "hcc", "colorectal", "pan-cancer", "transfer-unproven" ], "summary": "Eight oncology successes reduce to reusable property bundles; PDAC possesses only fragments of most bundles, so each transfer needs an experiment that recreates the missing property.", "content": "Matrix, ranked by near-term mission value. 1. EGFR-mutant lung adjuvant treatment → common-RAS after PDAC surgery: driver drug and trial exist; resistance and MRD measurement are missing. 2. CML chronic control → RAS suppression plus molecular switching: central driver partly matches; whole-body access and sensitive blood response do not. 3. Colorectal MRD → PDAC action trial: randomized decision design transfers; the positive-result treatment does not. 4. HER2 breast → HER2 biliary: assay-defined targeted route transfers directly to a small biliary subgroup, and only as method to most PDAC. 5. Neuroendocrine radioligands → FAPI PDAC: matched imaging, dosimetry and selection transfer; target specificity does not. 6. Melanoma checkpoint cure → measured immune repair in PDAC: durable immune memory is the destination; antigen visibility and cell entry are absent. 7. CAR-T blood cancers → engineered solid-tumour cells: manufacture and persistence monitoring transfer; safe antigen and tumour access do not. 8. Cervical HPV screening → risk-triggered PDAC work-up: test-to-action trial design transfers; accessible precursor and causal marker do not. Direct related-cancer successes already inside the atlas—HBV prevention in HCC and colorectal endoscopic prevention—remain stronger than distant analogies. Rule: a transfer enters a programme only after its required property is measured in the target disease. Shared vocabulary or pathway name is not enough.", "sourceIds": [ "tsuboi-2023-adaura-os", "hochhaus-2017-imatinib-iris", "tie-2022-dynamic-colon", "cameron-2017-hera-trastuzumab", "baudin-2026-oclurandom", "wolchok-2024-checkmate067-ten-year", "maude-2018-tisagenlecleucel", "sankaranarayanan-2009-hpv-screening-india" ], "links": [ { "target": "transfer", "relation": "part-of" }, { "target": "transfer-egfr-adjuvant-targeting", "relation": "ranks" }, { "target": "transfer-cml-target-addiction", "relation": "ranks" }, { "target": "transfer-colorectal-mrd-action", "relation": "ranks" }, { "target": "transfer-her2-selection", "relation": "ranks" }, { "target": "transfer-radioligand-selection", "relation": "ranks" }, { "target": "transfer-melanoma-immunity", "relation": "ranks" }, { "target": "transfer-car-t-lineage-access", "relation": "ranks" }, { "target": "transfer-hpv-screening-action-loop", "relation": "ranks" }, { "target": "research-frontier-ranking", "relation": "supports" }, { "target": "cross-disciplinary-transfer-map", "relation": "extended-by" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "strong", "randomised": "strong", "independentReplication": "strong", "sampleSize": "Eight property-level transfers backed by separate clinical programmes", "effect": "Decision matrix; target-disease effect stated separately for each row", "limits": [ "Ranking contains judgment", "Historical successes have sponsor and era effects", "Cross-cancer transfer remains unproved unless stated", "Several target-disease measurements do not yet exist", "Matrix omits lower-value analogies" ] }, "layout": { "x": 58, "y": 146 }, "updated": "2026-09-14" }, { "id": "treatment-long-tail-audit", "title": "Treatment long-tail: delivery and immune engineering must beat their payload and comparator", "shortTitle": "Long-tail treatments", "type": "system", "status": "supported", "scope": [ "pdac", "pnet", "cholangiocarcinoma", "hcc", "colorectal", "pan-cancer", "transfer-unproven" ], "summary": "The less mature modalities divide into small randomized signals, clear human failures, uncontrolled feasibility and model-only ideas; none yet approaches durable PDAC control.", "content": "Surgery, chemotherapy, targeted therapy, immunotherapy and RAS pathway targeting are covered in the main ranking. Surgical technique and neoadjuvant or adjuvant timing matter only when they increase complete resection or systemic control without selecting away the patients who progress. Local treatment has not solved systemic relapse. Stereotactic radiotherapy and irreversible electroporation were compared directly in CROSSFIRE after FOLFIRINOX; the randomized phase 2 trial stopped early for futility. SABER, a separate randomized phase 2 test of stereotactic body radiation after modified FOLFIRINOX, enrolled only 37 of 92 planned patients and stopped for slow accrual without a significant progression-free or overall-survival advantage. Ablation therefore needs a systemic-disease endpoint, not just local control. The remaining modalities are not one frontier. Physical-treatment signals now split. Ultrasound-mediated delivery increased median survival from 6.1 to 9.1 months in 78 Chinese patients, but a second randomized trial reports no overall PFS or OS difference in an accepted abstract. Tumour-treating fields increased survival by 2.0 months in PANOVA-3 and still needs independent survival reproduction. Human failures include broad unselected epigenetic treatment, VEGF-receptor inhibition with axitinib, broad stroma or hyaluronan targeting, napabucasin, PARP overall survival, unselected checkpoint therapy, cancer metabolism drug repurposing and several vaccines. Toxic or biologically incomplete: CCR2 inhibition reduced blood monocytes but added 24% pulmonary toxicity without excess efficacy. Feasible but uncontrolled: dendritic vaccines, TCR and CAR-T cells, NK cells, CLDN18.2 bispecific antibodies, multi-antigen T cells, EUS-delivered CYL-02 gene therapy, CHST15 RNA therapy, KRAS siRNA implants, organoid-guided treatment and local alpha emitters. Model or first-dose stage: antibody-drug conjugates, FAPI radiopharmaceuticals and radioligands, CRISPR-derived cells, in-vivo gene editing, most engineered viruses, most nanoparticles and adaptive combination therapy. Local drug delivery, implantable delivery and ultrasound-mediated delivery are useful only if they raise tumour exposure enough to change a controlled clinical outcome. Liposomes already work for irinotecan in NALIRIFOX and are being tested for HRS-4642; that does not validate nanoparticles as a class. DNA damage response and synthetic lethality work biologically in germline BRCA or PALB2 disease, but POLO did not show an overall-survival benefit and reaches a minority. Macrophage, other myeloid, fibroblast and tumour vasculature programmes must measure the intended human cell state because blanket removal can harm. Microbiome associations lack a reproducible intervention and cannot yet select therapy. RNA therapies, gene editing and CRISPR products face tumour access, safe target and manufacturing constraints before efficacy. Adaptive combination therapy needs at least two active, measurable options; PDAC usually lacks that condition. Ranking rule: randomized survival > randomized durable control > prospective contribution > pharmacology > response > immune or imaging change > animal model. A new platform does not inherit the evidence of its payload. KRAS siRNA also has a product-transfer problem: the first implant missed overall randomized survival, while the new soluble dual-route SIL204 programme plans up to 403 people without reported human tumour knockdown. Its 15-to-21-person first segment should become the target and spatial-delivery gate; no new cohort or capital is needed.", "sourceIds": [ "wang-2023-sonochemotherapy-pdac", "macarulla-2025-panova3", "kindler-2011-axitinib-pdac", "noel-2020-ccr2-pdac", "lau-2024-dendritic-resected-pdac", "buscaill-2015-cyl02-gene-pdac", "nakashima-2022-stnm01-pdac", "hingorani-2020-halo301", "oreilly-2019-durvalumab-tremelimumab", "wainberg-2023-napoli3", "timmer-2024-crossfire", "yoo-2026-saber", "alagesan-2026-sig12d-loder", "ctis-2026-sil204-phase23", "silexion-2026-sil204-site-initiation", "silexion-2026-halfyear-filing", "clinicaltrials-nct04821284-sonoporation", "ius-2026-sonoporation-nct04821284-abstract", "bazan-peregrino-2021-vcn01-intratumour", "garcia-carbonero-2022-vcn01-intravenous", "clinicaltrials-nct05673811-virage", "theriva-2023-virage-protocol", "theriva-2025-virage-topline", "esmo-2025-virage-2216mo", "ctis-2026-virage2", "theriva-2026-q2-10q", "richards-2006-ci994-pdac", "heumann-2022-azacitidine-pdac", "sohal-2020-thu-decitabine-pdac", "baretti-2024-entinostat-nivolumab-pdac", "clinicaltrials-nct03250273-entinostat-nivolumab", "safyan-2026-azacitidine-pembrolizumab-pdac", "clinicaltrials-nct03264404-azacitidine-pembrolizumab", "rodon-2024-amg193-prmt5", "matsumoto-2026-mtap-pancreatic", "pavlick-2026-mtap-genomics", "clinicaltrials-nct05732831-vopimetostat", "tango-2025-vopimetostat-monotherapy", "clinicaltrials-nct06922591-vopimetostat-ras", "tango-2026-vopimetostat-ras-data", "tango-2026-q2-10q", "mahalingam-2026-elraglusib-randomized", "clinicaltrials-nct03678883-elraglusib", "pathak-2026-elraglusib-folfirinox", "clinicaltrials-nct05077800-elraglusib-folfirinox", "actuate-2026-q2-10q", "sec-2026-actuate-submissions", "chiorean-2023-abemaciclib-pdac", "baghdadi-2019-palbociclib-cdkn2a", "hidalgo-2022-palbociclib-nabpaclitaxel", "laquente-2017-chk1-pdac", "huffman-2023-ly2880070-pdac", "cuneo-2019-adavosertib-pdac", "huffman-2024-azenosertib-pdac-design", "clinicaltrials-nct06015659-azenosertib", "kindler-2010-bevacizumab-calgb80303", "rougier-2013-aflibercept-vanilla", "van-cutsem-2016-maestro", "ramaswamy-2025-afpac-losartan", "clinicaltrials-nct03563248-losartan-localized", "geller-2017-bacterial-gemcitabine", "nakaoka-2024-kestose-pdac", "corty-2020-antibiotic-gemcitabine-toxicity", "sharma-2026-fmtpanc", "ctis-panoramix-2024-517766-41", "middleton-2014-telovac", "le-2019-eclipse", "elicio-2026-amplify7p-results", "nct05968326-imcode003", "wainberg-2026-quemliclustat-arc8", "coveler-2024-oleclumab-pdac", "clinicaltrials-nct06608927-prism1", "singhal-2021-sbp101-phase1", "carla-2024-ivospemin-retinal-atrophy", "clinicaltrials-nct05254171-aspire", "ctis-2024-514714-12-00-aspire", "clinicaltrials-nct03257033-tigerpac", "hatoum-2024-tamp-rr1-rr2", "novelli-2026-tigerpac-pk", "renovorx-2023-tigerpac-interim", "renovorx-2025-10k", "renovorx-2026-full-enrollment", "testoni-2021-hybridtherm-rct", "gannon-2026-pancreas-histotripsy-safety", "pusceddu-2026-percutaneous-cryoablation", "zhou-2024-pancreatic-mwa", "ross-2022-panco-p32", "huggett-2014-verteporfin-pdt", "dimatteo-2018-eus-laser-ablation", "clinicaltrials-nct07698613-asp1929-pancreas", "muragaki-2026-k912-sonodynamic", "clinicaltrials-nct05371223-nab-pipac", "padilla-valverde-2024-adjuvant-hipec-rct", "levy-2019-celiac-ganglia-versus-plexus-rct", "xie-2026-cnsi-fe-first-human", "fda-2024-bizengri-multidisciplinary-review", "maltoni-2016-systematic-palliative-qol", "heersche-2026-ugt1a1-guided-dosing", "pishvaian-2020-know-your-tumor", "ngohuang-2023-pancfit-rct" ], "links": [ { "target": "treatment", "relation": "part-of" }, { "target": "treatment-modality-ranking", "relation": "extends" }, { "target": "claim-sonochemotherapy-randomized-signal", "relation": "contains" }, { "target": "failure-ccr2-pulmonary-toxicity", "relation": "contains" }, { "target": "failure-axitinib-vascular-targeting", "relation": "contains" }, { "target": "claim-dendritic-cell-vaccine-uncontrolled", "relation": "contains" }, { "target": "failures", "relation": "depends-on" }, { "target": "failure-sonoporation-overall-replication-subgroup-rescue", "relation": "contains" }, { "target": "claim-sil204-trial-scales-before-human-knockdown", "relation": "contains" }, { "target": "trial-sil204-action-map", "relation": "constrained-by" }, { "target": "claim-vcn01-randomized-signal-not-survival-proof", "relation": "bounded-by" }, { "target": "failure-virage-analysis-and-survivor-selection", "relation": "learns-from" }, { "target": "unknown-vcn01-all-enrolled-survival", "relation": "leaves-open" }, { "target": "trial-vcn01-action-map", "relation": "implements" }, { "target": "hypothesis-vcn01-replication-to-survival-gate", "relation": "tests" }, { "target": "company-theriva-biologics", "relation": "contains" }, { "target": "failure-broad-epigenetic-treatment-pdac", "relation": "learns-from" }, { "target": "claim-mtap-prmt5-selected-human-signal", "relation": "bounded-by" }, { "target": "failure-vopimetostat-ras-contribution-unknown", "relation": "learns-from" }, { "target": "unknown-mtap-prmt5-ras-contribution", "relation": "leaves-open" }, { "target": "trial-mtap-prmt5-action-map", "relation": "implements" }, { "target": "hypothesis-mtap-prmt5-ras-deeper-control", "relation": "tests" }, { "target": "company-tango-therapeutics", "relation": "contains" }, { "target": "claim-elraglusib-randomized-survival-signal", "relation": "bounded-by" }, { "target": "failure-elraglusib-analysis-and-survival-discordance", "relation": "learns-from" }, { "target": "unknown-elraglusib-survival-replication", "relation": "leaves-open" }, { "target": "trial-elraglusib-action-map", "relation": "implements" }, { "target": "hypothesis-elraglusib-survival-confirmation-gate", "relation": "tests" }, { "target": "company-actuate-therapeutics", "relation": "contains" }, { "target": "failure-broad-cell-cycle-checkpoint-pdac", "relation": "learns-from" }, { "target": "unknown-wee1-replication-stress-pdac", "relation": "leaves-open" }, { "target": "trial-cell-cycle-checkpoint-action-map", "relation": "implements" }, { "target": "hypothesis-human-replication-stress-selector", "relation": "tests" }, { "target": "failure-vascular-starvation-pdac", "relation": "contains" }, { "target": "claim-hypoxia-control-without-survival-pdac", "relation": "contains" }, { "target": "failure-losartan-delivery-signal-not-replicated", "relation": "contains" }, { "target": "trial-pdac-vascular-hypoxia-action-map", "relation": "mapped-by" }, { "target": "claim-pdac-microbiome-mechanism-without-human-benefit", "relation": "contains" }, { "target": "failure-microbiome-proxy-without-patient-benefit", "relation": "learns-from" }, { "target": "unknown-microbiome-treatment-contribution-pdac", "relation": "contains" }, { "target": "trial-pdac-microbiome-action-map", "relation": "mapped-by" }, { "target": "claim-pdac-vaccine-immunity-without-proven-benefit", "relation": "contains" }, { "target": "failure-pdac-vaccine-escalation-reversals", "relation": "learns-from" }, { "target": "unknown-pdac-vaccine-regimen-contribution", "relation": "contains" }, { "target": "trial-pdac-vaccine-action-map", "relation": "mapped-by" }, { "target": "claim-cd73-blockade-signal-without-concurrent-control", "relation": "contains" }, { "target": "failure-adenosine-control-and-marker-pdac", "relation": "learns-from" }, { "target": "unknown-quemliclustat-survival-contribution", "relation": "contains" }, { "target": "trial-pdac-adenosine-action-map", "relation": "mapped-by" }, { "target": "claim-ivospemin-signal-without-controlled-benefit", "relation": "contains" }, { "target": "failure-aspire-trial-operational-collapse", "relation": "learns-from" }, { "target": "unknown-aspire-assigned-denominator-and-results", "relation": "contains" }, { "target": "trial-pdac-polyamine-action-map", "relation": "mapped-by" }, { "target": "claim-tamp-blood-exposure-not-tumour-delivery", "relation": "contains" }, { "target": "failure-tigerpac-component-and-interim-inference", "relation": "learns-from" }, { "target": "unknown-tigerpac-package-survival-and-delivery", "relation": "contains" }, { "target": "trial-pdac-intraarterial-delivery-action-map", "relation": "mapped-by" }, { "target": "trial-pdac-eus-rfa-action-map", "relation": "mapped-by" }, { "target": "trial-pdac-histotripsy-action-map", "relation": "mapped-by" }, { "target": "trial-pdac-cryoablation-action-map", "relation": "mapped-by" }, { "target": "trial-pdac-mwa-action-map", "relation": "mapped-by" }, { "target": "trial-pdac-p32-action-map", "relation": "mapped-by" }, { "target": "trial-pdac-pdt-action-map", "relation": "mapped-by" }, { "target": "trial-pdac-laser-ablation-action-map", "relation": "mapped-by" }, { "target": "trial-pdac-asp1929-action-map", "relation": "mapped-by" }, { "target": "trial-pdac-k912-sonodynamic-action-map", "relation": "mapped-by" }, { "target": "trial-pdac-pipac-action-map", "relation": "mapped-by" }, { "target": "trial-pdac-hipec-action-map", "relation": "mapped-by" }, { "target": "trial-pdac-celiac-pain-action-map", "relation": "mapped-by" }, { "target": "trial-pdac-ferroptosis-action-map", "relation": "mapped-by" }, { "target": "trial-pdac-nrg1-action-map", "relation": "mapped-by" }, { "target": "trial-pdac-palliative-care-action-map", "relation": "mapped-by" }, { "target": "trial-pdac-pharmacogenetic-safety-action-map", "relation": "mapped-by" }, { "target": "trial-pdac-kras-wildtype-action-map", "relation": "mapped-by" }, { "target": "trial-pdac-exercise-action-map", "relation": "mapped-by" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "moderate", "randomised": "strong", "independentReplication": "weak — mixed evidence", "sampleSize": "Evidence map across more than 25 treatment and delivery classes", "effect": "Class ranking; individual effects in linked records", "limits": [ "Some model-only modalities lack dedicated records", "Old trials use old drug backbones", "Platforms evolve", "Related cancers have different ranks", "Absence of evidence may reflect immature testing rather than invalid biology" ] }, "layout": { "x": 129, "y": 91 }, "updated": "2026-09-15", "contentSections": [ "Polyamine control has a human response signal, serious organ harm and a broken confirmation route. Ivospemin's phase 1 comparison used history rather than a randomized control. Its ASPIRE phase 3 had the right placebo design, but sponsor cash fell to $142,000, the contractor began termination for unpaid invoices, and the U.S. and European registries now report incompatible states. Rank this as data rescue, not a new drug trial.", "Oncolytic viruses require product-level judgment. Pelareorep remains a randomized failure: immune change did not improve PFS. VCN-01 is materially different because it copies in RB1-broken tumour cells, produces PH20 and was found in five of six post-intravenous tumour biopsies. VIRAGE adds a favourable randomized FAS result, OS HR 0.57 and PFS HR 0.55, but 112 enrolled becomes 96 analysed, public alpha rules conflict, and the strongest repeat-dose subgroup selects survivors. Recover the complete trial and use six-person VIRAGE2 only for repeat-dose pharmacology and safety. No new capital or efficacy cohort is added.", "Epigenetic treatment now splits by addressability. Broad DNA- or histone-control drugs have repeatedly failed to improve survival and two checkpoint combinations produced only three responses each. MTAP deletion is different: it creates a specific PRMT5 dependency in about 28% of PDAC, direct human tumour inhibition has been shown, and early vopimetostat plus RAS-inhibitor response is large. The current 25 responses among 39 evaluable people do not prove added effect because two RAS drugs, several doses and no RAS-alone control are combined. Mature every treated record, then randomize the identical RAS drug with or without vopimetostat. Mission capital is £0 because Tango reported $1.0bn available.", "Elraglusib now adds a genuine randomized survival signal, but not a proved treatment. In all 269 people assigned to the selected weekly schedule or control, median survival was 8.9 versus 7.2 months, HR 0.68. The larger 10.1-versus-7.2-month headline excludes 36 people before treatment, with removal rates of 9.4% versus 20.4%. PFS, response and disease control did not significantly improve, and severe events rose by 10.2 points. Recover the full randomization and analysis history before any blinded phase 3. Current mission capital is £0.", "Cell-cycle treatment splits at the checkpoint. Broad CDK4/6 and CHK1 approaches failed across selected cohorts, randomized comparisons and chemotherapy combinations. The cleanest randomized CDK4/6 study gave disease control of 15.2% and 12.1% versus 36.4% with chemotherapy; randomized CHK1 treatment gave survival of 7.8 versus 8.3 months. WEE1 is a different emergency-pause target, but its current PDAC study reports only the 12-person safety stage. Recover tumour measurements and outcomes for £0; do not start another cohort.", "Tumour vasculature now has a complete human boundary. Four phase 3 VEGF-starvation trials totaling 2,387 people failed survival. MAESTRO's low-oxygen-activated drug improved PFS but missed OS. Losartan vessel opening failed broad advanced disease after its uncontrolled localized signal, while a separate 168-person localized randomized record remains unpublished. These results reject broad starvation and new cohorts; they preserve only measured-flow reuse of the existing localized allocation and fixed-marker recovery from MAESTRO.", "The microbiome now separates five products that cannot borrow evidence: broad antibiotics, stool transfer, named live bacteria, fermentable supplements and antifungals. Human presence and model causality are strong, but the 40-person randomized supplement pilot did not improve response significantly; one live-bacterium trial ended at 13 after sponsor insolvency; and current single arms cannot isolate contribution. Randomized FMTPanc and PANORAMIX already own the next safety and efficacy questions.", "Vaccines separate into shared peptides, whole-cell constructs, bacterial vectors, dendritic cells, fixed KRAS peptides, personalized RNA and prevention. Shared-antigen and whole-cell routes failed later-stage survival tests. ELI-002 missed randomized disease-free survival. Personalized RNA remains open only because IMCODE003 already holds the relevant 260-person comparison. Prevention remains an immune-response study, not a cancer-incidence test. No route can borrow benefit from another product.", "Adenosine suppression separates mechanism, molecule and clinical contribution. Quemliclustat reaches the human CD73-to-NR4A route, but its survival estimate uses an external control and its randomized comparison tests checkpoint addition. Oleclumab supplies a negative direct class test. PRISM-1 now provides the missing placebo-controlled 610-person survival comparison, so no new cohort, marker-selected detour or capital is needed.", "Local artery delivery separates lower drug in blood, greater active drug in tumour cells and better patient outcomes. TIGeR-PaC has measured the first, not the second, and its enrolled phase 3 can estimate the third only for a package that also removes nab-paclitaxel and changes schedule. The final all-randomized result and total harm now outrank another cohort; transfer to another drug or cancer requires direct target-tissue evidence.", "Endoscopic radiofrequency treatment has a randomized negative control: HybridTherm did not improve patient outcomes and stopped short of target accrual. Newer studies can test safer delivery, but cannot inherit efficacy. Recover existing records before another cohort.", "Histotripsy uses focused sound to break tissue without heat. Pancreatic anatomy can block or distort the beam, and no controlled patient-benefit result exists. Treat the first human work as a visibility, coverage and safety test.", "Cryoablation freezes tissue and may reduce pain, but present reports are selected and uncontrolled. A bundled immune study cannot identify the freezing effect. Require a pragmatic pain and useful-life comparison.", "Microwave ablation heats tissue, often inside packages with artery treatment and chemotherapy. Technical heating does not prove safe complete coverage or survival. Existing data recovery outranks another cohort.", "Phosphorus-32 implants emit short-range radiation inside the tumour. Local control is plausible, but variable dose and liquefaction complicate surgery and survival inference. The randomized TRIPP-FFX result is the gate.", "Photodynamic treatment activates a light-sensitive drug. The limiting term is spatial: vessels, ducts, light penetration and drug distribution leave untreated tumour. Measure the full dose-to-viable-tumour map before claiming benefit.", "Laser ablation is technically feasible through an endoscope, but fistula, poor accrual and device continuity dominate the evidence. It remains an existing-data recovery task, not an investable treatment route.", "ASP-1929 joins an EGFR-targeting antibody to a light-activated payload. Common EGFR expression does not prove complete delivery or benefit, and pancreatic human evidence has not started. The first ten evaluable patients must show target, light and killing together.", "K-912 combines a sensitising drug, focused ultrasound and chemotherapy. The first human local signal cannot assign contribution to any component. Two existing trials must supply tissue exposure and all-lesion outcomes.", "PIPAC sprays pressurised chemotherapy into the abdominal cavity. Repeat procedures select patients who have not progressed or become too unwell. Analyse every enrollee, not only repeat recipients.", "HIPEC circulates heated chemotherapy during surgery. A small randomized study reduced local recurrence without extending survival, which defines the boundary between local control and useful life. Existing trials must answer the next question.", "Coeliac and splanchnic nerve ablation aim to interrupt pain signals. Pain results conflict and some direct comparisons raise survival harm. Routine use needs a death-aware net-benefit analysis.", "Ferroptosis kills through iron-driven damage to cell membranes. It can expose a tumour weakness or injure other pancreatic cells and accelerate disease, depending on cell type. Human cell-type coverage is the first gate.", "NRG1 fusion treatment is an approved rare-driver route, not a broad pancreatic cure. About 12 of 30 pancreatic patients responded in the FDA review, while most later stopped for progression. RNA test completion and durable control remain unresolved.", "Early palliative care actively treats symptoms, nutrition, mood, family burden and decisions alongside cancer treatment. Randomized evidence supports a small quality-of-life gain and less late chemotherapy, not longer survival. Measure quality-weighted days alive and at home.", "DPYD and UGT1A1 testing are treatment-control tools, not anticancer treatments. DPYD-guided 5-FU starts have regulator backing and repeated safety evidence; pancreatic UGT1A1 dosing remains regimen-specific and missed one tolerability rule. Repair the existing route rather than fund a drug or assay company.", "The KRAS-wild-type branch contains several rare driver routes, not one medicine. NTRK, MET, ROS1, RET, BRAF and NRG1 responses show why reflex RNA matters; Know Your Tumor shows why the all-referred denominator matters. Broaden the existing NRG1 route and add £0.", "Exercise is a supportive route whose actuator often moves while the cancer-pathway output does not. APACaP and PancFit missed their main between-group results; surgical studies did not repeat fewer major complications. Recover the existing randomized data for a low-function selector before funding more enrollment." ] }, { "id": "treatment-modality-ranking", "title": "Treatment modality map: what is proven, promising, failed or still only an experiment", "shortTitle": "Treatment ranking", "type": "system", "status": "supported", "scope": [ "pdac", "pnet", "cholangiocarcinoma", "gallbladder", "hcc", "colorectal", "pan-cancer", "transfer-unproven" ], "summary": "Common-RAS inhibition leads new PDAC systemic treatment; broad immune, vessel-starvation and vaccine routes failed, while CD40, hypoxia, vessel opening, microbiome, personalized vaccination and CD73 blockade remain bounded by existing controlled records.", "content": "PDAC ranking by current human evidence and mission reach. Tier A, proven survival effect: surgery for anatomically and biologically selected local disease; adjuvant mFOLFIRINOX; first-line FOLFIRINOX, NALIRIFOX or gemcitabine/nab-paclitaxel; second-line common-RAS daraxonrasib for eligible prior-treated disease. These extend life but do not durably control most patients. Tier B, randomized signals or live randomized tests with credible rationale: elraglusib plus gemcitabine/nab-paclitaxel, pending full-data audit and blinded phase 3; VCN-01 plus gemcitabine/nab-paclitaxel, pending complete all-enrolled analysis and blinded confirmation; first-line and post-resection common-RAS treatment; combined broad and G12D-selective RAS blockade, now linked to a 36% human KRAS-amplification escape class but without combination efficacy; G12D degradation; personalised neoantigen vaccine after surgery; biomarker-locked local consolidation; MRD-action trials. Tier C, early selected human signal without controlled contribution proof: MTAP-selected PRMT5 inhibition alone and with RAS blockade; HRD-selected PARP plus checkpoint maintenance, RAF/MEK plus FAK, NK cells, CLDN18.2 bispecifics, organoid-guided switching and pathology AI. Tier D, human failures or no survival gain: unselected checkpoint therapy; immune recruitment measured only by T-cell entry; broad stroma removal or HA/CTGF targeting; unselected vaccines; PARP maintenance OS; pelareorep; broad unselected epigenetic treatment including azacitidine maintenance; broad CDK4/6 and CHK1 checkpoint treatment; metformin; devimistat; hydroxychloroquine and its tested ERK or MEK combinations; napabucasin; routine local intensification. Tier E, model or dosimetry stage: other RAS-resistance combinations not anchored to a repeated human route; ADCs, FAPI radioligands, gene editing, CRISPR-derived cells, FAP-CD40 plus PD1-IL2v, nanoparticles beyond established liposomal chemotherapy, ultrasound delivery and most engineered viruses. Related cancers differ: molecular FGFR2 and HER2 routes matter in biliary cancer; immunochemotherapy changes HCC and biliary survival; pNET has targeted drugs and radioligands; colorectal cancer has prevention, MRD utility and selected liver-directed cure. Decision rule: a modality rises only with target-population evidence, direct human pharmacology, an independent control and durable outcome. Novel delivery cannot rescue a wrong target; T-cell entry cannot substitute for tumour killing; a biological response cannot substitute for survival. Ultrasound-plus-microbubble delivery remains below proven systemic treatment and below current physical-treatment evidence. One small randomized survival result did not reproduce overall in the second trial's accepted abstract. Its next value is complete data release and a planned fixed-backbone interaction, not clinical adoption or a third efficacy study. Dual-route SIL204 remains at human-mechanism stage, below an efficacy tier. The old implant missed overall survival; the new product has a live staged trial but no reported human dose or tumour knockdown. Its rank can rise only after multi-region target reduction, systemic exposure and preserved chemotherapy are shown publicly.", "sourceIds": [ "oreilly-2026-daraxonrasib", "aronchik-2026-daraxonrasib-resistance", "lin-2026-vertical-kras-inhibition", "conroy-2022-prodige24-five-year", "wainberg-2023-napoli3", "nct07409272-setidegrasib-first-line", "nct05968326-imcode003", "ho-2025-cxcr4-pd1-pdac", "bever-2026-vaccine-checkpoint-pdac", "park-2026-polar-pdac", "nguyen-2026-fap-cd40-pd1-il2v", "noonan-2016-pelareorep-pdac", "heumann-2022-azacitidine-pdac", "kordes-2015-metformin-pdac", "karasic-2019-hydroxychloroquine-pdac", "metrangolo-2025-upar-adc", "alistar-2017-devimistat-phase1", "philip-2024-avenger500", "surana-2025-erk-hcq-pdac", "clinicaltrials-nct05221320-ulixertinib-hcq", "manji-2023-mekiauto", "clinicaltrials-nct04214418-mekiauto", "clinicaltrials-nct04892017-inlexisertib", "deciphera-2022-dcc3116-phase1", "clinicaltrials-nct04524702-paricalcitol-hcq", "dimcevski-2016-sonoporation-phase1", "wang-2023-sonochemotherapy-pdac", "clinicaltrials-nct04821284-sonoporation", "ius-2026-sonoporation-nct04821284-abstract", "adler-2026-sonoporation-perfusion", "alagesan-2026-sig12d-loder", "ctis-2026-sil204-phase23", "silexion-2026-sil204-site-initiation", "silexion-2026-halfyear-filing", "bazan-peregrino-2021-vcn01-intratumour", "garcia-carbonero-2022-vcn01-intravenous", "clinicaltrials-nct05673811-virage", "theriva-2023-virage-protocol", "theriva-2025-virage-topline", "esmo-2025-virage-2216mo", "ctis-2026-virage2", "theriva-2026-q2-10q", "richards-2006-ci994-pdac", "sohal-2020-thu-decitabine-pdac", "baretti-2024-entinostat-nivolumab-pdac", "clinicaltrials-nct03250273-entinostat-nivolumab", "safyan-2026-azacitidine-pembrolizumab-pdac", "clinicaltrials-nct03264404-azacitidine-pembrolizumab", "rodon-2024-amg193-prmt5", "matsumoto-2026-mtap-pancreatic", "pavlick-2026-mtap-genomics", "clinicaltrials-nct05732831-vopimetostat", "tango-2025-vopimetostat-monotherapy", "clinicaltrials-nct06922591-vopimetostat-ras", "tango-2026-vopimetostat-ras-data", "tango-2026-q2-10q", "mahalingam-2026-elraglusib-randomized", "clinicaltrials-nct03678883-elraglusib", "pathak-2026-elraglusib-folfirinox", "clinicaltrials-nct05077800-elraglusib-folfirinox", "actuate-2026-q2-10q", "sec-2026-actuate-submissions", "chiorean-2023-abemaciclib-pdac", "baghdadi-2019-palbociclib-cdkn2a", "hidalgo-2022-palbociclib-nabpaclitaxel", "laquente-2017-chk1-pdac", "huffman-2023-ly2880070-pdac", "cuneo-2019-adavosertib-pdac", "huffman-2024-azenosertib-pdac-design", "clinicaltrials-nct06015659-azenosertib", "park-2026-gleam-zolbetuximab", "yu-2025-ibi343-pdac", "clinicaltrials-nct07066098-ibi343-phase3", "clinicaltrials-nct03023722-anetumab", "gou-2026-shr3821", "qi-2024-ct041-pdac", "xu-2026-kras-g12v-tcr", "leidner-2022-kras-tcr", "aznar-2025-mesothelin-car-t-resistance", "clinicaltrials-nct03323944-meso-cart", "clinicaltrials-nct04935359-danis2", "pelletier-2026-nis793-biomarkers", "padron-2022-prince", "clinicaltrials-nct03336216-cabiralizumab", "van-laethem-2025-optimize1-biomarkers", "clinicaltrials-nct04888312-optimize1", "alligator-2026-q2", "kindler-2010-bevacizumab-calgb80303", "rougier-2013-aflibercept-vanilla", "kindler-2011-axitinib-pdac", "van-cutsem-2016-maestro", "ramaswamy-2025-afpac-losartan", "clinicaltrials-nct03563248-losartan-localized", "geller-2017-bacterial-gemcitabine", "nakaoka-2024-kestose-pdac", "corty-2020-antibiotic-gemcitabine-toxicity", "sharma-2026-fmtpanc", "ctis-panoramix-2024-517766-41", "middleton-2014-telovac", "le-2019-eclipse", "elicio-2026-amplify7p-results", "wainberg-2026-quemliclustat-arc8", "coveler-2024-oleclumab-pdac", "clinicaltrials-nct06608927-prism1", "arcus-2026-q2-10q", "singhal-2021-sbp101-phase1", "carla-2024-ivospemin-retinal-atrophy", "clinicaltrials-nct05254171-aspire", "ctis-2024-514714-12-00-aspire", "clinicaltrials-nct03257033-tigerpac", "hatoum-2024-tamp-rr1-rr2", "novelli-2026-tigerpac-pk", "renovorx-2023-tigerpac-interim", "renovorx-2025-10k", "renovorx-2026-full-enrollment" ], "links": [ { "target": "treatment", "relation": "part-of" }, { "target": "claim-daraxonrasib-common-ras-survival-breakthrough", "relation": "ranks" }, { "target": "claim-adjuvant-mfolfirinox-improves-but-recurrence-remains", "relation": "ranks" }, { "target": "claim-setidegrasib-g12d-degradation", "relation": "ranks" }, { "target": "claim-neoantigen-vaccine-frontier", "relation": "ranks" }, { "target": "claim-extend-oligometastatic-local-therapy", "relation": "ranks" }, { "target": "claim-nk-cell-therapy-early", "relation": "ranks" }, { "target": "claim-cldn18-bispecific-early", "relation": "ranks" }, { "target": "claim-adc-pdac-preclinical", "relation": "ranks" }, { "target": "failures", "relation": "depends-on" }, { "target": "failure-immune-recruitment-without-response", "relation": "ranks" }, { "target": "claim-daraxonrasib-kras-amplification-resistance", "relation": "ranks" }, { "target": "claim-hcq-combinations-fail-pdac-not-autophagy", "relation": "ranks" }, { "target": "failure-metabolic-repurposing-pdac", "relation": "ranks" }, { "target": "hypothesis-tumour-verified-selective-autophagy-blockade", "relation": "ranks" }, { "target": "claim-sonochemotherapy-randomized-signal", "relation": "ranks" }, { "target": "failure-sonoporation-overall-replication-subgroup-rescue", "relation": "ranks" }, { "target": "hypothesis-measured-fixed-backbone-sonoporation", "relation": "ranks" }, { "target": "claim-sil204-trial-scales-before-human-knockdown", "relation": "ranks" }, { "target": "failure-sig12d-loder-randomized-os-not-shown", "relation": "ranks" }, { "target": "hypothesis-multicompartment-kras-sirna-delivery", "relation": "ranks" }, { "target": "claim-vcn01-randomized-signal-not-survival-proof", "relation": "bounded-by" }, { "target": "failure-virage-analysis-and-survivor-selection", "relation": "learns-from" }, { "target": "unknown-vcn01-all-enrolled-survival", "relation": "leaves-open" }, { "target": "trial-vcn01-action-map", "relation": "implements" }, { "target": "hypothesis-vcn01-replication-to-survival-gate", "relation": "tests" }, { "target": "company-theriva-biologics", "relation": "contains" }, { "target": "failure-broad-epigenetic-treatment-pdac", "relation": "learns-from" }, { "target": "claim-mtap-prmt5-selected-human-signal", "relation": "bounded-by" }, { "target": "failure-vopimetostat-ras-contribution-unknown", "relation": "learns-from" }, { "target": "unknown-mtap-prmt5-ras-contribution", "relation": "leaves-open" }, { "target": "trial-mtap-prmt5-action-map", "relation": "implements" }, { "target": "hypothesis-mtap-prmt5-ras-deeper-control", "relation": "tests" }, { "target": "company-tango-therapeutics", "relation": "contains" }, { "target": "claim-elraglusib-randomized-survival-signal", "relation": "bounded-by" }, { "target": "failure-elraglusib-analysis-and-survival-discordance", "relation": "learns-from" }, { "target": "unknown-elraglusib-survival-replication", "relation": "leaves-open" }, { "target": "trial-elraglusib-action-map", "relation": "implements" }, { "target": "hypothesis-elraglusib-survival-confirmation-gate", "relation": "tests" }, { "target": "company-actuate-therapeutics", "relation": "contains" }, { "target": "failure-broad-cell-cycle-checkpoint-pdac", "relation": "learns-from" }, { "target": "unknown-wee1-replication-stress-pdac", "relation": "leaves-open" }, { "target": "trial-cell-cycle-checkpoint-action-map", "relation": "implements" }, { "target": "hypothesis-human-replication-stress-selector", "relation": "tests" }, { "target": "claim-cldn182-adc-selected-response", "relation": "contains" }, { "target": "failure-antibody-target-and-payload-transfer-pdac", "relation": "contains" }, { "target": "trial-cldn182-antibody-action-map", "relation": "mapped-by" }, { "target": "claim-satricel-pdac-early-signal", "relation": "contains" }, { "target": "claim-kras-g12v-tcr-early-signal", "relation": "contains" }, { "target": "failure-cell-therapy-route-and-denominator-pdac", "relation": "contains" }, { "target": "trial-pdac-engineered-cell-action-map", "relation": "mapped-by" }, { "target": "claim-nis793-target-engagement-without-benefit", "relation": "contains" }, { "target": "claim-cd40-signal-without-contribution-proof", "relation": "contains" }, { "target": "failure-immune-reprogramming-without-patient-benefit", "relation": "learns-from" }, { "target": "trial-immune-reprogramming-action-map", "relation": "mapped-by" }, { "target": "failure-vascular-starvation-pdac", "relation": "learns-from" }, { "target": "claim-hypoxia-control-without-survival-pdac", "relation": "contains" }, { "target": "failure-losartan-delivery-signal-not-replicated", "relation": "learns-from" }, { "target": "trial-pdac-vascular-hypoxia-action-map", "relation": "mapped-by" }, { "target": "claim-pdac-microbiome-mechanism-without-human-benefit", "relation": "contains" }, { "target": "failure-microbiome-proxy-without-patient-benefit", "relation": "learns-from" }, { "target": "unknown-microbiome-treatment-contribution-pdac", "relation": "contains" }, { "target": "trial-pdac-microbiome-action-map", "relation": "mapped-by" }, { "target": "hypothesis-bacterial-gemcitabine-deactivation-gate", "relation": "tests" }, { "target": "claim-pdac-vaccine-immunity-without-proven-benefit", "relation": "contains" }, { "target": "failure-pdac-vaccine-escalation-reversals", "relation": "learns-from" }, { "target": "unknown-pdac-vaccine-regimen-contribution", "relation": "contains" }, { "target": "trial-pdac-vaccine-action-map", "relation": "mapped-by" }, { "target": "hypothesis-autogene-cevumeran-existing-trial-gate", "relation": "tests" }, { "target": "claim-cd73-blockade-signal-without-concurrent-control", "relation": "contains" }, { "target": "failure-adenosine-control-and-marker-pdac", "relation": "learns-from" }, { "target": "unknown-quemliclustat-survival-contribution", "relation": "contains" }, { "target": "trial-pdac-adenosine-action-map", "relation": "mapped-by" }, { "target": "hypothesis-quemliclustat-existing-phase3-gate", "relation": "tests" }, { "target": "claim-ivospemin-signal-without-controlled-benefit", "relation": "contains" }, { "target": "failure-aspire-trial-operational-collapse", "relation": "learns-from" }, { "target": "unknown-aspire-assigned-denominator-and-results", "relation": "contains" }, { "target": "trial-pdac-polyamine-action-map", "relation": "mapped-by" }, { "target": "hypothesis-aspire-data-rescue-gate", "relation": "tests" }, { "target": "claim-tamp-blood-exposure-not-tumour-delivery", "relation": "contains" }, { "target": "failure-tigerpac-component-and-interim-inference", "relation": "learns-from" }, { "target": "unknown-tigerpac-package-survival-and-delivery", "relation": "contains" }, { "target": "trial-pdac-intraarterial-delivery-action-map", "relation": "mapped-by" }, { "target": "hypothesis-tigerpac-existing-final-gate", "relation": "tests" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "strong", "randomised": "strong", "independentReplication": "moderate", "sampleSize": "Five evidence tiers spanning more than 30 modality classes", "effect": "Evidence ranking; individual clinical effects remain in linked records", "limits": [ "Ranking contains judgment", "Several 2026 trials are immature", "Not every named preclinical modality has its own node yet", "Related cancers cannot share one rank", "Clinical practice varies by jurisdiction" ] }, "layout": { "x": 124, "y": 86 }, "updated": "2026-09-15", "contentSections": [ "Ivospemin remains Tier D for proved benefit and Tier C only as a recoverable controlled question. Its 50-person one-arm study produced 12 responses among 28 evaluable patients but also six serious retinal and six serious liver toxicities. The correct 600-person masked trial lost its contractor after unpaid invoices; Europe now calls it ended while the stale U.S. record says recruiting. Fund no new patient or company. Spend at most £250,000 to test whether the assignment key, raw data and follow-up can be recovered.", "VCN-01 rises above the model-stage virus class because five of six intravenous phase 1 biopsies contained viral genomes and VIRAGE supplies a randomized directionally favourable estimate. It does not enter proved survival treatment: 112 enrolled becomes a 96-person FAS, the public protocol conflicts on alpha and population, no complete report exists and no independent trial reproduces it. Rank it as data recovery plus planned confirmation, with £0 mission capital now.", "Epigenetic treatment now splits by addressability. Broad DNA- or histone-control drugs have repeatedly failed to improve survival and two checkpoint combinations produced only three responses each. MTAP deletion is different: it creates a specific PRMT5 dependency in about 28% of PDAC, direct human tumour inhibition has been shown, and early vopimetostat plus RAS-inhibitor response is large. The current 25 responses among 39 evaluable people do not prove added effect because two RAS drugs, several doses and no RAS-alone control are combined. Mature every treated record, then randomize the identical RAS drug with or without vopimetostat. Mission capital is £0 because Tango reported $1.0bn available.", "Elraglusib now adds a genuine randomized survival signal, but not a proved treatment. In all 269 people assigned to the selected weekly schedule or control, median survival was 8.9 versus 7.2 months, HR 0.68. The larger 10.1-versus-7.2-month headline excludes 36 people before treatment, with removal rates of 9.4% versus 20.4%. PFS, response and disease control did not significantly improve, and severe events rose by 10.2 points. Recover the full randomization and analysis history before any blinded phase 3. Current mission capital is £0.", "Cell-cycle treatment splits at the checkpoint. Broad CDK4/6 and CHK1 approaches failed across selected cohorts, randomized comparisons and chemotherapy combinations. The cleanest randomized CDK4/6 study gave disease control of 15.2% and 12.1% versus 36.4% with chemotherapy; randomized CHK1 treatment gave survival of 7.8 versus 8.3 months. WEE1 is a different emergency-pause target, but its current PDAC study reports only the 12-person safety stage. Recover tumour measurements and outcomes for £0; do not start another cohort.", "IBI343 enters Tier C as one named antibody-drug carrier, not as proof that CLDN18.2 antibodies or drug conjugates work. Ten of 44 high-marker tumours responded versus none of 12 lower-marker tumours, while anaemia, low neutrophils and low white-cell counts show payload cost. GLEAM's OS HR 0.999 and the zero-response anetumab and SHR-3821 groups keep neighbouring products in Tier E. NCT07066098 already randomizes 201 people against placebo plus best supportive care, so the correct portfolio action is £0, protocol access and outcome watch.", "Engineered cells remain Tier C only as named products. Satri-cel CAR-T produced 4/24 responses with no PDAC control. KRAS TCR-T produced one response in two G12D patients and one in five G12V patients; repeat G12V dosing failed. Mesothelin CAR-T remains Tier E after zero responses, 3.9-week PFS and a registry-to-paper denominator of 54 versus nine. Existing studies already cover the next questions, so mission capital is £0 until full route accounting, repeatable durable response and a fixed human mechanism are shown.", "Immune reprogramming now splits three ways. Broad TGF-beta blocking is Tier D after daNIS-2 lost 2.0 median survival months and added 14.6 points of serious harm despite direct tissue change. Cabiralizumab CSF1R treatment is Tier D after a 205-person randomized study found no controlled survival gain. Mitazalimab CD40 treatment remains Tier C: 42.1% response and 14.9-month median survival in 57 evaluable people are worth a controlled test but cannot identify added effect. First recover all 94 records and product rights; any roughly 450-person survival trial needs a solvent partner funding at least half.", "Vascular treatment separates starvation, low-oxygen activation and vessel opening. VEGF starvation is Tier D after four phase 3 tests and 2,387 people produced no survival benefit. Evofosfamide remains a failed-primary-endpoint signal: PFS improved but OS HR 0.84 included no effect. Losartan is Tier D in broad advanced disease; localized use remains Tier C only because NCT03563248 already randomized 168 people and has not posted its contribution result. Spend £0 on new cohorts and recover the existing data.", "Microbiome treatment remains Tier C for mechanism and Tier D for patient benefit. Human tumours contain bacteria and bacterial drug inactivation is measurable, but the published 40-person randomized supplement pilot did not improve response significantly. Small single-arm antibiotic, stool-transfer and live-bacterium trials cannot prove contribution. The next decision belongs to the existing randomized FMTPanc and PANORAMIX programmes; cap support at £1m for fixed tumour assays and independent analysis, with £0 for another cohort.", "PDAC vaccines remain Tier D for proved patient benefit and Tier C for the personalized RNA regimen now under randomized test. Four later-stage programmes totaling 1,812 randomized participants missed survival or disease-free survival. Autogene cevumeran creates durable T cells in some patients, but the eight-responder versus eight-non-responder comparison is not a treatment effect. IMCODE003 already estimates 260 assigned patients and is no longer recruiting. Spend £0 on another cohort and require its all-assigned clinical endpoint, manufacture route and harm record.", "Quemliclustat enters Tier C because human tumour measurements support CD73 pathway change and its 122-person chosen-dose group had 15.7-month survival. It does not enter Tier B: the 9.8-month comparator came from other trials, response and PFS did not improve significantly, and the independent oleclumab programme missed its main randomized endpoint. PRISM-1 already completed enrollment for a masked 610-person survival test. Invest £0 before its expected first-half 2027 result.", "TAMP-delivered gemcitabine remains Tier C pending its enrolled phase 3 result. The first 45-person interim median was 16 versus 10 months, P=0.051, but no hazard ratio or confidence interval was reported and the second interim result is withheld. Blood pharmacology supports lower systemic active-drug exposure, not greater active drug inside tumour cells. Rank the final package outcome separately from the delivery mechanism and any transfer claim." ] }, { "id": "treatment", "title": "Treatment and delivery", "shortTitle": "Treatment", "type": "system", "status": "supported", "scope": [ "pdac", "pnet", "cholangiocarcinoma", "gallbladder", "hcc", "colorectal" ], "summary": "Therapies are ranked by human evidence, durable survival effect, reach, delivery limits, resistance and safety.", "content": "Curative-intent PDAC treatment still means complete surgical removal followed by systemic chemotherapy, usually modified FOLFIRINOX in fit patients. That is a real advance but not reliable cure: selected resected patients had 26.1% five-year disease-free survival. Neoadjuvant treatment can expose rapidly progressive disease and improve selection; it has not removed recurrence. In metastatic PDAC, combination chemotherapy produces median control measured in months. NALIRIFOX reached median overall survival of 11.1 months versus 9.2 with gemcitabine and nab-paclitaxel in NAPOLI-3. The first large targeted survival break is daraxonrasib after prior therapy: 13.2 versus 6.6 months in RAS-G12 disease, with median progression at 7.3 months and substantial toxicity. Paired phase 1/2 plasma now shows a recurrent escape route: treatment-emergent RAS-pathway changes in 26/44 patients, led by mutant KRAS amplification in 16/44. That supports deeper or matched pathway blockade as a test, not as proven care; the analysis is sponsor-led, several routes coexisted and no combination has randomized human benefit. Its first-line, postoperative and G12D-combination phase 3 programme now tests whether earlier timing and combinations deepen the initial result. Mutation-selective G12D inhibitors and degraders have early response signals; HRS-4642's 63.3% response came with chemotherapy, no controlled drug contribution and 90.3% grade 3 or worse treatment-related toxicity. G12C inhibitors prove KRAS targetability but reach few PDAC patients. Model lineage tracing warns that KRAS inhibition can enrich a classical epithelial reservoir that seeds relapse; immune treatment can select a different epigenetic escape state. Neither route has yet selected a human combination. PARP maintenance delays progression in germline-BRCA disease without demonstrated overall-survival gain. POLAR adds a bounded selected signal: pembrolizumab plus olaparib after platinum produced median overall survival of 28 months and estimated three-year survival of 44% in the core HRD cohort, but missed both prespecified phase 2 thresholds and had no randomized comparator. Checkpoint blockade alone is ineffective in unselected PDAC. More elaborate recruitment also fails as a simple rule. CXCR4 plus PD-1 treatment increased CD8 infiltration but produced no response in 21 reported patients; vaccine plus dual-checkpoint treatment generated tumour-specific T cells but only two responses among 57, with suppressive myeloid states and 68% severe related toxicity. Personalised and off-the-shelf KRAS vaccines can generate T cells, but randomized recurrence benefit is not established; ELI-002 missed its phase 2 main endpoint. CAR-T, TCR, NK, bispecific and antibody–drug conjugate work remains early and constrained by antigen coverage, tumour entry and toxicity. Drug delivery varies greatly between human PDAC tumours. Bulk stromal removal, soluble hyaluronidase, CTGF blockade, CCR2 inhibition, metabolic repurposing, pelareorep and several local-intensification strategies failed decisive human tests. VCN-01 is a narrower exception: tumour biopsies show viral replication and a randomized phase 2 points toward longer control and survival, but its 112 enrolled people reduce to a 96-person main analysis and independent confirmation is absent. PEGPH20 is the key warning: response increased, but progression and survival did not. Delivery engineering must measure tumour exposure, avoid removing protective tissue and accompany a drug capable of killing the exposed state. PANOVA-3 supplies a counterexample to blanket rejection of physical treatment: tumour-treating fields added two median survival months, but response and progression endpoints did not improve and skin toxicity was common. Cachexia treatment must be measured alongside tumour treatment: IL-6 blockade preserved muscle but missed its main survival endpoint and caused more severe toxicity. The related cancers show where biological selection can work. pNET has randomized radioligand and targeted control; cholangiocarcinoma has FGFR2, IDH1 and HER2 subsets plus a modest immunochemotherapy tail; HCC combines treatment with liver function and viral prevention; selected liver-only colorectal metastasis can sometimes be removed, ablated or transplanted. These results transfer trial and selection methods, not a single PDAC therapy. The treatment goal is extinction measured after treatment, not maximal initial shrinkage: kill all systemic clones at low burden, detect escape early, switch by a reproduced rule, and keep cumulative harm low enough for repeated intervention. Germline testing must occur early enough to change treatment, but its treatment yield is small and selected: one 992-patient service changed chemotherapy in 28 people, and a 1,214-consultation cohort reported matched treatment in 13 of 26 pathogenic-variant carriers with incurable disease. Those uncontrolled care changes do not establish a survival effect. Count the result-to-decision interval and the exact action; do not count a completed panel as treatment.", "contentSections": [ "Polyamine treatment is unresolved because its decisive experiment became an operating failure. Ivospemin produced 12 responses among 28 evaluable phase 1 patients without a control and six serious retinal toxicities among 50 treated. ASPIRE was designed as a 600-person masked survival comparison, but its contractor began termination for non-payment and its public registries now disagree. Preserve the existing randomization before exposing another patient.", "Hidden-spread staging is part of treatment selection, not a cure method. PREOPANC-2 supports starting a planned open operation with a short camera check because non-therapeutic laparotomy was 4.5% with the check and 17.1% without. The allocation was not randomized, and tumour cells found only in abdominal wash fluid do not yet have a proved treatment rule. The mission therefore measures the resulting treatment and patient outcome, not only the number of positive tests.", "Cholangiocarcinoma treatment must retain anatomical site and driver. BILCAP is compatible with adjuvant capecitabine benefit, but its primary intention-to-treat survival test was not significant; ACTICCA-1 has 789 actual participants and no posted result after its estimated 2025 completion. In advanced disease, durvalumab plus chemotherapy improves the survival tail across mixed biliary sites. FGFR2 inhibitors repeatedly produce responses in altered intrahepatic disease, but first-line pemigatinib improved progression-free survival without longer overall survival. Zanidatamab response differs sharply between HER2 IHC 3+ and 2+ tumours and still lacks randomized survival proof. Postoperative ctDNA identifies recurrence risk but does not yet select a treatment that improves outcome.", "Pancreatic enzyme replacement is symptom and malabsorption care, not a proved survival treatment. Two randomized trials of 67 and 88 patients missed short weight or BMI endpoints and did not establish survival. Later cohorts reported longer survival, including 17.1 versus 12.5 months and an adjusted one-year mortality OR of 0.49, but treatment selection can create those gaps; the chemotherapy-only subgroup in the larger study was not significant. Measure the enzyme-deficient state, dose, meal timing, adherence, symptoms, absorption, function and chemotherapy delivered. A weight-only improvement cannot be counted as durable cancer control.", "Venous-thrombosis prophylaxis is complication control, not tumour control. FRAGEM cut 12-week VTE from 23% to 3.4%; CONKO-004 cut three-month symptomatic VTE from 9.9% to 1.3%, yet its PFS HR was 1.06 and OS HR 1.01. Across five randomized studies and 1,003 patients, the absolute VTE reduction was eight points, about 12 people treated per event prevented. Major bleeding was uncommon but its confidence interval was wide. Offer prophylaxis early to suitable high-risk patients after an individual bleeding, interaction and burden check; record fill, use and stop reasons. Do not count fewer clots as cancer control or assume one agent fits every patient.", "Biliary drainage protects a route to treatment; it does not treat the tumour. Routine four-to-six-week drainage with mainly plastic stents raised serious complications from 39% to 74%. Modern metal-stent drainage was noninferior to early surgery for 120-day serious events in a 2026 trial, 29.0% versus 26.5%, but used a 15-point margin and still caused more pre-surgery serious events. When jaundice must be relieved for cholangitis, symptoms, neoadjuvant therapy or unavoidable delay, use a durable internal route and own it through treatment. During neoadjuvant therapy, metal stents reduce obstruction, cholangitis, reintervention and delays relative to plastic; covered and uncovered designs trade migration against ingrowth. Record treatment dates and serious harm. Do not count insertion, bilirubin fall or patency as cancer control.", "Surgery is now represented as a complete treatment route. England reports 96.4% 90-day survival after Whipple but only 62% chemotherapy or chemoradiotherapy within 14 weeks, and that field does not show course completion. Centre concentration, operation method, complication rescue and systemic-treatment delivery remain separate decisions. The funded experiment starts from every potentially curable scan, records the plan and every loss, and tests named ownership without paying for a surgical platform.", "Treatment timing is not a universal neoadjuvant rule. A nine-trial randomized analysis supports event-free survival, HR .77, but not yet overall survival, HR .85, 95% CI .68-1.05. PREOPANC and CISPD-1 favour treatment first; NORPACT-1 favours upfront surgery and exposes delivery failure. Keep resectable and borderline anatomy, regimen, radiation and completion separate. A021806's 358-person modern mFOLFIRINOX comparison is the next survival gate.", "Older and vulnerable patients require a separate treatment route, not an age cutoff. GIANT found no large survival difference between two reduced-intensity combinations and very short median survival in both. GrantPax showed that assessment labels can be ignored. Randomized mixed-cancer evidence supports a fast team that acts on nutrition, medicines, mobility, symptoms and social needs; it does not prove a PDAC survival gain.", "Initially locally advanced PDAC is a state to revisit, not a lifetime label. After at least four months of effective systemic treatment without distant progression, specialist review can find technically removable disease even when CT still shows vessel contact. PREOPANC-4 demonstrates expert feasibility but also 44% major morbidity and 3% 90-day mortality. The survival attributed to conversion surgery remains confounded by who survives and responds long enough to receive it. Offer reassessment; do not promise benefit from the operation until randomized useful-life evidence exists.", "Organoids are an executable tumour sample, not a complete patient model. They can help reject drugs in research now, but the current clinical route loses too many people and returns too slowly. Conditional accuracy among successful cultures cannot establish treatment benefit. Before guidance is used, require tumour identity, a result before the decision, cross-laboratory regimen-order agreement, available drugs and randomized survival plus quality-adjusted survival from the original enrolled denominator.", "FAP-targeted radiation is now a human-feasible treatment platform, not merely an imaging idea. FAP is carried mainly by tumour-support fibroblasts; the radioactive carrier can irradiate nearby malignant cells. LuMIERE established a 9.25-GBq phase 2 dose after 27 treated people, including nine pancreatic cancers. Four pancreatic cancers were stable and none had a reported partial response. GBq administered, PET brightness and Gy absorbed by a lesion are different measurements. The next decision is therefore data recovery and dose-response, not treatment adoption. Ongoing non-randomized PDAC expansion can supply signal; only a matched control can show added useful life.", "Autophagy is a cell recycling process, not a treatment label. HCQ disrupts the lysosome recycling compartment broadly; it does not prove that a selective autophagy target was blocked inside tumour. Three advanced-PDAC HCQ combinations failed their useful-control tests. The strongest direct comparison found four-month disease control of 5.0% with ERK plus HCQ and 5.3% with ERK alone, with the ERK dose halved in the combination. A newer direct ULK1/2 inhibitor reached 91 participants but its trial ended without posted final results. Stop HCQ combinations, recover the missing data, and permit a replacement only through paired tumour pharmacology and a later randomized contribution test.", "Sonoporation changes the transport channel, not the chemotherapy payload: ultrasound drives injected gas microbubbles and may briefly increase local permeability. A 78-person randomized trial reported median survival of 9.1 versus 6.1 months, but the independent NCT04821284 abstract reports no overall PFS or OS gain. Its 407.5 versus 234-day gemcitabine subgroup lacks the data needed to test whether the backbone truly changes the effect. Do not adopt or repeat the device. Release the full trial and measure tumour drug delivery before any fixed-backbone confirmation.", "SIL204 is a new mutant-KRAS siRNA, not a proven continuation of the failed LODER implant. The authorised plan combines injection into at least three tumour sites with under-skin dosing and may reach 403 participants. Its staged safety and dose design is useful, but plasma exposure cannot prove tumour-cell entry, RISC loading or mutant-RNA cutting. No human SIL204 result is reported. Put paired multi-region tumour knockdown, systemic coverage, chemotherapy preservation and public data before expansion; invest £0 now.", "VCN-01 is the strongest current PDAC engineered-virus route, not proved care. The virus was found in five of six sampled tumours after intravenous dosing, and VIRAGE reported OS 10.8 versus 8.6 months in a 96-person FAS. The registry says 112 enrolled, public alpha rules conflict, and the two-dose subgroup selects people who survived and stayed on treatment. Recover all participants and the final analysis plan before a blinded phase 3; current mission capital is £0.", "Epigenetic treatment now splits by addressability. Broad DNA- or histone-control drugs have repeatedly failed to improve survival and two checkpoint combinations produced only three responses each. MTAP deletion is different: it creates a specific PRMT5 dependency in about 28% of PDAC, direct human tumour inhibition has been shown, and early vopimetostat plus RAS-inhibitor response is large. The current 25 responses among 39 evaluable people do not prove added effect because two RAS drugs, several doses and no RAS-alone control are combined. Mature every treated record, then randomize the identical RAS drug with or without vopimetostat. Mission capital is £0 because Tango reported $1.0bn available.", "Elraglusib now adds a genuine randomized survival signal, but not a proved treatment. In all 269 people assigned to the selected weekly schedule or control, median survival was 8.9 versus 7.2 months, HR 0.68. The larger 10.1-versus-7.2-month headline excludes 36 people before treatment, with removal rates of 9.4% versus 20.4%. PFS, response and disease control did not significantly improve, and severe events rose by 10.2 points. Recover the full randomization and analysis history before any blinded phase 3. Current mission capital is £0.", "Cell-cycle treatment splits at the checkpoint. Broad CDK4/6 and CHK1 approaches failed across selected cohorts, randomized comparisons and chemotherapy combinations. The cleanest randomized CDK4/6 study gave disease control of 15.2% and 12.1% versus 36.4% with chemotherapy; randomized CHK1 treatment gave survival of 7.8 versus 8.3 months. WEE1 is a different emergency-pause target, but its current PDAC study reports only the 12-person safety stage. Recover tumour measurements and outcomes for £0; do not start another cohort.", "Antibody-delivered treatment is a three-part machine: a tumour-surface address, a carrier that reaches the cell and a payload that kills or recruits immunity. Each full product must earn its own human evidence. Zolbetuximab found CLDN18.2 but did not extend survival; mesothelin anetumab and CLDN18.2/4-1BB SHR-3821 produced no confirmed pancreatic responses in their reported evaluable groups. IBI343 is different enough to stay open because 10/44 high-marker tumours responded and 0/12 lower-marker tumours did, but this is Tier C evidence. Its existing randomized phase 3 must prove longer useful life and exact marker calibration before the class or product moves higher.", "Engineered cells split into surface readers and fragment readers. Satri-cel CAR-T reads CLDN18.2 directly and produced 4/24 responses, median PFS 3.3 months and OS 10.0 months without a control. KRAS TCR-T reads a mutant intracellular fragment through a matching HLA display socket: one of two G12D patients and one of five G12V patients had clear regression, while repeat G12V doses produced no response. Mesothelin CAR-T shows the engineering trap: cells reached tissue but did not expand or control disease. Keep satri-cel and KRAS TCR-T in Tier C, begin every denominator at target testing, and use existing cohorts before funding anything new.", "Tumour-immune treatment must traverse target binding, immune-system change, cancer-cell killing, durable control and useful life. NIS793 proves why the middle cannot stand for the end: it reduced TGF-beta, fibroblast and collagen signals, then produced OS 9.2 versus 11.2 months and 14.6 more serious-event percentage points in blinded phase 3. Cabiralizumab macrophage treatment also failed controlled benefit. Mitazalimab CD40 activation retains only a product-specific question after 42.1% response and 14.9-month median survival in 57 uncontrolled people. Recover all 94 registry records, then require identical mFOLFIRINOX and placebo control before any claim or clinical use.", "Blood-flow treatment is not one class. Starving tumour vessels failed four phase 3 survival tests totaling 2,387 people. Activating evofosfamide in low oxygen improved progression but missed survival in 693 people. Opening compressed vessels with losartan looked strong in a 49-person multi-treatment route, then failed an 88-person randomized advanced test. The unresolved localized question belongs to the existing 168-person NCT03563248 allocation. Any future claim must join working perfusion or oxygen, delivered cancer treatment, durable control, survival and harm.", "Microbiome treatment has a credible mechanism but no proved patient benefit. Bacteria in human tumours can inactivate gemcitabine in laboratory systems, while bacterial, fungal and stool-transfer interventions alter mouse tumours. The human randomized record is a 40-person supplement pilot with no significant response difference. Current antibiotic, stool-transfer and live-bacterium trials either lack a component control or remain unfinished. Finish randomized FMTPanc and PANORAMIX, join microbial function to drug exposure and outcome, and prescribe none of these approaches outside trials.", "Cancer vaccination is proven here as immune programming, not as patient benefit. Four randomized PDAC programmes covering 1,812 people missed survival or disease-free survival despite immune signals. Personalized RNA produced long-lived T cells in eight of 16 vaccinated patients, but only the existing 260-person IMCODE003 comparison can estimate benefit for its full vaccine, atezolizumab and chemotherapy package. It cannot isolate the vaccine component. Start no new cohort and use no PDAC vaccine outside a trial.", "CD73 blockade is a live phase 3 question, not a treatment. Quemliclustat changed an adenosine-linked tumour measurement and had 15.7-month survival in a 122-person treated set, but its comparator came from other trials and response and PFS did not improve. A different CD73 product failed its main randomized endpoint. PRISM-1 has already completed enrollment for the direct masked comparison in about 610 people; wait for all-randomized survival and use no CD73 drug outside a trial.", "Intra-arterial gemcitabine remains an unread package test. TIGeR-PaC changes delivery route, removes nab-paclitaxel and changes schedule together after selecting non-progressors through induction. Its first 45-person interim estimate was 16 versus 10 months, P=0.051, without a hazard ratio or confidence interval. A 16-person blood study supports lower systemic active-drug exposure but did not measure active drug in tumour cells. Wait for the enrolled final survival and total-harm record; use no TAMP package outside a trial.", "KRAS-wild-type is a routing flag for about 10.7% of PDAC, not a single treatment group. Reflex RNA can expose NTRK, ALK, ROS1, RET, FGFR2, BRAF or NRG1 fusions missed by a DNA-only route. Human responses exist but are mostly cases; each target keeps its own drug, denominator, durability and resistance result.", "Exercise and rehabilitation can raise steps, strength or short-term function, but randomized pancreatic trials have not shown longer survival or a repeated fall in major complications. Offer safe activity as supportive care; do not delay treatment or book a cancer-control effect. The next evidence step is death-aware participant-level recovery, not another all-patient programme." ], "sourceIds": [ "gudmundsdottir-2023-staging-laparoscopy", "rompen-2026-preopanc2-staging", "nice-2018-ng85-pancreatic", "conroy-2022-prodige24-five-year", "wainberg-2023-napoli3", "oreilly-2026-daraxonrasib", "aronchik-2026-daraxonrasib-resistance", "singhal-2024-classical-kras-resistance", "kim-2024-irf6-immune-plasticity", "koay-2014-human-pdac-drug-transport", "hingorani-2020-halo301", "chen-2025-tocilizumab-cachexia", "cui-2026-hrs4642-pdac", "kindler-2022-polo-overall-survival", "oreilly-2019-durvalumab-tremelimumab", "ho-2025-cxcr4-pd1-pdac", "bever-2026-vaccine-checkpoint-pdac", "park-2026-polar-pdac", "macarulla-2025-panova3", "baudin-2026-oclurandom", "pant-2026-zanidatamab-final", "primrose-2019-bilcap", "clinicaltrials-nct02170090-acticca1", "oh-2022-topaz1", "bekaii-saab-2026-fight302", "yoo-2025-extrahepatic-cca-ctdna", "finn-2020-imbrave150", "adam-2024-transmet", "woo-2016-pert-rct", "saito-2018-pert-rct", "picozzi-2025-pert-advanced-pdac", "dugic-2026-pert-mortality", "maraveyas-2012-fragem", "pelzer-2015-conko004", "vadhan-raj-2020-cassini-pdac", "frere-2020-pdac-thromboprophylaxis-meta", "ash-2023-primary-vte-prophylaxis-guide", "van-der-gaag-2010-pbd-rct", "costamagna-2026-sems-early-surgery-rct", "tamura-2021-fcsems-plastic-rct", "seo-2019-covered-uncovered-sems-rct", "lyu-2023-metal-plastic-nat-meta", "vehvilainen-2024-cholangitis-nat-interruption", "asco-2024-germline-panel-guideline", "chittenden-2021-systematic-germline", "sinan-2025-pdac-germline-implementation", "npaca-2026-state-of-nation", "nhs-england-2024-hpb-pancreatic-service", "de-wilde-2012-centralization", "lemmens-2011-centralization", "latenstein-2021-centralization", "coupland-2016-england-resection-volume", "labori-2016-adjuvant-completion", "henry-2023-complications-adjuvant", "kollbeck-2025-complications-chemo", "van-hilst-2019-leopard2", "korrel-2023-diploma", "bruna-2025-diploma-long-term", "klotz-2024-europa", "liu-2024-robotic-open-pd", "jin-2026-portal", "labori-2024-norpact1", "bai-2025-cispd1", "tan-2026-neoadjuvant-ipd-meta", "ghaneh-2023-espac5", "clinicaltrials-nct04340141-a021806", "dotan-2024-giant", "dotan-2025-giant-ga", "clinicaltrials-nct04233866-giant", "betge-2022-grantpax", "clinicaltrials-nct02143219-pamela70", "mohile-2021-gap70", "li-2021-gain", "ihorst-2026-geriatric-comanagement", "clinicaltrials-nct06040801-frail-gi", "stoop-2026-preopanc4-surgery", "dekker-2026-taps-restaging", "verkolf-2026-taps-resection-prediction", "kunzmann-2021-neolap", "guggenberger-2023-neolap-ct", "seelen-2026-pelican", "baum-2022-fap2286-firsthuman", "mcconathy-2026-lumiere-phase1", "clinicaltrials-nct04939610-lumiere", "baum-2026-3bp3940", "clinicaltrials-nct05432193-frontier", "clinicaltrials-nct07229768-ctr-fapi", "alistar-2017-devimistat-phase1", "philip-2024-avenger500", "surana-2025-erk-hcq-pdac", "clinicaltrials-nct05221320-ulixertinib-hcq", "manji-2023-mekiauto", "clinicaltrials-nct04214418-mekiauto", "clinicaltrials-nct04892017-inlexisertib", "deciphera-2022-dcc3116-phase1", "clinicaltrials-nct04524702-paricalcitol-hcq", "dimcevski-2016-sonoporation-phase1", "wang-2023-sonochemotherapy-pdac", "clinicaltrials-nct04821284-sonoporation", "ius-2026-sonoporation-nct04821284-abstract", "adler-2026-sonoporation-perfusion", "alagesan-2026-sig12d-loder", "ctis-2026-sil204-phase23", "silexion-2026-sil204-site-initiation", "silexion-2026-halfyear-filing", "bazan-peregrino-2021-vcn01-intratumour", "garcia-carbonero-2022-vcn01-intravenous", "clinicaltrials-nct05673811-virage", "theriva-2023-virage-protocol", "theriva-2025-virage-topline", "esmo-2025-virage-2216mo", "ctis-2026-virage2", "theriva-2026-q2-10q", "richards-2006-ci994-pdac", "heumann-2022-azacitidine-pdac", "sohal-2020-thu-decitabine-pdac", "baretti-2024-entinostat-nivolumab-pdac", "clinicaltrials-nct03250273-entinostat-nivolumab", "safyan-2026-azacitidine-pembrolizumab-pdac", "clinicaltrials-nct03264404-azacitidine-pembrolizumab", "rodon-2024-amg193-prmt5", "matsumoto-2026-mtap-pancreatic", "pavlick-2026-mtap-genomics", "clinicaltrials-nct05732831-vopimetostat", "tango-2025-vopimetostat-monotherapy", "clinicaltrials-nct06922591-vopimetostat-ras", "tango-2026-vopimetostat-ras-data", "tango-2026-q2-10q", "mahalingam-2026-elraglusib-randomized", "clinicaltrials-nct03678883-elraglusib", "pathak-2026-elraglusib-folfirinox", "clinicaltrials-nct05077800-elraglusib-folfirinox", "actuate-2026-q2-10q", "sec-2026-actuate-submissions", "chiorean-2023-abemaciclib-pdac", "baghdadi-2019-palbociclib-cdkn2a", "hidalgo-2022-palbociclib-nabpaclitaxel", "laquente-2017-chk1-pdac", "huffman-2023-ly2880070-pdac", "cuneo-2019-adavosertib-pdac", "huffman-2024-azenosertib-pdac-design", "clinicaltrials-nct06015659-azenosertib", "park-2026-gleam-zolbetuximab", "yu-2025-ibi343-pdac", "clinicaltrials-nct07066098-ibi343-phase3", "hao-2024-ibi389-pdac", "xu-2025-ibi389-phase1", "clinicaltrials-nct03023722-anetumab", "gou-2026-shr3821", "abou-alfa-2006-exatecan-pdac", "qi-2024-ct041-pdac", "xu-2026-kras-g12v-tcr", "leidner-2022-kras-tcr", "aznar-2025-mesothelin-car-t-resistance", "clinicaltrials-nct03323944-meso-cart", "clinicaltrials-nct04935359-danis2", "pelletier-2026-nis793-biomarkers", "clinicaltrials-nct03336216-cabiralizumab", "padron-2022-prince", "van-laethem-2025-optimize1-biomarkers", "clinicaltrials-nct04888312-optimize1", "kindler-2010-bevacizumab-calgb80303", "rougier-2013-aflibercept-vanilla", "van-cutsem-2016-maestro", "ramaswamy-2025-afpac-losartan", "clinicaltrials-nct03563248-losartan-localized", "geller-2017-bacterial-gemcitabine", "nakaoka-2024-kestose-pdac", "clinicaltrials-nct04193904-mrx0518-pdac", "sharma-2026-fmtpanc", "ctis-panoramix-2024-517766-41", "middleton-2014-telovac", "le-2019-eclipse", "elicio-2026-amplify7p-results", "nct05968326-imcode003", "wainberg-2026-quemliclustat-arc8", "coveler-2024-oleclumab-pdac", "clinicaltrials-nct06608927-prism1", "singhal-2021-sbp101-phase1", "carla-2024-ivospemin-retinal-atrophy", "clinicaltrials-nct05254171-aspire", "ctis-2024-514714-12-00-aspire", "clinicaltrials-nct03257033-tigerpac", "hatoum-2024-tamp-rr1-rr2", "novelli-2026-tigerpac-pk", "renovorx-2023-tigerpac-interim", "renovorx-2025-10k", "renovorx-2026-full-enrollment", "philip-2022-kras-wildtype-pdac", "neuzillet-2023-apacap-rct" ], "links": [ { "target": "claim-pdac-organoid-correlation-not-routing-utility", "relation": "contains" }, { "target": "trial-pdac-organoid-action-map", "relation": "contains" }, { "target": "claim-staging-laparoscopy-avoids-incision-not-proven-survival", "relation": "contains" }, { "target": "trial-pdac-occult-metastasis-action-map", "relation": "maps" }, { "target": "bottlenecks", "relation": "supports" }, { "target": "failures", "relation": "supports" }, { "target": "residual-disease", "relation": "part-of" }, { "target": "treatment-modality-ranking", "relation": "contains" }, { "target": "treatment-long-tail-audit", "relation": "contains" }, { "target": "claim-pdac-resistance-uses-reversible-cell-states", "relation": "contains" }, { "target": "claim-pdac-delivery-is-variable-not-single-barrier", "relation": "contains" }, { "target": "claim-cachexia-is-parallel-lethal-system", "relation": "contains" }, { "target": "failure-immune-recruitment-without-response", "relation": "contains" }, { "target": "claim-daraxonrasib-kras-amplification-resistance", "relation": "contains" }, { "target": "trial-cca-action-map", "relation": "contains" }, { "target": "claim-cca-ctdna-prognostic-not-action-guiding", "relation": "contains" }, { "target": "claim-pert-survival-observational-randomized-unproved", "relation": "contains" }, { "target": "trial-pert-pdac-action-map", "relation": "contains" }, { "target": "claim-anticoagulation-prevents-pdac-vte-not-survival", "relation": "contains" }, { "target": "trial-pdac-vte-action-map", "relation": "informed-by" }, { "target": "claim-biliary-drainage-is-route-not-cancer-treatment", "relation": "contains" }, { "target": "trial-pdac-biliary-drainage-action-map", "relation": "informed-by" }, { "target": "claim-universal-germline-testing-needs-action", "relation": "contains" }, { "target": "trial-pdac-germline-cascade-action-map", "relation": "informed-by" }, { "target": "claim-pdac-surgery-is-a-complete-treatment-route", "relation": "links-to" }, { "target": "system-england-pancreatic-surgery-network", "relation": "links-to" }, { "target": "failure-pdac-postoperative-treatment-loss", "relation": "links-to" }, { "target": "failure-leopard2-laparoscopic-whipple-safety", "relation": "links-to" }, { "target": "unknown-pdac-surgical-route-completion", "relation": "links-to" }, { "target": "trial-pdac-surgical-route-action-map", "relation": "links-to" }, { "target": "hypothesis-pdac-surgical-route-completion", "relation": "links-to" }, { "target": "claim-neoadjuvant-therapy-improves-selection-not-cure", "relation": "links-to" }, { "target": "failure-norpact1-neoadjuvant-delivery", "relation": "links-to" }, { "target": "unknown-pdac-perioperative-treatment-route", "relation": "links-to" }, { "target": "trial-pdac-perioperative-sequencing-action-map", "relation": "links-to" }, { "target": "claim-age-alone-cannot-route-pdac-treatment", "relation": "contains" }, { "target": "trial-older-pdac-action-map", "relation": "informed-by" }, { "target": "unknown-older-pdac-treatment-route", "relation": "contains" }, { "target": "claim-lapc-response-warrants-surgical-reassessment", "relation": "contains" }, { "target": "trial-lapc-conversion-surgery-action-map", "relation": "informed-by" }, { "target": "unknown-lapc-surgery-causal-benefit", "relation": "contains" }, { "target": "failure-conversion-surgery-selection-bias", "relation": "bounded-by" }, { "target": "claim-fap-radioligand-feasible-pdac-efficacy-unproven", "relation": "contains" }, { "target": "trial-pdac-fap-radioligand-action-map", "relation": "informed-by" }, { "target": "unknown-pdac-fap-radioligand-treatment-effect", "relation": "contains" }, { "target": "failure-fapi-scan-dose-benefit-collapse", "relation": "bounded-by" }, { "target": "claim-hcq-combinations-fail-pdac-not-autophagy", "relation": "contains" }, { "target": "failure-metabolic-repurposing-pdac", "relation": "bounded-by" }, { "target": "unknown-selective-autophagy-pdac-contribution", "relation": "leaves-open" }, { "target": "trial-pdac-autophagy-action-map", "relation": "maps" }, { "target": "hypothesis-tumour-verified-selective-autophagy-blockade", "relation": "tests" }, { "target": "claim-sonochemotherapy-randomized-signal", "relation": "contains" }, { "target": "failure-sonoporation-overall-replication-subgroup-rescue", "relation": "bounded-by" }, { "target": "unknown-pdac-sonoporation-treatment-effect", "relation": "leaves-open" }, { "target": "trial-pdac-sonoporation-action-map", "relation": "maps" }, { "target": "hypothesis-measured-fixed-backbone-sonoporation", "relation": "tests" }, { "target": "claim-sil204-trial-scales-before-human-knockdown", "relation": "contains" }, { "target": "failure-sig12d-loder-randomized-os-not-shown", "relation": "bounded-by" }, { "target": "unknown-sil204-human-target-delivery", "relation": "leaves-open" }, { "target": "trial-sil204-action-map", "relation": "maps" }, { "target": "hypothesis-multicompartment-kras-sirna-delivery", "relation": "tests" }, { "target": "claim-vcn01-randomized-signal-not-survival-proof", "relation": "bounded-by" }, { "target": "failure-virage-analysis-and-survivor-selection", "relation": "learns-from" }, { "target": "unknown-vcn01-all-enrolled-survival", "relation": "leaves-open" }, { "target": "trial-vcn01-action-map", "relation": "implements" }, { "target": "hypothesis-vcn01-replication-to-survival-gate", "relation": "tests" }, { "target": "company-theriva-biologics", "relation": "contains" }, { "target": "failure-broad-epigenetic-treatment-pdac", "relation": "learns-from" }, { "target": "claim-mtap-prmt5-selected-human-signal", "relation": "bounded-by" }, { "target": "failure-vopimetostat-ras-contribution-unknown", "relation": "learns-from" }, { "target": "unknown-mtap-prmt5-ras-contribution", "relation": "leaves-open" }, { "target": "trial-mtap-prmt5-action-map", "relation": "implements" }, { "target": "hypothesis-mtap-prmt5-ras-deeper-control", "relation": "tests" }, { "target": "company-tango-therapeutics", "relation": "contains" }, { "target": "claim-elraglusib-randomized-survival-signal", "relation": "bounded-by" }, { "target": "failure-elraglusib-analysis-and-survival-discordance", "relation": "learns-from" }, { "target": "unknown-elraglusib-survival-replication", "relation": "leaves-open" }, { "target": "trial-elraglusib-action-map", "relation": "implements" }, { "target": "hypothesis-elraglusib-survival-confirmation-gate", "relation": "tests" }, { "target": "company-actuate-therapeutics", "relation": "contains" }, { "target": "failure-broad-cell-cycle-checkpoint-pdac", "relation": "learns-from" }, { "target": "unknown-wee1-replication-stress-pdac", "relation": "leaves-open" }, { "target": "trial-cell-cycle-checkpoint-action-map", "relation": "implements" }, { "target": "hypothesis-human-replication-stress-selector", "relation": "tests" }, { "target": "claim-cldn182-adc-selected-response", "relation": "contains" }, { "target": "failure-antibody-target-and-payload-transfer-pdac", "relation": "contains" }, { "target": "trial-cldn182-antibody-action-map", "relation": "mapped-by" }, { "target": "hypothesis-cldn182-adc-survival-test", "relation": "tests" }, { "target": "claim-satricel-pdac-early-signal", "relation": "contains" }, { "target": "claim-kras-g12v-tcr-early-signal", "relation": "contains" }, { "target": "failure-cell-therapy-route-and-denominator-pdac", "relation": "contains" }, { "target": "trial-pdac-engineered-cell-action-map", "relation": "mapped-by" }, { "target": "hypothesis-kras-tcr-human-persistence-gate", "relation": "tests" }, { "target": "claim-nis793-target-engagement-without-benefit", "relation": "contains" }, { "target": "claim-cd40-signal-without-contribution-proof", "relation": "contains" }, { "target": "failure-immune-reprogramming-without-patient-benefit", "relation": "learns-from" }, { "target": "trial-immune-reprogramming-action-map", "relation": "mapped-by" }, { "target": "hypothesis-angiogenin-tgfb-existing-data-test", "relation": "tests" }, { "target": "hypothesis-mitazalimab-controlled-survival-gate", "relation": "tests" }, { "target": "failure-vascular-starvation-pdac", "relation": "learns-from" }, { "target": "claim-hypoxia-control-without-survival-pdac", "relation": "contains" }, { "target": "failure-losartan-delivery-signal-not-replicated", "relation": "learns-from" }, { "target": "trial-pdac-vascular-hypoxia-action-map", "relation": "mapped-by" }, { "target": "hypothesis-localized-losartan-existing-trial-gate", "relation": "tests" }, { "target": "claim-pdac-microbiome-mechanism-without-human-benefit", "relation": "contains" }, { "target": "failure-microbiome-proxy-without-patient-benefit", "relation": "learns-from" }, { "target": "unknown-microbiome-treatment-contribution-pdac", "relation": "contains" }, { "target": "trial-pdac-microbiome-action-map", "relation": "mapped-by" }, { "target": "hypothesis-bacterial-gemcitabine-deactivation-gate", "relation": "tests" }, { "target": "claim-pdac-vaccine-immunity-without-proven-benefit", "relation": "contains" }, { "target": "failure-pdac-vaccine-escalation-reversals", "relation": "learns-from" }, { "target": "unknown-pdac-vaccine-regimen-contribution", "relation": "contains" }, { "target": "trial-pdac-vaccine-action-map", "relation": "mapped-by" }, { "target": "hypothesis-autogene-cevumeran-existing-trial-gate", "relation": "tests" }, { "target": "claim-cd73-blockade-signal-without-concurrent-control", "relation": "contains" }, { "target": "failure-adenosine-control-and-marker-pdac", "relation": "learns-from" }, { "target": "unknown-quemliclustat-survival-contribution", "relation": "contains" }, { "target": "trial-pdac-adenosine-action-map", "relation": "mapped-by" }, { "target": "hypothesis-quemliclustat-existing-phase3-gate", "relation": "tests" }, { "target": "claim-ivospemin-signal-without-controlled-benefit", "relation": "contains" }, { "target": "failure-aspire-trial-operational-collapse", "relation": "learns-from" }, { "target": "unknown-aspire-assigned-denominator-and-results", "relation": "contains" }, { "target": "trial-pdac-polyamine-action-map", "relation": "mapped-by" }, { "target": "hypothesis-aspire-data-rescue-gate", "relation": "tests" }, { "target": "claim-tamp-blood-exposure-not-tumour-delivery", "relation": "contains" }, { "target": "failure-tigerpac-component-and-interim-inference", "relation": "learns-from" }, { "target": "unknown-tigerpac-package-survival-and-delivery", "relation": "contains" }, { "target": "trial-pdac-intraarterial-delivery-action-map", "relation": "mapped-by" }, { "target": "hypothesis-tigerpac-existing-final-gate", "relation": "tests" }, { "target": "trial-pdac-kras-wildtype-action-map", "relation": "mapped-by" }, { "target": "trial-pdac-exercise-action-map", "relation": "mapped-by" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "strong", "randomised": "strong", "independentReplication": "strong", "sampleSize": "Synthesis across phase 3 treatment trials, selected surgery cohorts and early modality studies", "effect": "Current PDAC treatment extends survival but does not reliably eliminate systemic disease; daraxonrasib provides the largest recent randomized targeted survival effect", "limits": [ "Cross-trial medians are not direct comparisons", "Related-cancer efficacy does not establish PDAC efficacy", "Several 2026 signals are sponsor-controlled or immature", "Durable survival and treatment-free time remain uncommon", "Toxicity limits multi-agent combinations" ] }, "layout": { "x": 69, "y": 78 }, "updated": "2026-09-15" }, { "id": "trial-2026-randomised-result-stress-test", "title": "Four 2026 randomized results sharpen selection, recruitment and endpoint rules", "shortTitle": "2026 trial stress test", "type": "trial", "status": "mixed", "scope": [ "hcc", "cholangiocarcinoma", "hepatobiliary", "treatment" ], "summary": "Four new randomized reports make HCC local treatment state-specific, expose rare-subgroup recruitment failure and keep progression-free time separate from survival.", "content": "TALENTOP supplies a response-gated local-treatment signal. It started with 489 people who had hepatocellular carcinoma invading major vessels but no spread outside the liver. Only 201 whose cancer responded or remained stable after atezolizumab plus bevacizumab and whose tumour was considered removable entered randomization. Surgery followed by the same medicines increased median time to treatment failure from 11.8 to 20.4 months: an 8.6-month difference and hazard ratio 0.60. The price was more grade 3-4 treatment-related harm, 39% versus 21%, and two treatment-related deaths in the surgery arm. This validates a response-and-anatomy gate; it does not yet prove longer overall survival or transfer the result to pancreatic cancer.\n\nTORCH supplies a survival result in a different HCC state. It randomized 241 people with liver-confined, unresectable BCLC stage B disease at two Chinese tertiary centres. Adding selective radiofrequency ablation after arterial chemoembolisation increased median progression-free survival from 7.3 to 17.7 months, HR 0.47, and median overall survival from 35.1 to 88.6 months, HR 0.50. Severe treatment-related events were 23.2% versus 18.3%. The result makes tumour burden, access geometry, liver reserve and centre capability part of the treatment rule. It does not establish ablation for extrahepatic disease or another cancer.\n\nPROOF 301 exposes a network failure. It screened 1,127 people at 120 sites over 40 months to find an FGFR2-rearranged cholangiocarcinoma group, but randomized only 48 of about 300 planned and stopped. Infigratinib did not beat chemotherapy in the truncated progression comparison, 7.4 versus 8.0 months, and severe adverse events were more frequent. The programme lesson is mathematical: 48 / 1,127 = 4.3% of screened people entered the trial. A rare-driver phase 3 must first prove testing coverage, referral speed and cross-border capacity; adding sites without a working route did not solve recruitment.\n\nEMERALD-3 shows why endpoints stay separate. Among 760 people with embolisation-eligible HCC, STRIDE immunotherapy plus lenvatinib plus arterial chemoembolisation increased median progression-free survival by 3.2 months, 13.0 versus 9.8. Overall survival was not significantly different in the reported analysis, hazard ratio 0.84 with a confidence interval crossing 1 and p=0.18. Serious adverse events were 64% versus 23%. The decision remains wait for mature survival and quality-of-life data; do not promote the progression result into durable control.\n\nPortfolio action: add no duplicate treatment trial. Use TORCH as a current treatment benchmark and require state-specific eligibility, delivered-treatment logs, severe-harm accounting, liver reserve and survival in any competing HCC route. The cholangiocarcinoma platform must publish the denominator from mutation testing to referral, consent and randomization. All four records preserve the separation between tumour-control time, useful life and cure.", "sourceIds": [ "sun-2026-talentop-hcc", "lyu-2026-torch-hcc", "abou-alfa-2026-proof301", "kudo-2026-emerald3" ], "links": [ { "target": "trials", "relation": "part-of" }, { "target": "hcc", "relation": "tests" }, { "target": "cholangiocarcinoma", "relation": "tests" }, { "target": "hypothesis-selected-local-consolidation", "relation": "informs" }, { "target": "hypothesis-cca-route-specific-action-platform", "relation": "informs" }, { "target": "claim-hcc-systemic-combination-improves-not-cures", "relation": "constrains" }, { "target": "unknown-hcc-liver-reserve-modifiability", "relation": "constrains" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong design evidence; mixed clinical conclusions", "independentReplication": "weak — each result is one sponsor-linked programme", "sampleSize": "TALENTOP 489 induced and 201 randomized; TORCH 241 randomized; PROOF 301 1,127 screened and 48 randomized; EMERALD-3 760 analysed", "effect": "TALENTOP time-to-failure HR 0.60; TORCH overall-survival HR 0.50; PROOF 301 stopped at 16% of target accrual; EMERALD-3 progression HR 0.70 but overall-survival HR 0.84, p=0.18", "limits": [ "TALENTOP selected only induction responders or stable disease judged resectable", "TALENTOP has not reported a mature overall-survival advantage", "TORCH enrolled at two Chinese tertiary centres and has no matching replication", "PROOF 301 was drastically underpowered", "EMERALD-3 overall survival remains immature or negative at the reported cut", "Commercial sponsorship applies to three programmes", "None establishes cure" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Each intervention follows a defined tumour or local-control route." }, "animalEvidence": { "rating": "not-applicable", "reason": "The decision rests on randomized human results." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Laboratory systems cannot answer these outcome and recruitment questions." }, "reproducibility": { "rating": "weak", "reason": "No result has an independent matching replication in the same patient state." }, "commercialConflicts": { "rating": "high", "reason": "Drug sponsors funded or materially supported TALENTOP, PROOF 301 and EMERALD-3; TORCH reported government and institutional funding." }, "recency": { "rating": "current", "reason": "All three reports were checked in September 2026." } } }, "updated": "2026-09-15" }, { "id": "trial-cca-action-map", "title": "Cholangiocarcinoma studies cover exposure, detection and recurrence but leave three action gaps", "shortTitle": "CCA action studies", "type": "trial", "status": "open", "scope": [ "cholangiocarcinoma", "hepatobiliary" ], "summary": "The evidence map separates liver-fluke prevention, PSC surveillance, endemic-area ultrasound, postoperative residual disease and treatment by anatomical and molecular group.", "content": "The study map has seven decisions. 1. Remove a proven cause: IARC classifies chronic Opisthorchis viverrini and Clonorchis sinensis infection as causes of cholangiocarcinoma. A Thai One Health programme reported infection falling from 6.0% to 0.3%, but without randomization or cancer follow-up. NCT05321992 randomizes 480 people to eCHEC or usual care, yet stops at 12-month behaviour and infection endpoints and has no posted result. 2. Watch PSC: annual MRI/MRCP plus CA19-9 in 512 people produced severe or progressive duct change in 23.8%, roughly nine non-malignant such findings for each malignancy, and median survival of 13 months after cholangiocarcinoma. A 298-person retrospective cohort linked surveillance to lower all-cause mortality but not longer survival after cancer diagnosis. 3. Screen endemic communities: Thai ultrasound found stage 0-II disease in 84.5% of 161 screen-detected cancers versus 21.6% of 601 walk-in cancers, but conditioned on diagnosed cancer and omitted population mortality and complete false-positive work. 4. Eliminate localized disease: among 1,221 intrahepatic resections, 59.4% recurred and median survival was 36.8 months. 5. Choose adjuvant treatment: BILCAP's primary intention-to-treat survival result was not significant. ACTICCA-1 randomized 789 people to gemcitabine-cisplatin or capecitabine, remains active but not recruiting after an estimated December 2025 completion and has no posted results. 6. Act on residual disease: ctDNA cohorts of 56, 89 and 18 enrolled repeatedly separate recurrence risk, but none randomized treatment or imaging from the result; NCT05743959 is an observational 100-person plan with unknown status and stale dates. 7. Treat advanced disease: TOPAZ-1 improves the survival tail across mixed biliary sites, while FGFR2 and IDH1 trials validate molecular subgroups without durable control for most patients. The three remaining action gaps are a PSC rule with tolerable benign work, a liver-fluke programme linked to long-term cancer outcomes, and a postoperative ctDNA-triggered intervention. They need separate protocols and can share data standards, independent statistics and outcome linkage.", "sourceIds": [ "iarc-2012-liver-flukes", "charoensuk-2024-one-health-opisthorchis", "clinicaltrials-nct05321992-echec", "villard-2023-psc-surveillance", "tan-2024-psc-mrcp-surveillance", "khuntikeo-2020-thailand-cca-screening", "hu-2020-icc-resection-recurrence", "primrose-2019-bilcap", "clinicaltrials-nct02170090-acticca1", "yu-2025-btc-ctdna-recurrence", "yoo-2025-extrahepatic-cca-ctdna", "park-2026-btc-ctdna-feasibility", "clinicaltrials-nct05743959-btc-mrd", "oh-2022-topaz1", "abou-alfa-2020-claridhy", "bekaii-saab-2026-fight302" ], "links": [ { "target": "trials", "relation": "part-of" }, { "target": "claim-cca-psc-surveillance-workload", "relation": "contains" }, { "target": "failure-psc-surveillance-no-curable-detection", "relation": "contains" }, { "target": "claim-fluke-control-prevents-exposure-not-yet-cancer", "relation": "contains" }, { "target": "failure-echec-cancer-outcome-mismatch", "relation": "contains" }, { "target": "claim-cca-ctdna-prognostic-not-action-guiding", "relation": "contains" }, { "target": "failure-cca-mrd-prediction-without-action", "relation": "contains" }, { "target": "unknown-cca-surveillance-utility", "relation": "exposes" }, { "target": "unknown-cca-postoperative-mrd-action", "relation": "exposes" }, { "target": "hypothesis-cca-route-specific-action-platform", "relation": "constrains" } ], "evidence": { "humanProspective": "strong — surveillance, infection and biomarker cohorts", "humanRetrospective": "strong — independent care and surgical cohorts", "randomised": "strong for selected treatments; planned for behaviour; none for surveillance or residual-disease action", "independentReplication": "moderate for components; none for a complete prevention or early-action route", "sampleSize": "PSC 512 prospective and 298 retrospective; Thai screen-detected cancer comparison 762; One Health 3,600; eCHEC planned 480; resection 1,221; ctDNA 56, 89 and 18 enrolled; ACTICCA-1 789", "effect": "Each route has a measurable signal, but none of the three early-action gaps has a randomized patient-benefit result", "limits": [ "Anatomical sites and causes differ", "Cancer endpoints need long follow-up", "Several registries are stale or have no results", "Molecular treatment trials select small subgroups", "Cross-study numbers cannot be pooled as one effect" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The map follows distinct causal and anatomical routes from exposure or duct injury to residual and advanced disease." }, "animalEvidence": { "rating": "moderate", "reason": "Models support liver-fluke carcinogenesis and drug mechanisms but cannot settle human action utility." }, "inVitroEvidence": { "rating": "moderate", "reason": "Molecular assays and target models support components, not the full care routes." }, "reproducibility": { "rating": "weak", "reason": "Signals repeat within some components, while no complete early-action pathway has reproduced better survival." }, "commercialConflicts": { "rating": "high", "reason": "Imaging, endoscopy, proprietary ctDNA assays and targeted medicines can each favour more use without net-benefit proof." }, "recency": { "rating": "current", "reason": "The map includes registry checks and human evidence through September 2026." } } }, "updated": "2026-09-14" }, { "id": "trial-cell-cycle-checkpoint-action-map", "title": "Cell-cycle action map: stop broad CDK work and recover the completed WEE1 safety stage", "shortTitle": "Cell-cycle action map", "type": "trial", "status": "mixed", "scope": [ "pdac", "cholangiocarcinoma" ], "summary": "No new trial is needed now: broad CDK4/6 and CHK1 routes failed, while the missing WEE1 answer sits in 12 completed records.", "content": "Closed routes. Do not run another unselected or CDKN2A-selected CDK4/6 monotherapy study. TAPUR closed both pancreatic and biliary cohorts for no response or 16-week stability. Do not repeat abemaciclib plus a broad resistance partner: both randomized arms controlled less disease and progressed faster than chemotherapy. Do not repeat CHK1 plus gemcitabine without a new human selector: one 99-person randomized trial was negative after adequate exposure and an 11-person replacement produced no response despite sensitive participant organoids. Open data task. NCT06015659 contains the only immediate WEE1 decision. The design promised up to 12 safety participants before a Simon two-stage efficacy expansion; the registry now reports actual n=12 and primary completion 6 August 2025. Obtain every participant's allocation date, drug exposure, tumour biopsy, blood, scan, six-month status, later treatment, survival, adverse events and the committee decision not to expand. Preserve failed and missing biopsies. Gate. At least ten paired tumours must be usable, at least eight must show the planned checkpoint and DNA-damage change, six-month PFS must reach at least 5/12, no treatment-related death may occur, and a candidate state must repeat across two assay methods. Hair-follicle target effect, plasma exposure or organoid killing cannot pass. If the gate fails, retire PDAC WEE1 work. If it passes, design a small randomized same-backbone contribution test, not a larger single-arm study. Capital. Mission cost is £0 now. Data recovery fits the existing decision core. No drug investment, cohort, licence or company rescue is justified before the 12-person record is public.", "sourceIds": [ "chiorean-2023-abemaciclib-pdac", "baghdadi-2019-palbociclib-cdkn2a", "hidalgo-2022-palbociclib-nabpaclitaxel", "laquente-2017-chk1-pdac", "huffman-2023-ly2880070-pdac", "cuneo-2019-adavosertib-pdac", "huffman-2024-azenosertib-pdac-design", "clinicaltrials-nct06015659-azenosertib" ], "links": [ { "target": "trials", "relation": "part-of" }, { "target": "failure-broad-cell-cycle-checkpoint-pdac", "relation": "maps" }, { "target": "unknown-wee1-replication-stress-pdac", "relation": "maps" }, { "target": "hypothesis-human-replication-stress-selector", "relation": "gates" }, { "target": "company-pfizer-cachexia", "relation": "learns-from" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong", "independentReplication": "moderate", "sampleSize": "Reuses CDK4/6 and CHK1 evidence and the completed 12-person WEE1 safety stage; no new cohort", "effect": "Stops duplicate broad work and defines a zero-cost tumour-mechanism and efficacy recovery gate", "limits": [ "Azenosertib data access untested", "Safety-stage sample is small", "Assay plan not public", "Thresholds are mission decisions", "Pfizer link is historical palbociclib learning, not current access" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "The remaining WEE1 question is narrow and starts with direct tumour measurements." }, "animalEvidence": { "rating": "strong", "reason": "Model support exists but is deliberately not a release gate." }, "inVitroEvidence": { "rating": "strong", "reason": "The programme explicitly rejects organoid sensitivity unless the same signal appears in treated tumours." }, "reproducibility": { "rating": "weak", "reason": "Failure is reproduced; the remaining WEE1 state is not." }, "commercialConflicts": { "rating": "moderate", "reason": "Multiple commercial agents are involved, but the next task uses existing academic-trial data." }, "recency": { "rating": "current", "reason": "The action map uses the completed 2025 primary endpoint and 2026 registry state." } } }, "layout": { "x": 214, "y": 132 }, "updated": "2026-09-15" }, { "id": "trial-cldn182-antibody-action-map", "title": "CLDN18.2 action map: let the funded phase 3 answer the IBI343 question", "shortTitle": "CLDN18.2 action map", "type": "trial", "status": "mixed", "scope": [ "pdac", "pan-cancer", "transfer-unproven" ], "summary": "No duplicate trial is needed: repair the live IBI343 phase 3 record, recover mature IBI389 data and stop broad product transfer.", "content": "Closed route. Do not repeat plain zolbetuximab plus gemcitabine/nab-paclitaxel in high-CLDN18.2 metastatic PDAC: 393 randomized people produced OS HR 0.999 and 19.8 points more serious treatment-emergent events. Do not advance mesothelin anetumab after zero responses among 14 evaluable patients. Do not infer efficacy from the four unconfirmed EBC-129 responses or from SHR-3821 disease stability after zero confirmed responses at its main doses. Existing-data task. Recover all 73 pancreatic IBI389 records and reconcile them with the preliminary 64-person report. Require treatment, dose, marker fraction and intensity, every scan, response confirmation, duration, progression, survival, cytokine-release syndrome, liver injury and reason for stopping. If mature confirmed ORR is below 15%, median response duration below six months or severe treatment-related harm above 40%, do not fund a randomized IBI389 test. Live randomized task. G-HOPE-002 already performs the required IBI343 contribution test. Before unmasking, publish the protocol, dated analysis plan, exact assay and cutoff, site plan, death target, effect target, missing-data rule, quality-of-life plan and participant-data-sharing route. Preserve all 201 assigned people. Read OS first, quality-adjusted survival second, then PFS and response. Go requires OS HR no higher than 0.75 with its 95% interval excluding 1.0, positive quality-adjusted survival, at least 99% survival accounting, no marker-rule change and no severe-harm excess above ten points. Stop for an interval including 1.0, a plausible missing-data result above HR 0.85, a post-outcome marker subgroup or intolerable hospital time. Capital. Mission investment is £0. Innovent reported RMB30.2bn cash reserves and Takeda owns development and manufacturing rights outside Greater China. The mission buys neither shares nor a duplicate cohort; it requests public, reusable evidence and monitors the already funded answer.", "sourceIds": [ "park-2026-gleam-zolbetuximab", "astellas-2026-gleam-medical-information", "clinicaltrials-nct03816163-gleam", "clinicaltrials-nct03023722-anetumab", "lentz-2025-ebc129-pdac", "gou-2026-shr3821", "hao-2024-ibi389-pdac", "xu-2025-ibi389-phase1", "clinicaltrials-nct05164458-ibi389", "yu-2025-ibi343-pdac", "clinicaltrials-nct05458219-ibi343", "clinicaltrials-nct07066098-ibi343-phase3", "abou-alfa-2006-exatecan-pdac", "innovent-2026-interim-results", "takeda-2025-ibi343-license" ], "links": [ { "target": "trials", "relation": "part-of" }, { "target": "claim-cldn182-adc-selected-response", "relation": "maps" }, { "target": "claim-cldn18-bispecific-early", "relation": "maps" }, { "target": "failure-antibody-target-and-payload-transfer-pdac", "relation": "maps" }, { "target": "unknown-cldn182-adc-survival-effect", "relation": "maps" }, { "target": "hypothesis-cldn182-adc-survival-test", "relation": "gates" }, { "target": "company-innovent-biologics", "relation": "monitors" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong", "independentReplication": "weak", "sampleSize": "Uses completed randomized n=393, active randomized planned n=201, mature IBI389 PDAC n=73 and several product controls", "effect": "Avoids a duplicate trial and converts live phase 3 into an explicit survival, quality and transparency gate", "limits": [ "IBI343 protocol unavailable", "Phase 3 currently lists one site", "IBI389 mature data access untested", "Company finance changes", "Thresholds are mission decisions" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The action map tests one address-carrier-payload product rather than treating antibodies as one class." }, "animalEvidence": { "rating": "moderate", "reason": "Model work does not release capital or change the human gate." }, "inVitroEvidence": { "rating": "strong", "reason": "Assay, uptake and payload measurements can explain a result but cannot replace survival." }, "reproducibility": { "rating": "weak", "reason": "Zolbetuximab failure is controlled; IBI343 and IBI389 signals are not yet independently reproduced." }, "commercialConflicts": { "rating": "high", "reason": "The decisive records are controlled by Innovent, Takeda and Astellas." }, "recency": { "rating": "current", "reason": "Uses active 2026 trial, company and rights records." } } }, "layout": { "x": 224, "y": 142 }, "updated": "2026-09-15" }, { "id": "trial-crc-action-map", "title": "Colorectal prevention, residual-disease and liver-control action map", "shortTitle": "CRC action trials", "type": "trial", "status": "open", "scope": [ "colorectal" ], "summary": "The colorectal map separates population invitation, treatment omission, treatment escalation, earlier imaging and organ-confined local control.", "content": "Population prevention. Repeated stool testing and one-time flexible sigmoidoscopy reduced cancer incidence and death. NordICC tests a different rule: one colonoscopy invitation. At 13 years it reduced incidence by 0.34 percentage points but did not show fewer colorectal-cancer deaths, with only 42.0% attendance. Do not replace invitation effects with estimates among attenders. Treatment omission. DYNAMIC stage II reduced chemotherapy and preserved two-year recurrence-free survival. DYNAMIC-III stage III reduced oxaliplatin but failed noninferiority, so omission is stage and rule specific. Treatment escalation. DYNAMIC-III found no recurrence-free-survival gain from more chemotherapy in ctDNA-positive patients. ALTAIR found no significant gain from trifluridine/tipiracil at molecular relapse and much more severe blood toxicity. CIRCULATE-US, NCT05174169, is recruiting 1,912 planned participants, tests no chemotherapy in ctDNA-negative disease and mFOLFIRINOX in ctDNA-positive disease, completes its primary endpoint in March 2029 and has no results. Earlier imaging. FIND randomized 584 eligible patients; ctDNA-triggered CT raised curative-intent treatment among recurrent patients from 23.6% to 48.1%, but long-term survival is pending. Organ-confined treatment. CLOCC and TransMet support long survival in narrow liver-only groups; EPOCH rejects broad liver radioembolization based on liver involvement alone. Capital rule. Do not fund another general prognostic cohort or repeat named chemotherapy questions. Fund assay comparison, serial resistance biology, mature FIND survival follow-up and a locked organ-confined selector that can route people to a randomized action.", "sourceIds": [ "mandel-2000-fobt-incidence", "atkin-2010-flexible-sigmoidoscopy", "bretthauer-2022-nordicc-10y", "kaminski-2026-nordicc-13y", "tie-2022-dynamic-colon", "tie-2025-dynamic3", "bando-2026-altair", "clinicaltrials-nct05174169-circulate-us", "mo-2026-find-crc-surveillance", "ruers-2017-clocc-crc-liver-ablation", "adam-2024-transmet", "mulcahy-2021-epoch-radioembolization" ], "links": [ { "target": "trials", "relation": "part-of" }, { "target": "claim-crc-colonoscopy-invitation-benefit-limited-by-participation", "relation": "tests" }, { "target": "claim-crc-ctdna-utility-is-action-specific", "relation": "tests" }, { "target": "unknown-crc-ctdna-positive-action", "relation": "constrains" }, { "target": "unknown-crc-liver-local-cure-selector", "relation": "constrains" }, { "target": "model-colorectal-causal-chain", "relation": "informs" } ], "evidence": { "humanProspective": "strong for completed rows; planned for the open row", "humanRetrospective": "not-applicable", "randomised": "strong for completed actions; pending for CIRCULATE-US", "independentReplication": "strong for prognosis and screening; mixed for treatment action", "sampleSize": "Completed studies from 119 to 170,432 participants; CIRCULATE-US plans 1,912", "effect": "Maps action-specific effects and failures; no pooled effect", "limits": [ "Different stages, assays and endpoints", "Registry status can change", "CIRCULATE-US has no result", "FIND survival is immature", "TransMet and CLOCC use narrow selection", "Population screening effects depend on uptake" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The map follows precursor removal, residual disease, action choice and organ-confined recurrence." }, "animalEvidence": { "rating": "not-applicable", "reason": "The map compares human decisions and trial endpoints." }, "inVitroEvidence": { "rating": "weak", "reason": "Laboratory work may nominate an action but cannot establish clinical utility." }, "reproducibility": { "rating": "strong", "reason": "Multiple randomized studies now test prevention, systemic action, imaging and local control." }, "commercialConflicts": { "rating": "high", "reason": "Commercial assay and treatment interests are material in several residual-disease rows." }, "recency": { "rating": "current", "reason": "The map includes 2026 reports and a registry checked in September 2026." } } }, "updated": "2026-09-14" }, { "id": "trial-daraxonrasib-expansion-programme", "title": "Three randomized phase 3 trials test whether common-RAS control transfers to first-line, post-resection and dual-RAS settings", "shortTitle": "Daraxonrasib phase III map", "type": "trial", "status": "open", "scope": [ "pdac" ], "summary": "Recruiting trials test first-line daraxonrasib with or without chemotherapy, post-resection daraxonrasib versus observation, and dual RAS inhibition for KRAS G12D disease.", "content": "The next three phase 3 tests directly span the main causal bottlenecks. NCT07491445 plans 900 untreated metastatic patients across daraxonrasib alone, daraxonrasib plus gemcitabine/nab-paclitaxel, and chemotherapy alone; PFS and OS are primary and completion is estimated in 2028. NCT07252232 plans 500 resected patients who completed perioperative multiagent chemotherapy, randomising daraxonrasib against observation; disease-free survival is primary and completion is estimated in 2029. NCT07805954 plans 400 untreated KRAS-G12D patients, comparing daraxonrasib plus the G12D-selective inhibitor zoldonrasib against chemotherapy; PFS and OS are primary and completion is estimated in 2029. The dual-inhibition rationale is now tied to human escape rather than models alone: paired phase 1/2 plasma found treatment-emergent RAS-pathway changes in 59% and mutant KRAS amplification in 36%, while models showed the G12D-selective plus broad inhibitor could delay resistance. The first-line trial separates replacement from combination. The resected trial asks whether suppressing occult systemic disease delays recurrence, but observation rather than placebo leaves symptoms and treatment exposure unblinded. The dual-inhibition trial tests depth, but may increase toxicity and still leave RTK, MAPK, PI3K, non-genomic or non-RAS resistance. Its serial plasma and tissue plan should be checked for independent access and enough samples to reproduce the 36% class. Registry designs are plans, not results, and dates can move.", "sourceIds": [ "nct07491445-daraxonrasib-first-line", "nct07252232-daraxonrasib-resected", "nct07805954-dual-ras-first-line", "oreilly-2026-daraxonrasib", "aronchik-2026-daraxonrasib-resistance", "lin-2026-vertical-kras-inhibition" ], "links": [ { "target": "claim-daraxonrasib-common-ras-survival-breakthrough", "relation": "extends" }, { "target": "claim-adjuvant-mfolfirinox-improves-but-recurrence-remains", "relation": "tests" }, { "target": "claim-kras-mutant-dosage-may-shape-resistance", "relation": "tests" }, { "target": "claim-daraxonrasib-kras-amplification-resistance", "relation": "must-reproduce" }, { "target": "residual-disease", "relation": "tests" }, { "target": "timeline", "relation": "part-of" }, { "target": "companies", "relation": "part-of" } ], "evidence": { "humanProspective": "none", "humanRetrospective": "not-applicable", "randomised": "none", "independentReplication": "none", "sampleSize": "Paired resistance evidence n=44; planned total 1,800 patients across three recruiting phase 3 trials", "effect": "No results yet", "limits": [ "Sponsor-run trials", "Open-label", "Recruiting", "Planned enrolment may change", "Completion dates are estimates", "No efficacy data", "Post-resection study is not biomarker-triggered" ] }, "layout": { "x": 93, "y": 58 }, "updated": "2026-09-14" }, { "id": "trial-elraglusib-action-map", "title": "Elraglusib action map: audit the randomized record before phase 3", "shortTitle": "Elraglusib action map", "type": "trial", "status": "mixed", "scope": [ "pdac" ], "summary": "Use a no-new-patient data gate first; only a robust all-randomized result can release a blinded survival confirmation.", "content": "Stage 0: recover the existing trial. Obtain participant-level NCT03678883 data, every protocol and statistical-analysis-plan version, randomization dates, treatment-start dates, survival follow-up and reasons for the 36 pre-treatment removals. Reconcile 286 randomized and 250 treated in the paper with the registry's estimated 350 total. Separate the discarded 17-person twice-weekly dose-selection arm from the 269-person weekly-versus-control comparison; do not silently erase it from the trial history. Require at least 99% survival accounting. Reproduce all-randomized OS, restricted mean survival and one-year survival. Test missing-outcome tipping points, early deaths, site, country, baseline albumin, performance status, liver metastasis, CA19-9, later treatment and treatment exposure. Publish code and negative results. The paper offers data to qualified applicants, while the registry says no sharing; submit and publish the access outcome. Release gate. Continue only if ITT HR is at most 0.75, the planned 95% interval excludes 1.0, no plausible missing-outcome case makes HR exceed 0.85, and the endpoint-change decision predates access to comparative unblinded survival. Stage 1: confirm survival. Run a double-blind placebo-controlled phase 3 of elraglusib plus gemcitabine/nab-paclitaxel versus matched placebo plus the same chemotherapy. Count all randomized people. Stratify region, performance status, liver metastasis and albumin. Use OS primary; blinded central PFS and response, restricted mean survival, days alive outside hospital, quality-adjusted survival, severe toxicity, visual function and chemotherapy dose intensity are required outputs. Ask participants and staff to guess assignment so masking failure is measured. With equal groups, two-sided alpha 0.05 and 80% power, detecting HR 0.78 requires about 508 deaths: 4×(1.96+0.84)^2/[ln(0.78)]^2=508. Allowing 10% without an event implies roughly 565 people; use 600-650 to absorb stratification and uncertainty. Capital. Commit £0 to phase 3 now. Stage 0 can use the existing independent-statistics team. Actuate's 30 June cash divided by its six-month monthly burn is 4.426/(9.422/6)=2.82 months, and no later company financing filing was found by 15 September. That creates urgency but not scientific validity. A phase 3 requires a well-funded partner to carry product, manufacturing and most trial cost; any later mission co-funding must displace a lower-ranked programme inside the £500m ceiling.", "sourceIds": [ "mahalingam-2026-elraglusib-randomized", "clinicaltrials-nct03678883-elraglusib", "pathak-2026-elraglusib-folfirinox", "clinicaltrials-nct05077800-elraglusib-folfirinox", "actuate-2026-q2-10q", "sec-2026-actuate-submissions" ], "links": [ { "target": "trials", "relation": "part-of" }, { "target": "claim-elraglusib-randomized-survival-signal", "relation": "maps" }, { "target": "failure-elraglusib-analysis-and-survival-discordance", "relation": "repairs" }, { "target": "unknown-elraglusib-survival-replication", "relation": "maps" }, { "target": "hypothesis-elraglusib-survival-confirmation-gate", "relation": "gates" }, { "target": "company-actuate-therapeutics", "relation": "requires-access-from" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "moderate", "independentReplication": "none", "sampleSize": "Stage 0 reuses n=269 randomized weekly or control records; phase 3 planning case about 600-650 people for roughly 508 deaths", "effect": "No-new-patient audit followed conditionally by an ITT OS confirmation powered near HR 0.78", "limits": [ "Data request not yet made", "Power calculation is a planning approximation", "Partner and protocol unnegotiated", "Visual effects can reveal treatment", "Finance snapshot may change" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "The plan starts from a human survival signal rather than requiring a settled mechanism." }, "animalEvidence": { "rating": "moderate", "reason": "Animal evidence is supportive but not a release gate." }, "inVitroEvidence": { "rating": "moderate", "reason": "Target assays can support mechanism but cannot replace survival confirmation." }, "reproducibility": { "rating": "weak", "reason": "Stage 0 can test internal robustness; only phase 3 can supply independent confirmation." }, "commercialConflicts": { "rating": "high", "reason": "Actuate controls the medicine, existing data and manufacturing route." }, "recency": { "rating": "current", "reason": "Uses September 2026 registries and SEC filing index." } } }, "layout": { "x": 204, "y": 122 }, "updated": "2026-09-15" }, { "id": "trial-enzeavour-japan-screening", "title": "Japan is testing a complete 10,000-person pancreatic screening workflow, not only stored samples", "shortTitle": "ENZEAVOUR Japan", "type": "trial", "status": "open", "scope": [ "pdac", "pancreatic-all" ], "summary": "A recruiting single-arm study screens asymptomatic adults and sends positive results to MRI/MRCP, EUS or CT; detection rate and positive predictive value are measured.", "content": "ENZEAVOUR-FS plans 10,000 asymptomatic adults attending routine health checks across Japan. The blood assay measures single-molecule enzyme activity. Positive participants receive investigator-selected MRI/MRCP, EUS or contrast CT. The primary endpoint is pancreatic-cancer detection within 12 months and the main secondary endpoint is positive predictive value. The design is more informative than a retrospective case-control study because it measures a real screening population and diagnostic pathway. It is still uncontrolled and mainly verifies positive tests, so sensitivity, interval cancers, overdiagnosis, procedure harm and mortality cannot be established. At an annual incidence near 14 per 100,000, an unselected 10,000-person cohort would expect roughly 1.4 cancers per year; Japanese age and clinic selection may raise this, but the study may still yield few cancers. It should be treated as the sample-size and workflow stage before a randomized utility trial, exactly as the registry states.", "sourceIds": [ "jrct-2026-enzeavour-feasibility", "nedo-2024-cosomil-profile", "seer-pancreas-statfacts" ], "links": [ { "target": "detection", "relation": "part-of" }, { "target": "claim-screening-base-rate-maths", "relation": "constrained-by" }, { "target": "company-cosomil", "relation": "uses-product-from" }, { "target": "claim-case-control-blood-tests-overstate-screening-readiness", "relation": "advances-beyond" }, { "target": "programme-500m-integrated-mission", "relation": "informs" } ], "evidence": { "humanProspective": "none — planned study, no result", "humanRetrospective": "not-applicable", "randomised": "none", "independentReplication": "none", "sampleSize": "Target 10,000 asymptomatic adults", "effect": "No results yet", "limits": [ "Single arm", "Partial verification", "Expected cancer count may be very small", "No mortality endpoint", "Product-provider conflict", "Assay accuracy not independently established in this population" ] }, "layout": { "x": 139, "y": 54 }, "updated": "2026-09-14" }, { "id": "trial-gallbladder-prevention-detection-action-map", "title": "Gallbladder prevention and detection studies cover fragments of the action path", "shortTitle": "Gallbladder action studies", "type": "trial", "status": "open", "scope": [ "gallbladder", "hepatobiliary" ], "summary": "Active cohorts can discover risk markers, diagnose visible lesions or triage pathology, but none yet tests whether risk-guided surgery prevents death.", "content": "The study map contains five different decisions. 1. Future risk: the planned Chinese 100,000-person benign-disease cohort would have measured cancer incidence but still says not yet recruiting and has no results. No live public study in this map follows a broad operated and unoperated population long enough to validate a future-cancer threshold. 2. Candidate discovery: EULAT Eradicate GBC is recruiting an estimated 15,000 people across South America and Europe, while Chile BiLS has recruited 313 prevalent or incident HGD-plus cases through August 2025 with gallstone controls. Their operated or diagnosed case-control structure can find candidate exposures and markers but cannot supply prospective population PPV by itself. 3. Before-diagnosis blood: six European cohorts produced only 73 prospectively sampled cancer-control pairs; miR-4533 repeated, but no locked classifier or action threshold exists. 4. Existing-lesion diagnosis: GBCseeker and GAIA-MIL show that multimodal CT and ultrasound models can discriminate selected cancer-enriched cases, with external performance loss and no screening or randomized action result. 5. After-removal detection: FANCY studied selective pathology in 10,041 Dutch operations. Routine histology found 28 malignancies; a selective approach would miss seven diagnoses with clinical consequences among 7,846 omitted specimens. Across the full cohort only one patient had clear treatment benefit from radical removal of residual disease and six had futile additional surgery. P-iGBC is now in feasibility and may scale to 30,000 UK operations to improve this pathology decision; it excludes suspicious disease, polyps at least 5 mm, PSC and other high-risk pathways and has no longitudinal follow-up.\n\nDecision: do not launch a stand-alone biomarker, imaging or universal-polyp trial. First join EULAT, Chile BiLS and P-iGBC fields around one pathology and outcome definition, obtain local cholecystectomy-harm data, and prospectively freeze external tests of the best imaging and blood candidates. The critical missing study is a country-specific action trial in which a locked rule changes whether or when an otherwise optional cholecystectomy occurs and measures HGD or invasive cancer, interval cancer, complications, quality of life and cancer death. Pathology triage after surgery is useful health-system work but cannot answer that prevention question.", "sourceIds": [ "clinicaltrials-nct03762837-gbc-risk-cohort", "clinicaltrials-nct06192719-eulat-gbc", "koshiol-2026-chile-bils", "blandino-2025-gallbladder-mirna", "yang-2025-gbcseeker", "gupta-2026-gaia-mil", "bastiaenen-2022-fancy", "clinicaltrials-nct06531408-pigbc", "wong-2025-ambrose-cholecystectomy" ], "links": [ { "target": "trials", "relation": "part-of" }, { "target": "failure-nct03762837-risk-cohort-never-started", "relation": "contains" }, { "target": "claim-gallbladder-polyp-guidelines-still-overtreat", "relation": "depends-on" }, { "target": "claim-gallbladder-diagnostic-models-existing-lesions-only", "relation": "depends-on" }, { "target": "claim-gallbladder-prevention-must-beat-surgery-harm", "relation": "depends-on" }, { "target": "hypothesis-gallbladder-two-stage-interception-rule", "relation": "constrains" }, { "target": "programme-500m-integrated-mission", "relation": "constrains" } ], "evidence": { "humanProspective": "moderate — multiple active or completed cohorts answer different parts", "humanRetrospective": "moderate", "randomised": "none for prevention action", "independentReplication": "weak for a complete pathway", "sampleSize": "EULAT planned n=15,000; Chile BiLS HGD-plus cases n=313; P-iGBC up to n=30,000; FANCY n=10,041; prospective serum case-control pairs n=73", "effect": "No study in the map reports fewer gallbladder-cancer deaths from a risk-guided surgery rule", "limits": [ "Registries state plans, not results", "Candidate cohorts are enriched by diagnosis or surgery", "Country incidence and operation harm differ", "No common prospective action threshold", "No randomized prevention result", "Long cancer latency delays definitive mortality evidence" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Removing a true dangerous precursor or high-risk organ before invasion could prevent cancer." }, "animalEvidence": { "rating": "weak", "reason": "Animal models may study progression but cannot validate human surgery selection." }, "inVitroEvidence": { "rating": "moderate", "reason": "Blood and tissue markers generate candidates but lack a validated action threshold." }, "reproducibility": { "rating": "weak", "reason": "Individual signals repeat, while no full risk-to-action pathway has been independently reproduced." }, "commercialConflicts": { "rating": "moderate", "reason": "Assay and AI suppliers can benefit from broad use; active cohorts are mainly academic or public." }, "recency": { "rating": "current", "reason": "Live registry status and studies were checked through September 2026." } } }, "layout": { "x": 104, "y": 81 }, "updated": "2026-09-14" }, { "id": "trial-gaspar-lacog-0222", "title": "GASPAR / LACOG 0222 — Latin American pancreatic and gastric cohort", "shortTitle": "GASPAR", "type": "trial", "status": "supported", "scope": [ "pdac", "pancreatic-all", "pan-cancer" ], "summary": "GASPAR is a completed observational route across seven sites and three countries, with 79 advanced pancreatic cases in its 2025 report and a material registry-count discrepancy.", "content": "Design: retrospective and prospective observational cohort of advanced gastric, gastro-oesophageal and pancreatic cancer diagnosed from 2019, registered as NCT05924789. The 2025 abstract reports 201 people from seven sites in Brazil, Argentina and Colombia: 122 gastric and 79 pancreatic. In the pancreatic cohort, 65/79 = 82% entered with metastatic disease, 67/79 = 85% received first-line treatment, 27/79 = 34% received second line and 10/79 = 13% third line. Median first-line progression-free survival was 5.4 months and median overall survival 10.2 months. These figures describe selected care; there is no randomized comparator and no population denominator. Representation is a hard limit: 76% of the combined cohort had private coverage, and the pancreatic cohort was reported as 90% white with no Black participants. Data-quality flag: ClinicalTrials.gov reports actual enrolment of 120, whereas the meeting abstract reports 201; do not pool or reuse the cohort until that difference is reconciled.", "sourceIds": [ "clinicaltrials-nct05924789-gaspar", "peixoto-2025-gaspar-results" ], "links": [ { "target": "trials", "relation": "part-of" }, { "target": "person-renata-dalpino-peixoto", "relation": "led-by" }, { "target": "lab-lacog-gaspar", "relation": "run-by" }, { "target": "datasets", "relation": "could-supply" }, { "target": "programme-unconstrained-global-system", "relation": "could-inform" } ], "evidence": { "humanProspective": "moderate — partly prospective observational cohort", "humanRetrospective": "moderate", "randomised": "none", "independentReplication": "none", "sampleSize": "Abstract n=201 overall and n=79 pancreatic; registry reports n=120 actual", "effect": "Pancreatic median PFS 5.4 months and OS 10.2 months; descriptive only", "limits": [ "No comparator", "Selected advanced-disease cohort", "Mostly private coverage", "Poor racial representation", "Registry and abstract totals conflict", "Meeting abstract without full paper" ], "dimensions": { "biologicalPlausibility": { "rating": "not-applicable", "reason": "The cohort describes care and outcomes rather than testing a biological mechanism." }, "animalEvidence": { "rating": "not-applicable", "reason": "Animal evidence does not affect this descriptive cohort." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Laboratory evidence is not part of the reported analysis." }, "reproducibility": { "rating": "weak", "reason": "Only a meeting abstract is available and its total conflicts with the registry." }, "commercialConflicts": { "rating": "moderate", "reason": "The registry lists Astellas as collaborator and includes commercial care sites." }, "recency": { "rating": "current", "reason": "Results were reported in 2025 and the completed registry was checked in 2026." } } }, "layout": { "x": 46, "y": 124 }, "updated": "2026-09-14" }, { "id": "trial-hcc-action-map", "title": "HCC prevention and surveillance action map", "shortTitle": "HCC action trials", "type": "trial", "status": "open", "scope": [ "hcc", "hepatobiliary" ], "summary": "The HCC trial map separates cause control, surveillance frequency, test substitution, late-stage prevention and mortality so that a new programme funds only the unanswered action.", "content": "Prevention. Taiwan's vaccination programme and the 651-person lamivudine trial establish viral cause control. ATTENTION moves the action earlier in noncirrhotic HBV, but its 2025 interim analysis had nine HCCs and did not cross the stop boundary; wait for mature follow-up before booking the effect. Surveillance frequency. NCT00190385 completed: ultrasound every three months found more tiny lesions than every six months but no more HCC 30 mm or smaller. Do not fund a general interval repeat. Marker add-on. NCT02272504 produced the 1,208-person randomized report: AFP, AFP-L3 and DCP added to ultrasound did not improve early-stage detection. Do not repeat its fixed add-on rule. MRI substitution. NCT02514434 produced the 414-person MIRACLE-HCC report: earlier BCLC stage and fewer false referrals with noncontrast MRI, but no mortality result. Treat this as a service and stage signal. Late-stage endpoint. NCT06084234 TRACER is recruiting 5,500 people and compares GALAD with ultrasound with or without AFP. It measures late-stage HCC, curative treatment and harms; primary completion is estimated for December 2029 and there are no results. Mortality endpoint. NCT05486572 PREMIUM is recruiting 4,700 VA participants and compares abbreviated contrast MRI plus AFP with ultrasound plus AFP. HCC mortality is primary; completion is estimated for September 2030 and there are no results. Capital rule. Do not launch another broad GALAD or MRI superiority trial while these remain viable. A £6m HCC first stage should instead build a route-stratified dataset that both trials can use: serial liver reserve, treatment eligibility, attendance, false work-up, cause control, decompensation and death. Release a larger co-control trial only if the reserve intervention changes a measured liver-risk state without worsening tumour control and does not duplicate a live randomized question.", "sourceIds": [ "chang-1997-hbv-vaccine-hcc", "liaw-2004-lamivudine-hbv", "lim-2025-attention-hbv", "zhang-2004-hcc-screening", "trinchet-2011-hcc-surveillance-interval", "hirode-2026-hcc-biomarker-rct", "rhee-2025-miracle-hcc", "singal-2024-tracer-protocol", "clinicaltrials-nct06084234-tracer", "ioannou-2026-premium-protocol", "clinicaltrials-nct05486572-premium" ], "links": [ { "target": "trials", "relation": "part-of" }, { "target": "claim-hcc-prevention-is-proven-for-viral-cause", "relation": "tests" }, { "target": "claim-hcc-surveillance-test-improvement-not-mortality", "relation": "tests" }, { "target": "unknown-hcc-surveillance-modality-utility", "relation": "resolves" }, { "target": "unknown-hcc-liver-reserve-modifiability", "relation": "constrains" }, { "target": "hypothesis-hcc-liver-reserve-co-control", "relation": "informs" } ], "evidence": { "humanProspective": "strong for completed rows; planned for open rows", "humanRetrospective": "not-applicable", "randomised": "strong for completed interval and add-on questions; pending for late-stage and mortality questions", "independentReplication": "moderate — mixed across actions and endpoints", "sampleSize": "Completed trials n=651, 18,816, 1,278, 1,208 and 414; live TRACER n=5,500 estimated; live PREMIUM n=4,700 estimated", "effect": "Maps tested actions and pending decision dates; no new pooled effect", "limits": [ "Registry status can change", "TRACER and PREMIUM results are not available", "The completed trials use different populations and endpoints", "No direct comparison of all modern routes", "ATTENTION is interim", "A co-control trial still needs an intervention with a measurable reserve effect" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The map follows the causal route from viral injury through detection, treatment eligibility, liver reserve and death." }, "animalEvidence": { "rating": "not-applicable", "reason": "The map concerns human trial actions and duplication." }, "inVitroEvidence": { "rating": "weak", "reason": "Laboratory assay performance cannot replace trial utility." }, "reproducibility": { "rating": "moderate", "reason": "Several randomized trials exist, but each tests a different link and modern mortality evidence remains pending." }, "commercialConflicts": { "rating": "moderate", "reason": "Industry ties affect the marker and antiviral rows; the large mortality trial is publicly sponsored." }, "recency": { "rating": "current", "reason": "TRACER and PREMIUM registry records were checked in September 2026." } } }, "updated": "2026-09-14" }, { "id": "trial-immune-reprogramming-action-map", "title": "PDAC immune-reprogramming action map: stop two routes, reuse one archive and gate one trial", "shortTitle": "Immune action map", "type": "trial", "status": "active", "scope": [ "pdac", "pan-cancer", "transfer-unproven" ], "summary": "Stop broad NIS793 and cabiralizumab, test the angiogenin claim in existing samples, and release mitazalimab data before any matched survival investment.", "content": "Route 1, TGF-beta. No new unselected NIS793 or broad TGF-beta cohort. Publish the daNIS-2 OS hazard ratio, patient function, hospital time and individual record. Test the published high-angiogenin threshold in stored blinded baseline samples linked to all 490 randomized outcomes. Budget up to £1m only after complete data and publication rights; otherwise £0. Route 2, macrophage CSF1R. Cabiralizumab plus nivolumab alone was worse than chemotherapy, its gemcitabine combination had PFS HR 1.00 and OS HR 1.04, and GemCaN enrolled two people. Close the route and spend £0 beyond data release. Route 3, CD40. First reconcile 88 screened, 70 initially enrolled, 57 efficacy-evaluable and 94 finally registered OPTIMIZE-1 participants. Publish every dose, chemotherapy exposure, response, duration, survival, severe event and later treatment. Lock any marker rule before using more outcomes. If the signal survives and a solvent owner supplies drug, manufacture, at least £25m and complete data rights, fund no more than £25m in stages toward a roughly 450-person placebo-controlled mFOLFIRINOX trial. The split is £5m for transfer, manufacturing audit and startup, then up to £20m against recruitment, data and safety milestones. No equity rescue. Logic. The archive study can remove a whole branch cheaply. The controlled CD40 trial can measure useful life but is too large and commercially captured for the mission to carry alone. Capital at risk before the data and partner gates is £1m, not £26m.", "sourceIds": [ "clinicaltrials-nct04935359-danis2", "pelletier-2026-nis793-biomarkers", "pietrobono-2026-angiogenin-alk5", "clinicaltrials-nct03336216-cabiralizumab", "clinicaltrials-nct03697564-gemcan", "van-laethem-2025-optimize1-biomarkers", "clinicaltrials-nct04888312-optimize1", "padron-2022-prince", "alligator-2026-q2" ], "links": [ { "target": "trials", "relation": "part-of" }, { "target": "claim-nis793-target-engagement-without-benefit", "relation": "maps" }, { "target": "claim-cd40-signal-without-contribution-proof", "relation": "maps" }, { "target": "failure-immune-reprogramming-without-patient-benefit", "relation": "repairs" }, { "target": "unknown-cd40-added-survival-pdac", "relation": "resolves" }, { "target": "hypothesis-angiogenin-tgfb-existing-data-test", "relation": "implements" }, { "target": "hypothesis-mitazalimab-controlled-survival-gate", "relation": "implements" }, { "target": "company-alligator-bioscience", "relation": "monitors" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "moderate", "randomised": "strong", "independentReplication": "moderate", "sampleSize": "Existing randomized records n=511, n=205, n=129 and n=156; mitazalimab registry n=94; proposed n about 450", "effect": "Capital decision map: £1m immediate ceiling, £25m later matched ceiling, £0 for stopped routes", "limits": [ "Archived samples may not exist", "Rights and drug supply are unconfirmed", "Trial cost is an estimate", "Different products do not prove a class-wide effect" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Each route follows a measurable target-to-patient chain." }, "animalEvidence": { "rating": "strong", "reason": "Model support exists but is held below randomized human outcomes." }, "inVitroEvidence": { "rating": "strong", "reason": "The proposed archive assay and product checks are measurable." }, "reproducibility": { "rating": "moderate", "reason": "Negative routes repeat; the positive CD40 route does not yet." }, "commercialConflicts": { "rating": "high", "reason": "Companies own the products, samples and most full records." }, "recency": { "rating": "current", "reason": "Uses trial and finance records through September 2026." } } }, "layout": { "x": 304, "y": 246 }, "updated": "2026-09-15" }, { "id": "trial-japan-ai-small-pdac-ct", "title": "A Japanese multicentre study will test CT AI on cancers 20 mm or smaller", "shortTitle": "Japan small-PDAC AI study", "type": "trial", "status": "open", "scope": [ "pdac" ], "summary": "The 400-record retrospective study includes carcinoma in situ and earlier scans, but cannot establish screening utility.", "content": "Kobe University registered a Fujifilm-funded multicentre retrospective study targeting 400 records. Eligible cancer cases were diagnosed from 2014 through 2024 with tumours 20 mm or smaller; a subgroup includes carcinoma in situ, high-grade PanIN or tumours 10 mm or smaller. The study compares the algorithm with pathology and expert radiologists and examines direct and indirect CT signs on current and earlier images. This is valuable non-US work on the hardest radiology target. Its case-control design, selected small-cancer archive and lack of a prospective clinical workflow mean that accuracy will not by itself establish population false-positive burden, lead time, treatment benefit or mortality reduction.", "sourceIds": [ "umin-2025-kobe-ai-early-pdac" ], "links": [ { "target": "technology-opportunistic-ct-ai-pancreas", "relation": "extends" }, { "target": "ai-programme", "relation": "supports" }, { "target": "detection", "relation": "part-of" }, { "target": "claim-case-control-blood-tests-overstate-screening-readiness", "relation": "shares-design-risk-with" } ], "evidence": { "humanProspective": "none", "humanRetrospective": "none — planned study, no result", "randomised": "none", "independentReplication": "none", "sampleSize": "Target 400 records", "effect": "No results yet", "limits": [ "Retrospective", "Enriched case-control archive", "Industry funded", "No patient-outcome endpoint", "No results posted" ] }, "layout": { "x": 148, "y": 60 }, "updated": "2026-09-14" }, { "id": "trial-lapc-conversion-surgery-action-map", "title": "Locally advanced PDAC surgery map: feasibility is ahead of causal proof", "shortTitle": "Conversion-surgery trial map", "type": "trial", "status": "supported", "scope": [ "pdac" ], "summary": "Modern studies show a route to resection and define its harm, but no current non-metastatic trial separates surgery from responder selection.", "content": "Treatment and reassessment. NEOLAP randomized 130 early non-progressors between two chemotherapy routes; 35.9% and 43.9% underwent complete visible removal, with no significant conversion or survival difference. Its CT subset shows why a static anatomy label fails after treatment: formal downstaging was 7.7%, yet any shrinkage was 90.4% and the imaging discriminator for complete-margin resection was only AUC 0.647. Its CA19-9 analysis supports serial blood measurement but derived a 61 U/mL cutoff from the same data and did not randomize the surgery decision. Feasibility and harm. PREOPANC-4 explored 180 selected Dutch patients after at least four months without scan progression; 155 were resected, 44% had major morbidity and 3% died within 90 days. The cohort validates a concentrated expert route, not survival benefit. Selection. TAPS found 171 resections among 663 selected locally advanced patients and a 36.1-month median survival, but the 20.2-month non-surgery group was not comparable. Its 2026 model predicts prognosis from the same fields used in treatment choice and explicitly rejects causal interpretation. A 312-person matched study reported another large surgery association, but route choice remained institutional or surgical. Direct tests. A 42-person 2004 trial favoured resection over old radiochemotherapy after open confirmation of technically resectable disease; it is too small and remote from modern induction care. CONKO-007 and NEOLAP randomize treatment before surgery, not surgery itself. PELICAN randomized a different invasive local action and found no survival benefit, more serious harm and worse quality of life. Live studies. NCT06714604 tests eight versus twelve FOLFIRINOX cycles or equivalent duration, with results estimated in 2032. NCT06132087 is single-arm with estimated enrolment two. Decision: offer every suitable responder expert reconsideration, concentrate any operation, and randomize only the point of real uncertainty. Do not count resection rate, a clean margin or selected survival as proof of useful life.", "sourceIds": [ "kunzmann-2021-neolap", "guggenberger-2023-neolap-ct", "hartlapp-2022-neolap-ca199", "stoop-2026-preopanc4-surgery", "theijse-2024-nontherapeutic-laparotomy", "dekker-2026-taps-restaging", "verkolf-2026-taps-resection-prediction", "napoli-2025-surgery-versus-oncology", "imamura-2004-locally-invasive-surgery-rct", "fietkau-2025-conko007", "seelen-2026-pelican", "clinicaltrials-nct06714604-duration", "clinicaltrials-nct06132087-proactive" ], "links": [ { "target": "trials", "relation": "part-of" }, { "target": "claim-lapc-response-warrants-surgical-reassessment", "relation": "supports" }, { "target": "unknown-lapc-surgery-causal-benefit", "relation": "leaves-open" }, { "target": "failure-conversion-surgery-selection-bias", "relation": "exposes" }, { "target": "hypothesis-randomized-lapc-resection-after-response", "relation": "motivates" }, { "target": "trial-pdac-occult-metastasis-action-map", "relation": "requires" }, { "target": "trial-pdac-surgical-route-action-map", "relation": "requires" } ], "evidence": { "humanProspective": "strong for feasibility and harm", "humanRetrospective": "strong", "randomised": "strong for adjacent treatment choices; absent for modern resection itself", "independentReplication": "moderate", "sampleSize": "NEOLAP randomized n=130; PREOPANC-4 explored n=180; TAPS n=663 and n=2,338; PELICAN n=188; old surgery trial n=42; live duration target n=432", "effect": "Reconsideration and technical feasibility supported; added survival effect of surgery unresolved", "limits": [ "Study populations overlap", "Definitions and treatment durations differ", "Only one old direct randomization", "Modern cohorts are selected", "PELICAN tests ablation rather than resection", "Live trials do not isolate surgery" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "Systemic treatment can reveal slower disease and shrink local tumour, but cannot prove absence of distant clones." }, "animalEvidence": { "rating": "not-applicable", "reason": "The decisive evidence is the human treatment route." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Cell assays cannot estimate surgical net benefit." }, "reproducibility": { "rating": "moderate", "reason": "Resection of a selected minority repeats; survival causality does not." }, "commercialConflicts": { "rating": "moderate", "reason": "Drug funding, centre expertise and procedural incentives differ across the mapped studies." }, "recency": { "rating": "current", "reason": "Includes 2026 trials, cohorts and live registry states." } } }, "layout": { "x": 114, "y": 180 }, "updated": "2026-09-15" }, { "id": "trial-linfu-action-map", "title": "LINFU registries can measure sampling yield but not causal patient benefit", "shortTitle": "LINFU study map", "type": "trial", "status": "open", "scope": [ "pdac", "pancreatic-all" ], "summary": "Two 500-person invitation-only observational registries are the existing route; neither randomizes the decision made after a positive result.", "content": "NCT04793633 combines low-intensity non-focused ultrasound, intravenous microbubbles and secretin to collect pancreatic cells in an estimated 500-person cohort with risk factors or possible disease. It started in January 2024, is enrolling by invitation, estimates completion in December 2030 and posts no results. NCT06276764 started in October 2024 and follows an estimated 500 people with IPMN through 2034, also without posted results. These studies can report specimen adequacy, cytology, pathology and later diagnoses. They cannot by themselves determine sensitivity unless negative participants receive adequate long follow-up, and they cannot assign whether LINFU-guided intervention reduces death or harm because there is no randomized action comparison. The investor's 4-of-90 claim must be reconciled to the registries: all enrolled people, every positive and negative call, reference standard, follow-up, complications, downstream endoscopy, biopsy and surgery, and whether the four lesions were inside the locked intended-use definition. Decision: launch no duplicate cohort. Offer only independent analysis after data access and a fixed statistical plan. A later randomized action test is justified only if at least 90% of participants have an adequate specimen and complete one-year negative follow-up, specificity against benign high-risk findings exceeds the locked work-up threshold, and positive results identify high-grade dysplasia or stage-I invasive disease that changes treatment. Stop if the procedure adds serious harm, produces excess low-risk operations, or results cannot be reproduced outside company-linked sites.", "sourceIds": [ "clinicaltrials-nct04793633-linfu", "clinicaltrials-nct06276764-linfu-ipmn", "igan-2026-adenocyte-financing" ], "links": [ { "target": "trials", "relation": "part-of" }, { "target": "company-adenocyte", "relation": "sponsored-by" }, { "target": "detection", "relation": "tests" }, { "target": "trial-pancreatic-precursor-action-map", "relation": "overlaps" } ], "evidence": { "humanProspective": "weak — registry protocols without posted results", "humanRetrospective": "none", "randomised": "none", "independentReplication": "none", "sampleSize": "Two observational registries each estimate n=500; sponsor reports 90 high-risk procedures", "effect": "No posted prospective accuracy, work-up-harm, stage-shift or mortality result", "limits": [ "Invitation-only selection", "Industry sponsorship", "No posted results", "Long estimated completion dates", "Reference standard and negative follow-up not publicly reported", "No randomized action comparison" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "A richer duct-cell sample could improve cytology, but sampling may still miss spatially separate lesions." }, "animalEvidence": { "rating": "weak", "reason": "Reported porcine development does not establish the human care path." }, "inVitroEvidence": { "rating": "weak", "reason": "Sample enrichment is an assay step, not a measure of downstream patient benefit." }, "reproducibility": { "rating": "none", "reason": "No independent site-level or published registry result is available." }, "commercialConflicts": { "rating": "high", "reason": "Adenocyte sponsors the studies and the checked performance report comes from its lead investor." }, "recency": { "rating": "current", "reason": "Registry status and financing were checked in September 2026." } } }, "layout": { "x": 145, "y": 88 }, "updated": "2026-09-15" }, { "id": "trial-mtap-prmt5-action-map", "title": "MTAP-PRMT5 action map: finish the denominator, then isolate contribution", "shortTitle": "MTAP-PRMT5 action map", "type": "trial", "status": "mixed", "scope": [ "pdac", "cholangiocarcinoma", "pan-cancer" ], "summary": "Existing trials establish a selected human route; the next useful step is mature all-treated reporting and a randomized RAS-drug contribution test.", "content": "Selection stage. Confirm homozygous MTAP loss directly in tumour tissue by genomic testing and MTAP protein staining where possible. Do not substitute CDKN2A loss: one large series found that 33.7% of CDKN2A-loss tumours retain MTAP. Do not use a negative liquid biopsy when tumour fraction is low. Record full versus partial exon loss because partial-loss sensitivity is unresolved. Mechanism stage. AMG 193 already showed complete tumour PRMT5 inhibition in paired biopsies at active doses. NCT05732831 should report paired tumour inhibition and exposure by deletion structure for vopimetostat, not only response. Maturity stage. For NCT06922591, report every treated participant when each reaches 14 weeks and again at 6, 12 and 18 months. Preserve the denominators 20 daraxonrasib-combination and 34 zoldonrasib-combination rather than only the current 12 and 27 evaluable sets. Publish confirmed response, duration, progression, death, toxicity and dose intensity. Contribution stage. Within MTAP-deleted RAS-mutant PDAC, randomize the same RAS inhibitor alone versus RAS inhibitor plus vopimetostat. A front-line package-versus-chemotherapy trial can test package value, but a second-line daraxonrasib-alone control is needed to measure vopimetostat's added value because daraxonrasib is now an active treatment. Lock MTAP assay, KRAS allele, line and liver-metastasis strata. Use blinded independent imaging, PFS, OS, quality-adjusted survival and full all-randomized accounting. Current registry arms are nonrandomized, including the added chemotherapy combinations. Capital. Mission investment is £0. Tango reported $1.0bn of cash and marketable securities at June 2026, compared with six-month operating cash use of $83.226m; 1000/(83.226/6)=72.1 months under a simple unchanged-burn calculation, far above the next readout. The company and partners can fund the proof. Mission leverage is public data, exact comparator design and refusal to let a package trial hide component value.", "sourceIds": [ "pavlick-2026-mtap-genomics", "rodon-2024-amg193-prmt5", "clinicaltrials-nct05732831-vopimetostat", "tango-2025-vopimetostat-monotherapy", "clinicaltrials-nct06922591-vopimetostat-ras", "tango-2026-vopimetostat-ras-data", "tango-2026-q2-10q" ], "links": [ { "target": "trials", "relation": "part-of" }, { "target": "claim-mtap-prmt5-selected-human-signal", "relation": "maps" }, { "target": "failure-vopimetostat-ras-contribution-unknown", "relation": "repairs" }, { "target": "unknown-mtap-prmt5-ras-contribution", "relation": "maps" }, { "target": "hypothesis-mtap-prmt5-ras-deeper-control", "relation": "gates" }, { "target": "company-tango-therapeutics", "relation": "run-by" }, { "target": "claim-daraxonrasib-common-ras-survival-breakthrough", "relation": "requires-as-comparator" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "strong", "randomised": "none", "independentReplication": "moderate", "sampleSize": "Uses existing trials: vopimetostat registry estimated n=225 and combination registry estimated n=183", "effect": "Generated selection, maturity and contribution gates; no new mission-funded cohort", "limits": [ "Future randomized protocol is not yet public", "Simple burn calculation is not company guidance", "Sponsor controls drugs and current data", "No participant-level data" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The trial route begins with a measurable deletion-created dependency." }, "animalEvidence": { "rating": "moderate", "reason": "Model evidence supports the route but is not a release gate." }, "inVitroEvidence": { "rating": "strong", "reason": "Exact deletion structure, drug exposure and PRMT5 inhibition can be tested." }, "reproducibility": { "rating": "weak", "reason": "Human target inhibition repeats across a second molecule, while the PDAC combination result does not." }, "commercialConflicts": { "rating": "high", "reason": "Tango and partners control the drugs, current data and future protocol." }, "recency": { "rating": "current", "reason": "Uses current registries and June 2026 company data." } } }, "layout": { "x": 190, "y": 111 }, "updated": "2026-09-15" }, { "id": "trial-new-onset-diabetes-workup-map", "title": "New-onset-diabetes studies show risk and feasibility, not a proven life-saving work-up", "shortTitle": "Diabetes work-up studies", "type": "trial", "status": "open", "scope": [ "pancreatic-all", "pdac" ], "summary": "A prospective risk cohort, two small imaging studies and the unfinished 8,869-person EDI programme leave stage shift, incidental-work-up burden and survival benefit unresolved.", "content": "The studies answer different questions. Prospective health-record surveillance of 18,838 adults aged at least 50 established a race-adjusted three-year pancreatic-cancer incidence of 0.62% after glycaemically defined new-onset diabetes. That is about 14.9 times the rough three-year general-population risk obtained from 3 × 13.9 per 100,000, but it does not show that work-up helps. A 93-scan prospective CT pilot found one PDAC and it was stage IV. Twelve participants had other pancreatic findings and 52 of 93, 56%, had extrapancreatic findings; 12 of 57 extrapancreatic findings prompted clinical evaluation. PANDOME screened 625 people for eligibility and enrolled 109 for MRI/MRCP. Four had pancreatic biopsy and one stage-I PDAC was found, in the 12-person deteriorating-diabetes subgroup. One cancer cannot estimate sensitivity, interval cancers or net benefit. EDI is the decisive active programme: 8,869 people enrolled, active but not recruiting, non-randomized parallel sites, no posted results and estimated completion in January 2028. Intervention sites calculate ENDPAC and invite people with a score above 0 for up to two CT or MRI studies; observation sites use health-record follow-up. This is not the score-at-least-3 threshold reported in the original END-PAC validation, and site allocation can mix pathway effects with health-system differences. The registry explicitly measures incidental findings and stage, but not a randomized mortality endpoint. Workload model, not an observed result: if the 0.62% three-year risk and the original 78% sensitivity and 85% specificity transported to 100,000 new-onset-diabetes patients, the rule would flag about 484 cancer-bearing people and 14,907 unaffected people: PPV 3.14% and 30.8 unaffected positives per cancer-bearing positive. If a later independent test achieved 90% sensitivity and 99% specificity inside that selected group, it would retain about 435 cancer-bearing people and 149 unaffected people, PPV 74.5%. Conditional performance at that level has not been shown, and correlated errors could make the calculation much worse. Decision: obtain complete EDI and PANDOME operating data before launching a 40,000–60,000-person duplicate. A new trial must test a distinct locked second-stage selector, individual or cluster randomization, complete negative follow-up, stage I/II aggressive-cancer yield, incidental procedures, quality of life, cost and pancreatic-cancer death. Stop if stage shift disappears after accounting for site differences, interval cancers erase sensitivity, or serious work-up harm per additional treatable cancer exceeds the locked net-benefit limit.", "sourceIds": [ "chari-2026-prospective-new-onset-diabetes", "sharma-2018-endpac", "wu-2022-nod-imaging-pilot", "frank-2025-pandome", "edi-nct04662879", "seer-pancreas-statfacts" ], "links": [ { "target": "trials", "relation": "part-of" }, { "target": "detection", "relation": "tests" }, { "target": "claim-new-onset-diabetes-enriches-risk", "relation": "extends" }, { "target": "hypothesis-one-time-risk-triggered-workup", "relation": "constrains" }, { "target": "programme-500m-integrated-mission", "relation": "must-not-duplicate" } ], "evidence": { "humanProspective": "moderate — risk cohort and small imaging studies", "humanRetrospective": "moderate — original END-PAC validation", "randomised": "none", "independentReplication": "weak", "sampleSize": "Risk cohort n=18,838 with 82 cancers; CT pilot n=93; PANDOME n=109 with one stage-I cancer; EDI n=8,869 with results pending", "effect": "Three-year risk 0.62%; one stage-IV cancer in CT pilot; one stage-I cancer and four biopsies in PANDOME; no completed utility effect", "limits": [ "No randomized utility or mortality result", "Very few cancers in completed imaging studies", "CT pilot found substantial incidental abnormalities", "PANDOME enrolled a selected minority", "EDI assigns pathways by site and uses ENDPAC above 0", "Race and health-system transportability", "The workload model assumes performance transports and does not model correlated errors" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Occult pancreatic cancer can cause changing glucose and weight before clinical diagnosis, creating a time-limited risk signal." }, "animalEvidence": { "rating": "not-applicable", "reason": "Only a human care-pathway comparison can establish stage shift, work-up harm and survival benefit." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Laboratory detection cannot measure health-system uptake, incidental procedures or patient outcomes." }, "reproducibility": { "rating": "weak", "reason": "Risk enrichment repeats, but the completed imaging studies contain only one pancreatic cancer each and use different entry rules." }, "commercialConflicts": { "rating": "low", "reason": "The cited studies are mainly academic and charity or public funded, though future blood-test selection could involve commercial assays." }, "recency": { "rating": "current", "reason": "The map includes a 2025 prospective report and EDI status verified in September 2026." } } }, "layout": { "x": 101, "y": 73 }, "updated": "2026-09-14" }, { "id": "trial-older-pdac-action-map", "title": "Older and vulnerable PDAC action map: stop comparing birthdays and finish the care route", "shortTitle": "Older-PDAC action map", "type": "trial", "status": "supported", "scope": [ "pdac" ], "summary": "Existing trials test drugs, assessment and support as separate objects; the next non-duplicate test connects assessment to fast action while retaining every patient in the route.", "content": "Drug-choice evidence. GIANT EA2186 randomized 176 vulnerable adults aged 70-90 across 92 US sites. Fortnightly reduced-dose gemcitabine plus nab-paclitaxel and fluorouracil plus liposomal irinotecan produced median survival of 4.7 and 4.4 months, P=.72; severe toxicity was 45.6% and 58.7%, P=.10. The study stopped for futility. This does not prove the regimens equivalent; it rejects the planned large survival difference. The current registry is active, not recruiting, with actual primary completion in March 2026 and final completion estimated December 2026. Single-route evidence. PAMELA-70 enrolled 72 and treated 69 adults aged at least 70 with adjusted FOLFIRINOX. Results appeared in April 2026, more than five years after completion, but report only one combined safety-and-response number, 65.8%, 95% CI 53.2-75.8. Serious adverse events affected 48/69, 69.6%. Without separate response, daily independence and survival, it cannot supply a treatment rule. Assessment-allocation evidence. GrantPax planned 135, enrolled 32 and assigned 15/32 inconsistently with its assessment; its nonrandom groups cannot compare drugs. Action evidence. GAP70+ randomized 40 practice clusters and 718 older adults with incurable mixed cancers; assessment plus recommendations reduced grade 3-5 toxicity from 71% to 51%, a 20-point difference, and falls from 21% to 12%. GAIN randomized 605 analyzable mixed-cancer patients; multidisciplinary action reduced severe toxicity from 60.6% to 50.5%, 10.1 points, without a survival gain. A 2026 randomized trial reduced severe toxicity from 30.8% to 15.5% but missed independence and quality-of-life primary outcomes. Live gap. NCT06040801 still says recruiting 138 people at one Taiwan site after its July 2026 estimated completion; it mixes gastric, biliary and pancreatic cancer and allocates intervention by baseline frailty without randomization, so it cannot estimate the action effect. Next move: obtain complete GIANT treatment-start, early-discontinuation, function, toxicity, hospital, quality-of-life and survival data; require full PAMELA outcomes; refresh the Taiwan registry; then run the drug-neutral assessment-to-action trial. Do not launch another two-regimen older-age comparison. The experiment must include a consecutive pre-decision denominator, state who chose no treatment and why, and test patient-valued outcomes as well as toxicity.", "sourceIds": [ "dotan-2024-giant", "dotan-2025-giant-ga", "clinicaltrials-nct04233866-giant", "betge-2022-grantpax", "clinicaltrials-nct02143219-pamela70", "mohile-2021-gap70", "li-2021-gain", "ihorst-2026-geriatric-comanagement", "clinicaltrials-nct06040801-frail-gi" ], "links": [ { "target": "trials", "relation": "part-of" }, { "target": "claim-age-alone-cannot-route-pdac-treatment", "relation": "tests" }, { "target": "failure-grantpax-geriatric-route", "relation": "includes" }, { "target": "unknown-older-pdac-treatment-route", "relation": "tests" }, { "target": "hypothesis-older-pdac-fast-support-route", "relation": "prioritizes" }, { "target": "programme-500m-integrated-mission", "relation": "constrains" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong but split across questions", "independentReplication": "moderate", "sampleSize": "GIANT n=176; PAMELA-70 n=72; GrantPax n=32; GAP70+ n=718; GAIN n=605; 2026 co-management n=217; live mixed-GI target n=138", "effect": "GIANT no regimen separation; replicated mixed-cancer severe-toxicity reductions of 10.1 to 20.0 points", "limits": [ "GIANT abstract reporting", "PAMELA incomplete results", "GrantPax allocation failure", "Mixed-cancer support evidence", "Taiwan study nonrandomized", "No PDAC survival-improving care rule" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Treatment benefit depends on disease control minus toxicity, delay, lost function and hospital burden." }, "animalEvidence": { "rating": "not-applicable", "reason": "Human comparative care data are decisive." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Laboratory response cannot measure function, preference or care delivery." }, "reproducibility": { "rating": "moderate", "reason": "Care action reduces toxicity in several randomized studies, but PDAC-specific regimen and survival results remain incomplete." }, "commercialConflicts": { "rating": "moderate", "reason": "Several studies use branded medicines; the proposed route is drug-neutral and demands independent analysis." }, "recency": { "rating": "current", "reason": "Registry states and the newest posted results were checked in September 2026." } } }, "updated": "2026-09-14" }, { "id": "trial-pancreatic-precursor-action-map", "title": "Pancreatic precursor trials have not yet shown that surveillance or interception prevents death", "shortTitle": "Precursor action trials", "type": "trial", "status": "open", "scope": [ "pdac" ], "summary": "EA2185 lost its full randomized surveillance comparison, SIPP-T3 accrued four people, 3P-C awaits results on a mixed endpoint, and ChARM shows cyst shrinkage in only 52 selected completers.", "content": "Four programmes cover different parts of the action path and none establishes cancer-mortality benefit. EA2185 originally planned 4,606 randomized patients to compare high- and low-intensity surveillance. Its current record is an estimated 770-person observational biomarker cohort; the randomized arms closed in August 2024 and no results are posted. 3P-C is the strongest completed-accrual prevention test visible: an estimated 100 high-risk or post-resection IPMN patients randomized to sulindac or placebo, primary follow-up completed in July 2026, no results posted as of 14 September 2026. Its primary composite mixes new cysts, cyst doubling, duct enlargement, resection and PDAC; even a positive result will need component outcomes and toxicity before it can support prevention. SIPP-T3 randomized tocotrienol against placebo but stopped recruiting with four actual participants versus 212 planned and cannot answer efficacy. ChARM chemoablation offers proof that some cysts can be locally treated without pancreatectomy: among 52 selected completers, 36 had complete and 11 partial volume response, with no cyst-associated malignancy observed. It does not prove prevention because follow-up selected completers, the sample was tiny for cancer, the outcome was mainly volume response and ablation would not address separate PDAC elsewhere. Decision: first acquire EA2185 amendment and randomized-subset data, then obtain 3P-C component and safety results. Do not launch another drug or ablation study around cyst size. A valid next trial needs a fixed HGD/invasive-cancer target, full adverse-event and pancreatic-function capture, the separate-PDAC route, and enough events to compare net patient benefit.", "sourceIds": [ "weinberg-2020-ea2185-protocol", "clinicaltrials-nct04239573-ea2185", "clinicaltrials-nct04207944-3pc", "clinicaltrials-nct06519097-sipp-t3", "moyer-2024-charm-followup", "ohtsuka-2024-bd-ipmn-surveillance" ], "links": [ { "target": "trials", "relation": "part-of" }, { "target": "failure-ea2185-randomized-question-closed", "relation": "contains" }, { "target": "failure-sipp-t3-accrual-collapse", "relation": "contains" }, { "target": "claim-ipmn-has-two-cancer-routes", "relation": "depends-on" }, { "target": "hypothesis-dynamic-cyst-interception-rule", "relation": "constrains" }, { "target": "programme-100m-randomized-platform", "relation": "constrains" } ], "evidence": { "humanProspective": "moderate — multiple prospective intervention or surveillance programmes", "humanRetrospective": "not-applicable", "randomised": "moderate in design, weak in available results", "independentReplication": "none for prevention utility", "sampleSize": "EA2185 current estimate 770 observational; 3P-C estimate 100; SIPP-T3 actual 4; ChARM follow-up 52 selected completers", "effect": "No posted HGD, invasive-cancer or mortality benefit from EA2185, 3P-C or SIPP-T3; ChARM 69.2% complete volume response without a powered cancer endpoint", "limits": [ "Registry data can change", "No 3P-C component results yet", "EA2185 amendment reason unknown", "SIPP-T3 is grossly under-enrolled", "ChARM selected completers and cannot prevent separate PDAC" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "Treating a true precursor could prevent its invasive descendant, but a local cyst action cannot remove the separate-PDAC route." }, "animalEvidence": { "rating": "weak", "reason": "Drug ideas have model support, while surveillance intensity and surgery require human evidence." }, "inVitroEvidence": { "rating": "weak", "reason": "Laboratory activity exists for some drugs but is not a clinical success condition." }, "reproducibility": { "rating": "none", "reason": "No intervention has an independently repeated cancer-prevention result." }, "commercialConflicts": { "rating": "moderate", "reason": "Most programmes are academic or public, but SIPP-T3 includes an ingredient-company collaborator and ablation expertise is concentrated." }, "recency": { "rating": "current", "reason": "Trial records were checked in September 2026 and include actual 2026 completion or update dates." } } }, "layout": { "x": 102, "y": 74 }, "updated": "2026-09-14" }, { "id": "trial-pdac-adenosine-action-map", "title": "PDAC adenosine action map: let the existing masked survival trial decide", "shortTitle": "Adenosine action map", "type": "trial", "status": "active", "scope": [ "pdac", "transfer-unproven" ], "summary": "ARC-8 establishes mechanism and feasibility, oleclumab supplies a negative class control, and PRISM-1 already holds the direct 610-person survival test.", "content": "Route 1, CD73 small-molecule blockade. ARC-8 supports tumour-pathway engagement but cannot estimate quemliclustat contribution because all randomized patients received it. Its 122-versus-122 survival comparison used a matched external control after the trial and disagreed with response and PFS. Do not use it for treatment or investment return estimates. Route 2, CD73 antibody blockade. In 170 randomized patients, oleclumab plus chemotherapy missed response and produced OS HR 1.26; the oleclumab, durvalumab and chemotherapy arm had HR 0.75, 95% CI 0.50-1.13. The favourable CD73-high subgroup is not a validated selector. Route 3, direct confirmation. PRISM-1 uses the clean comparison: quemliclustat plus gemcitabine/nab-paclitaxel versus placebo plus the same chemotherapy. It is quadruple masked, estimates 610 participants, uses all-randomized survival as the main endpoint and reports enrollment completed in September 2025. Results are expected in the first half of 2027. Obtain the protocol and analysis plan before readout; at release reconcile the estimated registry count with the final assigned count and audit later treatment and harm. Route 4, marker development. Do not open a selected cohort. Baseline NR4A did not repeat in two external datasets and treatment-linked NR4A change is measured too late to select therapy. Decision rule. Spend £0 on a new trial and £0 on Arcus equity now. The company reported $775m cash and investments and partner support, so mission subsidy would not accelerate the answer. Re-rank only after the masked survival result; conditional independent analysis may receive at most £250,000 if participant-level access is granted.", "sourceIds": [ "wainberg-2026-quemliclustat-arc8", "coveler-2024-oleclumab-pdac", "clinicaltrials-nct06608927-prism1", "arcus-2026-q2-10q" ], "links": [ { "target": "trials", "relation": "part-of" }, { "target": "claim-cd73-blockade-signal-without-concurrent-control", "relation": "maps" }, { "target": "failure-adenosine-control-and-marker-pdac", "relation": "maps" }, { "target": "unknown-quemliclustat-survival-contribution", "relation": "resolves" }, { "target": "hypothesis-quemliclustat-existing-phase3-gate", "relation": "implements" }, { "target": "company-arcus-biosciences", "relation": "involves" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "moderate", "randomised": "strong", "independentReplication": "moderate", "sampleSize": "ARC-8 n=138 across escalation and expansion; oleclumab trial n=195 total and n=170 randomized expansion; PRISM-1 estimated n=610", "effect": "No proved added benefit; existing phase 3 owns the answer and requires no mission trial capital", "limits": [ "PRISM-1 has no result", "Final assigned count and analysis plan are not public", "Registry and sponsor give different completion horizons", "Cross-product class inference is limited" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The map joins enzyme inhibition, tumour state and patient endpoints." }, "animalEvidence": { "rating": "strong", "reason": "Models support the route but do not trigger another cohort." }, "inVitroEvidence": { "rating": "strong", "reason": "Adenosine-linked gene expression is directly drug-responsive." }, "reproducibility": { "rating": "weak", "reason": "The only direct class control was negative overall and the named drug awaits phase 3." }, "commercialConflicts": { "rating": "high", "reason": "Arcus, Gilead and Taiho hold programme interests and Arcus controls the first readout." }, "recency": { "rating": "current", "reason": "The map uses the August 2026 company filing and current trial registry." } } }, "layout": { "x": 468, "y": 396 }, "updated": "2026-09-15" }, { "id": "trial-pdac-asp1929-action-map", "title": "PDAC ASP-1929 action map: instrument the first ten before the trial starts", "shortTitle": "ASP-1929 action map", "type": "trial", "status": "active", "scope": [ "pdac", "photoimmunotherapy", "local-treatment", "capital-allocation" ], "summary": "The first pancreatic trial has not recruited, leaving a narrow chance to add binding, light, safety and treatment-delay gates before irreversible spending and patient exposure.", "content": "Route 1, product. ASP-1929 joins an EGFR antibody to IR700 dye; local 690 nm light activates membrane damage. Route 2, transfer boundary. Japan conditionally approved the system for unresectable locally advanced or recurrent head and neck cancer with all-patient surveillance. That is operational evidence, not pancreatic evidence. Route 3, target. One tissue study found EGFR overexpression in 52/81 pancreatic cancers. Yet a 745-person trial found no benefit from cetuximab despite 90% EGFR positivity, so a stain is an address, not proof of dependence. Route 4, light. Outside light reaches about 2 cm in mouse tissue; deep pancreatic treatment needs inserted fibres and measured geometry. Route 5, trial. Thirty people are planned. All receive drug and surgery; only those found inoperable receive light. The first ten test safety and the expansion leads with six-month progression. There is no control. Route 6, capital. Buy £0 of equity, drug or devices for expansion. Offer at most £300,000 for independent monitoring and spatial measurement in the first ten, payable after raw release. Route 7, stop. Require at least 8/10 planned illuminations, 80% target binding-plus-light coverage, no more than one severe related event, no lost resection and no more than 14 extra days of systemic-treatment delay. Review after each of the first three and at ten. Route 8, interpretation. Passing permits descriptive completion, not an efficacy claim. Any later benefit trial must keep systemic care identical and randomize illumination as the only difference. Route 9, return. The near-term asset is a measured human pancreatic delivery chain and an honest stop decision, not cross-cancer approval or platform ownership.", "sourceIds": [ "clinicaltrials-nct07698613-asp1929-pancreas", "pmda-2020-akalux-review", "rakuten-medical-2026-asp1929", "kim-2015-egfr-expression-pdac", "philip-2010-swog-s0205-cetuximab", "maruoka-2018-interstitial-photoimmunotherapy" ], "links": [ { "target": "trials", "relation": "part-of" }, { "target": "claim-asp1929-pancreatic-human-test-not-started", "relation": "maps" }, { "target": "failure-asp1929-selection-light-and-endpoint-design", "relation": "maps" }, { "target": "unknown-asp1929-pancreatic-target-coverage-and-benefit", "relation": "resolves" }, { "target": "hypothesis-asp1929-first-ten-measurement-gate", "relation": "implements" }, { "target": "company-rakuten-medical", "relation": "involves" }, { "target": "trial-pdac-pdt-action-map", "relation": "compares-with" }, { "target": "trial-pdac-laser-ablation-action-map", "relation": "compares-with" } ], "evidence": { "humanProspective": "none — planned first-human pancreatic study", "humanRetrospective": "none in pancreatic cancer", "randomised": "none", "independentReplication": "none", "sampleSize": "First-ten safety gate within planned n=30; zero pancreatic outcomes now; cap £300,000", "effect": "No pancreatic benefit signal; unusually early chance to repair measurement before enrollment", "limits": [ "Sponsor amendment not agreed", "Trial may start before intervention", "Ten-person safety estimate is imprecise", "No causal efficacy result", "Cross-cancer data may not transfer", "Light and drug maps may be technically incomplete" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "A targetable surface marker and local switch exist, but human pancreatic overlap is untested." }, "animalEvidence": { "rating": "moderate", "reason": "Inserted-light studies explain depth constraints." }, "inVitroEvidence": { "rating": "moderate", "reason": "Cell killing supports first-human measurement, not benefit." }, "reproducibility": { "rating": "none", "reason": "No pancreatic clinical result exists." }, "commercialConflicts": { "rating": "high", "reason": "The programme is tied to one company's drug-device system." }, "recency": { "rating": "current", "reason": "The action map uses the July 2026 registry state." } } }, "layout": { "x": 1030, "y": 954 }, "updated": "2026-09-15" }, { "id": "trial-pdac-autophagy-action-map", "title": "PDAC autophagy action map: recover direct-inhibitor data and stop HCQ combinations", "shortTitle": "Autophagy action map", "type": "trial", "status": "mixed", "scope": [ "pdac", "cholangiocarcinoma", "pan-cancer", "transfer-unproven" ], "summary": "Existing trials answer the HCQ question and leave one missing dataset; another unselected combination trial would repeat failure without measuring the target in tumour.", "content": "Completed negative contribution test. NCT04386057 directly compared ERK inhibition plus HCQ with ERK inhibition alone after one or two earlier treatment lines. Its desired four-month disease-control rate was 20%, with 5% declared unacceptable and at least three controlled patients among 20 needed to reject that null. The combination produced one controlled patient among 20. It also used half the ERK-inhibitor dose and found neither clear synergy in eleven tested organoids nor consistent autophagy-marker change in blood cells. Independent failed routes. NCT05221320 enrolled all 18 planned pancreatic patients in stage one, found zero responses and did not expand. NCT04214418 treated 14, found limited tolerability and efficacy, and never opened its pancreatic phase 2 cohort. NCT04524702 reported three responses among seven evaluable patients after giving HCQ, paricalcitol and two active chemotherapy drugs to ten people, but no control and missing response and survival denominators prevent attribution. Closed information gap. NCT04892017 enrolled 91 people with the direct ULK1/2 inhibitor inlexisertib and ended for a stated business decision, not a stated safety or efficacy result. Its early sponsor report covered 18: target inhibition was measured in blood cells, best response was stable disease, and the later combination and pancreatic records remain unreported. Action now. Freeze new HCQ combinations. Request the NCT04892017 protocol history, analysis plan, participant flow, cancer-specific dose, tumour and blood pharmacology, combination exposure, scans, response and toxicity. Publish why the planned pancreatic expansion disappeared. Put missing and unevaluable people in every denominator. If there was no paired tumour measurement or no pancreatic combination cohort, say so; do not fill the gap with a new efficacy cohort. New human exposure is allowed only after an independently reproduced drug shows tumour target inhibition and recycling blockade at a tolerated dose, preserves the RAS drug's dose, and shows the predicted combination effect in patient-derived models. The later clinical test must randomize the added inhibitor against the same RAS treatment. Overall and quality-adjusted survival decide utility; target engagement only earns the right to test it. No new money is released now.", "sourceIds": [ "surana-2025-erk-hcq-pdac", "clinicaltrials-nct05221320-ulixertinib-hcq", "manji-2023-mekiauto", "clinicaltrials-nct04214418-mekiauto", "clinicaltrials-nct04524702-paricalcitol-hcq", "clinicaltrials-nct04892017-inlexisertib", "deciphera-2022-dcc3116-phase1", "alistar-2017-devimistat-phase1", "philip-2024-avenger500" ], "links": [ { "target": "trials", "relation": "part-of" }, { "target": "claim-hcq-combinations-fail-pdac-not-autophagy", "relation": "supports" }, { "target": "failure-metabolic-repurposing-pdac", "relation": "repairs" }, { "target": "unknown-selective-autophagy-pdac-contribution", "relation": "maps" }, { "target": "hypothesis-tumour-verified-selective-autophagy-blockade", "relation": "gates" }, { "target": "claim-pdac-metabolic-adaptation-is-redundant", "relation": "tests-boundary-of" } ], "evidence": { "humanProspective": "strong for failure of current combinations; weak for selective autophagy inhibition", "humanRetrospective": "none needed", "randomised": "moderate — small direct contribution test", "independentReplication": "moderate for HCQ-combination failure; none for selective tumour-confirmed inhibition", "sampleSize": "NCT04386057 n=52 enrolled with 39 randomized and treated; NCT05221320 pancreatic n=18; NCT04214418 treated n=14; NCT04524702 n=10; NCT04892017 n=91 without final results", "effect": "No HCQ combination crossed its useful-control gate; selective-inhibitor effect is unknown", "limits": [ "Small, mostly advanced-disease groups", "Different ERK and MEK drugs cannot be treated as identical", "Missing final inlexisertib data", "Single-arm multi-drug response cannot identify contribution", "Future RAS inhibitors may change the combination context" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "Adaptive recycling after RAS-pathway inhibition remains plausible, but the tested patient programmes did not establish tumour blockade." }, "animalEvidence": { "rating": "moderate", "reason": "Model synergy is sufficient for a measurement experiment, not another unmeasured patient expansion." }, "inVitroEvidence": { "rating": "weak", "reason": "Trial-derived organoids did not reproduce the expected HCQ combination effect." }, "reproducibility": { "rating": "moderate", "reason": "The negative HCQ direction repeats; a direct-inhibitor result does not." }, "commercialConflicts": { "rating": "high", "reason": "The missing direct-inhibitor dataset is sponsor-controlled, while drug replacement creates incentives to treat a molecule change as a reset." }, "recency": { "rating": "current", "reason": "Includes 2026 termination and current posted registry results." } } }, "layout": { "x": 102, "y": 158 }, "updated": "2026-09-15" }, { "id": "trial-pdac-biliary-drainage-action-map", "title": "PDAC biliary-drainage action map: choose the indication before the device", "shortTitle": "Biliary action map", "type": "trial", "status": "supported", "scope": [ "pdac", "cholangiocarcinoma", "transfer-unproven" ], "summary": "Completed randomized evidence rejects routine delayed plastic drainage, supports durable metal stents when drainage is needed, and leaves whole-route delivery as the next test.", "content": "Immediate-surgery boundary: ISRCTN31939699 randomized 202 patients and found serious complications of 39% with surgery within one week versus 74% after four to six weeks of mainly plastic drainage. The 2026 NCT01774019 report randomized 284 to metal-stent drainage or surgery within four weeks and found 29.0% versus 26.5% serious adverse events, meeting a wide 15-point noninferiority margin; it supports an option during real delay, not routine drainage. Neoadjuvant device boundary: UMIN000030473 randomized 22 and found dysfunction 18.2% with covered metal versus 72.8% with plastic. A review found lower reintervention, treatment delay, obstruction and cholangitis with metal, but most source studies were retrospective. Metal design boundary: NCT02238847 randomized 119 and found sustained drainage 72.2% covered versus 72.9% uncovered, with migration and ingrowth trading places. Access-route boundary: ELEMENT, NCT03870386, randomized 144 mixed malignant obstructions and found stent dysfunction 9.6% versus 9.9% for EUS-guided versus ERCP metal drainage. Six randomized studies pooled across malignant obstruction found similar success and survival, less reintervention and pancreatitis with EUS, but not a PDAC-specific survival result. Do not fund another broad metal-versus-plastic or EUS-versus-ERCP device trial. Test the whole route in consecutive PDAC patients: indication, 48-hour plan, durable device when needed, named rescue, infection response, treatment dates, reintervention, surgery, serious harm and survival. Preserve resectable, neoadjuvant and palliative states separately.", "sourceIds": [ "van-der-gaag-2010-pbd-rct", "costamagna-2026-sems-early-surgery-rct", "tamura-2021-fcsems-plastic-rct", "seo-2019-covered-uncovered-sems-rct", "lyu-2023-metal-plastic-nat-meta", "vehvilainen-2024-cholangitis-nat-interruption", "chen-2023-element-biliary-rct", "paik-2018-eus-ercp-biliary-rct", "barbosa-2024-eus-ercp-meta", "esge-2018-biliary-stenting-guideline", "asge-2024-therapeutic-eus-guideline" ], "links": [ { "target": "trials", "relation": "part-of" }, { "target": "claim-biliary-drainage-is-route-not-cancer-treatment", "relation": "uses" }, { "target": "unknown-pdac-biliary-drainage-route", "relation": "tests" }, { "target": "failure-routine-plastic-pbd", "relation": "learns-from" }, { "target": "failure-biliary-stent-treatment-interruption", "relation": "learns-from" }, { "target": "hypothesis-pdac-biliary-route-completion", "relation": "prioritizes" }, { "target": "company-boston-scientific-biliary", "relation": "includes-trial-by" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "moderate", "randomised": "strong", "independentReplication": "moderate", "sampleSize": "Randomized component trials n=202, 284, 119, 22 and 144; EUS-ERCP pool n=577", "effect": "Decision-oriented map; effects remain in linked records", "limits": [ "No full-route trial", "Mixed cancers", "Different treatment states", "Old plastic trial", "Expert EUS centres", "Manufacturer links" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Every mapped step has a direct path to infection, reintervention or treatment delay." }, "animalEvidence": { "rating": "not-applicable", "reason": "Human procedural trials define the action map." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Bench testing cannot settle clinical route selection." }, "reproducibility": { "rating": "moderate", "reason": "Component effects repeat, but no one study tests the complete route." }, "commercialConflicts": { "rating": "high", "reason": "Several pivotal records involve named stents and manufacturer authors or support." }, "recency": { "rating": "current", "reason": "The map includes the 2026 metal-stent versus surgery trial." } } }, "updated": "2026-09-14" }, { "id": "trial-pdac-cachexia-action-map", "title": "PDAC cachexia action map: use the live 982-person trial before designing another large study", "shortTitle": "Cachexia action map", "type": "trial", "status": "hypothesis", "scope": [ "pdac", "transfer-unproven" ], "summary": "Existing randomized evidence separates three actions: cheap appetite support now, failed toxic IL-6 blockade, and a large GDF-15 test whose primary endpoints still stop short of survival.", "content": "Question: what should be done now rather than launching another cachexia trial? First, routine supportive care can test low-dose olanzapine where clinically appropriate because a randomized 124-person Indian study found greater than 5% weight gain in 60% versus 9% and better appetite, quality of life and chemotherapy toxicity. Only 13 participants had hepatopancreaticobiliary cancer and survival was not measured, so this is supportive evidence, not PDAC control. Second, do not repeat broad IL-6 blockade with the same dose and population: the PDAC phase 2 missed six-month survival and raised grade 3 or worse treatment-related toxicity by 24.7 points. Third, monitor and seek data access for NCT06989437. It randomizes an estimated 982 metastatic-PDAC patients at 220 locations in 21 countries to ponsegromab 200 mg, 400 mg or placebo with first-line chemotherapy. Primary endpoints are week-12 weight and anorexia score. Overall survival, PFS, response, function and chemotherapy dose changes are secondary. The design excludes people with ECOG above 1, life expectancy below four months and those unable to complete the first chemotherapy cycle, so it cannot answer the frailest-patient problem. The mission should publish a prospective endpoint hierarchy before results: safety; function and symptoms; chemotherapy dose and completion; quality-adjusted survival; OS. It should obtain arm-level PDAC data, GDF-15 values, missing wearable data and reasons for chemotherapy changes. Do not duplicate the broad trial unless the sponsor study fails, denies usable data or leaves a named frail, adjuvant or low-resource state uncovered. No result is expected before the estimated January 2028 primary completion.", "sourceIds": [ "sandhya-2023-olanzapine-anorexia", "chen-2025-tocilizumab-cachexia", "groarke-2024-ponsegromab-cachexia", "clinicaltrials-nct06989437-ponsegromab-pdac", "pfizer-2026-ponsegromab-profile" ], "links": [ { "target": "trials", "relation": "part-of" }, { "target": "unknown-cachexia-treatment-utility", "relation": "tests" }, { "target": "claim-ponsegromab-gdf15-cachexia-driver-not-survival", "relation": "tests" }, { "target": "failure-tocilizumab-cachexia-muscle-without-survival", "relation": "learns-from" }, { "target": "company-pfizer-cachexia", "relation": "run-by" }, { "target": "trial-pert-pdac-action-map", "relation": "complements" }, { "target": "unknown-pert-pdac-utility", "relation": "separates-from" }, { "target": "programme-500m-integrated-mission", "relation": "must-not-duplicate" } ], "evidence": { "humanProspective": "strong for existing trials; none for this mission action rule", "humanRetrospective": "moderate", "randomised": "strong for intermediate outcomes; survival utility unresolved", "independentReplication": "none for the proposed hierarchy", "sampleSize": "Three completed randomized evidence sets of n=124, n=147 and n=187; live PDAC trial estimated n=982", "effect": "Decision map and duplication rule; not a treatment result", "limits": [ "Generated action rule", "Olanzapine trial contains 13 HPB patients", "Ponsegromab trial has no results", "Primary endpoints remain weight and appetite", "Frailer patients are excluded", "Data-access terms are unknown" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The map follows measured host, treatment and survival links rather than treating weight as cancer control." }, "animalEvidence": { "rating": "moderate", "reason": "Models support targets but do not decide which human action to fund." }, "inVitroEvidence": { "rating": "weak", "reason": "Laboratory systems do not reproduce treatment adherence, appetite or quality of life." }, "reproducibility": { "rating": "none", "reason": "The full endpoint hierarchy and data-access rule have not been tested." }, "commercialConflicts": { "rating": "high", "reason": "Pfizer controls the largest trial, drug and relevant access stack." }, "recency": { "rating": "current", "reason": "The trial status and endpoints were checked in September 2026." } } }, "layout": { "x": 141, "y": 67 }, "updated": "2026-09-14" }, { "id": "trial-pdac-celiac-pain-action-map", "title": "PDAC pain-nerve action map: relieve pain without hiding death", "shortTitle": "Pancreatic pain-nerve action map", "type": "trial", "status": "active", "scope": [ "pdac", "pain", "palliative-care", "celiac-plexus", "capital-allocation" ], "summary": "Pain relief is real for some people, routine early treatment is unproved and survival warnings make the existing 94-person trial the only justified gate.", "content": "Route 1, anatomy. Alcohol celiac-plexus neurolysis damages a nerve network behind the upper abdomen; direct ganglion treatment, upstream splanchnic treatment, radiofrequency heat and radiation are different interventions. Route 2, early alcohol evidence. A 96-person randomized trial improved three-month pain but not quality of life or survival. Route 3, modern comparator. A later 46-person analysed trial found four-week pain 1.3 versus 2.3, p=0.10, and no advantage at any time for pain, opioids or quality of life. Route 4, safety warning. Direct ganglion treatment did not improve pain over plexus treatment and produced median survival 5.59 versus 10.46 months, HR 1.49. Splanchnic alcohol improved pain by only 0.7/10 but produced survival HR 1.56. Treat these as unresolved safety signals. Route 5, other energy. Heat beat alcohol over four weeks in 26 people, too few and too short for adoption. Celiac radiation produced pain response in 48/90 evaluable but only 48/125 treated, with serious possible treatment harm and no control. Route 6, live decision. NCT06160323 randomizes 94 to early ganglion-or-plexus treatment versus guideline step-up care, but mixes the exact target and leads with a three-month pain score. Route 7, capital. Fund no new people or procedure. Release at most £250,000 for exact target and dose records, independent monitoring, data custody and death-aware analysis. Route 8, stop. Require vital status for at least 90/94, at least 90% opioid and pain coverage and joint pain-death reporting. Pause direct ganglion treatment at death HR 1.5 after 20 deaths or any related death; stop early use for less than one point pain-interference benefit, less than 15% opioid saving, ten points more severe harm or worse quality-adjusted survival. Route 9, sequencing. The active randomized result comes before any heat or radiation comparison. Give the new 52-person single-arm radiation study £0. Route 10, clinical boundary. Current care remains individualized; this research map does not advise a person to start, avoid or stop a pain procedure without their oncology and pain teams.", "sourceIds": [ "wyse-2011-early-eus-cpn-rct", "kanno-2020-eus-cpn-modern-opioid-rct", "levy-2019-celiac-ganglia-versus-plexus-rct", "dong-2021-splanchnic-neurolysis-rct", "bang-2019-celiac-rfa-versus-cpn-rct", "lawrence-2024-celiac-radiosurgery-phase2", "clinicaltrials-nct06160323-upfront-cgn-cpn", "clinicaltrials-nct07689955-celiac-radiotherapy" ], "links": [ { "target": "trials", "relation": "part-of" }, { "target": "claim-pdac-nerve-ablation-pain-benefit-inconsistent", "relation": "maps" }, { "target": "failure-pdac-pain-death-and-evaluable-denominators", "relation": "maps" }, { "target": "unknown-pdac-nerve-ablation-net-benefit", "relation": "resolves" }, { "target": "hypothesis-pdac-nerve-ablation-existing-trial-safety-gate", "relation": "implements" }, { "target": "trial-pdac-cryoablation-action-map", "relation": "compares-with" }, { "target": "trial-pdac-hifu-action-map", "relation": "compares-with" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "weak", "randomised": "strong", "independentReplication": "moderate — conflicting net results", "sampleSize": "Five key randomized studies total 374 analysed; active target n=94; zero new people; cap £250,000", "effect": "Variable short pain relief; no stable quality-of-life gain; unresolved survival harm with aggressive targets", "limits": [ "Different procedures", "Short endpoint windows", "Death-dependent missing data", "Modern opioid effect", "Small safety counts", "Rescue crossover", "Current trial excludes opioid users" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The route directly interrupts pancreatic visceral pain signalling." }, "animalEvidence": { "rating": "not-applicable", "reason": "The decision has direct randomized human tests." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "The route is palliative, not tumour-cell treatment." }, "reproducibility": { "rating": "moderate", "reason": "Pain effects vary and the safest timing and target have not repeated cleanly." }, "commercialConflicts": { "rating": "low", "reason": "The mission purchases only independent monitoring and analysis." }, "recency": { "rating": "current", "reason": "The randomized gate is recruiting in 2026 and a new radiation study is imminent." } } }, "layout": { "x": 1162, "y": 1086 }, "updated": "2026-09-15" }, { "id": "trial-pdac-cryoablation-action-map", "title": "PDAC cryoablation action map: test pain relief and reject survival shortcuts", "shortTitle": "Cryoablation action map", "type": "trial", "status": "active", "scope": [ "pdac", "unresectable", "pain", "local-treatment" ], "summary": "Cryoablation's investable question is short-term palliative value in safely reached tumours; survival, immune and imaging claims remain blocked.", "content": "Route 1, mechanism. Freeze-thaw cycles form ice, disrupt membranes and blood supply and kill local tissue; computed tomography shows the ice-ball boundary. Route 2, palliative signal. Eleven selected people had mean pain fall from 6.72 to 3.45 at one month; all reduced opioids and five stopped. Route 3, missing control. Pain can change with sedation, expectation, systemic treatment and selective follow-up. Opioid doses were qualitative, and twelve-month pain included only eight survivors. Route 4, route-specific harm. Open cryo produced pancreatic fluid leakage in 18/24, including four care-changing fistulas; percutaneous treatment excluded bowel invasion and used deliberately incomplete margins. Route 5, survival shortcut. A selected 24-versus-24 comparison reported 16.8 versus 11.4 months, while an older 118-person study found no significant prognosis difference. Neither assigned treatment. Route 6, combination trap. NCT07306286 joins cryo and PD-1/CTLA-4 blockade in 20 people without a control, so it cannot isolate either effect. Route 7, action. Spend up to £250,000 on protocol and site readiness. If at least 20% of screened patients have a safe path, run a 100-person palliative comparison capped at £3 million; buy no company equity. Route 8, release. Require at least a two-point pain-interference gain, 25% lower morphine-equivalent dose, no loss of days alive at home, bowel injury at or below 3%, severe pancreatitis at or below 5% and chemotherapy delay at or below seven days. Keep tumour cryo, open cryo and nerve cryo separate.", "sourceIds": [ "pusceddu-2026-percutaneous-cryoablation", "kang-2025-laparotomic-cryoablation", "song-2014-cryo-bypass-comparison", "clinicaltrials-nct07306286-cryo-immunotherapy" ], "links": [ { "target": "trials", "relation": "part-of" }, { "target": "claim-cryoablation-pain-signal-not-survival-proof", "relation": "maps" }, { "target": "failure-cryoablation-selection-attrition-and-route-harm", "relation": "maps" }, { "target": "unknown-cryoablation-randomized-pain-and-net-benefit", "relation": "resolves" }, { "target": "hypothesis-cryoablation-pragmatic-pain-trial", "relation": "implements" }, { "target": "trial-pdac-hifu-action-map", "relation": "compares-with" }, { "target": "trial-pdac-rfa-action-map", "relation": "compares-with" }, { "target": "trial-pdac-histotripsy-action-map", "relation": "compares-with" } ], "evidence": { "humanProspective": "weak", "humanRetrospective": "moderate", "randomised": "none — proposed", "independentReplication": "weak", "sampleSize": "Existing n=11, n=48 and n=118; proposed randomized n=100; maximum staged spend £3.25 million", "effect": "Strong uncontrolled pain signal; no reliable survival effect; palliative trial is the next action", "limits": [ "Small selected pain cohort", "Open and percutaneous routes differ", "No standard opioid units", "Survivor-only late scores", "Unassigned survival controls", "Mixed systemic regimens", "Palliative endpoint only" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Freezing can reduce viable local tissue and pressure around pain pathways." }, "animalEvidence": { "rating": "moderate", "reason": "Ice geometry and tissue injury are established across models." }, "inVitroEvidence": { "rating": "strong", "reason": "Freeze-thaw cell and vascular damage are direct effects." }, "reproducibility": { "rating": "weak", "reason": "No randomized palliative effect has been reproduced." }, "commercialConflicts": { "rating": "moderate", "reason": "Supplier-neutral procurement and independent analysis remain required." }, "recency": { "rating": "current", "reason": "The route includes May 2026 clinical data and current trial status." } } }, "layout": { "x": 868, "y": 792 }, "updated": "2026-09-15" }, { "id": "trial-pdac-ctdna-action-map", "title": "Three PDAC ctDNA programmes test different actions; none yet proves that acting improves survival", "shortTitle": "PDAC ctDNA action trials", "type": "trial", "status": "open", "scope": [ "pdac" ], "summary": "DYNAMIC-Pancreas, CIRCPAC and MAP-03 test adjuvant treatment choice, intensified surveillance and early treatment switching, but no completed randomized result supports routine ctDNA-directed care.", "content": "The decision is not whether ctDNA predicts recurrence; it is which action, in which patient state, produces a better outcome. DYNAMIC-Pancreas covers postoperative adjuvant choice. Its registry plans a randomized comparison in patients treated before surgery, but the reported 102-patient upfront-surgery cohort was a feasibility study: 40 were ctDNA-positive, 54 negative and only 24 of 54 negative patients de-escalated to three months of chemotherapy at clinician discretion. Median recurrence-free survival was 13 versus 22 months by ctDNA status, a prognostic comparison rather than a randomized action effect. CIRCPAC tests surveillance action: an estimated 1,000 Danish patients are randomized to ctDNA-triggered CT and EUS every three months or standard surveillance, with three-year disease-free and overall survival listed as primary outcomes and completion estimated in 2032. MAP-03 tests treatment switching in 100 Chinese patients with borderline-resectable or locally advanced disease: rising MRD without imaging progression triggers an unspecified later-line regimen versus continuing treatment. Its primary endpoint is resection rate, its registry was last verified in 2023 and its estimated primary-completion date passed without posted results. These trials cover different actions and stages, so their estimates cannot be pooled as one utility test. The mission should secure common assay, specimen, false-positive, quality-of-life and survival fields across them before paying for another overlapping trial. A new randomized study is justified only if none can answer the chosen treatment rule with a current comparator and adequate power.", "sourceIds": [ "dynamic-pancreas-registry", "lee-2024-dynamic-pancreas-feasibility", "clinicaltrials-nct05788744-circpac", "clinicaltrials-nct05802394-map03", "guo-2026-postsurgical-ctdna", "chen-2026-personalized-ctdna", "daamen-2024-radar-panc-protocol", "clinicaltrials-nct04875325-radar-panc" ], "links": [ { "target": "trials", "relation": "part-of" }, { "target": "unknown-ctdna-actionability", "relation": "tests" }, { "target": "residual-disease", "relation": "tests" }, { "target": "hypothesis-multimodal-mrd-action", "relation": "constrains" }, { "target": "programme-100m-randomized-platform", "relation": "must-not-duplicate" }, { "target": "trial-pdac-postresection-surveillance-action-map", "relation": "bounded-by" }, { "target": "unknown-postresection-surveillance-utility", "relation": "bounded-by" } ], "evidence": { "humanProspective": "moderate — one feasibility cohort and three active programme designs", "humanRetrospective": "moderate — two 2026 assay-performance cohorts", "randomised": "none — randomized utility results not reported", "independentReplication": "weak", "sampleSize": "DYNAMIC feasibility n=102; CIRCPAC planned n=1,000; MAP-03 planned n=100; 2026 performance cohorts n=51 and n=43; RADAR-PANC plans 306 randomized participants", "effect": "Prognostic separation and feasibility reported; treatment, surveillance and survival benefit from acting remain unknown", "limits": [ "No completed randomized utility result", "DYNAMIC reported cohort was not the randomized comparison", "CIRCPAC completion is estimated in 2032", "MAP-03 registry status is stale and its intervention is not fixed precisely", "Assays and patient settings differ", "Trial registries describe plans, not delivered results", "RADAR-PANC has no posted result and its registry was last verified in September 2024" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Residual PDAC can release tumour DNA before imaging recurrence, and several cohorts show prognosis differs by ctDNA status." }, "animalEvidence": { "rating": "not-applicable", "reason": "The clinical-action question must be answered in randomized human care pathways." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Laboratory detection does not establish that surveillance or treatment changes improve patient outcomes." }, "reproducibility": { "rating": "weak", "reason": "Prognostic direction repeats, but sensitivity, specificity and predictive values vary and no action effect has been reproduced." }, "commercialConflicts": { "rating": "moderate", "reason": "Some assay development and cohort work involve commercial test makers, while the listed randomized studies are institution-led." }, "recency": { "rating": "current", "reason": "The map includes 2026 cohorts and a trial status verified in March 2026." } } }, "layout": { "x": 97, "y": 65 }, "updated": "2026-09-14", "contentSections": [ "RADAR-PANC sets the standard-imaging comparator: NCT04875325 randomizes a planned 306 Dutch and UK patients to three-monthly CA19-9 plus CT or mainly symptom-led follow-up, with overall survival primary and no posted result before an estimated October 2026 completion. CIRCPAC tests a different ctDNA-directed intensity question. Recover RADAR-PANC's result before any biomarker is credited for the benefit of earlier imaging." ] }, { "id": "trial-pdac-electrochemotherapy-action-map", "title": "PDAC electrochemotherapy action map: recover the randomization, not another cohort", "shortTitle": "ECT action map", "type": "trial", "status": "active", "scope": [ "pdac", "locally-advanced", "resectable", "local-treatment" ], "summary": "Electrochemotherapy has passed delivery but not patient benefit; its 90-person randomized record and complete device logs are the next decision asset.", "content": "Route 1, mechanism. ECT uses short high-voltage pulses to open membranes temporarily while bleomycin is present; irreversible electroporation uses permanent membrane damage and is a separate treatment. Route 2, selected cohort. The 25-person study accrued about 3.1 people per year, selected chemotherapy non-progressors and reported 24/25 local control at one month. By six months 11/25 had died and lacked imaging; 14/25 remained controlled. Route 3, electrode claim. Fixed and variable geometry were surgeon-chosen. Median survival was 6 versus 12 months, P=0.18; the paper's stated 4.5-month difference fails 12 - 6 = 6. Route 4, cut-surface route. PanECT delivered treatment after resection in 10 people, with four complications, one serious fistula followed by death and seven later deaths. It has no control and posts no disease-free or overall-survival table. Route 5, decisive comparison. EudraCT 2018-003925-27 planned 90 people, identical FOLFOXIRI and a 40% to 70% response target. It transitioned to EU CT 2024-519067-17-00 without a public result. Route 6, action. Invest £0 in IGEA, £0 in equipment and £0 in another cohort. Spend at most £250,000 to recover at least 80 assignments only after raw access, independent analysis and negative-result publication are contracted. Route 7, release. Require 95% vital status, 90% pulse-log and image coverage, response counting death as failure, independent image agreement of 80%, preserved chemotherapy, less than five points excess severe harm and patient-valued or survival benefit. Route 8, separation. Do not pool tumour ECT, cut-surface ECT, irreversible electroporation and calcium electroporation. If the randomized record is too small or inaccessible, stop this mission branch.", "sourceIds": [ "izzo-2021-pancreatic-ect-phase12", "izzo-2021-laparoscopic-ect-protocol", "eudract-2018-003925-27-ect", "cebron-2023-panect-preliminary", "clinicaltrials-nct04281290-panect-results", "rudno-rudzinska-2021-irec-protocol", "igea-2026-cliniporator-vitae" ], "links": [ { "target": "trials", "relation": "part-of" }, { "target": "claim-electrochemotherapy-local-effect-without-patient-benefit", "relation": "maps" }, { "target": "failure-electrochemotherapy-endpoint-and-trial-drift", "relation": "maps" }, { "target": "unknown-electrochemotherapy-added-benefit-and-delivery", "relation": "resolves" }, { "target": "hypothesis-electrochemotherapy-randomized-data-gate", "relation": "implements" }, { "target": "company-igea", "relation": "involves" }, { "target": "trial-pdac-irreversible-electroporation-action-map", "relation": "compares-with" }, { "target": "trial-pdac-rfa-action-map", "relation": "compares-with" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "not-applicable", "randomised": "weak — protocol without result", "independentReplication": "weak", "sampleSize": "Single arms n=25 and n=10; randomized target n=90; zero new participants; recovery cap £250,000", "effect": "Feasible delivery; no comparative patient benefit; randomized-data recovery is the action", "limits": [ "Randomized actual enrollment unknown", "No survival-powered design", "Forty-four percent missing six-month scans through death", "Chosen electrode comparison", "Single manufacturer stack", "Open surgery burden", "PanECT outcome tables incomplete" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Electric-field drug entry is a direct physical mechanism." }, "animalEvidence": { "rating": "moderate", "reason": "Deep-tumour models support delivery, not human net benefit." }, "inVitroEvidence": { "rating": "strong", "reason": "Temporary membrane opening and drug uptake are measurable." }, "reproducibility": { "rating": "weak", "reason": "Feasibility repeats; comparative patient benefit does not." }, "commercialConflicts": { "rating": "high", "reason": "The main device maker touches equipment, planning, funding and authorship." }, "recency": { "rating": "current", "reason": "EU transition and 2026 PanECT results are included." } } }, "layout": { "x": 760, "y": 684 }, "updated": "2026-09-15" }, { "id": "trial-pdac-engineered-cell-action-map", "title": "PDAC engineered-cell trial map: recover lost patients and use existing cohorts", "shortTitle": "Engineered-cell action map", "type": "trial", "status": "active", "scope": [ "pdac", "pan-cancer" ], "summary": "No duplicate cell cohort is needed: three existing data routes can recover the mesothelin denominator, mature satri-cel, and test whether KRAS-G12V TCR responses repeat.", "content": "Route 1, terminated mesothelin CAR-T. Reconcile NCT03323944's 54 enrolled people with the nine infused people in the paper. Publish screen failure, collection, manufacture, release, conditioning, infusion, toxicity, response, survival and reason for termination for all 54. No new mesothelin cohort is justified by a nine-person zero-response result. Route 2, CLDN18.2 CAR-T. Release the full 134-person NCT03874897 record and identify every pancreatic patient before pooling. For the 24 infused PDAC patients, publish marker fraction and intensity by lesion, manufacturing time, bridging treatment, cell expansion, hospital days, response duration and survival. NCT05911217's 20-person adjuvant study cannot establish benefit without a concurrent group starting at the same disease-free landmark; six patients followed for a median 6.05 months are a safety and feasibility report. Route 3, KRAS-G12V TCR-T. Use the live 30-person NCT04146298 cohort; add no mission cohort. Preserve the full mutation-HLA screen and collection denominator, prospectively measure tumour presentation and cell state, and require at least five confirmed responses among at least 25 infused people with four lasting six months before a controlled design. Repeat infusion is suspended unless receptor immunogenicity is repaired and tested. Capital. Mission investment is £0 now. CARsgen had RMB1.400bn cash at 30 June 2026 and can finish its studies. The academic KRAS study is already recruiting. Data access and measurement are the missing assets, not another product bet.", "sourceIds": [ "clinicaltrials-nct03323944-meso-cart", "aznar-2025-mesothelin-car-t-resistance", "qi-2024-ct041-pdac", "clinicaltrials-nct03874897-ct041", "clinicaltrials-nct04581473-ct041", "clinicaltrials-nct05911217-satricel-adjuvant", "carsgen-2025-satricel-pdac-adjuvant", "xu-2026-kras-g12v-tcr", "clinicaltrials-nct04146298-kras-g12v-tcr", "leidner-2022-kras-tcr", "carsgen-2026-interim-results" ], "links": [ { "target": "trials", "relation": "part-of" }, { "target": "claim-satricel-pdac-early-signal", "relation": "maps" }, { "target": "claim-kras-g12v-tcr-early-signal", "relation": "maps" }, { "target": "failure-cell-therapy-route-and-denominator-pdac", "relation": "repairs" }, { "target": "unknown-engineered-cell-reproducible-benefit-pdac", "relation": "resolves" }, { "target": "hypothesis-kras-tcr-human-persistence-gate", "relation": "implements" }, { "target": "company-carsgen-therapeutics", "relation": "monitors" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "not-applicable", "randomised": "none", "independentReplication": "weak", "sampleSize": "Existing registries n=54, n=134, n=192, n=20 and n=30; published PDAC efficacy groups n=9, n=24, n=6 and n=5", "effect": "Decision map; product effects remain uncontrolled", "limits": [ "Some registry populations mix cancers", "Participant-level access is unconfirmed", "Protocol amendments may explain denominator differences", "No current randomized PDAC cell-therapy trial" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The map follows the necessary chain from target matching through durable tumour killing." }, "animalEvidence": { "rating": "moderate", "reason": "Models inform measurements but do not release clinical funding." }, "inVitroEvidence": { "rating": "strong", "reason": "Product potency and target recognition are measurable inputs." }, "reproducibility": { "rating": "weak", "reason": "The map exists because current responses have not reproduced within stable architectures." }, "commercialConflicts": { "rating": "high", "reason": "CARsgen controls satri-cel; academic developers control the KRAS receptor records." }, "recency": { "rating": "current", "reason": "Registry, publication and finance status were checked through September 2026." } } }, "layout": { "x": 250, "y": 174 }, "updated": "2026-09-15" }, { "id": "trial-pdac-eus-rfa-action-map", "title": "PDAC endoscopic radiofrequency action map: reconcile, measure, then stop or compare", "shortTitle": "EUS-RFA action map", "type": "trial", "status": "active", "scope": [ "pdac", "locally-advanced", "borderline-resectable", "local-treatment" ], "summary": "Endoscopic radiofrequency has passed technical delivery but failed to show added patient benefit; existing records, not another cohort, are the next asset.", "content": "Route 1, physical effect. An ultrasound-guided probe can place radiofrequency heat inside a pancreatic mass. HybridTherm used internal cooling; it was not tumour freezing despite the registry label. Route 2, delivery. First-session feasibility was 16/17, 94.1%; 11/37 procedures had an adverse event, none severe. Route 3, randomized output. Six-month freedom from progression was 41.2% versus 30%, P=0.48. Median survival was 13 versus 17 months, HR 1.18, P=0.79. Resection was 11.8% versus 20%. Route 4, evidence-system break. Only 40 of 66 planned phase 2 assignments were reported, three device assignments were excluded, the manufacturer held allocation and monitoring, the probe was withdrawn, and the registry now says 66 actual participants without results. Route 5, successor trap. PANCARDINAL-1 plans 60 people in one arm and asks whether they complete treatment and reach surgery. It cannot isolate RFA from chemotherapy, time selection or surgical selection. Route 6, action. Invest £0 in a manufacturer, £0 in equipment and £0 in new enrollment. Spend at most £300,000 to reconcile the old randomized record and harmonize existing probe logs only after independent access and publication rights. Route 7, release. Require complete assignment reconciliation, 95% vital status, 90% device logs, mapped heat coverage, preserved chemotherapy, severe device harm at or below five percent and a consistent patient-valued signal. Route 8, separation. Keep tumour RFA distinct from pain-directed coeliac nerve ablation, HIFU, open surgical RFA, irreversible electroporation and benign pancreatic lesions. If custody or outcome gates fail, stop this branch.", "sourceIds": [ "testoni-2021-hybridtherm-rct", "clinicaltrials-nct02336672-hybridtherm", "clinicaltrials-nct04990609-pancardinal1" ], "links": [ { "target": "trials", "relation": "part-of" }, { "target": "claim-hybridtherm-randomized-no-patient-benefit", "relation": "maps" }, { "target": "failure-hybridtherm-product-accrual-and-custody", "relation": "maps" }, { "target": "unknown-eus-rfa-added-benefit-after-hybridtherm", "relation": "resolves" }, { "target": "hypothesis-eus-rfa-existing-study-gate", "relation": "implements" }, { "target": "company-erbe-elektromedizin", "relation": "involves" }, { "target": "trial-pdac-rfa-action-map", "relation": "compares-with" }, { "target": "trial-pdac-hifu-action-map", "relation": "compares-with" }, { "target": "trial-pdac-irreversible-electroporation-action-map", "relation": "compares-with" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "moderate", "independentReplication": "weak", "sampleSize": "HybridTherm n=40 randomized; successor PANCARDINAL-1 target n=60 single arm; zero new participants; audit cap £300,000", "effect": "Feasible local ablation with no demonstrated survival or resection benefit; record recovery is the action", "limits": [ "Early stop", "Excluded assignments", "Registry mismatch", "Manufacturer-controlled allocation and monitoring", "Product withdrawal", "No independent randomized replication", "Successor designs cannot estimate added effect" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Controlled heat destroys local tissue through a direct physical mechanism." }, "animalEvidence": { "rating": "moderate", "reason": "Models support ablation and immune effects but cannot settle net human benefit." }, "inVitroEvidence": { "rating": "strong", "reason": "Temperature, impedance and tissue destruction are measurable." }, "reproducibility": { "rating": "weak", "reason": "Technical feasibility repeats, but patient benefit does not." }, "commercialConflicts": { "rating": "high", "reason": "The sole randomized record had concentrated manufacturer control." }, "recency": { "rating": "current", "reason": "Successor trial state was checked in July and September 2026." } } }, "layout": { "x": 796, "y": 720 }, "updated": "2026-09-15" }, { "id": "trial-pdac-exercise-action-map", "title": "Pancreatic exercise action map: preserve function without mistaking movement for survival", "shortTitle": "Exercise action map", "type": "trial", "status": "active", "scope": [ "pdac", "supportive-care", "surgery", "exercise", "quality-of-life", "capital-allocation" ], "summary": "Offer safe movement support, recover the randomized records, and require useful days or treatment delivery before making exercise a cancer-pathway investment.", "content": "Route 1, present care. Encourage safe, patient-chosen activity and rehabilitation as supportive care; adapt for pain, falls, cardiac risk, neuropathy, weight loss and treatment days. Do not promise cancer control. Route 2, advanced disease. APACaP randomized 313 and missed all three week-16 main quality endpoints in 172 analysable patients. Secondary longitudinal results among 259 remain a signal, not a reversal. Route 3, neoadjuvant care. PancFit increased strength sessions but produced similar aerobic activity and no significant quality or clinical difference. Route 4, surgery. Step calls, early walking, multicomponent preparation and goal-directed walking can change movement or selected recovery outputs. Trials of 40 to 180 people did not show a repeat reduction in overall major complications. Route 5, active duplication. HY-PANC plans ten unresectable patients and lean mass as its main outcome. Let it test feasibility; it cannot establish survival or route benefit. Route 6, existing-data gate. Spend at most £250,000 and enroll nobody. Recover at least 80% of assignments from five randomized studies, retain death and progression, and align function, received activity, treatment dose, operations, complications, admissions and quality-weighted days at home. Route 7, selector. Derive a low-function rule in one dataset and require the same interaction direction in two independent trials. Do not select on improvement after exercise. Route 8, success. Require at least five extra quality-weighted days over 16 weeks or a ten-point increase in receiving 80% of planned chemotherapy, without more falls, admissions, delay or early death. Route 9, failure. Stop the cancer-pathway claim if data recovery is below 80%, the interaction does not repeat, benefit is smaller, or burden erases it. Ordinary safe rehabilitation can continue. Route 10, capital. If the gate passes, place the fixed route inside the existing palliative-care or surgical-delivery work; added trial capital is £0 and no exercise platform or wearable company is purchased.", "sourceIds": [ "neuzillet-2023-apacap-rct", "ngohuang-2023-pancfit-rct", "ausania-2019-pancreatic-prehab-rct", "steindorf-2019-pancreatic-exercise-rct", "zarate-rodriguez-2023-pancreatectomy-steps-rct", "mizuno-2024-hpb-prehab-rct", "li-2024-pancreatic-early-mobilization-rct", "clinicaltrials-nct07470268-hypanc" ], "links": [ { "target": "trials", "relation": "part-of" }, { "target": "claim-pdac-exercise-function-not-survival", "relation": "maps" }, { "target": "failure-pdac-exercise-endpoint-and-attrition", "relation": "maps" }, { "target": "unknown-pdac-exercise-net-benefit-and-selector", "relation": "resolves" }, { "target": "hypothesis-pdac-exercise-existing-data-gate", "relation": "implements" }, { "target": "trial-pdac-palliative-care-action-map", "relation": "coordinates-with" }, { "target": "trial-pdac-surgical-route-action-map", "relation": "coordinates-with" }, { "target": "trial-pdac-cachexia-action-map", "relation": "coordinates-with" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "moderate", "randomised": "strong — multiple randomized comparisons define the action boundary", "independentReplication": "moderate — feasibility repeats but patient-level benefit does not", "sampleSize": "Randomized studies n=313, n=151, n=180, n=152, n=135, n=65 and n=40; recovery cap £250,000", "effect": "Supportive care remains reasonable; cancer-pathway investment waits for a repeated patient-level selector and useful-day effect", "limits": [ "Heterogeneous programmes", "High attrition", "Different comparators", "No survival power", "Data access", "Exercise cannot replace nutrition or symptom care" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Function is modifiable and can support care, while endpoints keep the claim bounded." }, "animalEvidence": { "rating": "not-applicable", "reason": "The action map depends on human adherence and patient-valued outcomes." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "No laboratory model can validate this care route." }, "reproducibility": { "rating": "moderate", "reason": "Movement changes reproduce, while durable patient benefit remains unconfirmed." }, "commercialConflicts": { "rating": "low", "reason": "The route uses ordinary care and independent analysis, not a proprietary platform." }, "recency": { "rating": "current", "reason": "Recent randomized surgical studies and the 2026 HY-PANC plan were checked." } } }, "layout": { "x": 2480, "y": 1720 }, "updated": "2026-09-15" }, { "id": "trial-pdac-fap-radioligand-action-map", "title": "PDAC FAP-radioligand action map: recover dose data before another trial", "shortTitle": "FAP treatment action map", "type": "trial", "status": "mixed", "scope": [ "pdac", "cholangiocarcinoma", "pan-cancer", "transfer-unproven" ], "summary": "Existing trials already cover dose escalation and single-arm PDAC response; the missing work is joined dose data followed, only if warranted, by a controlled treatment test.", "content": "Completed evidence. The 2022 compassionate-use FAP-2286 series treated 11 people, five with pancreatic cancer, and established biodistribution without a pancreatic response. LuMIERE phase 1 then scanned 35 and treated 27 across six centres, set 9.25 GBq as the phase 2 dose and reported four stable pancreatic cancers among nine treated. It did not publish tumour-lesion dosimetry. Live study. NCT04939610 remains recruiting with estimated n=222 and December 2027 primary completion. It already tests FAP-2286 alone after one or two advanced-PDAC regimens and with first-line modified FOLFIRINOX. Starting another uncontrolled PDAC response cohort would duplicate this work. Small study. NCT07229768 plans four people; it can add agent-specific delivery and immediate safety only. Stopped study. FRONTIER treated 20 and terminated for a business decision, with no public result. Data recovery has higher information value than assuming failure. Uncontrolled signal. The 3BP-3940 report treated 88 people with three isotopes across 21 cancers and reported response for 51 while allowing other care. Its one pancreatic dose example is useful for measurement design; its pooled response rate cannot size a PDAC benefit trial. Action now: request LuMIERE patient- and lesion-level scan, residence-time, absorbed-dose and response data; request FRONTIER results and screen failures; require all-denominator tables; and align lesion identifiers across scan, dose and response. Do this inside current data and imaging work. Gate to a future randomized PDAC test only if a dose-response reproduces, the PDAC phase 2 signal crosses its declared response bound, combination treatment preserves standard-drug dose intensity, and kidney, marrow and bowel rules remain acceptable. The future comparison must add the radioligand to the same standard regimen used in control and measure overall and quality-adjusted survival. No new treatment cohort is funded now.", "sourceIds": [ "baum-2022-fap2286-firsthuman", "mcconathy-2026-lumiere-phase1", "clinicaltrials-nct04939610-lumiere", "baum-2026-3bp3940", "clinicaltrials-nct05432193-frontier", "clinicaltrials-nct07229768-ctr-fapi", "pang-2022-fapi-pet", "lee-2024-prospective-fapi-resectable-pdac", "evans-2026-fapi-feasibility" ], "links": [ { "target": "trials", "relation": "part-of" }, { "target": "claim-fap-radioligand-feasible-pdac-efficacy-unproven", "relation": "supports" }, { "target": "failure-fapi-scan-dose-benefit-collapse", "relation": "repairs" }, { "target": "unknown-pdac-fap-radioligand-treatment-effect", "relation": "maps" }, { "target": "hypothesis-fapi-theranostic-pdac", "relation": "gates" }, { "target": "transfer-radioligand-selection", "relation": "tests-transfer-from" } ], "evidence": { "humanProspective": "moderate — completed dose escalation and active PDAC expansion", "humanRetrospective": "moderate — compassionate-use and mixed-cancer experience", "randomised": "none", "independentReplication": "moderate for lesion delivery, none for PDAC benefit", "sampleSize": "Published FAP-2286 n=11 and n=27; LuMIERE estimated n=222; FRONTIER actual n=20; CTR-FAPI estimated n=4; 3BP-3940 n=88 with 51 response-evaluable", "effect": "Human delivery and phase 2 dose established; no comparative PDAC efficacy result", "limits": [ "Current phase 2 is non-randomized", "Tumour dosimetry is missing from the LuMIERE paper", "FRONTIER results are absent", "Cross-agent results cannot be pooled without kinetics and isotope alignment", "Trial completion dates may change" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "Human imaging and therapy show target reach, while target function and lesion coverage remain variable." }, "animalEvidence": { "rating": "moderate", "reason": "Models support the route but add little once human dose data exist." }, "inVitroEvidence": { "rating": "moderate", "reason": "Carrier binding supports agent construction, not the patient decision." }, "reproducibility": { "rating": "moderate", "reason": "Delivery repeats across agents; PDAC response and useful-life benefit do not." }, "commercialConflicts": { "rating": "high", "reason": "Two major programmes are or were industry-led, and specialist-centre reports lack controls." }, "recency": { "rating": "current", "reason": "Includes June 2026 publications and July 2026 LuMIERE status." } } }, "layout": { "x": 82, "y": 148 }, "updated": "2026-09-15" }, { "id": "trial-pdac-ferroptosis-action-map", "title": "PDAC ferroptosis action map: prove selective cell death before efficacy", "shortTitle": "Ferroptosis action map", "type": "trial", "status": "active", "scope": [ "pdac", "pan-cancer", "ferroptosis", "nanoparticle-delivery", "capital-allocation" ], "summary": "Ferroptosis is compelling model biology with two pancreatic human exposures and no measured human mechanism; use the ongoing study as a tissue gate and buy no equity.", "content": "Route 1, mechanism. Iron-driven lipid oxidation breaks cell membranes; xCT supplies cystine for glutathione and GPX4 repairs lipid damage. Route 2, vulnerability. xCT loss kills PDAC cells but only delays xenografts because living tumours find other cysteine. Fentomycin kills iron-rich persister cells in culture. Route 3, direction risk. Pancreas-wide GPX4 loss or high iron accelerated KRAS tumour formation in mice through oxidized DNA, STING and macrophages. A tumour-selective GPX4 degrader instead supported CD8 immunity. Cell identity decides direction. Route 4, first human product. CNSI-Fe directly injects carbon-carried ferrous iron. Nineteen people across cancers were treated, fifteen completed and two had pancreatic cancer. Fourteen had grade 3-plus events, eight serious events and no reported treatment-related death. Route 5, efficacy boundary. The pancreatic long survivor had stable injected-lesion size, then later chemotherapy response. Ferroptosis was not measured; paired tumour-growth-rate change was non-significant, p=0.554. This is one case, not resistance reversal. Route 6, active expansion. NCT07433283 recruits 54 across many cancers, leads with 12-week safety and has no pancreatic quota or control. Route 7, company. Enray is a controlled subsidiary of listed Chongqing Lummy, owns a pending patent family and sponsors both trials. Its parent is already financing development; no mission equity is justified. Route 8, capital. Spend £0 on company equity, product or new participants. Release at most £400,000 for independent analysis of existing human samples and six model component tests. Route 9, release. Require at least twelve paired human cases, three pancreatic models, mapped iron and tumour coverage, two lipid-oxidation measures, repair failure, tumour-cell-selective injury and ferroptosis rescue. Stop below three paired pancreatic cases, below 70% coverage, immune injury at least half tumour injury or failed component and rescue tests. Route 10, future benefit. A later same-drug rechallenge comparison receives £0 until the human mechanism passes. Local necrosis, a patent and survival after later therapy are not substitutes.", "sourceIds": [ "daher-2019-xct-pdac-ferroptosis", "dai-2020-ferroptotic-damage-pdac", "li-2023-tumour-selective-gpx4-degrader", "caneque-2025-fentomycin-lysosomal-iron", "xie-2026-cnsi-fe-first-human", "clinicaltrials-nct06048367-cnsi-fe", "clinicaltrials-nct07433283-cnsi-fe-expansion", "patent-us20250144252a1-cnsi-fe", "enray-2026-company", "lummy-2025-annual-report-cnsi-fe" ], "links": [ { "target": "trials", "relation": "part-of" }, { "target": "claim-ferroptosis-pdac-selectivity-not-class-effect", "relation": "maps" }, { "target": "failure-cnsi-fe-resensitization-and-mechanism-claim", "relation": "maps" }, { "target": "unknown-ferroptosis-cell-type-coverage-and-human-benefit", "relation": "resolves" }, { "target": "hypothesis-cnsi-fe-current-study-mechanism-gate", "relation": "implements" }, { "target": "company-sichuan-enray-pharmaceutical", "relation": "involves" }, { "target": "trial-pdac-autophagy-action-map", "relation": "compares-with" }, { "target": "trial-pdac-intraarterial-delivery-action-map", "relation": "compares-with" } ], "evidence": { "humanProspective": "weak", "humanRetrospective": "weak", "randomised": "none", "independentReplication": "none in humans", "sampleSize": "Human product n=19, pancreatic n=2; expansion target n=54; zero new people; cap £400,000", "effect": "Local human necrosis without measured ferroptosis or controlled pancreatic benefit", "limits": [ "Two pancreatic cases", "Many cancers", "No human mechanism markers", "No control", "Post-treatment confounding", "Local injection", "Company-owned product and data", "Context-dependent tumour promotion" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The iron-lipid-repair mechanism is clear but direction depends on selective delivery." }, "animalEvidence": { "rating": "strong", "reason": "Mouse models show both intended killing and the exact context failure to guard against." }, "inVitroEvidence": { "rating": "strong", "reason": "PDAC xCT, GPX4 and lysosomal-iron vulnerabilities are established in model systems." }, "reproducibility": { "rating": "weak", "reason": "Different tools reproduce ferroptosis biology, not a human treatment effect." }, "commercialConflicts": { "rating": "high", "reason": "The human evidence, trials, patent and public claims centre on one sponsor-controlled product." }, "recency": { "rating": "current", "reason": "The action map includes 2025 chemistry, 2026 human reporting and a live expansion." } } }, "layout": { "x": 1192, "y": 1116 }, "updated": "2026-09-15" }, { "id": "trial-pdac-germline-cascade-action-map", "title": "PDAC germline action map: count the route from diagnosis to family action", "shortTitle": "Germline action map", "type": "trial", "status": "supported", "scope": [ "pdac", "pan-cancer", "transfer-unproven" ], "summary": "Completed studies show how to make testing easy after entry; the next trial must measure who never enters and whether a result changes care.", "content": "Patient-entry boundary: automatic referral lifted Dana-Farber completion from 16.5% to 38.0%; an embedded UCSF station reached 71%; a Johns Hopkins nurse-led offer reached about 69.8% of all 992 patients; a physician-led protocol reached 81.5%; and a prospective point-of-care cohort reached 97.4%. These are service comparisons, not randomized health-outcome trials. Time boundary: point-of-care testing cut median turnaround from 42 to 13 days, while 15/266 Dana-Farber patients died within 30 days of results. Result boundary: pathogenic-variant counselling completion was 64.7% in one mainstream route and 69.4% at Hopkins; a protocol that returned limited calls for negative and uncertain results generated variants of uncertain significance in 26.8%, showing the interpretation workload. Family boundary: 84% of relatives were told in one cohort but only 31% tested. GENERATE randomized 601 already motivated relatives and found covered remote testing completion of 87% with an added live session versus 93% with online education alone. GIFT randomized pan-cancer families and found that only 19.2% of reported relatives were invited; free testing helped, but the absolute family fraction tested increased only four points and human navigation alone did not clearly help. Policy boundary: England's current R367 rule is narrower than the ASCO all-patient indication, but no English all-comer study gives incremental action and cost. Do not fund another test-uptake report using only people who consented. Run one end-to-end programme with an all-patient denominator, R367-stratified yield, result-to-treatment time, family-cluster randomization, privacy events, inequalities, cost and gene-specific action. Keep PDAC, other exocrine cancers and neuroendocrine tumours separate.", "contentSections": [ "IGNITE-TX supplies a new transfer test. In 144 relatives from hereditary breast/ovarian or Lynch-syndrome families, six-month testing rose from 1/29 with printed information to 36/46 when proactive navigation, bilingual digital support and no-cost remote testing were combined. Navigation without free testing reached 17/30. The pilot supports active, consented, low-friction family contact, but is not PDAC-specific and does not measure cancer prevention." ], "sourceIds": [ "nhs-england-2026-r367", "asco-2024-germline-panel-guideline", "chittenden-2021-systematic-germline", "walker-2021-embedded-germline-station", "wang-2022-oncology-germline-cascade", "ramsey-2023-mainstreaming-germline", "rodriguez-2024-generate", "mcdonald-2024-germline-protocol", "sinan-2025-pdac-germline-implementation", "grewal-2026-pancan-germline-survey", "katz-2026-gift-cascade", "aguirre-2026-ignite-tx", "llach-2024-universal-germline" ], "links": [ { "target": "trials", "relation": "part-of" }, { "target": "claim-universal-germline-testing-needs-action", "relation": "uses" }, { "target": "failure-pdac-germline-route-dropout", "relation": "learns-from" }, { "target": "system-england-r367-germline-criteria", "relation": "tests-boundary-of" }, { "target": "unknown-pdac-germline-family-action-route", "relation": "tests" }, { "target": "hypothesis-pdac-germline-family-route-completion", "relation": "prioritizes" }, { "target": "claim-high-risk-surveillance", "relation": "routes-carriers-to" } ], "evidence": { "humanProspective": "strong for components", "humanRetrospective": "strong", "randomised": "moderate for family delivery", "independentReplication": "moderate for components; none for the whole route", "sampleSize": "Implementation cohorts n=245-1,214; family trials include GENERATE n=601 relatives, GIFT 414 probands with 4,946 reported relatives, and IGNITE-TX n=144 relatives", "effect": "Decision map; component effects remain in linked records", "limits": [ "No complete-route trial", "Mostly US academic care", "Selected family participants", "Pan-cancer transfer in GIFT", "Health outcomes not yet measured" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The map links inherited variants to specific patient and family decisions." }, "animalEvidence": { "rating": "not-applicable", "reason": "Human delivery studies define the route." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Bench work cannot estimate invitation, consent or action." }, "reproducibility": { "rating": "moderate", "reason": "Low-friction patient testing repeats, but no study covers the full English route." }, "commercialConflicts": { "rating": "moderate", "reason": "Some source routes used commercial or covered testing; the proposed trial should use product-neutral public procurement." }, "recency": { "rating": "current", "reason": "The map includes 2026 policy, survey and randomized cascade evidence." } } }, "updated": "2026-09-14" }, { "id": "trial-pdac-hifu-action-map", "title": "PDAC focused-ultrasound action map: test symptom value before expansion", "shortTitle": "HIFU action map", "type": "trial", "status": "active", "scope": [ "pancreatic-all", "pdac", "locally-advanced", "metastatic", "local-treatment" ], "summary": "Thermal HIFU has a credible pain signal but no trustworthy survival result; recover two existing randomizations and keep non-thermal drug-delivery ultrasound separate.", "content": "Route 1, mechanism. Thermal HIFU focuses acoustic energy into a pancreatic target, raising temperature enough to destroy tissue without a needle. Route 2, symptom evidence. A 50-person prospective single arm reported early pain relief in 84%. A 120-person retrospective comparison reported 57% versus 20% pain remission, a crude 37-point difference and possible number needed to treat of 2.7, but its allocation is not auditable. Route 3, survival boundary. That retrospective paper reported 10.3 versus 6.6 months and an adjusted death HR 0.193, yet provides no random sequence or registry and contradicts itself on design. Do not count it as randomized proof. Route 4, academic randomization. HIFU-PC15 started in February 2017, targets 40 and still says recruiting on a February 2024 record with no final count, no result and no participant-data sharing. Route 5, industry randomization. NCT05601323 targets 90, uses overall survival as primary, is active but not recruiting and estimates May 2027 completion; Sonire now says enrollment is complete. Route 6, U.S. transfer. SUNRISE-II is recruiting ten untreated people at Stanford, uses more than 50% treated-volume ablation as primary and has no control. Route 7, operating failures. NCT05262452 is unknown and stale with no result; PULS is suspended after its industrial partner permanently closed recruitment. Route 8, mechanism split. SonoPANIII plans 360 people but describes IMD10 as non-thermal and non-tissue-destructive. K-912 adds nanoparticle epirubicin to ultrasound. Neither future result can validate thermal ablation, and the reverse is also true. Route 9, action. Sonire reports $31 million of new private and government funding in 2026. Invest £0 in a device, £0 in another cohort and £0 before existing database locks. Spend at most £250,000 afterward only for independent HIFU-PC15 and NCT05601323 recovery and analysis with raw data and publication rights. Route 10, release. Require at least 95% and 99% assignment accounting respectively, at least 90% device-log coverage, less than five points excess serious harm, preserved chemotherapy and pain plus opioid improvement or positive quality-adjusted survival. If those gates fail, stop mission expansion; symptom use must still compete with ordinary pain care on burden and cost.", "sourceIds": [ "li-2015-hifu-s1", "marinova-2018-hifu-prospective", "marinova-2024-hifu-pc-protocol", "drks-00012367-hifu-pc", "clinicaltrials-nct05262452-hifu-folfirinox", "clinicaltrials-nct05601323-suizenji", "clinicaltrials-nct06211933-puls-hifu", "clinicaltrials-nct07325214-sonopan3", "clinicaltrials-nct07033689-sunrise2", "sonire-2026-series-a", "sonire-2026-nedo-award" ], "links": [ { "target": "trials", "relation": "part-of" }, { "target": "claim-hifu-pain-signal-without-survival-proof", "relation": "maps" }, { "target": "failure-hifu-randomization-and-trial-drift", "relation": "maps" }, { "target": "unknown-hifu-pain-survival-and-system-transfer", "relation": "resolves" }, { "target": "hypothesis-hifu-existing-randomized-gate", "relation": "implements" }, { "target": "company-sonire-therapeutics", "relation": "involves" }, { "target": "trial-pdac-rfa-action-map", "relation": "compares-with" }, { "target": "trial-pdac-irreversible-electroporation-action-map", "relation": "compares-with" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "moderate", "randomised": "weak", "independentReplication": "moderate — pain signal only", "sampleSize": "Published cohorts n=50 and n=120; randomized targets n=40 and n=90; zero new participants; analysis cap £250,000", "effect": "Possible symptom value; survival and quality-adjusted survival unresolved", "limits": [ "No published prospective randomized result", "Old allocation is contradictory", "Small mixed-population academic trial", "Industry owns the larger current trial", "Device and dose differ", "Open-label pain assessment", "Non-thermal ultrasound is a separate mechanism" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Focused heat can destroy local tissue and interrupt pain signalling." }, "animalEvidence": { "rating": "not-applicable", "reason": "Direct human symptom evidence is available." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "The route decision turns on human symptoms, harm and survival." }, "reproducibility": { "rating": "weak", "reason": "Pain signals repeat, while controlled treatment value has not been published." }, "commercialConflicts": { "rating": "high", "reason": "Several devices, sponsors and data owners sit across the evidence chain." }, "recency": { "rating": "current", "reason": "Registry status is current through September 2026." } } }, "layout": { "x": 730, "y": 654 }, "updated": "2026-09-15" }, { "id": "trial-pdac-hipec-action-map", "title": "PDAC HIPEC action map: local control does not license a survival claim", "shortTitle": "HIPEC action map", "type": "trial", "status": "active", "scope": [ "pdac", "hipec", "resectable", "peritoneal-metastases", "capital-allocation" ], "summary": "Adjuvant HIPEC changed local recurrence without extending survival; the active metastatic study needs independent route measurement, not mission-funded procedures.", "content": "Route 1, mechanism. After visible disease is removed, heated liquid chemotherapy circulates over abdominal surfaces to kill exposed microscopic deposits. Route 2, randomized boundary. In 42 resected people, local recurrence was 2/21 with gemcitabine HIPEC versus 11/21 without it. Median survival was 17.1 versus 18.0 months and disease-free survival 14 versus 10 months, neither different. One hospital death occurred in each arm. Route 3, selected metastatic signal. Eighteen people with low-volume peritoneal-only spread and at least six months of chemotherapy response underwent cytoreduction, HIPEC and mixed treatment of the pancreatic primary. Median survival was 26 months. Eight had severe complications, seven were readmitted and one died within 30 days. A ten-person retrospective series reported 24.2 months and two reoperations. Route 4, attribution. Long response, low burden, lack of disease elsewhere, complete surgery, HIPEC and other local treatment all moved together. No component can own the survival result. Route 5, live study. NCT04858009 is recruiting 40 at Mayo, repeats the selected single-arm design, permits up to five HIPEC procedures and estimates July 2029 completion. Route 6, capital. Put £0 into procedures, equipment or extra enrollment. Release at most £350,000 for an independent all-screened route register, procedure dosimetry, harm capture, systemic-treatment timing and recovery of the old randomization. Route 7, stop. Require 90% screened accounting, 100% enrolled accounting, 90% complete drug-temperature-time records, 90% systemic-treatment dates, no more than 5% 90-day mortality and no more than 35% severe harm. Route 8, future proof. Only after the live study reports may a randomized HIPEC component test compete with the existing local-consolidation programme. Hold the surgery decision and systemic care identical, randomize HIPEC alone and start the clock before operation. Route 9, endpoint. Quality-adjusted survival, hospital-free days, symptoms and systemic-dose intensity outrank local recurrence. Route 10, return. The asset is a verified selection-and-delivery dataset and a causal answer, not a heated-chemotherapy theatre programme.", "sourceIds": [ "padilla-valverde-2024-adjuvant-hipec-rct", "grotz-2022-crs-hipec-pdac-pilot", "clinicaltrials-nct04858009-hipec-pdac", "yan-2024-crs-hipec-pdac-retrospective" ], "links": [ { "target": "trials", "relation": "part-of" }, { "target": "claim-hipec-local-control-without-survival-proof", "relation": "maps" }, { "target": "failure-hipec-local-endpoint-and-package-selection", "relation": "maps" }, { "target": "unknown-hipec-added-value-peritoneal-pdac", "relation": "resolves" }, { "target": "hypothesis-hipec-existing-trial-measurement-gate", "relation": "implements" }, { "target": "trial-pdac-pipac-action-map", "relation": "compares-with" }, { "target": "trial-pdac-irreversible-electroporation-action-map", "relation": "compares-with" }, { "target": "trial-pdac-perioperative-sequencing-action-map", "relation": "compares-with" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "weak", "randomised": "moderate — small adjuvant trial only", "independentReplication": "none for survival", "sampleSize": "Adjuvant randomized n=42; metastatic pilot n=18; live target n=40; zero mission-funded participants; cap £350,000", "effect": "Local recurrence reduction without survival gain; selected metastatic survival unassigned", "limits": [ "Small randomized study", "Different resectable and metastatic questions", "No metastatic comparator", "Extreme selection", "Mixed local treatments", "High serious harm", "Live study ends 2029" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The local drug-and-heat effect is credible and consistent with lower local recurrence." }, "animalEvidence": { "rating": "not-applicable", "reason": "Human local-control evidence already exists." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Patient benefit and component attribution govern the decision." }, "reproducibility": { "rating": "weak", "reason": "Feasibility is repeated, but survival contribution is not." }, "commercialConflicts": { "rating": "low", "reason": "No equity or device purchase is justified; limited funds buy independent measurement." }, "recency": { "rating": "current", "reason": "The action map uses a September 2026 live registry update." } } }, "layout": { "x": 1132, "y": 1056 }, "updated": "2026-09-15" }, { "id": "trial-pdac-histotripsy-action-map", "title": "PDAC histotripsy action map: prove the control loop before clinical benefit", "shortTitle": "Histotripsy action map", "type": "trial", "status": "active", "scope": [ "pdac", "locally-advanced", "oligometastatic", "local-treatment" ], "summary": "Histotripsy has a credible non-thermal mechanism and a dangerous visibility failure; the first investment is independent control-loop proof inside the active study, not a survival claim.", "content": "Route 1, mechanism. Short ultrasound pulses create and collapse a bubble cloud, mechanically fragmenting tissue with little intended heat. Route 2, pancreas model. In nine healthy pigs, 6/9 planned zones landed in pancreas and bubble clouds were visible in 5/9. Route 3, failure. Poor visibility produced three off-target hits; septic peritonitis and gastrointestinal blockage required early killing in two pigs. Two possible pancreatitis events, one portal-vein clot and two skin burns were observed. Route 4, regulatory boundary. FDA authorized Edison to destroy liver tumours, while stating that safety and effectiveness for cancer treatment were not established. That is neither pancreatic authorization nor survival proof. Route 5, human study. GANNON is an industry-sponsored, 50-person, single-arm safety study. It selects visible targets at least one centimetre from bowel and reports 30-day complications as its only primary outcome. Route 6, action. Invest £0 in HistoSonics equity, £0 in a device and £0 in extra participants. Offer at most £400,000 for independent monitoring of the first ten existing cases only after raw export, pause authority and publication rights. Route 7, release. Require 9/10 cases with target and cloud visible for 95% of pulses, at least 90% safe planned-volume coverage and no severe pancreas or bowel event. Route 8, separation. Do not pool mechanical histotripsy with thermal HIFU, liver treatment, nerve ablation or inserted-probe ablation. Even a safe gate only licenses completion of feasibility; patient benefit still requires the same systemic care in both randomized arms.", "sourceIds": [ "gannon-2026-pancreas-histotripsy-safety", "clinicaltrials-nct06282809-gannon", "fda-den220087-edison" ], "links": [ { "target": "trials", "relation": "part-of" }, { "target": "claim-histotripsy-pancreas-visibility-limited", "relation": "maps" }, { "target": "failure-histotripsy-acoustic-window-and-translation", "relation": "maps" }, { "target": "unknown-histotripsy-human-coverage-and-benefit", "relation": "resolves" }, { "target": "hypothesis-gannon-independent-safety-gate", "relation": "implements" }, { "target": "company-histosonics", "relation": "involves" }, { "target": "trial-pdac-hifu-action-map", "relation": "compares-with" }, { "target": "trial-pdac-eus-rfa-action-map", "relation": "compares-with" } ], "evidence": { "humanProspective": "none — first trial recruiting", "humanRetrospective": "none", "randomised": "none", "independentReplication": "weak", "sampleSize": "Animal n=9; human target n=50; embedded gate first n=10; no added participants; cap £400,000", "effect": "Credible tissue fragmentation but no human pancreatic safety or benefit result", "limits": [ "Animal target was healthy and small", "Human study is selected and uncontrolled", "Short primary safety window", "Company sponsorship", "Unknown screening denominator", "No quality-of-life primary outcome", "No survival comparator" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Mechanical fragmentation by focused cavitation is observed and measurable." }, "animalEvidence": { "rating": "moderate", "reason": "The study directly demonstrates both targeting and severe mistargeting." }, "inVitroEvidence": { "rating": "strong", "reason": "The pressure field, bubble cloud and treatment volume can be instrumented." }, "reproducibility": { "rating": "weak", "reason": "Pancreas evidence remains concentrated in one sponsor-linked programme." }, "commercialConflicts": { "rating": "high", "reason": "The manufacturer sponsors the human study and supported the animal work." }, "recency": { "rating": "current", "reason": "The action map uses the July 2026 registry state." } } }, "layout": { "x": 832, "y": 756 }, "updated": "2026-09-15" }, { "id": "trial-pdac-intraarterial-delivery-action-map", "title": "PDAC intra-arterial delivery action map: finish the package test and cap the mechanism claim", "shortTitle": "Intra-arterial action map", "type": "trial", "status": "active", "scope": [ "pdac", "trial-delivery", "transfer-unproven" ], "summary": "TIGeR-PaC already owns the survival question; its final record can decide the treatment package, while a separate tissue measurement is required before claiming platform delivery.", "content": "Route 1, patient benefit. TIGeR-PaC randomizes only stable or responding non-surgical candidates after about four months of induction treatment. It compares intra-arterial gemcitabine alone every two weeks with intravenous gemcitabine plus nab-paclitaxel on a three-weeks-on schedule. Final overall survival from randomization can decide which whole package is better for that narrow patient state. It cannot isolate the catheter, route, missing second drug or schedule. Route 2, interim signal. Among the first 45 assigned people, median survival was 16 versus 10 months, P=0.051, without a reported hazard ratio or confidence interval. The second planned interim result is withheld. Two monitoring recommendations to continue are process checks, not efficacy results. Route 3, proposed mechanism. A 16-person blood study found lower active-gemcitabine exposure after artery delivery, 4.9 versus 8.8 hour-micrograms/mL, P=0.018. It did not measure tumour drug, spatial coverage or active intracellular metabolites, so the stronger delivery claim remains open. Route 4, harm. The early 23-person safety set included sepsis in 17.4% and an ischemic injury requiring surgery. Final comparison must add infection, access, vessel and organ injury to systemic chemotherapy harm, hospital time and quality of life. Route 5, action. Enrollment closed in August 2026 with 78 of 86 deaths reported and top-line data projected in the second half of 2027. Spend £0 on another cohort and £0 on RenovoRx equity. Use at most £250,000 for independent all-randomized analysis after lock and access. Adopt only if the locked primary rule and one mission benefit gate pass; do not extend the platform claim beyond gemcitabine in this selected state without direct human target-tissue evidence.", "sourceIds": [ "clinicaltrials-nct03257033-tigerpac", "hatoum-2024-tamp-rr1-rr2", "novelli-2026-tigerpac-pk", "renovorx-2023-tigerpac-interim", "renovorx-2025-10k", "renovorx-2026-full-enrollment", "renovorx-2026-q2-10q" ], "links": [ { "target": "trials", "relation": "part-of" }, { "target": "claim-tamp-blood-exposure-not-tumour-delivery", "relation": "maps" }, { "target": "failure-tigerpac-component-and-interim-inference", "relation": "maps" }, { "target": "unknown-tigerpac-package-survival-and-delivery", "relation": "resolves" }, { "target": "hypothesis-tigerpac-existing-final-gate", "relation": "implements" }, { "target": "company-renovorx", "relation": "involves" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "weak", "randomised": "strong", "independentReplication": "weak", "sampleSize": "Early pooled n=43, safety n=23, PK n=16, interim randomized n=45, final randomized target n=114", "effect": "Existing phase 3 can estimate package value; no new trial is justified before its final result", "limits": [ "No final result", "Final randomized denominator not disclosed", "Open-label design", "Package comparison cannot isolate delivery", "Tumour active-drug concentration unmeasured", "Sponsor controls current reporting" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The physical route and altered blood exposure support a testable delivery model." }, "animalEvidence": { "rating": "moderate", "reason": "Vessel transport work supports feasibility but not cancer-cell delivery or survival." }, "inVitroEvidence": { "rating": "strong", "reason": "Gemcitabine action is established and does not remove the route-measurement gap." }, "reproducibility": { "rating": "weak", "reason": "No independent modern randomized result reproduces the package or delivery claim." }, "commercialConflicts": { "rating": "high", "reason": "The sponsor owns the catheter, funded the supporting studies and controls the phase 3 result." }, "recency": { "rating": "current", "reason": "The map uses August 2026 registry, paper, filing and enrollment records." } } }, "layout": { "x": 544, "y": 468 }, "updated": "2026-09-15" }, { "id": "trial-pdac-irreversible-electroporation-action-map", "title": "PDAC irreversible electroporation action map: recover the randomized answer", "shortTitle": "IRE action map", "type": "trial", "status": "active", "scope": [ "pdac", "local-control", "trial-reporting" ], "summary": "NanoKnife has a sound physical mechanism and a selected registry signal, but treatment policy turns on the missing same-chemotherapy phase 3 result.", "content": "Route 1, mechanism. Irreversible electroporation uses needle electrodes and short high-voltage pulses to make permanent membrane holes and kill cells in a local volume. General anaesthesia, full muscle paralysis and heart-synchronized pulses make it an invasive procedure, not a harmless field. Route 2, completed random evidence. CROSSFIRE assigned 68 people after FOLFIRINOX to radiotherapy or IRE and stopped for futility; median survival was 16.1 versus 12.5 months, IRE HR 1.39, 95% CI 0.84-2.30. It does not contain a chemotherapy-only control. Route 3, observational claim. The DIRECT registry preprint reports 18 versus 10 months from enrollment, but IRE patients entered after eight months from diagnosis versus four for controls. Only ten of 38 controls were enrolled prospectively. This is useful for describing procedures and harm, not for causal survival. Route 4, decisive experiment. Separate study NCT03899636 randomizes the same mFOLFIRINOX route with or without added IRE after three months without progression. The public record still lists 528 estimated participants, an estimated April 2025 completion, active-not-recruiting status and no result. Estimated is not actual. Route 5, feasibility. UK LAP-PIE completed recruitment in November 2024 but has no posted result; its 50-person design tests whether a later independent comparison can recruit and deliver, not whether survival improves. Route 6, action. Spend £0 on a new cohort and £0 on AngioDynamics equity. Offer at most £250,000 for independent recovery and all-randomized analysis of NCT03899636 after data and publication rights are signed. Keep NICE's research-only boundary until the locked primary endpoint and mission survival, harm and quality-of-life gates pass. If the record cannot support a valid analysis, publish why and inspect LAP-PIE before designing or costing a replacement.", "sourceIds": [ "nice-2017-ire-pancreatic", "timmer-2024-crossfire", "clinicaltrials-nct03899636-direct-rct", "clinicaltrials-nct03899649-direct-registry", "martin-2024-direct-registry-safety", "martin-2026-direct-registry-survival-preprint", "rai-2022-lappie-protocol", "isrctn-14986389-lappie", "angiodynamics-2026-10k" ], "links": [ { "target": "trials", "relation": "part-of" }, { "target": "claim-ire-ablation-without-proven-added-survival", "relation": "maps" }, { "target": "failure-direct-selection-and-randomized-reporting-gap", "relation": "maps" }, { "target": "unknown-ire-added-survival-after-modern-chemotherapy", "relation": "resolves" }, { "target": "hypothesis-direct-randomized-data-recovery-gate", "relation": "implements" }, { "target": "company-angiodynamics", "relation": "involves" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "moderate", "randomised": "strong", "independentReplication": "weak", "sampleSize": "CROSSFIRE n=68; DIRECT registry survival n=137; phase 3 target n=528 estimated; LAP-PIE target n=50", "effect": "No new clinical capital before the existing randomized record is recovered", "limits": [ "No phase 3 result", "Actual phase 3 count unknown", "Registry is selected", "LAP-PIE result missing", "Procedure and access route vary", "Sponsor owns the device and evidence" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The device creates a measurable local ablation, but systemic disease can erase its value." }, "animalEvidence": { "rating": "moderate", "reason": "Animal tissue work helps define delivery and harm, not survival." }, "inVitroEvidence": { "rating": "strong", "reason": "Electric-field cell death is established." }, "reproducibility": { "rating": "weak", "reason": "The clean added-treatment comparison has not been reported." }, "commercialConflicts": { "rating": "high", "reason": "AngioDynamics funded the evidence and owns NanoKnife." }, "recency": { "rating": "current", "reason": "The map uses 2026 company, preprint and current registry records." } } }, "layout": { "x": 580, "y": 504 }, "updated": "2026-09-15" }, { "id": "trial-pdac-k912-sonodynamic-action-map", "title": "PDAC K-912 action map: identify the interaction before funding the package", "shortTitle": "K-912 sonodynamic action map", "type": "trial", "status": "active", "scope": [ "pdac", "cholangiocarcinoma", "sonodynamic-therapy", "capital-allocation" ], "summary": "The package has a first-human scan signal and no identified sound-drug interaction; spend at most £450,000 on component proof and zero on another patient cohort.", "content": "Route 1, mechanism claim. K-912 carries epirubicin in a polymer particle; focused ultrasound supplies heat, cavitation and possibly drug release. Route 2, human package. Eleven pancreatic and one cholangiocarcinoma participant received both parts. Nine had scan-defined coagulation, five tumour shrinkage, eight disease control and two of six pain improvement at one month. Route 3, safety. No grade 3-plus event appeared through 30 days; one mild burn and one moderate biliary laboratory change were linked to ultrasound. Repeat treatment and late cumulative drug injury were not tested. Route 4, patient outcome. All had peritoneal spread. Pancreatic median survival calculated from the individual table was 140 days. Complete local coagulation coexisted with progressive disease and 74-day survival in one person and 37-day survival in another. Route 5, attribution. Every person received drug plus ultrasound. All four complete coagulations occurred at 150 watts, while raising K-912 dose did not increase coagulation in these tiny groups. Route 6, capital. Buy £0 of company equity, drug or devices and enroll nobody. Release at most £450,000 for all-12 data recovery and independent four-way component testing. Route 7, release. Require twofold ultrasound-linked free-drug increase and at least 15 points of excess combination killing in three of four patient models after matching temperature, without loss of normal-cell safety. Failure retires the sonodynamic label. Route 8, separate assets. Sonire's thermal Suizenji programme is not the K-912 combination. Its 90-person randomized data and ten-person U.S. feasibility study must be judged under the HIFU route. Route 9, return. A real interaction is protectable know-how; a renamed sum of heat plus chemotherapy is not.", "sourceIds": [ "muragaki-2026-k912-sonodynamic", "clinicaltrials-nct05601323-suizenji", "clinicaltrials-nct07033689-sunrise2", "sonire-2026-series-a", "sonire-2026-nedo-award" ], "links": [ { "target": "trials", "relation": "part-of" }, { "target": "claim-k912-first-human-local-signal-no-component-proof", "relation": "maps" }, { "target": "failure-k912-component-and-local-endpoint", "relation": "maps" }, { "target": "unknown-k912-ultrasound-drug-interaction-and-benefit", "relation": "resolves" }, { "target": "hypothesis-k912-component-identification-gate", "relation": "implements" }, { "target": "company-sonire-therapeutics", "relation": "involves" }, { "target": "trial-pdac-hifu-action-map", "relation": "compares-with" }, { "target": "trial-pdac-sonoporation-action-map", "relation": "compares-with" } ], "evidence": { "humanProspective": "weak", "humanRetrospective": "none", "randomised": "none for K-912; separate thermal HIFU randomization exists", "independentReplication": "none", "sampleSize": "K-912 package n=12; four-model component gate; zero new people; cap £450,000", "effect": "Short-term local package activity without interaction proof or durable patient benefit", "limits": [ "Single centre", "Tiny sequential cohorts", "Mixed cancers", "No component controls", "Separate HIFU device programme", "Proprietary inputs may block testing", "Survival calculation is descriptive" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "Heat, drug and acoustic release are plausible but remain inseparable." }, "animalEvidence": { "rating": "moderate", "reason": "Animal work justifies component testing, not clinical expansion." }, "inVitroEvidence": { "rating": "weak", "reason": "Matched-temperature excess effect is not yet established." }, "reproducibility": { "rating": "none", "reason": "One human package study exists." }, "commercialConflicts": { "rating": "moderate", "reason": "Companies control important formulation and device inputs, though public funding supported the human study." }, "recency": { "rating": "current", "reason": "The action map includes the 2026 report and current Sonire programme." } } }, "layout": { "x": 1066, "y": 990 }, "updated": "2026-09-15" }, { "id": "trial-pdac-kras-wildtype-action-map", "title": "KRAS-wild-type pancreatic action map: find the alternative driver and deliver its route", "shortTitle": "KRAS-wild-type action map", "type": "trial", "status": "active", "scope": [ "pdac", "precision-treatment", "gene-fusion", "care-delivery", "capital-allocation" ], "summary": "Use one consecutive testing route, then keep every rare driver, drug, access decision and outcome separate.", "content": "Route 1, entry. Start with every advanced PDAC diagnosis and adequate tumour content. Record failed tissue and unconfirmed pathology. Route 2, branch. Establish KRAS status with a test capable of the common variants. A wild-type call triggers review for technical failure and an alternative-driver search; it is not itself a treatment marker. Route 3, RNA. Reflex valid KRAS-wild-type or otherwise driver-negative tumours to a structural RNA panel. Record exact coverage for NTRK, ALK, ROS1, RET, FGFR2, BRAF, NRG1 and other supported targets. Route 4, human evidence. NTRK, MET, ROS1, RET, BRAF and NRG1 pancreatic responses exist, but mostly as cases. Larotrectinib's 75% pooled response came from 55 people across 12 cancers and is not a pancreatic rate. Route 5, survival boundary. Know Your Tumor reported median survival of 2.58 versus 1.51 years for matched versus unmatched actionable cases, HR 0.42. Only 46/1,856 referrals reached the matched group and treatment was selected, so the estimate cannot prove profiling utility. Route 6, existing system. England's M219.2 pancreatic RNA route names NTRK1/2/3 as essential and lists several additional targets. NRG1 coverage must be checked at each laboratory. Route 7, scale. Reuse the existing 5,000-person, 20-centre NRG1 route. About 535 people would be KRAS wild type at 10.7%; a 20% fusion assumption suggests about 107 fusion candidates, but actual all-target yield and drug eligibility must be measured. Route 8, gates. Require 95% valid KRAS, 90% reflex RNA, 85% results inside 14 days, 90% seven-day access decisions, no more than 10% split-sample disagreement and no more than 5% standard-treatment delay. Route 9, outcomes. Report each target separately from diagnosis through medicine, response, resistance, quality of life and survival. A pooled matched group is secondary only. Route 10, capital. This broadens the existing £1.5m route and £150,000 NRG1 data line; added capital is £0. Buy no assay company, no drug stock and no new rare-driver cohort. Use existing approved access or target-specific studies and publish every route loss.", "sourceIds": [ "philip-2022-kras-wildtype-pdac", "fusco-2021-kras-wildtype-fusions", "singh-2023-kras-wildtype-drivers", "pishvaian-2020-know-your-tumor", "mehdi-2024-kras-wildtype-case-series", "oreilly-2019-ntrk-pdac-response", "fda-2018-larotrectinib-approval", "fda-2024-bizengri-multidisciplinary-review", "nhs-england-2026-cancer-genomic-test-directory" ], "links": [ { "target": "trials", "relation": "part-of" }, { "target": "trial-pdac-nrg1-action-map", "relation": "extends" }, { "target": "claim-kras-wildtype-pdac-needs-driver-route", "relation": "maps" }, { "target": "failure-kras-wildtype-matching-route-and-selection", "relation": "maps" }, { "target": "unknown-kras-wildtype-driver-route-completion", "relation": "resolves" }, { "target": "hypothesis-kras-wildtype-driver-route-completion", "relation": "implements" }, { "target": "trial-pdac-germline-cascade-action-map", "relation": "coordinates-with" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "strong", "randomised": "weak — no target-agnostic pancreatic utility trial exists", "independentReplication": "moderate — frequency and cases repeat, but treatment groups are tiny", "sampleSize": "Profiling n=2,483, n=795, n=100 and n=1,856 referred; existing route proposal n=5,000; added capital £0", "effect": "Immediate reflex-testing route for a 10.7% branch, without pooling target-specific efficacy", "limits": [ "Rare drivers", "Assay coverage varies", "Mostly case-level treatment evidence", "Tumour-wide approval transfer", "Drug access", "Selected survival comparison" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The route searches for a measured replacement driver when common KRAS is absent." }, "animalEvidence": { "rating": "moderate", "reason": "Models support some drivers, while action must follow human evidence." }, "inVitroEvidence": { "rating": "strong", "reason": "Selected alternative drivers show direct drug sensitivity in tumour models." }, "reproducibility": { "rating": "moderate", "reason": "Multiple cohorts reproduce enrichment, but patient benefit remains target-specific and sparse." }, "commercialConflicts": { "rating": "moderate", "reason": "Independent route analysis limits but cannot remove test and drug ownership interests." }, "recency": { "rating": "current", "reason": "The map uses current approvals and July 2026 English testing coverage." } } }, "layout": { "x": 2260, "y": 1720 }, "updated": "2026-09-15" }, { "id": "trial-pdac-laser-ablation-action-map", "title": "PDAC laser-ablation action map: recover failed studies, do not restart the hardware loop", "shortTitle": "Laser-ablation action map", "type": "trial", "status": "active", "scope": [ "pdac", "laser-ablation", "local-treatment", "capital-allocation" ], "summary": "The next decision costs at most £350,000 and uses existing records; a new laser cohort has no case until leakage, thermal coverage, recruitment and device continuity are resolved.", "content": "Route 1, mechanism. Near-infrared laser energy becomes heat around an inserted fibre and coagulates tissue. This is thermal ablation, not photodynamic chemistry. Route 2, endoscopic signal. Nine selected people completed one 1064 nm treatment without reported harm, producing 0.4 to 6.4 cubic centimetres of dead tissue. Route 3, surgical warning. Thirteen of 15 assigned people received open imILT; three developed pancreatic-fluid leaks and the authors judged the system unsatisfactory. Route 4, programme failure. A randomized comparison stopped at eight for slow recruitment and posted no result. A later endoscopic study stopped at five because the device was unavailable. A ten-person planning study is stale and mixes tumour types. Route 5, commercial reality. Clinical Laserthermia Systems currently leads with neurosurgery and repeat disposable revenue; pancreatic cancer is not named as a current focus. Route 6, capital. Put £0 into company equity, £0 into devices and £0 into a new cohort. Release at most £350,000 only for independent recovery and analysis of up to 47 existing records. Route 7, release conditions. Require every assignment, de-duplicated identity, 90% calibration and imaging, a dead-zone error no greater than 25%, explanation of 90% of failed procedures, and full reporting of fistula, admission, treatment delay and death. Stop at 10% serious procedure harm or any missed safety-boundary crossing. Route 8, future efficacy. If the gate passes, the only interpretable trial holds modern chemotherapy identical and randomizes laser as the sole difference. Do not replace effective systemic care with local heat and do not bundle an immune drug before local safety works. Route 9, business decision. The asset is the recovered result and tested heat model, not a laser box or disposable margin.", "sourceIds": [ "dimatteo-2018-eus-laser-ablation", "paiella-2021-imilt-pancreas", "clinicaltrials-nct02702986-imilt-verona", "clinicaltrials-nct02973217-imilt-marseille", "clinicaltrials-nct03187587-imilt-randomized", "clinicaltrials-nct03784417-eus-laser", "clinicaltrials-nct05549960-laser-model", "cls-2026-investment-case" ], "links": [ { "target": "trials", "relation": "part-of" }, { "target": "claim-pancreatic-laser-heat-feasibility-not-benefit", "relation": "maps" }, { "target": "failure-laser-fistula-accrual-and-device-continuity", "relation": "maps" }, { "target": "unknown-laser-ablation-delivery-and-net-benefit", "relation": "resolves" }, { "target": "hypothesis-laser-existing-study-recovery-gate", "relation": "implements" }, { "target": "company-clinical-laserthermia-systems", "relation": "involves" }, { "target": "trial-pdac-pdt-action-map", "relation": "compares-with" }, { "target": "trial-pdac-eus-rfa-action-map", "relation": "compares-with" }, { "target": "trial-pdac-mwa-action-map", "relation": "compares-with" } ], "evidence": { "humanProspective": "weak", "humanRetrospective": "none", "randomised": "weak — terminated with no result", "independentReplication": "weak", "sampleSize": "Five fragmented programmes, up to 47 records; zero new people now; audit cap £350,000", "effect": "Feasible local heat with a 23% fistula signal in one route and repeated programme failure", "limits": [ "Heterogeneous devices", "Small cohorts", "No current controlled outcome", "Some records may overlap", "Company focus can change", "Future trial not costed until the gate passes" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Laser-to-heat conversion is certain; selective coverage is the question." }, "animalEvidence": { "rating": "moderate", "reason": "Controlled pancreas lesions exist in animals, but human delivery has failed operationally." }, "inVitroEvidence": { "rating": "strong", "reason": "Thermal response is measurable, but does not answer clinical net benefit." }, "reproducibility": { "rating": "weak", "reason": "No stable pancreatic programme has repeated both delivery and benefit." }, "commercialConflicts": { "rating": "high", "reason": "The manufacturer is tied to the central studies and earns from equipment and disposables." }, "recency": { "rating": "current", "reason": "Trial records and company positioning were checked in September 2026." } } }, "layout": { "x": 994, "y": 918 }, "updated": "2026-09-15" }, { "id": "trial-pdac-microbiome-action-map", "title": "PDAC microbiome action map: finish controlled tests before funding another product", "shortTitle": "Microbiome action map", "type": "trial", "status": "active", "scope": [ "pdac", "transfer-unproven" ], "summary": "Two randomized programmes can answer the next questions; the small single-arm studies cannot establish treatment contribution.", "content": "Route 1, bacterial gemcitabine deactivation. PANORAMIX is now recruiting at 27 French sites and includes a masked ciprofloxacin-versus-placebo Step 2 with gemcitabine treatment, resistant-organism tracking and gemcitabine/dFdU measurements. Recover the complete protocol because CTIS says 142 planned across both steps while ClinicalTrials.gov says 206 and lists no Step 2 outcomes. Add a fixed tumour bacterial-enzyme assay only if samples, linkage and independent analysis are contractually available. Cap mission spending at £1m; add no participants. Route 2, stool transfer. FMTPanc is a randomized placebo-controlled phase 1 aiming for 28-60 and had enrolled 12 by January 2025; complete safety, engraftment and all-enrolled outcome reporting. CHASe-PDAC plans 20 in one arm and the suspended MD Anderson study plans 10; use them for feasibility and tissue change, not efficacy. Route 3, immune combinations. NCT05462496 plans 25 in one arm after chemotherapy and combines two antibiotics with pembrolizumab; its primary endpoint is change in any one of several T-cell markers. It cannot attribute response to microbes, antibiotic or checkpoint blockade. Route 4, named live bacteria. MRx0518 plus radiation ended at 13 after sponsor insolvency with no result; recover the record before any successor programme. Route 5, fungal treatment. Human tissue and mouse studies do not justify an antifungal clinical cohort. Decision rule. A treatment advances only when intervention, measured microbial function, drug or immune change, patient outcome and harm are joined in the same controlled record. New microbiome product trials receive £0 until the two current randomized programmes report.", "sourceIds": [ "geller-2017-bacterial-gemcitabine", "riquelme-2019-pdac-microbiome-survival", "aykut-2019-pdac-mycobiome", "nakaoka-2024-kestose-pdac", "clinicaltrials-nct03891979-microbiome-pdac", "clinicaltrials-nct04193904-mrx0518-pdac", "clinicaltrials-nct04975217-fmt-pdac", "clinicaltrials-nct05462496-antibiotic-pembro-pdac", "clinicaltrials-nct06393400-chase-pdac", "sharma-2026-fmtpanc", "ctis-panoramix-2024-517766-41", "clinicaltrials-nct07028424-panoramix" ], "links": [ { "target": "trials", "relation": "part-of" }, { "target": "claim-pdac-microbiome-mechanism-without-human-benefit", "relation": "maps" }, { "target": "failure-microbiome-proxy-without-patient-benefit", "relation": "maps" }, { "target": "unknown-microbiome-treatment-contribution-pdac", "relation": "resolves" }, { "target": "hypothesis-bacterial-gemcitabine-deactivation-gate", "relation": "implements" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "moderate", "randomised": "moderate", "independentReplication": "weak", "sampleSize": "Published randomized n=40; active randomized FMT n=28-60 and PANORAMIX n=142 or 206 across both steps; single arms planned n=45", "effect": "No proven clinical contribution; £1m maximum existing-trial assay support and £0 new-product cohorts", "limits": [ "PANORAMIX denominators conflict across registries", "FMTPanc has no result", "Single-arm trials mix several treatments", "Low-biomass tumour samples need contamination controls" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The routes separate drug metabolism, immune change, stool communities, named bacteria and fungi." }, "animalEvidence": { "rating": "strong", "reason": "Each route has model support, kept below human outcome evidence." }, "inVitroEvidence": { "rating": "strong", "reason": "Bacterial drug conversion is directly measurable." }, "reproducibility": { "rating": "weak", "reason": "No human treatment benefit has repeated under control." }, "commercialConflicts": { "rating": "high", "reason": "Live products and stool manufacture depend on sponsors, and one sponsor failure already stopped a trial." }, "recency": { "rating": "current", "reason": "The map uses February-September 2026 trial states." } } }, "layout": { "x": 386, "y": 336 }, "updated": "2026-09-15" }, { "id": "trial-pdac-mwa-action-map", "title": "PDAC microwave action map: recover coverage and outcomes before another trial", "shortTitle": "MWA action map", "type": "trial", "status": "active", "scope": [ "pdac", "unresectable", "pain", "local-treatment" ], "summary": "Microwave ablation has passed antenna placement but not complete coverage or comparative benefit; a £300,000 record audit is the only current investment.", "content": "Route 1, mechanism. An inserted antenna drives water molecules and heats tissue; this is a different field from radiofrequency current, freezing or mechanical histotripsy. Route 2, delivery gap. In 72 people, technical placement was 100% but complete ablation only 30/74 tumours, 40.5%; grade IIIa harm occurred in 7/72. Route 3, palliative signal. Pain improved in 32/33 without a control, while median follow-up was 4.6 months and survival 5.6 months. Route 4, selection shortcut. In a 150-person record, microwave cases started with tumours 31% smaller and metastases 20 points less often. Median survival was 14.6 versus 9.0 months, but local progression among imaged people was 10.6% versus 9.6% and 20/67 microwave scans were missing. Route 5, package harm. Of 12 people given dual immune drugs, biopsy and planned surgical microwave plus gemcitabine, eight completed the package and one died after a fistula and bleeding. Route 6, dormant comparison. NCT06231160 planned 185 prospectively observed cases without random assignment and has unknown status. Route 7, action. Invest £0 in equipment and £0 in new participants. Spend at most £300,000 to recover and align at least 150 existing records under independent custody. Route 8, release. Require 95% vital status, 90% imaging, 80% energy-log coverage and stable estimates across declared adjustment models. Even a passing audit only permits comparison against the cryo palliative route; it cannot release a survival claim.", "sourceIds": [ "zhou-2024-pancreatic-mwa", "vogl-2026-tace-mwa", "topal-2025-mwa-dual-ici", "clinicaltrials-nct06231160-mwa-cohort" ], "links": [ { "target": "trials", "relation": "part-of" }, { "target": "claim-mwa-pain-signal-no-added-benefit", "relation": "maps" }, { "target": "failure-mwa-selection-bundling-and-missing-imaging", "relation": "maps" }, { "target": "unknown-mwa-safe-coverage-and-patient-benefit", "relation": "resolves" }, { "target": "hypothesis-mwa-existing-data-gate", "relation": "implements" }, { "target": "trial-pdac-cryoablation-action-map", "relation": "compares-with" }, { "target": "trial-pdac-rfa-action-map", "relation": "compares-with" }, { "target": "trial-pdac-histotripsy-action-map", "relation": "compares-with" } ], "evidence": { "humanProspective": "weak", "humanRetrospective": "moderate", "randomised": "none", "independentReplication": "weak", "sampleSize": "Existing n=72, n=150 and n=12; dormant target n=185; zero new participants; cap £300,000", "effect": "Strong uncontrolled pain signal; low complete ablation; no causal survival estimate", "limits": [ "Allocation by tumour burden", "Microwave bundled with other treatments", "Imaging missing asymmetrically", "Short follow-up", "One fatal treatment-path event", "Unknown prospective cohort state", "Audit cannot eliminate unknown confounding" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Microwave heating is a direct local tissue-destruction mechanism." }, "animalEvidence": { "rating": "moderate", "reason": "Thermal field behavior is known, but the clinical delivery gap remains." }, "inVitroEvidence": { "rating": "strong", "reason": "Energy, temperature and cell death can be instrumented." }, "reproducibility": { "rating": "weak", "reason": "Technical access repeats; comparative patient benefit does not." }, "commercialConflicts": { "rating": "moderate", "reason": "Supplier-neutral specifications and independent custody are required." }, "recency": { "rating": "current", "reason": "The map includes 2025 and 2026 outcome reports." } } }, "layout": { "x": 898, "y": 822 }, "updated": "2026-09-15" }, { "id": "trial-pdac-nrg1-action-map", "title": "PDAC NRG1 action map: deploy testing and recover the overdue benefit record", "shortTitle": "NRG1 action map", "type": "trial", "status": "active", "scope": [ "pdac", "nrg1", "rna-testing", "regulatory", "capital-allocation" ], "summary": "The drug exists and the fusion is findable; capital belongs in route completion and independent follow-up, not another molecule or cohort.", "content": "Route 1, biology. An NRG1 fusion drives HER3-HER2 growth signalling in a rare, mostly KRAS-wild-type PDAC subgroup. Route 2, treatment. Zenocutuzumab blocks HER2 and HER3. FDA's 30-person pancreatic set had 12 responses, 40%, but 24 had stopped treatment by cutoff. Route 3, approval boundary. Accelerated approval permits use after prior systemic therapy; it does not prove survival. The at-least-50-person FDA follow-up report was due in August 2026 and remains listed as ongoing. Route 4, detection. DNA can miss NRG1's varied fusion partners and broad breakpoint space. RNA testing should follow every KRAS-wild-type or otherwise driver-negative result. Route 5, English system. NHS M219.2 already provides a pancreatic RNA structural-variant route, but NRG1 is not an explicit named target. Verify every laboratory before claiming absence. Route 6, denominator. In 5,000 advanced PDAC diagnoses, published rates imply about 535 KRAS-wild-type tumours and about 50 NRG1 fusions. Route 7, experiment. Spend at most £1.5m over two years to add explicit panel coverage and measure tissue, RAS, RNA, report, review, access, infusion and outcomes at at least 20 centres. Route 8, drug evidence. Spend at most £150,000 to obtain, independently analyse and publish the completed eNRGy pancreatic patient-level record, including all screened, treated and excluded cases, response dates, progression, later therapy, symptoms, quality of life and survival. Release it only if all patients can be accounted for and the sponsor permits negative publication. Route 9, company. Genmab owns the acquired asset and receives US royalties from Partner Therapeutics; separated pancreatic economics are absent. Buy £0 of either company or product inventory. Route 10, decision. Deploy the testing route now because it reuses an existing system and can find other fusions. Continue zenocutuzumab access for label-eligible positive patients while naming the uncertainty. Do not fund another efficacy cohort; read the overdue one.", "sourceIds": [ "fda-2024-bizengri-approval", "fda-2024-bizengri-multidisciplinary-review", "fda-2026-ongoing-bizengri-accelerated-approval", "clinicaltrials-nct02912949-enrgy", "schram-2025-zenocutuzumab-nrg1", "philip-2022-kras-wildtype-pdac", "fusco-2021-kras-wildtype-fusions", "schram-2026-nrg1-fusion-detection", "nhs-england-2026-cancer-genomic-test-directory", "genmab-2026-q1-bizengri" ], "links": [ { "target": "trials", "relation": "part-of" }, { "target": "claim-nrg1-approved-response-not-durable-control", "relation": "maps" }, { "target": "failure-nrg1-testing-and-confirmation-route", "relation": "maps" }, { "target": "unknown-nrg1-identification-and-survival-benefit", "relation": "resolves" }, { "target": "hypothesis-nrg1-rna-route-completion", "relation": "implements" }, { "target": "company-genmab-partner-bizengri", "relation": "involves" }, { "target": "trial-pdac-germline-cascade-action-map", "relation": "shares-infrastructure-with" }, { "target": "trial-cldn182-antibody-action-map", "relation": "contrasts-with" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "moderate", "randomised": "none", "independentReplication": "weak for treatment; moderate for fusion enrichment", "sampleSize": "Drug n=30 FDA set; later paper n=36 pancreatic; route proposal n=5,000 diagnoses; route cap £1.5m plus £150,000 data recovery", "effect": "Approved 40% response signal plus a measurable route to find roughly one eligible tumour per 100 PDAC diagnoses", "limits": [ "No comparative survival", "Rare target", "RNA access varies", "FDA report overdue", "UK drug route unconfirmed", "Company-controlled clinical record", "Historical centre comparison" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Driver, receptor pair, drug and clinical response align." }, "animalEvidence": { "rating": "not-applicable", "reason": "Human treatment evidence exists." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Care delivery and confirmatory follow-up now limit the decision." }, "reproducibility": { "rating": "moderate", "reason": "Detection can be tested across centres; efficacy still comes from one programme." }, "commercialConflicts": { "rating": "high", "reason": "Independent custody is required because product and pivotal trial are company-linked." }, "recency": { "rating": "current", "reason": "The map captures live policy and regulatory state through September 2026." } } }, "layout": { "x": 1234, "y": 1158 }, "updated": "2026-09-15" }, { "id": "trial-pdac-occult-metastasis-action-map", "title": "PDAC hidden-spread action map: direct view, wash cytology and timing answer different questions", "shortTitle": "PDAC hidden-spread trials", "type": "trial", "status": "supported", "scope": [ "pdac" ], "summary": "Modern cohorts show a useful staging signal, but none randomizes the complete test-to-treatment route or settles cytology-only action.", "content": "Before treatment. The Mayo series reviewed 1,004 scan-localized PDAC cases. Staging was positive in 180, 18%: 140 had visible biopsy-confirmed spread and 96 had positive abdominal-wash cytology, with overlap. Among 419 untreated people with wash sampling, 95, 23%, were positive. The separate procedure caused nine complications among 762 people, 1.2%, including one major complication, 0.1%. A 2026 Japanese series found hidden spread in 24/102, 23.5%, with no major complications, but treatment began after a mean 14.8 days and there was no comparison group. UMIN000061741 plans another 200-person single-arm early study whose primary endpoint is detection rate, not useful life. Before immediate surgery. PREOPANC found hidden spread in 13/133, 9.8%, at staging laparoscopy and 9/106, 8.5%, at direct open exploration. The laparoscopy group was more likely to receive later palliative chemotherapy, 76.9% versus 30.0%, but the staging method followed the treatment arm. PREOPANC-2 gives the strongest operational result after neoadjuvant treatment: among 322 explored patients, hidden spread occurred in 39, 12.1%, and caused 90.5% of abandoned resections. Same-session laparoscopy was associated with non-therapeutic laparotomy of 4.5% versus 17.1%, P=.002, reported NNT 8. The staging method was not randomized. Residual failure. In TAPS, 67/238 explorations after modified FOLFIRINOX, 28.2%, produced no resection; hidden spread caused 30/67, 44.8%, and 90-day death was 2/67, 3.0%. This does not show laparoscopy would prevent every failure because local vascular involvement caused the rest. Practice. NICE uses selective staging when small-volume liver or abdominal spread is suspected. A 617-surgeon global survey found 82% use but wide variation in timing, selection, wash cytology and ultrasound. Decision boundary. Do not fund another yield-only cohort. Use the existing English first-scan surgical-route study to compare current selective practice with a locked two-timepoint rule, separate biopsy-confirmed spread from cytology alone, and retain delay, harm, treatment, function and survival. Do not treat a negative camera check as absence of microscopic systemic disease or a positive cytology result as automatic proof that surgery cannot help.", "sourceIds": [ "gudmundsdottir-2023-staging-laparoscopy", "van-dongen-2023-preopanc-staging", "rompen-2026-preopanc2-staging", "theijse-2024-nontherapeutic-laparotomy", "stoop-2026-global-staging-survey", "kikuchi-2026-staging-laparoscopy", "umin-000061741-staging-laparoscopy", "nice-2018-ng85-pancreatic" ], "links": [ { "target": "trials", "relation": "part-of" }, { "target": "claim-staging-laparoscopy-avoids-incision-not-proven-survival", "relation": "tests" }, { "target": "failure-staging-yield-without-patient-utility", "relation": "includes" }, { "target": "unknown-pdac-staging-laparoscopy-rule", "relation": "tests" }, { "target": "hypothesis-pdac-two-timepoint-staging-route", "relation": "prioritizes" }, { "target": "trial-pdac-surgical-route-action-map", "relation": "extends" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "strong", "randomised": "weak — staging use was not randomized", "independentReplication": "moderate", "sampleSize": "Mayo n=1,004; PREOPANC explored n=239; PREOPANC-2 explored n=322; TAPS n=663; Japan n=102; live target n=200", "effect": "Hidden spread about 10% to 24%; PREOPANC-2 non-therapeutic laparotomy 4.5% versus 17.1%", "limits": [ "No randomized staging comparison", "Different disease states and timing", "Expert-centre selection", "Cytology grouped with visible spread", "No complete patient-utility endpoint", "Live study single-arm" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Direct visual inspection and biopsy can reveal surface metastases below scan resolution." }, "animalEvidence": { "rating": "not-applicable", "reason": "The action map depends on human imaging, operations and treatment decisions." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Laboratory performance cannot establish route benefit or procedure harm." }, "reproducibility": { "rating": "moderate", "reason": "Detection and avoided-incision signals repeat, but exact effect sizes and patient selection vary." }, "commercialConflicts": { "rating": "low", "reason": "The central evidence evaluates standard clinical procedures rather than a single marketed device." }, "recency": { "rating": "current", "reason": "The map includes 2026 PREOPANC-2, a current global survey and a live Japanese registry." } } }, "updated": "2026-09-15" }, { "id": "trial-pdac-organoid-action-map", "title": "PDAC organoid action map: useful model, incomplete clinical route", "shortTitle": "Organoid action map", "type": "trial", "status": "mixed", "scope": [ "pdac", "transfer-unproven" ], "summary": "The evidence supports laboratory drug triage and prospective shadow testing now; clinical treatment guidance still needs recovered randomized results or a new whole-route trial.", "content": "Model evidence. Tiriac established heterogeneous chemotherapy response and biological fidelity but did not test a treatment policy. Feasibility under routine sampling. Beutel obtained 28 profiles from 44 confirmed cancers in a median 53 days. HOPE obtained drug-testable cultures from 12 of 76 enrolled people in a median 96 days. ORGANOPREDICT improved take to 54 of 87, 62%, with mean turnaround 6.8 weeks and 34 fully evaluable patients; its high conditional sensitivity and specificity came from routine treatment rather than assigned guidance. Sampling. A paired randomized-order endoscopic study produced organoids in 17 of 50 people; fine-needle biopsy yielded nine unique successes versus two for aspiration, but the difference was not statistically secure and malignancy confirmation remained necessary. Current treatment context. PASS-01 randomized first-line chemotherapy, generated organoids in 50% and showed why second-line routing is late: 75 of 140 received a second line, median duration was 2.1 months, and the non-randomized correlate-guided survival comparison was 5.4 versus 4.4 months, P=.45. System warning. AVATAR routed only four of 81 assigned patients to matched treatment and did not improve assigned-group survival. Direct randomized utility. NCT04931381, planned n=100 advanced disease, and NCT04931394, planned n=200 postoperative disease, are both status unknown after stale July 2022 verification, passed May 2025 completion estimates and no posted results. Active and completed support. NCT05351983 completed 60 without results. PROSPER-PANC is an active observational n=200 pipeline. ADOPT is a recruiting single-site n=25 signal trial. UNITEPANC is a recruiting single-arm route that plans 92 generations, 38 treated and 34 analyzed, with a 12-week postoperative deadline and historical efficacy comparison. Decision: first recover the two randomized Chinese studies and the completed Swiss feasibility data, then join PASS-01, PROSPER-PANC and UNITEPANC under one all-patient route table. Do not start another cohort merely to report take rate or conditional accuracy.", "sourceIds": [ "tiriac-2018-organoid-profiling", "beutel-2021-pdac-organoids", "grossman-2022-hope-organoids", "boileve-2024-organopredict", "wiessner-2024-organoid-biopsy-randomized", "knox-2025-pass01", "sarno-2025-avatar", "clinicaltrials-nct04931381-organoid-advanced", "clinicaltrials-nct04931394-organoid-adjuvant", "clinicaltrials-nct05351983-organoid-screen", "clinicaltrials-nct05927298-prosper-panc", "clinicaltrials-nct06813079-adopt", "ettrich-2026-unitepanc" ], "links": [ { "target": "trials", "relation": "part-of" }, { "target": "claim-pdac-organoid-correlation-not-routing-utility", "relation": "supports" }, { "target": "failure-organoid-treatment-chain-attrition", "relation": "exposes" }, { "target": "unknown-pdac-organoid-guided-treatment-utility", "relation": "maps" }, { "target": "hypothesis-organoid-guided-early-switch", "relation": "gates" }, { "target": "technology-ai-organoid-active-learning", "relation": "gates" }, { "target": "lab-tuveson-organoid-models", "relation": "could-use" }, { "target": "person-david-tuveson", "relation": "could-use" } ], "evidence": { "humanProspective": "strong — multiple prospective feasibility and treatment-linked studies", "humanRetrospective": "moderate", "randomised": "moderate — treatment and sampling studies exist, but public organoid-policy results do not", "independentReplication": "moderate — culture-response correlation repeats but route performance varies", "sampleSize": "Mapped cohorts and trials range from n=25 to n=200; direct randomized organoid-policy plans total 300 estimated participants with no public results", "effect": "No demonstrated randomized useful-life gain; reported culture yield ranges from 15.8% of all enrolled in HOPE to 62% in ORGANOPREDICT", "limits": [ "Trial registries may be stale", "No patient-level data access confirmed", "Methods and drug panels differ", "Published correlations condition on successful cultures", "Adjuvant and advanced routes are not interchangeable", "Historical controls cannot isolate policy benefit" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Living tumour-cell response can plausibly discriminate between otherwise reasonable drugs." }, "animalEvidence": { "rating": "moderate", "reason": "Matched model studies support drug-response fidelity but omit the full clinical route." }, "inVitroEvidence": { "rating": "strong", "reason": "Independent groups reproduce heterogeneous response and some patient correlation." }, "reproducibility": { "rating": "moderate", "reason": "Biological signal repeats, while culture success, timing and complete route delivery do not." }, "commercialConflicts": { "rating": "moderate", "reason": "Independent comparison is needed because platforms, drug panels and laboratory services create adoption incentives." }, "recency": { "rating": "current", "reason": "Includes trials and results available through 15 September 2026." } } }, "layout": { "x": 124, "y": 178 }, "updated": "2026-09-15" }, { "id": "trial-pdac-p32-action-map", "title": "PDAC phosphorus-32 action map: read the completed randomization before spending again", "shortTitle": "P-32 action map", "type": "trial", "status": "active", "scope": [ "pdac", "locally-advanced", "brachytherapy", "local-treatment" ], "summary": "Modern phosphorus-32 has the most decision-ready local-treatment dataset found in this pass: 88 randomized assignments on identical chemotherapy, awaiting independent analysis.", "content": "Route 1, mechanism. An endoscopic needle deposits silicon phosphorus-32 particles; beta radiation gives a planned local dose while chemotherapy treats the rest of the body. Route 2, pilot. Of 50 people starting chemotherapy, 42 were implanted, early radiation was localized in 40, local control at 16 weeks was 82%, 10 were resected and median survival was 15.2 months. Three implanted people had grade 3-plus device or procedure events. Route 3, observational support. A three-month landmark analysis of 35 combination and 51 standard-care patients estimated 189 extra days within 30 months, but excludes early deaths and retains unknown selection. Route 4, historical warning. Older repeated colloidal phosphorus-32 produced 78% versus 8% tumour liquefaction, 4.2 versus 1.8 serious events per person and survival 7.4 versus 11.5 months. Route 5, decisive asset. TRIPP-FFX randomized 88 people to identical FOLFIRINOX with or without OncoSil. Primary completion was January 2026; no result is posted and final follow-up runs to July 2027. Route 6, action. Invest £0 in OncoSil equity, £0 in devices and £0 in a new cohort. Spend at most £350,000 for independent all-assigned analysis after raw access and publication rights. Route 7, release. Require all assignments, 95% vital status, 90% dose and quality-of-life records, at least 15 points more local control plus progression hazard ratio no worse than 0.75 or two quality-adjusted months, without five points excess severe harm. Route 8, separation. Keep modern microparticles distinct from old colloid, iodine seeds and external radiation.", "sourceIds": [ "clinicaltrials-nct05466799-tripp-ffx", "ross-2022-panco-p32", "lim-2026-p32-landmark", "rosemurgy-2008-p32-randomized" ], "links": [ { "target": "trials", "relation": "part-of" }, { "target": "claim-oncosil-local-control-signal-awaits-randomized-result", "relation": "maps" }, { "target": "failure-p32-liquefaction-survival-and-reporting-gap", "relation": "maps" }, { "target": "unknown-oncosil-randomized-survival-and-delivery", "relation": "resolves" }, { "target": "hypothesis-tripp-ffx-independent-analysis-gate", "relation": "implements" }, { "target": "company-oncosil-medical", "relation": "involves" }, { "target": "trial-pdac-eus-rfa-action-map", "relation": "compares-with" }, { "target": "trial-pdac-mwa-action-map", "relation": "compares-with" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "moderate", "randomised": "moderate — unreported modern dataset plus old negative study", "independentReplication": "weak", "sampleSize": "Pilot n=50; adjusted n=86; modern randomized n=88; old randomized n=30; audit cap £350,000", "effect": "Feasible localized delivery and local-control signal; modern comparative patient benefit unknown", "limits": [ "Industry concentration", "Primary endpoint is local control", "Final follow-up incomplete", "No public randomization result", "Pilot implantation attrition", "Old product differs", "Phase 2 sample limits survival precision" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Localized beta radiation produces a measured tumour dose." }, "animalEvidence": { "rating": "moderate", "reason": "Dosimetry is supported, while human randomization is the key asset." }, "inVitroEvidence": { "rating": "strong", "reason": "Decay, dose and radiation injury are measurable." }, "reproducibility": { "rating": "moderate", "reason": "Multicentre delivery exists; comparative outcome still needs independent analysis." }, "commercialConflicts": { "rating": "high", "reason": "The manufacturer funds, staffs and sponsors the central evidence chain." }, "recency": { "rating": "current", "reason": "TRIPP-FFX status was checked in September 2026." } } }, "layout": { "x": 928, "y": 852 }, "updated": "2026-09-15" }, { "id": "trial-pdac-palliative-care-action-map", "title": "PDAC palliative-care action map: measure good days, not referral counts", "shortTitle": "PDAC good-days action map", "type": "trial", "status": "active", "scope": [ "pdac", "palliative-care", "quality-of-life", "care-delivery", "capital-allocation" ], "summary": "Use early supportive care now, recover the hidden randomized result, and test a defined route with death-aware outcomes before scaling another service bundle.", "content": "Route 1, definition. Palliative care actively manages symptoms, function, mood, family load and treatment decisions while cancer treatment continues. Route 2, proven boundary. A 207-person pancreatic randomized trial improved quality scores and reduced final-month chemotherapy by 9.1 points; it did not extend survival. Route 3, contradiction. A 288-person pancreatic-and-biliary trial missed both main pain and depression endpoints. Component, trigger and comparator matter. Route 4, analysis. Fixed-week questionnaires select people still alive and able to answer. Use quality-weighted days alive and at home, plus all-assigned hospital, treatment and death records. Route 5, existing answer. NCT04632303 randomized an estimated 250 pancreatic patients, completed its main follow-up in November 2024 and has no public result. Spend at most £150,000 on full independent recovery before another efficacy study. Route 6, implementation. Spend at most £2m across 20 NHS centres and 1,500 consecutive unresectable PDAC patients for automatic multi-need screening, seven-day specialist response for high need, embedded review for all others and cycle-by-cycle checks. Route 7, safeguards. Do not delay or withhold beneficial tumour treatment. Audit opioid toxicity, psychiatric crisis, access gaps and caregiver burden. Route 8, active duplication. A 60-person digital community-worker pilot remained recruiting after its estimated completion date. Ask for its result; do not treat a digital layer as the intervention. Route 9, success. Require at least ten extra quality-weighted days over 24 weeks, an eight-point fall in chemotherapy in the last 30 days, at least 90% screening and no survival-harm signal. Route 10, capital. Palliative care is a care-system purchase, not an oncology-company investment. Fund teams, data and route ownership; buy no platform or branded referral product.", "sourceIds": [ "maltoni-2016-systematic-palliative-qol", "maltoni-2016-systematic-palliative-end-of-life", "woo-2019-pancreatobiliary-palliative-rct", "dufva-2026-home-palliative-pdac", "clinicaltrials-nct04632303-pdac-palliative", "clinicaltrials-nct07591571-pdac-community-worker", "sanders-2024-asco-palliative-guideline" ], "links": [ { "target": "trials", "relation": "part-of" }, { "target": "claim-pdac-palliative-care-improves-qol-not-survival", "relation": "maps" }, { "target": "failure-pdac-palliative-timing-and-analysis", "relation": "maps" }, { "target": "unknown-pdac-palliative-route-net-benefit", "relation": "resolves" }, { "target": "hypothesis-pdac-palliative-route-completion", "relation": "implements" }, { "target": "trial-pdac-cachexia-action-map", "relation": "coordinates-with" }, { "target": "trial-pert-pdac-action-map", "relation": "coordinates-with" }, { "target": "trial-pdac-celiac-pain-action-map", "relation": "coordinates-with" }, { "target": "trial-older-pdac-action-map", "relation": "coordinates-with" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "moderate", "randomised": "moderate and mixed", "independentReplication": "moderate — mixed results", "sampleSize": "Published randomized n=207 and n=288; completed-primary target n=250; route proposal n=1,500; cap £2.15m", "effect": "Bounded quality and end-of-life-care benefit, no proven survival effect, with a current unreleased randomized answer", "limits": [ "Heterogeneous packages", "Open-label", "Patient reporting attrition", "Mixed tumour cohorts", "Workforce capacity", "Historical survival confounding", "No current English completion denominator" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Pain, nutrition, mood, function and decisions are direct patient failure modes." }, "animalEvidence": { "rating": "not-applicable", "reason": "The intervention is human care." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Quality-weighted time and route completion require patient data." }, "reproducibility": { "rating": "moderate", "reason": "There are multiple trials, but the exact active package remains unresolved." }, "commercialConflicts": { "rating": "low", "reason": "The mission purchases independent care capacity and measurement." }, "recency": { "rating": "current", "reason": "Live and completed-primary trial status was checked in September 2026." } } }, "layout": { "x": 1270, "y": 1194 }, "updated": "2026-09-15" }, { "id": "trial-pdac-particle-therapy-action-map", "title": "PDAC particle-therapy action map: prove patient benefit beyond beam shape", "shortTitle": "Particle-therapy action map", "type": "trial", "status": "active", "scope": [ "pdac", "locally-advanced", "proton", "carbon-ion" ], "summary": "Particle physics supports better dose placement; pancreatic cancer evidence supports measurement and data release, not survival claims or new facility capital.", "content": "Route 1, physics. Protons and carbon ions slow in tissue and deposit a large share of energy near the end of their range. This can spare tissue beyond the target. Carbon ions have higher linear energy transfer, meaning denser energy deposited along a short path, and use a modelled relative biological effectiveness. Route 2, moving target. The pancreas moves with breathing, while stomach and bowel shape and gas change. Those changes alter the beam path and can move high dose away from tumour or into bowel. Daily delivered dose matters more than an ideal plan. Route 3, carbon evidence. Dose-escalation and multicentre cohorts report two-year local control near or above 80%, but no concurrent control. A 2026 84-person series reported 94.7% one-year local control and 44.5% freedom from distant spread. Route 4, proton evidence. A 15-person phase 2 produced 47% one-year survival and missed its 75% target. A 19-person prospective registry reported low severe toxicity and 14.6-month median survival without a photon control. Route 5, failed comparison. CIPHER, the only found phase 3 carbon-versus-photon plan, enrolled zero. It also changed fractionation and concurrent chemotherapy, so it would not have isolated particle type cleanly. Route 6, live evidence debt. PACK completed 25 carbon-ion participants without posted results. PAN009-18 remains a one-arm proton trial, targets 60 people and now estimates primary completion in 2039. Route 7, action. Spend £0 on a new facility, £0 on another cohort and at most £500,000 to recover complete existing outcomes and recalculate delivered dose under common rules. Require a locked systemic-risk selector because local control cannot beat distant failure. Route 8, future gate. A current-care randomized trial is considered only if existing data show a reproducible delivered-dose advantage, at least a five-point severe-harm or material treatment-delivery gain, transferable centre quality and two-year distant-failure risk below 40% in the selected group. It must keep systemic treatment, eligibility and follow-up identical. Until that gate passes, particle therapy is an engineering capability with unproved incremental patient value.", "sourceIds": [ "shinoto-2016-carbon-ion-gemcitabine", "kawashiro-2018-jcros-carbon-ion", "lai-2026-carbon-ion-risk-adapted", "rapp-2022-proton-phase2", "eckstein-2023-proton-registry", "clinicaltrials-nct02598349-proton", "clinicaltrials-nct03536182-cipher", "clinicaltrials-nct04194268-pack" ], "links": [ { "target": "trials", "relation": "part-of" }, { "target": "claim-particle-therapy-local-control-without-survival-proof", "relation": "maps" }, { "target": "failure-particle-therapy-comparator-and-accrual", "relation": "maps" }, { "target": "unknown-particle-therapy-added-patient-benefit", "relation": "resolves" }, { "target": "hypothesis-particle-therapy-existing-data-gate", "relation": "implements" }, { "target": "failure-local-intensification-pdac-survival-not-shown", "relation": "bounded-by" }, { "target": "hypothesis-selected-local-consolidation", "relation": "depends-on" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "moderate", "randomised": "none", "independentReplication": "moderate", "sampleSize": "Carbon cohorts up to 84; proton cohorts up to 19 reported; CIPHER n=0; PACK n=25 completed; PAN009-18 target n=60", "effect": "Local-control and delivery evidence without comparative overall or quality-adjusted survival", "limits": [ "No randomized result", "Decisive trial failed to start", "Particle comparisons change other treatment components", "Moving anatomy creates range uncertainty", "Carbon biological dose models differ", "Distant progression remains common", "Patient travel and facility opportunity cost are unmeasured" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Beam range and energy-density advantages can be measured directly." }, "animalEvidence": { "rating": "moderate", "reason": "Local damage models do not establish human net benefit." }, "inVitroEvidence": { "rating": "strong", "reason": "Particle radiation has distinct physical and cellular effects." }, "reproducibility": { "rating": "weak", "reason": "Several cohorts reproduce local control, but no comparative benefit is reproduced." }, "commercialConflicts": { "rating": "moderate", "reason": "Facility owners and specialist programmes have adoption incentives." }, "recency": { "rating": "current", "reason": "The map includes 2026 outcomes and live registry status through September 2026." } } }, "layout": { "x": 670, "y": 594 }, "updated": "2026-09-15" }, { "id": "trial-pdac-pdt-action-map", "title": "PDAC photodynamic action map: solve coverage before testing benefit", "shortTitle": "PDT action map", "type": "trial", "status": "active", "scope": [ "pdac", "photodynamic-therapy", "local-treatment", "capital-allocation" ], "summary": "PDT has a repeatable local-kill signal and no causal benefit result; the next £500,000 buys a coverage decision from existing data, not more devices or patients.", "content": "Route 1, physical mechanism. A drug accumulates unevenly, a fibre emits light over a small region and oxygen enables the killing reaction. All three fields must overlap. Route 2, human signal. Studies of 15, 12 and eight people created scan-visible dead tissue, but response varied and no study had a control. Route 3, scale check. In the eight-person series, mean dead-zone diameter was 15.7 mm against mean tumour diameter 33.3 mm; a sphere approximation gives about 10% volume coverage. Route 4, harm and selection. Multiple fibres produced inflammation outside the pancreas; one programme screened 623, started 13 and completed eight. Route 5, current asset. Twenty-four people are enrolled in a single group receiving FOLFIRINOX, pembrolizumab and verteporfin PDT. Main follow-up is planned for March 2028. It may describe the package but cannot isolate PDT. Route 6, capital now. Put £0 into a company, £0 into devices and £0 into another cohort. Release at most £500,000 for independent recovery and spatial analysis of up to 64 existing records, conditional on raw images, calibration files, every starter, blinded external testing and publication rights. Route 7, measurement release. Continue only if median coverage error is at most 20 points, no important off-target injury is missed and at least 70% reach 50% viable-tumour coverage. Route 8, efficacy. If the spatial gate passes and the active cohort has acceptable harm, design a later multi-centre randomization with identical systemic care and PDT as the only difference. A survival design targeting hazard ratio 0.70 needs about 4(1.96+0.84)^2/ln(0.70)^2 = 247 deaths, likely about 280 people after follow-up loss. Do not fund that trial until coverage passes. Route 9, decision. Judge symptoms, quality-adjusted time, distant failure, survival and all harm; dead tissue is an engineering output, not the business return.", "sourceIds": [ "huggett-2014-verteporfin-pdt", "dewitt-2019-eus-porfimer-pdt", "chandrasekhara-2021-eus-verteporfin-pdt", "clinicaltrials-nct03033225-vertpac02", "clinicaltrials-nct01770132-porfimer-pdt", "clinicaltrials-nct06381154-pdt-pembrolizumab" ], "links": [ { "target": "trials", "relation": "part-of" }, { "target": "claim-pdt-local-ablation-without-benefit-proof", "relation": "maps" }, { "target": "failure-pdt-geometry-selection-and-bundling", "relation": "maps" }, { "target": "unknown-pdt-spatial-coverage-and-patient-benefit", "relation": "resolves" }, { "target": "hypothesis-pdt-measured-spatial-gate", "relation": "implements" }, { "target": "trial-pdac-eus-rfa-action-map", "relation": "compares-with" }, { "target": "trial-pdac-histotripsy-action-map", "relation": "compares-with" }, { "target": "trial-pdac-p32-action-map", "relation": "compares-with" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "none", "randomised": "none", "independentReplication": "weak", "sampleSize": "Published n=35; up to 64 existing records for the gate; zero new people now; possible later survival trial about n=280", "effect": "Local dead tissue is repeatable; coverage, symptoms and survival benefit are unproved", "limits": [ "Current combination is single arm", "Existing data may not support common spatial reconstruction", "Death count is an approximation", "Later trial cost is not yet authorized", "PDT may be palliative rather than survival-improving", "Systemic disease lies outside the light field" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The programme starts from measured drug, light, oxygen and anatomy constraints." }, "animalEvidence": { "rating": "moderate", "reason": "Animal dose work supports local injury but not the capital decision." }, "inVitroEvidence": { "rating": "strong", "reason": "Photochemical killing is well established under controlled inputs." }, "reproducibility": { "rating": "weak", "reason": "No controlled multi-centre clinical benefit has been reproduced." }, "commercialConflicts": { "rating": "moderate", "reason": "The staged plan buys data access and independent testing, not product or equity." }, "recency": { "rating": "current", "reason": "The map incorporates the June 2026 trial update." } } }, "layout": { "x": 964, "y": 888 }, "updated": "2026-09-15" }, { "id": "trial-pdac-perioperative-sequencing-action-map", "title": "PDAC treatment-timing action map: preserve contradictions and finish A021806", "shortTitle": "Treatment-timing trial map", "type": "trial", "status": "supported", "scope": [ "pdac" ], "summary": "Completed trials disagree on survival and use different patient states and regimens; A021806 is the active modern test that should finish before another broad sequence trial.", "content": "Resectable timing evidence. PREOPANC randomized 246 eligible resectable or borderline patients: gemcitabine chemoradiotherapy before surgery produced five-year survival 20.5% versus 6.5% and OS HR .73, but timing, radiation and systemic regimen changed together. NORPACT-1 randomized 140 resectable pancreatic-head cases: only 79% assigned treatment first received it, severe events were 58% versus 40%, and median OS favoured upfront surgery, 38.5 versus 25.1 months, HR 1.52. CISPD-1 randomized 324 at one Hangzhou centre: a long sequential treatment-first regimen improved event-free survival, 15.3 versus 10.9 months, HR .71, and median OS 35.4 versus 27.2 months, HR .73 with 95% CI reaching 1.00; severe events were 47.6% versus 30.7%. The latest nine-trial analysis, 1,131 patients, resolves only part of the conflict: event-free survival HR .77 and noncurative exploration RR .90 favour treatment first, while OS HR .85, 95% CI .68-1.05 remains inconclusive. Borderline-route evidence. ESPAC5 randomized only 90 after screening 478; pooled neoadjuvant treatment did not raise resection, 55% versus 68%, but one-year survival was higher in its very small treatment arms. A021501 makes modified FOLFIRINOX a reference and rejects assuming extra radiation helps. Regimen evidence. PREOPANC-2 randomized 375 to all-neoadjuvant FOLFIRINOX or gemcitabine chemoradiotherapy plus postoperative gemcitabine; median OS 21.9 versus 21.3 months, HR .88, and treatment-related death occurred in two versus one. Live decision. NCT04340141 A021806 has 358 actual centrally reviewed, resectable participants and was updated in September 2026 as active, not recruiting. It compares the same twelve planned mFOLFIRINOX cycles split eight before and four after surgery against all twelve after upfront surgery. Overall survival is primary; disease-free survival, unresectability, R0, distant failure, dose intensity, cycles, severe events and eight-week quality of life are measured. Primary completion is estimated December 2028 and final completion November 2030. The next action is to recover complete route data and finish A021806, not start another broad timing trial. Any later study must name a patient state or treatment missing from A021806 and use overall survival, complete treatment exposure, operation loss and useful life.", "sourceIds": [ "versteijne-2022-preopanc-longterm", "preopanc2-2025", "labori-2024-norpact1", "bai-2025-cispd1", "tan-2026-neoadjuvant-ipd-meta", "ghaneh-2023-espac5", "katz-2022-a021501", "clinicaltrials-nct04340141-a021806" ], "links": [ { "target": "trials", "relation": "part-of" }, { "target": "claim-neoadjuvant-therapy-improves-selection-not-cure", "relation": "tests" }, { "target": "unknown-pdac-perioperative-treatment-route", "relation": "tests" }, { "target": "failure-norpact1-neoadjuvant-delivery", "relation": "includes" }, { "target": "trial-pdac-surgical-route-action-map", "relation": "complements" }, { "target": "programme-100m-randomized-platform", "relation": "must-not-duplicate" }, { "target": "programme-500m-integrated-mission", "relation": "constrains" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong but conflicting", "independentReplication": "moderate", "sampleSize": "PREOPANC n=246; NORPACT-1 n=140; CISPD-1 n=324; ESPAC5 n=90 allocated; PREOPANC-2 n=375; pooled nine trials n=1,131; A021806 actual n=358", "effect": "Action map; pooled EFS HR .77 and OS HR .85, 95% CI .68-1.05", "limits": [ "Different anatomy rules", "Different systemic regimens", "Radiation bundled in some trials", "Small and single-centre studies", "A021806 result pending", "No treatment-timing biomarker validated" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Early systemic exposure can treat occult disease and expose progression before surgery, while toxicity and delay can cause loss." }, "animalEvidence": { "rating": "not-applicable", "reason": "Human sequence delivery and survival are decisive." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Cell response cannot compare whole treatment routes." }, "reproducibility": { "rating": "moderate", "reason": "Trials reproduce event-free and pathology effects but not a stable overall-survival benefit." }, "commercialConflicts": { "rating": "low", "reason": "Most mapped trials are publicly or charitably funded academic comparisons." }, "recency": { "rating": "current", "reason": "Includes the 2025 phase 3 result, 2026 pooled analysis and September 2026 live registry state." } } }, "updated": "2026-09-14" }, { "id": "trial-pdac-pharmacogenetic-safety-action-map", "title": "Pancreatic pharmacogenetic safety action map: test, dose, measure, repair", "shortTitle": "Pancreatic dosing action map", "type": "trial", "status": "active", "scope": [ "pdac", "treatment-safety", "pharmacogenetics", "care-delivery", "capital-allocation" ], "summary": "Keep required DPYD testing, separate the UGT1A1 question, and join both to actual dosing, toxicity and tumour outcomes before expanding the panel or changing treatment.", "content": "Route 1, required care. Test DPD function before systemic 5-FU, capecitabine or tegafur. Do not give these drugs in complete deficiency; use a reduced start and monitored adjustment in partial deficiency. Route 2, evidence strength. Prospective mixed-cancer cohorts show fewer severe events among known DPYD carriers against historical controls. This is enough to act for safety, not enough to claim every variant, dose or cancer outcome is solved. Route 3, pancreatic boundary. FOLFIRINOX contains both 5-FU and irinotecan. In 199 pancreatic patients, UGT1A1 heterozygous risk differed between original and modified FOLFIRINOX; in a separate guided-dose study, 6/19 pancreatic patients had cycle-one dose-limiting toxicity and the prespecified tolerability bound failed. Route 4, current UGT1A1 evidence. A 501-person Dutch record found 77.8% severe toxicity in nine poor metabolizers mistakenly given full irinotecan versus 29.6% among 54 started at 70%, P=0.009. It supports avoiding full-dose errors but is not a randomized pancreatic result. Route 5, coverage. The four-variant DPYD panel leaves residual risk and may cover ancestries unequally. Store and study negative-test severe cases; add no variant to clinical action without reproduced function and outcome. Route 6, data unit. One row begins at treatment intent, not at a returned test. It ends with result time, dose action, escalation, delivered intensity, severe toxicity, admission, early death, tumour control and survival. Route 7, experiment. Spend at most £800,000 for 2,000 consecutive patients across 20 NHS centres over 18 months. Do not randomize away established DPYD testing and do not buy a proprietary platform. Route 8, release gates. Require at least 95% results before cycle one, less than 5% delay above three working days, at least 95% correct locked actions, no complete-deficiency exposure and less than 2% full-dose error in defined poor metabolizers. Route 9, harm gates. Stop repair rollout for severe toxicity above 35% in either identified high-risk group, reduced normal-metabolizer dose without a clinical reason, or a material upward 90-day-death signal. Route 10, capital. This is a small reliability purchase inside existing treatment delivery, not a drug-development bet. Invest £0 in an assay company and start no new efficacy trial.", "sourceIds": [ "ema-2020-dpd-testing", "mhra-2020-dpd-testing", "nhs-england-dpyd-implementation", "henricks-2018-dpyd-guided-dosing", "glewis-2024-pacific-pgx", "shirasu-2019-ugt1a1-folfirinox", "sharma-2019-ugt1a1-mfolfirinox", "heersche-2026-ugt1a1-guided-dosing", "chan-2024-dpyd-non-european-review" ], "links": [ { "target": "trials", "relation": "part-of" }, { "target": "claim-dpyd-testing-prevents-avoidable-fluoropyrimidine-harm", "relation": "maps" }, { "target": "failure-pharmacogenetic-panel-and-route-coverage", "relation": "maps" }, { "target": "unknown-pdac-pharmacogenetic-route-net-benefit", "relation": "resolves" }, { "target": "hypothesis-pdac-pharmacogenetic-route-completion", "relation": "implements" }, { "target": "trial-older-pdac-action-map", "relation": "coordinates-with" }, { "target": "trial-pdac-surgical-route-action-map", "relation": "coordinates-with" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "strong", "randomised": "weak — the action map audits and repairs standard care", "independentReplication": "moderate — component findings repeat across countries and regimens with important limits", "sampleSize": "Evidence n=1,181, n=462, n=501 and pancreatic n=199 plus n=19; proposed route n=2,000 and cap £800,000", "effect": "Immediate required DPYD safety action plus a bounded pancreatic route audit; UGT1A1 remains regimen-specific", "limits": [ "No pancreatic randomized net-benefit estimate", "Historical controls", "Rare variants", "Mixed populations", "Result timing and escalation data absent nationally" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Both genes change clearance and therefore exposure to active chemotherapy components." }, "animalEvidence": { "rating": "not-applicable", "reason": "Human pharmacology, dosing and toxicity are the decisive evidence." }, "inVitroEvidence": { "rating": "moderate", "reason": "Functional tests can support variant classification but not whole-route utility." }, "reproducibility": { "rating": "moderate", "reason": "Safety direction repeats, while pancreatic dose and panel boundaries remain unresolved." }, "commercialConflicts": { "rating": "low", "reason": "The map requires no proprietary assay and mandates independent public analysis." }, "recency": { "rating": "current", "reason": "Current English delivery and 2026 UGT1A1 evidence were checked in September 2026." } } }, "layout": { "x": 2040, "y": 1720 }, "updated": "2026-09-15" }, { "id": "trial-pdac-pipac-action-map", "title": "PDAC PIPAC action map: recover selection before testing added value", "shortTitle": "PIPAC action map", "type": "trial", "status": "active", "scope": [ "pdac", "peritoneal-metastases", "pipac", "capital-allocation" ], "summary": "PIPAC has local tissue activity and no controlled pancreatic benefit estimate; spend £300,000 at most on existing-data recovery before any randomized pilot.", "content": "Route 1, mechanism. Laparoscopy creates a pressurized chemotherapy mist over cancer deposits on the abdominal lining. It is a local surface-delivery system, not systemic control. Route 2, human signal. The largest registry includes 156 people and 350 procedures; older tiny series report tissue regression. Route 3, attrition. Only 55/156 reached three procedures. Progression or poor condition stopped 87/156. Average exposure was 2.24 procedures, so completer results cannot stand for all starters. Route 4, safety. Severe adverse events occurred after 10/350 procedures and 30-day mortality was 6/156, attributed to progression. Repeated anaesthesia and access remain part of the treatment cost. Route 5, survival. Median survival was nine months from first PIPAC. Associations with three-plus procedures and nab-paclitaxel are hypothesis-generating because repetition and drug choice were not randomized. Route 6, current trial. NCT05371223 targets 38 and bundles six cycles of intravenous nab-paclitaxel-gemcitabine with three nab-paclitaxel PIPAC procedures. Its record is unknown and stale since February 2024, with an estimated July 2025 completion and no posted results. Route 7, capital. Buy £0 of devices or procedure capacity and enroll nobody. Release at most £300,000 for all-screened, all-starter, dated data recovery across the registry and current trial. Route 8, stop. End mission funding if less than 95% of starters, 90% of screened candidates or 95% of procedure dates are recoverable, or systemic chemotherapy cannot be reconstructed. Route 9, later proof. Only after recovery, consider a randomized add-on pilot holding systemic drugs constant, with quality-adjusted survival, hospital-free days, symptoms, systemic-dose intensity and 90-day harm measured from assignment. Route 10, return. The value is a selection-corrected answer and reusable peritoneal-treatment dataset, not another uncontrolled local-response rate.", "sourceIds": [ "digiorgio-2026-pipac-pdac-registry", "clinicaltrials-nct05371223-nab-pipac", "digiorgio-2024-nab-pipac-protocol", "graversen-2017-pipac-pdac" ], "links": [ { "target": "trials", "relation": "part-of" }, { "target": "claim-pipac-local-delivery-signal-without-benefit-proof", "relation": "maps" }, { "target": "failure-pipac-repeat-procedure-selection-and-bundling", "relation": "maps" }, { "target": "unknown-pipac-added-survival-quality-and-systemic-control", "relation": "resolves" }, { "target": "hypothesis-pipac-existing-data-and-control-gate", "relation": "implements" }, { "target": "trial-pdac-regional-hyperthermia-action-map", "relation": "compares-with" }, { "target": "trial-pdac-intraarterial-delivery-action-map", "relation": "compares-with" } ], "evidence": { "humanProspective": "weak", "humanRetrospective": "moderate", "randomised": "none in pancreatic cancer", "independentReplication": "weak — procedure and tissue signals only", "sampleSize": "Registry n=156; current trial target n=38; zero new people; data-recovery cap £300,000", "effect": "Local delivery signal without attributable quality-adjusted survival benefit", "limits": [ "No controlled pancreatic result", "High repeat-procedure attrition", "Current study bundles treatments", "Stale registry", "Peritoneal-only reach", "Operative burden", "Drug and centre variation" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Direct aerosol exposure of abdominal-surface disease is physically credible." }, "animalEvidence": { "rating": "not-applicable", "reason": "Human feasibility exists; benefit attribution is the missing layer." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "The action map is governed by patient-valued outcomes and causal controls." }, "reproducibility": { "rating": "weak", "reason": "Several centres can deliver PIPAC, but none has produced a controlled pancreatic benefit estimate." }, "commercialConflicts": { "rating": "moderate", "reason": "No equity case exists; restricted data payments reduce procedure-owner bias." }, "recency": { "rating": "current", "reason": "The route includes the August 2026 multicentre report and current registry status." } } }, "layout": { "x": 1102, "y": 1026 }, "updated": "2026-09-15" }, { "id": "trial-pdac-polyamine-action-map", "title": "PDAC polyamine action map: rescue the comparison before testing the drug again", "shortTitle": "Polyamine action map", "type": "trial", "status": "unknown", "scope": [ "pdac", "trial-delivery", "transfer-unproven" ], "summary": "The drug's early signal is uncontrolled, its severe-eye risk is unresolved, and its existing phase 3 record is more valuable than a duplicate cohort.", "content": "Route 1, mechanism. Polyamines are required for cell growth; ivospemin imitates spermine and disturbs polyamine supply. This supports measurement, not clinical use. Route 2, early human signal. NCT03412799 treated 50 people in one group. Twelve of 28 evaluable patients responded, while six serious liver and six serious retinal toxicities were attributed to ivospemin. A single independent case documents profound bilateral sight loss after treatment despite a healthy baseline retina. Route 3, decisive comparison. ASPIRE was designed correctly: ivospemin plus gemcitabine and nab-paclitaxel versus placebo plus the same drugs, quadruple masked, with survival primary. Its operating chain then broke when the contractor began termination for non-payment. The U.S. registry still says recruiting and 600 planned; Europe says ended in March 2025 and 547 planned. Neither posts results or the actual assigned count. Route 4, capital. Fund no new ivospemin patient, trial, licence or Panbela equity. Release at most £250,000 from the existing data and independent-statistics line for three-month custody and completeness diligence. Require at least 400 reconciled assignments, a usable blind and key, all protocol and analysis-plan versions, at least 90% 12-month vital status and arm-missingness difference no greater than ten points. Only then price a survival follow-up and analysis rescue; any rescue must displace an equal sum inside the fixed £500m programme. If the data fail, publish the failure and preserve the severe-harm record. The useful asset is a trustworthy comparison, not the company shell.", "sourceIds": [ "clinicaltrials-nct05254171-aspire", "ctis-2024-514714-12-00-aspire", "singhal-2021-sbp101-phase1", "clinicaltrials-nct03412799-sbp101", "carla-2024-ivospemin-retinal-atrophy", "panbela-2024-q3-10q", "panbela-2025-note-default", "sec-2025-panbela-registration-abandoned" ], "links": [ { "target": "trials", "relation": "part-of" }, { "target": "claim-ivospemin-signal-without-controlled-benefit", "relation": "maps" }, { "target": "failure-aspire-trial-operational-collapse", "relation": "maps" }, { "target": "unknown-aspire-assigned-denominator-and-results", "relation": "resolves" }, { "target": "hypothesis-aspire-data-rescue-gate", "relation": "implements" }, { "target": "company-panbela-therapeutics", "relation": "involves" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "weak", "randomised": "strong", "independentReplication": "weak", "sampleSize": "Phase 1 n=50; ASPIRE planned n=600 U.S. versus 547 Europe; safety database n=395; actual assigned count unknown", "effect": "No proved benefit; current action is a capped data-custody audit with zero new treatment exposure", "limits": [ "ASPIRE results absent", "Trial state conflict", "Data ownership unknown", "Severe-eye frequency unresolved", "Rescue cost cannot be estimated before diligence" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "The cell-growth route is plausible but exact action and therapeutic window remain uncertain." }, "animalEvidence": { "rating": "moderate", "reason": "Models support testing but not another uncontrolled escalation." }, "inVitroEvidence": { "rating": "strong", "reason": "Polyamine pools and cell growth respond to ivospemin." }, "reproducibility": { "rating": "weak", "reason": "The controlled human answer is trapped in an unresolved trial record." }, "commercialConflicts": { "rating": "high", "reason": "Sponsor distress and vendor claims can shape custody and disclosure." }, "recency": { "rating": "current", "reason": "The phase 3 milestones have passed and registry records conflict." } } }, "layout": { "x": 508, "y": 432 }, "updated": "2026-09-15" }, { "id": "trial-pdac-postresection-surveillance-action-map", "title": "Post-resection surveillance map: finish RADAR-PANC before adding another scan trial", "shortTitle": "Post-resection follow-up map", "type": "trial", "status": "supported", "scope": [ "pdac" ], "summary": "Observational imaging signals are strong but biased; RADAR-PANC tests routine scans directly, while CIRCPAC tests a different ctDNA-directed schedule.", "content": "Present evidence answers three different questions. Association: among 333 patients with recurrence across 33 centres and 13 countries, routine imaging tracked with recurrence treatment OR 2.57 and survival HR 0.75. In 1,311 Dutch resections, imaging tracked with asymptomatic recurrence OR 3.79, recurrence treatment OR 2.50 and survival HR 0.56. Neither comparison randomized the schedule, and the first conditioned on diagnosed recurrence. Direct scan utility: NCT04875325 RADAR-PANC plans 306 Dutch and UK patients after R0-R1 PDAC resection. It offers the intervention arm clinical review, CA19-9 and contrast CT chest/abdomen every three months for two years; control receives non-standardized, mainly symptom-triggered follow-up. Overall survival is primary; quality of life, worry, recurrence treatment, compliance and cost are secondary. The protocol expected 20% intervention refusal and 5% control imaging, assumed median survival 29.4 versus 19.7 months and reported 219 enrolled by January 2024. The registry was last verified in September 2024, remains enrolling by invitation, estimates October 2026 completion and has no posted result. Molecularly directed intensity: NCT05788744 CIRCPAC is recruiting a planned 1,000 Danish patients, with 410 originally described for the randomized surveillance part; ctDNA changes scan and EUS frequency, and completion extends to 2032. DYNAMIC-Pancreas and MAP-03 test drug decisions, not the same scan question. The next action is data recovery and common outcome fields, not a new broad trial. Ask RADAR-PANC for actual randomized count, acceptance, crossover, recurrence treatments, events, scan and false-positive burden; ask CIRCPAC for the randomized denominator and exact analysis population; align overall survival, quality of life, radiation, procedures and cost. A new study becomes justified only if the live trials fail operationally or leave a named patient state, action or comparator uncovered.", "sourceIds": [ "andel-2025-routine-imaging", "van-goor-2025-nationwide-imaging", "daamen-2024-radar-panc-protocol", "clinicaltrials-nct04875325-radar-panc", "clinicaltrials-nct05788744-circpac", "dynamic-pancreas-registry", "clinicaltrials-nct05802394-map03", "tai-2025-pan-asian-esmo-pancreatic" ], "links": [ { "target": "trials", "relation": "part-of" }, { "target": "claim-routine-imaging-after-pdac-resection-not-yet-proven", "relation": "tests" }, { "target": "unknown-postresection-surveillance-utility", "relation": "tests" }, { "target": "trial-pdac-ctdna-action-map", "relation": "complements" }, { "target": "unknown-ctdna-actionability", "relation": "sets-imaging-comparator-for" }, { "target": "person-lois-daamen", "relation": "led-by" }, { "target": "lab-dpcg-radar-panc-network", "relation": "run-by" }, { "target": "programme-100m-randomized-platform", "relation": "must-not-duplicate" } ], "evidence": { "humanProspective": "strong for live randomized design; moderate for completed association", "humanRetrospective": "strong for association", "randomised": "strong design; no result", "independentReplication": "none for utility", "sampleSize": "RADAR-PANC planned n=306; CIRCPAC planned n=1,000; observational comparisons n=333 and n=1,311", "effect": "Decision map; no completed randomized survival result", "limits": [ "RADAR-PANC registry verification is stale", "Optimistic survival assumption", "Treatment at recurrence remains chosen", "Crossover can dilute contrast", "CIRCPAC adds a biomarker and different schedule", "No posted results" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "Earlier detection has value only through an effective action at recurrence." }, "animalEvidence": { "rating": "not-applicable", "reason": "Human trial outcomes decide the schedule." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Bench evidence cannot compare follow-up routes." }, "reproducibility": { "rating": "none", "reason": "The randomized utility result is pending." }, "commercialConflicts": { "rating": "low", "reason": "The mapped trials are academic strategy studies; any assay access in CIRCPAC still requires disclosure." }, "recency": { "rating": "current", "reason": "Registry status was refreshed through March 2026 for CIRCPAC and September 2026 for this audit." } } }, "updated": "2026-09-14" }, { "id": "trial-pdac-regional-hyperthermia-action-map", "title": "PDAC regional hyperthermia action map: recover the clean experiment first", "shortTitle": "Hyperthermia action map", "type": "trial", "status": "active", "scope": [ "pdac", "resected", "locally-advanced", "regional-hyperthermia" ], "summary": "The only reported randomized package is negative and cannot isolate heat; an unpublished heat-only randomization must be recovered before new recruitment.", "content": "Route 1, treatment. Regional hyperthermia aims to hold deep pancreatic tissue near 40-43 degrees C for about 60 minutes so chemotherapy or radiation may work better. It does not burn the tumour like thermal ablation. Route 2, reported human evidence. HEAT randomized 117 people after surgery but changed heat, cisplatin, gemcitabine schedule and control chemotherapy together. One contrast therefore contains at least five causal terms rather than an isolated heat effect. Disease-free survival was 12.7 versus 11.2 months, P=0.394; overall survival was 33.2 versus 25.2 months, P=0.099. Neither test established benefit. Route 3, trial integrity. Enrollment reached 117 / 336 = 34.8% of plan. The European record says prematurely ended without giving a reason. The U.S. registry and paper disagree about how much control capecitabine was delivered. Route 4, cleaner experiment. HEATPAC planned identical drug and radiation treatment in both groups and changed only heat, with real-time temperature probes. That design can estimate added heat. Route 5, missing result. HEATPAC's registry has been unknown since 2017, lists 78 rather than the protocol's 86 planned people, estimated completion in 2021 and has no posted result. Route 6, action. Spend £0 on a new cohort and at most £250,000 on independent recovery of both randomized records, including assignment, survival, serious harm and temperature traces. Publish failure to recover as a result. Route 7, release gate. Consider a current-care randomized trial only if at least 95% of assignments and deaths are reconciled, at least 80% of heat-session temperature records survive, the all-assigned estimate is compatible with a worthwhile heat-specific gain, severe harm is acceptable and the new trial changes heat alone. Until then, rank regional hyperthermia below proven modern systemic treatments and do not market the secondary post-recurrence signal as efficacy.", "sourceIds": [ "issels-2023-heat-rct", "clinicaltrials-nct01077427-heat", "eudract-2008-004802-14-heat", "datta-2017-heatpac-protocol", "clinicaltrials-nct02439593-heatpac" ], "links": [ { "target": "trials", "relation": "part-of" }, { "target": "claim-heat-regional-hyperthermia-primary-endpoint-negative", "relation": "maps" }, { "target": "failure-heat-component-and-registry-drift", "relation": "maps" }, { "target": "unknown-regional-hyperthermia-added-benefit-modern-care", "relation": "resolves" }, { "target": "hypothesis-heat-existing-randomization-recovery-gate", "relation": "implements" }, { "target": "claim-adjuvant-mfolfirinox-improves-but-recurrence-remains", "relation": "ranked-below" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "not-applicable", "randomised": "moderate", "independentReplication": "none", "sampleSize": "HEAT n=117 randomized; HEATPAC actual enrollment and result unknown", "effect": "No established heat-specific benefit; recover existing randomized data before new enrollment", "limits": [ "Reported primary endpoint negative", "Reported trial confounded by component changes", "Both registries are stale or incomplete", "HEATPAC result and actual enrollment are unknown", "Older treatment backbones", "Temperature measurement may be incomplete", "No evidence of cure" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "Heat could sensitize treatment, but the clinical estimate is unresolved." }, "animalEvidence": { "rating": "moderate", "reason": "Laboratory evidence supports measurement, not adoption." }, "inVitroEvidence": { "rating": "moderate", "reason": "Heat-treatment interactions are plausible in model systems." }, "reproducibility": { "rating": "none", "reason": "There is no reported independent heat-only randomized result in pancreatic cancer." }, "commercialConflicts": { "rating": "low", "reason": "No company owns a single treatment claim here; incomplete academic records remain a serious bias risk." }, "recency": { "rating": "mixed", "reason": "The map uses the 2023 HEAT publication and current registries but does not treat older regimens as current care." } } }, "layout": { "x": 640, "y": 564 }, "updated": "2026-09-15" }, { "id": "trial-pdac-residual-cell-state-action-map", "title": "PDAC residual-cell action map: join ancestry, state, time and treatment before intervention", "shortTitle": "Residual-cell study map", "type": "trial", "status": "supported", "scope": [ "pdac", "transfer-unproven" ], "summary": "No human study yet observes a PDAC cell during the disease-free interval, links it to recurrence and proves that a safe treatment removes it.", "content": "Human risk signals. The 175-person surgical marrow study found 13.7% cytokeratin-positive and OS HR 2.283; the 48-person unresectable study found 31% positive before chemotherapy and OS HR 2.0. Both are prognostic, not dormant-cell lineage tests. Human route signals. In 744 resected patients, 333 recurred; selected primary-tumour single-nucleus data from 21 cases linked different expression states to later liver or lung recurrence. Rapid-autopsy spatial work across 49 sections from 13 people found multiple lineage states and strong organ effects, but only at treatment-refractory death. Microscopic local persistence was invisible on imaging in 5/14 autopsies after extended surgery and radiation. Model mechanisms. Reversible MYC models prove that residual cells can regrow; the 2026 chemotherapy-senescence-NET-CCDC25 study proves one liver reawakening chain in models; a separate 2026 model finds EREG-high quiescent cells after CAR-T treatment. These are different treatments, compartments and states, not one marker. Missing human chain: a sample obtained while the person has no radiographic disease; proof that the cell is malignant and low-cycling; private alterations linking it to the later recurrence; a state that adds information beyond stage, treatment and ctDNA; and a clinically achievable intervention. The existing £25m observatory can run the next action without a new budget: use a 250-person resected subset, fixed blood and imaging, archived primary tissue, safe recurrence biopsy before new therapy and rapid autopsy. Use marrow only in a nested consented comparison, not as an assumed reservoir. Conceal outcomes during state calling and reproduce the state in a second laboratory. Go to an intervention only if the state appears before imaging in at least 25 recurrence cases, ancestry is supported in at least 80% of paired samples, assay agreement is at least 80%, lead time is at least eight weeks, and prediction improves beyond stage plus ctDNA. Stop for serious biopsy harm above 10%, fewer than 100 complete timelines, no added prediction, a handling-created state, or a target needed for normal repair or antimicrobial defence.", "sourceIds": [ "effenberger-2012-pdac-bone-marrow-dtc", "nordgard-2022-unresectable-dtc", "chalabi-hajkarim-2025-organotropism", "pei-2025-spatial-metastatic-pdac", "mcclellan-2026-quiescent-pdac-car-t", "tanaka-2012-microscopic-local-recurrence", "lin-2013-reversible-pdac-dormancy", "shi-2026-chemo-dormant-dtc", "botta-2024-tumor-informed-ctdna" ], "links": [ { "target": "trials", "relation": "part-of" }, { "target": "claim-human-pdac-dtc-prognostic-not-dormancy", "relation": "maps" }, { "target": "failure-pdac-dormancy-sample-gap", "relation": "repairs" }, { "target": "claim-pdac-dormancy-model-proven-human-uncertain", "relation": "tests" }, { "target": "unknown-dormant-cell-actionability", "relation": "tests" }, { "target": "hypothesis-human-dormancy-reawakening-gate", "relation": "operationalizes" }, { "target": "programme-25m-prospective-observatory", "relation": "part-of" }, { "target": "trial-pdac-ctdna-action-map", "relation": "uses-comparator-from" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "strong", "randomised": "none", "independentReplication": "none for the complete chain", "sampleSize": "Existing cohorts n=175, n=48 and n=744; autopsy studies n=13 and n=14; proposed nested subset n=250", "effect": "Action map and gates; no intervention effect", "limits": [ "No direct human residual-organ sample", "Recurrence biopsy selection", "Rapid-autopsy selection", "Low event count", "State instability", "No safe intervention established" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Human recurrence, occult-cell signals and model lineage tracing support persistence, while exact states remain unknown." }, "animalEvidence": { "rating": "strong", "reason": "Independent models show residual survival, reawakening and treatment-specific quiescent states." }, "inVitroEvidence": { "rating": "moderate", "reason": "Human-derived systems can test state dependence but cannot preserve the whole organ and immune route." }, "reproducibility": { "rating": "none", "reason": "No group has reproduced ancestry, state, timing and action in one human PDAC cohort." }, "commercialConflicts": { "rating": "moderate", "reason": "ctDNA, spatial assays and later target access can create product control unless data and methods remain open." }, "recency": { "rating": "current", "reason": "The map includes 2025 human spatial work and two 2026 model mechanisms." } } }, "updated": "2026-09-14" }, { "id": "trial-pdac-rfa-action-map", "title": "PDAC radiofrequency-ablation action map: stop treatment, salvage evidence", "shortTitle": "RFA action map", "type": "trial", "status": "active", "scope": [ "pdac", "locally-advanced", "radiofrequency-ablation" ], "summary": "A modern randomized trial closes the open surgical RFA route; only capped reuse of existing material remains justified.", "content": "Route 1, treatment. Pancreatic RFA places electrodes into tumour during open surgery and heats tissue enough to destroy it. Route 2, decisive comparison. PELICAN randomized 190 people after about two months of stable or responding multiagent chemotherapy; 188 remained analysable. Median survival was 12.1 months with RFA versus 11.6 with chemotherapy alone, HR 1.07, 95% CI 0.80-1.45. Progression-free survival was 5.8 versus 6.9 months. Route 3, harm. Grade 3 or worse serious events were 27% versus 11%, a 16-point excess and approximate number needed to harm of 6.25. Quality of life fell beyond the ten-point important threshold at one, three and 18 months only in the RFA route. One RFA-related death occurred. Route 4, full route. Only 70/95, 74%, assigned to RFA received it versus 91/93, 98%, controls receiving chemotherapy. Twenty-one ablation-assigned patients had hidden metastases found during surgery. Keep those losses in the treatment estimate. Route 5, subgroup boundary. A secondary CA19-9 interaction is counterintuitive and cannot rescue treatment without fixed independent reproduction. Route 6, registry. ClinicalTrials.gov still says status unknown, target 228 and last update 2018; the paper supplies the current result and 190 randomized count. Route 7, action. Spend £0 on RFA equipment, £0 on another RFA cohort and do not use open RFA for this indication outside a truly distinct ethical research question. At most £150,000 may reuse existing PELICAN specimens to test the claimed immune mechanism, with no treatment release. Route 8, lesson. Every invasive local-treatment estimate starts at assignment and includes occult spread, non-delivery, surgery, treatment delay, severe harm, quality of life and survival. Technical ablation is not patient benefit.", "sourceIds": [ "seelen-2026-pelican", "clinicaltrials-nct03690323-pelican" ], "links": [ { "target": "trials", "relation": "part-of" }, { "target": "claim-pelican-rfa-no-survival-and-more-harm", "relation": "maps" }, { "target": "failure-rfa-selection-and-route-attrition", "relation": "maps" }, { "target": "unknown-rfa-human-mechanism-after-negative-trial", "relation": "resolves" }, { "target": "hypothesis-pelican-specimen-salvage-gate", "relation": "implements" }, { "target": "failure-local-intensification-pdac-survival-not-shown", "relation": "supports" }, { "target": "trial-pdac-smart-action-map", "relation": "contrasts-with" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong", "independentReplication": "moderate", "sampleSize": "190 randomized; 188 analysed; 95 RFA route and 93 control", "effect": "No survival benefit; severe serious events +16 percentage points; clinically important quality-of-life loss", "limits": [ "Open-label", "Two control withdrawals excluded", "Induction chemotherapy was short", "Does not test endoscopic RFA", "Does not exclude every future molecularly selected use", "Stored specimens are not confirmed", "Registry remains stale" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The procedure destroys local tissue but did not improve net outcome." }, "animalEvidence": { "rating": "moderate", "reason": "Model immune effects cannot justify renewed exposure." }, "inVitroEvidence": { "rating": "strong", "reason": "Thermal destruction is established and is not the clinical bottleneck." }, "reproducibility": { "rating": "moderate", "reason": "Two randomized routes have not shown survival gain; PELICAN is the current decisive result." }, "commercialConflicts": { "rating": "low", "reason": "The result argues against procedure adoption." }, "recency": { "rating": "current", "reason": "PELICAN published in March 2026." } } }, "layout": { "x": 736, "y": 660 }, "updated": "2026-09-15" }, { "id": "trial-pdac-smart-action-map", "title": "PDAC SMART action map: preserve precise delivery, rebuild the survival test", "shortTitle": "SMART action map", "type": "trial", "status": "active", "scope": [ "pdac", "locally-advanced", "borderline-resectable", "radiation" ], "summary": "MRI-guided adaptation is a useful delivery system, but its survival trial, vendor continuity and all-patient route must be repaired before mission adoption.", "content": "Route 1, engineering. SMART uses live MRI to see soft tissue, pause for motion and rebuild a radiation plan while the patient is on the table. It can deliver 50 Gy in five treatments while constraining nearby stomach and bowel. Route 2, human evidence. A 136-person single-arm phase 2 selected people after at least three months of chemotherapy without distant progression and CA19-9 at most 500. Two-year survival was 53.6% from diagnosis and 40.5% from SMART; 34.6% later had surgery. Route 3, safety boundary. Acute severe gastrointestinal toxicity definitely attributed to treatment was zero, but later severe toxicity was 0% definitely, 4.6% probably and 11.5% possibly related. Preserve all events and attribution levels. Route 4, comparison boundary. CROSSFIRE compared SMART with irreversible electroporation and found no survival difference. It cannot show that either local addition beats continued chemotherapy. Route 5, decisive trial failure. LAP-ABLATE planned the correct 267-person treatment-versus-no-treatment comparison, but remains not yet recruiting on a record last updated in 2022. MASPAC terminated after nine people. Route 6, business cause. ViewRay sponsored the phase 2 and LAP-ABLATE, burned $53.6m of operating cash in one quarter against $81.3m unrestricted cash and filed Chapter 11 in July 2023. Route 7, action. Invest £0 in ViewRay, £0 in a machine, £0 in a new cohort and at most £500,000 to reconcile phase 2 data and replace the branded protocol with testable performance requirements across maintained systems. Route 8, release gate. Require complete harms, 99% survival accounting, positive quality-adjusted survival, preserved chemotherapy, no postoperative-death excess, at least ten committed centres and cross-system dose differences within 5%. If any gate fails, retire mission expansion while retaining SMART as a selected clinical delivery option where independently justified.", "sourceIds": [ "parikh-2023-smart-phase2-safety", "chuong-2024-smart-phase2-survival", "clinicaltrials-nct03621644-smart", "clinicaltrials-nct05585554-lap-ablate", "clinicaltrials-nct05114213-maspac", "viewray-2023-q1-10q", "viewray-2023-chapter11-8k", "timmer-2024-crossfire" ], "links": [ { "target": "trials", "relation": "part-of" }, { "target": "claim-smart-delivery-and-safety-without-added-survival-proof", "relation": "maps" }, { "target": "failure-smart-trial-and-vendor-dependency", "relation": "maps" }, { "target": "unknown-smart-added-survival-and-surgery-harm", "relation": "resolves" }, { "target": "hypothesis-smart-vendor-independent-randomized-gate", "relation": "implements" }, { "target": "company-viewray", "relation": "involves" }, { "target": "trial-pdac-irreversible-electroporation-action-map", "relation": "compares-with" }, { "target": "hypothesis-randomized-lapc-resection-after-response", "relation": "interacts-with" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "weak", "independentReplication": "none", "sampleSize": "Phase 2 n=136; CROSSFIRE n=68; LAP-ABLATE target n=267 not started; MASPAC n=9", "effect": "Delivery and safety signal without added-survival proof; recovery capped at £500,000 and new trial receives £0 now", "limits": [ "Selected single-arm survival", "No continued-chemotherapy comparison", "Late harm depends on attribution", "Surgical effect and harm are entangled", "Vendor bankruptcy", "Dormant phase 3 registry", "Cross-system equivalence is unproved" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Daily image guidance directly improves the ability to deliver local dose around moving bowel." }, "animalEvidence": { "rating": "not-applicable", "reason": "Human dose planning and toxicity are available." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "The unresolved value is clinical route benefit." }, "reproducibility": { "rating": "weak", "reason": "Delivery repeats, while treatment-versus-no-treatment survival has not been tested." }, "commercialConflicts": { "rating": "high", "reason": "The former vendor sponsored the core phase 2 and planned phase 3." }, "recency": { "rating": "current", "reason": "Current records include the bankruptcy, stale phase 3 and 2025 randomized-trial closure." } } }, "layout": { "x": 700, "y": 624 }, "updated": "2026-09-15" }, { "id": "trial-pdac-sonoporation-action-map", "title": "PDAC sonoporation action map: release the completed randomized trial before another device study", "shortTitle": "Sonoporation action map", "type": "trial", "status": "mixed", "scope": [ "pdac", "pan-cancer", "transfer-unproven" ], "summary": "Two randomized trials now bound the question; the next action is complete NCT04821284 reporting and a planned interaction test, not duplicate recruitment.", "content": "Origin. NCT01674556 showed that commercial ultrasound and microbubbles could be delivered after gemcitabine in ten people, but its survival comparison used 63 historical controls. First contribution test. ChiCTR2100044721 randomized 78 of 82 recruited people and reported overall survival of 9.1 versus 6.1 months, P=0.037, disease control of 73.68% versus 42.50%, P=0.005, and progression-free survival of 5.5 versus 3.5 months, P=0.080. Second contribution test. NCT04821284 randomizes stage II to IV unresectable PDAC in the United States and Norway to identical gemcitabine/nab-paclitaxel or FOLFIRINOX with or without perflubutane microbubbles and contrast-enhanced ultrasound. Its primary endpoint is PFS; OS, chemotherapy cycles, tumour volume, perfusion, pressure and stiffness are secondary. The registry estimates 120 participants, reports actual primary completion on 31 October 2025, estimated study completion on 28 February 2026, remains active not recruiting and has no results. An accepted 2026 abstract says 97 enrolled, neither overall PFS nor OS differed, P>0.36, and highlights a gemcitabine subgroup without enough information to test it. Action now. Do not recruit another efficacy cohort. Reconcile 97 with 120; release the protocol versions, dated analysis plan, participant flow, arm and backbone counts, stage, treatment exposure, adverse events, quality of life, scans and all survival events. Report overall estimates and the treatment-by-backbone interaction with confidence intervals and the number of subgroup tests. Keep the 13-person perfusion analysis as an exploratory instrument-development dataset: four negative slopes tracked improvement, six positive slopes tracked deterioration and three were discordant, so 10/13 = 76.9% agreed in direction. It neither estimates randomized benefit nor confirms that chemotherapy concentration rose inside tumour. No programme money is released. A fixed-backbone confirmation may compete later only if a planned interaction survives full reporting and a direct or validated drug-delivery measure works.", "sourceIds": [ "dimcevski-2016-sonoporation-phase1", "wang-2023-sonochemotherapy-pdac", "clinicaltrials-nct04821284-sonoporation", "ius-2026-sonoporation-nct04821284-abstract", "adler-2026-sonoporation-perfusion" ], "links": [ { "target": "trials", "relation": "part-of" }, { "target": "claim-sonochemotherapy-randomized-signal", "relation": "tests" }, { "target": "failure-sonoporation-overall-replication-subgroup-rescue", "relation": "repairs" }, { "target": "unknown-pdac-sonoporation-treatment-effect", "relation": "maps" }, { "target": "hypothesis-measured-fixed-backbone-sonoporation", "relation": "gates" }, { "target": "failure-local-intensification-pdac-survival-not-shown", "relation": "constrains" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "not-applicable", "randomised": "moderate", "independentReplication": "weak — negative overall result available as accepted abstract", "sampleSize": "Feasibility n=10; first randomized n=78; second randomized n=97 reported enrolled versus n=120 estimated; imaging n=13", "effect": "Positive first-trial OS signal; no overall second-trial PFS or OS difference, both P>0.36", "limits": [ "No complete NCT04821284 report", "Enrollment mismatch", "Open label", "Mixed stage and backbone", "Incomplete subgroup", "No direct drug-delivery measure" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "The device could change delivery locally, which must be measured rather than assumed." }, "animalEvidence": { "rating": "moderate", "reason": "Enough to justify the completed human trials, not a third one before reporting." }, "inVitroEvidence": { "rating": "weak", "reason": "A vessel and tissue transport problem requires human measurement." }, "reproducibility": { "rating": "weak", "reason": "Overall benefit did not reproduce in the available second-trial result." }, "commercialConflicts": { "rating": "moderate", "reason": "Full device, contrast-agent, funding and analysis roles await the final report." }, "recency": { "rating": "current", "reason": "Uses the latest registry state, 2026 imaging paper and accepted trial abstract." } } }, "layout": { "x": 157, "y": 151 }, "updated": "2026-09-15" }, { "id": "trial-pdac-surgical-route-action-map", "title": "PDAC surgical action map: test the route, not a device race", "shortTitle": "Surgical action map", "type": "trial", "status": "supported", "scope": [ "pdac", "pancreatic-all" ], "summary": "Centralization, operation method and treatment completion have different evidence; the next experiment should join delivery while preserving procedure-specific safety gates.", "content": "1. Centre concentration: Dutch observational series linked movement into higher-volume centres with in-hospital mortality falling from 9.8% to 5.1% nationally and, in a separate region and era, from 24.4% to 3.6%. These studies changed referral, selection, surgery and rescue together. An English cohort of 31,973 diagnoses found lower adjusted mortality where population resection rates were higher, HR 0.82 highest versus lowest quintile, but no clear hospital-volume trend among the 2,580 operated patients, P=0.177. Volume is a screen for capability, not the active ingredient. 2. Current English route: the commissioned team threshold is at least 80 neoplastic or suspected-neoplastic pancreatic operations yearly. The 2026 audit reports 23 centres, 96.4% 90-day survival after Whipple, and only 62% postoperative systemic-treatment receipt within 14 weeks, centre IQR 54-68%. It does not count full-course completion. 3. Laparoscopic Whipple: NTR5689 LEOPARD-2 stopped after 105 randomizations because complication-related deaths were 5/50 versus 1/49 and recovery was not faster. 4. Robotic Whipple: DRKS00020407 EUROPA was exploratory and found more grade B/C pancreas-specific complications, 58.6% versus 33.3%, in only 62 analysed patients. ChiCTR2200056809 randomized 164 selected patients and reduced median stay by two days after surgeons passed stated learning curves, with one death per arm. NCT04400357 PORTAL randomized 268 across seven high-volume Chinese centres and reduced restricted mean recovery time by 3.9 days, with 1/132 versus 3/122 deaths, but cost about ¥22,834 more per admission and has no long-term PDAC result. These trials are not interchangeable and mix indications. 5. Left-sided PDAC: ISRCTN44897265 DIPLOMA randomized 258 and passed its -7% R0 noninferiority margin. At 38 months, overall-survival HR was 1.02, disease-free-survival HR 0.96, and adjuvant treatment 70% versus 72%. This supports qualified minimally invasive left pancreatectomy, not Whipple transfer. 6. Treatment completion: current audit gives a timely-start denominator; historical PDAC cohorts link route loss to progression, complications, poor fitness and age but cannot estimate the avoidable fraction. Do not fund another broad robot-versus-open race. Fund a national route study in the existing 23-centre network, with a first-scan denominator, locked plan, procedure-specific proficiency, rescue, recovery, full systemic-treatment delivery, patient-reported recovery, costs and long-term cancer outcomes.", "sourceIds": [ "npaca-2026-state-of-nation", "nhs-england-2024-hpb-pancreatic-service", "de-wilde-2012-centralization", "lemmens-2011-centralization", "latenstein-2021-centralization", "coupland-2016-england-resection-volume", "labori-2016-adjuvant-completion", "henry-2023-complications-adjuvant", "kollbeck-2025-complications-chemo", "van-hilst-2019-leopard2", "korrel-2023-diploma", "bruna-2025-diploma-long-term", "klotz-2024-europa", "liu-2024-robotic-open-pd", "jin-2026-portal", "labori-2024-norpact1", "bai-2025-cispd1", "tan-2026-neoadjuvant-ipd-meta", "ghaneh-2023-espac5", "clinicaltrials-nct04340141-a021806" ], "links": [ { "target": "trials", "relation": "part-of" }, { "target": "claim-pdac-surgery-is-a-complete-treatment-route", "relation": "supports" }, { "target": "system-england-pancreatic-surgery-network", "relation": "uses" }, { "target": "failure-pdac-postoperative-treatment-loss", "relation": "learns-from" }, { "target": "failure-leopard2-laparoscopic-whipple-safety", "relation": "learns-from" }, { "target": "unknown-pdac-surgical-route-completion", "relation": "tests" }, { "target": "hypothesis-pdac-surgical-route-completion", "relation": "prioritizes" }, { "target": "trial-pdac-perioperative-sequencing-action-map", "relation": "bounded-by" }, { "target": "unknown-pdac-perioperative-treatment-route", "relation": "bounded-by" } ], "evidence": { "humanProspective": "strong for randomized procedure components", "humanRetrospective": "strong for service and completion associations", "randomised": "strong for procedure-specific comparisons; none for whole route", "independentReplication": "moderate for procedure components; none for whole route", "sampleSize": "National audit n=17,672 diagnoses; centralization cohorts n=2,129-31,973; procedure trials n=81-268; completion cohorts n=203-1,071", "effect": "Decision map; exact component effects remain attached to their source and operation", "limits": [ "Different operations cannot be pooled", "Mixed benign and malignant trial populations", "High-volume expert-centre transport", "Short-term recovery is not cancer control", "Centralization is not randomized", "Full treatment completion is missing nationally" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The map joins the clinical steps required to turn resection into completed multimodality treatment." }, "animalEvidence": { "rating": "not-applicable", "reason": "Human operations and delivery data are decisive." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Laboratory work cannot rank service or operative routes." }, "reproducibility": { "rating": "moderate", "reason": "Several randomized procedure trials exist, but they differ by operation, platform, centre and endpoint; no full-route trial exists." }, "commercialConflicts": { "rating": "moderate", "reason": "Some minimally invasive trials received device-company funding, and robot adoption has direct commercial value." }, "recency": { "rating": "current", "reason": "The map includes the July 2026 PORTAL trial and September 2026 national audit." } } }, "updated": "2026-09-14", "contentSections": [ "Treatment timing is a separate choice inside the surgical route. PREOPANC, NORPACT-1 and CISPD-1 conflict on survival, and the pooled event-free gain does not settle overall survival. A021806 has completed accrual to the same-regimen timing test. Capture timing, every cycle and why treatment or surgery was lost in the English route study, but do not randomize timing again before A021806 reports." ] }, { "id": "trial-pdac-ttfields-action-map", "title": "PDAC tumour-treating fields action map: use narrowly, test independently", "shortTitle": "TTFields action map", "type": "trial", "status": "active", "scope": [ "pdac", "locally-advanced", "metastatic", "post-approval" ], "summary": "A positive phase 3 result justifies one narrow use; it does not justify a platform claim, metastatic expansion or an unrestricted price.", "content": "Route 1, mechanism. Optune Pax sends 150 kHz alternating electric fields through four insulated arrays on the abdomen. It is worn at home, not implanted, but it still demands batteries, cables, skin shaving, array replacement every three to four days and at least 12 hours of average daily use under the U.S. instructions. Route 2, decisive human evidence. PANOVA-3 randomized 571 people with unresectable locally advanced pancreatic adenocarcinoma to gemcitabine and nab-paclitaxel with or without the device. It improved median survival by 2.0 months and one-year survival by 7.9 points. Route 3, boundary. Progression-free survival missed its test, surgery was 7.0% versus 10.1%, median use was 11.2 hours per day, 76.3% had device-related skin toxicity and 8.4% stopped for a device-related adverse event. Route 4, regulation. FDA approval on 11 February 2026 covers adults with locally advanced disease only when the device is used with gemcitabine and nab-paclitaxel. Do not infer benefit with FOLFIRINOX, NALIRIFOX, after surgery or in metastatic disease. Route 5, failed extension. PANOVA-4 combined the device and atezolizumab in one metastatic cohort and compared disease control with an old trial. With one contrast and at least five causal terms, it cannot isolate either addition. Route 6, economics. Two 2026 models disagree on added quality-adjusted life-years, 0.148 versus 0.34, but both find poor value at current U.S. cost. Route 7, action. Recognize the FDA-approved option in shared decisions, showing the two-month median gain and daily burden together. Spend £0 on Novocure equity, £0 on product subsidy and £0 on a new company cohort. Offer at most £500,000 for independent PANOVA-3 analysis and complete prospective route measurement only after data access, analysis independence and public negative-result rights are signed. Require Novocure to fund randomized modern-backbone replication and any metastatic study. Apply the mission price formula only to independently verified health gain.", "sourceIds": [ "macarulla-2025-panova3", "clinicaltrials-nct03377491-panova3", "fda-2026-optune-pax-ssed", "clinicaltrials-nct06390059-panova4", "novocure-2026-panova4-topline", "novocure-2026-q2-10q", "guzauskas-2026-ttfields-cost", "liu-2026-ttfields-cost" ], "links": [ { "target": "trials", "relation": "part-of" }, { "target": "claim-panova3-ttfields-small-os-benefit", "relation": "maps" }, { "target": "failure-panova4-component-and-historical-control", "relation": "maps" }, { "target": "unknown-ttfields-survival-replication-and-access", "relation": "resolves" }, { "target": "hypothesis-optune-pax-independent-value-gate", "relation": "implements" }, { "target": "regulatory-fda-optune-pax-2026", "relation": "bounded-by" }, { "target": "company-novocure", "relation": "involves" }, { "target": "claim-modern-metastatic-chemotherapy-remains-short-control", "relation": "compares-with" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong", "independentReplication": "none", "sampleSize": "PANOVA-3 n=571; PANOVA-4 n=84 enrolled and 78 in topline analysis", "effect": "Narrow approved median OS gain of 2.0 months; no causal metastatic estimate", "limits": [ "One pivotal survival trial", "Open-label", "No independent replication", "No benefit established with another backbone", "High daily burden", "Current price appears poor value", "PANOVA-4 has no component control" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "A direct physical effect exists, but the clinical pathway and treatment interaction remain uncertain." }, "animalEvidence": { "rating": "moderate", "reason": "Animal evidence informed safety and field delivery but not comparative survival." }, "inVitroEvidence": { "rating": "strong", "reason": "Alternating fields can disturb division in pancreatic cancer cells." }, "reproducibility": { "rating": "none", "reason": "PANOVA-3 remains the only pancreatic randomized survival result." }, "commercialConflicts": { "rating": "high", "reason": "Novocure owns the device and sponsored both pancreatic studies." }, "recency": { "rating": "current", "reason": "The map includes the February 2026 FDA approval, March 2026 PANOVA-4 release and July 2026 filing." } } }, "layout": { "x": 610, "y": 534 }, "updated": "2026-09-15" }, { "id": "trial-pdac-vaccine-action-map", "title": "PDAC vaccine action map: finish the randomized tests already in hand", "shortTitle": "Vaccine action map", "type": "trial", "status": "active", "scope": [ "pdac", "transfer-unproven" ], "summary": "Every active route already has a bounded question; no new vaccine cohort is justified before IMCODE003 and complete AMPLIFY-7P reporting.", "content": "Route 1, common shared antigens. Stop GV1001: TeloVac randomized 1,062 and found no survival gain. Stop algenpantucel-L: 303 randomized patients had survival of 14.3 versus 14.9 months. Route 2, whole-cell and bacterial-vector vaccines. The 90-person GVAX/CRS-207 signal reversed in ECLIPSE; do not revive the product from immune markers or from the newer 57-person checkpoint study's two responses and 68% severe related toxicity. Route 3, off-the-shelf KRAS peptide vaccine. AMPLIFY-7P reports that its 144-person randomized PDAC stage missed the main disease-free-survival endpoint; its registry counts 158 actual participants across all stages. Require the complete intention-to-treat estimate and November 2026 study completion before any phase 3. The new 20-person ELI-002 plus-or-minus tislelizumab pilot has severe toxicity as its primary endpoint and cannot establish survival. Route 4, personalized RNA vaccine. IMCODE003 is active but no longer recruiting, estimates 260 participants at 87 sites and compares the vaccine/atezolizumab/chemotherapy package with the same chemotherapy. Wait for its all-assigned disease-free and overall survival record; report manufacture failures from first tissue submission. Route 5, dendritic-cell vaccine. The 38-person one-arm result is feasibility, not efficacy; fund no repeat cohort. Route 6, prevention. The ongoing single-centre KRAS study estimates 37 people and ends in 2031, but its main outcomes are toxicity and T-cell response rather than cancer incidence. It can establish safety and immunity, not prevention. Decision rule. A vaccine advances only through an all-assigned comparison against the same effective care, with recurrence or survival, complete manufacture and delivery denominators, and harm. Immune response is a diagnostic measurement inside the system, not the business outcome. Mission allocation is £0 for another trial, at most £250,000 for independent IMCODE003 analysis after data lock and access, and no off-trial clinical use.", "sourceIds": [ "middleton-2014-telovac", "le-2015-gvax-crs207", "le-2019-eclipse", "hewitt-2022-algenpantucel-l", "bever-2026-vaccine-checkpoint-pdac", "pant-2024-eli002", "wainberg-2025-eli002-final", "elicio-2026-amplify7p-results", "clinicaltrials-nct05726864-amplify7p", "clinicaltrials-nct07671339-eli002-tislelizumab", "rojass-2025-neoantigen-followup", "nct05968326-imcode003", "lau-2024-dendritic-resected-pdac", "haldar-2026-kras-interception", "clinicaltrials-nct05013216-kras-prevention" ], "links": [ { "target": "trials", "relation": "part-of" }, { "target": "claim-pdac-vaccine-immunity-without-proven-benefit", "relation": "maps" }, { "target": "failure-pdac-vaccine-escalation-reversals", "relation": "maps" }, { "target": "unknown-pdac-vaccine-regimen-contribution", "relation": "resolves" }, { "target": "hypothesis-autogene-cevumeran-existing-trial-gate", "relation": "implements" }, { "target": "claim-kras-interception-vaccine-immune-response-not-prevention", "relation": "contains" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong", "independentReplication": "moderate", "sampleSize": "Failed randomized programmes n=1,812; active IMCODE003 estimated n=260; current ELI-002 pilot n=20; prevention study estimated n=37", "effect": "No proven clinical benefit; £0 new cohorts and at most £250,000 conditional independent analysis", "limits": [ "IMCODE003 result is years away", "AMPLIFY-7P lacks a full report", "Different products cannot borrow efficacy", "Prevention follow-up is too short and lacks a cancer-incidence endpoint" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The map separates antigen choice, immune response, manufacture and patient outcome." }, "animalEvidence": { "rating": "strong", "reason": "Platform biology is mature but not allowed to override human failures." }, "inVitroEvidence": { "rating": "strong", "reason": "T-cell specificity, function and persistence are measurable for each product." }, "reproducibility": { "rating": "weak", "reason": "No PDAC vaccine benefit has repeated in a valid standard-care comparison." }, "commercialConflicts": { "rating": "high", "reason": "Current vaccines, assays and datasets are sponsor-controlled." }, "recency": { "rating": "current", "reason": "Trial states are current through September 2026." } } }, "layout": { "x": 428, "y": 366 }, "updated": "2026-09-15" }, { "id": "trial-pdac-vascular-hypoxia-action-map", "title": "PDAC blood-flow and hypoxia action map: stop starvation, recover two existing records", "shortTitle": "Vascular and hypoxia action map", "type": "trial", "status": "active", "scope": [ "pdac", "pan-cancer", "transfer-unproven" ], "summary": "No new cohort is justified: VEGF starvation failed, MAESTRO needs marker-linked reuse, and NCT03563248 already owns the localized losartan answer.", "content": "Route 1, vessel starvation. Four phase 3 tests totaling 2,387 people found no overall-survival benefit from bevacizumab, aflibercept or axitinib. Spend £0 and start no unselected VEGF-blocking PDAC trial. Route 2, low-oxygen activation. MAESTRO improved PFS and response but its OS HR 0.84 interval included no effect and the primary test failed. First recover the 693-person outcome record and determine whether fixed baseline tumour-hypoxia tissue exists. A retrospective test has value only if the assay is fixed outside MAESTRO and outcome linkage is at least 80%; no new evofosfamide cohort or capital is proposed. Route 3, vessel opening. AFPAC rejected 50 mg daily losartan in broad advanced disease at 88 randomized patients. Localized disease remains distinct, but NCT03563248 has already randomized 168 people. Publish factorial contribution estimates, working-flow or oxygen measures, full treatment exposure, R0 resection, PFS, OS, function and serious harm. Mission capital is £0 before those existing results. Decision rule. A scan, blood TGF-beta change, clean margin or PFS shift can keep a mechanism open but cannot release clinical use. A route advances only when working flow, drug or radiation delivery and patient benefit move together under a concurrent control.", "sourceIds": [ "kindler-2010-bevacizumab-calgb80303", "van-cutsem-2009-bevacizumab-erlotinib", "rougier-2013-aflibercept-vanilla", "kindler-2011-axitinib-pdac", "clinicaltrials-nct01746979-maestro", "van-cutsem-2016-maestro", "murphy-2019-losartan-lapc", "ramaswamy-2025-afpac-losartan", "clinicaltrials-nct03563248-losartan-localized" ], "links": [ { "target": "trials", "relation": "part-of" }, { "target": "failure-vascular-starvation-pdac", "relation": "maps" }, { "target": "claim-hypoxia-control-without-survival-pdac", "relation": "maps" }, { "target": "failure-losartan-delivery-signal-not-replicated", "relation": "maps" }, { "target": "unknown-localized-losartan-contribution-pdac", "relation": "resolves" }, { "target": "hypothesis-localized-losartan-existing-trial-gate", "relation": "implements" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong", "independentReplication": "strong", "sampleSize": "Four VEGF phase 3 trials n=2,387; MAESTRO n=693; AFPAC n=88; localized trial n=168", "effect": "Decision map; £0 new-cohort capital and two existing-data recovery routes", "limits": [ "Older treatment backbones", "Hypoxia sample access unconfirmed", "Localized trial result not posted", "Different vascular interventions cannot share one efficacy estimate" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Starvation, oxygen activation and vessel opening have different physical predictions." }, "animalEvidence": { "rating": "strong", "reason": "All three routes have model support, held below human results." }, "inVitroEvidence": { "rating": "moderate", "reason": "Hypoxia activation is testable in cells; perfusion and pressure need intact tissue." }, "reproducibility": { "rating": "strong", "reason": "The starvation failure repeats; the remaining questions are bound to existing controlled records." }, "commercialConflicts": { "rating": "moderate", "reason": "Anti-VEGF and hypoxia products were sponsor controlled; losartan is generic." }, "recency": { "rating": "current", "reason": "Old decisive phase 3 results are joined to 2025-2026 losartan evidence." } } }, "layout": { "x": 346, "y": 302 }, "updated": "2026-09-15" }, { "id": "trial-pdac-vte-action-map", "title": "PDAC clot-prevention action map: deliver known prophylaxis and do not duplicate active questions", "shortTitle": "PDAC VTE action map", "type": "trial", "status": "supported", "scope": [ "pdac" ], "summary": "Completed trials settle that prophylaxis prevents VTE; active and missing studies now divide implementation, novel oral prevention, surgery timing and survival claims.", "content": "Completed efficacy anchors: FRAGEM n=123, CONKO-004 n=312 and the CASSINI pancreatic subgroup n=273 repeat a large VTE reduction with dalteparin, enoxaparin and rivaroxaban. A five-study n=1,003 pool estimates an eight-point absolute reduction and NNT 11.9, without a clear major-bleeding increase; CONKO-004 found no PFS or OS gain. Action now is an early guideline-based offer to eligible high-risk patients with an individual bleeding, interaction and burden check, plus audited fill and persistence. Do not launch another broad conventional anticoagulant-versus-no-treatment efficacy trial. First recover imPaCT-PRO: NCT05178628 planned 450 patients and directly tested PFS and VTE, but status is unknown after its December 2024 completion estimate and no result is posted. CATIQ phase 3, NCT06861088, is recruiting an estimated 480 patients at three listed sites and tests Kinisoquin at two doses versus placebo for 16-week thromboembolic events, with one dose chosen after 26 events; PFS and OS are secondary. Its earlier 57-person study had no control, so the phase 3 is the clinical test. PREPOSTEROUS Pancreas, NCT05245877, is separately recruiting an estimated 800 surgical patients to LMWH before versus after surgery; the August 2026 primary date has passed, no result is posted and eligibility includes non-cancer operations. PINPOINT, NCT07196020, observes 380 mixed-cancer patients for VTE markers through 2029 and cannot establish an action. Refresh all four registries before a decision. Join no datasets unless treatment, clot, bleed, interruption and negative follow-up definitions can be aligned. New capital should test delivery or a named patient gap, not rediscover that anticoagulants reduce clots.", "sourceIds": [ "maraveyas-2012-fragem", "pelzer-2015-conko004", "vadhan-raj-2020-cassini-pdac", "frere-2020-pdac-thromboprophylaxis-meta", "ash-2023-primary-vte-prophylaxis-guide", "clinicaltrials-nct05178628-impact-pro", "clinicaltrials-nct06861088-catiq-p3", "clinicaltrials-nct05245877-preposterous-pancreas", "clinicaltrials-nct07196020-pinpoint-vte", "zwicker-2019-isoquercetin-catiq" ], "links": [ { "target": "trials", "relation": "part-of" }, { "target": "claim-anticoagulation-prevents-pdac-vte-not-survival", "relation": "uses" }, { "target": "unknown-pdac-vte-prophylaxis-rule", "relation": "tests" }, { "target": "failure-pdac-vte-prophylaxis-delivery-gap", "relation": "learns-from" }, { "target": "failure-impact-pro-status-unknown", "relation": "learns-from" }, { "target": "claim-kinisoquin-phase2-marker-not-clinical-proof", "relation": "tests" }, { "target": "company-quercis-pharma", "relation": "includes-trial-by" }, { "target": "hypothesis-pdac-vte-route-completion", "relation": "prioritizes" }, { "target": "programme-500m-integrated-mission", "relation": "must-not-duplicate" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "moderate for delivery", "randomised": "strong for conventional prophylaxis; none for the generated portfolio rule", "independentReplication": "strong for VTE prevention; none for implementation", "sampleSize": "Completed randomized n=123, n=312 and n=273; pooled n=1,003; live or unresolved registry plans n=450, n=480, n=800 and n=380", "effect": "Decision and duplication map; individual trial effects remain in linked records", "limits": [ "Registry status changes", "Active studies have no results", "Different drugs and settings", "Surgical and ambulatory questions differ", "Implementation design is generated", "No complete survival answer" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The map starts from repeated clinical clot prevention and keeps distinct patient settings apart." }, "animalEvidence": { "rating": "not-applicable", "reason": "The mapped decisions have direct human tests." }, "inVitroEvidence": { "rating": "weak", "reason": "Markers support Kinisoquin mechanism but cannot rank clinical actions." }, "reproducibility": { "rating": "moderate", "reason": "Conventional prophylaxis repeats, while delivery and novel oral treatment remain unresolved." }, "commercialConflicts": { "rating": "moderate", "reason": "Several drug trials have sponsor control, countered by academic studies and independent checks." }, "recency": { "rating": "current", "reason": "All live registry states were checked in September 2026." } } }, "layout": { "x": 160, "y": 84 }, "updated": "2026-09-14" }, { "id": "trial-pert-pdac-action-map", "title": "PDAC enzyme-replacement action map: finish the selected trial and measure delivery before another efficacy study", "shortTitle": "PERT action map", "type": "trial", "status": "hypothesis", "scope": [ "pdac", "pancreatic-all", "transfer-unproven" ], "summary": "Two small randomized trials are negative for short weight outcomes; three later studies test feasibility, dose or delayed treatment but still do not make survival the main decision.", "content": "Question: what evidence can change care beyond treating obvious maldigestion? Completed randomized anchors: NCT01587534 enrolled 67 and found no weight, nutrition, quality-of-life or OS benefit; the Japanese 88-person trial found no BMI benefit and an uncertain 19.0 versus 12.0-month survival signal. PANCAX-3 enrolled 36 at one centre, reports 30 evaluable and 29 meeting the adherence endpoint, but has one treatment arm and cannot estimate efficacy. NCT05642962 compares two PERT doses in 17 stage-IV FOLFIRINOX patients; it has no untreated arm, BMI is its primary outcome, it is active not recruiting and completion is estimated November 2026. PERTseverance is recruiting an estimated 100 across European centres and randomizes immediate versus three-month-delayed Creon. It selects head tumours, duct dilation, at least 5% weight loss and six-month life expectancy, but its estimated June 2026 primary date has passed, no result is posted and weight remains primary. Action now: refresh PERTseverance monthly, request its protocol and arm-level data, and join the three later study datasets under one field set. Do not launch another broad PERT efficacy trial before these data are obtained or a documented failure persists. The pooled audit must not pool treatment effects across controlled and single-arm designs. It should identify whether a declared maldigestion state and adequate meal-level dose predict stool, absorption and function response. Only a remaining named state with a plausible ten-point treatment-delivery gain can justify a new randomized trial. Separately, a prescribing programme must measure recommendation, fill and treatment gaps because the English audit and national shortage show that efficacy and delivery are different failures.", "sourceIds": [ "woo-2016-pert-rct", "saito-2018-pert-rct", "clinicaltrials-nct04098237-pancax3", "clinicaltrials-nct05642962-pertzye-dose", "clinicaltrials-nct06099119-pertseverance", "lemanska-2023-opensafely-pert-audit", "dhsc-2024-pert-shortage-alert" ], "links": [ { "target": "trials", "relation": "part-of" }, { "target": "unknown-pert-pdac-utility", "relation": "tests" }, { "target": "failure-pert-short-randomized-trials", "relation": "learns-from" }, { "target": "failure-pert-prescribing-and-supply-gap", "relation": "constrained-by" }, { "target": "claim-pert-survival-observational-randomized-unproved", "relation": "tests" }, { "target": "trial-pdac-cachexia-action-map", "relation": "extends" }, { "target": "programme-500m-integrated-mission", "relation": "must-not-duplicate" } ], "evidence": { "humanProspective": "strong for mapped studies", "humanRetrospective": "strong for delivery gaps", "randomised": "moderate for weight; weak for survival utility", "independentReplication": "none for the generated action rule", "sampleSize": "Completed randomized n=67 and n=88; single-arm n=36; dose comparison n=17; live delayed-control trial estimated n=100; prescribing audit n=22,860", "effect": "Action and duplication map; no pooled PERT effect", "limits": [ "Generated design", "Small heterogeneous trials", "Different formulations and doses", "Live trial result absent", "Weight-heavy endpoints", "Supply state changes", "No trial tests the entire prescription-to-survival path" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The route starts with measured maldigestion and adequate replacement rather than treating all weight loss as enzyme failure." }, "animalEvidence": { "rating": "not-applicable", "reason": "The mapped decision is directly testable in patients." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Bench activity cannot establish adherence, symptoms or cancer treatment delivery." }, "reproducibility": { "rating": "none", "reason": "No independent group has tested the full data-joining and release rule." }, "commercialConflicts": { "rating": "moderate", "reason": "Some studies include manufacturer collaboration; the action map requires independent analysis." }, "recency": { "rating": "current", "reason": "The live trial records and supply context were checked in September 2026." } } }, "layout": { "x": 157, "y": 81 }, "updated": "2026-09-14" }, { "id": "trial-regional-route-completion-action-map", "title": "Regional HPB route-completion action map", "shortTitle": "Regional route action map", "type": "trial", "status": "hypothesis", "scope": [ "pdac", "cholangiocarcinoma", "gallbladder", "hcc", "hepatobiliary", "pancreatic-all", "transfer-unproven" ], "summary": "The regional evidence is sufficient to justify a shared route audit, but not to choose one continent-wide intervention or infer treatment benefit from selected cohorts.", "content": "Question: what action can the mission test without pretending that Africa or Latin America is one health system? Evidence map. African pancreatic cancer: 20 studies and 26,850 cases show late presentation but lack stage for nearly 40% and complete treatment paths in more than half of studies. Sub-Saharan HCC: 39 reports across 15 of 48 countries show 6% curative treatment and 84% supportive care, but no common route denominator. Cape Town: a public/private cohort shows 68 versus 703 days median survival and a 54-point gap in supportive-care-only use, with major baseline differences; a South Africa/Sweden cohort shows 9.2% versus 42.5% curative-intent use and lower survival within treatment groups. Latin American CCA: five expert centres contribute 309 eligible cases and treatment detail, but exclude mandatory-data failures and people never reaching referral. GASPAR: seven sites in three countries contribute 79 pancreatic cases but are concentrated in private care. Common next action: one route dictionary and one-year consecutive capture. Site-specific next action: only after the audit identifies a repairable loss of at least 15 points, randomize the matching repair. Required controls are contemporaneous usual care, locked eligibility, independent adjudication of non-delivery, public missingness and loss-to-follow-up reports, and outcomes measured from population entry rather than expert-centre arrival. Do not pool cancers or countries for the primary result. Do not use post-treatment survival to claim treatment benefit. Do not start a new drug trial to solve a route that has not been measured.", "sourceIds": [ "gismalla-2026-africa-pancreatic-review", "sobnach-2024-ssa-hcc-systematic-review", "sobnach-2025-south-africa-hcc-disparities", "sobnach-2026-global-hcc-disparities", "da-fonseca-2024-latam-cca", "clinicaltrials-nct05924789-gaspar", "peixoto-2025-gaspar-results" ], "links": [ { "target": "trials", "relation": "part-of" }, { "target": "unknown-regional-route-loss", "relation": "tests" }, { "target": "hypothesis-regional-route-completion", "relation": "tests" }, { "target": "lab-africa-hpb-cancer-consortium", "relation": "could-run" }, { "target": "lab-uct-groote-schuur-hpb", "relation": "could-run" }, { "target": "lab-latam-cca-registry", "relation": "could-run" }, { "target": "lab-lacog-gaspar", "relation": "could-run" } ], "evidence": { "humanProspective": "weak — the linked networks exist, but the common audit and intervention do not", "humanRetrospective": "strong for large route gaps, weak for their causes", "randomised": "none", "independentReplication": "none for a common action rule", "sampleSize": "Evidence map spans five regional datasets; proposed trial size depends on the measured route loss and clustering", "effect": "Decision map and proposed design, not an intervention result", "limits": [ "Generated design", "Health systems and cancers differ", "Existing cohorts start at different route points", "No common outcome dictionary", "Cluster count may limit power", "Treatment completion can improve without survival benefit" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Time, stage, performance and organ reserve link route completion to treatment eligibility." }, "animalEvidence": { "rating": "not-applicable", "reason": "This is a human delivery design." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Laboratory evidence cannot validate the route action." }, "reproducibility": { "rating": "none", "reason": "No shared action rule has been reproduced across the named sites." }, "commercialConflicts": { "rating": "moderate", "reason": "Some networks include commercial centres or industry funding; governance must prevent product-led route selection." }, "recency": { "rating": "current", "reason": "The map uses reports through 2026." } } }, "updated": "2026-09-14" }, { "id": "trial-sil204-action-map", "title": "SIL204 action map: make human KRAS knockdown a gate inside the live trial", "shortTitle": "SIL204 action map", "type": "trial", "status": "planned", "scope": [ "pdac", "colorectal", "pan-cancer", "transfer-unproven" ], "summary": "The authorised adaptive trial can answer survival only if its early segments first prove target delivery and preserve complete denominators across mutation, route and chemotherapy.", "content": "First-generation boundary. NCT01676259 enrolled 59 people and randomized 37. The overall randomized median survival difference was 22.7 versus 21.9 months and was not significant. Mutation status was available for 31/59 = 52.5%. The G12D/V randomized subgroup contained 11 treated and five controls; its HR 0.59 had a 95% confidence interval from 0.18 to 1.96 and P=0.39. That cannot validate either benefit or the second-generation product. New trial. CTIS 2025-525157-39-00 authorises an integrated phase 2/3 of intratumour plus subcutaneous SIL204 and chemotherapy against chemotherapy alone, with 403 planned participants. Segment 1 admits stage III or limited metastatic disease with any KRAS mutation, tests three subcutaneous doses with fixed local dosing and measures safety plus plasma concentration. Segments 2 and 3 admit non-resectable stage III disease with any KRAS mutation. The primary survival comparison uses G12D/V participants receiving the selected dose across Segments 2 and 3. All other enrolled mutation groups, failed biopsies and unconfirmed results require explicit denominators. Public status. CTIS says authorised with German recruitment pending. A 29 July sponsor release says one Israeli site was cleared to start screening, first dosing still anticipated, about 18 in the run-in and about 166 randomized. The later filing and CTIS say about 403 total. No treated SIL204 participant or human target result is reported. Data rule. CTIS says participant-level data will not be shared. Mission action. Offer £0 and no separate cohort. Seek a contract for the protocol history, data dictionary, assay plan, participant flow, mutation result, all tumour and plasma measurements, chemotherapy delivery, harms, scans, quality of life and survival. Insert the paired multi-region target gate before Segment 2 and an independent statistics team before Segment 3. Reconcile 166 and 403. Report survival for the complete randomized population and the locked G12D/V primary set; do not pool Segment 1 safety participants into efficacy. Stop support if the sponsor refuses the data contract, no human knockdown is shown, systemic delivery remains unmeasured, procedure harm or chemotherapy loss crosses the gate, or the adaptive analysis can choose a favourable population after outcome access.", "sourceIds": [ "alagesan-2026-sig12d-loder", "ctis-2026-sil204-phase23", "silexion-2026-sil204-site-initiation", "silexion-2026-halfyear-filing" ], "links": [ { "target": "trials", "relation": "part-of" }, { "target": "claim-sil204-trial-scales-before-human-knockdown", "relation": "maps" }, { "target": "failure-sig12d-loder-randomized-os-not-shown", "relation": "repairs" }, { "target": "company-silexion-therapeutics", "relation": "run-by" }, { "target": "unknown-sil204-human-target-delivery", "relation": "maps" }, { "target": "hypothesis-multicompartment-kras-sirna-delivery", "relation": "gates" } ], "evidence": { "humanProspective": "weak — authorised and activated at one site without reported dosing", "humanRetrospective": "not-applicable", "randomised": "none — planned but not yet reported", "independentReplication": "none", "sampleSize": "Segment 1 plans 15-21; CTIS and filing plan about 403 total; site release describes about 166 randomized", "effect": "No SIL204 human effect; trial plan only", "limits": [ "Sponsor-declared status outside Germany", "Conflicting public sample descriptions", "No human target result", "No participant-level data-sharing plan", "Adaptive population and dose choices require strict firewalls" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "The intervention targets a validated driver with a plausible cutting mechanism but unproved human delivery." }, "animalEvidence": { "rating": "moderate", "reason": "Sponsor models support entry into Segment 1, not automatic expansion." }, "inVitroEvidence": { "rating": "moderate", "reason": "Mutant RNA reduction is reported in cell systems." }, "reproducibility": { "rating": "weak", "reason": "The second-generation product has no human dataset." }, "commercialConflicts": { "rating": "high", "reason": "The listed sponsor owns the product and trial, and does not plan individual-data sharing." }, "recency": { "rating": "current", "reason": "Reflects June authorisation and July site activation." } } }, "layout": { "x": 190, "y": 148 }, "updated": "2026-09-15" }, { "id": "trial-small-pnet-action-map", "title": "Small pNET studies cover detection, observation and treatment but not lifetime net benefit", "shortTitle": "Small pNET action studies", "type": "trial", "status": "open", "scope": [ "pnet" ], "summary": "ASPEN, PANDORA and DutchMEN studies each answer a fragment; no randomized pathway links early detection to safer surgery timing and less metastatic disease.", "content": "The study map has six decisions. 1. When to start inherited-risk imaging: DutchMEN found five clinically relevant non-functioning pNETs before age 18 among 350 people and modelled 2.5% risk at age 13.5, but did not compare schedules or mortality. 2. Whether a small sporadic tumour can be observed: PANDORA followed 76 people for median 17 months, while ASPEN observed 406 of 500 for median 25 months. Neither reported a new distant metastasis in surveillance, but both are too short for lifetime safety. 3. Whether surgery improves outcome: ASPEN was non-randomized and strongly centre- and preference-selected; its registry now says unknown status with a passed 2024 completion estimate and no final result. DutchMEN1 found no significant surgery association with liver metastasis or death among 152, but confidence intervals were very wide. 4. Which operation is acceptable: a 460-person small-tumour surgery cohort found 14.8% major morbidity and 0.9% 90-day death, with procedure-specific fistula and lymph-node trade-offs. 5. Whether medicine can delay surgery in MEN1: a 42-person prospective non-randomized comparison found much less progression with lanreotide over 73 months, but no proof of fewer metastases, operations or deaths. 6. Whether a blood test selects future danger: a 110-person MEN1 mRNA study failed prediction of new tumour, growth or metastasis. Repeat biopsy and dual PET at metastatic progression are a separate later decision and have changed treatment without proving outcome benefit. The missing trial starts with the actual choice: among 1-2 cm tumours without a mandatory operation feature, does a locked time-updated rule improve quality-adjusted life without increasing metastatic progression? It must retain every crossover, use procedure-specific harm, link at least five years of outcomes and keep sporadic and MEN1 strata separate.", "sourceIds": [ "partelli-2022-aspen-interim", "clinicaltrials-nct03084770-aspen", "heidsma-2021-pandora-pnet", "maekawa-2026-small-pnet-surgery-morbidity", "klein-haneveld-2021-men1-young-screening", "nell-2018-men1-surgery-watchful-waiting", "faggiano-2020-men1-lanreotide", "van-vliembergen-2026-men1-mnet-assay", "mollazadegan-2025-pnet-recharacterization" ], "links": [ { "target": "trials", "relation": "part-of" }, { "target": "claim-small-nfpnet-surveillance-surgery-tradeoff", "relation": "contains" }, { "target": "claim-men1-surveillance-finds-lesions-not-danger", "relation": "contains" }, { "target": "failure-aspen-cannot-estimate-surgery-benefit", "relation": "contains" }, { "target": "failure-men1-mnet-assay-no-progression-prediction", "relation": "contains" }, { "target": "unknown-small-pnet-surgery-selector", "relation": "exposes" }, { "target": "unknown-pnet-reclassification-actionability", "relation": "distinct-from" }, { "target": "hypothesis-pnet-time-updated-surgery-rule", "relation": "constrains" } ], "evidence": { "humanProspective": "strong — multiple prospective cohorts", "humanRetrospective": "strong — national and multicentre comparison cohorts", "randomised": "none for the surgery or surveillance action", "independentReplication": "moderate — component findings repeat, the full action path does not", "sampleSize": "ASPEN 500; PANDORA 76; DutchMEN young screening 350; DutchMEN surgery 152; MEN1 lanreotide 42; mNET assay 110; surgery harm 460", "effect": "Short observation appears feasible and operation harm is material; mortality-improving action selection remains unproved", "limits": [ "Different populations and endpoints", "Short surveillance follow-up", "No randomized surgery comparison", "Rare metastasis", "Current grade can differ from future grade", "Registry completion is unresolved" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The mapped decisions match inherited susceptibility, slow localized growth, occasional spread and changing metastatic state." }, "animalEvidence": { "rating": "weak", "reason": "Models do not settle human surveillance intervals or operation harm." }, "inVitroEvidence": { "rating": "moderate", "reason": "Molecular candidates exist but have not completed the risk-to-action path." }, "reproducibility": { "rating": "weak", "reason": "Observation feasibility repeats, while no full pathway has reproduced improved patient outcomes." }, "commercialConflicts": { "rating": "moderate", "reason": "Drugs, imaging, assays, surgery and ablation each create incentives to favour a route." }, "recency": { "rating": "current", "reason": "The map includes live registry status and 2026 biomarker and surgical evidence." } } }, "updated": "2026-09-14" }, { "id": "trial-vcn01-action-map", "title": "VCN-01 action map: recover VIRAGE before funding confirmation", "shortTitle": "VCN-01 action map", "type": "trial", "status": "mixed", "scope": [ "pdac" ], "summary": "VIRAGE supplies a credible randomized signal and a human delivery mechanism; the next actions are complete data recovery, six-person dosing work and an independently analysed phase 3, in that order.", "content": "Mechanism stage. NCT02045589 and NCT02045602 show that VCN-01 can replicate in human PDAC, release PH20 and alter the tumour barrier. Intravenous viral genomes were found in five of six day-eight biopsies. High Ad5-neutralising antibodies excluded 26% screened, and one phase 1 dose level produced a treatment-related death, so access and dose safety remain material. Contribution stage. NCT05673811 VIRAGE is completed with 112 actual enrollment and no posted results. Its sponsor reports 96 people in the FAS: OS 10.8 versus 8.6 months, HR 0.57, 95% CI 0.34-0.96, log-rank P=0.0546; PFS 7.0 versus 4.6 months, HR 0.55, P=0.0105. The protocol planned 92 and 57 deaths, uses treated rather than all-randomized definitions, and contains conflicting 0.10 and 0.05 two-sided alpha logic. ESMO also reports a baseline liver-metastasis imbalance. The strongest repeat-dose estimate selects people after they survived and received four cycles, so it is not a dose contribution test. Dosing stage. VIRAGE2, 2026-525566-21-00 and NCT07701486, recruits six people at one Spanish centre for three VCN-01 doses 56 days apart. Its main outputs are safety and blood viral genomes; it has no control and cannot confirm survival. CTIS says no participant-level sharing. Confirmation stage. FDA and EMA discussions support a future double-blind placebo-controlled adaptive phase 3, but Theriva says current cash does not fund it. Mission action. Invest £0 now. First recover all 112 VIRAGE records, final analysis plan and full protocol history for independent reproduction. Use VIRAGE2 only to test repeat-dose feasibility, neutralising-antibody escape, chemotherapy delivery and serious harm; do not read its response or survival descriptively as efficacy. A phase 3 may compete for capital only after the all-randomized result survives the gate and a partner supplies the product and most trial cost under public-data and negative-publication terms. It must stratify liver metastasis and baseline Ad5 antibodies, use identical chemotherapy, all-randomized OS, quality-adjusted survival and a bound on severe toxicity. No effect is assigned to the second or third dose without randomized dose allocation or a valid exposure-response method fixed before outcome access.", "sourceIds": [ "bazan-peregrino-2021-vcn01-intratumour", "garcia-carbonero-2022-vcn01-intravenous", "clinicaltrials-nct05673811-virage", "theriva-2023-virage-protocol", "theriva-2025-virage-topline", "esmo-2025-virage-2216mo", "ctis-2026-virage2", "theriva-2026-q2-10q" ], "links": [ { "target": "trials", "relation": "part-of" }, { "target": "claim-vcn01-randomized-signal-not-survival-proof", "relation": "maps" }, { "target": "failure-virage-analysis-and-survivor-selection", "relation": "repairs" }, { "target": "unknown-vcn01-all-enrolled-survival", "relation": "maps" }, { "target": "hypothesis-vcn01-replication-to-survival-gate", "relation": "gates" }, { "target": "company-theriva-biologics", "relation": "run-by" }, { "target": "failure-oncolytic-virus-pelareorep", "relation": "learns-from" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "not-applicable", "randomised": "moderate", "independentReplication": "weak", "sampleSize": "Phase 1 PDAC n=26, VIRAGE registry n=112 and FAS n=96, VIRAGE2 planned n=6", "effect": "Randomized OS HR 0.57 and PFS HR 0.55 in the reported FAS; confirmation absent", "limits": [ "No full phase 2 report", "No all-randomized result", "Analysis rules conflict", "Repeat-dose trial is single-arm", "Future phase 3 is unfunded" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "Human biopsies show tumour delivery and replication, and the randomized direction is favourable." }, "animalEvidence": { "rating": "moderate", "reason": "Models support lysis and barrier change but have been overtaken by human tests." }, "inVitroEvidence": { "rating": "moderate", "reason": "The engineered mechanism is measurable after cell entry." }, "reproducibility": { "rating": "weak", "reason": "No independent controlled study reproduces the phase 2 effect." }, "commercialConflicts": { "rating": "high", "reason": "Theriva controls the agent and all current trial datasets." }, "recency": { "rating": "current", "reason": "Includes first VIRAGE2 dosing and August 2026 financing limits." } } }, "layout": { "x": 176, "y": 96 }, "updated": "2026-09-15" }, { "id": "trials", "title": "Clinical trial map: results, live decisions and registry-only plans", "shortTitle": "Trials", "type": "trial", "status": "supported", "scope": [ "pdac", "pnet", "cholangiocarcinoma", "gallbladder", "hcc", "colorectal", "pan-cancer" ], "summary": "The useful trial map is organized by the decision each study can settle, with completed randomized evidence kept separate from recruiting or registry-only programmes.", "content": "Completed PDAC survival anchors: PRODIGE-24 established adjuvant modified FOLFIRINOX but left most patients at risk of recurrence; NAPOLI-3 moved first-line median survival by about two months; RASolute 302 made common-RAS inhibition the first large targeted survival break after prior treatment. Completed negative or bounded tests include HALO-301 stromal hyaluronan removal, POLO PARP maintenance overall survival, LAP07 local radiation, pancreatic checkpoint blockade, ELI-002 phase 2, pamrevlumab and multiple repurposed drugs. Current systemic-control decisions: NCT07491445 tests first-line daraxonrasib alone and with chemotherapy; NCT07252232 tests post-resection daraxonrasib after multiagent chemotherapy; NCT07805954 tests dual broad and G12D-selective RAS inhibition; NCT07409272 tests first-line setidegrasib; NCT07232875 tests HRS-4642 plus chemotherapy. These are sponsor-run plans with estimated completion dates, not evidence. Current low-burden and immune decisions: IMCODE003 randomizes personalised neoantigen vaccine after resection. Three PDAC ctDNA programmes now test distinct actions. DYNAMIC-Pancreas covers adjuvant choice, but its published 102-patient upfront-surgery cohort established feasibility and prognosis rather than a randomized action effect. CIRCPAC randomizes an estimated 1,000 Danish patients to ctDNA-triggered intensified imaging or standard surveillance, with completion estimated in 2032. MAP-03 plans an early treatment switch in 100 Chinese patients, but its registry has not been verified since 2023 and its estimated primary-completion date has passed without results. Two 2026 cohorts show postoperative ctDNA sensitivity and predictive values vary materially, so the active studies should not be treated as interchangeable or as validation of a ctDNA-only rule. Detection decisions: ENZEAVOUR follows 10,000 Japanese screening participants through blood test and diagnostic imaging, while the Kobe CT study is retrospective and cannot measure clinical utility. New-onset-diabetes work-up has its own active sequence. A prospective 18,838-person cohort established 0.62% three-year risk; a 93-scan CT pilot found one stage-IV PDAC and incidental extrapancreatic findings in 56%; PANDOME found one stage-I PDAC among 109 MRI participants; EDI enrolled 8,869 but uses a non-randomized site comparison, has no posted result and completes in 2028. These do not justify a duplicate large utility trial until the unresolved action and selector are named. Delivery evidence now includes GASPAR, an observational LACOG cohort across seven sites and three Latin American countries. Its 79 advanced pancreatic cases expose treatment attrition, selection and representation limits but cannot test treatment effect; its abstract and registry enrolment totals conflict. Related-cancer reference trials show properties that change outcomes: OCLURANDOM supports receptor-directed radiation in pNET; SEQTOR found no clear survival winner between everolimus-first and chemotherapy-first pNET sequences; KEYNOTE-177 tests a checkpoint-sensitive colorectal subgroup; CLOCC and TransMet test aggressively selected liver-only colorectal disease; SURF compares resection with ablation in small HCC; TOPAZ-1 and ClarIDHy test treatment across biliary or molecularly selected cholangiocarcinoma. Trial priority is not enrolment count. A trial ranks higher when it tests a bottleneck, uses a current comparator, locks a selector before analysis, captures serial tissue and blood, measures overall survival or a validated patient benefit, and publishes failed arms. Registries establish intent, planned sample size and endpoints only. Every live status and completion date must be refreshed before an investment or clinical decision. Germline delivery now has its own action map. Automatic, embedded and oncology-led patient testing can raise completion, but service reports use different denominators and do not prove health benefit. GENERATE randomized remote delivery only after motivated relatives entered and achieved 90% testing; GIFT found the larger population loss was invitation, with only 19.2% of reported relatives invited. England's R367 rule remains narrower than all-patient US guidance. The next trial should not repeat a consented-participant uptake report: it should count all newly diagnosed PDAC patients, measure variant and action yield inside versus outside R367, and randomize consented direct family contact against the standard letter.", "contentSections": [ "Electrochemotherapy has passed a delivery test, not a patient-benefit test. The 25-person cohort lost 11 six-month scans through death and compared surgeon-chosen electrode layouts. PanECT enrolled ten without a control and leaves survival outcomes unposted. The needed 90-person same-chemotherapy randomization transitioned to EU CT 2024-519067-17-00 without a public result. Recover it before another cohort or equipment purchase.", "Thermal focused ultrasound has a repeated pain signal but no trustworthy survival result. A 120-person paper reports large pain and survival differences but contradicts itself about random assignment. HIFU-PC15 has targeted 40 since 2017 without a public result; the 90-person Suizenji randomized study estimates completion in 2027. Recover those records before another cohort. Keep SonoPANIII separate because its 360-person IMD10 trial is explicitly non-thermal and non-tissue-destructive.", "Radiofrequency ablation has a complete randomized answer for the tested route. PELICAN analysed 188 people after induction treatment: median survival was 12.1 versus 11.6 months, HR 1.07, while grade 3 or worse serious events were 27% versus 11%. Only 70 of 95 assigned to ablation received it, largely because surgery exposed hidden spread. Stop new treatment cohorts and equipment spending; at most £150,000 may reuse existing specimens to explain tissue effects without reviving the failed treatment claim.", "MRI-guided adaptive radiation has passed a selected delivery and safety test, not a survival test. The 136-person phase 2 excluded early distant progressors and had no chemotherapy-only arm. LAP-ABLATE planned the correct 267-person comparison but remains not yet recruiting on a stale vendor-owned record; MASPAC stopped at nine. Rebuild around measurable cross-system performance and full harm before any patient or capital commitment.", "Particle therapy has real beam advantages but no randomized pancreatic result. Carbon-ion cohorts report high local control while distant spread remains common; proton cohorts are small. CIPHER enrolled zero, completed PACK has no posted result and PAN009-18 now estimates primary completion in 2039. The current action is £0 for another cohort and at most £500,000 to recover complete outcomes and recalculate delivered dose under one system before any randomized survival proposal.", "Regional hyperthermia has one reported randomized package and one cleaner but unpublished randomization. HEAT enrolled 117 of 336 planned people, missed disease-free survival and changed heat, cisplatin, gemcitabine schedule and control treatment together. HEATPAC was designed to change heat alone while measuring temperature, but its registry is unknown, stale since 2017 and has no result. The trial action is recovery, not another cohort: £0 for recruitment and at most £250,000 to reconstruct assignments, deaths, harms and thermal delivery.", "Polyamine action map. Ivospemin's early 12-of-28 response signal is uncontrolled and includes serious retinal and liver harm. ASPIRE supplies the needed masked placebo comparison, but its contractor began termination for non-payment; ClinicalTrials.gov still says recruiting while Europe says ended. Release at most £250,000 for custody and completeness diligence. Require 400 reconciled assignments, a usable blind and key, and at least 90% 12-month vital status before pricing any data rescue.", "The hidden-spread action map joins modern evidence without treating it as one trial. Mayo found an 18% combined visible-spread or cytology yield with 0.1% major harm from a separate procedure. PREOPANC-2 associated same-session staging with non-therapeutic laparotomy falling from 17.1% to 4.5%, but did not randomize staging. The live 200-person Japanese study remains single-arm and yield-led. The non-duplicate experiment belongs inside the English surgical route and must separate result type, timing, delay and patient action.", "Pancreatic precursor action is not a solved trial class. EA2185 planned 4,606 randomized patients but now lists a 770-person observational biomarker cohort, with both randomized arms closed from August 2024. The 100-person 3P-C sulindac study has completed primary follow-up but has no posted results and mixes cyst growth, resection and cancer in one endpoint. SIPP-T3 stopped recruiting at four of 212 planned participants. ChARM shows that some cysts can shrink after EUS-guided ablation, but its long-term report contains 52 selected completers and no powered cancer endpoint. These programmes do not yet show prevention of HGD, invasive cancer or death.", "Gallbladder prevention has active recruitment but no complete action trial. EULAT estimates 15,000 cancer, dysplasia and gallstone participants and P-iGBC may scale from feasibility to 30,000 routine UK operations; both end at diagnosis or pathology rather than future mortality. Chile BiLS supplies 313 HGD-plus cases and matched gallstone controls for candidate discovery. NCT03762837 planned 100,000 people with benign disease or healthy controls, but still says not yet recruiting, retains a passed 2022 primary date and has no results. Imaging and blood classifiers must first validate in these consecutive pathways; only a later country-specific randomized or staged rollout can show whether changing optional surgery prevents death.", "Small non-functioning pNET has prospective observation but no action trial. ASPEN assigned 406 of 500 to surveillance and 94 to surgery by centre choice; patient or centre preference explained 79 operations, severe complications affected 13% and median follow-up was 25 months. PANDORA observed 76 for 17 months. Neither had a new surveillance metastasis, but neither settles lifetime net benefit. ASPEN's registry targets 1,000, now reports unknown status after a passed 2024 completion estimate and has no final posted result. A valid next study must retain crossover, model metastatic progression and procedure-specific harm together, and randomize only the 1-2 cm group for whom both routes remain acceptable.", "Cholangiocarcinoma has three separate action gaps. PSC surveillance found severe or progressive duct change in 23.8% of 512 people but did not create durable curable detection. Liver-fluke control removes a proven cause, yet NCT05321992 ends at 12-month behaviour and infection rather than cancer. Postoperative ctDNA detected 15 of 16 recurrences in one cohort and strongly separated risk in STAMP, but no trial assigned treatment or imaging from the result. ACTICCA-1 enrolled 789 people and remains without posted results after its estimated 2025 completion. The next programme should rank these routes under shared data checks while keeping each population, action and endpoint separate.", "HCC prevention and surveillance now have an explicit action map. ATTENTION's early-antiviral interim result did not cross its stopping boundary. NCT00190385 rejects a general three-month ultrasound interval; NCT02272504 rejects its tested three-marker add-on; NCT02514434 supports an MRI stage and false-referral signal without mortality proof. TRACER, NCT06084234, is recruiting an estimated 5,500 people and tests late-stage HCC plus physical, financial and psychological harm. PREMIUM, NCT05486572, is recruiting 4,700 and tests HCC mortality. Do not duplicate either modern question before its result or a documented failure.", "Colorectal evidence is now mapped by action. NordICC shows an invitation effect: 42.0% attended, incidence fell, and mortality did not fall significantly at 13 years. DYNAMIC stage II validates one treatment-omission rule. DYNAMIC-III rejects its tested stage III omission and escalation rules. ALTAIR rejects trifluridine/tipiracil at molecular relapse. FIND increased curative-intent treatment after ctDNA-triggered CT but has no mature survival result. CIRCULATE-US, NCT05174169, is recruiting 1,912 planned participants, completes its primary endpoint in March 2029 and has no results. Do not collapse these into an assay claim or duplicate the live chemotherapy questions.", "Regional delivery does not yet have an action trial. African pancreatic and HCC evidence shows large late-stage and treatment losses but lacks a common dated route. The Cape Town and South Africa/Sweden HCC comparisons cannot isolate a treatment, insurance or referral effect. LATAM-CCA and GASPAR prove that multi-country records can be joined, but both begin in selected centres and neither has a population denominator. The proposed action map first requires one year of consecutive capture at five African and five Latin American sites. Only a repairable loss of at least 15 percentage points at two sites or 100 patients can release a site-matched route-completion trial.", "Cachexia now has a live PDAC action trial. NCT06989437 randomizes an estimated 982 metastatic-PDAC patients at 220 locations in 21 countries to ponsegromab 200 mg, 400 mg or placebo with first-line chemotherapy. Week-12 weight and anorexia are primary; OS, PFS, response, function and chemotherapy dose changes are secondary. It excludes ECOG above 1 and patients unable to complete the first chemotherapy cycle. Do not duplicate it before the estimated January 2028 primary completion unless the trial fails, usable data are unavailable or a named frail, adjuvant or low-resource state remains uncovered.", "Pancreatic enzyme replacement has two small completed randomized failures and three incomplete action studies. Trials of 67 and 88 patients missed eight-week weight or BMI endpoints and did not establish survival. PANCAX-3 enrolled 36 and has no untreated control. NCT05642962 compares standard with high-dose Pertzye in 17 patients and cannot estimate treatment versus no treatment. PERTseverance randomizes an estimated 100 selected patients to immediate or three-month-delayed Creon, remains recruiting after its estimated June 2026 primary date, has no result and uses six-month weight as its primary endpoint. Do not launch another broad PERT trial. Obtain these data and test symptoms, absorption, function, chemotherapy delivered, quality-adjusted survival and supply continuity before a new efficacy study.", "PDAC venous-thrombosis evidence now separates a settled drug-class effect from four live decisions. FRAGEM, CONKO-004 and the CASSINI pancreatic subgroup repeat a large fall in VTE, and a five-trial pool of 1,003 patients estimates an eight-point absolute reduction and NNT 11.9; CONKO-004 found no PFS or OS benefit. imPaCT-PRO planned a 450-person PFS and VTE phase 3 but has unknown status after a passed completion estimate and no posted result. CATIQ phase 3 recruits an estimated 480 for Kinisoquin versus placebo; its earlier 57-person study changed clotting markers without a control. PREPOSTEROUS Pancreas separately tests LMWH timing around surgery, and PINPOINT is observational. Do not repeat conventional anticoagulant efficacy. Recover missing data, refresh registries and test a seven-day assessment-to-use route with clot and bleeding outcomes.", "Biliary obstruction now has an action map rather than a generic stent claim. A 202-patient randomized trial found serious complications of 39% with surgery within one week versus 74% after four to six weeks of mainly plastic drainage. A 2026 trial of 284 patients found 29.0% versus 26.5% with metal-stent drainage and surgery within four weeks, meeting a wide 15-point noninferiority margin. During neoadjuvant therapy, metal stents reduce obstruction and reintervention; covered and uncovered designs fail differently. Randomized malignant-obstruction evidence makes EUS-guided drainage a credible expert-centre alternative, especially after failed ERCP, but does not show a PDAC survival gain. Do not repeat broad device comparisons. Test indication, 48-hour plan, durable drainage, rescue, infection response and treatment completion as one route.", "The PDAC surgical-route map separates observational centralization from randomized procedure evidence and route completion. LEOPARD-2 rejects laparoscopic Whipple as a recovery shortcut under its tested conditions. DIPLOMA supports qualified minimally invasive left pancreatectomy, while EUROPA, the 2024 Chinese trial and PORTAL give mixed and setting-bound robotic Whipple evidence. None answers whether England's 23 specialist centres can raise postoperative treatment from 62% within 14 weeks or complete the planned course without more 90-day harm. The next trial tests that route, not another broad device contest.", "Post-resection follow-up now has its own action map. Two observational comparisons favour routine imaging but cannot remove treatment-selection and centre bias. NCT04875325 RADAR-PANC already randomizes 306 planned Dutch and UK patients to three-monthly CA19-9 plus CT or mainly symptom-led follow-up, with overall survival primary. Its registry is stale, has no result and estimates October 2026 completion. CIRCPAC asks a different ctDNA-directed intensity question. The programme first finishes and aligns these trials rather than launching another broad scan study.", "Perioperative timing now has its own action map. PREOPANC, NORPACT-1 and CISPD-1 give conflicting survival results under different regimens and settings. Across nine randomized resectable-disease trials, event-free survival improves, HR .77, while overall survival remains inconclusive, HR .85, 95% CI .68-1.05. NCT04340141 A021806 has 358 actual participants and directly compares the same twelve planned mFOLFIRINOX cycles split before and after surgery against all after surgery. Finish its 2028 survival follow-up before another broad timing trial.", "The residual-cell action map distinguishes human prognosis, route association, end-stage state and model mechanism. Marrow studies of 175 and 48 people predict outcome but do not prove dormancy. A 744-person recurrence cohort, 21 sequenced primaries and 13-person rapid-autopsy atlas show route and state diversity without sampling the disease-free cell. The existing £25m observatory runs a 250-person nested ancestry-state-time test; no anti-NET, EREG or other dormancy treatment advances from model evidence alone.", "Older and vulnerable PDAC now has a joined action map. GIANT randomized 176 people and found median survival 4.7 versus 4.4 months for two reduced-intensity regimens. GrantPax enrolled 32 of 135 and misrouted 15/32 against its assessment. PAMELA-70 posted only one combined outcome after treating 69. Three mixed-cancer randomized trials reproduce lower severe toxicity when assessment triggers care. The next test is the support route, not another birthday-defined drug comparison.", "Locally advanced surgery now has a separate action map. NEOLAP shows that roughly one third of early non-progressors can reach resection under protocol, while PREOPANC-4 shows 44% major morbidity and 3% 90-day mortality in a concentrated expert route. TAPS models predict technical selection and prognosis but explicitly cannot estimate surgery's effect. PELICAN shows why local treatment needs randomization. NCT06714604 answers treatment duration and the two-person estimated PROACTIVE pilot answers neither comparative survival nor selection. The uncovered trial is resection versus continued treatment after the same response gate.", "Organoid guidance now has its own action map. HOPE, Beutel and ORGANOPREDICT support conditional response correlation but took 53 to 96 days and lost substantial denominators. PASS-01 shows second-line routing is late. NCT04931381 and NCT04931394 are the direct randomized utility tests, but both have unknown status, stale July 2022 verification, passed May 2025 completion estimates and no posted result. NCT05351983 completed 60 without results; PROSPER-PANC is observational; ADOPT and UNITEPANC are single-arm. Recover existing data before any duplicate cohort or trial.", "FAP-radioligand map. The 2022 compassionate-use report and LuMIERE phase 1 establish delivery and a 9.25-GBq phase 2 dose, not PDAC benefit. NCT04939610 already covers non-randomized PDAC monotherapy and first-line modified-FOLFIRINOX combination work through an estimated 222-person programme. NCT07229768 plans four people and cannot estimate efficacy. FRONTIER stopped after 20 for a business decision without results. The 3BP-3940 series mixes 21 cancers, three isotopes and other care. Recover lesion dose and every denominator before considering a controlled test; do not add another single-arm PDAC cohort.", "Autophagy action map. NCT04386057 directly rejected the tested HCQ contribution: four-month disease control was 1/20 with ERK plus HCQ and 1/19 with ERK alone. NCT05221320 found 0/18 pancreatic responses and NCT04214418 never opened phase 2 after limited tolerability and efficacy. NCT04524702 is an uninterpretable ten-person, four-drug, single-arm signal. NCT04892017 enrolled 91 with a direct ULK1/2 inhibitor, then terminated for a stated business decision without final results. Freeze new HCQ trials and recover that dataset before a tumour-measurement study; no current treatment trial is funded.", "Sonoporation now has a completed randomized decision rather than an open replication request. ChiCTR2100044721 reported OS 9.1 versus 6.1 months among 78 randomized people. NCT04821284 estimates 120, reports 97 enrolled in an accepted abstract and no overall PFS or OS difference, both P>0.36. Its gemcitabine subgroup cannot be interpreted without counts, confidence intervals and a planned interaction. Release the complete trial, reconcile enrollment and keep the 13-person perfusion study separate from treatment effect; do not open another efficacy cohort.", "SIL204 has an authorised trial but no reported treated participant. CTIS 2025-525157-39-00 plans 403 across a 15-to-21-person safety stage, randomized dose selection and confirmation. One Israeli site was cleared to screen in July 2026; first dosing was still anticipated, and German recruitment is pending. The public release describes about 166 randomized people while CTIS and the later filing say 403 total. Make paired multi-region KRAS knockdown and public individual data gates before expansion; do not add a mission cohort.", "VCN-01 action map. Early studies show human tumour delivery and replication. VIRAGE is completed with 112 actual enrollment but reports only a 96-person FAS: OS 10.8 versus 8.6 months, HR 0.57, log-rank P=0.0546, and PFS 7.0 versus 4.6 months, HR 0.55. The public protocol conflicts on alpha and analysis population, and no registry results or full paper are posted. VIRAGE2 recruits six people for three doses and can test repeat-dose safety and blood virus only. Recover all 112 before a blinded phase 3 is funded; no new mission trial starts.", "MTAP-PRMT5 action map. NCT05732831 recruits an estimated 225 people and NCT06922591 an estimated 183, but both are open-label and nonrandomized and neither posts results. Interim sponsor data cover 54 combination-treated PDAC patients and 39 with at least 14 weeks since first dose. Preserve all denominators, confirm exact MTAP loss and add a RAS-drug-alone arm before a component claim. A package-versus-chemotherapy phase 3 answers package value, not vopimetostat's added value.", "CLDN18.2 antibody action map. GLEAM has already answered the plain-antibody question negatively in 393 high-marker patients: OS HR 0.999. Anetumab, EBC-129 and SHR-3821 cannot transfer efficacy to another carrier or payload. IBI389 showed 7/23 preliminary responses at its selected dose and marker, but the mature 121-person paper does not report efficacy for its 73 pancreatic patients. IBI343 showed 10/44 responses in a high-marker analysis and 0/12 below it; the existing 201-person placebo-controlled NCT07066098 now owns the survival question. Require exact assay rules, full denominators, survival and toxicity; fund no duplicate trial.", "Engineered-cell action map. Recover all 54 NCT03323944 participant routes before accepting its nine-person mesothelin paper as complete. Release the full 134-person NCT03874897 record and identify every pancreatic patient before pooling satri-cel. Treat NCT05911217's six-person adjuvant result as safety and feasibility until a concurrent group starts at the same disease-free landmark. Use the existing 30-person NCT04146298 G12V-TCR trial to test repeatable response, tumour presentation, entry, persistence and receptor rejection. Mission capital is £0: the programmes already exist and CARsgen reported RMB1.400bn cash.", "Immune-reprogramming action map. Stop broad NIS793 and cabiralizumab cohorts. For up to £1m, test the fixed angiogenin claim in stored daNIS-2 samples only with complete randomized outcome linkage and publication rights. Before mitazalimab money, release all 94 OPTIMIZE-1 records, audit manufacture and secure a solvent owner. A roughly 450-person placebo-controlled mFOLFIRINOX survival trial can receive up to £25m only if a partner supplies drug and at least the other £25m.", "Vascular and hypoxia action map. Stop unselected VEGF starvation after four failed phase 3 survival tests. Recover MAESTRO's 693-person record and any fixed baseline hypoxia material before considering the route. Use the existing 168-person NCT03563248 allocation to settle localized losartan; require flow, treatment delivery, recurrence, survival and harm rather than clean-margin resection alone. New-cohort mission cost is £0.", "Microbiome action map. Finish randomized FMTPanc for stool-transfer safety and PANORAMIX Step 2 for antibiotic contribution. The latter must reconcile its 142-versus-206 whole-trial denominator and join tumour bacterial function to active gemcitabine, six-month survival, progression, severe toxicity and resistant organisms. Treat the 25- and 20-person single arms as mechanism studies, recover the terminated 13-person MRx0518 record, and start no new product cohort.", "Vaccine action map. Stop shared-peptide, whole-cell and bacterial-vector products that failed standard-care survival comparisons. Recover the complete 158-person AMPLIFY-7P record after its randomized PDAC endpoint miss. Treat the new 20-person ELI-002 checkpoint pilot as safety only and the 37-person KRAS prevention study as immunity only. IMCODE003 is active but no longer recruiting and estimates 260 people; its all-assigned recurrence, survival, manufacture and harm record owns the next personalized-vaccine decision. Fund no duplicate cohort.", "Adenosine action map. ARC-8 is a mechanism and feasibility study because every randomized patient received quemliclustat; its survival comparator came from other trials. Oleclumab missed its randomized response endpoint and did not establish overall survival. PRISM-1 already completed enrollment for a quadruple-masked, placebo-controlled 610-person all-randomized survival comparison. Obtain its protocol and analysis plan, then wait for the expected first-half 2027 result. Mission trial capital is £0.", "Intra-arterial action map. TIGeR-PaC closed enrollment in August 2026 and reported 78 of 86 required deaths. Its final comparison can settle whole-package survival and total harm after induction, not the catheter's separate contribution because route, nab-paclitaxel and schedule differ together. The 16-person blood study does not measure active drug in tumour cells. Enrol nobody new; recover the final plan and every randomized record, then use at most £250,000 for independent analysis after access.", "Irreversible-electroporation action map. CROSSFIRE stopped for futility against radiotherapy and cannot estimate benefit over chemotherapy alone. The DIRECT registry's 18-versus-10-month association follows a median eight-versus-four-month diagnosis-to-enrollment filter and mostly reconstructed controls. Separate phase 3 NCT03899636 already tests the same chemotherapy with or without IRE, but its count remains estimated and its result is missing after estimated completion. Enrol nobody new; recover its complete all-randomized record for at most £250,000 after access rights.", "Tumour-treating-fields action map. PANOVA-3 supports the FDA-approved locally advanced use with gemcitabine and nab-paclitaxel: 16.2 versus 14.2 months median survival in 571 assigned people. Keep the unchanged PFS, 11.2-hour median daily use, skin harm and lack of independent reproduction visible beside it. PANOVA-4 cannot isolate either the device or atezolizumab in metastatic disease. Spend no product, equity or company trial capital; cap independent analysis and route measurement at £500,000 and require the manufacturer to fund replication and expansion.", "LINFU action map. NCT04793633 and NCT06276764 each target 500 people and can measure pancreatic-fluid sampling and later diagnoses, but both are observational, invitation-only and have no posted results. Do not duplicate them. Require all-participant yield, negative follow-up, complications and central pathology before considering a randomized action study." ], "sourceIds": [ "gudmundsdottir-2023-staging-laparoscopy", "van-dongen-2023-preopanc-staging", "rompen-2026-preopanc2-staging", "umin-000061741-staging-laparoscopy", "conroy-2022-prodige24-five-year", "wainberg-2023-napoli3", "oreilly-2026-daraxonrasib", "nct07491445-daraxonrasib-first-line", "nct07252232-daraxonrasib-resected", "nct07805954-dual-ras-first-line", "nct07409272-setidegrasib-first-line", "nct07232875-hrs4642-phase3", "nct05968326-imcode003", "dynamic-pancreas-registry", "lee-2024-dynamic-pancreas-feasibility", "clinicaltrials-nct05788744-circpac", "clinicaltrials-nct05802394-map03", "guo-2026-postsurgical-ctdna", "chen-2026-personalized-ctdna", "jrct-2026-enzeavour-feasibility", "umin-2025-kobe-ai-early-pdac", "chari-2026-prospective-new-onset-diabetes", "wu-2022-nod-imaging-pilot", "frank-2025-pandome", "edi-nct04662879", "weinberg-2020-ea2185-protocol", "clinicaltrials-nct04239573-ea2185", "clinicaltrials-nct04207944-3pc", "clinicaltrials-nct06519097-sipp-t3", "moyer-2024-charm-followup", "clinicaltrials-nct06192719-eulat-gbc", "clinicaltrials-nct06531408-pigbc", "clinicaltrials-nct03762837-gbc-risk-cohort", "koshiol-2026-chile-bils", "yang-2025-gbcseeker", "gupta-2026-gaia-mil", "baudin-2026-oclurandom", "capdevila-2025-seqtor-pnet", "partelli-2022-aspen-interim", "clinicaltrials-nct03084770-aspen", "heidsma-2021-pandora-pnet", "maekawa-2026-small-pnet-surgery-morbidity", "iarc-2012-liver-flukes", "charoensuk-2024-one-health-opisthorchis", "clinicaltrials-nct05321992-echec", "villard-2023-psc-surveillance", "tan-2024-psc-mrcp-surveillance", "khuntikeo-2020-thailand-cca-screening", "clinicaltrials-nct02170090-acticca1", "yu-2025-btc-ctdna-recurrence", "yoo-2025-extrahepatic-cca-ctdna", "park-2026-btc-ctdna-feasibility", "clinicaltrials-nct05743959-btc-mrd", "bretthauer-2022-nordicc-10y", "kaminski-2026-nordicc-13y", "tie-2022-dynamic-colon", "tie-2025-dynamic3", "bando-2026-altair", "clinicaltrials-nct05174169-circulate-us", "mo-2026-find-crc-surveillance", "andre-2025-keynote177-five-year", "ruers-2017-clocc-crc-liver-ablation", "adam-2024-transmet", "kawaguchi-2025-surf-hcc", "oh-2022-topaz1", "abou-alfa-2020-claridhy", "clinicaltrials-nct05924789-gaspar", "peixoto-2025-gaspar-results", "lim-2025-attention-hbv", "trinchet-2011-hcc-surveillance-interval", "hirode-2026-hcc-biomarker-rct", "rhee-2025-miracle-hcc", "singal-2024-tracer-protocol", "clinicaltrials-nct06084234-tracer", "ioannou-2026-premium-protocol", "clinicaltrials-nct05486572-premium", "gismalla-2026-africa-pancreatic-review", "sobnach-2024-ssa-hcc-systematic-review", "sobnach-2025-south-africa-hcc-disparities", "sobnach-2026-global-hcc-disparities", "da-fonseca-2024-latam-cca", "sandhya-2023-olanzapine-anorexia", "chen-2025-tocilizumab-cachexia", "groarke-2024-ponsegromab-cachexia", "clinicaltrials-nct06989437-ponsegromab-pdac", "woo-2016-pert-rct", "saito-2018-pert-rct", "clinicaltrials-nct04098237-pancax3", "clinicaltrials-nct05642962-pertzye-dose", "clinicaltrials-nct06099119-pertseverance", "lemanska-2023-opensafely-pert-audit", "dhsc-2024-pert-shortage-alert", "maraveyas-2012-fragem", "pelzer-2015-conko004", "vadhan-raj-2020-cassini-pdac", "frere-2020-pdac-thromboprophylaxis-meta", "ash-2023-primary-vte-prophylaxis-guide", "clinicaltrials-nct05178628-impact-pro", "clinicaltrials-nct06861088-catiq-p3", "clinicaltrials-nct05245877-preposterous-pancreas", "clinicaltrials-nct07196020-pinpoint-vte", "zwicker-2019-isoquercetin-catiq", "van-der-gaag-2010-pbd-rct", "costamagna-2026-sems-early-surgery-rct", "tamura-2021-fcsems-plastic-rct", "seo-2019-covered-uncovered-sems-rct", "lyu-2023-metal-plastic-nat-meta", "vehvilainen-2024-cholangitis-nat-interruption", "chen-2023-element-biliary-rct", "paik-2018-eus-ercp-biliary-rct", "barbosa-2024-eus-ercp-meta", "esge-2018-biliary-stenting-guideline", "asge-2024-therapeutic-eus-guideline", "nhs-england-2026-r367", "asco-2024-germline-panel-guideline", "llach-2024-universal-germline", "chittenden-2021-systematic-germline", "walker-2021-embedded-germline-station", "wang-2022-oncology-germline-cascade", "ramsey-2023-mainstreaming-germline", "rodriguez-2024-generate", "mcdonald-2024-germline-protocol", "sinan-2025-pdac-germline-implementation", "grewal-2026-pancan-germline-survey", "katz-2026-gift-cascade", "chen-2023-fdr-cancer-risk", "npaca-2026-state-of-nation", "nhs-england-2024-hpb-pancreatic-service", "nice-2018-ng85-pancreatic", "de-wilde-2012-centralization", "lemmens-2011-centralization", "latenstein-2021-centralization", "coupland-2016-england-resection-volume", "labori-2016-adjuvant-completion", "henry-2023-complications-adjuvant", "kollbeck-2025-complications-chemo", "van-hilst-2019-leopard2", "korrel-2023-diploma", "bruna-2025-diploma-long-term", "klotz-2024-europa", "liu-2024-robotic-open-pd", "jin-2026-portal", "andel-2025-routine-imaging", "van-goor-2025-nationwide-imaging", "daamen-2024-radar-panc-protocol", "clinicaltrials-nct04875325-radar-panc", "tai-2025-pan-asian-esmo-pancreatic", "labori-2024-norpact1", "bai-2025-cispd1", "tan-2026-neoadjuvant-ipd-meta", "ghaneh-2023-espac5", "clinicaltrials-nct04340141-a021806", "effenberger-2012-pdac-bone-marrow-dtc", "nordgard-2022-unresectable-dtc", "chalabi-hajkarim-2025-organotropism", "pei-2025-spatial-metastatic-pdac", "mcclellan-2026-quiescent-pdac-car-t", "tanaka-2012-microscopic-local-recurrence", "dotan-2024-giant", "dotan-2025-giant-ga", "clinicaltrials-nct04233866-giant", "betge-2022-grantpax", "clinicaltrials-nct02143219-pamela70", "mohile-2021-gap70", "li-2021-gain", "ihorst-2026-geriatric-comanagement", "clinicaltrials-nct06040801-frail-gi", "stoop-2026-preopanc4-surgery", "dekker-2026-taps-restaging", "verkolf-2026-taps-resection-prediction", "seelen-2026-pelican", "kunzmann-2021-neolap", "hartlapp-2022-neolap-ca199", "guggenberger-2023-neolap-ct", "napoli-2025-surgery-versus-oncology", "imamura-2004-locally-invasive-surgery-rct", "clinicaltrials-nct06714604-duration", "clinicaltrials-nct06132087-proactive", "baum-2022-fap2286-firsthuman", "mcconathy-2026-lumiere-phase1", "clinicaltrials-nct04939610-lumiere", "baum-2026-3bp3940", "clinicaltrials-nct05432193-frontier", "clinicaltrials-nct07229768-ctr-fapi", "alistar-2017-devimistat-phase1", "philip-2024-avenger500", "surana-2025-erk-hcq-pdac", "clinicaltrials-nct05221320-ulixertinib-hcq", "manji-2023-mekiauto", "clinicaltrials-nct04214418-mekiauto", "clinicaltrials-nct04892017-inlexisertib", "deciphera-2022-dcc3116-phase1", "clinicaltrials-nct04524702-paricalcitol-hcq", "dimcevski-2016-sonoporation-phase1", "wang-2023-sonochemotherapy-pdac", "clinicaltrials-nct04821284-sonoporation", "ius-2026-sonoporation-nct04821284-abstract", "adler-2026-sonoporation-perfusion", "alagesan-2026-sig12d-loder", "ctis-2026-sil204-phase23", "silexion-2026-sil204-site-initiation", "silexion-2026-halfyear-filing", "bazan-peregrino-2021-vcn01-intratumour", "garcia-carbonero-2022-vcn01-intravenous", "clinicaltrials-nct05673811-virage", "theriva-2023-virage-protocol", "theriva-2025-virage-topline", "esmo-2025-virage-2216mo", "ctis-2026-virage2", "theriva-2026-q2-10q", "richards-2006-ci994-pdac", "heumann-2022-azacitidine-pdac", "sohal-2020-thu-decitabine-pdac", "baretti-2024-entinostat-nivolumab-pdac", "clinicaltrials-nct03250273-entinostat-nivolumab", "safyan-2026-azacitidine-pembrolizumab-pdac", "clinicaltrials-nct03264404-azacitidine-pembrolizumab", "rodon-2024-amg193-prmt5", "matsumoto-2026-mtap-pancreatic", "pavlick-2026-mtap-genomics", "clinicaltrials-nct05732831-vopimetostat", "tango-2025-vopimetostat-monotherapy", "clinicaltrials-nct06922591-vopimetostat-ras", "tango-2026-vopimetostat-ras-data", "tango-2026-q2-10q", "park-2026-gleam-zolbetuximab", "clinicaltrials-nct03816163-gleam", "yu-2025-ibi343-pdac", "clinicaltrials-nct05458219-ibi343", "clinicaltrials-nct07066098-ibi343-phase3", "hao-2024-ibi389-pdac", "xu-2025-ibi389-phase1", "clinicaltrials-nct05164458-ibi389", "lentz-2025-ebc129-pdac", "clinicaltrials-nct03023722-anetumab", "gou-2026-shr3821", "clinicaltrials-nct03323944-meso-cart", "aznar-2025-mesothelin-car-t-resistance", "qi-2024-ct041-pdac", "clinicaltrials-nct03874897-ct041", "clinicaltrials-nct04581473-ct041", "clinicaltrials-nct05911217-satricel-adjuvant", "carsgen-2025-satricel-pdac-adjuvant", "xu-2026-kras-g12v-tcr", "clinicaltrials-nct04146298-kras-g12v-tcr", "leidner-2022-kras-tcr", "carsgen-2026-interim-results", "clinicaltrials-nct04935359-danis2", "pietrobono-2026-angiogenin-alk5", "clinicaltrials-nct03336216-cabiralizumab", "van-laethem-2025-optimize1-biomarkers", "clinicaltrials-nct04888312-optimize1", "alligator-2026-q2", "kindler-2010-bevacizumab-calgb80303", "rougier-2013-aflibercept-vanilla", "van-cutsem-2016-maestro", "ramaswamy-2025-afpac-losartan", "clinicaltrials-nct03563248-losartan-localized", "clinicaltrials-nct04193904-mrx0518-pdac", "clinicaltrials-nct05462496-antibiotic-pembro-pdac", "clinicaltrials-nct06393400-chase-pdac", "sharma-2026-fmtpanc", "ctis-panoramix-2024-517766-41", "middleton-2014-telovac", "le-2019-eclipse", "clinicaltrials-nct05726864-amplify7p", "clinicaltrials-nct07671339-eli002-tislelizumab", "clinicaltrials-nct05013216-kras-prevention", "wainberg-2026-quemliclustat-arc8", "coveler-2024-oleclumab-pdac", "clinicaltrials-nct06608927-prism1", "arcus-2026-q2-10q", "singhal-2021-sbp101-phase1", "clinicaltrials-nct03412799-sbp101", "carla-2024-ivospemin-retinal-atrophy", "clinicaltrials-nct05254171-aspire", "ctis-2024-514714-12-00-aspire", "panbela-2024-q3-10q", "clinicaltrials-nct03257033-tigerpac", "hatoum-2024-tamp-rr1-rr2", "novelli-2026-tigerpac-pk", "renovorx-2023-tigerpac-interim", "renovorx-2025-10k", "renovorx-2026-q2-10q", "renovorx-2026-full-enrollment", "timmer-2024-crossfire", "clinicaltrials-nct03899636-direct-rct", "clinicaltrials-nct03899649-direct-registry", "martin-2024-direct-registry-safety", "martin-2026-direct-registry-survival-preprint", "rai-2022-lappie-protocol", "isrctn-14986389-lappie", "nice-2017-ire-pancreatic", "macarulla-2025-panova3", "clinicaltrials-nct03377491-panova3", "fda-2026-optune-pax-ssed", "clinicaltrials-nct06390059-panova4", "novocure-2026-panova4-topline", "novocure-2026-q2-10q", "guzauskas-2026-ttfields-cost", "liu-2026-ttfields-cost", "issels-2023-heat-rct", "clinicaltrials-nct01077427-heat", "eudract-2008-004802-14-heat", "datta-2017-heatpac-protocol", "clinicaltrials-nct02439593-heatpac", "shinoto-2016-carbon-ion-gemcitabine", "kawashiro-2018-jcros-carbon-ion", "lai-2026-carbon-ion-risk-adapted", "rapp-2022-proton-phase2", "eckstein-2023-proton-registry", "clinicaltrials-nct02598349-proton", "clinicaltrials-nct03536182-cipher", "clinicaltrials-nct04194268-pack", "parikh-2023-smart-phase2-safety", "chuong-2024-smart-phase2-survival", "clinicaltrials-nct03621644-smart", "clinicaltrials-nct05585554-lap-ablate", "clinicaltrials-nct05114213-maspac", "viewray-2023-q1-10q", "viewray-2023-chapter11-8k", "clinicaltrials-nct03690323-pelican", "li-2015-hifu-s1", "marinova-2018-hifu-prospective", "marinova-2024-hifu-pc-protocol", "drks-00012367-hifu-pc", "clinicaltrials-nct05262452-hifu-folfirinox", "clinicaltrials-nct05601323-suizenji", "clinicaltrials-nct06211933-puls-hifu", "clinicaltrials-nct07325214-sonopan3", "izzo-2021-pancreatic-ect-phase12", "izzo-2021-laparoscopic-ect-protocol", "eudract-2018-003925-27-ect", "cebron-2023-panect-preliminary", "clinicaltrials-nct04281290-panect-results", "rudno-rudzinska-2021-irec-protocol", "igea-2026-cliniporator-vitae", "clinicaltrials-nct04793633-linfu", "clinicaltrials-nct06276764-linfu-ipmn" ], "links": [ { "target": "trial-pdac-organoid-action-map", "relation": "contains" }, { "target": "trial-pdac-occult-metastasis-action-map", "relation": "contains" }, { "target": "trial-daraxonrasib-expansion-programme", "relation": "contains" }, { "target": "trial-enzeavour-japan-screening", "relation": "contains" }, { "target": "trial-japan-ai-small-pdac-ct", "relation": "contains" }, { "target": "trial-gaspar-lacog-0222", "relation": "contains" }, { "target": "trial-pdac-ctdna-action-map", "relation": "contains" }, { "target": "trial-new-onset-diabetes-workup-map", "relation": "contains" }, { "target": "trial-pancreatic-precursor-action-map", "relation": "contains" }, { "target": "trial-gallbladder-prevention-detection-action-map", "relation": "contains" }, { "target": "trial-small-pnet-action-map", "relation": "contains" }, { "target": "trial-cca-action-map", "relation": "contains" }, { "target": "trial-hcc-action-map", "relation": "contains" }, { "target": "trial-crc-action-map", "relation": "contains" }, { "target": "trial-regional-route-completion-action-map", "relation": "contains" }, { "target": "trial-pdac-cachexia-action-map", "relation": "contains" }, { "target": "trial-pert-pdac-action-map", "relation": "contains" }, { "target": "treatment-modality-ranking", "relation": "supports" }, { "target": "detection-modality-ranking", "relation": "supports" }, { "target": "failures", "relation": "learns-from" }, { "target": "timeline", "relation": "part-of" }, { "target": "programme-100m-randomized-platform", "relation": "informs" }, { "target": "trial-pdac-vte-action-map", "relation": "contains" }, { "target": "trial-pdac-biliary-drainage-action-map", "relation": "contains" }, { "target": "trial-pdac-germline-cascade-action-map", "relation": "contains" }, { "target": "unknown-pdac-germline-family-action-route", "relation": "constrained-by" }, { "target": "hypothesis-pdac-germline-family-route-completion", "relation": "informs" }, { "target": "claim-pdac-surgery-is-a-complete-treatment-route", "relation": "links-to" }, { "target": "system-england-pancreatic-surgery-network", "relation": "links-to" }, { "target": "failure-pdac-postoperative-treatment-loss", "relation": "links-to" }, { "target": "failure-leopard2-laparoscopic-whipple-safety", "relation": "links-to" }, { "target": "unknown-pdac-surgical-route-completion", "relation": "links-to" }, { "target": "trial-pdac-surgical-route-action-map", "relation": "links-to" }, { "target": "hypothesis-pdac-surgical-route-completion", "relation": "links-to" }, { "target": "trial-pdac-postresection-surveillance-action-map", "relation": "contains" }, { "target": "unknown-postresection-surveillance-utility", "relation": "contains" }, { "target": "trial-pdac-perioperative-sequencing-action-map", "relation": "contains" }, { "target": "unknown-pdac-perioperative-treatment-route", "relation": "contains" }, { "target": "trial-pdac-residual-cell-state-action-map", "relation": "contains" }, { "target": "unknown-dormant-cell-actionability", "relation": "contains" }, { "target": "trial-older-pdac-action-map", "relation": "contains" }, { "target": "unknown-older-pdac-treatment-route", "relation": "contains" }, { "target": "trial-lapc-conversion-surgery-action-map", "relation": "contains" }, { "target": "unknown-lapc-surgery-causal-benefit", "relation": "contains" }, { "target": "hypothesis-randomized-lapc-resection-after-response", "relation": "constrained-by" }, { "target": "trial-pdac-fap-radioligand-action-map", "relation": "contains" }, { "target": "unknown-pdac-fap-radioligand-treatment-effect", "relation": "constrained-by" }, { "target": "hypothesis-fapi-theranostic-pdac", "relation": "informs" }, { "target": "trial-pdac-autophagy-action-map", "relation": "contains" }, { "target": "unknown-selective-autophagy-pdac-contribution", "relation": "constrained-by" }, { "target": "hypothesis-tumour-verified-selective-autophagy-blockade", "relation": "informs" }, { "target": "trial-pdac-sonoporation-action-map", "relation": "contains" }, { "target": "unknown-pdac-sonoporation-treatment-effect", "relation": "constrained-by" }, { "target": "hypothesis-measured-fixed-backbone-sonoporation", "relation": "informs" }, { "target": "trial-sil204-action-map", "relation": "contains" }, { "target": "unknown-sil204-human-target-delivery", "relation": "constrained-by" }, { "target": "hypothesis-multicompartment-kras-sirna-delivery", "relation": "informs" }, { "target": "claim-vcn01-randomized-signal-not-survival-proof", "relation": "bounded-by" }, { "target": "failure-virage-analysis-and-survivor-selection", "relation": "learns-from" }, { "target": "unknown-vcn01-all-enrolled-survival", "relation": "leaves-open" }, { "target": "trial-vcn01-action-map", "relation": "implements" }, { "target": "hypothesis-vcn01-replication-to-survival-gate", "relation": "tests" }, { "target": "company-theriva-biologics", "relation": "contains" }, { "target": "failure-broad-epigenetic-treatment-pdac", "relation": "learns-from" }, { "target": "claim-mtap-prmt5-selected-human-signal", "relation": "bounded-by" }, { "target": "failure-vopimetostat-ras-contribution-unknown", "relation": "learns-from" }, { "target": "unknown-mtap-prmt5-ras-contribution", "relation": "leaves-open" }, { "target": "trial-mtap-prmt5-action-map", "relation": "implements" }, { "target": "hypothesis-mtap-prmt5-ras-deeper-control", "relation": "tests" }, { "target": "company-tango-therapeutics", "relation": "contains" }, { "target": "claim-cldn182-adc-selected-response", "relation": "contains" }, { "target": "failure-antibody-target-and-payload-transfer-pdac", "relation": "contains" }, { "target": "unknown-cldn182-adc-survival-effect", "relation": "contains" }, { "target": "trial-cldn182-antibody-action-map", "relation": "contains" }, { "target": "hypothesis-cldn182-adc-survival-test", "relation": "tests" }, { "target": "company-innovent-biologics", "relation": "contains" }, { "target": "claim-satricel-pdac-early-signal", "relation": "contains" }, { "target": "claim-kras-g12v-tcr-early-signal", "relation": "contains" }, { "target": "failure-cell-therapy-route-and-denominator-pdac", "relation": "contains" }, { "target": "unknown-engineered-cell-reproducible-benefit-pdac", "relation": "contains" }, { "target": "trial-pdac-engineered-cell-action-map", "relation": "contains" }, { "target": "hypothesis-kras-tcr-human-persistence-gate", "relation": "tests" }, { "target": "company-carsgen-therapeutics", "relation": "contains" }, { "target": "claim-nis793-target-engagement-without-benefit", "relation": "contains" }, { "target": "failure-immune-reprogramming-without-patient-benefit", "relation": "contains" }, { "target": "unknown-cd40-added-survival-pdac", "relation": "contains" }, { "target": "trial-immune-reprogramming-action-map", "relation": "contains" }, { "target": "hypothesis-angiogenin-tgfb-existing-data-test", "relation": "tests" }, { "target": "hypothesis-mitazalimab-controlled-survival-gate", "relation": "tests" }, { "target": "company-alligator-bioscience", "relation": "contains" }, { "target": "failure-vascular-starvation-pdac", "relation": "contains" }, { "target": "claim-hypoxia-control-without-survival-pdac", "relation": "contains" }, { "target": "failure-losartan-delivery-signal-not-replicated", "relation": "contains" }, { "target": "unknown-localized-losartan-contribution-pdac", "relation": "contains" }, { "target": "trial-pdac-vascular-hypoxia-action-map", "relation": "contains" }, { "target": "hypothesis-localized-losartan-existing-trial-gate", "relation": "tests" }, { "target": "trial-pdac-microbiome-action-map", "relation": "contains" }, { "target": "claim-pdac-microbiome-mechanism-without-human-benefit", "relation": "contains" }, { "target": "failure-microbiome-proxy-without-patient-benefit", "relation": "learns-from" }, { "target": "unknown-microbiome-treatment-contribution-pdac", "relation": "resolves" }, { "target": "hypothesis-bacterial-gemcitabine-deactivation-gate", "relation": "tests" }, { "target": "trial-pdac-vaccine-action-map", "relation": "contains" }, { "target": "claim-pdac-vaccine-immunity-without-proven-benefit", "relation": "contains" }, { "target": "failure-pdac-vaccine-escalation-reversals", "relation": "learns-from" }, { "target": "unknown-pdac-vaccine-regimen-contribution", "relation": "resolves" }, { "target": "hypothesis-autogene-cevumeran-existing-trial-gate", "relation": "tests" }, { "target": "trial-pdac-adenosine-action-map", "relation": "contains" }, { "target": "claim-cd73-blockade-signal-without-concurrent-control", "relation": "contains" }, { "target": "failure-adenosine-control-and-marker-pdac", "relation": "learns-from" }, { "target": "unknown-quemliclustat-survival-contribution", "relation": "resolves" }, { "target": "hypothesis-quemliclustat-existing-phase3-gate", "relation": "tests" }, { "target": "trial-pdac-polyamine-action-map", "relation": "contains" }, { "target": "claim-ivospemin-signal-without-controlled-benefit", "relation": "contains" }, { "target": "failure-aspire-trial-operational-collapse", "relation": "learns-from" }, { "target": "unknown-aspire-assigned-denominator-and-results", "relation": "resolves" }, { "target": "hypothesis-aspire-data-rescue-gate", "relation": "tests" }, { "target": "trial-pdac-intraarterial-delivery-action-map", "relation": "contains" }, { "target": "claim-tamp-blood-exposure-not-tumour-delivery", "relation": "contains" }, { "target": "failure-tigerpac-component-and-interim-inference", "relation": "learns-from" }, { "target": "unknown-tigerpac-package-survival-and-delivery", "relation": "resolves" }, { "target": "hypothesis-tigerpac-existing-final-gate", "relation": "tests" }, { "target": "trial-pdac-irreversible-electroporation-action-map", "relation": "contains" }, { "target": "claim-ire-ablation-without-proven-added-survival", "relation": "contains" }, { "target": "failure-direct-selection-and-randomized-reporting-gap", "relation": "learns-from" }, { "target": "unknown-ire-added-survival-after-modern-chemotherapy", "relation": "resolves" }, { "target": "hypothesis-direct-randomized-data-recovery-gate", "relation": "tests" }, { "target": "trial-pdac-ttfields-action-map", "relation": "contains" }, { "target": "claim-panova3-ttfields-small-os-benefit", "relation": "contains" }, { "target": "failure-panova4-component-and-historical-control", "relation": "learns-from" }, { "target": "unknown-ttfields-survival-replication-and-access", "relation": "resolves" }, { "target": "hypothesis-optune-pax-independent-value-gate", "relation": "tests" }, { "target": "regulatory-fda-optune-pax-2026", "relation": "bounded-by" }, { "target": "trial-pdac-regional-hyperthermia-action-map", "relation": "contains" }, { "target": "claim-heat-regional-hyperthermia-primary-endpoint-negative", "relation": "contains" }, { "target": "failure-heat-component-and-registry-drift", "relation": "learns-from" }, { "target": "unknown-regional-hyperthermia-added-benefit-modern-care", "relation": "resolves" }, { "target": "hypothesis-heat-existing-randomization-recovery-gate", "relation": "tests" }, { "target": "trial-pdac-particle-therapy-action-map", "relation": "contains" }, { "target": "claim-particle-therapy-local-control-without-survival-proof", "relation": "contains" }, { "target": "failure-particle-therapy-comparator-and-accrual", "relation": "learns-from" }, { "target": "unknown-particle-therapy-added-patient-benefit", "relation": "resolves" }, { "target": "hypothesis-particle-therapy-existing-data-gate", "relation": "tests" }, { "target": "trial-pdac-smart-action-map", "relation": "contains" }, { "target": "claim-smart-delivery-and-safety-without-added-survival-proof", "relation": "contains" }, { "target": "failure-smart-trial-and-vendor-dependency", "relation": "learns-from" }, { "target": "unknown-smart-added-survival-and-surgery-harm", "relation": "resolves" }, { "target": "hypothesis-smart-vendor-independent-randomized-gate", "relation": "tests" }, { "target": "trial-pdac-rfa-action-map", "relation": "contains" }, { "target": "claim-pelican-rfa-no-survival-and-more-harm", "relation": "contains" }, { "target": "failure-rfa-selection-and-route-attrition", "relation": "learns-from" }, { "target": "unknown-rfa-human-mechanism-after-negative-trial", "relation": "resolves" }, { "target": "hypothesis-pelican-specimen-salvage-gate", "relation": "tests" }, { "target": "trial-pdac-hifu-action-map", "relation": "contains" }, { "target": "claim-hifu-pain-signal-without-survival-proof", "relation": "contains" }, { "target": "failure-hifu-randomization-and-trial-drift", "relation": "learns-from" }, { "target": "unknown-hifu-pain-survival-and-system-transfer", "relation": "resolves" }, { "target": "hypothesis-hifu-existing-randomized-gate", "relation": "tests" }, { "target": "trial-pdac-electrochemotherapy-action-map", "relation": "contains" }, { "target": "claim-electrochemotherapy-local-effect-without-patient-benefit", "relation": "contains" }, { "target": "failure-electrochemotherapy-endpoint-and-trial-drift", "relation": "learns-from" }, { "target": "unknown-electrochemotherapy-added-benefit-and-delivery", "relation": "resolves" }, { "target": "hypothesis-electrochemotherapy-randomized-data-gate", "relation": "tests" }, { "target": "trial-linfu-action-map", "relation": "contains" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "moderate", "randomised": "strong", "independentReplication": "moderate", "sampleSize": "Primary trial reports and registry records selected from more than 120 preserved source-level trial identifiers", "effect": "Decision-oriented trial portfolio map; individual effects remain in linked records", "limits": [ "Not every registered study has a dedicated node", "Recruitment status can change", "Sponsor registry records are plans", "Cross-trial comparisons are not randomized", "Diabetes imaging studies are too small for utility inference", "Related-cancer results do not establish PDAC effects", "Germline studies mostly measure delivery rather than cancer outcomes" ] }, "layout": { "x": 86, "y": 61 }, "updated": "2026-09-15" }, { "id": "unknown-actionable-interception-window", "title": "Unknown: how long is PDAC detectable while still biologically curable?", "shortTitle": "Interception window", "type": "unknown", "status": "open", "scope": [ "pdac" ], "summary": "Slow evolutionary estimates, early dissemination models and punctuated genomic change describe different routes; no single lead-time number is defensible.", "content": "Class: conflicting evidence plus missing longitudinal human data. Decision affected: whether money should favour annual high-risk surveillance, one-time risk-triggered work-up, repeated liquid biopsy or systemic interception. A radiographic lead time is not automatically an actionable biological window: a small primary can already have seeded lethal clones, while some precursor lesions never progress. Required experiment: link serial blood and routine imaging from high-risk, new-onset-diabetes and incidental-cyst cohorts to complete pathology, recurrence and survival; model aggressive PDAC separately from indolent lesions; validate on a second health system. The useful endpoint is the fraction found before systemic competence and treatable with current or trial therapy, not months of earlier diagnosis. Stop a marker route if lead time is created mainly by indolent lesions or if aggressive tumours remain below detection until dissemination.", "sourceIds": [ "yachida-2010-metastatic-timing", "rhim-2012-early-dissemination", "notta-2016-punctuated-evolution", "blackford-2024-high-risk-surveillance", "chari-2026-prospective-new-onset-diabetes", "effenberger-2012-pdac-bone-marrow-dtc", "chalabi-hajkarim-2025-organotropism" ], "links": [ { "target": "unknowns", "relation": "part-of" }, { "target": "claim-pdac-progression-timing-uncertain", "relation": "extends" }, { "target": "hypothesis-one-time-risk-triggered-workup", "relation": "tested-by" }, { "target": "detection", "relation": "controls" }, { "target": "claim-human-pdac-dtc-prognostic-not-dormancy", "relation": "constrained-by" }, { "target": "trial-pdac-residual-cell-state-action-map", "relation": "constrained-by" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "moderate", "randomised": "none", "independentReplication": "weak — conflicting evidence", "sampleSize": "Multiple small evolutionary and surveillance cohorts; no direct intervention-window cohort", "effect": "No settled estimate", "limits": [ "Different methods estimate different events", "Lead-time and length bias", "High-risk cohorts differ from ordinary-risk disease", "Pathology is available only after intervention" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Precursor evolution and occult dissemination make an interception interval plausible, but its end differs by route." }, "animalEvidence": { "rating": "moderate", "reason": "Mouse lineage tracing shows early dissemination but cannot date lethal spread in people." }, "inVitroEvidence": { "rating": "weak", "reason": "Laboratory timing does not reproduce years of human evolution." }, "reproducibility": { "rating": "weak", "reason": "Human clock, mouse lineage and punctuated-genome evidence remain method-dependent." }, "commercialConflicts": { "rating": "low", "reason": "Assay and later target access require declared ownership, open methods and independent analysis." }, "recency": { "rating": "mixed", "reason": "The evidence set includes 2025 human spatial studies and 2026 treatment-specific models." } } }, "layout": { "x": 33, "y": 99 }, "updated": "2026-09-14", "contentSections": [ "Recent human cellular-state work still does not provide a clock. In 744 resections, 333 recurred and selected primary tumours carried liver- or lung-linked programmes, but only 21 entered single-nucleus analysis and recurrence association cannot date dissemination. Bone-marrow tumour-cell signals show occult spread at diagnosis in some patients but do not prove when the lethal clone left or whether it remained inactive. The useful window remains action-linked, not a single number." ] }, { "id": "unknown-asp1929-pancreatic-target-coverage-and-benefit", "title": "Will ASP-1929 bind and illuminate enough pancreatic cancer to matter?", "shortTitle": "ASP-1929 pancreatic coverage unknown", "type": "unknown", "status": "unresolved", "scope": [ "pdac", "photoimmunotherapy", "egfr", "dosimetry" ], "summary": "The open chain is EGFR density to antibody uptake to light coverage to local kill, immune response, systemic control and patient-valued time.", "content": "Known quantities. Many pancreatic cancers express EGFR. ASP-1929 combines an EGFR-binding antibody with IR700 dye. Japan has conditional clinical experience in head and neck cancer. Unknown quantities. What fraction of each pancreatic tumour has enough surface EGFR for binding? Does prior chemotherapy change that fraction? Can the large antibody reach cancer cells through dense scar-like tissue? What fraction of bound dye receives enough 690 nm light when the pancreas sits behind stomach, bowel, vessels and other tissue? Do resectable participants who receive drug but no light have any drug harm? Among inoperable participants, what tissue volume is illuminated and what critical structures receive spill? Does rapid local membrane damage release tumour signals that produce a useful immune response beyond the illuminated region, or only inflammation? Does the required break from systemic treatment increase distant failure? Does six-month progression translate into less pain, fewer obstructions, more quality-adjusted time or longer survival? Resolve each layer in all enrolled people. Measure EGFR on whole biopsy sections, fluorescent drug distribution where possible, fibre coordinates, light output, dead-zone coverage, circulating immune changes, local and distant failure, planned versus delivered chemotherapy, hospital days, symptoms and survival. A head-and-neck approval and a pancreatic EGFR stain are boundary evidence, not substitutes.", "sourceIds": [ "clinicaltrials-nct07698613-asp1929-pancreas", "pmda-2020-akalux-review", "rakuten-medical-2026-asp1929", "kim-2015-egfr-expression-pdac", "maruoka-2018-interstitial-photoimmunotherapy", "philip-2010-swog-s0205-cetuximab" ], "links": [ { "target": "unknowns", "relation": "part-of" }, { "target": "claim-asp1929-pancreatic-human-test-not-started", "relation": "bounded-by" }, { "target": "failure-asp1929-selection-light-and-endpoint-design", "relation": "exposed-by" }, { "target": "hypothesis-asp1929-first-ten-measurement-gate", "relation": "tested-by" }, { "target": "trial-pdac-asp1929-action-map", "relation": "mapped-by" }, { "target": "company-rakuten-medical", "relation": "constrained-by" }, { "target": "unknown-pdt-spatial-coverage-and-patient-benefit", "relation": "compares-with" } ], "evidence": { "humanProspective": "none in pancreatic cancer", "humanRetrospective": "none in pancreatic cancer", "randomised": "none", "independentReplication": "none", "sampleSize": "Pancreatic treatment n=0; planned n=30", "effect": "No human pancreatic estimate for binding, illumination, safety, immune response or benefit", "limits": [ "Cross-cancer regulatory data", "Variable EGFR", "No pancreatic pharmacology", "No human light map", "No control", "Systemic-treatment interruption", "Surgery-dependent treatment assignment" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "Each component works in another setting; their overlap in human pancreatic anatomy is untested." }, "animalEvidence": { "rating": "moderate", "reason": "Animal models support antibody-light killing and expose the need for internal illumination." }, "inVitroEvidence": { "rating": "moderate", "reason": "Model systems show targeted killing, but not human tumour penetration or spatial coverage." }, "reproducibility": { "rating": "none", "reason": "There is no pancreatic human result to repeat." }, "commercialConflicts": { "rating": "high", "reason": "The central drug, dye and light platform are controlled by one company." }, "recency": { "rating": "current", "reason": "The unknown is anchored to the first planned pancreatic study in 2026." } } }, "layout": { "x": 1018, "y": 942 }, "updated": "2026-09-15" }, { "id": "unknown-aspire-assigned-denominator-and-results", "title": "How many ASPIRE patients were assigned, followed and analysed?", "shortTitle": "ASPIRE data state", "type": "unknown", "status": "open", "scope": [ "pdac", "trial-delivery", "transfer-unproven" ], "summary": "Planned counts differ, the reported safety set reached 395, and no source gives the actual assignment, follow-up, blind, survival or harm record.", "content": "Decision affected: whether ASPIRE can still settle ivospemin benefit and harm, and whether public money should preserve its data. Required fields are screened, eligible, assigned by arm, treated by arm, last known alive date, death source, progression, response, dose exposure, chemotherapy exposure, eye examinations, liver injury, serious harm, withdrawals, site closure and reason for missing follow-up. The U.S. registry estimates 600 while Europe lists 547 planned. Panbela reported a 395-person safety database, but that is not an actual randomization count and does not reveal arm balance. First resolution gate: identify the legal data controller and assignment-key custodian; hash and copy the raw database, image archive, safety reports and site-payment ledger; obtain the locked protocol and every analysis-plan version. Second gate: reconcile every randomization number with a consented patient and site, and obtain vital status at 12 months for at least 90% of assigned patients with no more than ten percentage points difference in missingness between arms. Third gate: publish all-randomized overall survival, progression, response, treatment exposure, eye harm, liver harm and serious harm by arm. A continuation recommendation from a monitoring board is a safety process result, not evidence that treatment works. If the blind, consent, assignment key or outcome chain cannot be recovered, label the efficacy question unresolved and publish the failure rather than infer benefit from the pooled event rate.", "sourceIds": [ "clinicaltrials-nct05254171-aspire", "ctis-2024-514714-12-00-aspire", "panbela-2024-q3-10q", "singhal-2021-sbp101-phase1", "carla-2024-ivospemin-retinal-atrophy" ], "links": [ { "target": "unknowns", "relation": "part-of" }, { "target": "claim-ivospemin-signal-without-controlled-benefit", "relation": "limits" }, { "target": "failure-aspire-trial-operational-collapse", "relation": "caused-by" }, { "target": "hypothesis-aspire-data-rescue-gate", "relation": "tested-by" }, { "target": "trial-pdac-polyamine-action-map", "relation": "tracked-by" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong", "independentReplication": "moderate", "sampleSize": "U.S. planned n=600; European planned n=547; sponsor safety database n=395; actual assigned and followed denominators unknown", "effect": "Cannot estimate efficacy, attrition or arm-specific severe harm from public records", "limits": [ "No participant flow", "No posted results", "No protocol or analysis plan in the linked U.S. record", "Participant-level sharing is marked no", "Trial custody may have changed" ], "dimensions": { "biologicalPlausibility": { "rating": "not-applicable", "reason": "The uncertainty concerns patient and data accounting." }, "animalEvidence": { "rating": "not-applicable", "reason": "Models do not resolve missing trial state." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Laboratory activity does not resolve missing trial state." }, "reproducibility": { "rating": "weak", "reason": "The two public registries disagree and no all-randomized result is available." }, "commercialConflicts": { "rating": "high", "reason": "The sponsor controlled the missing counts and first analysis." }, "recency": { "rating": "current", "reason": "Both projected readout and completion dates have passed." } } }, "layout": { "x": 496, "y": 420 }, "updated": "2026-09-15" }, { "id": "unknown-cachexia-treatment-utility", "title": "Does treating PDAC cachexia preserve function and chemotherapy enough to extend useful life?", "shortTitle": "Cachexia utility gap", "type": "unknown", "status": "open", "scope": [ "pdac" ], "summary": "Weight and muscle can change without proof of survival; the unresolved decision is whether a low-toxicity intervention changes treatment delivery and patient-valued time.", "content": "Decision: should a pancreatic-cancer programme fund a separate cachexia intervention, and for whom? Muscle loss predicts death, but prediction is not proof that reversing it changes death. Tocilizumab preserved muscle without improving its main survival endpoint and added severe toxicity. Ponsegromab increased weight and activity over twelve weeks in a mixed-cancer, GDF-15-selected trial, but did not test survival as the main result. Low-dose olanzapine cheaply improved weight and appetite in a mixed-cancer Indian trial, but only 13 patients had hepatopancreaticobiliary cancer and survival was not tested. NCT06989437 now supplies the large PDAC-specific test: estimated n=982, randomized ponsegromab versus placebo with first-line chemotherapy, with survival, function, tumour outcomes and chemotherapy dose changes measured. Its two primary endpoints remain week-12 weight and anorexia score. Resolve the unknown by checking a locked hierarchy: first no unacceptable toxicity; then better function and patient symptoms; then at least ten percentage points more patients receiving at least 80% of planned chemotherapy dose intensity at twelve weeks or an equally patient-important treatment-completion gain; finally survival or quality-adjusted survival. Weight alone cannot pass. Stop or classify as supportive care only if weight improves but dose delivery, function and patient-valued time do not, or if benefit is absent in the PDAC and high-GDF-15 groups. Do not duplicate the 982-person trial before its data and access terms are audited.", "sourceIds": [ "babic-2019-muscle-loss-survival", "chen-2025-tocilizumab-cachexia", "groarke-2024-ponsegromab-cachexia", "clinicaltrials-nct06989437-ponsegromab-pdac", "sandhya-2023-olanzapine-anorexia" ], "links": [ { "target": "unknowns", "relation": "part-of" }, { "target": "trial-pdac-cachexia-action-map", "relation": "could-resolve" }, { "target": "claim-ponsegromab-gdf15-cachexia-driver-not-survival", "relation": "constrained-by" }, { "target": "failure-tocilizumab-cachexia-muscle-without-survival", "relation": "constrained-by" }, { "target": "unknown-pert-pdac-utility", "relation": "separates-from" }, { "target": "trial-pert-pdac-action-map", "relation": "complements" }, { "target": "programme-500m-integrated-mission", "relation": "constrains" } ], "evidence": { "humanProspective": "strong for intermediate outcomes; absent for the full utility chain", "humanRetrospective": "strong for prognosis", "randomised": "strong for weight and muscle; unresolved for durable utility", "independentReplication": "weak across different interventions; none for PDAC ponsegromab", "sampleSize": "Evidence includes a 164-person PDAC cohort, 147-person PDAC tocilizumab trial, 187-person mixed-cancer ponsegromab trial, 124-person olanzapine trial and an estimated 982-person live PDAC trial", "effect": "Host measures can improve; the effect on chemotherapy delivery, quality-adjusted survival and OS is unknown", "limits": [ "Different cachexia definitions", "Mixed-cancer trials", "Short weight endpoints", "Post-randomization frailty exclusions", "No completed PDAC GDF-15 utility result", "Generated ten-point delivery gate" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Wasting, reduced function and treatment tolerance form a plausible route, but each link needs human proof." }, "animalEvidence": { "rating": "moderate", "reason": "Models support inflammatory and GDF-15 routes but cannot quantify patient benefit." }, "inVitroEvidence": { "rating": "weak", "reason": "Cell systems poorly represent appetite, movement, frailty and treatment completion." }, "reproducibility": { "rating": "none", "reason": "No intervention has yet reproduced the complete PDAC chain from host recovery to longer useful life." }, "commercialConflicts": { "rating": "high", "reason": "The largest live test is sponsored by the company that owns the drug and relevant patent families." }, "recency": { "rating": "current", "reason": "The live trial record was updated in August 2026." } } }, "layout": { "x": 139, "y": 65 }, "updated": "2026-09-14" }, { "id": "unknown-cca-postoperative-mrd-action", "title": "Unknown: which postoperative cholangiocarcinoma action benefits a ctDNA-positive patient?", "shortTitle": "CCA MRD action", "type": "unknown", "status": "open", "scope": [ "cholangiocarcinoma", "hepatobiliary" ], "summary": "ctDNA can identify a high-risk state months early, but neither intensified imaging nor systemic escalation has proved that it changes survival.", "content": "Class: prognostic evidence without clinical utility. Decision affected: whether a patient with no visible disease after cholangiocarcinoma resection should receive a different adjuvant regimen, longer treatment, a molecularly selected drug, earlier imaging or observation when ctDNA is positive. The current evidence does not identify one action. STAMP randomized two adjuvant regimens but not by ctDNA result and found no significant disease-free or overall-survival difference between those regimens. The real-world 56-person cohort supplied 3.7 months of average lead time, but lead time is useful only if an intervention works in that molecular-only state. Resolving experiment: after site-specific assay validation and a 21-day result-return run-in, randomize ctDNA-positive patients between current standard care and one named escalation with overall survival, recurrence-free survival, serious toxicity and quality-adjusted life as endpoints. A separate randomization must test imaging frequency; it cannot be combined with treatment escalation. Stop if panel construction succeeds in fewer than 90%, fewer than 85% receive the result before action, the selected regimen lacks an independently supported effect in the same anatomical and molecular group, or serious treatment harm exceeds the lower credible absolute recurrence benefit.", "sourceIds": [ "yoo-2025-extrahepatic-cca-ctdna", "yu-2025-btc-ctdna-recurrence", "park-2026-btc-ctdna-feasibility", "clinicaltrials-nct05743959-btc-mrd", "clinicaltrials-nct02170090-acticca1" ], "links": [ { "target": "unknowns", "relation": "part-of" }, { "target": "claim-cca-ctdna-prognostic-not-action-guiding", "relation": "depends-on" }, { "target": "trial-cca-action-map", "relation": "exposes" }, { "target": "hypothesis-cca-route-specific-action-platform", "relation": "could-change" }, { "target": "unknown-ctdna-actionability", "relation": "parallel-to" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "moderate", "randomised": "none for ctDNA-directed action", "independentReplication": "none for utility", "sampleSize": "Prognostic cohorts of 56, 89 and 18 enrolled; no action cohort", "effect": "No estimate of benefit or harm from acting on ctDNA", "limits": [ "The best action may differ by site and driver", "Rare disease makes site-specific randomization difficult", "Adjuvant standards are still changing", "Assay failure and turnaround affect eligibility", "Earlier recurrence knowledge can increase anxiety without extending life" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Molecular disease should precede some visible recurrences, but treatment sensitivity at that state is unknown." }, "animalEvidence": { "rating": "weak", "reason": "Models cannot choose the human adjuvant action." }, "inVitroEvidence": { "rating": "moderate", "reason": "Tumour-informed detection is feasible, while treatment response cannot be inferred from detection alone." }, "reproducibility": { "rating": "none", "reason": "No randomized group has reproduced a benefit from a ctDNA-directed biliary action." }, "commercialConflicts": { "rating": "high", "reason": "Both testing volume and treatment escalation have commercial incentives." }, "recency": { "rating": "current", "reason": "The uncertainty incorporates evidence and registry status through September 2026." } } }, "updated": "2026-09-14" }, { "id": "unknown-cca-surveillance-utility", "title": "Unknown: which cholangiocarcinoma risk groups gain life from surveillance?", "shortTitle": "CCA surveillance utility", "type": "unknown", "status": "open", "scope": [ "cholangiocarcinoma", "hepatobiliary" ], "summary": "PSC and liver-fluke regions produce different risks and detectable lesions; existing studies show stage or association signals without randomized mortality proof.", "content": "Class: conflicting route-specific evidence plus missing randomized utility data. Decision affected: whether money should fund annual MRI/MRCP and CA19-9 in PSC, ultrasound in liver-fluke regions, or a different risk and test sequence. A prospective Swedish PSC programme did not produce reliable early curable detection, while retrospective Australian surveillance was associated with lower all-cause mortality but not better survival after cancer diagnosis. Thailand's ultrasound programme shifted stage among cancers found but has not yet established population mortality benefit. These results cannot be pooled because disease cause, anatomy, base rate and available surgery differ. Resolving experiment: maintain two separate protocols. In PSC, use a multi-country stepped-wedge comparison of a locked imaging and escalation route, recording transplant, false procedures, interval cancer and cause-specific mortality. In endemic fluke areas, randomize districts to risk reduction plus ultrasound navigation versus risk reduction alone, with population denominators. Stop a route if fewer than 60% complete confirmatory work-up, false invasive procedures exceed ten per additional resectable cancer, or no credible mortality or advanced-stage signal remains after adequate follow-up.", "sourceIds": [ "villard-2023-psc-surveillance", "tan-2024-psc-mrcp-surveillance", "khuntikeo-2020-thailand-cca-screening", "iarc-2012-liver-flukes", "charoensuk-2024-one-health-opisthorchis", "clinicaltrials-nct05321992-echec" ], "links": [ { "target": "unknowns", "relation": "part-of" }, { "target": "claim-cca-surveillance-depends-on-risk-population", "relation": "depends-on" }, { "target": "model-cholangiocarcinoma-causal-chain", "relation": "could-change" }, { "target": "detection", "relation": "controls" }, { "target": "trial-cca-action-map", "relation": "tested-by" }, { "target": "hypothesis-cca-route-specific-action-platform", "relation": "tested-by" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "moderate", "randomised": "none for modern route-specific mortality", "independentReplication": "weak — conflicting evidence", "sampleSize": "PSC prospective and retrospective cohorts plus population ultrasound screening in endemic Thailand", "effect": "No settled net-benefit or mortality estimate", "limits": [ "Different causal populations", "Lead-time and referral bias", "Rare outcome in PSC", "Changing transplant and surgery access", "No common action threshold" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The risk states are real but create different lesions, base rates and available interventions." }, "animalEvidence": { "rating": "moderate", "reason": "Liver-fluke models support causation but do not select a human surveillance route." }, "inVitroEvidence": { "rating": "weak", "reason": "Candidate markers have not completed the risk-to-action path." }, "reproducibility": { "rating": "weak", "reason": "Prospective PSC, retrospective PSC and endemic ultrasound results do not supply one repeatable utility result." }, "commercialConflicts": { "rating": "moderate", "reason": "Imaging, endoscopy and assays can expand use before patient benefit is known." }, "recency": { "rating": "current", "reason": "The uncertainty uses evidence checked through September 2026." } } }, "layout": { "x": 35, "y": 104 }, "updated": "2026-09-14" }, { "id": "unknown-cd40-added-survival-pdac", "title": "Does adding mitazalimab to mFOLFIRINOX extend useful life in metastatic PDAC?", "shortTitle": "CD40 survival effect unknown", "type": "unknown", "status": "open", "scope": [ "pdac", "pan-cancer", "transfer-unproven" ], "summary": "A 57-person uncontrolled signal and a failed test of another CD40 drug leave one answerable question: the added survival effect of mitazalimab on a fixed chemotherapy backbone.", "content": "Known. Mitazalimab plus mFOLFIRINOX produced a 42.1% confirmed response rate, 12.6-month median response duration and 14.9-month median survival among 57 evaluable participants. Known weakness. The trial had no mFOLFIRINOX-only group, its registry now counts 94 participants without final posted results, and the developer stopped its own phase 3 work. Another CD40 agonist, sotigalimab, did not pass its historical survival test in PRINCE; that study also lacked chemotherapy-only control and used gemcitabine/nab-paclitaxel. Unknown. The added mitazalimab effect, the people who gain it, and its full toxicity remain unidentified. Resolving evidence. First publish all 94 OPTIMIZE-1 participants and lock a candidate fibrosis and immune-state score without refitting it to later outcomes. Then randomize about 450 untreated metastatic-PDAC patients 1:1 to mitazalimab or placebo with identical mFOLFIRINOX, stratified by performance, liver disease volume, region and the fixed score. The primary endpoint is overall survival; patient function, hospital-free days, PFS, confirmed response, response duration and severe harm are required. A useful claim requires OS HR at most 0.75 with its 95% CI below 1, no treatment-related mortality excess above two points and benefit not confined to an invented subgroup. Stop at planned interim futility, a serious-event excess above ten points, loss of drug supply or failure to publish all randomized participants. Until then, clinical use and equity investment are £0.", "sourceIds": [ "van-laethem-2024-optimize1", "van-laethem-2025-optimize1-biomarkers", "clinicaltrials-nct04888312-optimize1", "padron-2022-prince", "alligator-2026-q2" ], "links": [ { "target": "unknowns", "relation": "part-of" }, { "target": "claim-cd40-signal-without-contribution-proof", "relation": "supported-by" }, { "target": "failure-immune-reprogramming-without-patient-benefit", "relation": "bounded-by" }, { "target": "hypothesis-mitazalimab-controlled-survival-gate", "relation": "tested-by" }, { "target": "trial-immune-reprogramming-action-map", "relation": "resolved-by" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "not-applicable", "randomised": "weak", "independentReplication": "weak", "sampleSize": "Mitazalimab efficacy n=57 of n=70 reported and n=94 registered; proposed controlled trial about n=450", "effect": "Question about added OS, function and harm; no assigned effect", "limits": [ "Proposed effect threshold is a mission decision", "Required event count depends on follow-up", "Final product supply and rights are uncertain", "A different CD40 drug is only indirect contrary evidence" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Human responses and immune activation make the question credible." }, "animalEvidence": { "rating": "strong", "reason": "PDAC models support CD40 activation with chemotherapy." }, "inVitroEvidence": { "rating": "strong", "reason": "Receptor activation and immune response can be tested." }, "reproducibility": { "rating": "weak", "reason": "No controlled mitazalimab result exists and another CD40 drug did not pass its gate." }, "commercialConflicts": { "rating": "high", "reason": "The insolvent-looking owner needs a transaction and controls the product records." }, "recency": { "rating": "current", "reason": "The company stopped independent development in July 2026." } } }, "layout": { "x": 286, "y": 222 }, "updated": "2026-09-15" }, { "id": "unknown-cldn182-adc-survival-effect", "title": "Does IBI343 extend useful life in marker-selected PDAC?", "shortTitle": "IBI343 survival gap", "type": "unknown", "status": "open", "scope": [ "pdac" ], "summary": "An uncontrolled marker-response gradient is credible enough to test, while the live 201-person phase 3 has missing design details and can only settle a large effect.", "content": "Known. At 6 mg/kg, IBI343 produced ten confirmed responses among 44 high-marker pancreatic cancers and none among 12 below the analysis boundary. The company is already recruiting G-HOPE-002: 201 people after at least two systemic regimens, allocated 2:1 to IBI343 plus best supportive care or placebo plus best supportive care, with participants, clinicians, investigators and outcome readers masked. Overall survival is primary. Unknown. The registry does not state the CLDN18.2 assay or cutoff, lists only one Chinese site, has no participant-data-sharing plan, and says eligible performance status is 0 or 2 rather than 0 through 2. It does not publish the expected death count, effect size, false-positive rate, power, analysis populations, missing-data rule, marker-stratification rule or quality-of-life endpoint. Power boundary. Under ordinary two-sided 5% error and 80% power assumptions with 2:1 allocation, detecting HR 0.70 needs about 278 deaths: (1.96+0.84)^2 / [(2/3)(1/3)(ln 0.70)^2] = 278, more deaths than 201 people can supply. HR 0.65 needs about 191 deaths and is just feasible if nearly everyone has an event. This is planning arithmetic, not the sponsor's design. Resolving evidence. Obtain the protocol and dated analysis plan before unmasking. Require the same central assay and fixed cutoff used before allocation, all-randomized survival accounting of at least 99%, country and site expansion beyond one centre, response and progression by independent review, patient-reported function, dose intensity, payload toxicity and later treatment. A positive claim requires OS HR at or below 0.75 with its 95% interval excluding 1.0, positive quality-adjusted survival and no severe-harm excess above ten points. Stop or call the effect unresolved if the interval includes 1.0, missing outcomes can move HR above 0.85, the marker rule changes after outcomes, or the survival gain is purchased by intolerable time in hospital.", "sourceIds": [ "yu-2025-ibi343-pdac", "clinicaltrials-nct05458219-ibi343", "clinicaltrials-nct07066098-ibi343-phase3", "park-2026-gleam-zolbetuximab", "abou-alfa-2006-exatecan-pdac" ], "links": [ { "target": "unknowns", "relation": "part-of" }, { "target": "claim-cldn182-adc-selected-response", "relation": "bounded-by" }, { "target": "failure-antibody-target-and-payload-transfer-pdac", "relation": "constrained-by" }, { "target": "trial-cldn182-antibody-action-map", "relation": "resolved-by" }, { "target": "hypothesis-cldn182-adc-survival-test", "relation": "tests" }, { "target": "company-innovent-biologics", "relation": "controlled-by" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "not-applicable", "randomised": "none", "independentReplication": "none", "sampleSize": "Phase 1 selected analysis n=56 at 6 mg/kg; active phase 3 planned n=201 with 2:1 allocation", "effect": "Required result is all-randomized OS HR at most 0.75 plus positive quality-adjusted survival", "limits": [ "Protocol and analysis plan unavailable", "Marker rule absent from registry", "One listed site", "No data-sharing statement", "Power calculation uses assumptions", "Late-line population", "Registry performance-status wording may be an error" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The phase 1 response gradient matches the carrier's surface-address mechanism." }, "animalEvidence": { "rating": "moderate", "reason": "Model support exists but cannot settle the survival question." }, "inVitroEvidence": { "rating": "strong", "reason": "Target binding, uptake and payload damage can be measured and should be linked to patient records." }, "reproducibility": { "rating": "none", "reason": "No randomized or independent PDAC benefit has reported." }, "commercialConflicts": { "rating": "high", "reason": "Innovent controls the active Greater China trial and Takeda controls rights elsewhere." }, "recency": { "rating": "current", "reason": "Uses the live phase 3 registry and 2025 phase 1 result." } } }, "layout": { "x": 222, "y": 140 }, "updated": "2026-09-15" }, { "id": "unknown-crc-ctdna-positive-action", "title": "Unknown: which action helps a ctDNA-positive colorectal-cancer patient?", "shortTitle": "CRC positive-MRD action", "type": "unknown", "status": "open", "scope": [ "colorectal", "transfer-unproven" ], "summary": "Blood-positive residual disease strongly predicts recurrence, but more standard chemotherapy and trifluridine/tipiracil have not improved the tested endpoints.", "content": "Class: strong risk marker, missing effective treatment action. Decision affected: whether a blood-positive, scan-negative patient should receive more chemotherapy, a different systemic treatment, earlier imaging, local treatment after a lesion appears, or observation inside a trial. DYNAMIC-III found no recurrence-free-survival gain from chemotherapy escalation. ALTAIR found no significant disease-free-survival gain from trifluridine/tipiracil and added 69.7 percentage points of grade 3 or worse hematologic toxicity. FIND doubled the proportion of recurrent patients receiving curative-intent treatment by triggering CT earlier, but survival is not mature. CIRCULATE-US is recruiting 1,912 planned participants and tests both ctDNA-negative treatment omission and ctDNA-positive mFOLFIRINOX; estimated primary completion is March 2029 and no results are posted. Resolving experiment: first show activity in a blood-positive window with serial clearance and resistance sampling, then randomize the fixed action against standard care. Overall survival, quality-adjusted survival, severe toxicity and treatment-free time must decide release. Stop a stage for less than a 15-point improvement in durable six-month ctDNA clearance, more than a 10-point increase in grade 3 or worse toxicity without a clearance gain, or early survival harm. Do not duplicate CIRCULATE-US chemotherapy questions while that trial remains viable.", "sourceIds": [ "tie-2025-dynamic3", "bando-2026-altair", "mo-2026-find-crc-surveillance", "clinicaltrials-nct05174169-circulate-us" ], "links": [ { "target": "unknowns", "relation": "part-of" }, { "target": "claim-crc-ctdna-utility-is-action-specific", "relation": "depends-on" }, { "target": "trial-crc-action-map", "relation": "tested-by" }, { "target": "experiments", "relation": "resolved-by" }, { "target": "unknown-ctdna-actionability", "relation": "may-transfer-to" } ], "evidence": { "humanProspective": "strong for tested actions", "humanRetrospective": "strong for prognosis", "randomised": "strong for failed named actions; pending for CIRCULATE-US", "independentReplication": "moderate", "sampleSize": "DYNAMIC-III n=968; ALTAIR n=243; FIND n=584; CIRCULATE-US plans n=1,912", "effect": "No validated positive-MRD systemic action; earlier imaging increased curative-intent treatment but survival is unknown", "limits": [ "Assays and populations differ", "A failed drug does not refute the marker", "Clearance is not a final patient endpoint", "CIRCULATE-US has no result", "FIND follow-up is short", "Stop thresholds are programme choices" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "A positive postoperative result indicates residual tumour, but its drug sensitivity and geography remain unknown." }, "animalEvidence": { "rating": "weak", "reason": "Models may rank treatments but cannot establish the human action benefit." }, "inVitroEvidence": { "rating": "moderate", "reason": "Patient-derived models could screen treatments after a positive result, but this has not validated a care rule." }, "reproducibility": { "rating": "moderate", "reason": "The risk signal repeats and several action trials now exist, with mixed utility results." }, "commercialConflicts": { "rating": "high", "reason": "Assay-company involvement is common and CIRCULATE-US names a commercial test." }, "recency": { "rating": "current", "reason": "The ledger includes 2026 results and a September 2026 registry status." } } }, "updated": "2026-09-14" }, { "id": "unknown-crc-liver-local-cure-selector", "title": "Unknown: which colorectal liver metastases are still a local curable state?", "shortTitle": "CRC liver cure selector", "type": "unknown", "status": "open", "scope": [ "colorectal", "hepatobiliary", "transfer-unproven" ], "summary": "Ablation and transplant can create long survival in narrow liver-only groups, while broader liver-directed treatment can improve local control without extending life.", "content": "Class: known treatment effect in selected groups, missing biological selector. Decision affected: who should receive resection, ablation, hepatic infusion or transplant instead of systemic treatment alone. CLOCC randomized 119 people with fewer than ten unresectable liver metastases and no extrahepatic disease and found an eight-year survival difference favouring local treatment. TransMet found a large transplant benefit in a much smaller, tightly selected liver-only group. EPOCH, in a broader second-line population, improved hepatic progression by 1.9 months without improving overall survival and increased severe harm. Anatomy and scan negativity are therefore necessary but insufficient. Resolving experiment: within a current systemic-treatment platform, lock a selector using disease-free interval, molecular subtype, ctDNA clearance, response duration and repeated whole-body imaging, then randomize selector-positive patients to comprehensive local control or systemic care. Overall survival, treatment-free time, quality of life and serious procedure harm are co-primary decision measures. Stop if extrahepatic relapse within 12 months exceeds 35%, peri-treatment death exceeds 3%, or local control improves without a credible survival or treatment-free-time signal.", "sourceIds": [ "ruers-2017-clocc-crc-liver-ablation", "adam-2024-transmet", "mulcahy-2021-epoch-radioembolization", "tie-2022-dynamic-colon" ], "links": [ { "target": "unknowns", "relation": "part-of" }, { "target": "colorectal-liver-metastasis", "relation": "depends-on" }, { "target": "hypothesis-selected-local-consolidation", "relation": "resolved-by" }, { "target": "transferability-matrix", "relation": "controls" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "strong", "randomised": "strong for selected interventions; none for a validated selector", "independentReplication": "weak — mixed evidence", "sampleSize": "CLOCC n=119, TransMet small randomized liver-only population, EPOCH broader randomized second-line population", "effect": "No validated probability threshold separating curable local-dominant from occult systemic disease", "limits": [ "Small selected trials", "Different local methods", "Changing systemic comparators", "Transplant scarcity", "Selector must be tested inside randomization" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Disease-free interval, molecular state, treatment response and repeated staging can jointly estimate occult systemic risk." }, "animalEvidence": { "rating": "moderate", "reason": "Models can test routes of spread but cannot validate the clinical selector." }, "inVitroEvidence": { "rating": "moderate", "reason": "Drug response models may inform systemic control but cannot establish organ confinement." }, "reproducibility": { "rating": "weak", "reason": "Local benefit repeats only under differing narrow selection rules; no locked selector has randomized validation." }, "commercialConflicts": { "rating": "moderate", "reason": "Imaging, assay, device and transplant programmes each create different incentives." }, "recency": { "rating": "recent", "reason": "The open question incorporates the 2024 TransMet result." } } }, "layout": { "x": 38, "y": 107 }, "updated": "2026-09-14" }, { "id": "unknown-cryoablation-randomized-pain-and-net-benefit", "title": "Does tumour cryoablation relieve pain better than existing palliation without costing useful life?", "shortTitle": "Cryo palliative benefit unknown", "type": "unknown", "status": "unresolved", "scope": [ "pdac", "unresectable", "pain", "quality-of-life" ], "summary": "The decision question is palliative and comparative: pain interference, opioid burden, days at home, treatment delivery and serious harm versus the best available pain route.", "content": "Known quantity. Selected people report a large fall in pain after the tumour is frozen, and the ice ball can be watched on computed tomography. Unknown quantities. How much of that fall comes from cryoablation rather than sedation, concurrent cancer treatment, natural pain variation, expectation or selective follow-up? Does freezing tumour outperform coeliac plexus neurolysis, which chemically damages the nearby nerve network, or high-quality medical pain care? Does the benefit last? Does inserting needles through abdomen or stomach cause pancreatitis, bleeding, infection, fistula or treatment delay? Does conservative partial freezing relieve pain without controlling the tumour, and is that trade acceptable? The answer needs a randomized palliative comparison, not a tumour-response trial. Measure daily 0-to-10 pain, how pain disrupts sleep and activity, morphine-equivalent opioid dose, constipation and sedation, validated quality of life, days alive at home, chemotherapy dose and timing, all procedures, local and distant progression and survival. Death before the main four-week assessment must count as failure, not disappear. Stratify metastatic versus locally advanced disease and planned needle route. Keep tumour freezing separate from freezing the coeliac or splanchnic nerves and from open laparotomy.", "sourceIds": [ "pusceddu-2026-percutaneous-cryoablation", "kang-2025-laparotomic-cryoablation", "song-2014-cryo-bypass-comparison" ], "links": [ { "target": "unknowns", "relation": "part-of" }, { "target": "claim-cryoablation-pain-signal-not-survival-proof", "relation": "bounded-by" }, { "target": "failure-cryoablation-selection-attrition-and-route-harm", "relation": "exposed-by" }, { "target": "hypothesis-cryoablation-pragmatic-pain-trial", "relation": "tested-by" }, { "target": "trial-pdac-cryoablation-action-map", "relation": "mapped-by" }, { "target": "unknown-hifu-pain-survival-and-system-transfer", "relation": "compares-with" } ], "evidence": { "humanProspective": "weak", "humanRetrospective": "moderate", "randomised": "none", "independentReplication": "weak", "sampleSize": "Pain signal n=11; comparative survival records n=48 and n=118", "effect": "Causal pain relief, opioid reduction, quality-adjusted time and survival remain unknown", "limits": [ "Different cryo routes", "Different disease stages", "No common pain comparator", "No blinded assessment", "Concurrent treatment varies", "Survivor conditioning", "Rare severe harm needs larger follow-up" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Local tissue and nerve-pressure reduction can plausibly reduce pain." }, "animalEvidence": { "rating": "moderate", "reason": "Freeze geometry and tissue injury are known, but pain is a human outcome." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "A cell assay cannot answer palliative value." }, "reproducibility": { "rating": "weak", "reason": "The uncontrolled clinical signal has not been tested against a fixed comparator." }, "commercialConflicts": { "rating": "moderate", "reason": "A device trial needs supplier-neutral specifications and independent data custody." }, "recency": { "rating": "current", "reason": "The uncertainty is framed around the May 2026 pain series." } } }, "layout": { "x": 856, "y": 780 }, "updated": "2026-09-15" }, { "id": "unknown-ctdna-actionability", "title": "Postoperative ctDNA predicts PDAC recurrence risk; whether acting on it improves survival is unknown", "shortTitle": "ctDNA utility gap", "type": "unknown", "status": "open", "scope": [ "pdac" ], "summary": "Tumour-informed ctDNA repeatedly marks higher recurrence risk, but its sensitivity and predictive values vary and no completed randomised trial shows that changing treatment or surveillance from the result helps patients.", "content": "A multicentre retrospective commercial series analysed 1,329 samples from 298 patients. CtDNA was detected in 29% during the 2-to-12-week postoperative residual-disease window and 29.6% in later surveillance. Positive tests were associated with much shorter disease-free survival: 6.37 versus 33.31 months in the postoperative window and 11.40 months versus not reached during surveillance. In multivariable analysis, surveillance positivity had hazard ratio 24.28 for recurrence. Two 2026 cohorts confirm the direction but expose unstable operating performance. In 51 resected patients, postoperative ctDNA sensitivity was 35.7% and specificity 88.9%; later testing reached 62.5% sensitivity and 95.5% specificity. In a Chinese cohort, personalized panels were built for 35 of 43 patients; after chemotherapy the positive predictive value was 75.0%, negative predictive value 82.4% and median molecular lead time 4.59 months. A negative result therefore cannot replace imaging, and a positive result is not automatically a treatment indication. Three programmes test different actions: DYNAMIC-Pancreas changes adjuvant duration or regimen, CIRCPAC triggers more frequent imaging, and MAP-03 switches therapy before radiographic progression. None has reported a completed randomized utility result. The missing proof remains a fixed action against a concealed-result or standard-care control, measuring overall survival, quality of life, false-positive harm and treatment burden. The existence of these trials changes the funding decision: align assays and outcomes or fill a clear gap instead of duplicating them.", "sourceIds": [ "botta-2024-tumor-informed-ctdna", "dynamic-pancreas-registry", "lee-2024-dynamic-pancreas-feasibility", "guo-2026-postsurgical-ctdna", "chen-2026-personalized-ctdna", "clinicaltrials-nct05788744-circpac", "clinicaltrials-nct05802394-map03", "daamen-2024-radar-panc-protocol", "clinicaltrials-nct04875325-radar-panc" ], "links": [ { "target": "residual-disease", "relation": "supports" }, { "target": "experiments", "relation": "resolves" }, { "target": "unknowns", "relation": "part-of" }, { "target": "trial-pdac-ctdna-action-map", "relation": "tested-by" }, { "target": "pdac", "relation": "part-of" }, { "target": "unknown-postresection-surveillance-utility", "relation": "bounded-by" }, { "target": "trial-pdac-postresection-surveillance-action-map", "relation": "bounded-by" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "moderate", "randomised": "none", "independentReplication": "moderate", "sampleSize": "Largest retrospective series n=298 and 1,329 samples; additional 2026 cohorts n=51 and n=43; active action trials plan 438, 1,000 and 100 participants in differing settings", "effect": "Largest series: surveillance positivity recurrence HR 24.28 after adjustment. Smaller cohorts: sensitivity 35.7% to 62.5%, specificity 88.9% to 95.5%, and post-chemotherapy PPV 75.0%. Clinical-utility effect unknown", "limits": [ "Largest evidence is retrospective and commercial", "Testing schedules and assays differ", "Small 2026 cohorts", "Some authors were employed by test manufacturers", "Low sensitivity leaves blood-negative recurrence", "Risk prediction does not show benefit from a treatment decision", "No mature randomised utility result" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Residual tumour can shed DNA before imaging recurrence, and positive results consistently mark higher recurrence risk." }, "animalEvidence": { "rating": "not-applicable", "reason": "The utility claim concerns human surveillance and treatment decisions, not an animal mechanism." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Analytical detection in a laboratory cannot show that acting on a result improves survival or quality of life." }, "reproducibility": { "rating": "moderate", "reason": "The prognostic direction repeats across cohorts, but sensitivity, specificity and predictive values vary and no action effect has been repeated." }, "commercialConflicts": { "rating": "high", "reason": "The largest cohort used a commercial assay and several studies include test-company employees or commercial assay interests." }, "recency": { "rating": "current", "reason": "The record includes two 2026 cohorts and a randomized surveillance trial status verified in 2026." } } }, "layout": { "x": 26, "y": 103 }, "updated": "2026-09-14", "contentSections": [ "RADAR-PANC supplies the missing randomized standard-imaging comparator without ctDNA. Its result should set the control schedule and expected benefit before any biomarker is credited for earlier detection." ] }, { "id": "unknown-dangerous-precursor-selection", "title": "Unknown: which pancreatic precursor lesions will become lethal?", "shortTitle": "Dangerous precursor selection", "type": "unknown", "status": "open", "scope": [ "pdac" ], "summary": "Visible IPMN and largely invisible PanIN require different routes, and IPMN itself splits into current advanced disease, future cyst progression and cancer elsewhere in the pancreas.", "content": "Class: known and clinically urgent, with several strong cohorts but no validated action rule. The old question—'which precursor is dangerous?'—contains four different decisions. PanIN is usually invisible and common driver mutations appear in low-grade lesions, so a population action needs a host or shed-signal route that does not require locating every PanIN. Visible IPMN then splits into: whether advanced disease is already present; whether a low-grade cyst will progress later; and whether a separate PDAC will arise elsewhere. Japan's 2,104-person prospective cohort observed 1.90% HGD or invasive disease within the IPMN and 2.11% separate PDAC over 5.17 years, with different predictors. PancreaSeq GC detected current advanced neoplasia at 86.6% sensitivity and 97.9% specificity in an enriched 241-person sampled cohort, but did not test future progression in routine low-risk surveillance. Surgery is not a harmless reference standard: a 1,137-operation German mixed-cyst registry reported 28.4% severe morbidity and 2.6% 30-day mortality. Yet waiting until high-risk stigmata improved pathology yield while 45% of recent post-surveillance resections were already invasive. Required experiment: three separate locked IPMN models for present advanced disease, future within-cyst progression and separate PDAC, followed by randomization of the action. Keep PanIN as a distinct high-risk-cohort discovery lane. The rule must beat current care at the same missed-invasive-cancer rate, count operation harm and pancreatic function, and validate unchanged outside its development systems. Stop a sampled-marker route if fewer than one in four positive operations contains HGD or invasive disease, if interval cancers cross the locked bound, or if performance uses information available only after resection.", "contentSections": [ "Late surveillance evidence narrows but does not close the low-risk rule. PACYFIC followed 975 people without baseline warning features across 44 centres and observed only 20 HGD or cancer events; its comparison between small stable cysts and population incidence had a 95% confidence interval from 0.01 to 6.30, compatible with both much lower and materially higher risk. A separate 1,000-person maintained database found invasive cancer in 1.4% of subcentimetre cysts versus 1.8% of larger cysts, while growth rate did not select high-risk pathology. The Kyoto guideline therefore presents both stopping and continuing surveillance after five unchanged years as options. Size, stability and elapsed time reduce concern but do not yet define a safe universal stop rule." ], "sourceIds": [ "kanda-2012-early-panin-mutations", "wu-2011-gnas-ipmn", "blackford-2024-high-risk-surveillance", "ohtsuka-2024-bd-ipmn-surveillance", "mirzaian-2026-low-risk-cysts", "singhi-2026-pancreaseq-gc", "pea-2026-post-surveillance-resection", "henn-2023-cyst-surgery-registry", "clinicaltrials-nct04239573-ea2185", "levink-2025-pacyfic", "ohtsuka-2024-kyoto-ipmn-guideline", "wehrle-2025-subcentimeter-ipmn" ], "links": [ { "target": "unknowns", "relation": "part-of" }, { "target": "claim-early-panin-driver-mutations-not-progression-markers", "relation": "depends-on" }, { "target": "claim-ipmn-mutations-identify-lineage-not-progression", "relation": "depends-on" }, { "target": "claim-ipmn-has-two-cancer-routes", "relation": "depends-on" }, { "target": "claim-cyst-classifier-current-disease-not-future-risk", "relation": "depends-on" }, { "target": "claim-ipmn-surgery-timing-tradeoff", "relation": "depends-on" }, { "target": "hypothesis-dynamic-cyst-interception-rule", "relation": "tested-by" }, { "target": "detection", "relation": "controls" } ], "evidence": { "humanProspective": "strong for natural history; weak for action utility", "humanRetrospective": "strong for surgery selection and timing", "randomised": "none", "independentReplication": "moderate", "sampleSize": "Prospective IPMN cohort n=2,104; low-risk cyst cohort n=6,064; classifier n=241; surgery registry n=1,137; no validated intervention rule", "effect": "Two IPMN cancer routes, good current-disease classification in an enriched cohort, and no proven future-progression or mortality rule", "limits": [ "Surgical pathology selects higher-risk lesions", "Long follow-up and competing death", "PanIN is usually not directly visible", "Procedure harms constrain sampling", "Separate PDAC is not controlled by treating the indexed cyst" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Human pathology and surveillance show real precursor lineages, but lesion presence is much more common than lethal progression." }, "animalEvidence": { "rating": "moderate", "reason": "Models support stepwise progression and interception, but cannot set human surgery thresholds." }, "inVitroEvidence": { "rating": "moderate", "reason": "Cyst-fluid and tissue markers can identify lineage and advanced disease, while prospective timing remains unproved." }, "reproducibility": { "rating": "moderate", "reason": "Natural-history patterns repeat across cohorts, but no action rule has independent validation." }, "commercialConflicts": { "rating": "moderate", "reason": "Key natural-history studies are academic, while leading molecular classifiers have intellectual-property and royalty interests." }, "recency": { "rating": "current", "reason": "The record includes prospective, molecular and surgical evidence through September 2026." } } }, "layout": { "x": 30, "y": 102 }, "updated": "2026-09-14" }, { "id": "unknown-dormant-cell-actionability", "title": "Unknown: can human PDAC residual cells be identified and eliminated before recurrence?", "shortTitle": "Dormant-cell actionability", "type": "unknown", "status": "open", "scope": [ "pdac", "transfer-unproven" ], "summary": "Models expose several persistence and reawakening routes, but there is no validated human dormant-cell state, sampling method or intervention rule.", "content": "Class: unknown because of missing human data and difficult sampling. Decision affected: whether residual-disease treatment should kill quiescent cells, hold them inactive, block organ-specific reawakening or simply intensify common-RAS treatment. The same late recurrence could arise from a dormant single cell, a slowly growing micrometastasis or repeated treatment-sensitive and resistant state changes. Required experiment: build a consented post-resection cohort with serial tumour-informed blood, immune and host-injury measurements, complete recurrence biopsies and rapid autopsy; add liver tissue only when another operation makes sampling safe. Lock marker definitions before outcome analysis. Reproduce candidate states in two human-derived systems and perturb them without first forcing proliferation. A go decision requires a state found in human residual or recurrence material, association with later relapse independent of stage and ctDNA burden, and a clinically achievable intervention that removes it in two systems. Stop a route if its marker appears only after tissue handling, if awakening is required to make the cells drug-sensitive, or if target expression is equally necessary in normal repair cells.", "sourceIds": [ "lin-2013-reversible-pdac-dormancy", "shi-2026-chemo-dormant-dtc", "botta-2024-tumor-informed-ctdna", "effenberger-2012-pdac-bone-marrow-dtc", "nordgard-2022-unresectable-dtc", "chalabi-hajkarim-2025-organotropism", "pei-2025-spatial-metastatic-pdac", "tanaka-2012-microscopic-local-recurrence" ], "links": [ { "target": "unknowns", "relation": "part-of" }, { "target": "claim-pdac-dormancy-model-proven-human-uncertain", "relation": "follows" }, { "target": "unknown-ctdna-actionability", "relation": "complements" }, { "target": "programme-25m-prospective-observatory", "relation": "resolved-by" }, { "target": "claim-human-pdac-dtc-prognostic-not-dormancy", "relation": "resolved-by" }, { "target": "failure-pdac-dormancy-sample-gap", "relation": "resolved-by" }, { "target": "trial-pdac-residual-cell-state-action-map", "relation": "resolved-by" } ], "evidence": { "humanProspective": "weak", "humanRetrospective": "weak", "randomised": "none", "independentReplication": "none", "sampleSize": "No complete human longitudinal residual-cell cohort", "effect": "Unresolved decision, not an effect estimate", "limits": [ "Residual tissue is rarely accessible", "Tissue handling can create false cell states", "Blood-negative disease may remain in organs", "Long follow-up is required", "Intervention may awaken rather than eliminate cells" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "Human occult-cell signals and recurrence support persistence, while dormancy and useful targeting are unproved." }, "animalEvidence": { "rating": "strong", "reason": "Several models directly trace residual cells and later regrowth." }, "inVitroEvidence": { "rating": "moderate", "reason": "Human-derived systems can test candidate dependence but may create or erase quiescence." }, "reproducibility": { "rating": "none", "reason": "No complete human ancestry-state-action chain has been reproduced." }, "commercialConflicts": { "rating": "moderate", "reason": "Assay and later target access require declared ownership, open methods and independent analysis." }, "recency": { "rating": "current", "reason": "The evidence set includes 2025 human spatial studies and 2026 treatment-specific models." } } }, "layout": { "x": 50, "y": 108 }, "updated": "2026-09-14", "contentSections": [ "Human marrow studies supply a prognostic signal, not a dormant-cell definition: 24/175 were cytokeratin-positive at surgery with OS HR 2.283, and 15/48 unresectable cases were RNA-panel positive before treatment with OS HR 2.0. The later on-treatment result was not prognostic. The action now requires ancestry, low proliferation, lead time and added value beyond stage plus ctDNA in the same timeline. End-stage or primary-only expression signatures do not pass." ] }, { "id": "unknown-durable-ras-control", "title": "Unknown: can broad RAS suppression become durable without unacceptable toxicity?", "shortTitle": "Durable RAS control", "type": "unknown", "status": "open", "scope": [ "pdac", "pan-cancer", "transfer-unproven" ], "summary": "Daraxonrasib establishes broad human targetability, and mutant KRAS amplification now supplies a repeated human escape class; durable, independently reproduced and safely actionable control remains unsolved.", "content": "Class: known and scientifically difficult, but narrowed. Decision affected: whether systemic RAS control remains the first programme priority or becomes only another temporary line of treatment. Supporting evidence: randomized survival doubled against later-line chemotherapy. A paired phase 1/2 plasma study now identifies treatment-emergent RAS-pathway changes in 26/44 = 59% of progressing patients, led by mutant KRAS amplification in 16/44 = 36%. That clears the programme's first gate that one class explain at least 25% of progression. Contrary evidence: median progression remained 7.3 months; 11/44 patients acquired more than one alteration; the resistance study was sponsor-led; serious gastrointestinal and lung toxicities occurred; and no matched combination has randomized human benefit. Required experiment: reproduce KRAS amplification and the wider pathway map in baseline, week-four, progression and post-treatment blood plus multi-lesion tumour from the randomized first-line and post-resection trials. Test whether amplification appears before clinical progression and predicts response to a fixed tolerable combination. Go to a matched trial only if the class repeats at a decision-relevant rate, two laboratories agree, routing completes within 14 days and the combination suppresses it at clinically achieved exposure. Stop a route for failure to reproduce, no tumour target engagement, overlapping grade 3–5 toxicity above its bound, or a bypass that remains active at clinical exposure.", "sourceIds": [ "oreilly-2026-daraxonrasib", "aronchik-2026-daraxonrasib-resistance", "dorbin-2026-allele-specific-daraxonrasib-resistance", "lin-2026-vertical-kras-inhibition", "varghese-2025-kras-dosage", "nct07491445-daraxonrasib-first-line", "nct07252232-daraxonrasib-resected" ], "links": [ { "target": "unknowns", "relation": "part-of" }, { "target": "claim-daraxonrasib-common-ras-survival-breakthrough", "relation": "follows" }, { "target": "hypothesis-prospective-ras-resistance-routing", "relation": "tested-by" }, { "target": "claim-daraxonrasib-kras-amplification-resistance", "relation": "narrowed-by" }, { "target": "programme-100m-randomized-platform", "relation": "controls" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "moderate", "randomised": "strong", "independentReplication": "weak", "sampleSize": "Randomized later-line trial n=500; paired phase 1/2 resistance cohort n=44; first-line and post-resection trials ongoing", "effect": "Median PFS 7.3 versus 3.5 months in RAS-G12 disease; emergent RAS-pathway changes 59% and mutant KRAS amplification 36% in the paired resistance cohort", "limits": [ "Resistance cohort was small and sponsor-led", "Resistance specimens not yet reported from phase 3", "Plasma can miss spatial or low-shedding disease", "Combination tolerability and predictive benefit unknown", "Residual-disease biology may differ from metastatic disease" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Human progression is frequently accompanied by renewed RAS-pathway output, led by increased mutant KRAS dosage." }, "animalEvidence": { "rating": "strong", "reason": "Several human and mouse PDAC models reproduce pathway reactivation and test candidate combinations." }, "inVitroEvidence": { "rating": "strong", "reason": "Resistance systems reproduce amplification, RTK reactivation and allele-specific escape, while leaving clinical prediction unproved." }, "reproducibility": { "rating": "weak", "reason": "The main human frequency comes from one sponsor-led phase 1/2 cohort and requires independent phase 3 confirmation." }, "commercialConflicts": { "rating": "high", "reason": "The drug sponsor generated the main resistance dataset and controls the key compounds and combination programme." }, "recency": { "rating": "current", "reason": "The pivotal efficacy and paired resistance reports were published in 2026." } } }, "layout": { "x": 38, "y": 96 }, "updated": "2026-09-14" }, { "id": "unknown-electrochemotherapy-added-benefit-and-delivery", "title": "Does pancreatic electrochemotherapy add useful life, and was the whole target electrically covered?", "shortTitle": "ECT benefit and coverage unknown", "type": "unknown", "status": "unresolved", "scope": [ "pdac", "locally-advanced", "resectable", "drug-delivery" ], "summary": "The open question joins a clinical effect to a delivery check: identical systemic care, full electric-field coverage, intracellular drug effect, survival, quality of life and surgical harm.", "content": "Known quantity. Short electric pulses can temporarily open cell membranes, and the Cliniporator records voltage and current. Unknown quantities. Did every part of the pancreatic target receive enough field while vessels, ducts and bowel remained safe? Did bleomycin enter tumour cells rather than merely reach blood? Does the laparoscopic operation delay chemotherapy or expose occult spread? Does treating a cut surface after removal lower local recurrence without adding fistula, bleeding or death? Most important, does either route increase overall or quality-adjusted survival over the identical operation and systemic treatment? The current evidence cannot answer those questions. The 25-person cohort selected only chemotherapy non-progressors, changed electrode geometry by surgeon choice and lost 44% of six-month scans through death. PanECT has no comparator. The planned 90-person randomization changes ECT alone on a fixed FOLFOXIRI backbone, but powers a scan-response endpoint rather than survival. Resolve its all-assigned dataset before building anything new. Require electrode coordinates, pulse waveforms, planned and achieved field coverage, bleomycin timing, tumour and normal-tissue measures where stored material permits, operation time and complications, chemotherapy delivery, imaging under a fixed rule, local and distant failure, quality of life and survival. A scan-density change is not enough. Keep locally advanced tumour treatment separate from cut-surface treatment after resection and from irreversible or calcium electroporation.", "sourceIds": [ "izzo-2021-pancreatic-ect-phase12", "izzo-2021-laparoscopic-ect-protocol", "eudract-2018-003925-27-ect", "cebron-2023-panect-preliminary", "clinicaltrials-nct04281290-panect-results", "rudno-rudzinska-2021-irec-protocol", "igea-2026-cliniporator-vitae" ], "links": [ { "target": "unknowns", "relation": "part-of" }, { "target": "claim-electrochemotherapy-local-effect-without-patient-benefit", "relation": "bounded-by" }, { "target": "failure-electrochemotherapy-endpoint-and-trial-drift", "relation": "exposed-by" }, { "target": "hypothesis-electrochemotherapy-randomized-data-gate", "relation": "tested-by" }, { "target": "trial-pdac-electrochemotherapy-action-map", "relation": "mapped-by" }, { "target": "company-igea", "relation": "constrained-by" }, { "target": "unknown-ire-added-survival-after-modern-chemotherapy", "relation": "compares-with" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "not-applicable", "randomised": "weak — protocol only", "independentReplication": "weak", "sampleSize": "Existing uncontrolled n=25 and n=10; randomized target n=90", "effect": "Added survival, recurrence prevention, quality of life and full target delivery remain unknown", "limits": [ "No published randomized result", "No direct intracellular bleomycin measure", "No control for cut-surface treatment", "Electrode geometry varies", "Operation and drug effects are joined", "The randomized trial is response-powered", "Device data access is unconfirmed" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Temporary membrane opening can increase uptake of a poorly permeant drug." }, "animalEvidence": { "rating": "moderate", "reason": "Deep-tumour delivery has model support but does not identify net human benefit." }, "inVitroEvidence": { "rating": "strong", "reason": "Pulse-driven membrane permeability and bleomycin uptake are directly testable." }, "reproducibility": { "rating": "weak", "reason": "Human feasibility repeats but field coverage and patient benefit do not." }, "commercialConflicts": { "rating": "high", "reason": "The generator, software and electrodes come from one involved manufacturer." }, "recency": { "rating": "current", "reason": "The uncertainty includes 2026 registry results and current EU transition state." } } }, "layout": { "x": 748, "y": 672 }, "updated": "2026-09-15" }, { "id": "unknown-elraglusib-survival-replication", "title": "Does elraglusib cause a repeatable survival gain in metastatic PDAC?", "shortTitle": "Elraglusib replication", "type": "unknown", "status": "open", "scope": [ "pdac" ], "summary": "One randomized phase 2 gives a survival signal, but the treatment effect, mechanism, safety tradeoff and repeatability remain unresolved.", "content": "Known. The all-randomized phase 2 estimate favours elraglusib plus gemcitabine/nab-paclitaxel: OS 8.9 versus 7.2 months, HR 0.68, with a 19.8-point one-year survival difference. Unknown. Whether this repeats under blinding; why OS improved without PFS, response or disease-control improvement; whether unequal pre-treatment withdrawals, site effects, country, early deaths or later treatment explain part of the result; whether the relevant state can be selected before treatment; and whether added neutropenia, fatigue and visual disturbance buy useful life. Resolving evidence has two stages. Stage 0 is independent analysis of all 269 weekly-or-control randomized records with at least 99% survival accounting, dated protocol and analysis-plan versions, missing-outcome tipping points, site and region tests, post-progression treatment, quality of life and cause of death. Continue only if all-randomized HR remains at or below 0.75, its 95% interval excludes 1.0 and no plausible missing-data case moves HR above 0.85. Stage 1 is a double-blind placebo-controlled phase 3 of elraglusib plus the same chemotherapy versus placebo plus chemotherapy. Primary analysis includes every randomized person. Stop the route if OS HR exceeds 0.80, its planned interval includes 1.0, grade 3 or worse harm rises by more than 10 percentage points without positive quality-adjusted survival, chemotherapy dose intensity falls by more than 10 points, or benefit is confined to a subgroup found after seeing outcomes. The transient visual effect may reveal assignment, so masking success must be measured rather than assumed.", "sourceIds": [ "mahalingam-2026-elraglusib-randomized", "clinicaltrials-nct03678883-elraglusib", "pathak-2026-elraglusib-folfirinox", "clinicaltrials-nct05077800-elraglusib-folfirinox" ], "links": [ { "target": "unknowns", "relation": "part-of" }, { "target": "claim-elraglusib-randomized-survival-signal", "relation": "limits" }, { "target": "failure-elraglusib-analysis-and-survival-discordance", "relation": "caused-by" }, { "target": "trial-elraglusib-action-map", "relation": "resolved-by" }, { "target": "hypothesis-elraglusib-survival-confirmation-gate", "relation": "tests" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "moderate", "independentReplication": "none", "sampleSize": "Existing confirmation candidate: 269 relevant randomized people; future phase 3 not yet fixed", "effect": "Required output is a repeatable all-randomized survival effect with positive quality-adjusted life", "limits": [ "Thresholds are mission decisions", "Participant data access untested", "Future protocol and partner are not fixed", "Masking may be weakened by visual effects" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "A wide signaling-node effect is plausible but the survival-only pattern is not explained." }, "animalEvidence": { "rating": "moderate", "reason": "Models support testing but cannot reproduce the human survival result." }, "inVitroEvidence": { "rating": "moderate", "reason": "Target effects are measurable but do not settle clinical utility." }, "reproducibility": { "rating": "none", "reason": "The central survival result has not repeated in an independent randomized trial." }, "commercialConflicts": { "rating": "high", "reason": "Data and product access depend on a capital-constrained sponsor." }, "recency": { "rating": "current", "reason": "The question uses 2026 paper, abstract and registry evidence." } } }, "layout": { "x": 202, "y": 120 }, "updated": "2026-09-15" }, { "id": "unknown-engineered-cell-reproducible-benefit-pdac", "title": "Which engineered-cell architecture can produce repeatable useful life in PDAC?", "shortTitle": "Cell-therapy benefit unknown", "type": "unknown", "status": "open", "scope": [ "pdac", "pan-cancer", "transfer-unproven" ], "summary": "CLDN18.2 CAR-T has 4/24 uncontrolled responses and mutation-HLA-matched TCR-T has 2 clear responses across seven reported patients; neither has reproducible survival evidence.", "content": "Known. Surface-reading CAR-T can reach some pancreatic cancers: satri-cel produced four responses among 24 infused patients, with median response duration 9.5 months. Mutation-fragment-reading TCR-T can also work: one of two reported KRAS-G12D patients and one of five KRAS-G12V patients had clear regression. Known failure. Mesothelin CAR-T produced no objective response in nine reported patients despite local delivery and tissue detection. Three repeat KRAS-G12V TCR infusions produced no response, and two were rapidly rejected. Unknown. The records do not identify which combination of address coverage, HLA presentation, cell product state, lesion entry, expansion, persistence, tumour killing and conditioning yields repeatable benefit. They also do not give a complete population denominator from marker or HLA screening through manufacture and outcome. Resolving evidence. First recover all 54 NCT03323944 participants and the full 134-person NCT03874897 record. In the existing 30-person NCT04146298 study, preserve every screened and collected patient; measure HLA display machinery, target fragment, paired tumour entry, cell state, receptor persistence, anti-receptor antibodies and every lesion response. A mechanism passes only if at least 25 patients are infused, at least five have confirmed responses, at least four responses last six months, paired tumour measurement succeeds in at least 80%, and a fixed measurement separates response from failure. It still does not establish useful-life benefit; that requires a later concurrent control. Stop for treatment-related death, severe normal-tissue attack, fewer than 70% of collected patients infused, no durable response or a selector invented after outcomes.", "sourceIds": [ "qi-2024-ct041-pdac", "clinicaltrials-nct03874897-ct041", "xu-2026-kras-g12v-tcr", "clinicaltrials-nct04146298-kras-g12v-tcr", "leidner-2022-kras-tcr", "aznar-2025-mesothelin-car-t-resistance", "clinicaltrials-nct03323944-meso-cart" ], "links": [ { "target": "unknowns", "relation": "part-of" }, { "target": "claim-satricel-pdac-early-signal", "relation": "supported-by" }, { "target": "claim-kras-g12v-tcr-early-signal", "relation": "supported-by" }, { "target": "failure-cell-therapy-route-and-denominator-pdac", "relation": "bounded-by" }, { "target": "trial-pdac-engineered-cell-action-map", "relation": "resolved-by" }, { "target": "hypothesis-kras-tcr-human-persistence-gate", "relation": "tested-by" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "not-applicable", "randomised": "none", "independentReplication": "weak", "sampleSize": "Satri-cel n=24 infused; KRAS TCR n=7 reported across G12D and G12V; mesothelin CAR-T n=9 reported versus n=54 registry enrollment", "effect": "Mechanism discrimination and route-completion question; no controlled useful-life effect", "limits": [ "Architectures are not directly compared", "Very small mutation-HLA groups", "No randomized PDAC cell-therapy outcome", "Product and conditioning differ", "Population reach is unmeasured" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Both surface and presented-driver routes have produced human regression, while failure modes are measurable." }, "animalEvidence": { "rating": "strong", "reason": "Many engineered fixes work in models, but translation is the unresolved question." }, "inVitroEvidence": { "rating": "strong", "reason": "Address density, HLA presentation, cell potency and receptor loss can be tested." }, "reproducibility": { "rating": "weak", "reason": "Responses are sparse, uncontrolled and split across different products." }, "commercialConflicts": { "rating": "high", "reason": "Product owners control most trial records and manufacture." }, "recency": { "rating": "current", "reason": "Includes the 2026 KRAS-G12V report and current trial records." } } }, "layout": { "x": 244, "y": 166 }, "updated": "2026-09-15" }, { "id": "unknown-eus-rfa-added-benefit-after-hybridtherm", "title": "Can any endoscopic radiofrequency probe add useful life to modern systemic treatment?", "shortTitle": "EUS-RFA added benefit unknown", "type": "unknown", "status": "unresolved", "scope": [ "pdac", "locally-advanced", "borderline-resectable", "local-treatment" ], "summary": "HybridTherm proves access and local heating, not added survival; newer probes remain in small single-arm studies that cannot separate device effect from treatment and selection.", "content": "Known quantity. A probe passed through an endoscope can heat a pancreatic mass, and HybridTherm completed 37 procedures with no severe device adverse event. Unknown quantities. How much viable tumour remains outside the heated zone? Does heating release useful immune signals or only inflammation? Does repeated treatment delay or reduce chemotherapy? Which people have disease that will remain local long enough for local control to matter? Most important, does the same modern systemic treatment plus EUS-guided radiofrequency ablation improve overall or quality-adjusted survival over systemic treatment alone? HybridTherm's randomized point estimates do not answer yes: median survival was four months lower and resection was eight points lower with the device, both very uncertain. PANCARDINAL-1 plans 60 people in one arm and makes completion of chemotherapy, ablation and later resection its main endpoint. It can test delivery but cannot estimate the device's added effect. Resolve complete existing participant-level records, including the 66-versus-40 HybridTherm mismatch, before opening another efficacy trial. Separate pain-directed nerve ablation, benign endocrine tumours, cyst ablation and tumour ablation; they solve different problems.", "sourceIds": [ "testoni-2021-hybridtherm-rct", "clinicaltrials-nct02336672-hybridtherm", "clinicaltrials-nct04990609-pancardinal1" ], "links": [ { "target": "unknowns", "relation": "part-of" }, { "target": "claim-hybridtherm-randomized-no-patient-benefit", "relation": "bounded-by" }, { "target": "failure-hybridtherm-product-accrual-and-custody", "relation": "exposed-by" }, { "target": "hypothesis-eus-rfa-existing-study-gate", "relation": "tested-by" }, { "target": "trial-pdac-eus-rfa-action-map", "relation": "mapped-by" }, { "target": "company-erbe-elektromedizin", "relation": "constrained-by" }, { "target": "unknown-rfa-human-mechanism-after-negative-trial", "relation": "compares-with" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "moderate", "independentReplication": "weak", "sampleSize": "HybridTherm 40 randomized; PANCARDINAL-1 target 60 single arm", "effect": "Added overall survival, quality-adjusted survival and resection benefit remain unknown", "limits": [ "HybridTherm was underpowered", "Its product was withdrawn", "New probes have different geometry", "Successor trials are single arm", "Systemic regimens vary", "Local and distant progression compete", "The data mismatch is unresolved" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The probe can generate a local zone of heat-driven tissue death." }, "animalEvidence": { "rating": "moderate", "reason": "Local and immune effects are plausible in models but do not show human benefit." }, "inVitroEvidence": { "rating": "strong", "reason": "Thermal destruction is physically direct and measurable." }, "reproducibility": { "rating": "weak", "reason": "No modern independent randomized survival comparison has repeated the route." }, "commercialConflicts": { "rating": "high", "reason": "Device-specific data and operating knowledge remain manufacturer-dependent." }, "recency": { "rating": "current", "reason": "The uncertainty includes a July 2026 recruiting registry record." } } }, "layout": { "x": 784, "y": 708 }, "updated": "2026-09-15" }, { "id": "unknown-ferroptosis-cell-type-coverage-and-human-benefit", "title": "Can ferroptosis be confined to pancreatic cancer cells and improve systemic control?", "shortTitle": "Ferroptosis selectivity unknown", "type": "unknown", "status": "unresolved", "scope": [ "pdac", "ferroptosis", "drug-delivery", "immune-microenvironment", "resistance" ], "summary": "The missing chain runs from target-cell iron chemistry through spatial tumour kill and immune response to control of uninjected disease and longer useful life.", "content": "Known quantities. PDAC cells can depend on cystine import and GPX4 repair. Direct iron injection can create a dead-looking centre in an injected human lesion. Two pancreatic people have received the reported product. Unknown quantities. Which PDAC cell states are iron-rich or dependent enough to die at human exposure? Which normal pancreatic, liver, marrow, endothelial and immune cells share the dependency? Does direct injection distribute through the whole viable tumour or leave a growing rim? Are lipid peroxides, depleted GPX4 or glutathione and ferroptosis-specific membrane changes present in human tumour cells rather than only general necrosis? Does cell death activate useful CD8 T cells or the macrophage route that accelerated KRAS tumours in mice? Can any local effect alter uninjected lesions? Does later chemotherapy response exceed what would happen after an ordinary treatment break? Can a systemically delivered GPX4 or lysosomal-iron drug reach all lesions without harming immune cells and organs? Does sublethal exposure select a resistant state? Resolve injected and uninjected lesions, tumour and immune cells, drug distribution, ferroptosis markers, circulating immune changes, later therapy, all lesions, survival and harm in the same people. A black scan centre, stable disease or one 28-month survivor cannot close that chain.", "sourceIds": [ "daher-2019-xct-pdac-ferroptosis", "dai-2020-ferroptotic-damage-pdac", "li-2023-tumour-selective-gpx4-degrader", "caneque-2025-fentomycin-lysosomal-iron", "xie-2026-cnsi-fe-first-human", "clinicaltrials-nct07433283-cnsi-fe-expansion" ], "links": [ { "target": "unknowns", "relation": "part-of" }, { "target": "claim-ferroptosis-pdac-selectivity-not-class-effect", "relation": "bounded-by" }, { "target": "failure-cnsi-fe-resensitization-and-mechanism-claim", "relation": "exposed-by" }, { "target": "hypothesis-cnsi-fe-current-study-mechanism-gate", "relation": "tested-by" }, { "target": "trial-pdac-ferroptosis-action-map", "relation": "mapped-by" }, { "target": "company-sichuan-enray-pharmaceutical", "relation": "constrained-by" }, { "target": "unknown-selective-autophagy-pdac-contribution", "relation": "compares-with" } ], "evidence": { "humanProspective": "weak — delivery and safety only", "humanRetrospective": "weak", "randomised": "none", "independentReplication": "none in humans", "sampleSize": "Reported human n=19 mixed cancers, including n=2 pancreatic; expansion target n=54", "effect": "Human ferroptosis, cell selectivity, systemic response and benefit remain unmeasured", "limits": [ "Local injection", "Sparse pancreatic exposure", "No paired biopsies reported", "No human ferroptosis marker", "Immune direction can reverse", "No systemic tool", "Company-controlled product and data" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Every link is measurable and model evidence supports both vulnerability and risk." }, "animalEvidence": { "rating": "strong", "reason": "Mouse studies demonstrate why cell type and immune context decide direction." }, "inVitroEvidence": { "rating": "strong", "reason": "PDAC dependencies and resistant-state selection can be tested directly." }, "reproducibility": { "rating": "none", "reason": "No human PDAC mechanism result exists to repeat." }, "commercialConflicts": { "rating": "high", "reason": "The only human product is sponsor-owned and its mechanism was not independently measured." }, "recency": { "rating": "current", "reason": "The unknown incorporates a recruiting 2026 expansion and current chemical tools." } } }, "layout": { "x": 1180, "y": 1104 }, "updated": "2026-09-15" }, { "id": "unknown-gallbladder-prevention-selector", "title": "Unknown: who should have a gallbladder removed to prevent cancer?", "shortTitle": "Gallbladder prevention selector", "type": "unknown", "status": "open", "scope": [ "gallbladder", "hepatobiliary" ], "summary": "Gallstones and small polyps are common while cancer is rare; existing studies discover candidates, diagnose visible lesions or inspect removed organs, but do not validate future-risk surgery.", "content": "Class: known removable target organ but missing future-risk and action selector. Four questions must remain separate. 1. Who will develop HGD or invasive cancer if the gallbladder remains? The only 100,000-person public cohort designed around incidence still says not yet recruiting seven years after its estimated start. 2. Who already has cancer or HGD in a visible lesion? GBCseeker and GAIA-MIL can classify selected lesions, but their external sets are small and cancer-enriched; predictive values cannot be carried into low-incidence screening. 3. Who should receive follow-up or optional surgery? In POLYP, the revised European rule found one HGD and no cancer among 68 guideline-indicated operations, while most specimens had no neoplastic polyp. 4. Which removed organs require pathology or re-resection? FANCY and P-iGBC address this after-surgery decision, not prevention. EULAT, Chile BiLS and before-diagnosis serum cohorts can generate candidates but do not yet supply prospective country-level PPV or surgery benefit.\n\nDecision affected: whether prevention should use symptoms, stone burden, fixed polyp morphology, PSC or congenital disease, ancestry where appropriate, geography, inflammation or molecular markers to recommend surgery. Resolve it by extending active high-incidence and routine-surgery cohorts with an unoperated comparison, registry-linked future cancer, local elective-operation harms and fixed measurements. Validate a country-specific two-stage rule silently before it changes care, then randomize or stage its rollout. Stop if local calibration error exceeds five percentage points, sub-10-mm sensitivity stays below 80%, fewer than five expected cancer deaths are prevented per operative death, false-positive invasive work-up exceeds ten per HGD or stage-I cancer, or serious harm removes net benefit. A North Indian referral cohort showing 80.2% metastatic presentation strengthens urgency but supplies no population denominator.", "contentSections": [ "A 2025 review found the same selection problem across 788,214 people who underwent gallbladder removal. Older age, female sex, raised alkaline phosphatase, polyps over 10 mm and open surgery were associated with incidental cancer, while stones over 3 cm were not in the pooled analysis. Open surgery can be a consequence of operative difficulty or cancer suspicion, not a cause available for prevention selection. The negative pooled large-stone result conflicts with an operated-specimen association and blocks a stone-size-only rule." ], "sourceIds": [ "szpakowski-2020-gallbladder-polyps", "vineet-2026-north-india-gallbladder-cohort", "van-dijk-2025-polyp-study", "yang-2025-gbcseeker", "gupta-2026-gaia-mil", "clinicaltrials-nct03762837-gbc-risk-cohort", "clinicaltrials-nct06192719-eulat-gbc", "clinicaltrials-nct06531408-pigbc", "wong-2025-ambrose-cholecystectomy", "ramamoorthy-2025-incidental-gbc-meta" ], "links": [ { "target": "unknowns", "relation": "part-of" }, { "target": "claim-gallbladder-prevention-selection-bottleneck", "relation": "depends-on" }, { "target": "claim-gallbladder-polyp-selector", "relation": "depends-on" }, { "target": "model-gallbladder-causal-chain", "relation": "could-change" }, { "target": "trial-gallbladder-prevention-detection-action-map", "relation": "depends-on" }, { "target": "hypothesis-gallbladder-two-stage-interception-rule", "relation": "tested-by" } ], "evidence": { "humanProspective": "moderate for component studies; none for the final action rule", "humanRetrospective": "strong for polyp natural history and risk candidates", "randomised": "none", "independentReplication": "moderate for low polyp yield; none for a complete selector", "sampleSize": "622,227-person health-system cohort; prospective POLYP n=302; AMBROSE n=21,706 operations; active EULAT estimate n=15,000 and P-iGBC estimate up to n=30,000", "effect": "No validated preventive-surgery rule; one HGD and no cancer among 68 revised-guideline polyp operations", "limits": [ "Ultrasound measurement error", "Cancer is rare", "Surgical selection", "Regional incidence varies sharply", "High-grade precursor capture is incomplete", "Diagnostic models begin after pathology is visible", "Country-specific operation harm is not yet joined to cancer risk" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Removing a truly dangerous gallbladder before invasion can prevent the disease." }, "animalEvidence": { "rating": "weak", "reason": "Models cannot validate who should undergo human surgery." }, "inVitroEvidence": { "rating": "moderate", "reason": "Blood and tissue markers provide candidates without a future-risk action threshold." }, "reproducibility": { "rating": "none", "reason": "No country has independently validated a complete risk-to-surgery rule with clinical outcomes." }, "commercialConflicts": { "rating": "moderate", "reason": "Assay and imaging suppliers can benefit from broad testing, requiring independent validation and data rights." }, "recency": { "rating": "current", "reason": "The open question is defined by studies and registry states checked through September 2026." } } }, "layout": { "x": 36, "y": 105 }, "updated": "2026-09-14" }, { "id": "unknown-hcc-liver-reserve-modifiability", "title": "Unknown: can preserving liver reserve during HCC treatment extend survival?", "shortTitle": "HCC reserve modifiability", "type": "unknown", "status": "open", "scope": [ "hcc", "hepatobiliary" ], "summary": "Liver function strongly predicts HCC outcome, but prognosis does not prove that an active liver-preservation route changes survival.", "content": "Class: strong prognostic evidence, missing intervention evidence. Decision affected: whether a major HCC programme should fund tumour drugs alone or a coupled tumour-and-liver control trial. Early HCC cohorts assign a substantial share of deaths to non-cancer liver disease, and advanced-treatment survival falls steeply across Child-Pugh and ALBI groups. Those comparisons are confounded: patients with better liver function also differ in tumour burden, treatment access and general health. Resolving experiment: the proposed 24-centre cluster-randomized trial compares a timed liver-reserve protocol with standardized usual multidisciplinary care and measures death or first decompensation, tumour control, dose intensity, later-line access and quality of life. Cause-specific outcomes are adjudicated blind to arm. Stop if fewer than 85% receive required actions, cancer-treatment intensity falls by more than ten percentage points, early mortality rises, or the adequately delivered intervention produces no joint-endpoint signal. This unknown remains separate from whether one cancer drug works in Child-Pugh B.", "sourceIds": [ "seif-el-dahan-2026-early-hcc-causes-death", "storandt-2024-atezo-bev-liver-function", "finn-2020-imbrave150" ], "links": [ { "target": "unknowns", "relation": "part-of" }, { "target": "claim-hcc-tumour-and-liver-are-competing-failure-axes", "relation": "depends-on" }, { "target": "hypothesis-hcc-liver-reserve-co-control", "relation": "resolved-by" }, { "target": "model-hcc-causal-chain", "relation": "could-change" } ], "evidence": { "humanProspective": "none for co-control", "humanRetrospective": "strong", "randomised": "none for co-control", "independentReplication": "moderate for prognosis", "sampleSize": "Early HCC n=1,336; advanced treatment n=322; proposed trial about 600", "effect": "No causal estimate for preserving reserve", "limits": [ "Liver function is both cause and marker", "Bundle attribution", "Competing risks", "Changing usual care", "Cluster trial contamination" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Preserved liver reserve can prevent direct liver death and retain access to later tumour treatment." }, "animalEvidence": { "rating": "weak", "reason": "Models can change liver injury but cannot establish a safe clinical bundle across HCC treatments." }, "inVitroEvidence": { "rating": "weak", "reason": "Cell systems omit organ reserve and treatment sequencing." }, "reproducibility": { "rating": "weak", "reason": "Liver reserve repeatedly predicts outcome, but no randomized co-control route proves that changing it improves survival." }, "commercialConflicts": { "rating": "moderate", "reason": "The proposed bundle mixes low-cost care with medicines and procedures that can carry provider or company interests." }, "recency": { "rating": "current", "reason": "The main cause-of-death and treatment-state evidence was checked through 2026." } } }, "layout": { "x": 37, "y": 106 }, "updated": "2026-09-14" }, { "id": "unknown-hcc-surveillance-modality-utility", "title": "Unknown: which HCC surveillance route prevents late disease and death at acceptable cost and harm?", "shortTitle": "HCC surveillance utility gap", "type": "unknown", "status": "open", "scope": [ "hcc", "hepatobiliary" ], "summary": "MRI, ultrasound and blood panels differ in attendance, stage and false work-up, but comparative mortality evidence is not yet available.", "content": "The unresolved choice is not which test has the highest isolated sensitivity. It is which complete route gets used, produces fewer late HCCs, reaches curative treatment, avoids unnecessary CT or MRI, preserves liver function and reduces death. MIRACLE-HCC supports fewer false referrals and earlier stage with noncontrast MRI in 414 selected people, but reports no mortality comparison. The three-month interval trial and the serum-marker add-on trial show that more testing can fail at the next clinical step. TRACER should estimate late-stage disease, curative treatment and physical, financial and psychological harm for GALAD versus ultrasound with or without AFP; its primary completion is estimated for December 2029. PREMIUM should estimate HCC mortality for abbreviated contrast MRI plus AFP versus ultrasound plus AFP; its primary completion is estimated for September 2030. Until then, local changes should be treated as measured service trials, with explicit scan capacity, attendance, false work-up and treatment access, rather than assumed mortality improvements.", "sourceIds": [ "rhee-2025-miracle-hcc", "hirode-2026-hcc-biomarker-rct", "trinchet-2011-hcc-surveillance-interval", "singal-2024-tracer-protocol", "clinicaltrials-nct06084234-tracer", "ioannou-2026-premium-protocol", "clinicaltrials-nct05486572-premium", "zhang-2004-hcc-screening" ], "links": [ { "target": "unknowns", "relation": "part-of" }, { "target": "claim-hcc-surveillance-test-improvement-not-mortality", "relation": "depends-on" }, { "target": "trial-hcc-action-map", "relation": "resolved-by" }, { "target": "detection", "relation": "could-change" } ], "evidence": { "humanProspective": "strong for intermediate endpoints", "humanRetrospective": "not-applicable for the main comparison", "randomised": "strong for completed intermediate-endpoint trials; pending for modern utility trials", "independentReplication": "weak for mortality", "sampleSize": "Completed randomized evidence n=18,816, 1,278, 1,208 and 414; pending trials target 5,500 and 4,700", "effect": "No current comparative mortality estimate for GALAD or abbreviated MRI against modern ultrasound routes", "limits": [ "TRACER uses late-stage disease rather than death as its primary endpoint", "PREMIUM is restricted to VA participants with cirrhosis and MRI eligibility", "Results are years away", "Cause mix, obesity and local treatment access alter transport", "No one trial directly compares GALAD, noncontrast MRI and contrast MRI" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Surveillance can help only through earlier actionable disease and must compete with liver failure and work-up harm." }, "animalEvidence": { "rating": "not-applicable", "reason": "This is a human care-route and mortality question." }, "inVitroEvidence": { "rating": "weak", "reason": "Assay studies cannot determine attendance, false work-up, treatment or mortality." }, "reproducibility": { "rating": "weak", "reason": "Modern mortality results are not available and the completed trials answer different questions." }, "commercialConflicts": { "rating": "moderate", "reason": "Some biomarker investigators and sponsors have assay-company ties, while PREMIUM is publicly sponsored." }, "recency": { "rating": "current", "reason": "Trial status and estimated completion dates were checked in September 2026." } } }, "updated": "2026-09-14" }, { "id": "unknown-hifu-pain-survival-and-system-transfer", "title": "Does thermal HIFU improve pain, useful life or both, and do devices transfer?", "shortTitle": "HIFU patient benefit unknown", "type": "unknown", "status": "unresolved", "scope": [ "pancreatic-all", "pdac", "locally-advanced", "metastatic", "patient-reported-outcomes" ], "summary": "Pain response, survival, chemotherapy delivery and device performance must be measured as separate outputs under random assignment.", "content": "Known quantity. Thermal HIFU can destroy a visible pancreatic target without an incision, and uncontrolled cohorts repeatedly report less pain. Unknown quantities. How much pain improvement exceeds sham, expectation, medication change and ordinary chemotherapy response? Does it reduce opioid burden, hospital days or local complications? Does local destruction extend life, or does distant spread dominate? Which acoustic window, tumour depth, bowel position and device settings make treatment deliverable? Does pausing or delaying chemotherapy erase benefit? A HIFU machine is not a single intervention: guidance method, frequency, acoustic power, pulse pattern, focal scan, cooling, anaesthesia and thermal measurement all change delivered energy. The question also splits at mechanism. Thermal HIFU deliberately heats and destroys tissue. SonoPANIII's IMD10 is explicitly non-thermal and non-tissue-destructive and is intended to alter drug delivery; its future result cannot validate thermal ablation, and the reverse is also true. Resolve thermal HIFU with every screened and assigned person, a fixed same-treatment comparator, delivered acoustic and temperature data, pain plus analgesic use, quality-adjusted survival, chemotherapy dose and delay, local complications, distant failure and full harm. Analyze locally advanced and metastatic disease separately. Require the 40-person HIFU-PC15 result and the 90-person Suizenji overall-survival study before considering another trial. Device transfer requires a common delivered-energy and safety test, not brand names or scan brightness.", "sourceIds": [ "marinova-2018-hifu-prospective", "marinova-2024-hifu-pc-protocol", "drks-00012367-hifu-pc", "clinicaltrials-nct05601323-suizenji", "clinicaltrials-nct06211933-puls-hifu", "clinicaltrials-nct07325214-sonopan3", "clinicaltrials-nct07033689-sunrise2", "sonire-2026-nedo-award" ], "links": [ { "target": "unknowns", "relation": "part-of" }, { "target": "claim-hifu-pain-signal-without-survival-proof", "relation": "bounded-by" }, { "target": "failure-hifu-randomization-and-trial-drift", "relation": "exposed-by" }, { "target": "hypothesis-hifu-existing-randomized-gate", "relation": "tested-by" }, { "target": "trial-pdac-hifu-action-map", "relation": "mapped-by" }, { "target": "company-sonire-therapeutics", "relation": "constrained-by" }, { "target": "hypothesis-selected-local-consolidation", "relation": "extends" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "moderate", "randomised": "weak", "independentReplication": "moderate — symptoms; none for survival", "sampleSize": "Existing pain cohorts; HIFU-PC15 target n=40; Suizenji target n=90; SonoPANIII target n=360 but different non-thermal mechanism", "effect": "Controlled pain, quality-adjusted survival, chemotherapy preservation and device transfer remain unresolved", "limits": [ "No posted prospective randomized HIFU result", "Pain and analgesic endpoints vary", "No sham control", "Stage and systemic treatment differ", "Treatment physics differ by device", "Non-thermal focused ultrasound is a different route", "Travel and access burden are poorly measured" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Thermal tissue destruction and nerve interruption can plausibly reduce local pain." }, "animalEvidence": { "rating": "not-applicable", "reason": "The open question is comparative patient benefit." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Cell killing cannot settle pain, treatment delivery or survival." }, "reproducibility": { "rating": "weak", "reason": "Symptom signals repeat, but device-normalized randomized effects do not." }, "commercialConflicts": { "rating": "high", "reason": "Several live questions depend on company devices and sponsor-controlled data." }, "recency": { "rating": "current", "reason": "Live and stopped studies were checked in September 2026." } } }, "layout": { "x": 718, "y": 642 }, "updated": "2026-09-15" }, { "id": "unknown-hipec-added-value-peritoneal-pdac", "title": "Which part of cytoreduction plus HIPEC helps selected peritoneal PDAC?", "shortTitle": "HIPEC added value unknown", "type": "unknown", "status": "unresolved", "scope": [ "pdac", "hipec", "peritoneal-metastases", "surgery", "systemic-control" ], "summary": "The open question is not whether selected people can live longer, but whether heated chemotherapy adds anything to prior response and complete surgery.", "content": "Known quantities. A very small group with peritoneal-only disease, low burden and prolonged chemotherapy response can undergo complete cytoreduction plus HIPEC. Severe complications are common. In resectable PDAC, randomized HIPEC changed local recurrence but not survival. Unknown quantities. Of all people with PDAC peritoneal spread, how many satisfy the response and anatomy gates? How many begin assessment but progress before surgery? Does complete removal of visible disease account for the selected survival, does heated chemotherapy add to surgery, or is both merely a marker of unusually slow disease? Does HIPEC add value beyond unheated intraperitoneal drug? Which drug and temperature matter? Does repeating HIPEC improve peritoneal control or only add operative exposure? Does control of the abdominal lining reduce bowel blockage, fluid, pain and hospital time while preserving systemic chemotherapy? Do later liver and lung failures erase the local gain? Resolve the chain from first eligibility screen, through chemotherapy response, laparoscopy, assignment, complete cytoreduction, delivered heat and drug exposure, complications, systemic-treatment restart, local and distant recurrence, symptoms, quality-adjusted survival and death. Start every outcome clock at the decision to pursue the strategy, not the operation.", "sourceIds": [ "padilla-valverde-2024-adjuvant-hipec-rct", "grotz-2022-crs-hipec-pdac-pilot", "clinicaltrials-nct04858009-hipec-pdac", "yan-2024-crs-hipec-pdac-retrospective" ], "links": [ { "target": "unknowns", "relation": "part-of" }, { "target": "claim-hipec-local-control-without-survival-proof", "relation": "bounded-by" }, { "target": "failure-hipec-local-endpoint-and-package-selection", "relation": "exposed-by" }, { "target": "hypothesis-hipec-existing-trial-measurement-gate", "relation": "tested-by" }, { "target": "trial-pdac-hipec-action-map", "relation": "mapped-by" }, { "target": "unknown-pipac-added-survival-quality-and-systemic-control", "relation": "compares-with" }, { "target": "unknown-lapc-surgery-causal-benefit", "relation": "supports" } ], "evidence": { "humanProspective": "weak", "humanRetrospective": "weak", "randomised": "none for peritoneal metastatic PDAC", "independentReplication": "none for HIPEC contribution", "sampleSize": "Peritoneal prospective n=18, retrospective n=10, current single-arm target n=40", "effect": "No causal estimate of heated intraperitoneal chemotherapy beyond selection and surgery", "limits": [ "Tiny eligible population", "Long qualification period", "No untreated concurrent control", "Mixed primary-tumour treatment", "High operative harm", "Current study completes in 2029", "Unknown all-screened denominator" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "Surface drug exposure can add local kill, but prior chemotherapy and complete surgery remove much of the same measurable burden." }, "animalEvidence": { "rating": "not-applicable", "reason": "Human component attribution is now the binding question." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Heat-drug killing cannot price selection, systemic relapse or operative harm." }, "reproducibility": { "rating": "none", "reason": "No controlled peritoneal estimate exists." }, "commercialConflicts": { "rating": "low", "reason": "The main constraint is concentrated surgical expertise, not a central proprietary drug." }, "recency": { "rating": "current", "reason": "The question is anchored to the September 2026 Mayo record." } } }, "layout": { "x": 1120, "y": 1044 }, "updated": "2026-09-15" }, { "id": "unknown-histotripsy-human-coverage-and-benefit", "title": "Which pancreatic tumours can histotripsy fully and safely reach, and does that help the person?", "shortTitle": "Histotripsy coverage and benefit unknown", "type": "unknown", "status": "unresolved", "scope": [ "pdac", "locally-advanced", "oligometastatic", "local-treatment" ], "summary": "The missing answer joins an engineering boundary to a clinical one: visible acoustic access, complete tumour coverage, preserved systemic treatment and useful life.", "content": "Known quantities. Histotripsy can generate a visible cavitation cloud and a planned zone of tissue breakup. In favourable pig views, treatment stayed in the pancreas and some vessels remained open. Unknown quantities. What fraction of real human pancreatic tumours have a stable acoustic path after breathing, stomach and bowel motion? Can a device cover an irregular two-to-five-centimetre tumour plus margin while staying at least one centimetre from bowel, ducts and vessels? What happens when visibility changes during hundreds of pulses? Does breaking one local tumour change pain, chemotherapy delivery, distant spread, quality of life or survival? GANNON selects only people with visible targets, favourable clearance, good function and at least eight weeks of chemotherapy, and treats one pancreatic tumour. It can estimate safety in that narrow operating region. It cannot show who was screened out, how much viable tumour remained, or the added effect over the same chemotherapy without histotripsy unless those records are deliberately captured. Required measurements are the full screening denominator; exclusion geometry; breathing-resolved visibility; planned and achieved treatment volume; continuous bubble-cloud location; immediate and later viable tumour; bowel, duct and vessel injury; pancreatitis; chemotherapy interruption; pain; quality of life; local and distant progression; and survival. Mechanical destruction and any immune change must be reported separately.", "sourceIds": [ "gannon-2026-pancreas-histotripsy-safety", "clinicaltrials-nct06282809-gannon", "fda-den220087-edison" ], "links": [ { "target": "unknowns", "relation": "part-of" }, { "target": "claim-histotripsy-pancreas-visibility-limited", "relation": "bounded-by" }, { "target": "failure-histotripsy-acoustic-window-and-translation", "relation": "exposed-by" }, { "target": "hypothesis-gannon-independent-safety-gate", "relation": "tested-by" }, { "target": "trial-pdac-histotripsy-action-map", "relation": "mapped-by" }, { "target": "company-histosonics", "relation": "constrained-by" }, { "target": "unknown-hifu-pain-survival-and-system-transfer", "relation": "compares-with" } ], "evidence": { "humanProspective": "none — first feasibility study recruiting", "humanRetrospective": "none", "randomised": "none", "independentReplication": "weak", "sampleSize": "Nine healthy pigs; planned human n=50", "effect": "Human eligibility fraction, complete coverage, harm and added patient benefit are unknown", "limits": [ "No human pancreas results", "One small animal study", "Healthy tissue differs from tumour", "Only one lesion treated", "Strong anatomic selection", "Short primary safety window", "No comparator" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Focused cavitation mechanically fragments soft tissue." }, "animalEvidence": { "rating": "moderate", "reason": "The same system produced on-target and off-target outcomes in pigs." }, "inVitroEvidence": { "rating": "strong", "reason": "Bubble-cloud position and tissue fractionation are directly measurable." }, "reproducibility": { "rating": "weak", "reason": "Independent pancreatic replication is missing." }, "commercialConflicts": { "rating": "high", "reason": "The first human trial and key animal work depend on one manufacturer." }, "recency": { "rating": "current", "reason": "The first human trial remains active with no posted result." } } }, "layout": { "x": 820, "y": 744 }, "updated": "2026-09-15" }, { "id": "unknown-ire-added-survival-after-modern-chemotherapy", "title": "Does adding IRE after modern chemotherapy extend useful life?", "shortTitle": "IRE added benefit unknown", "type": "unknown", "status": "unresolved", "scope": [ "pdac", "local-control", "decision-critical" ], "summary": "The missing quantity is the all-randomized survival and total-harm difference between the same chemotherapy route with and without IRE.", "content": "Unknown quantity. For people with unresectable stage 3 pancreatic ductal adenocarcinoma who remain fit and progression-free after three months of mFOLFIRINOX, what is the causal effect of adding one NanoKnife IRE procedure on overall survival, two-year survival, hospital-free time, pain, quality of life and severe harm? Affected decision. This determines research-only use versus adoption, not whether electric pulses can kill local tissue. It also determines whether any later trial is needed. Why unresolved. CROSSFIRE compared IRE with radiotherapy rather than chemotherapy alone and stopped for futility. The DIRECT registry reports a large survival association after unequal selection routes and includes mostly reconstructed controls. The separate DIRECT phase 3 trial has the correct same-chemotherapy comparison, but its 528-person count is estimated, its April 2025 completion is estimated, and neither actual assignment nor results are public. LAP-PIE is a completed UK feasibility study planned for 50 people, not a powered efficacy answer, and its public registry has no result. Resolving experiment. Recover the complete NCT03899636 protocol, dated analysis plan, screening log, actual assignments, treatment delivery, vital status, quality of life, later treatment and every procedure and chemotherapy harm. Analyse everyone as assigned from randomization. Use the registry for rare-harm detection and route description, never as the causal survival estimate. Publish LAP-PIE recruitment, technical success and retention to decide whether a later independent trial is operationally possible if DIRECT is unusable.", "sourceIds": [ "timmer-2024-crossfire", "clinicaltrials-nct03899636-direct-rct", "clinicaltrials-nct03899649-direct-registry", "martin-2024-direct-registry-safety", "martin-2026-direct-registry-survival-preprint", "rai-2022-lappie-protocol", "isrctn-14986389-lappie", "nice-2017-ire-pancreatic" ], "links": [ { "target": "unknowns", "relation": "part-of" }, { "target": "claim-ire-ablation-without-proven-added-survival", "relation": "bounded-by" }, { "target": "failure-direct-selection-and-randomized-reporting-gap", "relation": "exposed-by" }, { "target": "hypothesis-direct-randomized-data-recovery-gate", "relation": "tested-by" }, { "target": "trial-pdac-irreversible-electroporation-action-map", "relation": "mapped-by" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "moderate", "randomised": "strong", "independentReplication": "weak", "sampleSize": "One randomized n=68 comparison without chemotherapy-only control; one phase 3 target n=528 estimated; one UK feasibility target n=50", "effect": "The answer may already exist in an unreported randomized record", "limits": [ "Actual phase 3 assignment count unknown", "No phase 3 result", "LAP-PIE result missing", "Observational control is small and selected", "Quality-of-life completeness unknown" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Local killing could help a local-dominant subgroup after systemic control." }, "animalEvidence": { "rating": "moderate", "reason": "Model evidence supports local ablation, not useful-life benefit." }, "inVitroEvidence": { "rating": "strong", "reason": "Electric-field killing is established outside patients." }, "reproducibility": { "rating": "weak", "reason": "No independent same-chemotherapy randomized survival result is available." }, "commercialConflicts": { "rating": "high", "reason": "AngioDynamics owns the product and the missing phase 3 record." }, "recency": { "rating": "current", "reason": "Both current registries and the 2026 survival preprint were checked." } } }, "layout": { "x": 568, "y": 492 }, "updated": "2026-09-15" }, { "id": "unknown-k912-ultrasound-drug-interaction-and-benefit", "title": "Does ultrasound change K-912 action beyond adding heat to chemotherapy?", "shortTitle": "K-912 sound-drug interaction unknown", "type": "unknown", "status": "unresolved", "scope": [ "pdac", "cholangiocarcinoma", "sonodynamic-therapy", "drug-delivery" ], "summary": "The central unknown is whether sound changes drug release or chemistry enough to produce benefit beyond the sum of ordinary epirubicin and thermal HIFU.", "content": "Known quantities. K-912 carries epirubicin in polymer micelles. MS-2 focused ultrasound at 75 to 150 watts can heat or mechanically disturb a small target. The combined package altered short-term scans in 12 people. Unknown quantities. How much K-912 reached the pancreatic tumour and peritoneal deposits? Did ultrasound release more active epirubicin, create reactive chemistry or merely add heat? Was the 150-watt response pattern a thermal dose effect? Would matched heat without acoustic pulses produce the same result? Does the package control disease beyond the focal target? Can it be repeated without cumulative heart, marrow, liver, skin or pancreatic injury? Would ordinary modern chemotherapy outperform or safely combine with it? Does pain improve after standardizing opioid use? The existing participant files should be recovered first, including acoustic waveforms, temperature estimates, K-912 blood levels, scan volumes, distant lesions, later therapy and complete survival. The mechanism then needs a four-way comparison: neither component, K-912 alone, matched thermal injury alone and the combination. A useful interaction means the combination exceeds the sum expected from each component, not merely that the combination kills cells.", "sourceIds": [ "muragaki-2026-k912-sonodynamic" ], "links": [ { "target": "unknowns", "relation": "part-of" }, { "target": "claim-k912-first-human-local-signal-no-component-proof", "relation": "bounded-by" }, { "target": "failure-k912-component-and-local-endpoint", "relation": "exposed-by" }, { "target": "hypothesis-k912-component-identification-gate", "relation": "tested-by" }, { "target": "trial-pdac-k912-sonodynamic-action-map", "relation": "mapped-by" }, { "target": "unknown-hifu-pain-survival-and-system-transfer", "relation": "compares-with" } ], "evidence": { "humanProspective": "weak", "humanRetrospective": "none", "randomised": "none", "independentReplication": "none", "sampleSize": "One combined-treatment cohort n=12", "effect": "Local package activity is visible; the interaction term and patient benefit are unidentified", "limits": [ "No component arms", "No direct tumour drug measurement reported", "No matched heat condition", "All participants had disseminated disease", "One-month outcome", "Repeat dose untested", "Mixed cancer types" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "Ultrasound can heat, cavitate and alter particles, but the claimed added interaction was not measured." }, "animalEvidence": { "rating": "moderate", "reason": "Animal package studies support testing but do not identify the human interaction." }, "inVitroEvidence": { "rating": "weak", "reason": "The exact K-912 release and matched-temperature comparison remains missing from the decision record." }, "reproducibility": { "rating": "none", "reason": "Only one clinical group has reported the combination." }, "commercialConflicts": { "rating": "moderate", "reason": "Drug, device ancestry and company-supported writing require independent component work." }, "recency": { "rating": "current", "reason": "The unknown follows directly from the 2026 first-human result." } } }, "layout": { "x": 1054, "y": 978 }, "updated": "2026-09-15" }, { "id": "unknown-kras-wildtype-driver-route-completion", "title": "How many KRAS-wild-type pancreatic patients reach a useful driver-matched treatment?", "shortTitle": "Rare-driver route unknown", "type": "unknown", "status": "open", "scope": [ "pdac", "precision-treatment", "gene-fusion", "care-delivery" ], "summary": "Driver frequency is measurable, but current records do not show the full tissue-to-RNA-to-drug route or target-specific survival benefit.", "content": "The route has three unanswered layers. Detection: among every new advanced PDAC, how many have adequate tumour, a valid KRAS result, reflex RNA when required and a result inside 14 days? Interpretation: how many results are strong activating drivers with a treatment that has pancreatic or credible tumour-wide evidence, rather than a variant of uncertain meaning? Delivery: how many patients receive specialist review, obtain the medicine before fitness loss, start it and remain on it with measured response, quality of life, resistance and survival? Each target stays separate. NTRK, RET, ALK, ROS1, FGFR2, BRAF, ERBB2, MSI-high and NRG1 do not share one drug or evidence level. The present data show rare dramatic cases and one NRG1 cohort, not the number helped across all eligible pancreatic patients. Resolve with a consecutive national route, not another selected case series.", "sourceIds": [ "philip-2022-kras-wildtype-pdac", "fusco-2021-kras-wildtype-fusions", "singh-2023-kras-wildtype-drivers", "pishvaian-2020-know-your-tumor", "mehdi-2024-kras-wildtype-case-series", "oreilly-2019-ntrk-pdac-response", "nhs-england-2026-cancer-genomic-test-directory" ], "links": [ { "target": "unknowns", "relation": "part-of" }, { "target": "unknown-nrg1-identification-and-survival-benefit", "relation": "contains" }, { "target": "failure-kras-wildtype-matching-route-and-selection", "relation": "learns-from" }, { "target": "hypothesis-kras-wildtype-driver-route-completion", "relation": "tested-by" }, { "target": "trial-pdac-kras-wildtype-action-map", "relation": "mapped-by" } ], "evidence": { "humanProspective": "weak", "humanRetrospective": "strong", "randomised": "weak — no all-route target-specific comparison exists", "independentReplication": "moderate — frequency repeats but completed treatment delivery does not", "sampleSize": "Profiling cohorts n=2,483, n=795, n=100 and n=1,856 referred; target-specific treatments remain small", "effect": "Decision-critical missing all-eligible and target-specific utility evidence", "limits": [ "Rare targets", "Tissue failure", "RNA availability", "Drug access", "Changing approvals", "Long survival follow-up" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Alternative drivers can replace KRAS and select different treatments." }, "animalEvidence": { "rating": "moderate", "reason": "Models support selected drivers but do not measure the human route." }, "inVitroEvidence": { "rating": "strong", "reason": "Direct pathway sensitivity is shown for some driver classes." }, "reproducibility": { "rating": "weak", "reason": "The national all-eligible route and target-specific outcomes have not been repeated." }, "commercialConflicts": { "rating": "moderate", "reason": "Both testing and several drugs have commercial owners." }, "recency": { "rating": "current", "reason": "Current English coverage and recent case data define the question." } } }, "layout": { "x": 2180, "y": 1600 }, "updated": "2026-09-15" }, { "id": "unknown-lapc-surgery-causal-benefit", "title": "Which treated locally advanced PDAC patient gains useful life from resection itself?", "shortTitle": "Conversion-surgery benefit unknown", "type": "unknown", "status": "supported", "scope": [ "pdac" ], "summary": "Current data can estimate prognosis and technical removal, but cannot identify the added survival and quality-of-life effect of surgery in today's responder state.", "content": "Decision: after initially locally advanced PDAC has remained confined on scans through systemic treatment, choose expert resection or continued non-surgical treatment. Three estimates must not be substituted for one another. Technical probability asks whether the surgeon can remove all visible disease. Prognosis asks how long a person with this response pattern is likely to live. Treatment effect asks how much longer or better that same person lives because surgery was added. NEOLAP, PREOPANC-4 and the TAPS calculators improve the first two. None answers the third. CA19-9 fall, CT shrinkage, lack of spread, fitness and time under treatment should define and stratify the decision state, but no value is a proved treatment-effect cutoff. The search found no modern registered trial that randomizes resection against continued treatment for non-metastatic locally advanced responders. NCT06714604 randomizes treatment duration before the surgical route; NCT06132087 is a two-person estimated single-arm pilot; neither isolates surgery. Required result: overall survival from randomization, quality-adjusted survival, time outside hospital, recurrence route, systemic treatment received, 90-day death, major complications, endocrine and digestive loss, pain, work and patient regret. Complete-margin resection is an intermediate result, not the answer.", "sourceIds": [ "dekker-2026-taps-restaging", "verkolf-2026-taps-resection-prediction", "kunzmann-2021-neolap", "hartlapp-2022-neolap-ca199", "guggenberger-2023-neolap-ct", "stoop-2026-preopanc4-surgery", "clinicaltrials-nct06714604-duration", "clinicaltrials-nct06132087-proactive" ], "links": [ { "target": "unknowns", "relation": "part-of" }, { "target": "trial-lapc-conversion-surgery-action-map", "relation": "mapped-by" }, { "target": "failure-conversion-surgery-selection-bias", "relation": "exposed-by" }, { "target": "hypothesis-randomized-lapc-resection-after-response", "relation": "tested-by" }, { "target": "claim-pdac-death-can-be-local-or-metastatic", "relation": "depends-on" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "strong", "randomised": "none — absent for the modern surgery decision", "independentReplication": "none — absent for a causal treatment selector", "sampleSize": "Modern cohorts n=130 to 2,338; direct old randomized trial n=42", "effect": "Unknown added useful-life effect of resection", "limits": [ "No modern randomized comparison", "Treatment and surgical techniques vary", "Definitions of locally advanced disease differ", "No validated surgery-benefit marker", "Crossovers will be unavoidable", "Expertise and complication rescue modify net benefit" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "Removing the last dominant local reservoir could help, but occult systemic clones may still determine death." }, "animalEvidence": { "rating": "weak", "reason": "Models cannot reproduce human vascular surgery and long-term systemic relapse." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Cell killing cannot value an operation against continued therapy." }, "reproducibility": { "rating": "weak", "reason": "The observational association repeats, but causal benefit has not been reproduced." }, "commercialConflicts": { "rating": "moderate", "reason": "Referral and procedural incentives can affect who is offered surgery and how results are reported." }, "recency": { "rating": "current", "reason": "Registry and publication search was refreshed on 15 September 2026." } } }, "layout": { "x": 110, "y": 172 }, "updated": "2026-09-15" }, { "id": "unknown-laser-ablation-delivery-and-net-benefit", "title": "Can laser heat cover pancreatic cancer without unacceptable leakage or delay?", "shortTitle": "Laser coverage and net benefit unknown", "type": "unknown", "status": "unresolved", "scope": [ "pdac", "laser-ablation", "thermal-dosimetry", "quality-of-life" ], "summary": "The missing quantities are complete thermal coverage, temperature at critical structures, procedure completion, systemic-treatment delivery and patient-valued benefit.", "content": "Known quantities. A narrow endoscopic needle can place a laser fibre inside selected tumours. Heat produces visible local coagulation. Nine early endoscopic procedures completed without reported harm. Unknown quantities. How much viable tumour was covered? What temperature reached the pancreatic duct, duodenum, stomach, major vessels and nerves? Can planning software predict the treated shape across tumour locations and blood-flow cooling? What causes a procedure to be attempted but cancelled? Can hardware and disposables remain available across sites? Does laser treatment delay chemotherapy or reduce its dose? Do local effects reduce pain and obstruction, or add pancreatic-fluid leaks and hospital time? Does any claimed immune response change distant disease? The eight-person randomized record could at least reveal assignment flow and early outcomes, but nothing is posted. The later five- and ten-person endoscopic records could test reproducibility and planning, but one terminated and one is stale. Recover every record and separate bare-fibre endoscopic ablation from open surgical imILT. Only random assignment with the same chemotherapy in both groups can test added patient benefit.", "sourceIds": [ "dimatteo-2018-eus-laser-ablation", "paiella-2021-imilt-pancreas", "clinicaltrials-nct03187587-imilt-randomized", "clinicaltrials-nct03784417-eus-laser", "clinicaltrials-nct05549960-laser-model" ], "links": [ { "target": "unknowns", "relation": "part-of" }, { "target": "claim-pancreatic-laser-heat-feasibility-not-benefit", "relation": "bounded-by" }, { "target": "failure-laser-fistula-accrual-and-device-continuity", "relation": "exposed-by" }, { "target": "hypothesis-laser-existing-study-recovery-gate", "relation": "tested-by" }, { "target": "trial-pdac-laser-ablation-action-map", "relation": "mapped-by" }, { "target": "unknown-pdt-spatial-coverage-and-patient-benefit", "relation": "compares-with" } ], "evidence": { "humanProspective": "weak", "humanRetrospective": "none", "randomised": "weak — data unavailable", "independentReplication": "weak", "sampleSize": "Known human programmes n=9, 15, 8 and 5; planning study target n=10 with unknown status", "effect": "Local heating is feasible; reliable coverage, leakage risk and net benefit remain unknown", "limits": [ "Sparse participant-level data", "Different devices and routes", "No common thermal map", "No controlled patient outcomes", "Programme attrition", "Company priorities have moved elsewhere" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Local heat injury is certain; selective whole-tumour coverage is the unresolved engineering task." }, "animalEvidence": { "rating": "moderate", "reason": "Animal studies support controlled lesions but do not resolve diseased human anatomy." }, "inVitroEvidence": { "rating": "strong", "reason": "Thermal dose and cell death are measurable under controlled conditions." }, "reproducibility": { "rating": "weak", "reason": "Tiny programmes have not produced a complete shared result." }, "commercialConflicts": { "rating": "high", "reason": "Some central data and device continuity depend on the manufacturer." }, "recency": { "rating": "current", "reason": "No newer completed pancreatic treatment programme was found in the current registry search." } } }, "layout": { "x": 982, "y": 906 }, "updated": "2026-09-15" }, { "id": "unknown-localized-losartan-contribution-pdac", "title": "Does losartan improve the complete localized-PDAC treatment route?", "shortTitle": "Localized losartan unknown", "type": "unknown", "status": "open", "scope": [ "pdac", "transfer-unproven" ], "summary": "Advanced disease is negative, while the localized question can be answered from the existing 168-person NCT03563248 trial.", "content": "Known. The 49-person losartan/FOLFIRINOX/radiation/surgery route reported high clean-margin resection and long survival but cannot isolate losartan. AFPAC found no efficacy signal in advanced disease. Unknown. Losartan may still open vessels and alter radiation response while disease remains localized, yet that effect must improve the whole route rather than just resection pathology. Existing answer. NCT03563248 has actual enrollment 168 and randomizes FOLFIRINOX with or without losartan, then SBRT with or without nivolumab before attempted surgery. Its primary endpoint is clean-margin resection; PFS, OS, complete response and serious harm are listed. Primary collection ended in February 2022, no results are posted and individual data sharing says no. Resolution. Publish all assigned participants, factorial contribution estimates for losartan and nivolumab, chemotherapy completion, blood pressure, perfusion or oxygen measurement, surgery attempted, severe harm, recurrence, PFS, OS and patient function. A losartan claim requires a controlled treatment-route improvement with preserved survival and harm, not more R0 resections alone. No new cohort and £0 mission capital are justified before this record is released.", "sourceIds": [ "murphy-2019-losartan-lapc", "ramaswamy-2025-afpac-losartan", "clinicaltrials-nct03563248-losartan-localized" ], "links": [ { "target": "unknowns", "relation": "part-of" }, { "target": "failure-losartan-delivery-signal-not-replicated", "relation": "bounded-by" }, { "target": "hypothesis-localized-losartan-existing-trial-gate", "relation": "tested-by" }, { "target": "trial-pdac-vascular-hypoxia-action-map", "relation": "resolved-by" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "moderate", "independentReplication": "moderate", "sampleSize": "Localized single-arm n=49; existing localized randomized n=168; advanced randomized n=88", "effect": "Setting-specific treatment-route question; no assigned benefit", "limits": [ "Four trial groups complicate contribution estimates", "Primary endpoint is resection margin, not survival", "No data-sharing plan", "Final follow-up continues", "Blood-flow measures may be incomplete" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Localized compressed tumours may differ from widespread advanced disease." }, "animalEvidence": { "rating": "strong", "reason": "Models support pressure reduction and improved delivery." }, "inVitroEvidence": { "rating": "weak", "reason": "The physical flow mechanism needs intact tissue and circulation." }, "reproducibility": { "rating": "weak", "reason": "The positive localized signal is uncontrolled and the randomized localized result is unavailable." }, "commercialConflicts": { "rating": "low", "reason": "Generic drug economics reduce product-owner pressure, though investigators control data." }, "recency": { "rating": "current", "reason": "The existing trial record was checked through April 2026." } } }, "layout": { "x": 334, "y": 286 }, "updated": "2026-09-15" }, { "id": "unknown-microbiome-treatment-contribution-pdac", "title": "Does changing a measured pancreatic microbe state improve treatment benefit enough to outweigh infection and resistance harm?", "shortTitle": "Microbiome treatment contribution", "type": "unknown", "status": "open", "scope": [ "pdac", "transfer-unproven" ], "summary": "Human causality requires target change, drug or immune change, and patient benefit in the same controlled record.", "content": "The unresolved question is narrower than whether PDAC has microbes. It asks whether a defined intervention changes a measured microbial function in people, whether that change alters active chemotherapy or immune entry, and whether patients live or function longer without unacceptable infection, gut injury or antibiotic resistance. Separate three products. Antibiotics remove a broad set of bacteria and may increase active gemcitabine exposure, toxicity and resistant organisms. Stool transfer supplies a changing community whose donor, manufacture and engraftment vary. A live bacterial product supplies a named strain but still needs tumour entry or a measurable distant immune route. Current human evidence does not allow these products to borrow from one another. Resolve the antibiotic route first in PANORAMIX Step 2 because it has placebo, masking, gemcitabine/dFdU drug measurements and resistance endpoints. Resolve stool transfer through the randomized FMTPanc safety study before efficacy. Treat the 25-person antibiotic-pembrolizumab and 20-person CHASe-PDAC studies as mechanism and safety records only. A causal pass requires all assigned participants, a fixed microbial assay, target change before outcome, at least a ten-point improvement in six-month survival or PFS HR at most 0.75 with a compatible OS estimate, and no more than a five-point excess in severe infection or treatment-limiting toxicity. Stop for missing randomized denominators, no target change, marker-only movement, resistant-organism excess of ten points, severe-harm excess, or a benefit found only after selecting completers.", "sourceIds": [ "geller-2017-bacterial-gemcitabine", "corty-2020-antibiotic-gemcitabine-toxicity", "sharma-2026-fmtpanc", "ctis-panoramix-2024-517766-41", "clinicaltrials-nct05462496-antibiotic-pembro-pdac", "clinicaltrials-nct06393400-chase-pdac" ], "links": [ { "target": "unknowns", "relation": "part-of" }, { "target": "claim-pdac-microbiome-mechanism-without-human-benefit", "relation": "bounded-by" }, { "target": "failure-microbiome-proxy-without-patient-benefit", "relation": "learns-from" }, { "target": "hypothesis-bacterial-gemcitabine-deactivation-gate", "relation": "resolved-by" }, { "target": "trial-pdac-microbiome-action-map", "relation": "resolved-by" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "moderate", "randomised": "weak", "independentReplication": "weak", "sampleSize": "Two controlled programmes active or recruiting; no reported randomized clinical result", "effect": "Unknown treatment contribution", "limits": [ "The exact PANORAMIX Step 2 sample is not public on ClinicalTrials.gov", "FMT donor and engraftment can vary", "Tumour microbial measurement may be missing", "Thresholds are mission decisions" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The microbe-to-drug and microbe-to-immune chains are measurable." }, "animalEvidence": { "rating": "strong", "reason": "Multiple microbial interventions change tumour growth in models." }, "inVitroEvidence": { "rating": "strong", "reason": "Gemcitabine inactivation supplies a direct biochemical measurement." }, "reproducibility": { "rating": "weak", "reason": "No human clinical contribution has been reproduced." }, "commercialConflicts": { "rating": "moderate", "reason": "Stool manufacture and live products need locked product and data rights." }, "recency": { "rating": "current", "reason": "The decision uses trials recruiting in 2026." } } }, "layout": { "x": 370, "y": 324 }, "updated": "2026-09-15" }, { "id": "unknown-mtap-prmt5-ras-contribution", "title": "How much does PRMT5 inhibition add to RAS inhibition in MTAP-deleted PDAC?", "shortTitle": "PRMT5-RAS added effect", "type": "unknown", "status": "open", "scope": [ "pdac" ], "summary": "The combination response is large, but no controlled experiment separates vopimetostat, the RAS drug and their interaction.", "content": "Known. MTAP loss occurs in roughly 28% of PDAC, vopimetostat has single-agent activity, and 25 of 39 evaluable people responded to one of two early RAS combinations. Unknown. The extra response probability, duration and survival caused by adding vopimetostat; whether benefit differs by exact MTAP deletion, KRAS allele, line of care or liver metastasis; and whether added toxicity reduces RAS-drug exposure. Why it matters. In business terms, the package has apparent product-market fit in a selected segment but no component attribution. A positive package-versus-chemotherapy trial could support use while still leaving an avoidable drug, toxicity and price inside the package. Resolving evidence. All-treated mature data can test whether the headline survives follow-up but cannot prove contribution. The decisive experiment is randomized RAS inhibitor alone versus the identical RAS inhibitor plus vopimetostat, stratified by exact MTAP loss, line and liver disease, with progression-free and overall survival, response duration, quality-adjusted survival and dose intensity. Stop treating the interaction as established if the added arm gives PFS HR above 0.80, its 95% interval includes 1.0 after the planned event count, confirmed response improves by less than 15 points, severe harm rises by more than 10 points or RAS dose intensity falls by more than 10 points.", "sourceIds": [ "matsumoto-2026-mtap-pancreatic", "pavlick-2026-mtap-genomics", "clinicaltrials-nct06922591-vopimetostat-ras", "tango-2026-vopimetostat-ras-data" ], "links": [ { "target": "unknowns", "relation": "part-of" }, { "target": "claim-mtap-prmt5-selected-human-signal", "relation": "limits" }, { "target": "failure-vopimetostat-ras-contribution-unknown", "relation": "caused-by" }, { "target": "trial-mtap-prmt5-action-map", "relation": "resolved-by" }, { "target": "hypothesis-mtap-prmt5-ras-deeper-control", "relation": "tests" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "strong", "randomised": "none", "independentReplication": "none", "sampleSize": "Current combination evidence: 54 treated and 39 response evaluable PDAC", "effect": "Required output is the randomized added effect of vopimetostat", "limits": [ "Thresholds are mission decisions", "Future protocol not fixed", "Drug access depends on sponsors" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "A two-dependency interaction is plausible but its size and duration are unknown." }, "animalEvidence": { "rating": "moderate", "reason": "Models cannot replace the missing component comparison." }, "inVitroEvidence": { "rating": "strong", "reason": "Both targets and the selected genotype are measurable." }, "reproducibility": { "rating": "none", "reason": "No randomized contribution estimate exists." }, "commercialConflicts": { "rating": "high", "reason": "Resolving evidence depends on company-controlled products and data." }, "recency": { "rating": "current", "reason": "The uncertainty is defined against June 2026 data." } } }, "layout": { "x": 188, "y": 109 }, "updated": "2026-09-15" }, { "id": "unknown-mwa-safe-coverage-and-patient-benefit", "title": "Can pancreatic microwave ablation cover enough tumour without adding heat injury or treatment delay?", "shortTitle": "MWA coverage and benefit unknown", "type": "unknown", "status": "unresolved", "scope": [ "pdac", "unresectable", "local-treatment", "drug-delivery" ], "summary": "Technical placement succeeds more often than complete ablation; the missing answer connects heat coverage, local and distant failure, pain, systemic treatment and survival.", "content": "Known quantity. A microwave antenna can heat pancreatic tissue quickly, and selected people often report less pain. Unknown quantities. Why did technical success reach 100% while complete ablation reached only 40.5%? Was the missing 59.5% a deliberate safety margin near bowel, bile duct and vessels, inadequate energy or irregular tumour geometry? How does the achieved temperature field relate to viable tumour, pancreatitis, fistula and bleeding? Does local heating preserve or delay chemotherapy? Does it improve pain more safely than cryoablation or HIFU? Most importantly, does it change survival when entry burden and systemic care are identical? Existing records cannot answer. The 150-person study assigns easier cases to microwave and lacks imaging in 20 of 67. The prospective 12-person package changes immune drugs, biopsy, operation, microwave and chemotherapy, then analyzes eight completers. NCT06231160 could provide a larger two-cohort record if it enrolled, but non-probability sampling still prevents a clean effect estimate. Recover energy, antenna track, planned and achieved heat coverage, direct biopsy, complications, systemic treatment, pain, quality of life, local and distant progression and survival for every eligible person. Treat fatal fistula and missing scans as outcomes, not exclusions.", "sourceIds": [ "zhou-2024-pancreatic-mwa", "vogl-2026-tace-mwa", "topal-2025-mwa-dual-ici", "clinicaltrials-nct06231160-mwa-cohort" ], "links": [ { "target": "unknowns", "relation": "part-of" }, { "target": "claim-mwa-pain-signal-no-added-benefit", "relation": "bounded-by" }, { "target": "failure-mwa-selection-bundling-and-missing-imaging", "relation": "exposed-by" }, { "target": "hypothesis-mwa-existing-data-gate", "relation": "tested-by" }, { "target": "trial-pdac-mwa-action-map", "relation": "mapped-by" }, { "target": "unknown-rfa-human-mechanism-after-negative-trial", "relation": "compares-with" }, { "target": "unknown-cryoablation-randomized-pain-and-net-benefit", "relation": "compares-with" } ], "evidence": { "humanProspective": "weak", "humanRetrospective": "moderate", "randomised": "none", "independentReplication": "weak", "sampleSize": "Existing n=72, n=150 and n=12; planned observational n=185", "effect": "Complete target coverage, added pain relief and survival remain unresolved", "limits": [ "Energy systems vary", "Imaging records incomplete", "Ablation often partial", "Systemic treatments differ", "Multiple bundled procedures", "Strong allocation bias", "No quality-adjusted survival estimate" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "A measured microwave field can heat and kill tissue." }, "animalEvidence": { "rating": "moderate", "reason": "Thermal safety margins can be modeled, but human anatomy and treatment interactions dominate." }, "inVitroEvidence": { "rating": "strong", "reason": "Temperature and cell death are direct physical outputs." }, "reproducibility": { "rating": "weak", "reason": "Coverage and clinical outcome are not consistently reported across systems." }, "commercialConflicts": { "rating": "moderate", "reason": "Raw energy and device records must be portable across suppliers." }, "recency": { "rating": "current", "reason": "The uncertainty includes 2026 outcomes and trial status." } } }, "layout": { "x": 886, "y": 810 }, "updated": "2026-09-15" }, { "id": "unknown-nrg1-identification-and-survival-benefit", "title": "Can every NRG1-fusion PDAC be found in time, and does zenocutuzumab extend useful life?", "shortTitle": "NRG1 route and benefit unknown", "type": "unknown", "status": "open", "scope": [ "pdac", "nrg1", "rna-testing", "survival", "access" ], "summary": "The treatment signal is actionable now, while population detection, UK access, resistance and comparative survival remain unresolved.", "content": "Unknown 1: among all consecutive advanced PDAC diagnoses, what fractions receive adequate tissue sampling, KRAS status, reflex RNA fusion testing and an NRG1 result before the next treatment decision? Unknown 2: which assays and fusion partners produce false negative or disputed results? Unknown 3: how many positive patients are fit, referred, authorized, infused and followed, and how long does each step take? Unknown 4: does the required 50-person FDA follow-up reproduce the 40% response and show enough duration to verify benefit? Unknown 5: what changes cause resistance after initial response? Unknown 6: is zenocutuzumab licensed, commissioned and obtainable in the UK, or available only by a special route? Unknown 7: does treatment improve overall survival, quality-adjusted survival or symptom time against the treatment a comparable patient would otherwise receive? A response scan answers none of the last three questions by itself.", "sourceIds": [ "fda-2024-bizengri-multidisciplinary-review", "fda-2026-ongoing-bizengri-accelerated-approval", "clinicaltrials-nct02912949-enrgy", "schram-2026-nrg1-fusion-detection", "nhs-england-2026-cancer-genomic-test-directory" ], "links": [ { "target": "unknowns", "relation": "part-of" }, { "target": "claim-nrg1-approved-response-not-durable-control", "relation": "qualifies" }, { "target": "failure-nrg1-testing-and-confirmation-route", "relation": "caused-by" }, { "target": "hypothesis-nrg1-rna-route-completion", "relation": "tested-by" }, { "target": "trial-pdac-nrg1-action-map", "relation": "tracked-by" } ], "evidence": { "humanProspective": "moderate for response; none for route completion", "humanRetrospective": "moderate for RNA detection", "randomised": "none", "independentReplication": "none for survival benefit", "sampleSize": "Thirty-person FDA efficacy set and unresolved at-least-50-person requirement", "effect": "Known response probability, unknown found-and-treated probability and comparative useful-life gain", "limits": [ "Rare event", "No common national denominator", "Current regulatory deadline passed", "UK access not established", "No randomized control", "Evolving assay coverage" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "NRG1 is a measured driver and the drug directly blocks its receptor pair." }, "animalEvidence": { "rating": "not-applicable", "reason": "Human treatment evidence exists." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "The remaining questions are detection, access, resistance and useful life." }, "reproducibility": { "rating": "weak", "reason": "One pivotal clinical programme supplies the treatment result." }, "commercialConflicts": { "rating": "high", "reason": "Company ownership spans the product and pivotal evidence." }, "recency": { "rating": "current", "reason": "The uncertainty includes the current overdue FDA follow-up state." } } }, "layout": { "x": 1222, "y": 1146 }, "updated": "2026-09-15" }, { "id": "unknown-older-pdac-treatment-route", "title": "Unknown: which older or vulnerable person gains useful life from each PDAC treatment route?", "shortTitle": "Older-PDAC treatment route", "type": "unknown", "status": "supported", "scope": [ "pdac" ], "summary": "No completed study joins a pre-treatment vulnerability state to a tested choice among combination treatment, gentler treatment, supportive care and rapid corrective support.", "content": "Class: missing action evidence. Keep five choices separate. First, whether anticancer treatment is likely to add useful life over supportive care. Second, which drug family best matches the tumour. Third, the starting dose and schedule. Fourth, which correctable problem—nutrition, medicines, mobility, mood, cognition, pain, pancreatic enzyme shortage, transport or caregiver load—must be repaired immediately. Fifth, when worsening disease or harm should stop treatment. GIANT compared two reduced-intensity drug combinations after selecting vulnerable patients; it did not randomize the decision to treat or the support system. GrantPax let clinicians override its categories and was too small. PAMELA-70 treated 69 people with one adjusted FOLFIRINOX route; its April 2026 registry result reports only a composite 65.8%, without separate response, independence or survival, and 48/69 had a serious adverse event. Mixed-cancer trials show that acting on assessment can reduce severe toxicity but have not proved a PDAC survival gain. Resolving experiment: register every consecutive older or screen-positive patient before the plan; record patient priorities and a short domain inventory; then randomize people choosing systemic treatment to the assessment report alone or the same report plus a team that completes named actions within 72 hours. Preserve the untreated group as a complete observed route rather than deleting it. The primary test is severe treatment toxicity; key patient tests are days alive outside hospital, function, symptoms, goal-concordant care, dose delivered and survival. Go if at least 90% are assessed before the final plan, at least 85% of named actions occur within 72 hours, severe toxicity falls ten points, and people gain at least seven more days alive outside hospital by week 16 without reducing treatment starts by more than five points. Stop for median treatment delay above three days, 90-day death up three points, seven-day emergency care up five points, an inequality gap above ten points, or no patient-valued gain. This experiment cannot prove which molecular drug is best; it tests whether the routing and support layer makes chosen treatment safer and more usable.", "sourceIds": [ "dotan-2024-giant", "dotan-2025-giant-ga", "clinicaltrials-nct04233866-giant", "betge-2022-grantpax", "clinicaltrials-nct02143219-pamela70", "mohile-2021-gap70", "li-2021-gain", "ihorst-2026-geriatric-comanagement", "clinicaltrials-nct06040801-frail-gi" ], "links": [ { "target": "unknowns", "relation": "part-of" }, { "target": "claim-age-alone-cannot-route-pdac-treatment", "relation": "bounded-by" }, { "target": "failure-grantpax-geriatric-route", "relation": "exposed-by" }, { "target": "trial-older-pdac-action-map", "relation": "tested-by" }, { "target": "hypothesis-older-pdac-fast-support-route", "relation": "resolved-by" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong for pieces; absent for the complete PDAC route", "independentReplication": "moderate", "sampleSize": "PDAC trials n=176, n=72 and n=32; mixed-cancer randomized trials n=718, n=605 and n=217", "effect": "No validated selection rule; proposed ten-point severe-toxicity and seven-day home-time gates", "limits": [ "Generated action thresholds", "Untreated choice remains observed", "Regimens evolve", "Mixed-cancer transfer", "Rapid PDAC decline", "Contamination between care teams" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Correctable host problems can change tolerance and exposure, while disease speed can erase the time available to act." }, "animalEvidence": { "rating": "not-applicable", "reason": "The unknown concerns human priorities, function and care delivery." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "A tumour assay cannot select supportive actions or measure useful life." }, "reproducibility": { "rating": "moderate", "reason": "Toxicity reduction repeats across cancers, but a full PDAC route and survival effect do not." }, "commercialConflicts": { "rating": "low", "reason": "The resolving experiment is drug-neutral and requires independent analysis." }, "recency": { "rating": "current", "reason": "Includes 2026 PAMELA-70 posting, GIANT follow-up and a current mixed-GI registry." } } }, "updated": "2026-09-14" }, { "id": "unknown-oncosil-randomized-survival-and-delivery", "title": "Does modern phosphorus-32 add useful life on the same FOLFIRINOX backbone?", "shortTitle": "OncoSil randomized benefit unknown", "type": "unknown", "status": "unresolved", "scope": [ "pdac", "locally-advanced", "brachytherapy", "quality-of-life" ], "summary": "TRIPP-FFX contains the needed contrast: identical chemotherapy with or without implantation; the unresolved quantities are all-assigned delivery, survival, symptoms, quality of life, surgery and harm.", "content": "Known quantities. Modern microparticles can be injected endoscopically, early imaging usually shows activity at the intended tumour, and 88 people have been randomly assigned on a fixed FOLFIRINOX backbone. Unknown quantities. How many assigned to OncoSil were actually implanted, and why were any cancelled? Did measured radiation cover the viable tumour without reaching stomach, duodenum or blood? Did implantation delay or reduce chemotherapy? Does a 16-week local scan difference become longer progression-free or overall survival? Does it reduce pain and improve quality-adjusted time, or merely increase procedures and adverse events? Does any extra resection produce a clear margin and useful life after surgical harm? The old colloidal product shows why every layer matters: a 70-point liquefaction difference coexisted with more serious events and a four-month worse survival point estimate. Resolve TRIPP-FFX in all 88 as assigned. Report the randomization flow, implantation success, dose and spatial distribution, chemotherapy dose intensity, local and distant progression, symptoms, validated quality of life, operations, 90-day surgical outcomes and survival. Present local control as a mechanism output, not the decision endpoint. Separate the modern silicon-particle estimate from old colloid and from iodine-125 seeds.", "sourceIds": [ "clinicaltrials-nct05466799-tripp-ffx", "ross-2022-panco-p32", "rosemurgy-2008-p32-randomized", "lim-2026-p32-landmark" ], "links": [ { "target": "unknowns", "relation": "part-of" }, { "target": "claim-oncosil-local-control-signal-awaits-randomized-result", "relation": "bounded-by" }, { "target": "failure-p32-liquefaction-survival-and-reporting-gap", "relation": "exposed-by" }, { "target": "hypothesis-tripp-ffx-independent-analysis-gate", "relation": "tested-by" }, { "target": "trial-pdac-p32-action-map", "relation": "mapped-by" }, { "target": "company-oncosil-medical", "relation": "constrained-by" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "moderate", "randomised": "moderate — complete enrollment without result", "independentReplication": "weak", "sampleSize": "Modern randomized n=88; pilot n=50; adjusted n=86; old randomized n=30", "effect": "Modern all-assigned survival, quality of life and net benefit remain unknown", "limits": [ "Primary local-control endpoint", "Open-label design", "Industry sponsor", "Final follow-up runs to 2027", "No public result", "Some implants may be cancelled", "Old formulation is not directly transferable" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The device should deposit a localized radiation dose inside tumour." }, "animalEvidence": { "rating": "moderate", "reason": "Distribution and dosimetry are plausible; the human comparison is decisive." }, "inVitroEvidence": { "rating": "strong", "reason": "Beta-radiation dose and cell injury are measurable." }, "reproducibility": { "rating": "moderate", "reason": "Delivery is multicentre, but independent outcome replication is absent." }, "commercialConflicts": { "rating": "high", "reason": "The company sponsors and controls access to the pivotal dataset." }, "recency": { "rating": "current", "reason": "Primary completion occurred in January 2026." } } }, "layout": { "x": 916, "y": 840 }, "updated": "2026-09-15" }, { "id": "unknown-particle-therapy-added-patient-benefit", "title": "Do proton or carbon-ion beams add useful life over current care?", "shortTitle": "Particle patient benefit unknown", "type": "unknown", "status": "unresolved", "scope": [ "pdac", "locally-advanced", "proton", "carbon-ion", "value" ], "summary": "Dose geometry, local control, total burden, survival and value have not been measured together in a valid pancreatic cancer comparison.", "content": "Known quantity. Particle beams can reduce or reshape dose outside the intended target, carbon ions can deliver denser ionization, and several centres can treat locally advanced pancreatic cancer with high local control. Unknown quantities. Does this reduce serious stomach and bowel injury compared with modern photon plans? Does it preserve chemotherapy dose and time at home? Does improved local control extend overall or quality-adjusted survival when distant spread is common? Which moving anatomy receives the planned dose each day? Breathing, bowel filling and gas change the material a particle crosses and therefore where it stops. Carbon-ion dose also includes a modelled relative biological effectiveness, meaning an estimate of damage compared with standard X-rays; models and centres are not automatically interchangeable. Who can travel to a scarce centre without delaying systemic treatment, and what is the full patient and capital cost? The largest recent uncontrolled contrast inside one report is local control 94.7% versus freedom from distant spread 44.5% at one year, a 50.2-point gap. That gap argues for a systemic-risk selector before local escalation. Resolve the immediate evidence debt by publishing every participant and endpoint from completed PACK, maturing the 60-person proton registry without another extension, and joining delivered rather than planned dose, anatomy changes, chemotherapy, distant spread, severe harm, quality of life and survival. No cross-cohort survival claim passes without a concurrent current-care comparator.", "sourceIds": [ "shinoto-2016-carbon-ion-gemcitabine", "kawashiro-2018-jcros-carbon-ion", "lai-2026-carbon-ion-risk-adapted", "rapp-2022-proton-phase2", "eckstein-2023-proton-registry", "clinicaltrials-nct02598349-proton", "clinicaltrials-nct03536182-cipher", "clinicaltrials-nct04194268-pack" ], "links": [ { "target": "unknowns", "relation": "part-of" }, { "target": "claim-particle-therapy-local-control-without-survival-proof", "relation": "bounded-by" }, { "target": "failure-particle-therapy-comparator-and-accrual", "relation": "exposed-by" }, { "target": "hypothesis-particle-therapy-existing-data-gate", "relation": "tested-by" }, { "target": "trial-pdac-particle-therapy-action-map", "relation": "mapped-by" }, { "target": "claim-pdac-death-can-be-local-or-metastatic", "relation": "depends-on" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "moderate", "randomised": "none", "independentReplication": "moderate", "sampleSize": "Several uncontrolled cohorts from 15 to 84 people; no randomized comparative participant", "effect": "Incremental overall survival, quality-adjusted survival, severe-harm reduction and value are unknown", "limits": [ "No randomized comparison", "Distant failure dominates several cohorts", "Dose definitions differ", "Pancreatic and bowel motion change particle range", "Travel selection and treatment delay are unmeasured", "Current proton and carbon registries lack results", "Facility cost and opportunity cost are not joined to patient outcomes" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The radiation-physics advantages are real and measurable." }, "animalEvidence": { "rating": "moderate", "reason": "Local biological effect is plausible, while systemic patient benefit requires human comparison." }, "inVitroEvidence": { "rating": "strong", "reason": "Carbon ions produce dense ionization and complex damage in cell systems." }, "reproducibility": { "rating": "weak", "reason": "Local-control direction repeats, but incremental clinical benefit has never been identified." }, "commercialConflicts": { "rating": "moderate", "reason": "Treatment centres have infrastructure and referral incentives even in academic studies." }, "recency": { "rating": "current", "reason": "The unknown incorporates 2026 carbon-ion outcomes and registry updates." } } }, "layout": { "x": 658, "y": 582 }, "updated": "2026-09-15" }, { "id": "unknown-pdac-biliary-drainage-route", "title": "Which obstructed PDAC patient needs drainage, which route, and who owns treatment continuity?", "shortTitle": "Biliary route rule", "type": "unknown", "status": "supported", "scope": [ "pdac", "cholangiocarcinoma", "transfer-unproven" ], "summary": "The decision is a sequence—prompt surgery versus drainage, metal-stent design, first route, rescue and follow-up—not a single device comparison.", "content": "Decision: prevent cholangitis, reintervention and cancer-treatment loss without exposing prompt-surgery patients to an unnecessary procedure. Keep five choices separate. First, immediate surgery versus drainage: cholangitis requires urgent relief; severe symptoms, neoadjuvant treatment and unavoidable delay support drainage; routine delay does not. Second, when drainage must last through neoadjuvant treatment, metal rather than plastic has the stronger patency evidence. Third, covered versus uncovered metal stents have similar overall sustained drainage but different migration, ingrowth and cholecystitis routes. Fourth, ERCP remains the common first route, while randomized malignant-obstruction evidence makes EUS-guided drainage a credible expert-centre alternative and guidelines favour it over percutaneous rescue after failed ERCP when expertise exists. Fifth, insertion is not completion: assign an owner for bilirubin recovery, fever, antibiotic use, scheduled treatment, fill or device record, reintervention and surgery. The unresolved part is the net route in current PDAC care, because trials mix resectable, palliative and periampullary populations, and the 2026 surgery comparison used a 15-point noninferiority margin. Resolve it with a cluster-randomized route-completion study, not another simple metal-versus-plastic trial. Pass if at least 90% have a documented indication and plan within 48 hours, metal-stent use reaches at least 90% when drainage must span neoadjuvant treatment, and cholangitis or obstruction causing treatment interruption falls from about 16% to 10%. Stop for serious procedure harm up five points, 30-day death imbalance, surgery delayed more than seven additional days, or no improvement in treatment delivered.", "sourceIds": [ "van-der-gaag-2010-pbd-rct", "costamagna-2026-sems-early-surgery-rct", "tamura-2021-fcsems-plastic-rct", "seo-2019-covered-uncovered-sems-rct", "lyu-2023-metal-plastic-nat-meta", "vehvilainen-2024-cholangitis-nat-interruption", "chen-2023-element-biliary-rct", "barbosa-2024-eus-ercp-meta", "esge-2018-biliary-stenting-guideline", "asge-2024-therapeutic-eus-guideline" ], "links": [ { "target": "unknowns", "relation": "part-of" }, { "target": "trial-pdac-biliary-drainage-action-map", "relation": "could-resolve" }, { "target": "hypothesis-pdac-biliary-route-completion", "relation": "could-resolve" }, { "target": "failure-routine-plastic-pbd", "relation": "constrained-by" }, { "target": "failure-biliary-stent-treatment-interruption", "relation": "constrained-by" } ], "evidence": { "humanProspective": "strong for components; weak for the full route", "humanRetrospective": "moderate", "randomised": "strong for components; none for the complete rule", "independentReplication": "none for the complete rule", "sampleSize": "Component trials n=202, 284, 119, 144 and 577 pooled; generated implementation rule", "effect": "Drainage and device boundaries partly known; whole-route treatment completion unresolved", "limits": [ "Generated thresholds", "Mixed cancer states", "Expert-centre EUS evidence", "Local surgery delays", "Rare serious harms", "Device availability" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The rule removes an unnecessary procedure and makes necessary drainage durable enough to protect treatment." }, "animalEvidence": { "rating": "not-applicable", "reason": "Patient procedures and route completion directly test the decision." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Laboratory patency cannot establish infection, delay or treatment delivery." }, "reproducibility": { "rating": "none", "reason": "No study has reproduced the full indication-to-surgery route." }, "commercialConflicts": { "rating": "high", "reason": "Metal and lumen-apposing stent trials involve manufacturers and named devices." }, "recency": { "rating": "current", "reason": "The route uses a 2026 randomized trial and current procedural guidance." } } }, "updated": "2026-09-14" }, { "id": "unknown-pdac-exercise-net-benefit-and-selector", "title": "Which pancreatic patients gain useful days from a defined exercise or rehabilitation route?", "shortTitle": "Exercise selector unknown", "type": "unknown", "status": "open", "scope": [ "pdac", "supportive-care", "surgery", "exercise", "quality-of-life" ], "summary": "Broad programmes change some movement measures, but the useful patient state, active component and effect on treatment completion remain unknown.", "content": "Resolve three questions from existing randomized records. Selector: does low baseline walk distance, strength, recent decline, age, cachexia or planned surgery identify a group with more capacity to gain, or merely a group less able to complete the programme? Component: are supervised strength, aerobic work, nutrition, pancreatic enzymes, breathing work or regular contact responsible for any effect? Patient return: does the route add quality-weighted days alive and at home, deliver more planned chemotherapy, reduce major complications or preserve independence? APACaP's primary failure and secondary longitudinal signal make death-aware missing-data analysis mandatory. PancFit's similar aerobic exposure between arms makes actual received activity more useful for mechanism but not causal benefit. HY-PANC plans ten participants and lean mass as its main outcome; it can show feasibility, not solve patient benefit. A new broad exercise trial has low value until participant-level data identify a reproducible state and outcome.", "sourceIds": [ "neuzillet-2023-apacap-rct", "ngohuang-2023-pancfit-rct", "ausania-2019-pancreatic-prehab-rct", "steindorf-2019-pancreatic-exercise-rct", "zarate-rodriguez-2023-pancreatectomy-steps-rct", "mizuno-2024-hpb-prehab-rct", "li-2024-pancreatic-early-mobilization-rct", "clinicaltrials-nct07470268-hypanc" ], "links": [ { "target": "unknowns", "relation": "part-of" }, { "target": "failure-pdac-exercise-endpoint-and-attrition", "relation": "learns-from" }, { "target": "unknown-cachexia-treatment-utility", "relation": "coordinates-with" }, { "target": "hypothesis-pdac-exercise-existing-data-gate", "relation": "tested-by" }, { "target": "trial-pdac-exercise-action-map", "relation": "mapped-by" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "moderate", "randomised": "strong — several comparisons exist but do not resolve the selector", "independentReplication": "weak — no repeated low-function treatment interaction exists", "sampleSize": "Several randomized cohorts totaling more than 900 assignments, without a harmonized participant-level selector analysis", "effect": "Decision-critical missing subgroup, component and useful-life evidence", "limits": [ "Heterogeneous programmes", "Different stages", "Missing questionnaires", "Post-treatment subgroup bias", "No locked selector" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Functional reserve can affect treatment tolerance and independence." }, "animalEvidence": { "rating": "not-applicable", "reason": "The selector and useful-life effect require patient data." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "No cell system can determine programme adherence or quality-adjusted time." }, "reproducibility": { "rating": "weak", "reason": "A patient state with repeated causal benefit has not been identified." }, "commercialConflicts": { "rating": "low", "reason": "The candidate intervention does not require a proprietary platform." }, "recency": { "rating": "current", "reason": "The current field includes a 2026 ten-person live protocol." } } }, "layout": { "x": 2400, "y": 1600 }, "updated": "2026-09-15" }, { "id": "unknown-pdac-fap-radioligand-treatment-effect", "title": "Unknown: which FAP radiation dose improves useful life in PDAC?", "shortTitle": "FAP dose and benefit", "type": "unknown", "status": "unknown", "scope": [ "pdac", "cholangiocarcinoma", "pan-cancer", "transfer-unproven" ], "summary": "The field lacks both a PDAC tumour dose-response curve and a controlled estimate of treatment benefit.", "content": "Decision: add FAP-targeted radiation to pancreatic-cancer treatment, or keep standard systemic treatment alone. Two unknowns must be closed in order. Physics unknown: for each lesion, how do scan uptake, residence time and absorbed Gy relate, and what tumour dose predicts control while kidney, marrow, bowel and healing tissue remain safe? LuMIERE reports normal-organ dose but withholds tumour-lesion dosimetry from its phase 1 paper. Its scan threshold selected visible disease but was too small and varied to link uptake with response. The 3BP-3940 report gives one pancreatic primary at 2.2 Gy after one cycle, not a dose-response curve. Clinical unknown: compared with the same standard treatment in the same disease state, does the radioligand improve overall survival or quality-adjusted survival? LuMIERE phase 2 is non-randomized and uses objective response as its main efficacy endpoint. Its monotherapy and first-line modified-FOLFIRINOX combination groups can establish a PDAC signal and safety, but cannot isolate the added treatment effect. A valid lesion model does not prove patient benefit. NCT07229768 plans four people and can only check delivery. FRONTIER terminated for a business decision and posts no result. Required result: first recover participant- and lesion-level LuMIERE and FRONTIER data, including every failed screen. Then require a reproduced PDAC dose-response and a controlled trial of radioligand plus standard care versus the same standard care, measured from random allocation through death, quality of life, renal and marrow injury and later treatment access. Until then, the honest answer is that the drug reaches selected people, while its useful dose and net benefit remain unknown.", "sourceIds": [ "mcconathy-2026-lumiere-phase1", "clinicaltrials-nct04939610-lumiere", "baum-2026-3bp3940", "clinicaltrials-nct05432193-frontier", "clinicaltrials-nct07229768-ctr-fapi", "baum-2022-fap2286-firsthuman" ], "links": [ { "target": "unknowns", "relation": "part-of" }, { "target": "claim-fap-radioligand-feasible-pdac-efficacy-unproven", "relation": "bounded-by" }, { "target": "failure-fapi-scan-dose-benefit-collapse", "relation": "caused-by" }, { "target": "trial-pdac-fap-radioligand-action-map", "relation": "resolved-by" }, { "target": "hypothesis-fapi-theranostic-pdac", "relation": "tested-by" } ], "evidence": { "humanProspective": "moderate for delivery and safety, weak for PDAC dose-response", "humanRetrospective": "moderate for feasibility", "randomised": "none", "independentReplication": "none for PDAC benefit", "sampleSize": "Fourteen pancreatic patients across the 2022 and 2026 FAP-2286 reports, plus one published pancreatic dosimetry example for 3BP-3940", "effect": "No controlled PDAC survival or quality-of-life estimate and no reported pancreatic partial response in LuMIERE phase 1", "limits": [ "Patient-level data access not confirmed", "Agent and isotope properties differ", "No validated PDAC tumour-dose target", "Non-randomized response cannot isolate added benefit", "Late-line and first-line disease states are not exchangeable" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "Radiation can kill nearby cells, but the target sits mainly on support cells and may vary across lesions." }, "animalEvidence": { "rating": "moderate", "reason": "Models support delivery and killing but cannot set a safe, useful human PDAC dose." }, "inVitroEvidence": { "rating": "moderate", "reason": "Binding and radiation response do not include whole-body clearance or normal-organ exposure." }, "reproducibility": { "rating": "none", "reason": "No PDAC treatment benefit or lesion dose-response has independent human reproduction." }, "commercialConflicts": { "rating": "high", "reason": "Most current data come from agent developers or specialist treatment centres." }, "recency": { "rating": "current", "reason": "Includes registry status checked on 15 September 2026." } } }, "layout": { "x": 75, "y": 143 }, "updated": "2026-09-15" }, { "id": "unknown-pdac-germline-family-action-route", "title": "Which germline route gets a useful result to every patient and consenting at-risk relative?", "shortTitle": "Germline-to-family route", "type": "unknown", "status": "open", "scope": [ "pdac" ], "summary": "The unresolved decision is not whether inherited variants exist, but whom England should test and how the route reaches treatment and relatives without coercion or misuse of uncertain results.", "content": "Uncertainty: England lacks a prospective denominator showing pathogenic-variant yield and documented actions inside versus outside R367, while completed testing and family uptake remain incomplete even in specialist services. Decision affected: whether to commission all-patient testing for exocrine pancreatic cancer, how much genetics capacity to add, and whether consented direct family contact plus free home testing should replace reliance on a proband letter. Current range: patient completion is about 38%-82% in several real services, 97.4% in one prospective point-of-care cohort, and 31% of informed first-degree relatives tested in that cohort. GENERATE shows 90% completion among motivated relatives already enrolled; GIFT shows only 19.2% of all reported relatives were invited. Resolving experiment: an English all-comer prospective route with R367-stratified yield and action, followed by family-cluster randomization of standard letters versus consented direct contact, multilingual navigation and no-cost home testing. Required denominators are every eligible patient, every pathogenic result, every eligible adult relative, every invitation, every returned test and every gene-specific action. Flip rule: expand routine eligibility and the direct-contact route only if excluded patients produce a prespecified positive net benefit and relative testing rises without privacy, distress or inequality harm. Retain or narrow the route if actionable yield outside R367 is too low, variants of uncertain significance drive care, results arrive after treatment decisions, or gains concentrate in already well-served groups.", "sourceIds": [ "nhs-england-2026-r367", "llach-2024-universal-germline", "chittenden-2021-systematic-germline", "walker-2021-embedded-germline-station", "wang-2022-oncology-germline-cascade", "rodriguez-2024-generate", "sinan-2025-pdac-germline-implementation", "katz-2026-gift-cascade", "grewal-2026-pancan-germline-survey" ], "links": [ { "target": "unknowns", "relation": "part-of" }, { "target": "claim-universal-germline-testing-needs-action", "relation": "constrained-by" }, { "target": "failure-pdac-germline-route-dropout", "relation": "caused-by" }, { "target": "system-england-r367-germline-criteria", "relation": "could-change" }, { "target": "trial-pdac-germline-cascade-action-map", "relation": "mapped-by" }, { "target": "hypothesis-pdac-germline-family-route-completion", "relation": "resolved-by" } ], "evidence": { "humanProspective": "moderate for component delivery", "humanRetrospective": "strong for service loss", "randomised": "moderate for relative-facing components", "independentReplication": "none for the complete English route", "sampleSize": "Unresolved system decision informed by cohorts and family trials", "effect": "No complete-route effect estimate", "limits": [ "English all-comer data absent", "Country-specific law and commissioning", "Gene-specific actions differ", "Family denominator often missing", "Long-term health outcomes require follow-up" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "A known inherited variant can guide distinct care for a patient and blood relatives." }, "animalEvidence": { "rating": "not-applicable", "reason": "The uncertainty concerns human eligibility, consent and delivery." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Assay performance does not resolve reach or clinical action." }, "reproducibility": { "rating": "none", "reason": "The proposed whole route has not been tested in England." }, "commercialConflicts": { "rating": "low", "reason": "The question can be answered through NHS laboratories with product-neutral analysis." }, "recency": { "rating": "current", "reason": "The uncertainty uses the current 2026 NHS rule and 2026 family evidence." } } }, "updated": "2026-09-14" }, { "id": "unknown-pdac-nerve-ablation-net-benefit", "title": "Who gains useful time from pancreatic pain-nerve ablation?", "shortTitle": "Nerve-ablation net benefit unknown", "type": "unknown", "status": "unresolved", "scope": [ "pdac", "pain", "palliative-care", "celiac-plexus", "quality-of-life" ], "summary": "The unresolved decision joins pain interference, opioid burden, function, treatment delivery, complications and survival for each exact nerve target.", "content": "Known quantities. Alcohol plexus treatment can lower pain in some people. Modern opioids can also lower it. Direct ganglion, plexus, splanchnic, heat and radiation routes are anatomically and physically different. Unknown quantities. Which baseline pain pattern predicts benefit? Does visible tumour invasion of the nerve region matter? Does early treatment outperform waiting until standard pain medicines fail? How much improvement exceeds placebo, sedation, chemotherapy response and ordinary dose adjustment? Does opioid reduction restore alertness, bowel function and activity, or simply replace one burden with diarrhoea, low blood pressure or procedure pain? Why did two randomized forms of more aggressive nerve destruction show worse survival? Were those chance findings, disease imbalances, treatment effects or consequences of altered autonomic function? How many people die or become unable to report before each pain visit? Can heat or radiation improve duration without raising bowel, vascular or systemic-treatment harm? Resolve each technique separately with daily pain interference, oral-morphine-equivalent dose, constipation, sedation, food intake, function, quality of life, hospital-free days, systemic-treatment delivery, every randomized death and patient preference. A two-point pain change with stable survival can be valuable palliation; a lower score among survivors is not enough.", "sourceIds": [ "wyse-2011-early-eus-cpn-rct", "kanno-2020-eus-cpn-modern-opioid-rct", "levy-2019-celiac-ganglia-versus-plexus-rct", "dong-2021-splanchnic-neurolysis-rct", "bang-2019-celiac-rfa-versus-cpn-rct", "lawrence-2024-celiac-radiosurgery-phase2", "clinicaltrials-nct06160323-upfront-cgn-cpn", "clinicaltrials-nct07689955-celiac-radiotherapy" ], "links": [ { "target": "unknowns", "relation": "part-of" }, { "target": "claim-pdac-nerve-ablation-pain-benefit-inconsistent", "relation": "bounded-by" }, { "target": "failure-pdac-pain-death-and-evaluable-denominators", "relation": "exposed-by" }, { "target": "hypothesis-pdac-nerve-ablation-existing-trial-safety-gate", "relation": "tested-by" }, { "target": "trial-pdac-celiac-pain-action-map", "relation": "mapped-by" }, { "target": "unknown-cryoablation-randomized-pain-and-net-benefit", "relation": "compares-with" }, { "target": "unknown-hifu-pain-survival-and-system-transfer", "relation": "compares-with" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "weak", "randomised": "strong", "independentReplication": "weak for a single decision rule", "sampleSize": "Several small randomized studies; live early-treatment trial target n=94", "effect": "No stable selector for pain benefit without loss of quality-adjusted survival", "limits": [ "Target and technique heterogeneity", "Changing background pain care", "Short expected survival", "Competing death", "Uncertain survival mechanism", "Sparse patient preference", "Limited long-term safety" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The treatments interrupt known visceral pain pathways." }, "animalEvidence": { "rating": "not-applicable", "reason": "Human randomized evidence is available but clinically inconsistent." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "The unknown is person-level palliation and survival." }, "reproducibility": { "rating": "weak", "reason": "No exact technique and timing rule has a replicated net-benefit result." }, "commercialConflicts": { "rating": "low", "reason": "The route mostly uses generic inputs, though radiation and endoscopy capacity can shape adoption." }, "recency": { "rating": "current", "reason": "Two active 2026 programmes test early endoscopy and radiation." } } }, "layout": { "x": 1150, "y": 1074 }, "updated": "2026-09-15" }, { "id": "unknown-pdac-organoid-guided-treatment-utility", "title": "Unknown: does organoid-guided treatment improve useful life from the original patient denominator?", "shortTitle": "Organoid routing utility", "type": "unknown", "status": "unknown", "scope": [ "pdac", "transfer-unproven" ], "summary": "Feasibility, analytical agreement, response prediction and treatment-policy benefit are four different questions.", "content": "Decision: use an organoid result to choose or switch systemic treatment, or keep current clinical selection. Keep four functions separate. Feasibility asks whether a tumour-confirmed result returns before the decision for every consecutive eligible person. Analytical agreement asks whether two laboratories rank the same current regimens in the same order. Prediction asks whether treatment effect differs when a patient receives a regimen called sensitive rather than resistant; general prognosis does not answer that interaction. Utility asks whether the whole assay-guided policy improves survival or quality-adjusted life after failed cultures, unavailable drugs, default routes and non-adherence are retained. No published PDAC study closes all four. NCT04931381 and NCT04931394 are the relevant randomized advanced and postoperative trials, but both are status unknown, were last verified in July 2022, retain estimated May 2025 completion dates and post no results. NCT05351983 completed 60 people and posts no result; it tests feasibility, not routing. PROSPER-PANC is active but observational. ADOPT is recruiting about 25 people and UNITEPANC plans 92 for generation and 38 for organoid-chosen treatment; both are single-arm. Required result: a randomized policy comparison from enrollment, with default care for failed or late assays, overall survival and quality-adjusted survival, exact treatment access, biopsy harm, time, cost and every pipeline loss. Until then, use organoids to reject drug ideas and collect shadow results, not to promise better care.", "sourceIds": [ "clinicaltrials-nct04931381-organoid-advanced", "clinicaltrials-nct04931394-organoid-adjuvant", "clinicaltrials-nct05351983-organoid-screen", "clinicaltrials-nct05927298-prosper-panc", "clinicaltrials-nct06813079-adopt", "ettrich-2026-unitepanc", "knox-2025-pass01" ], "links": [ { "target": "unknowns", "relation": "part-of" }, { "target": "claim-pdac-organoid-correlation-not-routing-utility", "relation": "bounded-by" }, { "target": "failure-organoid-treatment-chain-attrition", "relation": "caused-by" }, { "target": "trial-pdac-organoid-action-map", "relation": "resolved-by" }, { "target": "hypothesis-organoid-guided-early-switch", "relation": "tested-by" } ], "evidence": { "humanProspective": "moderate — several prospective pipelines exist", "humanRetrospective": "moderate", "randomised": "weak — relevant randomized trials are stale and have no public result", "independentReplication": "none — no repeated randomized policy benefit", "sampleSize": "Randomized plans n=100 and n=200; observational or single-arm plans n=25, n=60, n=92 and n=200", "effect": "No published estimate of incremental survival or quality-adjusted-life benefit from an organoid-guided PDAC policy", "limits": [ "Two potentially decisive randomized trials may have unpublished data", "Registry status can be stale", "Assays and drug panels differ", "Treatment context ranges from first line to refractory and postoperative", "Default care after assay failure is often unclear" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Functional drug response could add information beyond anatomy and gene sequence." }, "animalEvidence": { "rating": "moderate", "reason": "Model fidelity supports the mechanism but not the clinical policy." }, "inVitroEvidence": { "rating": "strong", "reason": "Patient-derived cultures repeatedly show different drug-response patterns." }, "reproducibility": { "rating": "weak", "reason": "No independent randomized patient-benefit result is public." }, "commercialConflicts": { "rating": "moderate", "reason": "Platform owners and specialist laboratories may favour adoption before whole-route utility is shown." }, "recency": { "rating": "current", "reason": "Trial records and the 2026 UNITEPANC protocol were checked on 15 September 2026." } } }, "layout": { "x": 116, "y": 170 }, "updated": "2026-09-15" }, { "id": "unknown-pdac-palliative-route-net-benefit", "title": "Which early palliative-care route gives people with PDAC more good days?", "shortTitle": "Palliative route unknown", "type": "unknown", "status": "open", "scope": [ "pdac", "palliative-care", "quality-of-life", "care-delivery", "health-economics" ], "summary": "Quality support is justified, but the trigger, service components, coverage and quality-weighted time gained in a modern PDAC route are not known.", "content": "Unknown 1: should every unresectable PDAC patient receive a specialist visit, or should an automatic symptom, function, caregiver or hospital-use trigger prioritize scarce teams? Unknown 2: which parts add value: symptom prescribing, goals and treatment explanation, mental-health care, dietetics, exercise, home visits, remote follow-up or caregiver support? Unknown 3: what fraction of all eligible people are screened, referred, seen and followed before death or fitness loss? Unknown 4: does the route increase quality-weighted days alive and at home, not merely questionnaire scores among survivors? Unknown 5: does it reduce emergency use and chemotherapy near death without withholding beneficial treatment or shortening life? Unknown 6: what is the result of the completed-primary 250-person Danish randomized study? Unknown 7: do language, deprivation, age, geography and digital access change completion or benefit? Unknown 8: what is the cost per added quality-weighted day and per hospital-free day?", "sourceIds": [ "maltoni-2016-systematic-palliative-qol", "maltoni-2016-systematic-palliative-end-of-life", "woo-2019-pancreatobiliary-palliative-rct", "clinicaltrials-nct04632303-pdac-palliative", "sanders-2024-asco-palliative-guideline" ], "links": [ { "target": "unknowns", "relation": "part-of" }, { "target": "claim-pdac-palliative-care-improves-qol-not-survival", "relation": "qualifies" }, { "target": "failure-pdac-palliative-timing-and-analysis", "relation": "caused-by" }, { "target": "hypothesis-pdac-palliative-route-completion", "relation": "tested-by" }, { "target": "trial-pdac-palliative-care-action-map", "relation": "tracked-by" } ], "evidence": { "humanProspective": "strong but heterogeneous", "humanRetrospective": "moderate", "randomised": "moderate", "independentReplication": "moderate — mixed results", "sampleSize": "Published pancreatic randomized n=207; completed-primary Danish target n=250 without results", "effect": "Possible quality and care-use gain; unknown complete-route quality-weighted time and cost", "limits": [ "No common package", "No common endpoint", "Death and dropout", "Changing systemic therapy", "Workforce limits", "Caregiver outcomes sparse", "Danish data unavailable" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Direct management of symptoms and decisions can increase good days even without tumour change." }, "animalEvidence": { "rating": "not-applicable", "reason": "The question concerns patient experience and service delivery." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "No model system can substitute for patient-reported and care-use outcomes." }, "reproducibility": { "rating": "moderate", "reason": "Some patient-centred effects recur, but the route and size vary." }, "commercialConflicts": { "rating": "low", "reason": "The main uncertainty is public service design." }, "recency": { "rating": "current", "reason": "A completed-primary randomized study remains unread in 2026." } } }, "layout": { "x": 1258, "y": 1182 }, "updated": "2026-09-15" }, { "id": "unknown-pdac-perioperative-treatment-route", "title": "Which resectable PDAC patient should receive systemic treatment before surgery?", "shortTitle": "Treatment timing unknown", "type": "unknown", "status": "supported", "scope": [ "pdac" ], "summary": "Event-free survival favours treatment first across trials, but overall survival, the useful regimen and the patients who should avoid delay remain unresolved.", "content": "Decision: choose treatment first or surgery first for a person whose PDAC is technically resectable, without losing either effective systemic therapy or a curative operation. Keep four choices separate. 1. Anatomy: clearly resectable and borderline resectable disease have different operation risks and trial evidence. 2. Systemic regimen: PREOPANC used gemcitabine chemoradiotherapy, NORPACT-1 used four FOLFIRINOX cycles, and CISPD-1 used sequential gemcitabine/nab-paclitaxel then modified FOLFIRINOX. A timing result does not validate every regimen. 3. Local radiation: PREOPANC bundled timing with radiation, while PREOPANC-2 and A021501 did not show that adding or choosing radiation improved survival. 4. Delivery: biopsy, biliary control, start time, dose intensity, progression, surgery and all postoperative cycles decide what treatment is actually received. The 2026 nine-trial analysis found event-free survival HR .77, 95% CI .65-.90, and fewer noncurative explorations, RR .90, but overall survival remained inconclusive, HR .85, 95% CI .68-1.05. CISPD-1 improved event-free survival and reported a borderline overall-survival result at one expert Chinese centre; NORPACT-1 favoured surgery first. NCT04340141 A021806 is the decisive modern multicentre comparison: 358 actual participants, active but not recruiting, eight preoperative plus four postoperative mFOLFIRINOX cycles against surgery then twelve cycles, overall survival primary, estimated primary completion December 2028. Obtain complete intention-to-treat route data and wait for that survival result before another broad timing trial. A useful result must report overall and disease-free survival, unresectability, R0, distant failure, every chemotherapy cycle, severe harm and quality of life. Do not promote a universal treatment-first rule from R0 rate, node-negative pathology or event-free survival alone. Stop a follow-on route if it lowers the operation rate without improving overall survival, adds ten points of severe toxicity without useful-life gain, or leaves more than 15% of allocated patients without their assigned first treatment.", "sourceIds": [ "versteijne-2022-preopanc-longterm", "preopanc2-2025", "labori-2024-norpact1", "bai-2025-cispd1", "tan-2026-neoadjuvant-ipd-meta", "ghaneh-2023-espac5", "katz-2022-a021501", "clinicaltrials-nct04340141-a021806" ], "links": [ { "target": "unknowns", "relation": "part-of" }, { "target": "claim-neoadjuvant-therapy-improves-selection-not-cure", "relation": "constrains" }, { "target": "failure-norpact1-neoadjuvant-delivery", "relation": "learns-from" }, { "target": "trial-pdac-perioperative-sequencing-action-map", "relation": "mapped-by" }, { "target": "claim-pdac-surgery-is-a-complete-treatment-route", "relation": "part-of" }, { "target": "hypothesis-pdac-surgical-route-completion", "relation": "constrains" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong but conflicting", "independentReplication": "moderate for event-free survival; none for overall-survival rule", "sampleSize": "Nine-trial analysis n=1,131; PREOPANC n=246; NORPACT-1 n=140; CISPD-1 n=324; A021806 actual n=358", "effect": "Pooled EFS HR .77; pooled OS HR .85, 95% CI .68-1.05", "limits": [ "Regimens differ", "Resectable and borderline states differ", "Several trials are small", "One positive modern trial is single-centre", "A021806 survival pending", "No validated patient selector" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Occult systemic disease is common and treatment before operative recovery can increase early systemic exposure." }, "animalEvidence": { "rating": "not-applicable", "reason": "Human route timing and survival decide the question." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Laboratory response cannot value delay, surgery loss or treatment completion." }, "reproducibility": { "rating": "weak", "reason": "Event-free survival repeats across pooled trials, while overall-survival trials disagree." }, "commercialConflicts": { "rating": "low", "reason": "The central studies are academic strategy trials using established chemotherapy." }, "recency": { "rating": "current", "reason": "The 2025 Chinese phase 3 result, 2026 meta-analysis and September 2026 A021806 update are current." } } }, "updated": "2026-09-14" }, { "id": "unknown-pdac-pharmacogenetic-route-net-benefit", "title": "Does pancreatic pharmacogenetic dosing reduce severe harm without delay or lost cancer control?", "shortTitle": "Pancreatic dosing route unknown", "type": "unknown", "status": "open", "scope": [ "pdac", "treatment-safety", "pharmacogenetics", "care-delivery" ], "summary": "The safety mechanism is established, but the complete pancreatic path from eligibility through result, dose, escalation, toxicity and tumour control is not published.", "content": "The unknown is not whether DPD deficiency can cause catastrophic 5-FU exposure. It can. The unknown is the reliability and net effect of the pancreatic care route. Count all patients planned for FOLFIRINOX, NALIRIFOX, chemoradiation or another systemic fluoropyrimidine; record sample date, result date, first dose, exact variant or phenotype, initial dose, every later adjustment, treatment intensity, severe toxicity, admission, early death, response and survival. Run a separate UGT1A1 branch for irinotecan rather than calling either test positive or negative in the same way. Resolve four boundaries: the fraction treated before a result; residual severe toxicity among standard-panel negatives; whether broader or functional testing improves prediction across ancestries; and whether carrier dose escalation recovers exposure without renewed harm. A safe test that delays a time-critical regimen or silently lowers effective treatment is not a successful system.", "sourceIds": [ "ema-2020-dpd-testing", "mhra-2020-dpd-testing", "nhs-england-dpyd-implementation", "glewis-2024-pacific-pgx", "shirasu-2019-ugt1a1-folfirinox", "sharma-2019-ugt1a1-mfolfirinox", "heersche-2026-ugt1a1-guided-dosing", "chan-2024-dpyd-non-european-review" ], "links": [ { "target": "unknowns", "relation": "part-of" }, { "target": "failure-pharmacogenetic-panel-and-route-coverage", "relation": "learns-from" }, { "target": "hypothesis-pdac-pharmacogenetic-route-completion", "relation": "tested-by" }, { "target": "trial-pdac-pharmacogenetic-safety-action-map", "relation": "mapped-by" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "strong", "randomised": "weak — no pancreatic all-route randomized utility result", "independentReplication": "moderate — pharmacology repeats but route-level pancreatic outcomes do not", "sampleSize": "No published all-eligible English pancreatic denominator; component studies range from 19 pancreatic patients to 1,181 mixed-cancer enrollees", "effect": "Decision-critical missing route and net-benefit evidence", "limits": [ "Testing is already standard care", "Withholding established DPYD testing would be unsafe", "Rare genotypes", "Regimen changes confound historical comparisons", "Survival requires long follow-up" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Dose, clearance and toxicity form a direct measurable control loop." }, "animalEvidence": { "rating": "not-applicable", "reason": "Only complete human treatment routes resolve this question." }, "inVitroEvidence": { "rating": "moderate", "reason": "Functional assays can classify variants but not route timing or useful life." }, "reproducibility": { "rating": "weak", "reason": "No repeated pancreatic whole-route result was found." }, "commercialConflicts": { "rating": "low", "reason": "The question can be answered across NHS laboratories without one vendor." }, "recency": { "rating": "current", "reason": "Current NHS delivery and 2026 irinotecan evidence define the open question." } } }, "layout": { "x": 1960, "y": 1600 }, "updated": "2026-09-15" }, { "id": "unknown-pdac-sonoporation-treatment-effect", "title": "Does sonoporation add useful life on any fixed PDAC chemotherapy backbone?", "shortTitle": "Sonoporation contribution", "type": "unknown", "status": "open", "scope": [ "pdac", "pan-cancer", "transfer-unproven" ], "summary": "Overall randomized replication is negative in an abstract; complete data must show whether any planned backbone interaction is real and whether the device actually increased tumour drug delivery.", "content": "Unknown. Does ultrasound-driven microbubble motion add overall or quality-adjusted survival when both groups receive the same modern chemotherapy, and is any effect restricted to gemcitabine/nab-paclitaxel? A first randomized trial was positive for overall survival, while the second trial's accepted abstract reports no overall PFS or OS difference and an incomplete gemcitabine subgroup. A second unknown sits inside the first: did the device increase active drug concentration in tumour, rather than only change ultrasound perfusion images? Decision affected. A valid fixed-backbone contribution could justify a low-cost physical delivery route; a failed interaction or no delivery measurement should end new clinical investment. Resolving work. First recover NCT04821284's locked protocol and analysis history, randomization list, all 120 planned or 97 reported participant records, stage, backbone, dose, device exposure, scans, toxicity, quality of life and event follow-up. Report why enrollment differs and preserve every randomized person in the intended-to-treat analysis. Estimate overall and stage- and backbone-stratified hazard ratios with 95% confidence intervals, then test the treatment-by-backbone interaction. Analyse the 13-person perfusion method separately: 10/13 = 76.9% directional agreement is exploratory, three were discordant, and clinical outcome was not blinded drug exposure. No inference from perfusion alone. If tissue sampling is clinically safe, quantify tumour and plasma drug in a nested timed subset; otherwise use a separately validated labelled-drug or pharmacokinetic delivery measure. Stop the branch if the overall result is null and the planned interaction fails, if the subgroup was selected after looking at outcomes, or if a confirmatory fixed-backbone result fails. Reopen only for the complete trial report, posted results, a prespecified interaction with uncertainty, or direct human tumour-delivery evidence.", "sourceIds": [ "wang-2023-sonochemotherapy-pdac", "clinicaltrials-nct04821284-sonoporation", "ius-2026-sonoporation-nct04821284-abstract", "adler-2026-sonoporation-perfusion" ], "links": [ { "target": "unknowns", "relation": "part-of" }, { "target": "claim-sonochemotherapy-randomized-signal", "relation": "qualifies" }, { "target": "failure-sonoporation-overall-replication-subgroup-rescue", "relation": "created-by" }, { "target": "trial-pdac-sonoporation-action-map", "relation": "resolved-by" }, { "target": "hypothesis-measured-fixed-backbone-sonoporation", "relation": "tested-by" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "none needed", "randomised": "moderate — two randomized trials, but the independent result is abstract-only", "independentReplication": "weak — overall effect not reproduced in the available abstract", "sampleSize": "First trial n=78 randomized; NCT04821284 n=97 reported enrolled against n=120 registry estimate; perfusion analysis n=13", "effect": "Overall PFS and OS P>0.36 in the second-trial abstract; subgroup difference lacks uncertainty and interaction", "limits": [ "No complete second-trial report", "No posted results", "Unknown missing-person flow", "No direct tumour-drug measurement", "Two disease stages and two backbones can create effect mixing" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "The physical mechanism is plausible but does not establish net drug delivery or survival." }, "animalEvidence": { "rating": "moderate", "reason": "Models can guide exposure measures, not settle the clinical contribution." }, "inVitroEvidence": { "rating": "weak", "reason": "The relevant system includes vessels, acoustic field, bubbles and whole-body drug clearance." }, "reproducibility": { "rating": "weak", "reason": "The overall randomized survival signal did not reproduce." }, "commercialConflicts": { "rating": "moderate", "reason": "Device and contrast-agent adoption value makes independent analysis important." }, "recency": { "rating": "current", "reason": "The uncertainty is defined by the 2026 abstract and current registry." } } }, "layout": { "x": 150, "y": 137 }, "updated": "2026-09-15" }, { "id": "unknown-pdac-staging-laparoscopy-rule", "title": "Unknown: who needs PDAC staging laparoscopy, when, and what should each result change?", "shortTitle": "PDAC staging-laparoscopy rule", "type": "unknown", "status": "supported", "scope": [ "pdac" ], "summary": "No validated rule joins timing, risk, visible spread, cytology, delay and patient benefit from first potentially treatable scan through operation.", "content": "Class: missing action evidence. Keep four decisions separate. First, an early separate camera examination before chemotherapy can redirect a falsely localized case, but adds another anaesthetic and may delay treatment. Second, a same-session examination immediately before the planned open operation can prevent the large incision if hidden spread is seen, but cannot recover treatment time already spent. Third, a visible lesion confirmed by biopsy establishes metastatic disease more directly than positive wash cytology alone. Fourth, a low-risk omission rule must miss very few people: PREOPANC-2 still found 2.8% hidden spread in its zero-factor group, with a wide 95% interval from 0.9% to 8.0%, while Mayo's zero-factor estimate was 1.6% in only 63 people. Resolving route: inside England's existing 23-centre surgical study, register every potentially curable scan and both planned staging times. For the first 500 intended operations, sites crossing to the route start every operation with a standard short inspection, biopsy suspicious lesions, collect wash fluid, and report direct view, pathology and cytology separately; current-care sites retain NICE-selected staging until crossover. Lock the risk rule before outcomes. Primary staging result is an open exploration avoided because biopsy confirmed spread. Patient results are treatment start, delay, procedure harm, days alive outside hospital, function, resection, recurrence and survival. Go if at least 95% of intended operations have complete staging fields, the route reduces non-therapeutic laparotomy due hidden spread by at least eight points, adds no more than 30 minutes median anaesthetic time, causes no more than 1.5% major procedure harm, and does not delay systemic treatment or surgery by more than three days. Stop or amend for major harm above 2%, a treatment delay above seven days in more than 5%, a false-positive denial of surgery, a false-negative hidden metastasis rate above 3% at open exploration, or any automatic no-surgery decision from cytology alone without a recorded specialist and patient review. A negative camera check cannot prove absence of microscopic systemic disease, and a route gain cannot by itself claim longer survival.", "sourceIds": [ "gudmundsdottir-2023-staging-laparoscopy", "van-dongen-2023-preopanc-staging", "rompen-2026-preopanc2-staging", "theijse-2024-nontherapeutic-laparotomy", "stoop-2026-global-staging-survey", "kikuchi-2026-staging-laparoscopy", "nice-2018-ng85-pancreatic" ], "links": [ { "target": "unknowns", "relation": "part-of" }, { "target": "claim-staging-laparoscopy-avoids-incision-not-proven-survival", "relation": "bounded-by" }, { "target": "failure-staging-yield-without-patient-utility", "relation": "exposed-by" }, { "target": "trial-pdac-occult-metastasis-action-map", "relation": "tested-by" }, { "target": "hypothesis-pdac-two-timepoint-staging-route", "relation": "resolved-by" }, { "target": "unknown-pdac-surgical-route-completion", "relation": "part-of" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "strong", "randomised": "weak", "independentReplication": "moderate", "sampleSize": "Modern relevant cohorts n=1,004, n=322, n=239, n=663 and n=102", "effect": "Known detection and incision-avoidance signal; unresolved timing, omission threshold and cytology-only action", "limits": [ "Generated route gates", "Cluster learning", "Rare low-risk events", "No randomized survival effect", "Cytology treatment uncertainty", "Current England use not measured" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Surface spread can be invisible to scans yet visible through a camera, while microscopic systemic disease remains beyond direct sight." }, "animalEvidence": { "rating": "not-applicable", "reason": "Human route timing, pathology and patient outcomes define the uncertainty." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Laboratory tests cannot determine the net value of another procedure or omitted operation." }, "reproducibility": { "rating": "moderate", "reason": "Hidden-spread yield repeats, but exact low-risk rules and resulting actions do not." }, "commercialConflicts": { "rating": "low", "reason": "The proposed record uses standard tools and independent analysis rather than a proprietary imaging product." }, "recency": { "rating": "current", "reason": "The uncertainty uses 2026 comparative, practice and live-study records." } } }, "updated": "2026-09-15" }, { "id": "unknown-pdac-surgical-route-completion", "title": "Which parts of the PDAC surgical route are avoidably losing patients?", "shortTitle": "Surgical-route loss", "type": "unknown", "status": "open", "scope": [ "pdac" ], "summary": "Current audit shows mortality and timely treatment receipt, but not the planned route, full-course completion or reason for each loss.", "content": "Class: observed whole-route gap with mixed biological, clinical and service causes. Five decisions must remain separate. 1. Access: did every potentially curable scan receive an expert review, and did travel, age or deprivation change the chance of reaching it? 2. Selection: was surgery, treatment before surgery or no operation chosen from a fixed anatomy and systemic-risk record? 3. Procedure: was the chosen open, laparoscopic or robotic operation supported by procedure-specific team proficiency and rescue capacity? 4. Recovery: which complication, nutrition, pancreatic-enzyme, fitness or handoff event prevented the next treatment? 5. Completion: how much of the planned systemic treatment was delivered, when, and why did it stop? England records 96.4% 90-day survival after Whipple and 62% chemotherapy or chemoradiotherapy within 14 weeks, but the second number is not full-course completion and neither starts from all potentially curable patients. Resolve this with a national first-scan denominator, a plan locked at specialist MDT, procedure and surgeon proficiency fields, 90-day complications and death, dose-level systemic-treatment data, progression, patient choice and fixed avoidable-loss adjudication. The action gate is at least 95% complete route data, timely postoperative treatment rising from 62% to at least 72%, no more than a one-point absolute increase in 90-day mortality, no more than five points more major complications, and no greater than ten-point gap by age, sex, ethnicity, deprivation, language, region or travel time after clinically justified differences are shown. Full-course completion must improve by at least ten points from the measured run-in baseline. Stop or amend if route recording falls below 90%, 90-day mortality exceeds 5%, an operation-rate drop exceeds five points without more progression or patient refusal, serious complications rise five points, or any inequality gap widens by more than ten points. A survival claim waits for recurrence and cause-specific death; a process gain alone is not cure.", "sourceIds": [ "npaca-2026-state-of-nation", "nhs-england-2024-hpb-pancreatic-service", "nice-2018-ng85-pancreatic", "coupland-2016-england-resection-volume", "latenstein-2021-centralization", "labori-2016-adjuvant-completion", "henry-2023-complications-adjuvant", "van-hilst-2019-leopard2", "jin-2026-portal" ], "links": [ { "target": "unknowns", "relation": "part-of" }, { "target": "claim-pdac-surgery-is-a-complete-treatment-route", "relation": "constrains" }, { "target": "system-england-pancreatic-surgery-network", "relation": "tests" }, { "target": "failure-pdac-postoperative-treatment-loss", "relation": "separates-causes-of" }, { "target": "failure-leopard2-laparoscopic-whipple-safety", "relation": "learns-from" }, { "target": "trial-pdac-surgical-route-action-map", "relation": "mapped-by" }, { "target": "hypothesis-pdac-surgical-route-completion", "relation": "tested-by" } ], "evidence": { "humanProspective": "moderate for audit fields", "humanRetrospective": "strong for associations", "randomised": "none for whole route", "independentReplication": "none for the complete English route", "sampleSize": "Current national and multi-centre records; proposed resolving study in linked hypothesis", "effect": "Unknown avoidable fraction within the observed 38% without timely postoperative treatment", "limits": [ "No present full-course denominator", "Reasons for non-treatment overlap", "Patient preference must not be labelled failure", "Treatment-before-surgery changes the route", "Centre comparisons remain confounded" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Each route step has a direct path to treatment loss or harm." }, "animalEvidence": { "rating": "not-applicable", "reason": "The uncertainty is about human care delivery." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Bench systems cannot resolve access, skill or handoff." }, "reproducibility": { "rating": "none", "reason": "No linked national record yet spans first expert review through full planned treatment with adjudicated loss." }, "commercialConflicts": { "rating": "moderate", "reason": "Robot makers can benefit from procedure adoption, so platform evaluation needs independent data and analysis." }, "recency": { "rating": "current", "reason": "The missing fields are visible in current 2026 audit and commissioning records." } } }, "updated": "2026-09-14" }, { "id": "unknown-pdac-vaccine-regimen-contribution", "title": "Can any PDAC vaccine regimen prevent recurrence, and which component would cause the benefit?", "shortTitle": "Vaccine contribution unknown", "type": "unknown", "status": "open", "scope": [ "pdac", "transfer-unproven" ], "summary": "The 260-person IMCODE003 trial can test its full RNA-vaccine regimen, but it cannot isolate the vaccine from atezolizumab because both are added together.", "content": "Keep three questions separate. First, does the complete regimen help? IMCODE003 compares autogene cevumeran plus atezolizumab plus mFOLFIRINOX with mFOLFIRINOX alone after surgery. Its all-assigned disease-free-survival result can answer that. Disease-free survival is the time until recurrence, a new cancer or death. Second, does the vaccine itself help? The experimental arm adds both the personalized RNA vaccine and atezolizumab, an antibody that releases an immune brake. A positive comparison would prove the package, not the fraction attributable to either added component. Third, who benefits? The 16-person phase 1 split patients into eight immune responders and eight non-responders after vaccination. That is useful mechanism evidence but not a treatment-selector proof. Manufacturing creates another denominator: how many consented patients supplied usable tumour, received a vaccine on time, completed doses and remained recurrence-free must be reported from the first screened person, not only from vaccinated people. IMCODE003 is active but no longer recruiting, lists an estimated 260 participants and has primary completion in January 2031. Until its all-assigned result, no immune assay, manufactured-product rate or response-defined subgroup closes the efficacy question.", "sourceIds": [ "rojass-2025-neoantigen-followup", "nct05968326-imcode003", "middleton-2014-telovac", "le-2019-eclipse", "elicio-2026-amplify7p-results" ], "links": [ { "target": "unknowns", "relation": "part-of" }, { "target": "claim-pdac-vaccine-immunity-without-proven-benefit", "relation": "bounded-by" }, { "target": "failure-pdac-vaccine-escalation-reversals", "relation": "learns-from" }, { "target": "hypothesis-autogene-cevumeran-existing-trial-gate", "relation": "resolved-by" }, { "target": "trial-pdac-vaccine-action-map", "relation": "resolved-by" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "moderate", "independentReplication": "none", "sampleSize": "Phase 1 n=16 vaccinated; current randomized phase 2 estimated n=260", "effect": "Regimen efficacy, component contribution and treatment selector remain distinct unknowns", "limits": [ "IMCODE003 has no result", "Enrollment remains labelled estimated", "Open-label design", "Vaccine and atezolizumab contributions are confounded by design", "Public registry does not list manufacturing success as an outcome" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The vaccine creates long-lived target-specific T cells in some patients." }, "animalEvidence": { "rating": "strong", "reason": "Models support vaccination at low disease burden but cannot estimate human recurrence benefit." }, "inVitroEvidence": { "rating": "strong", "reason": "Individual neoantigens and vaccine-induced T cells are measurable." }, "reproducibility": { "rating": "none", "reason": "No randomized PDAC efficacy result exists for this regimen." }, "commercialConflicts": { "rating": "high", "reason": "Genentech and BioNTech control the trial, product and much of the data route." }, "recency": { "rating": "current", "reason": "Registry status was verified in September 2026." } } }, "layout": { "x": 412, "y": 354 }, "updated": "2026-09-15" }, { "id": "unknown-pdac-vte-prophylaxis-rule", "title": "Which PDAC patients should receive which clot prophylaxis, when and for how long?", "shortTitle": "PDAC clot-prevention rule", "type": "unknown", "status": "supported", "scope": [ "pdac" ], "summary": "The drug-class effect is known; the open question is an early patient rule that joins clot risk, bleeding risk, interactions, burden, duration and delivery.", "content": "Decision: turn a conditional high-risk recommendation into a reliable offer without treating every patient alike. Pancreatic cancer automatically contributes two Khorana points, yet a 40,218-person validation found that the score-at-least-two threshold did not stratify risk within pancreatic or hepatobiliary cancer. Tumour site identifies a broad high-risk group, not the individual balance. A usable rule must be recorded before or at the first systemic-treatment cycle and include existing or screen-detected VTE, tumour invasion of stomach or bowel, recent bleeding or surgery, varices, platelet count, kidney and liver function, interacting medicines, route and cost preference, expected treatment duration and whether injections or an oral agent are feasible. It must keep four decisions separate: ambulatory prophylaxis during systemic treatment; postoperative timing; extended postoperative duration; and treatment of an established clot. Resolve the ambulatory gap with a cluster-randomized delivery study, not another drug-versus-nothing efficacy trial. Report symptomatic VTE, incidental VTE, VTE hospitalization and VTE death separately from major and clinically relevant bleeding; record offer, choice, fill, use, interruption and reason. A programme passes if at least 80% receive a documented decision within seven days, prophylaxis use rises by at least 30 points, symptomatic VTE falls by at least three points at six months and major bleeding does not rise by two points or more. Stop for a fatal-bleeding imbalance, no VTE reduction, or benefit confined to incidental scans. Kinisoquin, the unknown imPaCT-PRO result and surgical timing each remain separate questions.", "sourceIds": [ "ash-2023-primary-vte-prophylaxis-guide", "overvad-2022-khorana-validation", "heffley-2024-pdac-vte-uptake", "noor-2026-pdac-vte-implementation", "clinicaltrials-nct05245877-preposterous-pancreas", "clinicaltrials-nct06861088-catiq-p3", "clinicaltrials-nct05178628-impact-pro" ], "links": [ { "target": "unknowns", "relation": "part-of" }, { "target": "trial-pdac-vte-action-map", "relation": "could-resolve" }, { "target": "hypothesis-pdac-vte-route-completion", "relation": "could-resolve" }, { "target": "failure-pdac-vte-prophylaxis-delivery-gap", "relation": "constrained-by" }, { "target": "failure-impact-pro-status-unknown", "relation": "constrained-by" }, { "target": "claim-kinisoquin-phase2-marker-not-clinical-proof", "relation": "constrained-by" } ], "evidence": { "humanProspective": "strong for drug efficacy; weak for the generated care rule", "humanRetrospective": "moderate", "randomised": "strong for VTE reduction; none for the complete rule", "independentReplication": "none for the complete rule", "sampleSize": "Randomized pooled n=1,003; risk validation n=40,218; implementation cohort n=243; live studies estimated n=480 and n=800", "effect": "Prophylaxis reduces VTE; individual net-benefit and delivery rule unresolved", "limits": [ "Generated thresholds", "Bleeding events are uncommon", "Risk tools incompletely stratify PDAC", "Drug interactions", "Different care settings", "Patient burden and preferences" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The rule builds on a reproduced drug-class effect and explicit causes of bleeding and non-delivery." }, "animalEvidence": { "rating": "not-applicable", "reason": "Patient events and care delivery directly test the decision." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Coagulation markers cannot establish individual net benefit or adherence." }, "reproducibility": { "rating": "none", "reason": "No study has reproduced the full seven-day selection and delivery rule." }, "commercialConflicts": { "rating": "moderate", "reason": "Agent choice involves branded and generic drugs, while the proposed implementation analysis is independent." }, "recency": { "rating": "current", "reason": "The rule uses current guidance, a 2024 cohort and 2026 implementation and trial records." } } }, "layout": { "x": 158, "y": 82 }, "updated": "2026-09-14" }, { "id": "unknown-pdt-spatial-coverage-and-patient-benefit", "title": "Can pancreatic PDT cover enough viable tumour to improve a person's life?", "shortTitle": "PDT coverage and benefit unknown", "type": "unknown", "status": "unresolved", "scope": [ "pdac", "photodynamic-therapy", "dosimetry", "quality-of-life" ], "summary": "The unresolved chain runs from drug, light and oxygen coverage to whole-tumour control, symptoms, spread, survival and treatment burden.", "content": "Known quantities. A fibre can be placed into selected pancreatic tumours. Verteporfin plus 690 nm light and porfimer plus 630 nm light can create image-visible dead tissue. Verteporfin needs only a short period of light avoidance. Unknown quantities. What fraction of living tumour receives enough drug, photons and oxygen? Which imaging signal best estimates each field before treatment? Can multiple fibres enlarge coverage without overlap injury? Does the local dead zone remain controlled, or does viable tissue around it regrow? Does killing part of the primary tumour reduce pain, obstruction and hospital time? Does it delay distant spread or improve survival when systemic chemotherapy is held equal? Does adding pembrolizumab, a drug that releases an immune brake called PD-1, produce a body-wide response, and is PDT necessary for it? The active 24-person study can describe the combined package, not answer the last causal question. Resolve the engineering chain first with participant-level fibre geometry, measured light output, contrast perfusion, oxygen-sensitive imaging where feasible, two-day dead-zone maps, later local failure and every attempted and cancelled procedure. Resolve patient benefit only through random assignment on an identical systemic treatment backbone. Report all starters, not only people who complete illumination.", "sourceIds": [ "huggett-2014-verteporfin-pdt", "chandrasekhara-2021-eus-verteporfin-pdt", "clinicaltrials-nct03033225-vertpac02", "clinicaltrials-nct06381154-pdt-pembrolizumab", "dewitt-2019-eus-porfimer-pdt" ], "links": [ { "target": "unknowns", "relation": "part-of" }, { "target": "claim-pdt-local-ablation-without-benefit-proof", "relation": "bounded-by" }, { "target": "failure-pdt-geometry-selection-and-bundling", "relation": "exposed-by" }, { "target": "hypothesis-pdt-measured-spatial-gate", "relation": "tested-by" }, { "target": "trial-pdac-pdt-action-map", "relation": "mapped-by" }, { "target": "unknown-histotripsy-human-coverage-and-benefit", "relation": "compares-with" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "none", "randomised": "none", "independentReplication": "weak", "sampleSize": "Published pancreatic PDT n=35; completed VERTPAC-02 starters n=13; active combination n=24", "effect": "Local injury is real; whole-tumour coverage and patient-valued benefit remain unknown", "limits": [ "No complete spatial dose map", "No randomized efficacy comparison", "Different photosensitizers", "Small selected samples", "Systemic disease remains outside the light field", "Current combination cannot isolate PDT" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Local chemistry is established, but coverage and disease beyond the beam remain limiting." }, "animalEvidence": { "rating": "moderate", "reason": "Animal studies guide tissue dose but do not establish useful human coverage." }, "inVitroEvidence": { "rating": "strong", "reason": "Photochemical killing is established under controlled drug, light and oxygen." }, "reproducibility": { "rating": "weak", "reason": "No multi-centre controlled evidence connects coverage to patient outcomes." }, "commercialConflicts": { "rating": "moderate", "reason": "Access depends mainly on academic trial groups, with one drug-company collaborator in the older study." }, "recency": { "rating": "current", "reason": "The active combination study was updated in June 2026." } } }, "layout": { "x": 952, "y": 876 }, "updated": "2026-09-15" }, { "id": "unknown-pert-pdac-utility", "title": "Which PDAC patients gain function or treatment capacity from PERT, at what dose, and does that change survival?", "shortTitle": "PERT utility gap", "type": "unknown", "status": "open", "scope": [ "pdac", "pancreatic-all" ], "summary": "The useful unknown is a risk-to-dose-to-delivery rule, not whether enzymes digest fat in general.", "content": "Decision: prescribe broadly on a low threshold, or require a symptom, imaging or function-defined exocrine-insufficiency state and adjust dose? Existing randomized trials missed short weight outcomes. Cohorts report survival associations but cannot remove selection. PANCAX-3 shows that 29 of 30 evaluable patients can take at least half of a high planned dose, but it has no untreated control. NCT05642962 compares standard and high doses in only 17 stage-IV patients and has no no-PERT arm. PERTseverance is the best live causal test: an estimated 100 selected patients with unresectable head tumours, duct dilation, at least 5% weight loss and expected survival of six months receive immediate or three-month-delayed Creon. It remains recruiting after its estimated June 2026 primary date, has no result and uses weight as the primary endpoint. Resolve the unknown with the existing studies first. Require individual data on tumour site and duct obstruction, symptoms, faecal elastase or a declared function test, actual lipase units per meal, acid suppression, adherence, product gaps, stool and absorption response, function, chemotherapy dose delivered, quality of life and survival. A selected effect passes only if symptoms or absorption improve and at least ten percentage points more patients receive 80% of planned chemotherapy dose at twelve weeks, or quality-adjusted survival improves, without serious treatment-related harm. A weight-only result classifies PERT as nutrition or symptom care, not durable cancer control. If the live 100-person study does not answer treatment delivery, embed the missing fields in routine care before funding another efficacy trial.", "sourceIds": [ "woo-2016-pert-rct", "saito-2018-pert-rct", "clinicaltrials-nct04098237-pancax3", "clinicaltrials-nct05642962-pertzye-dose", "clinicaltrials-nct06099119-pertseverance", "picozzi-2025-pert-advanced-pdac", "dugic-2026-pert-mortality" ], "links": [ { "target": "unknowns", "relation": "part-of" }, { "target": "unknown-cachexia-treatment-utility", "relation": "part-of" }, { "target": "trial-pert-pdac-action-map", "relation": "could-resolve" }, { "target": "failure-pert-short-randomized-trials", "relation": "constrained-by" }, { "target": "failure-pert-prescribing-and-supply-gap", "relation": "constrained-by" }, { "target": "programme-500m-integrated-mission", "relation": "constrains" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "strong", "randomised": "moderate for short weight endpoints; weak for selected utility", "independentReplication": "none for a complete selected action rule", "sampleSize": "Completed randomized n=67 and n=88, single-arm n=36, dose trial n=17, live delayed-control trial estimated n=100, observational n=501 and n=1,247", "effect": "Maldigestion relief is plausible; selected effects on treatment delivery and survival remain unknown", "limits": [ "Different definitions and formulations", "Small trials", "Weight-centred endpoints", "Confounded survival cohorts", "Live study delayed", "Generated ten-point treatment-delivery gate" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "A measurable enzyme-deficient state should respond to adequate replacement, but cancer wasting can persist independently." }, "animalEvidence": { "rating": "not-applicable", "reason": "Human digestion, symptoms and treatment delivery are the direct test." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Bench enzyme activity cannot select patients or show useful life gained." }, "reproducibility": { "rating": "none", "reason": "No complete patient-selection, dose and clinical-utility rule has been reproduced." }, "commercialConflicts": { "rating": "moderate", "reason": "Several formulations and studies involve enzyme manufacturers, while the most useful live randomized trial is academic." }, "recency": { "rating": "current", "reason": "Trial statuses and observational evidence were checked through September 2026." } } }, "layout": { "x": 155, "y": 79 }, "updated": "2026-09-14" }, { "id": "unknown-pipac-added-survival-quality-and-systemic-control", "title": "Does PIPAC add good-quality life beyond systemic chemotherapy alone?", "shortTitle": "PIPAC added benefit unknown", "type": "unknown", "status": "unresolved", "scope": [ "pdac", "peritoneal-metastases", "pipac", "quality-of-life", "systemic-control" ], "summary": "Local biopsy change is known; the added effect on symptoms, hospital time, distant spread and survival is not.", "content": "Known quantities. PIPAC is technically repeatable in selected people, can produce regression in sampled peritoneal deposits and has a low reported rate of severe procedure events. The 2026 registry gives an all-starter denominator and honest discontinuation counts. Unknown quantities. How many screened people never reach the first laparoscopy? Among everyone assigned to a PIPAC strategy, what fraction receives one, two and three procedures? Does aerosol delivery add response beyond the same systemic drugs, or merely mark people whose cancer was already controlled? Does it relieve fluid buildup, bowel symptoms and pain enough to repay operations, anaesthesia, recovery and time off systemic treatment? Does it prevent obstruction? Does peritoneal control matter when liver, lung and other distant disease remain reachable only by systemic therapy? Are 30-day deaths and complications fully captured across centres? Does nab-paclitaxel outperform the older aerosol drugs after matching disease burden, prior treatment and calendar time? The decisive output is quality-adjusted survival from assignment, with hospital-free days and all-starter flow beside it, not biopsy regression among repeat-procedure survivors.", "sourceIds": [ "digiorgio-2026-pipac-pdac-registry", "clinicaltrials-nct05371223-nab-pipac", "digiorgio-2024-nab-pipac-protocol", "graversen-2017-pipac-pdac" ], "links": [ { "target": "unknowns", "relation": "part-of" }, { "target": "claim-pipac-local-delivery-signal-without-benefit-proof", "relation": "bounded-by" }, { "target": "failure-pipac-repeat-procedure-selection-and-bundling", "relation": "exposed-by" }, { "target": "hypothesis-pipac-existing-data-and-control-gate", "relation": "tested-by" }, { "target": "trial-pdac-pipac-action-map", "relation": "mapped-by" }, { "target": "evidence-quality", "relation": "supports" }, { "target": "unknown-regional-hyperthermia-added-benefit-modern-care", "relation": "compares-with" } ], "evidence": { "humanProspective": "weak", "humanRetrospective": "moderate", "randomised": "none in pancreatic cancer", "independentReplication": "none for added benefit", "sampleSize": "Registry n=156; current uncontrolled trial target n=38", "effect": "No causal estimate for added quality-adjusted survival, symptoms or systemic control", "limits": [ "Rare selected population", "Changing systemic standards", "Repeat operations", "Local and distant disease interact", "Current trial is stale", "Quality-of-life data not yet posted" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "Better surface drug exposure could control peritoneal deposits, but cannot treat distant disease." }, "animalEvidence": { "rating": "not-applicable", "reason": "The open question is net patient benefit." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Drug sensitivity does not price surgical burden or systemic failure." }, "reproducibility": { "rating": "none", "reason": "No controlled pancreatic estimate exists to reproduce." }, "commercialConflicts": { "rating": "moderate", "reason": "A specialist procedure ecosystem has incentives to expand on local-response signals." }, "recency": { "rating": "current", "reason": "The unknown uses the largest 2026 registry and September 2026 registry status." } } }, "layout": { "x": 1090, "y": 1014 }, "updated": "2026-09-15" }, { "id": "unknown-pnet-reclassification-actionability", "title": "Unknown: does repeated pNET classification improve outcomes or only change labels?", "shortTitle": "pNET reclassification utility", "type": "unknown", "status": "open", "scope": [ "pnet" ], "summary": "Progressive pNET often changes Ki-67, grade or receptor pattern, but no controlled study shows that acting on repeat biopsy and dual PET extends useful life.", "content": "Class: missing intervention evidence. Decision affected: whether every progressive metastatic pNET should receive repeat core biopsy, FDG PET and somatostatin-receptor PET before the next treatment. A 21-person prospective cohort found a clinically significant change in 13 people and altered treatment in seven, but had no comparator or outcome test. SEQTOR did not establish one best order for everolimus and chemotherapy, and it did not route patients using current grade or receptor state. Resolving experiment: randomize progressive pNET patients whose original tissue is older than 12 months to locked re-characterization with a fixed action table versus standard restaging; compare quality-adjusted time without progression or grade 3-4 toxicity, overall survival, biopsy harm, treatment delay and cost. Stratify by prior receptor-directed radiation and need for rapid shrinkage. Stop if adequate tissue is obtained in fewer than 75%, median treatment is delayed by more than 14 days, the action table changes treatment in fewer than 20%, or conditional power for the clinical endpoint falls below 20%. Until then, repeat classification is biologically plausible and often decision-changing, not proven beneficial.", "sourceIds": [ "mollazadegan-2025-pnet-recharacterization", "capdevila-2025-seqtor-pnet", "grillo-2016-gepnet-grade-metastasis" ], "links": [ { "target": "unknowns", "relation": "part-of" }, { "target": "claim-pnet-state-changes-at-progression", "relation": "depends-on" }, { "target": "claim-pnet-seqtor-no-sequence-winner", "relation": "depends-on" }, { "target": "model-pnet-causal-chain", "relation": "could-change" } ], "evidence": { "humanProspective": "moderate for state change; none for action benefit", "humanRetrospective": "moderate", "randomised": "strong for two treatment orders; none for reclassification routing", "independentReplication": "moderate for grade discordance", "sampleSize": "Prospective re-characterization n=21; randomized SEQTOR n=141; retrospective multi-sample GEP-NET n=60", "effect": "No settled clinical-utility estimate", "limits": [ "Rare disease", "Biopsy selection", "Mixed imaging and tissue changes", "Treatment sequence changes over time", "No validated action table" ] }, "layout": { "x": 34, "y": 103 }, "updated": "2026-09-14" }, { "id": "unknown-postresection-surveillance-utility", "title": "Does scheduled follow-up after PDAC resection improve survival or only move the diagnosis date?", "shortTitle": "Post-resection follow-up utility", "type": "unknown", "status": "open", "scope": [ "pdac" ], "summary": "The missing proof is whether scheduled CA19-9 and imaging lead to treatment that improves life enough to outweigh scans, false alarms, anxiety and cost.", "content": "Class: consistent observational association with a live randomized answer. The decision is not simply scan or no scan. Four choices must remain separate. 1. Schedule: symptoms alone, CA19-9, CT, or a joined rule. 2. Action: which local or systemic treatment follows an asymptomatic recurrence, and does low burden change its effect? 3. Patient state: recurrence pattern, fitness, prior treatment, CA19-9 secretion and willingness can change net benefit. 4. Harm: contrast injury, radiation, false-positive procedures, anxiety, travel, cost and earlier toxic treatment must be counted. RADAR-PANC directly tests clinical review, CA19-9 and CT chest/abdomen every three months for two years against non-standardized mainly symptom-led care, with overall survival primary and quality of life, worry, treatment and cost secondary. Its 306-person power model assumes median survival 29.4 versus 19.7 months after expected crossover, a very large effect inherited from observational data. The registry was last verified in September 2024, says enrolling by invitation, estimates October 2026 completion and posts no result. CIRCPAC separately tests ctDNA-directed scan intensity in 1,000 planned Danish patients through 2032. Resolve the near-term question by obtaining RADAR-PANC actual randomization, acceptance, crossover, event and outcome dates; preserve intention-to-treat; publish survival, quality of life, recurrence treatment, false-positive work-up, radiation and cost. Do not start another broad scheduled-imaging trial before this result unless the study is underpowered, cannot recover its control contrast or does not complete. A positive gate is an overall-survival confidence interval excluding no benefit plus no clinically important quality-of-life loss and an acceptable cost per quality-adjusted life-year. A negative or imprecise survival result blocks routine three-month CT expansion; it does not block a new test tied to a different effective action. Stop any follow-on route for a ten-point quality-of-life loss, more than five additional invasive false-positive procedures per 100 patients, or earlier treatment without longer survival or useful life.", "sourceIds": [ "andel-2025-routine-imaging", "van-goor-2025-nationwide-imaging", "daamen-2024-radar-panc-protocol", "clinicaltrials-nct04875325-radar-panc", "clinicaltrials-nct05788744-circpac", "tai-2025-pan-asian-esmo-pancreatic" ], "links": [ { "target": "unknowns", "relation": "part-of" }, { "target": "claim-routine-imaging-after-pdac-resection-not-yet-proven", "relation": "constrains" }, { "target": "trial-pdac-postresection-surveillance-action-map", "relation": "mapped-by" }, { "target": "trial-pdac-ctdna-action-map", "relation": "overlaps" }, { "target": "unknown-ctdna-actionability", "relation": "distinguishes-from" }, { "target": "programme-100m-randomized-platform", "relation": "constrains" }, { "target": "lab-dpcg-radar-panc-network", "relation": "tested-by" } ], "evidence": { "humanProspective": "moderate for association", "humanRetrospective": "strong for association", "randomised": "none reported; live trial pending", "independentReplication": "none for utility", "sampleSize": "Observational n=333 and n=1,311; RADAR-PANC planned n=306; CIRCPAC planned n=1,000", "effect": "Unknown causal survival and quality-of-life effect", "limits": [ "Live randomized result pending", "Optimistic trial effect assumption", "Control crossover and intervention refusal", "Treatment at recurrence not fixed", "Different ctDNA question in CIRCPAC", "Imaging harms incompletely measured" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "Earlier treatment could help selected low-burden recurrence, but most recurrent PDAC may already be systemic and resistant." }, "animalEvidence": { "rating": "not-applicable", "reason": "Human surveillance utility is the question." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "No model can estimate the linked scan-to-treatment benefit and patient burden." }, "reproducibility": { "rating": "none", "reason": "No completed randomized trial has shown utility." }, "commercialConflicts": { "rating": "low", "reason": "RADAR-PANC is an academic care-strategy trial without a named product sponsor." }, "recency": { "rating": "current", "reason": "Live registry state was checked on 14 September 2026, before its estimated October 2026 completion." } } }, "updated": "2026-09-14" }, { "id": "unknown-quemliclustat-survival-contribution", "title": "Does quemliclustat add useful life to first-line chemotherapy in metastatic PDAC?", "shortTitle": "Quemliclustat survival unknown", "type": "unknown", "status": "open", "scope": [ "pdac", "transfer-unproven" ], "summary": "PRISM-1 has completed enrollment and directly randomizes the missing component, so no surrogate, subgroup or new cohort should replace its survival result.", "content": "The unknown is now narrow. ARC-8 established that quemliclustat can be given with gemcitabine and nab-paclitaxel, changes an adenosine-linked tumour measurement and has an encouraging external-control survival estimate. It did not randomize quemliclustat versus no quemliclustat. PRISM-1 does: the experimental group receives quemliclustat plus gemcitabine/nab-paclitaxel and the control receives placebo plus identical chemotherapy. It is quadruple masked, lists overall survival in every randomized patient as the primary endpoint, estimates 610 participants across 111 sites and allows qualified requests for participant-level data, protocol, analysis plan and clinical report. Arcus says enrollment completed in September 2025 and results are expected in the first half of 2027. The registry still labels enrollment estimated and primary completion November 2030, so result timing and final denominator must be checked at release. A baseline NR4A or CD73 subgroup cannot answer the all-comer question because neither selector has been validated. An on-treatment NR4A fall cannot answer selection at all. Wait for the direct comparison.", "sourceIds": [ "wainberg-2026-quemliclustat-arc8", "coveler-2024-oleclumab-pdac", "clinicaltrials-nct06608927-prism1", "arcus-2026-q2-10q" ], "links": [ { "target": "unknowns", "relation": "part-of" }, { "target": "claim-cd73-blockade-signal-without-concurrent-control", "relation": "bounded-by" }, { "target": "failure-adenosine-control-and-marker-pdac", "relation": "learns-from" }, { "target": "hypothesis-quemliclustat-existing-phase3-gate", "relation": "resolved-by" }, { "target": "trial-pdac-adenosine-action-map", "relation": "resolved-by" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "moderate", "randomised": "strong", "independentReplication": "weak", "sampleSize": "Existing masked phase 3 estimated n=610 at 111 sites; enrollment reported complete", "effect": "Direct added survival effect not yet reported", "limits": [ "Registry enrollment remains estimated", "Sponsor readout timing precedes registry completion estimate", "No public power assumptions or analysis plan located", "Participant-level access is conditional", "No validated treatment selector" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Human tissue supports target-pathway change." }, "animalEvidence": { "rating": "strong", "reason": "Models support the route but do not resolve survival." }, "inVitroEvidence": { "rating": "strong", "reason": "Drug-linked inhibition of adenosine-regulated expression is reproducible in cells." }, "reproducibility": { "rating": "weak", "reason": "Another CD73 product did not show overall-population benefit." }, "commercialConflicts": { "rating": "high", "reason": "Arcus controls the study while Gilead and Taiho hold commercial interests." }, "recency": { "rating": "current", "reason": "Enrollment and financial state were checked through August 2026." } } }, "layout": { "x": 452, "y": 384 }, "updated": "2026-09-15" }, { "id": "unknown-regional-hyperthermia-added-benefit-modern-care", "title": "Does measured regional heating add benefit to current pancreatic cancer care?", "shortTitle": "Regional heating value unknown", "type": "unknown", "status": "unresolved", "scope": [ "pdac", "locally-advanced", "resected", "regional-hyperthermia" ], "summary": "The reported randomized package missed its main endpoint; the cleaner heat-only randomized trial has a stale registry and no public result.", "content": "Known quantity. HEAT did not improve its main disease-free-survival endpoint, but its intervention cannot isolate heat because chemotherapy also changed. Unknown quantities. Did each intended treatment actually hold the pancreatic region at 40-43 degrees C for 60 minutes? Does achieved thermal dose predict biological effect without merely identifying fitter people? Does heat add survival or local-control benefit when systemic treatment, radiation, follow-up and later care are identical? Does it add pain, burns, bleeding, tube complications, treatment delay or travel burden? Does any result transfer to current chemotherapy? HEATPAC was designed as the cleaner test: after FOLFIRINOX, both randomized groups received the same gemcitabine chemoradiotherapy, while one also received six weekly heat sessions with real-time temperature measurement. Its protocol planned 86 people; ClinicalTrials.gov later listed 78. The registry still says status unknown, was last updated on 1 August 2017, estimated completion in 2021 and contains no results. Before funding a new cohort, recover the participant count, allocation, temperature traces, harms, scans, progression, surgery and survival from HEATPAC, plus the complete HEAT record. If the data no longer exist or cannot be independently analysed and published, record that loss as the result.", "sourceIds": [ "issels-2023-heat-rct", "clinicaltrials-nct01077427-heat", "eudract-2008-004802-14-heat", "datta-2017-heatpac-protocol", "clinicaltrials-nct02439593-heatpac" ], "links": [ { "target": "unknowns", "relation": "part-of" }, { "target": "claim-heat-regional-hyperthermia-primary-endpoint-negative", "relation": "bounded-by" }, { "target": "failure-heat-component-and-registry-drift", "relation": "exposed-by" }, { "target": "hypothesis-heat-existing-randomization-recovery-gate", "relation": "tested-by" }, { "target": "trial-pdac-regional-hyperthermia-action-map", "relation": "mapped-by" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "not-applicable", "randomised": "moderate", "independentReplication": "none", "sampleSize": "HEAT n=117 randomized; HEATPAC planned 86 in protocol and 78 in registry, actual number unknown", "effect": "No isolated and reported randomized estimate of regional heat in pancreatic cancer", "limits": [ "HEAT changed several components", "HEAT primary endpoint failed", "HEATPAC actual enrollment is unknown", "HEATPAC has no posted result", "HEATPAC registry has been stale since 2017", "Thermal delivery may vary between people and sessions", "Both treatment settings are older than current care" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "Mild heat can affect perfusion, oxygenation and treatment sensitivity, but a human net benefit remains unmeasured." }, "animalEvidence": { "rating": "moderate", "reason": "Laboratory and other-cancer evidence support a test, not adoption in pancreatic cancer." }, "inVitroEvidence": { "rating": "moderate", "reason": "Heat can enhance radiation and gemcitabine effects in model systems." }, "reproducibility": { "rating": "none", "reason": "The only reported randomized pancreatic trial did not isolate heat; the cleaner trial has no public result." }, "commercialConflicts": { "rating": "low", "reason": "The identified trials are academic; the main risks are incomplete records and investigator attachment." }, "recency": { "rating": "mixed", "reason": "The HEAT paper appeared in 2023, but both trial designs and the unresolved HEATPAC registry are old." } } }, "layout": { "x": 628, "y": 552 }, "updated": "2026-09-15" }, { "id": "unknown-regional-route-loss", "title": "Which step removes the most patients from a curative or life-extending HPB route in each region?", "shortTitle": "Regional route-loss gap", "type": "unknown", "status": "open", "scope": [ "pdac", "cholangiocarcinoma", "gallbladder", "hcc", "hepatobiliary", "pancreatic-all", "transfer-unproven" ], "summary": "Current African and Latin American cohorts show late stage and low treatment use, but cannot divide loss among symptom recognition, first contact, imaging, pathology, referral, eligibility, supply, cost and treatment completion.", "content": "Decision: which single repair should release intervention money at each site? Competing answers are earlier symptom recognition, risk surveillance, faster imaging, pathology completion, multidisciplinary referral, transport and cost support, operation capacity, systemic-treatment supply, or follow-up. Current records start at different points and omit many dates, so a low surgery rate cannot identify the cause. Resolve this with consecutive incident registration linked from district or population catchment to expert centre. For every patient, time-stamp first symptom or risk trigger, first contact, imaging, pathology, stage, referral, multidisciplinary decision, treatment recommendation, start, completion, reason for non-delivery and vital status. Classify a lost opportunity only after independent review using the standard available at that site and date. First gate: at least five African and five Latin American sites must capture at least 80% of eligible cases, keep required-field missingness at or below 20%, and follow at least 80% for twelve months. Second gate: no intervention trial releases unless one repairable step accounts for at least 15 percentage points of route loss at two sites or 100 patients, whichever is larger. A low observed loss, poor capture or disagreement between reviewers stops or redesigns the test.", "sourceIds": [ "gismalla-2026-africa-pancreatic-review", "sobnach-2024-ssa-hcc-systematic-review", "sobnach-2025-south-africa-hcc-disparities", "sobnach-2026-global-hcc-disparities", "da-fonseca-2024-latam-cca", "peixoto-2025-gaspar-results" ], "links": [ { "target": "unknowns", "relation": "part-of" }, { "target": "hypothesis-regional-route-completion", "relation": "could-resolve" }, { "target": "trial-regional-route-completion-action-map", "relation": "could-resolve" }, { "target": "programme-unconstrained-global-system", "relation": "blocks" } ], "evidence": { "humanProspective": "weak — GASPAR is a prospective and retrospective route but too selected for this question", "humanRetrospective": "moderate — several cohorts show the loss without locating it", "randomised": "none", "independentReplication": "moderate for the existence of route loss; none for the dominant repairable step", "sampleSize": "Evidence spans 26,850 pancreatic cases in a scoping review, 3,989 HCC cases in a systematic review, two HCC comparisons, 309 CCA cases and 79 GASPAR pancreatic cases", "effect": "The size of the lost route is visible; its timing and cause remain unresolved", "limits": [ "Different diseases and health systems", "Missing population denominators", "Referral selection", "Incompatible fields", "No common definition of avoidable loss", "No randomized repair" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Stage, performance and organ reserve narrow treatment eligibility over time." }, "animalEvidence": { "rating": "not-applicable", "reason": "The unknown is a human-system counterfactual." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Models cannot locate real-world route loss." }, "reproducibility": { "rating": "none", "reason": "No common prospective route audit has identified the same repairable step across regions." }, "commercialConflicts": { "rating": "moderate", "reason": "Most sources are academic, but the Latin American CCA and GASPAR routes include commercial centres or funders." }, "recency": { "rating": "current", "reason": "The evidence set includes 2026 African reports and the current Latin American cohorts." } } }, "updated": "2026-09-14" }, { "id": "unknown-rfa-human-mechanism-after-negative-trial", "title": "Did pancreatic RFA create any useful human immune or tissue signal?", "shortTitle": "RFA mechanism salvage unknown", "type": "unknown", "status": "unresolved", "scope": [ "pdac", "locally-advanced", "radiofrequency-ablation", "tissue-reuse" ], "summary": "The treatment route should stop, but stored PELICAN material could still test whether local destruction changed immunity or only added injury.", "content": "The clinical decision is settled for the studied route: do not add open RFA after short induction chemotherapy. A different question remains. Did heat-destroyed tumour release antigens and activate useful tumour-specific immunity, or did the operation mainly generate inflammation and suppressive repair? The paper reports an interaction by CA19-9 direction: among people whose marker decreased, the survival HR was 1.80, while the stable-or-increased group had HR 0.70, P=0.01 for interaction. This counterintuitive subgroup can arise from chance, missing marker data, competing surgery findings or altered chemotherapy delivery. It must not select RFA patients without reproduction and a credible mechanism. Resolve only from material and records already created: all stored blood and tumour, exact sampling time, inflammatory and tumour-specific immune state, chemotherapy interruption, occult metastasis and survival. Require the marker rule and immune analyses to be fixed before outcome access. If no paired material exists, if fewer than 60 paired participants are usable, if laboratories disagree or if any immune change does not associate with tumour control beyond operation injury and chemotherapy delivery, close the mechanism branch. A biological signal would justify a non-ablative laboratory hypothesis, not another patient RFA trial.", "sourceIds": [ "seelen-2026-pelican", "clinicaltrials-nct03690323-pelican" ], "links": [ { "target": "unknowns", "relation": "part-of" }, { "target": "claim-pelican-rfa-no-survival-and-more-harm", "relation": "bounded-by" }, { "target": "failure-rfa-selection-and-route-attrition", "relation": "constrained-by" }, { "target": "hypothesis-pelican-specimen-salvage-gate", "relation": "tested-by" }, { "target": "trial-pdac-rfa-action-map", "relation": "mapped-by" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong", "independentReplication": "none", "sampleSize": "PELICAN n=188 analysed; paired biospecimen count unknown", "effect": "Clinical net benefit is negative; immune or tissue mechanism is unknown", "limits": [ "Subgroup is secondary", "Marker data are incomplete", "Direction is counterintuitive", "Specimen existence and timing are unknown", "Surgery itself changes immunity", "Chemotherapy delivery can mediate outcome", "A mechanism cannot rescue the failed clinical route" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "Thermal cell death can release tumour material, but useful systemic immunity is unproved." }, "animalEvidence": { "rating": "moderate", "reason": "Model immune effects motivate reuse, not renewed patient exposure." }, "inVitroEvidence": { "rating": "moderate", "reason": "Heat-induced antigen release is plausible but incomplete as an immune route." }, "reproducibility": { "rating": "none", "reason": "No fixed human immune result or subgroup reproduction exists." }, "commercialConflicts": { "rating": "low", "reason": "Independent stored-sample work can be separated from device sales." }, "recency": { "rating": "current", "reason": "The question is created by the 2026 randomized result." } } }, "layout": { "x": 724, "y": 648 }, "updated": "2026-09-15" }, { "id": "unknown-selective-autophagy-pdac-contribution", "title": "Can selective, tumour-confirmed autophagy blockade add to RAS treatment in PDAC?", "shortTitle": "Selective autophagy unknown", "type": "unknown", "status": "open", "scope": [ "pdac", "pan-cancer", "transfer-unproven" ], "summary": "HCQ combinations failed, but no published study has shown that a selective inhibitor blocks recycling inside PDAC tumours while preserving an active RAS-drug dose and improving outcomes.", "content": "What is unknown. Does a selective inhibitor of the machinery that starts autophagy create enough blockade inside a patient's pancreatic tumour to make RAS-pathway treatment materially more effective without forcing the active drug dose down? Why it changes a decision. If the answer is no, the autophagy combination branch should be retired and capital moved to measured human resistance routes. If yes, it supplies a plausible escape-blocking partner for common-RAS or allele-specific treatment. Existing evidence cannot answer it. HCQ combinations produced no useful PDAC control, and the randomized combination halved the ERK-inhibitor dose. Blood-cell LC3B and NBR1 did not consistently show autophagy inhibition and are not tumour measurements. Early inlexisertib data showed reduced phosphorylated ATG14 in blood cells, not final PDAC efficacy; the 91-person study then terminated without posted results. Resolving work. First recover the complete inlexisertib dataset: every dose, cancer type, RAS alteration, paired tissue and blood measurement, scan, response, adverse event, treatment stop and missing denominator. Determine whether the planned pancreatic expansion opened and whether tumour material measured ULK target inhibition and completed recycling blockade. Only if those data show a tolerable exposure and tumour effect should a different molecule be considered. A future study must use paired tumour biopsies, quantify both direct target inhibition and material accumulating because recycling stopped, record the RAS drug's delivered dose, and keep nonresponders in the denominator. No inference from blood cells alone, stable disease alone or a company decision is allowed.", "sourceIds": [ "surana-2025-erk-hcq-pdac", "clinicaltrials-nct04892017-inlexisertib", "deciphera-2022-dcc3116-phase1", "clinicaltrials-nct05221320-ulixertinib-hcq", "clinicaltrials-nct04214418-mekiauto" ], "links": [ { "target": "unknowns", "relation": "part-of" }, { "target": "claim-hcq-combinations-fail-pdac-not-autophagy", "relation": "opened-by" }, { "target": "failure-metabolic-repurposing-pdac", "relation": "constrained-by" }, { "target": "trial-pdac-autophagy-action-map", "relation": "mapped-by" }, { "target": "hypothesis-tumour-verified-selective-autophagy-blockade", "relation": "tested-by" } ], "evidence": { "humanProspective": "weak for selective inhibition — early whole-tumour programme data are incomplete", "humanRetrospective": "none", "randomised": "none for a selective inhibitor; randomized evidence rejects the HCQ combination tested", "independentReplication": "none", "sampleSize": "Inlexisertib n=91 across cancers with no final posted result; early sponsor cutoff n=18, including about five pancreatic cancers", "effect": "Unknown; blood-cell target inhibition and stable disease do not establish tumour blockade or added benefit", "limits": [ "Final participant-level data are unavailable", "Earlier and final registry plans differ", "Blood and tumour pharmacology may differ", "Direct ULK inhibition may not block all autophagy routes", "Advanced cancers contain several simultaneous escape routes" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "The target is closer to autophagy initiation than HCQ, but the relevant tumour dependency remains unmeasured in people." }, "animalEvidence": { "rating": "moderate", "reason": "Combination activity exists in models and has not transported to a controlled patient result." }, "inVitroEvidence": { "rating": "weak", "reason": "Different model systems disagree on combination synergy." }, "reproducibility": { "rating": "weak", "reason": "No selective inhibitor has a repeated PDAC tumour-pharmacology and response result." }, "commercialConflicts": { "rating": "high", "reason": "The only selective-inhibitor human signal is sponsor-reported and the final dataset is not public." }, "recency": { "rating": "current", "reason": "Registry status checked through September 2026." } } }, "layout": { "x": 86, "y": 151 }, "updated": "2026-09-15" }, { "id": "unknown-sil204-human-target-delivery", "title": "Does dual-route SIL204 reduce mutant KRAS in both primary and distant human tumour?", "shortTitle": "SIL204 target delivery", "type": "unknown", "status": "open", "scope": [ "pdac", "colorectal", "pan-cancer", "transfer-unproven" ], "summary": "Plasma concentration, local injection and mouse activity do not establish cytosolic delivery, mutation-specific RNA cutting or multi-region tumour coverage in people.", "content": "Unknown. Does intratumour plus subcutaneous SIL204 reach the cytosol of enough PDAC cells to reduce the intended mutant KRAS messenger RNA and protein across the primary tumour and occult or visible distant disease? The route has at least five serial gates: drug leaves its depot, crosses tissue, enters tumour cells, escapes the endosome into cytosol, loads the RNA-cutting complex and removes the matched message. Plasma pharmacokinetics tests circulation, not the final four gates. A needle deposit at three primary-tumour sites does not establish coverage between sites. The first-generation implant had no measurable plasma siRNA and no published randomized tumour-wide knockdown result. Decision affected. Without direct human target evidence, a 403-person survival programme is testing delivery, biology and clinical benefit in one expensive black box. Resolving experiment. Before randomized expansion, use Segment 1 and no separate mission cohort. Predefine mutation-aware digital RNA and protein assays in paired samples from the three injected regions; use an accessible metastasis only when clinically safe. Retain every dose and sample failure. Go requires paired primary-tumour results in at least 80% of exposed participants, at least 50% median mutant-RNA reduction with concordant KRAS-protein reduction in at least 70%, agreement across at least two sampled regions in at least 70%, and systemic exposure above the locked active range without wild-type RAS suppression beyond the safety limit. In metastatic Segment 1 participants, lack of an accessible distant sample leaves systemic target delivery unknown; it is not counted as success. At least 75% must receive 80% or more of planned chemotherapy, and severe EUS or injection harm above 5% stops repeat local dosing. Independent laboratories must reproduce the assays. Publish the all-patient table before Segment 2. Stop the mission route if the sponsor will not share the necessary data, if target reduction fails, if spatial disagreement exceeds 30%, if systemic exposure is absent, or if delivery forces chemotherapy loss. Reopen for first-patient pharmacology, a revised public protocol, or independent human tumour data.", "sourceIds": [ "alagesan-2026-sig12d-loder", "ctis-2026-sil204-phase23", "silexion-2026-sil204-site-initiation", "silexion-2026-halfyear-filing" ], "links": [ { "target": "unknowns", "relation": "part-of" }, { "target": "claim-sil204-trial-scales-before-human-knockdown", "relation": "created-by" }, { "target": "failure-sig12d-loder-randomized-os-not-shown", "relation": "learns-from" }, { "target": "trial-sil204-action-map", "relation": "resolved-by" }, { "target": "hypothesis-multicompartment-kras-sirna-delivery", "relation": "tested-by" } ], "evidence": { "humanProspective": "weak", "humanRetrospective": "none", "randomised": "none for SIL204", "independentReplication": "none", "sampleSize": "Designed for the planned 15-21-person SIL204 safety and pharmacology segment", "effect": "Generated target and delivery gates; no human SIL204 measurement exists", "limits": [ "Thresholds are mission decisions", "Repeated tumour sampling may be unsafe", "A primary-tumour result cannot prove occult-metastasis coverage", "KRAS protein assays may mix malignant and normal cells", "Sponsor does not plan participant-level data sharing" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "The molecular cutting mechanism is established generally, while delivery through every human tissue barrier is not." }, "animalEvidence": { "rating": "moderate", "reason": "Sponsor models support both routes and require human reproduction." }, "inVitroEvidence": { "rating": "moderate", "reason": "Cell killing and RNA reduction support target activity after artificial access." }, "reproducibility": { "rating": "weak", "reason": "No human SIL204 target dataset exists." }, "commercialConflicts": { "rating": "high", "reason": "The sponsor controls the agent, trial, samples and stated no plan to share participant-level data." }, "recency": { "rating": "current", "reason": "The question is defined by the authorised 2026 protocol." } } }, "layout": { "x": 182, "y": 132 }, "updated": "2026-09-15" }, { "id": "unknown-small-pnet-surgery-selector", "title": "Unknown: which small non-functioning pNET should be removed?", "shortTitle": "Small pNET surgery selector", "type": "unknown", "status": "open", "scope": [ "pnet" ], "summary": "Short-term observation is often safe and surgery is measurably harmful, but current measurements do not predict lifetime spread well enough to choose every 1-2 cm tumour.", "content": "Class: missing future-risk and action selector. Decision affected: observe, ablate or resect a sporadic or MEN1-associated non-functioning pNET no larger than 2 cm, and choose the operation if resection is necessary. Evidence for observation: ASPEN and PANDORA reported no new distant metastasis during median 25 and 17 months, with only 9/406 and 2/76 crossing to surgery. Evidence for caution: ASPEN found four synchronous metastatic cases among 500 and 19 aggressive pathology findings among 94 selected operations; national and surgical registries find a minority of small tumours with nodes or adverse grade. Evidence for harm: severe complications were 13% in ASPEN surgery, while a 2026 series of 460 small-tumour resections reported 25.0% clinically relevant fistula, 14.8% major morbidity and 0.9% 90-day mortality. Current size, growth, Ki-67 from a small biopsy, duct dilation, FDG uptake, receptor imaging, MEN1 status and patient fitness each contain information, but none has been prospectively combined and compared with the actual procedure-specific harm. Resolve it with complete ASPEN and PANDORA follow-up, a prospective consecutive 1-2 cm cohort with locked imaging and tissue measurements, central pathology, every operation and crossover retained, and registry-linked metastasis, pancreatic function, quality of life and death. Then randomize the action among patients for whom both routes remain acceptable. Stop if model calibration differs by more than five percentage points across centres, post-resection variables drive the score, fewer than 85% retain outcome linkage, or the lower net-benefit bound does not exceed a size-plus-duct rule.", "sourceIds": [ "partelli-2022-aspen-interim", "clinicaltrials-nct03084770-aspen", "heidsma-2021-pandora-pnet", "maekawa-2026-small-pnet-surgery-morbidity", "nell-2018-men1-surgery-watchful-waiting", "van-vliembergen-2026-men1-mnet-assay" ], "links": [ { "target": "unknowns", "relation": "part-of" }, { "target": "unknown-pnet-reclassification-actionability", "relation": "distinct-from" }, { "target": "claim-small-nfpnet-surveillance-surgery-tradeoff", "relation": "supported-by" }, { "target": "trial-small-pnet-action-map", "relation": "mapped-by" }, { "target": "hypothesis-pnet-time-updated-surgery-rule", "relation": "tested-by" } ], "evidence": { "humanProspective": "moderate — short prospective observation and selected surgery cohorts", "humanRetrospective": "strong — national and multicentre surgery cohorts", "randomised": "none", "independentReplication": "moderate for the trade-off; none for a complete selector", "sampleSize": "ASPEN n=500; PANDORA n=76; small-tumour surgical harm n=460; DutchMEN1 management n=152", "effect": "No validated pre-operation rule estimates both metastatic risk and procedure-specific harm", "limits": [ "Low event rate", "Long disease course", "Tumour grade sampling error", "Post-surgery pathology leakage", "Sporadic and MEN1 disease differ", "Randomization may be difficult" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Tumour state and the required operation both plausibly determine the balance of future cancer harm and immediate treatment harm." }, "animalEvidence": { "rating": "weak", "reason": "Models can test progression biology but not a full human surgical decision." }, "inVitroEvidence": { "rating": "moderate", "reason": "Molecular candidate markers exist, but none has prospectively improved the action rule." }, "reproducibility": { "rating": "none", "reason": "No complete pre-operation selector has been independently validated with long-term outcomes." }, "commercialConflicts": { "rating": "moderate", "reason": "Imaging, assay, ablation and surgical routes create competing incentives around the decision." }, "recency": { "rating": "current", "reason": "The unknown incorporates evidence and surgical harm reported through August 2026." } } }, "updated": "2026-09-14" }, { "id": "unknown-smart-added-survival-and-surgery-harm", "title": "Does SMART add useful life, and what happens if surgery follows it?", "shortTitle": "SMART net benefit unknown", "type": "unknown", "status": "unresolved", "scope": [ "pdac", "locally-advanced", "borderline-resectable", "radiation" ], "summary": "Selected-cohort survival cannot separate radiation from induction response, surgery and later care; severe late harm and the surgical route need complete comparison.", "content": "Known quantity. Daily MRI and plan adaptation can deliver an ablative five-fraction course to selected pancreatic tumours with little definitely attributed acute severe gastrointestinal toxicity. Unknown quantities. Compared with continued systemic treatment, does SMART increase overall survival, days without treatment, pain control or quality-adjusted survival? Does it merely move local failure later while distant disease determines death? Which patients need elective regional coverage? How much chemotherapy is delayed or omitted? Among the 34.6% who underwent surgery after phase 2 SMART, did radiation cause or prevent vascular, gastrointestinal or healing complications? The safety report itself advised caution around surgery, especially vascular resection, because contribution to postoperative toxicity was unclear. Late grade 3 or worse toxicity ranges from 0% definitely related to 11.5% possibly related. Both numbers are true labels for different attribution rules; neither alone is a complete harm count. Operational uncertainty matters too. The phase 3 registry is stale and names a bankrupt sponsor's machine, while another randomized study recruited only nine. Resolve the question with a brand-independent performance specification, every screened and randomized denominator, identical systemic treatment, overall and quality-adjusted survival, chemotherapy delivery, hospital-free days, full adverse events without filtering by attribution, and operation-specific outcomes. Manufacturer continuity, service time and data export must be treated as trial variables rather than procurement footnotes.", "sourceIds": [ "parikh-2023-smart-phase2-safety", "chuong-2024-smart-phase2-survival", "clinicaltrials-nct03621644-smart", "clinicaltrials-nct05585554-lap-ablate", "clinicaltrials-nct05114213-maspac", "viewray-2023-chapter11-8k", "timmer-2024-crossfire" ], "links": [ { "target": "unknowns", "relation": "part-of" }, { "target": "claim-smart-delivery-and-safety-without-added-survival-proof", "relation": "bounded-by" }, { "target": "failure-smart-trial-and-vendor-dependency", "relation": "exposed-by" }, { "target": "hypothesis-smart-vendor-independent-randomized-gate", "relation": "tested-by" }, { "target": "trial-pdac-smart-action-map", "relation": "mapped-by" }, { "target": "hypothesis-randomized-lapc-resection-after-response", "relation": "interacts-with" }, { "target": "company-viewray", "relation": "constrained-by" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "weak", "independentReplication": "none", "sampleSize": "Phase 2 n=136 uncontrolled; LAP-ABLATE n=267 planned but not started; MASPAC n=9", "effect": "Added survival, quality-adjusted survival and postoperative net harm are unknown", "limits": [ "Strong induction-response selection", "No chemotherapy-only phase 2 control", "Later surgery and treatment vary", "Harm attribution changes reported rate", "Phase 3 has not started", "Named vendor entered bankruptcy", "Cross-machine equivalence is not established" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "Accurate high-dose local treatment can plausibly improve local control and symptoms." }, "animalEvidence": { "rating": "not-applicable", "reason": "The unresolved question is net human benefit and route harm." }, "inVitroEvidence": { "rating": "not-applicable", "reason": "Radiation sensitivity cannot resolve surgery, distant spread or quality of life." }, "reproducibility": { "rating": "none", "reason": "No randomized comparison against no local radiation exists." }, "commercialConflicts": { "rating": "high", "reason": "The main study and dormant phase 3 plan were sponsored by the machine vendor." }, "recency": { "rating": "current", "reason": "The record includes the 2025 MASPAC termination and current registry status." } } }, "layout": { "x": 688, "y": 612 }, "updated": "2026-09-15" }, { "id": "unknown-tigerpac-package-survival-and-delivery", "title": "Does the TIGeR-PaC replacement package extend life or preserve it with less total harm?", "shortTitle": "TIGeR-PaC package value", "type": "unknown", "status": "open", "scope": [ "pdac", "trial-delivery", "transfer-unproven" ], "summary": "The completed enrollment should answer package survival and harm, but cannot by itself prove greater active drug inside tumours or transfer to another medicine or cancer.", "content": "Decision affected: whether to use intra-arterial gemcitabine after induction for the narrow group who have not progressed and remain unable to have surgery. Required result fields are screened, inducted, progressed during induction, converted to surgery, randomized by arm, treated by arm, chemotherapy dose intensity, crossover, later treatment, death, progression, local versus distant first failure, hospital days, quality of life, neuropathy, low blood counts, infection, bleeding, clot, artery injury, ischemia and every serious event. The final randomized count must be reconciled with the protocol target of 114 and the registry estimate of 190 people including induction. The locked original and changed protocols, analysis plans, interim boundaries and regulator correspondence must show how the two interim looks affect the final P value. Two useful outcomes remain possible. The package may extend survival enough to justify the invasive route, or it may preserve survival while materially lowering systemic toxicity and improving quality-adjusted life. Neither is established. A separate mechanism question asks whether more active drug enters cancer cells. That needs tumour or local-vein sampling of gemcitabine and its active intracellular metabolites; peripheral blood and CA19-9 cannot answer it. Keep package adoption, catheter mechanism and transfer to other cancers as three different decisions.", "sourceIds": [ "clinicaltrials-nct03257033-tigerpac", "renovorx-2023-tigerpac-interim", "renovorx-2025-10k", "renovorx-2026-full-enrollment", "novelli-2026-tigerpac-pk", "hatoum-2024-tamp-rr1-rr2" ], "links": [ { "target": "unknowns", "relation": "part-of" }, { "target": "claim-tamp-blood-exposure-not-tumour-delivery", "relation": "limits" }, { "target": "failure-tigerpac-component-and-interim-inference", "relation": "caused-by" }, { "target": "hypothesis-tigerpac-existing-final-gate", "relation": "tested-by" }, { "target": "trial-pdac-intraarterial-delivery-action-map", "relation": "tracked-by" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "weak", "randomised": "strong", "independentReplication": "weak", "sampleSize": "Protocol target n=114 randomized and 86 deaths; registry estimate n=190 including induction; final randomized count not public", "effect": "Final package survival, total harm, quality of life and local active-drug delivery remain unresolved", "limits": [ "No final result", "Final assigned count not public", "Open-label treatment", "Components change together", "Induction selects non-progressors", "No direct human tumour drug measurement" ], "dimensions": { "biologicalPlausibility": { "rating": "strong", "reason": "The proposed transport mechanism is plausible and blood exposure changes." }, "animalEvidence": { "rating": "moderate", "reason": "Animal artery-wall delivery cannot answer survival or total human harm." }, "inVitroEvidence": { "rating": "strong", "reason": "Gemcitabine mechanism is established but does not settle route benefit." }, "reproducibility": { "rating": "weak", "reason": "The final randomized result and independent reproduction do not exist." }, "commercialConflicts": { "rating": "high", "reason": "RenovoRx owns the device, funded linked studies and controls trial reporting." }, "recency": { "rating": "current", "reason": "Enrollment closed and 78 of 86 deaths were reported in August 2026." } } }, "layout": { "x": 532, "y": 456 }, "updated": "2026-09-15" }, { "id": "unknown-ttfields-survival-replication-and-access", "title": "Will the Optune Pax survival gain reproduce and justify its burden and price?", "shortTitle": "TTFields net value unknown", "type": "unknown", "status": "unresolved", "scope": [ "pdac", "locally-advanced", "value", "access" ], "summary": "One sponsor-run trial supports a narrow FDA-approved use, but independent survival, full patient experience and affordable net value remain unknown.", "content": "Known quantity. In PANOVA-3, adding Optune Pax to gemcitabine and nab-paclitaxel changed median survival from 14.2 to 16.2 months and one-year survival from 60.2% to 68.1%. Unknown quantities. Would an independent randomized trial on the same or a current systemic-treatment backbone reproduce the overall-survival effect? Why did survival improve while progression-free survival, local progression-free survival, response and surgery did not establish corresponding gains? Which patients can and will carry the device for a median 11.2 hours each day, replace arrays every three to four days and manage skin injury? What are quality of life and hospital-free time after local progression, when PANOVA-3 stopped collecting some quality-of-life measures? What price makes the health gain a better use of limited cancer funds? The two published U.S. models disagree on benefit size but agree on poor value at current cost. One company-funded model estimated 0.34 extra quality-adjusted life-years, meaning the equivalent of 0.34 years in full health, for $131,400 extra and $387,300 per quality-adjusted life-year. An independent model estimated 0.148 extra quality-adjusted life-years for $147,766 extra and $1,001,995 per quality-adjusted life-year. These are models, not observed trial outcomes. A mission ceiling of £50,000 per verified quality-adjusted life-year would allow only £7,400 to £17,000 of total extra cost across that benefit range: £50,000 x 0.148 = £7,400; £50,000 x 0.34 = £17,000. That is a mission purchasing rule, not national guidance. Resolving work needs independent participant-level PANOVA-3 analysis, complete patient-experience collection through progression, a full referral-to-use denominator, and company-funded randomized reproduction before use beyond the FDA label.", "sourceIds": [ "macarulla-2025-panova3", "clinicaltrials-nct03377491-panova3", "fda-2026-optune-pax-ssed", "guzauskas-2026-ttfields-cost", "liu-2026-ttfields-cost" ], "links": [ { "target": "unknowns", "relation": "part-of" }, { "target": "claim-panova3-ttfields-small-os-benefit", "relation": "bounded-by" }, { "target": "failure-panova4-component-and-historical-control", "relation": "exposed-by" }, { "target": "hypothesis-optune-pax-independent-value-gate", "relation": "tested-by" }, { "target": "trial-pdac-ttfields-action-map", "relation": "mapped-by" }, { "target": "company-novocure", "relation": "constrained-by" } ], "evidence": { "humanProspective": "strong", "humanRetrospective": "not-applicable", "randomised": "strong", "independentReplication": "none", "sampleSize": "PANOVA-3 n=571; two cost models with no new patients", "effect": "Randomized median OS +2.0 months, but independent reproduction and affordable net value remain unknown", "limits": [ "One pivotal trial", "Open-label", "Sponsor-run", "Progression endpoints did not align", "Quality of life after local progression was not captured", "Cost studies are models", "U.S. costs do not directly transfer to the United Kingdom", "Patient preference and access route are not quantified" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "The field can affect dividing cells, but the clinical path from exposure to survival without scan-based control is unresolved." }, "animalEvidence": { "rating": "moderate", "reason": "Animal and cell work informed the device frequency but cannot explain the human endpoint pattern." }, "inVitroEvidence": { "rating": "strong", "reason": "Cell-division effects are reproducible outside patients." }, "reproducibility": { "rating": "none", "reason": "No independent pancreatic randomized survival reproduction was found." }, "commercialConflicts": { "rating": "high", "reason": "The device owner sponsored the pivotal trial and one of the two value models." }, "recency": { "rating": "current", "reason": "FDA, trial, cost and commercial records were checked through September 2026." } } }, "layout": { "x": 598, "y": 522 }, "updated": "2026-09-15" }, { "id": "unknown-vcn01-all-enrolled-survival", "title": "Does VCN-01 improve survival across everyone assigned to VIRAGE?", "shortTitle": "VCN-01 survival unknown", "type": "unknown", "status": "open", "scope": [ "pdac" ], "summary": "The unknown is not whether the curves favour VCN-01 in 96 treated people; it is whether that effect survives complete participant accounting, the locked test and independent confirmation.", "content": "Unknown. Does adding VCN-01 to gemcitabine and nab-paclitaxel improve overall and quality-adjusted survival for all eligible people assigned treatment, and which baseline state changes that effect? Four uncertainties must stay separate. First, accounting: what happened to the 16 people between 112 actual enrollment and the 96-person FAS? Second, statistics: which alpha, endpoint order and analysis population were fixed in the final dated analysis plan before the database was unlocked? Third, balance: does the estimate persist after true all-randomized analysis and the reported 83% versus 65% liver-metastasis imbalance? Fourth, confirmation: does a double-blind phase 3 reproduce survival without sacrificing chemotherapy, quality of life or safe access? Resolving work now. No new mission cohort is needed. Obtain the all-112 flow, allocation status, arm, reasons, outcomes, participant-level time-to-event data, chemotherapy doses, both VCN-01 doses, harms, quality of life, neutralising antibodies, blood viral genomes and PH20. An independent team must reproduce every named analysis and report both the locked primary test and an all-randomized estimate. Go to phase 3 support only if at least 99% of randomized participants are accounted for, the all-randomized OS hazard ratio remains at or below 0.75 with its 95% interval excluding 1.0 or the sponsor shows a regulator-agreed phase 2 success rule fixed before access, no plausible missing-outcome assignment erases the direction, and quality-adjusted survival does not worsen. The phase 3 must be placebo-controlled, stratify baseline liver metastasis and neutralising-antibody state, keep gemcitabine/nab-paclitaxel identical, use all-randomized OS as primary and publish every denominator. Stop mission support for data refusal, an all-randomized HR above 0.85, major severe-toxicity excess above ten points, chemotherapy dose loss above ten points, or a design that conditions efficacy on reaching later cycles. Reopen for posted results, a peer-reviewed full report, the final analysis plan or a funded phase 3 protocol.", "sourceIds": [ "clinicaltrials-nct05673811-virage", "theriva-2023-virage-protocol", "theriva-2025-virage-topline", "esmo-2025-virage-2216mo", "ctis-2026-virage2", "theriva-2026-q2-10q" ], "links": [ { "target": "unknowns", "relation": "part-of" }, { "target": "claim-vcn01-randomized-signal-not-survival-proof", "relation": "created-by" }, { "target": "failure-virage-analysis-and-survivor-selection", "relation": "caused-by" }, { "target": "trial-vcn01-action-map", "relation": "resolved-by" }, { "target": "hypothesis-vcn01-replication-to-survival-gate", "relation": "tested-by" }, { "target": "company-theriva-biologics", "relation": "owned-by" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "not-applicable", "randomised": "moderate", "independentReplication": "none", "sampleSize": "Registry n=112; sponsor FAS n=96; future phase 3 size not public", "effect": "Defines data, survival, harm and design gates; no new effect estimate", "limits": [ "Gate thresholds are mission decisions", "Regulator discussions are sponsor summaries", "Full data and final analysis plan are unavailable", "Future phase 3 funding is absent" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "Human tumour replication supports a causal route, while survival still requires a clean comparison." }, "animalEvidence": { "rating": "moderate", "reason": "Models support entry to human testing, not a phase 3 conclusion." }, "inVitroEvidence": { "rating": "moderate", "reason": "Engineered lysis and PH20 activity are measurable but not sufficient." }, "reproducibility": { "rating": "none", "reason": "No independent randomized VCN-01 survival result exists." }, "commercialConflicts": { "rating": "high", "reason": "Product, phase 2 data and planned phase 3 are sponsor-controlled." }, "recency": { "rating": "current", "reason": "Built from current trial, registry and finance status." } } }, "layout": { "x": 174, "y": 94 }, "updated": "2026-09-15" }, { "id": "unknown-wee1-replication-stress-pdac", "title": "Does WEE1 inhibition exploit a measured replication-stress state in PDAC?", "shortTitle": "WEE1 utility gap", "type": "unknown", "status": "open", "scope": [ "pdac" ], "summary": "An uncontrolled local-disease study and a 12-person completed safety stage leave WEE1 efficacy, tumour target effect and patient selection unresolved.", "content": "Known. WEE1 is an emergency pause before cell division. Gemcitabine damages DNA while it is copied; WEE1 inhibition may force a highly stressed cancer cell through division before repair. In 34 locally advanced patients, adavosertib with gemcitabine and radiation produced median OS 21.7 months and PFS 9.4 months. The study had no control, eight people had a dose-limiting toxicity and target effect was shown in hair follicles rather than pancreatic tumour. NCT06015659 planned a 12-person azenosertib safety lead-in followed by a two-stage efficacy test of six-month PFS at least 44% versus an unacceptable 24%. The current registry reports actual enrollment of 12, actual primary completion in August 2025, no public results and no efficacy expansion. Unknown. Whether tumour WEE1 was inhibited, whether replication stress rose, why expansion did not occur, the six-month control rate, response, survival, severe harm and whether a baseline or early-change marker identifies benefit. Why it matters. CHK1 inhibition already failed a randomized PDAC test, so a nearby checkpoint cannot inherit efficacy from the same model logic. Resolving evidence. Release all 12 records, paired tumour results, dose intensity and the documented expand-or-stop decision. Continue the route only if at least ten have usable paired tumour samples, at least eight show the planned WEE1 and DNA-damage change, six-month PFS reaches at least 5/12, no treatment-related death occurs and the result can define a reproducible selection rule. Do not infer efficacy from hair, blood, organoids or historical survival. If those gates fail, stop PDAC WEE1 investment. If they pass, the next test must randomize identical chemotherapy with or without the WEE1 inhibitor and count every assigned person.", "sourceIds": [ "cuneo-2019-adavosertib-pdac", "huffman-2024-azenosertib-pdac-design", "clinicaltrials-nct06015659-azenosertib", "laquente-2017-chk1-pdac", "huffman-2023-ly2880070-pdac" ], "links": [ { "target": "unknowns", "relation": "part-of" }, { "target": "failure-broad-cell-cycle-checkpoint-pdac", "relation": "bounded-by" }, { "target": "trial-cell-cycle-checkpoint-action-map", "relation": "resolved-by" }, { "target": "hypothesis-human-replication-stress-selector", "relation": "tests" } ], "evidence": { "humanProspective": "weak", "humanRetrospective": "not-applicable", "randomised": "none", "independentReplication": "none", "sampleSize": "Uncontrolled adavosertib n=34; completed azenosertib safety stage n=12", "effect": "Required output is tumour target effect, six-month control, harm and a reproducible selector in all 12 existing records", "limits": [ "No posted azenosertib results", "Small denominator", "Historical comparator", "Different drugs and disease stages", "Proposed thresholds are mission decisions" ], "dimensions": { "biologicalPlausibility": { "rating": "moderate", "reason": "Forcing a DNA-damaged cell through division is coherent, but PDAC checkpoint translation has failed before." }, "animalEvidence": { "rating": "strong", "reason": "Models show chemotherapy interaction but have not selected human benefit." }, "inVitroEvidence": { "rating": "strong", "reason": "Replication stress, checkpoint activity and DNA damage can be measured directly." }, "reproducibility": { "rating": "none", "reason": "No controlled PDAC WEE1 efficacy result exists." }, "commercialConflicts": { "rating": "moderate", "reason": "Zentalis supplied the product, while academic and charity groups sponsor the current study." }, "recency": { "rating": "current", "reason": "Uses the May 2026 registry state after actual primary completion." } } }, "layout": { "x": 212, "y": 130 }, "updated": "2026-09-15" }, { "id": "unknowns", "title": "Uncertainty ledger", "shortTitle": "Unknowns", "type": "unknown", "status": "supported", "scope": [ "pdac", "pnet", "cholangiocarcinoma", "gallbladder", "hcc", "colorectal", "pan-cancer" ], "summary": "72 unresolved questions can reverse programme decisions; each names the uncertainty, affected decision and resolving experiment.", "content": "PDAC questions. 1 — Can broad RAS suppression be made deep and durable without intolerable normal-tissue toxicity? Median progression on daraxonrasib was 7.3 months, so initial targetability is settled but durable control is not. 2 — Are the cells that cause late recurrence dormant, slowly growing, repeatedly changing state or a mixture, and where do they reside? Models prove possible routes, not prevalence in people. 3 — How long is the useful interception window? Slow phylogenetic estimates, early dissemination and punctuated evolution cannot be used as one timetable. 4 — Which PanIN or IPMN will become lethal? KRAS and GNAS identify lineage but do not select dangerous progression. 5 — Does acting on postoperative molecular residual disease improve survival? Prognosis is not utility. 6 — Which patients have a local-dominant course that justifies radiation, fields or ablation? 7 — Can immune exclusion be changed without removing protective stroma or causing systemic toxicity? 8 — Can tumour-specific delivery be measured fast enough to route treatment? 9 — Does treating cachexia preserve function and chemotherapy enough to extend useful life, rather than weight alone? 10 — Which PDAC patients gain function or treatment capacity from pancreatic enzyme replacement, at what dose, and does that change survival rather than weight alone? 11 — Which PDAC patients should receive which clot prophylaxis, when and for how long? The class prevents VTE, while individual bleeding, interactions, burden, delivery and duration remain unresolved. 12 — Which obstructed PDAC patient needs prompt surgery or drainage, which device and rescue route, and who owns infection-free treatment continuity? 13 — Which English germline-testing rule and consented family-contact route turns inherited findings into timely patient and relative action without privacy, inequality or uncertain-variant harm? 14 — Which parts of the PDAC surgical route are avoidably losing expert review, safe recovery or planned systemic treatment, and can a named route owner repair them without increasing operative harm? 15 — Does scheduled CA19-9 and CT after PDAC resection improve survival and useful life, or only detect recurrence earlier and select fitter patients for treatment? 16 — Which technically resectable PDAC patient should receive systemic treatment before surgery, with which regimen, without losing either effective chemotherapy or a curative operation? Each linked record states the resolving human experiment and stop rule. 17 — Older and vulnerable PDAC: which person gains useful life from combination treatment, gentler treatment or supportive care, and which correctable need must be acted on before toxicity removes that gain? 18 — Hidden-spread staging: which patient needs a separate camera examination before treatment, should every planned open operation begin with a same-session check, and when should visible spread, biopsy or abdominal-wash cytology change the plan? 19 — Organoid treatment choice: can a tumour-confirmed result return before first-line treatment, reproduce across laboratories and improve useful life from the original enrolled denominator? 20 — FAP radiation dose: which lesion-level absorbed dose is safe and predicts control, and does adding the radioligand to the same standard treatment improve overall or quality-adjusted survival? 21 — Selective autophagy: can a direct inhibitor block the intended recycling process inside PDAC tumour, preserve an active RAS-drug dose and improve useful life? HCQ combinations failed; final inlexisertib data are missing. 22 — Sonoporation: does ultrasound plus microbubbles add useful life on one fixed chemotherapy backbone, and can increased drug delivery into tumour be measured directly? The second randomized trial is negative overall in an accepted abstract; its highlighted subgroup lacks the information needed to test an interaction. 23 — KRAS siRNA: can combined tumour injection and under-skin SIL204 reduce mutant KRAS across separated primary-tumour regions and distant disease without sacrificing chemotherapy? The new product has no reported human target result. 24 — VCN-01: does its favourable 96-person result persist across everyone assigned from the 112 enrolled, under the final locked analysis, and then reproduce in a blinded trial without added harm or chemotherapy loss?\n\nOther cancers and delivery. 25 — pNET: does repeat biopsy plus dual PET improve outcomes, or only change labels and treatment plans? 26 — pNET: which small non-functioning tumour should be observed, ablated or removed after procedure-specific harm is counted? 27 — Cholangiocarcinoma: which separate PSC and liver-fluke risk routes produce net benefit from surveillance? 28 — Gallbladder: can a selector find cancer or high-grade dysplasia often enough to justify preventive surgery when stones and small polyps are common? 29 — HCC: can active preservation of liver reserve reduce death or decompensation without weakening tumour treatment? 30 — Colorectal liver metastasis: which molecular and time-dependent state is still locally curable by ablation, resection or transplant? 31 — Cholangiocarcinoma: which treatment or imaging action benefits a postoperative ctDNA-positive patient? 32 — HCC: which surveillance route prevents late disease and death at acceptable cost and work-up harm? 33 — Colorectal: which treatment or imaging action improves survival for a postoperative ctDNA-positive patient? 34 — Regional delivery: which dated step removes the most otherwise eligible patients from a curative or life-extending HPB route at each site? Current African and Latin American reports expose large loss but begin after different selection points. The shared audit must locate it before a local repair is randomized. These ten related-cancer and delivery questions repair the prior PDAC-only bias. Questions 1 and 5 still have the highest expected effect on the central PDAC programme, while question 32 blocks global intervention money that would otherwise be spent on an unmeasured route.", "sourceIds": [ "gudmundsdottir-2023-staging-laparoscopy", "van-dongen-2023-preopanc-staging", "rompen-2026-preopanc2-staging", "oreilly-2026-daraxonrasib", "lin-2013-reversible-pdac-dormancy", "shi-2026-chemo-dormant-dtc", "yachida-2010-metastatic-timing", "rhim-2012-early-dissemination", "notta-2016-punctuated-evolution", "wu-2011-gnas-ipmn", "kanda-2012-early-panin-mutations", "botta-2024-tumor-informed-ctdna", "macarulla-2025-panova3", "feig-2013-cxcl12-immune-exclusion", "koay-2014-human-pdac-drug-transport", "hingorani-2020-halo301", "mollazadegan-2025-pnet-recharacterization", "partelli-2022-aspen-interim", "maekawa-2026-small-pnet-surgery-morbidity", "villard-2023-psc-surveillance", "khuntikeo-2020-thailand-cca-screening", "yu-2025-btc-ctdna-recurrence", "yoo-2025-extrahepatic-cca-ctdna", "clinicaltrials-nct05743959-btc-mrd", "szpakowski-2020-gallbladder-polyps", "seif-el-dahan-2026-early-hcc-causes-death", "ruers-2017-clocc-crc-liver-ablation", "rhee-2025-miracle-hcc", "hirode-2026-hcc-biomarker-rct", "trinchet-2011-hcc-surveillance-interval", "clinicaltrials-nct06084234-tracer", "clinicaltrials-nct05486572-premium", "tie-2025-dynamic3", "bando-2026-altair", "mo-2026-find-crc-surveillance", "clinicaltrials-nct05174169-circulate-us", "gismalla-2026-africa-pancreatic-review", "sobnach-2024-ssa-hcc-systematic-review", "sobnach-2025-south-africa-hcc-disparities", "sobnach-2026-global-hcc-disparities", "da-fonseca-2024-latam-cca", "babic-2019-muscle-loss-survival", "chen-2025-tocilizumab-cachexia", "groarke-2024-ponsegromab-cachexia", "clinicaltrials-nct06989437-ponsegromab-pdac", "sandhya-2023-olanzapine-anorexia", "woo-2016-pert-rct", "saito-2018-pert-rct", "picozzi-2025-pert-advanced-pdac", "dugic-2026-pert-mortality", "clinicaltrials-nct04098237-pancax3", "clinicaltrials-nct05642962-pertzye-dose", "clinicaltrials-nct06099119-pertseverance", "maraveyas-2012-fragem", "pelzer-2015-conko004", "vadhan-raj-2020-cassini-pdac", "frere-2020-pdac-thromboprophylaxis-meta", "ash-2023-primary-vte-prophylaxis-guide", "overvad-2022-khorana-validation", "heffley-2024-pdac-vte-uptake", "clinicaltrials-nct06861088-catiq-p3", "clinicaltrials-nct05178628-impact-pro", "van-der-gaag-2010-pbd-rct", "costamagna-2026-sems-early-surgery-rct", "tamura-2021-fcsems-plastic-rct", "seo-2019-covered-uncovered-sems-rct", "lyu-2023-metal-plastic-nat-meta", "vehvilainen-2024-cholangitis-nat-interruption", "chen-2023-element-biliary-rct", "paik-2018-eus-ercp-biliary-rct", "barbosa-2024-eus-ercp-meta", "esge-2018-biliary-stenting-guideline", "asge-2024-therapeutic-eus-guideline", "nhs-england-2026-r367", "llach-2024-universal-germline", "wang-2022-oncology-germline-cascade", "rodriguez-2024-generate", "katz-2026-gift-cascade", "npaca-2026-state-of-nation", "nhs-england-2024-hpb-pancreatic-service", "nice-2018-ng85-pancreatic", "de-wilde-2012-centralization", "lemmens-2011-centralization", "latenstein-2021-centralization", "coupland-2016-england-resection-volume", "labori-2016-adjuvant-completion", "henry-2023-complications-adjuvant", "kollbeck-2025-complications-chemo", "van-hilst-2019-leopard2", "korrel-2023-diploma", "bruna-2025-diploma-long-term", "klotz-2024-europa", "liu-2024-robotic-open-pd", "jin-2026-portal", "andel-2025-routine-imaging", "van-goor-2025-nationwide-imaging", "daamen-2024-radar-panc-protocol", "clinicaltrials-nct04875325-radar-panc", "tai-2025-pan-asian-esmo-pancreatic", "labori-2024-norpact1", "bai-2025-cispd1", "tan-2026-neoadjuvant-ipd-meta", "ghaneh-2023-espac5", "clinicaltrials-nct04340141-a021806", "effenberger-2012-pdac-bone-marrow-dtc", "nordgard-2022-unresectable-dtc", "chalabi-hajkarim-2025-organotropism", "pei-2025-spatial-metastatic-pdac", "mcclellan-2026-quiescent-pdac-car-t", "tanaka-2012-microscopic-local-recurrence", "dotan-2024-giant", "dotan-2025-giant-ga", "clinicaltrials-nct04233866-giant", "betge-2022-grantpax", "clinicaltrials-nct02143219-pamela70", "mohile-2021-gap70", "li-2021-gain", "ihorst-2026-geriatric-comanagement", "clinicaltrials-nct06040801-frail-gi", "stoop-2026-preopanc4-surgery", "dekker-2026-taps-restaging", "verkolf-2026-taps-resection-prediction", "kunzmann-2021-neolap", "clinicaltrials-nct06714604-duration", "clinicaltrials-nct06132087-proactive", "baum-2022-fap2286-firsthuman", "mcconathy-2026-lumiere-phase1", "clinicaltrials-nct04939610-lumiere", "baum-2026-3bp3940", "clinicaltrials-nct05432193-frontier", "clinicaltrials-nct07229768-ctr-fapi", "alistar-2017-devimistat-phase1", "philip-2024-avenger500", "surana-2025-erk-hcq-pdac", "clinicaltrials-nct05221320-ulixertinib-hcq", "manji-2023-mekiauto", "clinicaltrials-nct04214418-mekiauto", "clinicaltrials-nct04892017-inlexisertib", "deciphera-2022-dcc3116-phase1", "clinicaltrials-nct04524702-paricalcitol-hcq", "dimcevski-2016-sonoporation-phase1", "wang-2023-sonochemotherapy-pdac", "clinicaltrials-nct04821284-sonoporation", "ius-2026-sonoporation-nct04821284-abstract", "adler-2026-sonoporation-perfusion", "alagesan-2026-sig12d-loder", "ctis-2026-sil204-phase23", "silexion-2026-sil204-site-initiation", "silexion-2026-halfyear-filing", "bazan-peregrino-2021-vcn01-intratumour", "garcia-carbonero-2022-vcn01-intravenous", "clinicaltrials-nct05673811-virage", "theriva-2023-virage-protocol", "theriva-2025-virage-topline", "esmo-2025-virage-2216mo", "ctis-2026-virage2", "theriva-2026-q2-10q", "richards-2006-ci994-pdac", "heumann-2022-azacitidine-pdac", "sohal-2020-thu-decitabine-pdac", "baretti-2024-entinostat-nivolumab-pdac", "clinicaltrials-nct03250273-entinostat-nivolumab", "safyan-2026-azacitidine-pembrolizumab-pdac", "clinicaltrials-nct03264404-azacitidine-pembrolizumab", "rodon-2024-amg193-prmt5", "matsumoto-2026-mtap-pancreatic", "pavlick-2026-mtap-genomics", "clinicaltrials-nct05732831-vopimetostat", "tango-2025-vopimetostat-monotherapy", "clinicaltrials-nct06922591-vopimetostat-ras", "tango-2026-vopimetostat-ras-data", "tango-2026-q2-10q", "mahalingam-2026-elraglusib-randomized", "clinicaltrials-nct03678883-elraglusib", "pathak-2026-elraglusib-folfirinox", "clinicaltrials-nct05077800-elraglusib-folfirinox", "actuate-2026-q2-10q", "sec-2026-actuate-submissions", "chiorean-2023-abemaciclib-pdac", "baghdadi-2019-palbociclib-cdkn2a", "hidalgo-2022-palbociclib-nabpaclitaxel", "laquente-2017-chk1-pdac", "huffman-2023-ly2880070-pdac", "cuneo-2019-adavosertib-pdac", "huffman-2024-azenosertib-pdac-design", "clinicaltrials-nct06015659-azenosertib", "yu-2025-ibi343-pdac", "clinicaltrials-nct05458219-ibi343", "clinicaltrials-nct07066098-ibi343-phase3", "park-2026-gleam-zolbetuximab", "abou-alfa-2006-exatecan-pdac", "qi-2024-ct041-pdac", "clinicaltrials-nct03874897-ct041", "xu-2026-kras-g12v-tcr", "clinicaltrials-nct04146298-kras-g12v-tcr", "leidner-2022-kras-tcr", "aznar-2025-mesothelin-car-t-resistance", "clinicaltrials-nct03323944-meso-cart", "van-laethem-2025-optimize1-biomarkers", "clinicaltrials-nct04888312-optimize1", "padron-2022-prince", "alligator-2026-q2", "murphy-2019-losartan-lapc", "ramaswamy-2025-afpac-losartan", "clinicaltrials-nct03563248-losartan-localized", "geller-2017-bacterial-gemcitabine", "corty-2020-antibiotic-gemcitabine-toxicity", "sharma-2026-fmtpanc", "ctis-panoramix-2024-517766-41", "middleton-2014-telovac", "le-2019-eclipse", "elicio-2026-amplify7p-results", "nct05968326-imcode003", "wainberg-2026-quemliclustat-arc8", "coveler-2024-oleclumab-pdac", "clinicaltrials-nct06608927-prism1", "arcus-2026-q2-10q", "clinicaltrials-nct05254171-aspire", "ctis-2024-514714-12-00-aspire", "panbela-2024-q3-10q", "singhal-2021-sbp101-phase1", "carla-2024-ivospemin-retinal-atrophy", "clinicaltrials-nct03257033-tigerpac", "hatoum-2024-tamp-rr1-rr2", "novelli-2026-tigerpac-pk", "renovorx-2023-tigerpac-interim", "renovorx-2025-10k", "renovorx-2026-full-enrollment", "timmer-2024-crossfire", "clinicaltrials-nct03899636-direct-rct", "clinicaltrials-nct03899649-direct-registry", "martin-2024-direct-registry-safety", "martin-2026-direct-registry-survival-preprint", "rai-2022-lappie-protocol", "isrctn-14986389-lappie", "nice-2017-ire-pancreatic", "clinicaltrials-nct03377491-panova3", "fda-2026-optune-pax-ssed", "clinicaltrials-nct06390059-panova4", "novocure-2026-panova4-topline", "novocure-2026-q2-10q", "guzauskas-2026-ttfields-cost", "liu-2026-ttfields-cost", "issels-2023-heat-rct", "clinicaltrials-nct01077427-heat", "eudract-2008-004802-14-heat", "datta-2017-heatpac-protocol", "clinicaltrials-nct02439593-heatpac", "shinoto-2016-carbon-ion-gemcitabine", "kawashiro-2018-jcros-carbon-ion", "lai-2026-carbon-ion-risk-adapted", "rapp-2022-proton-phase2", "eckstein-2023-proton-registry", "clinicaltrials-nct02598349-proton", "clinicaltrials-nct03536182-cipher", "clinicaltrials-nct04194268-pack", "parikh-2023-smart-phase2-safety", "chuong-2024-smart-phase2-survival", "clinicaltrials-nct03621644-smart", "clinicaltrials-nct05585554-lap-ablate", "clinicaltrials-nct05114213-maspac", "viewray-2023-q1-10q", "viewray-2023-chapter11-8k", "seelen-2026-pelican", "clinicaltrials-nct03690323-pelican", "li-2015-hifu-s1", "marinova-2018-hifu-prospective", "marinova-2024-hifu-pc-protocol", "drks-00012367-hifu-pc", "clinicaltrials-nct05262452-hifu-folfirinox", "clinicaltrials-nct05601323-suizenji", "clinicaltrials-nct06211933-puls-hifu", "clinicaltrials-nct07325214-sonopan3", "izzo-2021-pancreatic-ect-phase12", "izzo-2021-laparoscopic-ect-protocol", "eudract-2018-003925-27-ect", "cebron-2023-panect-preliminary", "clinicaltrials-nct04281290-panect-results", "rudno-rudzinska-2021-irec-protocol", "igea-2026-cliniporator-vitae", "testoni-2021-hybridtherm-rct", "gannon-2026-pancreas-histotripsy-safety", "pusceddu-2026-percutaneous-cryoablation", "zhou-2024-pancreatic-mwa", "ross-2022-panco-p32", "huggett-2014-verteporfin-pdt", "dimatteo-2018-eus-laser-ablation", "clinicaltrials-nct07698613-asp1929-pancreas", "muragaki-2026-k912-sonodynamic", "clinicaltrials-nct05371223-nab-pipac", "padilla-valverde-2024-adjuvant-hipec-rct", "levy-2019-celiac-ganglia-versus-plexus-rct", "xie-2026-cnsi-fe-first-human", "fda-2024-bizengri-multidisciplinary-review", "maltoni-2016-systematic-palliative-qol", "chan-2024-dpyd-non-european-review", "pishvaian-2020-know-your-tumor", "neuzillet-2023-apacap-rct" ], "links": [ { "target": "unknown-pdac-organoid-guided-treatment-utility", "relation": "contains" }, { "target": "unknown-pdac-staging-laparoscopy-rule", "relation": "contains" }, { "target": "experiments", "relation": "resolved-by" }, { "target": "unknown-durable-ras-control", "relation": "contains" }, { "target": "unknown-actionable-interception-window", "relation": "contains" }, { "target": "unknown-dangerous-precursor-selection", "relation": "contains" }, { "target": "unknown-ctdna-actionability", "relation": "contains" }, { "target": "unknown-dormant-cell-actionability", "relation": "contains" }, { "target": "unknown-pnet-reclassification-actionability", "relation": "contains" }, { "target": "unknown-small-pnet-surgery-selector", "relation": "contains" }, { "target": "unknown-cca-surveillance-utility", "relation": "contains" }, { "target": "unknown-cca-postoperative-mrd-action", "relation": "contains" }, { "target": "unknown-gallbladder-prevention-selector", "relation": "contains" }, { "target": "unknown-hcc-liver-reserve-modifiability", "relation": "contains" }, { "target": "unknown-hcc-surveillance-modality-utility", "relation": "contains" }, { "target": "unknown-crc-liver-local-cure-selector", "relation": "contains" }, { "target": "unknown-crc-ctdna-positive-action", "relation": "contains" }, { "target": "unknown-regional-route-loss", "relation": "contains" }, { "target": "unknown-cachexia-treatment-utility", "relation": "contains" }, { "target": "unknown-pert-pdac-utility", "relation": "contains" }, { "target": "claim-pdac-delivery-is-variable-not-single-barrier", "relation": "could-change" }, { "target": "ranking-provisional-pdac-bottlenecks", "relation": "could-change" }, { "target": "unknown-pdac-vte-prophylaxis-rule", "relation": "contains" }, { "target": "unknown-pdac-biliary-drainage-route", "relation": "contains" }, { "target": "unknown-pdac-germline-family-action-route", "relation": "contains" }, { "target": "claim-pdac-surgery-is-a-complete-treatment-route", "relation": "links-to" }, { "target": "system-england-pancreatic-surgery-network", "relation": "links-to" }, { "target": "failure-pdac-postoperative-treatment-loss", "relation": "links-to" }, { "target": "failure-leopard2-laparoscopic-whipple-safety", "relation": "links-to" }, { "target": "unknown-pdac-surgical-route-completion", "relation": "links-to" }, { "target": "trial-pdac-surgical-route-action-map", "relation": "links-to" }, { "target": "hypothesis-pdac-surgical-route-completion", "relation": "links-to" }, { "target": "unknown-postresection-surveillance-utility", "relation": "links-to" }, { "target": "claim-routine-imaging-after-pdac-resection-not-yet-proven", "relation": "links-to" }, { "target": "trial-pdac-postresection-surveillance-action-map", "relation": "links-to" }, { "target": "unknown-pdac-perioperative-treatment-route", "relation": "links-to" }, { "target": "trial-pdac-perioperative-sequencing-action-map", "relation": "links-to" }, { "target": "failure-norpact1-neoadjuvant-delivery", "relation": "links-to" }, { "target": "claim-human-pdac-dtc-prognostic-not-dormancy", "relation": "links-to" }, { "target": "failure-pdac-dormancy-sample-gap", "relation": "links-to" }, { "target": "trial-pdac-residual-cell-state-action-map", "relation": "links-to" }, { "target": "unknown-older-pdac-treatment-route", "relation": "contains" }, { "target": "claim-age-alone-cannot-route-pdac-treatment", "relation": "links-to" }, { "target": "trial-older-pdac-action-map", "relation": "links-to" }, { "target": "hypothesis-older-pdac-fast-support-route", "relation": "links-to" }, { "target": "unknown-lapc-surgery-causal-benefit", "relation": "contains" }, { "target": "trial-lapc-conversion-surgery-action-map", "relation": "links-to" }, { "target": "hypothesis-randomized-lapc-resection-after-response", "relation": "links-to" }, { "target": "unknown-pdac-fap-radioligand-treatment-effect", "relation": "contains" }, { "target": "failure-fapi-scan-dose-benefit-collapse", "relation": "caused-by" }, { "target": "trial-pdac-fap-radioligand-action-map", "relation": "resolved-by" }, { "target": "unknown-selective-autophagy-pdac-contribution", "relation": "contains" }, { "target": "trial-pdac-autophagy-action-map", "relation": "mapped-by" }, { "target": "unknown-pdac-sonoporation-treatment-effect", "relation": "contains" }, { "target": "trial-pdac-sonoporation-action-map", "relation": "mapped-by" }, { "target": "hypothesis-measured-fixed-backbone-sonoporation", "relation": "resolved-by" }, { "target": "unknown-sil204-human-target-delivery", "relation": "contains" }, { "target": "trial-sil204-action-map", "relation": "mapped-by" }, { "target": "hypothesis-multicompartment-kras-sirna-delivery", "relation": "resolved-by" }, { "target": "claim-vcn01-randomized-signal-not-survival-proof", "relation": "bounded-by" }, { "target": "failure-virage-analysis-and-survivor-selection", "relation": "learns-from" }, { "target": "unknown-vcn01-all-enrolled-survival", "relation": "leaves-open" }, { "target": "trial-vcn01-action-map", "relation": "implements" }, { "target": "hypothesis-vcn01-replication-to-survival-gate", "relation": "tests" }, { "target": "company-theriva-biologics", "relation": "contains" }, { "target": "failure-broad-epigenetic-treatment-pdac", "relation": "learns-from" }, { "target": "claim-mtap-prmt5-selected-human-signal", "relation": "bounded-by" }, { "target": "failure-vopimetostat-ras-contribution-unknown", "relation": "learns-from" }, { "target": "unknown-mtap-prmt5-ras-contribution", "relation": "leaves-open" }, { "target": "trial-mtap-prmt5-action-map", "relation": "implements" }, { "target": "hypothesis-mtap-prmt5-ras-deeper-control", "relation": "tests" }, { "target": "company-tango-therapeutics", "relation": "contains" }, { "target": "claim-elraglusib-randomized-survival-signal", "relation": "bounded-by" }, { "target": "failure-elraglusib-analysis-and-survival-discordance", "relation": "learns-from" }, { "target": "unknown-elraglusib-survival-replication", "relation": "leaves-open" }, { "target": "trial-elraglusib-action-map", "relation": "implements" }, { "target": "hypothesis-elraglusib-survival-confirmation-gate", "relation": "tests" }, { "target": "company-actuate-therapeutics", "relation": "contains" }, { "target": "failure-broad-cell-cycle-checkpoint-pdac", "relation": "learns-from" }, { "target": "unknown-wee1-replication-stress-pdac", "relation": "leaves-open" }, { "target": "trial-cell-cycle-checkpoint-action-map", "relation": "implements" }, { "target": "hypothesis-human-replication-stress-selector", "relation": "tests" }, { "target": "unknown-cldn182-adc-survival-effect", "relation": "contains" }, { "target": "hypothesis-cldn182-adc-survival-test", "relation": "tests" }, { "target": "unknown-engineered-cell-reproducible-benefit-pdac", "relation": "contains" }, { "target": "hypothesis-kras-tcr-human-persistence-gate", "relation": "tests" }, { "target": "unknown-cd40-added-survival-pdac", "relation": "contains" }, { "target": "hypothesis-mitazalimab-controlled-survival-gate", "relation": "tests" }, { "target": "unknown-localized-losartan-contribution-pdac", "relation": "contains" }, { "target": "hypothesis-localized-losartan-existing-trial-gate", "relation": "tests" }, { "target": "unknown-microbiome-treatment-contribution-pdac", "relation": "contains" }, { "target": "hypothesis-bacterial-gemcitabine-deactivation-gate", "relation": "tests" }, { "target": "trial-pdac-microbiome-action-map", "relation": "mapped-by" }, { "target": "unknown-pdac-vaccine-regimen-contribution", "relation": "contains" }, { "target": "hypothesis-autogene-cevumeran-existing-trial-gate", "relation": "tests" }, { "target": "trial-pdac-vaccine-action-map", "relation": "mapped-by" }, { "target": "unknown-quemliclustat-survival-contribution", "relation": "contains" }, { "target": "hypothesis-quemliclustat-existing-phase3-gate", "relation": "tests" }, { "target": "trial-pdac-adenosine-action-map", "relation": "mapped-by" }, { "target": "unknown-aspire-assigned-denominator-and-results", "relation": "contains" }, { "target": "hypothesis-aspire-data-rescue-gate", "relation": "tests" }, { "target": "trial-pdac-polyamine-action-map", "relation": "mapped-by" }, { "target": "unknown-tigerpac-package-survival-and-delivery", "relation": "contains" }, { "target": "hypothesis-tigerpac-existing-final-gate", "relation": "tests" }, { "target": "trial-pdac-intraarterial-delivery-action-map", "relation": "mapped-by" }, { "target": "unknown-ire-added-survival-after-modern-chemotherapy", "relation": "contains" }, { "target": "hypothesis-direct-randomized-data-recovery-gate", "relation": "tests" }, { "target": "trial-pdac-irreversible-electroporation-action-map", "relation": "mapped-by" }, { "target": "unknown-ttfields-survival-replication-and-access", "relation": "contains" }, { "target": "hypothesis-optune-pax-independent-value-gate", "relation": "tests" }, { "target": "trial-pdac-ttfields-action-map", "relation": "mapped-by" }, { "target": "unknown-regional-hyperthermia-added-benefit-modern-care", "relation": "contains" }, { "target": "hypothesis-heat-existing-randomization-recovery-gate", "relation": "tests" }, { "target": "trial-pdac-regional-hyperthermia-action-map", "relation": "mapped-by" }, { "target": "unknown-particle-therapy-added-patient-benefit", "relation": "contains" }, { "target": "hypothesis-particle-therapy-existing-data-gate", "relation": "tests" }, { "target": "trial-pdac-particle-therapy-action-map", "relation": "mapped-by" }, { "target": "unknown-smart-added-survival-and-surgery-harm", "relation": "contains" }, { "target": "hypothesis-smart-vendor-independent-randomized-gate", "relation": "tests" }, { "target": "trial-pdac-smart-action-map", "relation": "mapped-by" }, { "target": "unknown-rfa-human-mechanism-after-negative-trial", "relation": "contains" }, { "target": "hypothesis-pelican-specimen-salvage-gate", "relation": "tests" }, { "target": "trial-pdac-rfa-action-map", "relation": "mapped-by" }, { "target": "unknown-hifu-pain-survival-and-system-transfer", "relation": "contains" }, { "target": "hypothesis-hifu-existing-randomized-gate", "relation": "tests" }, { "target": "trial-pdac-hifu-action-map", "relation": "mapped-by" }, { "target": "unknown-electrochemotherapy-added-benefit-and-delivery", "relation": "contains" }, { "target": "hypothesis-electrochemotherapy-randomized-data-gate", "relation": "tests" }, { "target": "trial-pdac-electrochemotherapy-action-map", "relation": "mapped-by" }, { "target": "trial-pdac-eus-rfa-action-map", "relation": "mapped-by" }, { "target": "trial-pdac-histotripsy-action-map", "relation": "mapped-by" }, { "target": "trial-pdac-cryoablation-action-map", "relation": "mapped-by" }, { "target": "trial-pdac-mwa-action-map", "relation": "mapped-by" }, { "target": "trial-pdac-p32-action-map", "relation": "mapped-by" }, { "target": "trial-pdac-pdt-action-map", "relation": "mapped-by" }, { "target": "trial-pdac-laser-ablation-action-map", "relation": "mapped-by" }, { "target": "trial-pdac-asp1929-action-map", "relation": "mapped-by" }, { "target": "trial-pdac-k912-sonodynamic-action-map", "relation": "mapped-by" }, { "target": "trial-pdac-pipac-action-map", "relation": "mapped-by" }, { "target": "trial-pdac-hipec-action-map", "relation": "mapped-by" }, { "target": "trial-pdac-celiac-pain-action-map", "relation": "mapped-by" }, { "target": "trial-pdac-ferroptosis-action-map", "relation": "mapped-by" }, { "target": "trial-pdac-nrg1-action-map", "relation": "mapped-by" }, { "target": "trial-pdac-palliative-care-action-map", "relation": "mapped-by" }, { "target": "unknown-pdac-pharmacogenetic-route-net-benefit", "relation": "contains" }, { "target": "unknown-kras-wildtype-driver-route-completion", "relation": "contains" }, { "target": "unknown-pdac-exercise-net-benefit-and-selector", "relation": "contains" } ], "evidence": { "humanProspective": "moderate", "humanRetrospective": "strong", "randomised": "moderate", "independentReplication": "moderate", "sampleSize": "72 decision-critical questions linked to human trials, cohorts, tissue studies, route audits and models", "effect": "Uncertainty ranking; no pooled effect", "limits": [ "Unknowns cannot be assigned conventional confidence", "Some conflicts reflect different patient states rather than incompatible results", "Proposed resolving experiments may reveal new branches", "Priority is programme judgement" ] }, "layout": { "x": 30, "y": 94 }, "updated": "2026-09-15", "contentSections": [ "Question 43 asks whether bleomycin electrochemotherapy adds useful life and whether the whole target received the planned electric field. Recover the 90-person randomization with pulse logs, death-aware imaging, chemotherapy delivery, operation harm, quality of life and survival. Keep tumour treatment, cut-surface treatment, irreversible electroporation and calcium electroporation separate.", "Question 42 asks whether thermal HIFU improves pain, useful life or both and whether results transfer across devices. Pain response, opioid use, chemotherapy delivery, local failure, distant failure and quality-adjusted survival must remain separate. SonoPANIII's non-thermal drug-delivery device is a different mechanism and cannot validate thermal ablation.", "Question 41 asks whether radiofrequency ablation caused a reproducible human tissue or immune change despite failing as treatment. PELICAN already settles the clinical route: no survival benefit and more severe serious events. Its counterintuitive CA19-9 subgroup cannot select patients. Only existing paired specimens may be used to explain mechanism, with two-laboratory agreement and no authority to reopen an efficacy trial.", "Question 35 asks whether ASPIRE still contains a trustworthy randomized answer. The U.S. record estimates 600 and says recruiting; Europe lists 547 planned and says ended; the sponsor reported a 395-person safety set before its contractor exit. Resolve legal custody, every assignment, the blind and key, protocol versions, 12-month vital status and arm-specific serious eye and liver harm before interpreting pooled survival or funding rescue.", "Question 18 separates early staging before treatment from the same-session camera check before an open operation. It also prevents a pathology error: a visible biopsy-confirmed deposit and tumour cells found only in abdominal wash fluid are not the same decision state. The resolving 500-person cohort is nested in the existing English surgical-route study, with delay, procedure harm, avoided incision and later treatment all retained.", "Questions 2 and 3 now have a stricter human boundary. Bone-marrow tumour-cell assays predict poor outcome but do not establish viability, low cycling, ancestry or a useful action. Primary-tumour and end-stage spatial studies miss the disease-free interval. The resolving experiment must join those times in one patient and outperform stage plus ctDNA before any dormancy intervention is funded.", "Question 17 prevents age from becoming a treatment exclusion or a false safety claim. GIANT rejects a large difference between two reduced-intensity regimens, while GrantPax shows that assessment labels can be ignored. The missing experiment tests whether named support completed within 72 hours reduces severe toxicity and increases days alive outside hospital without delaying or covertly denying treatment.", "Question 6 now has a specific surgery branch: which initially locally advanced responder gains useful life from resection itself. Technical-removal calculators and survival calculators answer different questions. NEOLAP, PREOPANC-4, TAPS and live duration studies do not randomize surgery against continued treatment in the same response state. A clean cut edge, resection rate or selected survival gap cannot close this question.", "Question 22 separates a physical delivery mechanism from a survival claim. NCT04821284 reports no overall PFS or OS difference in its accepted abstract, while the gemcitabine subgroup lacks a denominator, uncertainty and interaction test. The resolving work is full trial release, planned interaction analysis and a direct or separately validated tumour-drug measure. Perfusion direction in 10/13 people is exploratory and cannot stand in for drug exposure.", "Question 23 treats siRNA as a six-gate delivery system: depot exit, tissue transport, tumour-cell entry, endosomal escape, RISC loading and matched RNA cutting. SIL204's public Segment 1 measures plasma concentration but does not make multi-region tumour knockdown a release gate. Use the existing trial, retain failed samples and stop support if public target data, spatial agreement, systemic exposure or chemotherapy preservation fails.", "Question 24 separates a real virus-delivery result from a confirmed survival result. VCN-01 was found in five of six sampled tumours and VIRAGE favoured virus plus chemotherapy, but 112 enrolled becomes a 96-person FAS and public alpha rules conflict. Recover the complete allocation and outcome flow, then require all-randomized survival and a blinded independent confirmation. The six-person repeat-dose study can test safety and blood pharmacology only.", "Question 25 separates a selected biological route from a proved drug interaction. MTAP loss reaches about 28% of PDAC and two PRMT5 inhibitors show human activity, but vopimetostat combination data have no RAS-alone control. Resolve the added response, duration, survival, harm and dose intensity by randomizing the identical RAS inhibitor with or without vopimetostat inside exact MTAP-loss strata.", "Question 26 asks whether elraglusib causes a repeatable survival gain. The first randomized phase 2 remains positive when every weekly-or-control randomized person counts, but the effect shrinks from 2.9 to 1.7 median months and has no matching PFS or response result. Resolve allocation, missing outcomes, later treatment, quality-adjusted life and masking in the existing record before a blinded all-randomized phase 3.", "Question 27 asks whether WEE1 has a human tumour-defined state distinct from failed CDK4/6 and CHK1 routes. The only immediate answer sits in 12 completed azenosertib records originally intended as a safety lead-in. Recover paired tumours, six-month status, harm and the decision not to expand. Stop unless at least 10 pairs are usable, eight show checkpoint plus damage change and at least 5/12 remain progression-free at six months.", "Question 28 asks whether IBI343 converts a CLDN18.2 address into longer useful life in pancreatic cancer. The phase 1 marker contrast is real but small: 10/44 responses when at least 60% of tumour cells stained at least 1+, versus 0/12 below that analysis cut. Entry used a different cut, at least 40% staining at intensity 2+ or 3+, and the live 201-person phase 3 registry does not disclose its assay rule, power assumptions or data-sharing plan. Resolve overall and quality-adjusted survival, marker calibration, every randomized denominator and payload-specific harm in NCT07066098; do not create a duplicate cohort.", "Question 29 asks which engineered-cell architecture can repeat its pancreatic response. Surface-reading satri-cel has 4/24 uncontrolled responses. Mutation-fragment-reading TCR-T has two clear responses across seven reported G12D and G12V patients, while five did not respond and repeat G12V infusions failed. Mesothelin CAR-T found tissue without control, and its registry-to-paper count is 54 versus nine. Recover every route denominator and use the existing 30-person G12V study to join target presentation, tumour entry, cell state, persistence and durable response before any controlled survival trial.", "Question 30 asks whether mitazalimab adds survival to mFOLFIRINOX. Its 57-person efficacy set reports 42.1% response and 14.9-month median survival, but no control; its registry now counts 94 without final results. A different CD40 drug missed its historical survival gate. Resolve the question only after all 94 records are released and a solvent owner funds at least half of a roughly 450-person placebo-controlled survival trial.", "Question 31 asks whether losartan improves the complete localized-PDAC route. A 49-person uncontrolled programme reported 34 clean-margin resections and 31.4-month survival, while an 88-person randomized advanced study found no efficacy signal. NCT03563248 already randomized 168 localized patients across losartan and no-losartan groups. Publish its factorial treatment contribution, flow, treatment completion, surgery, recurrence, survival and harm; spend £0 on another cohort.", "Question 32 asks whether changing a measured pancreatic microbial function improves treatment benefit enough to outweigh toxicity, infection and resistance. PANORAMIX can test ciprofloxacin with placebo, drug measurements and resistant-organism tracking; FMTPanc can test stool-transfer safety against placebo. Require all assigned people, target change before outcome, survival or progression benefit and explicit harm limits. Fund no new cohort.", "Question 33 asks whether a PDAC vaccine package prevents recurrence and which component causes any benefit. IMCODE003 already compares personalized RNA vaccine plus atezolizumab plus mFOLFIRINOX with the same chemotherapy in an estimated 260 people. It can settle package benefit, but not vaccine-versus-atezolizumab contribution. Require every assigned person, the full tissue-to-vaccine manufacture route, disease-free and overall survival, and harm; an immune-responder subgroup cannot answer the question.", "Question 34 asks whether quemliclustat adds useful survival to first-line chemotherapy. ARC-8 joined human target change to an encouraging external-control survival estimate, while an independent CD73 antibody failed overall. PRISM-1 already holds the clean answer: about 610 people, identical chemotherapy, placebo, quadruple masking and all-randomized survival. Do not substitute post-treatment NR4A change, a marker subgroup or another cohort for that result.", "Question 35 asks whether the TIGeR-PaC replacement package extends life or preserves it with less total harm after induction. The existing final result can answer that narrow use question. It cannot isolate delivery because the experimental arm also removes nab-paclitaxel and changes schedule, and it cannot prove target-tissue exposure because the 16-person study sampled blood only. Keep package benefit, procedure harm and platform mechanism separate.", "Question 36 asks whether adding irreversible electroporation after three months of mFOLFIRINOX extends useful life. CROSSFIRE compared it with radiotherapy, while the DIRECT registry selects device patients later and mostly reconstructs its small control group. NCT03899636 already has the correct same-chemotherapy randomization but no public actual count or result after its estimated completion. Recover every assignment, outcome, harm and dated plan; use completed LAP-PIE only to judge feasibility if the phase 3 record is unusable.", "Question 37 asks whether Optune Pax's two-month survival gain will reproduce, who can sustain its daily burden and what price the benefit justifies. PANOVA-3 is positive and supports the narrow FDA label, but PFS did not improve, follow-up missed quality of life after local progression and no independent randomized trial exists. Two cost models estimate 0.148 to 0.34 extra quality-adjusted life-years yet both find poor value. Resolve this with participant-level reanalysis, an all-eligible care route and manufacturer-funded modern-backbone replication before any wider use.", "Question 38 asks whether measured regional heating adds benefit to otherwise identical current care. HEAT missed its main endpoint and changed at least five causal terms in one comparison. HEATPAC was built to isolate heat but has no public result and a registry stale since 2017. Recover both randomized records before enrolling another person; require assignment, death, harm and temperature accounting and publish loss of data as a result.", "Question 39 asks whether proton or carbon-ion beams improve useful life over current photon treatment or systemic treatment alone. Beam geometry is established; comparative patient benefit is not. The only planned phase 3 comparison enrolled zero, and current cohorts leave local control separated from distant failure. Resolve delivered dose, severe harm, chemotherapy preservation, quality of life, travel and survival under identical care before facility or trial capital.", "Question 40 asks whether SMART adds useful life over continued chemotherapy and whether later surgery adds or exposes harm. The selected single arm cannot answer either. LAP-ABLATE never began on its stale record after its machine sponsor failed, and MASPAC stopped at nine. Recovery needs a brand-independent performance rule, every denominator, unfiltered harm and identical systemic care.", "Question 44 asks whether endoscopic radiofrequency treatment adds useful life after HybridTherm's negative randomized result. Resolve existing assignments, treatment delivery, pain, survival and harm before a new cohort.", "Question 45 asks whether histotripsy can reach enough of a pancreatic tumour without injuring bowel, stomach or vessels. Resolve the first independent procedures with planned-versus-delivered volume and all harms.", "Question 46 asks whether cryoablation improves pain and useful days over otherwise identical care. Selection and loss before repeat procedures must remain in the denominator.", "Question 47 asks whether microwave ablation adds patient benefit beyond its bundled artery treatment and chemotherapy. Complete imaging and one-component comparisons are required.", "Question 48 asks whether phosphorus-32 improves survival or only local control. TRIPP-FFX's all-assigned randomized result is the next answer.", "Question 49 asks whether photodynamic treatment covers enough viable tumour to matter. Map light dose, drug exposure, untreated volume, recurrence and harm in the same patient.", "Question 50 asks whether pancreatic laser ablation has a reproducible safe route after fistula, accrual and device-continuity failures. Recover the existing treated and screened denominators first.", "Question 51 asks whether ASP-1929 reaches EGFR-positive pancreatic tumour, receives enough light and kills viable tumour without unacceptable injury. The first ten evaluable patients are the gate.", "Question 52 asks which part of K-912 sonodynamic treatment causes any observed effect. Separate ultrasound, drug and chemotherapy contribution before expansion.", "Question 53 asks whether PIPAC improves quality-adjusted survival rather than selecting people fit enough for repeat procedures. Count every enrolled patient and every stop reason.", "Question 54 asks whether HIPEC adds survival in peritoneal pancreatic disease after a small trial changed local recurrence but not survival. Resolve the live study and systemic-failure boundary.", "Question 55 asks whether nerve ablation's pain effect outweighs any survival, opioid, procedure and neurological harm. Use death-aware analysis of existing randomized records.", "Question 56 asks which cells undergo ferroptosis in human pancreatic tumours and surrounding pancreas. Tumour exposure, cell type, membrane damage and clinical outcome must align.", "Question 57 asks whether routine RNA fusion testing finds NRG1-positive patients fast enough to deliver treatment and whether response becomes durable control. Measure the entire test-to-access route.", "Question 58 asks whether needs-triggered palliative care adds quality-weighted days alive and at home without withholding useful cancer treatment. Release the completed Danish trial before building another route.", "Question 59 asks whether pancreatic DPYD and UGT1A1 dosing reduces severe harm without delaying treatment or losing cancer control. Join every intended patient to result timing, exact dose, escalation, toxicity, admission, early death and tumour outcome; keep negative-test failures and ancestry coverage visible.", "Question 60 asks how many KRAS-wild-type pancreatic patients move from adequate tissue through valid DNA, reflex RNA, target-specific review and actual treatment before fitness is lost. NTRK, RET, ALK, ROS1, FGFR2, BRAF, ERBB2, MSI-high and NRG1 remain separate routes and outcomes.", "Question 61 asks which baseline pancreatic patient state gains useful days, treatment delivery or fewer complications from a defined exercise route. Recover randomized records, retain death and non-completion, derive the selector once and require it to repeat independently before another broad trial." ] } ], "sources": [ { "id": "abou-alfa-2006-exatecan-pdac", "type": "paper", "title": "Randomized phase III study of exatecan and gemcitabine compared with gemcitabine alone in untreated advanced pancreatic cancer", "authors": [ "Abou-Alfa, Ghassan K.", "Letourneau, Roger", "Harker, W. G.", "et al." ], "publisher": "Journal of Clinical Oncology", "date": "2006-09-20", "url": "https://pubmed.ncbi.nlm.nih.gov/16983112/", "doi": "10.1200/JCO.2006.07.0201", "pmid": "16983112", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Primary randomized phase 3 payload control. Exatecan plus gemcitabine versus gemcitabine alone in 349 assigned people produced median OS 6.7 versus 6.2 months, P=.52, response 6.9% versus 5.2%, and more severe neutropenia and thrombocytopenia. Targeted delivery could change the payload's concentration and therapeutic ratio, so this rejects free exatecan addition rather than IBI343." }, { "id": "abou-alfa-2020-claridhy", "type": "paper", "title": "Ivosidenib in IDH1-mutant, chemotherapy-refractory cholangiocarcinoma (ClarIDHy)", "authors": [ "Abou-Alfa, Ghassan K.", "Macarulla, Teresa", "Javle, Milind M.", "et al." ], "publisher": "The Lancet Oncology", "date": "2020-05-13", "url": "https://pubmed.ncbi.nlm.nih.gov/32416072/", "doi": "10.1016/S1470-2045(20)30157-1", "pmid": "32416072", "trialId": "NCT02989857", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Randomised double-blind phase 3 in 185 previously treated IDH1-mutant cholangiocarcinoma patients. Placebo-to-ivosidenib crossover was permitted. Funded by Agios." }, { "id": "abou-alfa-2020-pemigatinib", "type": "paper", "title": "Pemigatinib for previously treated, locally advanced or metastatic cholangiocarcinoma", "authors": [ "Abou-Alfa, Ghassan K.", "Sahai, Vaibhav", "Hollebecque, Antoine", "et al." ], "publisher": "The Lancet Oncology", "date": "2020-03-20", "url": "https://pubmed.ncbi.nlm.nih.gov/32203698/", "doi": "10.1016/S1470-2045(20)30109-1", "pmid": "32203698", "trialId": "NCT02924376", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Open-label single-arm phase 2 at 146 sites; 107 participants had FGFR2 fusions or rearrangements. Funded by Incyte." }, { "id": "abou-alfa-2026-proof301", "type": "paper", "title": "Phase III trial of infigratinib versus gemcitabine/cisplatin in adults with advanced cholangiocarcinoma with FGFR2 gene fusion or rearrangement: results and reflections on early termination of PROOF 301", "authors": [ "Abou-Alfa, Ghassan K.", "Borbath, Ivan", "Roychowdhury, Sameek", "et al." ], "publisher": "ESMO Open", "date": "2026-04-01", "url": "https://pubmed.ncbi.nlm.nih.gov/41850040/", "doi": "10.1016/j.esmoop.2026.106306", "pmid": "41850040", "trialId": "NCT03773302", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Randomized first-line phase 3 trial for advanced FGFR2-rearranged cholangiocarcinoma. It screened 1,127 people at 120 sites over 40 months but randomized only 48 of about 300 planned, then stopped for poor recruitment. Median progression-free survival was 7.4 months with infigratinib and 8.0 months with gemcitabine-cisplatin; grade 3-4 adverse events were 79.3% and 58.8%. The small truncated comparison cannot establish equivalence or superiority. It is direct evidence that a rare molecular trial needs a global identification and referral system before it needs another drug arm." }, { "id": "abrams-2026-rtog0848", "type": "paper", "title": "Adjuvant chemotherapy plus or minus chemoradiotherapy for adenocarcinoma of the pancreatic head: results of the radiotherapy random assignment of NRG Oncology/RTOG 0848", "authors": [ "Abrams, Ross A.", "Winter, Kathryn A.", "Goodman, Karyn A.", "et al." ], "publisher": "Journal of Clinical Oncology", "date": "2026-08-20", "url": "https://pubmed.ncbi.nlm.nih.gov/42456089/", "doi": "10.1200/JCO-25-02520", "pmid": "42456089", "trialId": "NCT01013649", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Multicentre phase 3 randomization after resection and five chemotherapy cycles without progression. Among 354 people, adding fluoropyrimidine-sensitized radiation did not improve overall survival, HR 0.96, 90% CI 0.79-1.18, p=0.77 two-sided, or disease-free survival. Grade 3 toxicity increased from 19% to 38%. A node-negative subgroup interaction was reported and requires confirmation with current systemic treatment rather than promotion to a general rule." }, { "id": "abramson-2024-alphafold3", "type": "paper", "title": "Accurate structure prediction of biomolecular interactions with AlphaFold 3", "authors": [ "Abramson, Josh", "Adler, Jonas", "Dunger, Jack", "et al." ], "publisher": "Nature", "date": "2024-05-08", "url": "https://pubmed.ncbi.nlm.nih.gov/38718835/", "doi": "10.1038/s41586-024-07487-w", "pmid": "38718835", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Primary model paper for joint structures containing proteins, nucleic acids, small molecules, ions and modified residues. Used as interaction-prediction evidence, not evidence of drug efficacy." }, { "id": "actuate-2026-q2-10q", "type": "company-filing", "title": "Actuate Therapeutics quarterly report for the period ended 30 June 2026", "authors": [ "Actuate Therapeutics, Inc." ], "publisher": "U.S. Securities and Exchange Commission", "date": "2026-08-14", "url": "https://www.sec.gov/Archives/edgar/data/1652935/000168316826006472/actuate_i10q-063026.htm", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Primary current finance and listing-risk source. Cash was $4.426m and working capital $0.943m at 30 June 2026; six-month operating cash use was $9.422m. The unchanged-burn ratio is 4.426/(9.422/6)=2.82 months, not company guidance. Management said cash would not fund operations beyond September without new capital and reported substantial going-concern doubt. The filing also records a Nasdaq market-value compliance deadline of 11 January 2027." }, { "id": "adam-2024-transmet", "type": "paper", "title": "Liver transplantation plus chemotherapy versus chemotherapy alone in patients with permanently unresectable colorectal liver metastases (TransMet)", "authors": [ "Adam, René", "Piedvache, Céline", "Chiche, Laurence", "et al." ], "publisher": "The Lancet", "date": "2024-09-21", "url": "https://pubmed.ncbi.nlm.nih.gov/39306468/", "doi": "10.1016/S0140-6736(24)01595-2", "pmid": "39306468", "trialId": "NCT02597348", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "European open-label randomised trial in 94 highly selected patients aged 18 to 65 with BRAF-non-mutated, chemotherapy-responsive, liver-only disease and no extrahepatic cancer. Publicly funded." }, { "id": "adenocyte-company-linfu", "type": "company", "title": "Adenocyte — LINFU platform and company history", "authors": [ "Adenocyte" ], "publisher": "Adenocyte", "date": "2026-09-15", "url": "https://www.adenocyte.com/what-we-do-1", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Company source states a 2018 founding and describes LINFU as microbubble-assisted low-intensity non-focused ultrasound intended to increase pancreatic duct-cell release into pancreatic fluid for cytology. It is a technology and origin source, not independent evidence of accuracy or benefit." }, { "id": "adler-2026-sonoporation-perfusion", "type": "paper", "title": "Perfusion Assessment in CEUS Imaging for Estimating Pancreatic Cancer Response to Sonoporation-Enhanced Chemotherapy", "authors": [ "Omri Adler", "Priscilla Machado", "Trang Vu", "Flemming Forsberg", "Odd Helge Gilja", "Dan Adam" ], "publisher": "Ultrasonic Imaging", "date": "2026-01", "url": "https://pubmed.ncbi.nlm.nih.gov/41133726/", "doi": "10.1177/01617346251367758", "pmid": "41133726", "trialId": "NCT04821284", "patentId": null, "grantId": "R01 CA199646", "accessed": "2026-09-15", "notes": "Exploratory perfusion analysis from 13 NCT04821284 participants. Imaging direction agreed with reported clinical direction in 10 of 13 and disagreed in three; this is not a randomized treatment-effect result or a direct measure of tumour drug concentration." }, { "id": "aguirre-2026-ignite-tx", "type": "paper", "title": "A multilevel intervention to identify individuals for genetic testing and treatment: the IGNITE-TX pilot randomized clinical trial", "authors": [ "Jorgensen, Kelsey A.", "Agusti, Nuria", "Coffin, Thomas", "et al." ], "publisher": "JAMA Network Open", "date": "2026-08-03", "url": "https://pubmed.ncbi.nlm.nih.gov/42658496/", "doi": "10.1001/jamanetworkopen.2026.30668", "pmid": "42658496", "trialId": "NCT05677048", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "US pan-cancer 2x2 factorial pilot. Sixty people with hereditary breast/ovarian or Lynch syndromes and 144 relatives were enrolled. Six-month cascade-test completion was 1/29 with printed material, 8/39 with free remote counselling/testing, 17/30 with proactive navigator plus bilingual digital support, and 36/46 when both were combined. The large pilot effect supports proactive consented contact and removal of cost and language barriers; it is not PDAC-specific and does not establish cancer prevention." }, { "id": "alagesan-2026-sig12d-loder", "type": "paper", "title": "A Phase II Trial of an Extended-Release siRNA Implant Targeting KRASG12D/V in Locally Advanced Pancreatic Cancer", "authors": [ "Alagesan, Brinda", "Varghese, Anna M.", "Ang, Celina", "et al." ], "publisher": "Clinical Cancer Research", "date": "2026-03-16", "url": "https://pubmed.ncbi.nlm.nih.gov/41528981/", "doi": "10.1158/1078-0432.CCR-25-3189", "pmid": "41528981", "trialId": "NCT01676259", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Two-cohort, open-label phase 2 study with 59 patients. Cohort 1 randomized an unselected population; cohort 2 was single-arm. Two authors were affiliated with Silexion Therapeutics. The mutation-selected analysis was exploratory and had a very wide confidence interval." }, { "id": "alistar-2017-devimistat-phase1", "type": "paper", "title": "Safety and tolerability of CPI-613 with modified FOLFIRINOX in metastatic pancreatic cancer: phase 1", "authors": [ "Alistar, Angela", "Morris, Bonny B.", "Desnoyer, Rodwige", "et al." ], "publisher": "The Lancet Oncology", "date": "2017-06-01", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC5635818/", "doi": "10.1016/S1470-2045(17)30314-5", "pmid": "28495639", "trialId": "NCT01835041", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Single-centre, open-label, dose-escalation study of 20 people. Among 18 treated at the selected 500 mg/m2 dose, 11, or 61%, had a complete or partial response. The study was built to set dose and safety, had no concurrent control, and explicitly said the activity required validation. It supplied the early signal later tested by AVENGER 500." }, { "id": "alligator-2026-q2", "type": "company-filing", "title": "Alligator Bioscience financial results for 1 January to 30 June 2026 and business update", "authors": [ "Alligator Bioscience" ], "publisher": "Alligator Bioscience", "date": "2026-08-26", "url": "https://alligatorbioscience.se/en/mfn_news/alligator-bioscience-ab-reports-financial-results-for-the-period-1-january-30-june-2026-and-provides-a-business-update/", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Issuer report. Cash was SEK16.6m at 30 June 2026 after first-half cash flow of minus SEK45.5m. On 23 July the company stopped all independent mitazalimab development and phase 3 support, planned to reduce to minimum staffing, and sought an out-license or sale. A proposed SEK125.6m rights issue was only 47% covered by commitments and guarantees; SEK19m bridge loans were added." }, { "id": "almoguera-1988-kras-pdac", "type": "paper", "title": "Most human carcinomas of the exocrine pancreas contain mutant c-K-ras genes", "authors": [ "Almoguera, C", "Shibata, D", "Forrester, K", "Martin, J", "Arnheim, N", "Perucho, M" ], "publisher": "Cell", "date": "1988-05-20", "url": "https://pubmed.ncbi.nlm.nih.gov/2453289/", "doi": "10.1016/0092-8674(88)90571-5", "pmid": "2453289", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Primary tumour study: 21 of 22 exocrine pancreatic carcinomas carried a codon-12 c-K-ras mutation; seven matched metastases retained it. Small historical cohort and histological naming predates current PDAC classification. It established prevalence, not druggability." }, { "id": "andel-2025-routine-imaging", "type": "paper", "title": "Routine Imaging or Symptomatic Follow-Up After Resection of Pancreatic Adenocarcinoma", "authors": [ "Andel, Pepijn C M", "van Goor, Iris W J M", "Augustinus, Simone", "et al." ], "publisher": "JAMA Surgery", "date": "2025-01-01", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC11541741/", "doi": "10.1001/jamasurg.2024.5024", "pmid": "39504033", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Prospective international cross-sectional study of 333 patients with PDAC recurrence from 33 centres in 13 countries. Routine imaging was associated with recurrence-focused treatment, adjusted OR 2.57, and overall survival from surgery, adjusted HR 0.75; median overall survival was 28 versus 23 months. Follow-up strategy was not randomized, and inclusion was conditioned on diagnosed recurrence, so selection, centre practice and earlier detection can explain some or all of the association. Authors disclosed limited outside grants and fees; no direct product test." }, { "id": "ando-2019-smoking-japan", "type": "paper", "title": "Smoking and Pancreatic Cancer Incidence: A Pooled Analysis of 10 Population-Based Cohort Studies in Japan", "authors": [ "Ando, et al." ], "publisher": "Cancer Epidemiology, Biomarkers & Prevention", "date": "2019-05-01", "url": "https://pubmed.ncbi.nlm.nih.gov/31113869/", "doi": "10.1158/1055-9965.EPI-18-1327", "pmid": "31113869", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Pooled primary-data analysis of 10 Japanese cohorts, 354,154 people, 4.70 million person-years and 1,779 pancreatic cancers. Sex difference in risk attenuation after cessation may reflect power or residual differences." }, { "id": "andre-2025-keynote177-five-year", "type": "paper", "title": "Pembrolizumab versus chemotherapy in MSI-high or mismatch-repair-deficient metastatic colorectal cancer: five-year KEYNOTE-177 follow-up", "authors": [ "Andre, Thierry", "et al." ], "publisher": "Annals of Oncology", "date": "2024-11-30", "url": "https://pubmed.ncbi.nlm.nih.gov/39631622/", "doi": "10.1016/j.annonc.2024.11.012", "pmid": "39631622", "trialId": "NCT02563002", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Five-year follow-up of randomized phase 3 KEYNOTE-177, 307 first-line MSI-high or mismatch-repair-deficient metastatic colorectal cancer patients. Median OS 77.5 versus 36.7 months, HR 0.73, 95% CI 0.53-0.99; five-year OS 54.8% versus 44.2%. Median response duration 75.4 versus 10.6 months. Effective checkpoint crossover from chemotherapy was 62%, complicating OS attribution. Merck funded the trial." }, { "id": "angiodynamics-2026-10k", "type": "financial", "title": "AngioDynamics annual report for the year ended May 31, 2026", "authors": [ "AngioDynamics, Inc." ], "publisher": "U.S. Securities and Exchange Commission", "date": "2026-07-10", "url": "https://www.sec.gov/Archives/edgar/data/1275187/000162828026048138/ango-20260531.htm", "doi": null, "pmid": null, "trialId": "NCT03899636", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Primary company filing. AngioDynamics reported $320.2m annual revenue, $36.7m net loss, $3.1m cash provided by operations, $53.9m cash and no outstanding debt at 31 May 2026. NanoKnife sales increased $8.6m and the Med Tech segment containing NanoKnife grew 18.4%. The filing says full paralytic anaesthesia is required for safe IRE delivery. The company has sufficient commercial capacity to fund its product evidence; mission subsidy or equity is not justified." }, { "id": "arcus-2026-q2-10q", "type": "financial", "title": "Arcus Biosciences quarterly report for the period ended June 30, 2026", "authors": [ "Arcus Biosciences, Inc." ], "publisher": "U.S. Securities and Exchange Commission", "date": "2026-08-06", "url": "https://www.sec.gov/Archives/edgar/data/1724521/000172452126000044/rcus-20260630.htm", "doi": null, "pmid": null, "trialId": "NCT06608927", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Primary financial and programme source. Arcus reported $775m cash, cash equivalents and marketable securities at June 30, 2026 and $257m net operating cash use for the first six months. It says resources cover planned operations into at least the second half of 2028. PRISM-1 enrollment completed in September 2025; the August results release says results are expected in the first half of 2027. Gilead licenses quemliclustat and Taiho participates in PRISM-1." }, { "id": "aronchik-2026-daraxonrasib-resistance", "type": "paper", "title": "Acquired resistance to the RAS(ON) multi-selective inhibitor daraxonrasib guides rational combination therapy strategies in pancreatic cancer", "authors": [ "Aronchik, Ida", "Kar, Sumit", "Zhuang, Yongxian", "et al." ], "publisher": "Nature Medicine", "date": "2026-08-11", "url": "https://pubmed.ncbi.nlm.nih.gov/42581230/", "doi": "10.1038/s41591-026-04537-w", "pmid": "42581230", "trialId": "NCT05379985", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Paired pretreatment and end-of-treatment ctDNA study from 44 phase 1/2 daraxonrasib patients. Treatment-emergent RAS-pathway alterations appeared in 26/44, 59%: mutant KRAS amplification 16/44, 36%; MAPK alterations 11/44, 25%; RTK and PI3K alterations 4/44 each, 9%. No acquired secondary KRAS mutations were seen. Models supported concordant mechanisms and combinations with DNA-damage-response, RTK or G12D-selective agents, but no randomized combination benefit was tested. Most laboratory authors were current or former Revolution Medicines employees with equity, and several were named inventors on company patent applications." }, { "id": "arpa-h-adapt-2026", "type": "government-programme", "title": "ARPA-H Advanced Analysis for Precision Cancer Therapy programme", "authors": [ "Advanced Research Projects Agency for Health" ], "publisher": "ARPA-H", "date": "2024-03-07", "url": "https://arpa-h.gov/news-and-events/arpa-h-launches-program-transform-cancer-care", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Primary programme announcement. ADAPT funds therapy-recommendation methods, an evolutionary clinical trial and a treatment-and-analysis platform for cancers that change under therapy. Existing programme rather than an open general grant; a new PDAC proposal must avoid duplicating funded work and use the rolling office route only for a distinct problem." }, { "id": "arpa-h-open-funding-2026", "type": "funder-guidance", "title": "ARPA-H open funding opportunities", "authors": [ "Advanced Research Projects Agency for Health" ], "publisher": "ARPA-H", "date": "2026-09-14", "url": "https://arpa-h.gov/explore-funding/open-funding-opportunities", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Primary live-opportunity page. Four Mission Office Innovative Solution Openings accept required concept summaries on a rolling basis: Health Science Futures, Proactive Health, Resilient Systems and Scalable Solutions. Programme-specific solicitations are time-limited. ARPA-H may reject concepts already covered by its portfolio. Snapshot only; availability changes." }, { "id": "asco-2019-pancreatic-susceptibility", "type": "guideline", "title": "Evaluating Susceptibility to Pancreatic Cancer: ASCO Provisional Clinical Opinion", "authors": [ "Stoffel, Elena M.", "McKernin, Shannon E.", "Brand, Randall", "et al." ], "publisher": "Journal of Clinical Oncology", "date": "2018-11-20", "url": "https://ascopubs.org/doi/10.1200/JCO.18.01489", "doi": "10.1200/JCO.18.01489", "pmid": "30457921", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Primary ASCO guidance. It recommends early germline-risk assessment and shared decision-making, explains that many mutation carriers lack a typical family history, and warns that a variant of uncertain significance is not evidence of inherited cancer risk. The panel states that randomized trial data were unavailable for these guidance statements." }, { "id": "asco-2024-germline-panel-guideline", "type": "guideline", "title": "Selection of Germline Genetic Testing Panels in Patients With Cancer: ASCO Guideline", "authors": [ "Tung, Nadine", "Ricker, Charité", "Messersmith, Hope", "et al." ], "publisher": "Journal of Clinical Oncology", "date": "2024-05-17", "url": "https://ascopubs.org/doi/10.1200/JCO.24.00662", "doi": "10.1200/JCO.24.00662", "pmid": "38759122", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Primary ASCO guideline on panel selection. It treats exocrine pancreatic cancer as an all-patient germline-testing indication, recommends multigene panels when several genes are relevant, and says tumour testing cannot replace germline testing because it misses some inherited variants. It does not prove that testing or family cascade delivery improves survival." }, { "id": "asge-2024-therapeutic-eus-guideline", "type": "guideline", "title": "ASGE guideline on therapeutic EUS in biliary tract disorders", "authors": [ "ASGE Standards of Practice Committee", "Pawa, Swati", "Marya, Neil B", "et al." ], "publisher": "American Society for Gastrointestinal Endoscopy", "date": "2024-12-01", "url": "https://www.asge.org/home/resources/publications/guidelines/american-society-for-gastrointestinal-endoscopy-guideline-on-the-role-of-therapeutic-eus-in-the-management-of-biliary-tract-disorders--summary-and-recommendations", "doi": "10.1016/j.gie.2024.03.027", "pmid": "39078360", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Primary society guideline using GRADE. After failed ERCP it conditionally suggests EUS-guided rather than percutaneous drainage, based on low-quality evidence, with route selection driven by anatomy, stability, anaesthesia and local expertise. Several panel members disclosed device-company relationships. The guideline does not establish EUS as the universal first route for resectable PDAC." }, { "id": "ash-2023-primary-vte-prophylaxis-guide", "type": "guideline", "title": "Primary prophylaxis of venous thromboembolism in patients with cancer: pocket guide", "authors": [ "Wang, Tzu-Fei", "Schaefer, Jordan K.", "Carrier, Marc" ], "publisher": "American Society of Hematology", "date": "2023-10-01", "url": "https://www.hematology.org/-/media/hematology/files/clinicians/guidelines/vte/vte-update-03_24/19894_vte-prophylaxis-pocket-guide-8-panel___0001_high_res_proof-(2).pdf", "doi": "10.1182/bloodadvances.2020003442", "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Current ASH pocket guide based on the 2021 guideline. For high-risk ambulatory patients receiving systemic therapy it conditionally suggests LMWH, apixaban or rivaroxaban over no prophylaxis. Risk tools must be complemented by clinical judgment and bleeding risk; the recommendation is not a universal treatment order." }, { "id": "asombang-2026-africa-hpb-registry", "type": "journal-article", "title": "Establishing cancer registries in Africa — focus on hepatopancreatobiliary cancers", "authors": [ "Asombang, Akwi W", "Antwi, Samuel O", "Omonisi, Abidemi", "Nartey, Yvonne A", "Desalegn, Hailemichael", "Okeke, Edith", "Rubagumya, Fidel", "Africa HepatoPancreatoBiliary Cancer Consortium", "Roberts, Lewis R" ], "publisher": "Nature Reviews Gastroenterology & Hepatology", "date": "2026-02", "url": "https://pubmed.ncbi.nlm.nih.gov/41249658/", "doi": "10.1038/s41575-025-01138-9", "pmid": "41249658", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Roadmap from the Africa HepatoPancreatoBiliary Cancer Consortium. It says African HPB risk factors, tumour features, treatment and survival are poorly characterized and argues for clinical and population registries. Authors span Nigeria, Ghana, Ethiopia, Rwanda, Burkina Faso, Egypt, Zambia and US partners. It does not report a complete continent-wide registry." }, { "id": "assawasirisin-2025-long-term-ipmn", "type": "paper", "title": "Unraveling the Long-term Natural History of Branch Duct Intraductal Papillary Mucinous Neoplasm: Beyond 10 years", "authors": [ "Assawasirisin, Charnwit", "Fagenholz, Peter", "Qadan, Motaz", "et al." ], "publisher": "Annals of Surgery", "date": "2024-09-10", "url": "https://pubmed.ncbi.nlm.nih.gov/39253809/", "doi": "10.1097/SLA.0000000000006535", "pmid": "39253809", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Retrospective Massachusetts General Hospital cohort selected for at least 10 years without intervention, n=316, median follow-up 13.5 years. Pancreatic malignancy occurred in 8.2%. Long-surviving, non-operated selection limits comparison with newly diagnosed cysts." }, { "id": "astellas-2026-gleam-medical-information", "type": "company-report", "title": "Astellas medical information summary of the GLEAM trial", "authors": [ "Astellas Pharma US, Inc." ], "publisher": "Astellas Medical Information", "date": "2026-06-16", "url": "https://www.astellasanswers.com/us/zolbetuximab/information-on-zolbetuximab-and-gleam-trial", "doi": null, "pmid": null, "trialId": "NCT03816163", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Sponsor medical-information page reproducing the final GLEAM presentation. It reports no OS or PFS gain, a 7.3-point ORR increase, two-month longer response duration, serious treatment-emergent events in 73.7% versus 53.9%, and permanent discontinuation in 37.1% of the zolbetuximab arm, most often from nausea or vomiting." }, { "id": "astellas-2026-setidegrasib-phase3", "type": "company-release", "title": "Astellas doses first patient in phase 3 study of setidegrasib for KRAS G12D-mutated metastatic pancreatic ductal adenocarcinoma", "authors": [ "Astellas Pharma Inc." ], "publisher": "Astellas Pharma Inc.", "date": "2026-04-15", "url": "https://newsroom.astellas.com/2026-04-15-astellas-doses-first-patient-in-phase-3-study-of-setidegrasib-for-kras-g12d-mutated-metastatic-pancreatic-ductal-adenocarcinoma", "doi": null, "pmid": null, "trialId": "NCT07409272", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Sponsor announcement used to verify that setidegrasib was discovered internally and that the randomized, double-blind first-line phase 3 began dosing. Efficacy comes from the paper and registry, not this release." }, { "id": "astrazeneca-2025-annual-report", "type": "company", "title": "Annual Report and Form 20-F Information 2025", "authors": [ "AstraZeneca PLC" ], "publisher": "AstraZeneca PLC", "date": "2026-02-10", "url": "https://www.astrazeneca.com/investor-relations/annual-reports.html", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Official annual-report index confirming the 2025 report. Company-wide revenue, R&D and pipeline values are not assigned to durvalumab, biliary cancer or HCC without a product-specific disclosure." }, { "id": "astrazeneca-2025-history", "type": "company", "title": "A timeline of our company history", "authors": [ "AstraZeneca PLC" ], "publisher": "AstraZeneca PLC", "date": "2025-06-01", "url": "https://www.astrazeneca.com/our-company/history.html", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Company source used only for corporate history: AstraZeneca formed in 1999 from Astra AB and Zeneca Group. The page also reports group-wide facilities and investment plans, which cannot be assigned to biliary cancer or HCC." }, { "id": "atkin-2010-flexible-sigmoidoscopy", "type": "paper", "title": "Once-only flexible sigmoidoscopy screening in prevention of colorectal cancer", "authors": [ "Atkin, Wendy S.", "Edwards, Rob", "Kralj-Hans, Ivana", "et al." ], "publisher": "The Lancet", "date": "2010-04-27", "url": "https://pubmed.ncbi.nlm.nih.gov/20430429/", "doi": "10.1016/S0140-6736(10)60551-X", "pmid": "20430429", "trialId": "ISRCTN28352761", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "UK multicentre randomised trial of 170,432 people who had indicated willingness to be screened. Intention-to-treat attendance was 71%; median follow-up 11.2 years. Public and charity funded." }, { "id": "ausania-2019-pancreatic-prehab-rct", "type": "paper", "title": "Prehabilitation in patients undergoing pancreaticoduodenectomy: a randomized controlled trial", "authors": [ "Ausania, Fabio", "Senra, Paula", "Melendez, Reyes", "et al." ], "publisher": "Revista Espanola de Enfermedades Digestivas", "date": "2019-08-01", "url": "https://doi.org/10.17235/reed.2019.6182/2019", "doi": "10.17235/reed.2019.6182/2019", "pmid": "31232076", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Forty pancreatic or periampullary surgical patients were analysed after randomization: 18 multicomponent prehabilitation and 22 standard care. Overall and major complications did not differ. Delayed gastric emptying was 5.6% versus 40.9%, p=0.01, but it was secondary, the study was tiny and neoadjuvant patients were excluded." }, { "id": "aykut-2019-pdac-mycobiome", "type": "paper", "title": "The fungal mycobiome promotes pancreatic oncogenesis via activation of MBL", "authors": [ "Aykut, Berk", "Pushalkar, Smruti", "Chen, Raphael", "et al." ], "publisher": "Nature", "date": "2019-10-03", "url": "https://pubmed.ncbi.nlm.nih.gov/31578522/", "doi": "10.1038/s41586-019-1608-2", "pmid": "31578522", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Human tissue and mouse mechanism study. Fungal signal was about 3,000-fold higher in PDAC than normal pancreas and Malassezia was enriched. Antifungal removal, fungal replacement and mannose-binding-lectin pathway tests changed tumour growth in mice. No human antifungal treatment effect was tested." }, { "id": "aznar-2025-mesothelin-car-t-resistance", "type": "paper", "title": "Clinical and molecular dissection of CAR T cell resistance in pancreatic cancer", "authors": [ "Aznar, Maria A.", "Good, Christopher R.", "Barber-Rotenberg, Jennifer S.", "et al." ], "publisher": "Cell Reports Medicine", "date": "2025-09-16", "url": "https://pubmed.ncbi.nlm.nih.gov/40829595/", "doi": "10.1016/j.xcrm.2025.102301", "pmid": "40829595", "trialId": "NCT03323944", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Phase 1 translational study of nine infused patients using intravenous, intraperitoneal or hepatic-arterial mesothelin CAR T cells. One had stable disease at day 28 then progression; the other eight had progression at their first stated scan. Median PFS was 3.9 weeks and OS 6.7 weeks. Tissue showed low entry, expansion, receptor surface loss and exhaustion. The proposed ID3/SOX4 solution is still mouse evidence. The trial registry now counts 54 enrolled people, which this paper does not reconcile." }, { "id": "babic-2019-muscle-loss-survival", "type": "paper", "title": "Postdiagnosis loss of skeletal muscle, but not adipose tissue, is associated with shorter survival of patients with advanced pancreatic cancer", "authors": [ "Babic, Ana", "Rosenthal, Michael H.", "Bamlet, William R.", "et al." ], "publisher": "Cancer Epidemiology, Biomarkers and Prevention", "date": "2019-09-18", "url": "https://pubmed.ncbi.nlm.nih.gov/31533940/", "doi": "10.1158/1055-9965.EPI-19-0370", "pmid": "31533940", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Longitudinal CT body-composition analysis in 164 people with advanced pancreatic cancer. The quartile with greatest muscle loss had death HR 2.01 versus the quartile that gained muscle. Association does not show that preventing loss would extend survival." }, { "id": "baghdadi-2019-palbociclib-cdkn2a", "type": "paper", "title": "Palbociclib in patients with pancreatic and biliary cancer with CDKN2A alterations: results from the TAPUR study", "authors": [ "Al Baghdadi, Tarek", "Halabi, Susan", "Garrett-Mayer, Elizabeth", "et al." ], "publisher": "JCO Precision Oncology", "date": "2019-12-01", "url": "https://pubmed.ncbi.nlm.nih.gov/35100714/", "doi": "10.1200/PO.19.00124", "pmid": "35100714", "trialId": "NCT02693535", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Primary biomarker-selected basket report. Twelve pancreatic and ten biliary patients with CDKN2A loss or mutation received palbociclib. No evaluable patient responded or remained stable for 16 weeks, so both cohorts closed. Pancreatic median PFS was 7.2 weeks and OS 12.4 weeks. CDKN2A loss alone did not select useful CDK4/6 inhibition." }, { "id": "bai-2025-cispd1", "type": "paper", "title": "Neoadjuvant nab-paclitaxel plus gemcitabine followed by modified FOLFIRINOX for resectable pancreatic cancer", "authors": [ "Bai, Xueli", "Li, Xiang", "Chen, Yiwen", "et al." ], "publisher": "Cancer Cell", "date": "2025-12-08", "url": "https://pubmed.ncbi.nlm.nih.gov/41072417/", "doi": "10.1016/j.ccell.2025.09.009", "pmid": "41072417", "trialId": "NCT03750669", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Single-centre randomized phase 3 trial in Hangzhou, China, with 324 resectable pancreatic-cancer patients. Sequential neoadjuvant gemcitabine/nab-paclitaxel then modified FOLFIRINOX produced median event-free survival 15.3 versus 10.9 months, HR .71, 95% CI .54-.93, and median OS 35.4 versus 27.2 months, HR .73, 95% CI .53-1.00, nominal P=.0477. Grade 3 or worse adverse events were 47.6% versus 30.7%. Half the cohort had body or tail tumours. The complex sequence, single expert centre, different tumour-location mix and borderline OS result require outside reproduction." }, { "id": "ball-2017-transient-tic-clones", "type": "paper", "title": "Succession of transiently active tumor-initiating cell clones in human pancreatic cancer xenografts", "authors": [ "Ball, Christina R.", "Oppel, Felix", "Ehrenberg, Katharina R.", "et al." ], "publisher": "EMBO Molecular Medicine", "date": "2017-07-01", "url": "https://pubmed.ncbi.nlm.nih.gov/28526679/", "doi": "10.15252/emmm.201607354", "pmid": "28526679", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Genetically marked clones in serial human PDAC xenografts. Different tumour-initiating clones became active in successive passages, supporting plastic activation rather than one permanently active hierarchy. Xenografting imposes selection and is not direct lineage tracing in a patient." }, { "id": "bando-2026-altair", "type": "paper", "title": "Post-adjuvant chemotherapy in ctDNA-positive patients with resected colorectal cancer: a randomized phase 3 trial", "authors": [ "Bando, Hideaki", "Watanabe, Jun", "Takahashi, Yusuke", "et al." ], "publisher": "Nature Medicine", "date": "2026-07-01", "url": "https://pubmed.ncbi.nlm.nih.gov/42260101/", "doi": "10.1038/s41591-026-04428-0", "pmid": "42260101", "trialId": "NCT04457297", "patentId": null, "grantId": "19ck0106447h0002; 21ck0106711h0001; 21ck0106710h0001; 22lk0201164h0001; 22lk0201148h0002", "accessed": "2026-09-14", "notes": "ALTAIR randomized 243 patients who became tumour-informed ctDNA positive after standard treatment but had no radiological disease. Six months of trifluridine/tipiracil did not significantly improve disease-free survival and caused grade 3 or worse hematologic adverse events in 73.0% versus 3.3%. Six authors were Natera employees and several investigators reported assay or drug-company relationships." }, { "id": "bang-2019-celiac-rfa-versus-cpn-rct", "type": "paper", "title": "EUS-guided celiac ganglion radiofrequency ablation versus celiac plexus neurolysis for palliation of pain in pancreatic cancer: a randomized controlled trial", "authors": [ "Bang, Ji Young", "Sutton, Bryce", "Hawes, Robert H", "Varadarajulu, Shyam" ], "publisher": "Gastrointestinal Endoscopy", "date": "2018-08-16", "url": "https://doi.org/10.1016/j.gie.2018.08.005", "doi": "10.1016/j.gie.2018.08.005", "pmid": "30120957", "trialId": "NCT03152487", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Single-blind randomized 26-person comparison of heat ablation around visible celiac ganglia or plexus versus alcohol plexus neurolysis. The heat group reported less pain and better quality-of-life domains at two to four weeks. The study had 12 heat and 14 alcohol participants, no medication-only control and no durable survival result; it can rank two procedures only weakly." }, { "id": "barbosa-2024-eus-ercp-meta", "type": "paper", "title": "EUS- versus ERCP-guided biliary drainage for malignant obstruction: meta-analysis of randomized trials", "authors": [ "Barbosa, Eduardo Cerchi", "Espírito Santo, Paula Arruda", "Baraldo, Stefano", "et al." ], "publisher": "Gastrointestinal Endoscopy", "date": "2024-09-01", "url": "https://pubmed.ncbi.nlm.nih.gov/38648989/", "doi": "10.1016/j.gie.2024.04.019", "pmid": "38648989", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Random-effects meta-analysis of six randomized trials and 577 patients with malignant biliary obstruction. Technical and clinical success, stent patency, overall adverse events, cholangitis and survival were similar. EUS-guided drainage reduced reintervention, RR .57, pancreatitis, RR .15, tumour ingrowth or overgrowth, RR .28, and hospital stay by 1.03 days. Cancer types, resectability, devices and expert settings varied; the survival difference confidence interval spanned -34.23 to 43.40 days." }, { "id": "baretti-2024-entinostat-nivolumab-pdac", "type": "paper", "title": "Entinostat in combination with nivolumab in metastatic pancreatic ductal adenocarcinoma", "authors": [ "Baretti, Marina", "Danilova, Ludmila", "Durham, Jennifer N.", "et al." ], "publisher": "Nature Communications", "date": "2024-11-12", "url": "https://pubmed.ncbi.nlm.nih.gov/39532835/", "doi": "10.1038/s41467-024-52528-7", "pmid": "39532835", "trialId": "NCT03250273", "patentId": null, "grantId": "5P01CA247886", "accessed": "2026-09-15", "notes": "Primary PDAC phase 2 report. Three of 27 participants had partial responses, ORR 11% (95% CI 2.4-29.2), median response duration 10.2 months, PFS 1.89 months and OS 2.729 months. Grade 3 or worse treatment-related events occurred in 19 patients (63%). The main response endpoint was not met despite measured immune-state changes." }, { "id": "bastiaenen-2022-fancy", "type": "paper", "title": "Safety and economic analysis of selective histopathology following cholecystectomy: multicentre, prospective, cross-sectional FANCY study", "authors": [ "Bastiaenen, Vivian P.", "van Vliet, Jaap L. P.", "Bemelman, Willem A." ], "publisher": "British Journal of Surgery", "date": "2022-02-28", "url": "https://academic.oup.com/bjs/article/109/4/355/6538872", "doi": "10.1093/bjs/znab469", "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Full report checked. Prospective Dutch study included 10,041 cholecystectomy specimens. Routine histology found 28 malignancies; a surgeon-led selective policy would have omitted 7,846 specimens and missed eight malignancies, seven with clinical consequences. Only one of 28 patients had clear treatment benefit from complete resection of residual disease, while six experienced futile additional surgery. It evaluates pathology after the gallbladder is removed, not prevention or population screening." }, { "id": "baudin-2026-oclurandom", "type": "paper", "title": "177Lu-Lu-DOTA-TATE versus sunitinib in patients with metastatic progressive neuroendocrine tumours of the pancreas (OCLURANDOM)", "authors": [ "Baudin, Eric", "Durand, Alice", "Beron, Amandine", "et al." ], "publisher": "The Lancet Oncology", "date": "2026-06-01", "url": "https://pubmed.ncbi.nlm.nih.gov/42225102/", "doi": "10.1016/S1470-2045(26)00108-7", "pmid": "42225102", "trialId": "NCT02230176", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "French publicly funded randomized, open-label, non-comparative phase 2 in 84 pretreated, progressive, somatostatin-receptor-positive metastatic pNET patients. Cross-over was allowed." }, { "id": "baum-2022-fap2286-firsthuman", "type": "paper", "title": "Feasibility, Biodistribution, and Preliminary Dosimetry in Peptide-Targeted Radionuclide Therapy of Diverse Adenocarcinomas Using 177Lu-FAP-2286: First-in-Humans Results", "authors": [ "Baum, Richard P.", "Schuchardt, Christiane", "Singh, Aviral", "et al." ], "publisher": "Journal of Nuclear Medicine", "date": "2022-03-01", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC8978187/", "doi": "10.2967/jnumed.120.259192", "pmid": "34168013", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Retrospective compassionate-use report in 11 end-stage adenocarcinomas, including five pancreatic cancers, selected for FAP-positive imaging. Two patients had stable disease and nine progressed after the first cycle; the stable cases were breast cancers. Grade 3 events occurred in three people and no grade 4 event was reported. Kidney dose averaged 1.0 Gy/GBq and red-marrow dose 0.05 Gy/GBq; tumour dosimetry was based on 13 bone lesions and one liver lesion, not pancreatic primary tumours." }, { "id": "baum-2026-3bp3940", "type": "paper", "title": "Fibroblast Activation Protein-Targeted Radiopharmaceutical Therapy Using 177Lu-, 90Y-, and 225Ac-Labeled 3BP-3940: First Experience in 21 Different Advanced Malignancies", "authors": [ "Baum, Richard P.", "Zhao, Tianzhi", "Jakobsson, Vivianne", "et al." ], "publisher": "Journal of Nuclear Medicine", "date": "2026-06-01", "url": "https://pubmed.ncbi.nlm.nih.gov/42225357/", "doi": "10.2967/jnumed.125.271129", "pmid": "42225357", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Uncontrolled first experience: 88 people with 21 cancers received 227 cycles using three isotopes alone or in tandem; some also received standard treatments. Best-response data were reported for only 51 people: 34 complete or partial responses, 7 stable, 8 progressive and 2 mixed. Median survival for all 88 was seven months. One pancreatic-cancer dosimetry example gave 2.2 Gy to the primary tumour and 0.265 Gy to kidneys after 9.73 GBq, a tumour-to-kidney ratio of 8.3. The mixed diseases, mixed isotopes, other treatments and 37 people absent from the best-response denominator prevent a PDAC treatment-effect estimate." }, { "id": "bayne-2012-gmcsf-myeloid", "type": "paper", "title": "Tumor-derived granulocyte-macrophage colony-stimulating factor regulates myeloid inflammation and T cell immunity in pancreatic cancer", "authors": [ "Bayne, Lauren J.", "Beatty, Gregory L.", "Jhala, Nirag", "et al." ], "publisher": "Cancer Cell", "date": "2012-06-01", "url": "https://pubmed.ncbi.nlm.nih.gov/22698406/", "doi": "10.1016/j.ccr.2012.04.025", "pmid": "22698406", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Genetically engineered mouse evidence that tumour-derived GM-CSF recruits suppressive myeloid cells; human tumour cells expressed GM-CSF, but therapeutic efficacy was tested in mice." }, { "id": "bazan-peregrino-2021-vcn01-intratumour", "type": "paper", "title": "VCN-01 disrupts pancreatic cancer stroma and exerts antitumor effects", "authors": [ "Bazan-Peregrino, Miriam", "Garcia-Carbonero, Rocio", "Laquente, Berta", "et al." ], "publisher": "Journal for ImmunoTherapy of Cancer", "date": "2021-11-01", "url": "https://pubmed.ncbi.nlm.nih.gov/35149591/", "doi": "10.1136/jitc-2021-003254", "pmid": "35149591", "trialId": "NCT02045589", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Primary mechanism paper joining models to an intratumour PDAC phase 1 study. Post-treatment biopsy and secondary blood peaks detected VCN-01; serum PH20 rose and tumour stiffness fell. The small uncontrolled study supports human replication and stromal disruption, not survival benefit." }, { "id": "beaufils-2026-pancpraid", "type": "paper", "title": "Developing a Histology-Based Artificial Intelligence Biomarker to Predict Adjuvant Chemotherapy Benefit in Pancreatic Cancer", "authors": [ "Beaufils, Audrey", "De Martino, Julien", "Jiang, Xiaofeng", "et al." ], "publisher": "Journal of Clinical Oncology", "date": "2026-09-10", "url": "https://pubmed.ncbi.nlm.nih.gov/42507965/", "doi": "10.1200/JCO-26-00327", "pmid": "42507965", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Retrospective model development in 231 resected PDAC patients, followed by external analysis of 313 archived slides from the randomized PRODIGE-24 trial. The reported treatment interactions support a prospective treatment-selection test; retrospective reuse of a randomized trial does not itself prove that model-guided care improves outcome." }, { "id": "bekaii-saab-2023-adagrasib-solid-tumors", "type": "paper", "title": "Adagrasib in Advanced Solid Tumors Harboring a KRAS G12C Mutation", "authors": [ "Bekaii-Saab, Tanios S.", "Yaeger, Rona", "Spira, Alexander I.", "et al." ], "publisher": "Journal of Clinical Oncology", "date": "2023-04-26", "url": "https://pubmed.ncbi.nlm.nih.gov/37099736/", "doi": "10.1200/JCO.23.00434", "pmid": "37099736", "trialId": "NCT03785249", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Single-arm phase 2 cohort in rare KRAS G12C solid tumours other than lung and colorectal cancer. Mirati Therapeutics sponsored the trial; pancreatic and biliary subgroup counts were small." }, { "id": "bekaii-saab-2023-napabucasin", "type": "paper", "title": "Napabucasin plus nab-paclitaxel with gemcitabine versus nab-paclitaxel with gemcitabine in previously untreated metastatic pancreatic adenocarcinoma", "authors": [ "Bekaii-Saab, Tanios", "Okusaka, Takuji", "Goldstein, David", "et al." ], "publisher": "eClinicalMedicine", "date": "2023-03-16", "url": "https://pubmed.ncbi.nlm.nih.gov/36969338/", "doi": "10.1016/j.eclinm.2023.101897", "pmid": "36969338", "trialId": "NCT02993731", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Open-label phase 3 at 165 sites across 20 countries; 1,134 randomised patients. Terminated for futility. Funded by Sumitomo Pharma Oncology." }, { "id": "bekaii-saab-2026-fight302", "type": "paper", "title": "Pemigatinib for Unresectable or Metastatic Cholangiocarcinoma With Fibroblast Growth Factor Receptor-2 Rearrangement: Results From the Phase III FIGHT-302 Trial", "authors": [ "Bekaii-Saab, Tanios S.", "Melisi, Davide", "Wilmink, Hanneke", "et al." ], "publisher": "Journal of Clinical Oncology", "date": "2026-06-01", "url": "https://pubmed.ncbi.nlm.nih.gov/42223137/", "doi": "10.1200/JCO-26-00788", "pmid": "42223137", "trialId": "NCT03656536", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "First-line global randomised phase 3 in FGFR2-rearranged advanced cholangiocarcinoma. Closed early after 167 randomisations because the standard of care changed. Chemotherapy crossover to pemigatinib was allowed. Incyte employees were authors." }, { "id": "betge-2022-grantpax", "type": "paper", "title": "Neglected geriatric assessment and overtreatment of older patients with pancreatic cancer: GrantPax results", "authors": [ "Betge, Johannes", "Schulte, Nadine", "Belle, Sebastian", "et al." ], "publisher": "Journal of Geriatric Oncology", "date": "2022-06-01", "url": "https://pubmed.ncbi.nlm.nih.gov/34991998/", "doi": "10.1016/j.jgo.2021.12.018", "pmid": "34991998", "trialId": "NCT02812992", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Prospective open-label German phase 4 study in adults aged at least 70 with metastatic PDAC. It enrolled 32 of 135 planned at six centres and stopped for poor recruitment. Geriatric assessment classified 29/32 as Slow-Go and three as Go-Go, yet clinicians placed 15/32, 47%, on a treatment route inconsistent with the assessment. Serious adverse events affected 11/14, 78.6%, in the combination safety group and 8/15, 53.3%, in the gemcitabine group. Median time to quality-of-life deterioration was 13 versus 29 days. The groups were assigned by clinician choice rather than randomized, so treatment effects are not causal." }, { "id": "beutel-2021-pdac-organoids", "type": "paper", "title": "A prospective feasibility trial to challenge patient-derived pancreatic cancer organoids in predicting treatment response", "authors": [ "Beutel, Alica K.", "Schütte, Lena", "Scheible, Jeanette", "Roger, Elodie", "Müller, Martin", "Perkhofer, Lukas", "Kleger, Alexander" ], "publisher": "Cancers", "date": "2021-05-21", "url": "https://pubmed.ncbi.nlm.nih.gov/34064221/", "doi": "10.3390/cancers13112539", "pmid": "34064221", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Primary prospective feasibility study. Organoids were attempted from 54 patients; 28 pharmacotyping profiles were produced, 63.6%, after a median 53 days, range 21–126. Reported treatment-response prediction was exploratory and nonrandomized; performance fell to 40% in pretreated patients." }, { "id": "bever-2026-vaccine-checkpoint-pdac", "type": "paper", "title": "A Randomized Phase II Study: CRS207/GVAX plus Anti-PD-1 and Anti-CTLA4 Recruits Mesothelin- and mKRAS-Specific T Cells into PDAC", "authors": [ "Bever, Katherine M.", "Huff, Amanda L.", "Danilova, Ludmila", "et al." ], "publisher": "Cancer Immunology Research", "date": "2026-03-04", "url": "https://pubmed.ncbi.nlm.nih.gov/41524570/", "doi": "10.1158/2326-6066.CIR-25-0545", "pmid": "41524570", "trialId": "NCT03190265", "patentId": null, "grantId": "P30 CA006973; R50 CA243627; R21 CA264004; P01 CA247886; U01 CA253403", "accessed": "2026-09-14", "notes": "Randomized phase 2 comparison of CRS-207 with nivolumab and ipilimumab, with or without GVAX, in metastatic PDAC after chemotherapy. Fifty-seven patients received treatment; two partial responses occurred, both in the arm without GVAX, for 4% overall response, and response rates did not differ significantly. Grade 3 or worse related events occurred in 39/57, 68%. Antigen-specific T-cell clones reached tumours, but myeloid and regulatory T-cell signatures rose and correlated with poor response. Both arms received experimental immune combinations, so this was not a standard-care efficacy comparison." }, { "id": "biontech-2026-cevumeran-crc-termination", "type": "company-announcement", "title": "BioNTech provides update on phase 2 autogene cevumeran trial in resected colorectal cancer", "authors": [ "BioNTech SE" ], "publisher": "BioNTech SE", "date": "2026-08-28", "url": "https://www.biontech.com/au/en/home/mediaroom/news/statements/2026/08/BioNTech-Provides-Update-on-Phase-2-Clinical-Trial-of-Autogene-Cevumeran-in-Resected-Colorectal-Cancer.html", "doi": null, "pmid": null, "trialId": "NCT04486378", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Company source: BioNTech terminated the ctDNA-positive resected colorectal phase 2 monotherapy trial while stating that the separate PDAC combination trial continues. The public statement does not provide a mature randomized effect estimate." }, { "id": "blackford-2024-high-risk-surveillance", "type": "paper", "title": "Pancreatic Cancer Surveillance and Survival of High-Risk Individuals", "authors": [ "Blackford, Amanda L.", "Canto, Marcia Irene", "Dbouk, Mohamad", "et al." ], "publisher": "JAMA Oncology", "date": "2024-07-03", "url": "https://jamanetwork.com/journals/jamaoncology/fullarticle/2820691", "doi": "10.1001/jamaoncol.2024.1930", "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Comparative cohort: 26 high-risk individuals with PDAC versus 1,504 matched SEER controls. The cohort was small, all 26 were White, it was not randomised, and the authors ran lead-time sensitivity analyses." }, { "id": "blandino-2025-gallbladder-mirna", "type": "paper", "title": "Small-RNA sequencing reveals potential serum biomarkers for gallbladder cancer: Results from a three-stage collaborative study of large European prospective cohorts", "authors": [ "Blandino, Alice", "Scherer, Dominique", "Lorenzo Bermejo, Justo" ], "publisher": "European Journal of Cancer", "date": "2025-01-01", "url": "https://pubmed.ncbi.nlm.nih.gov/39579640/", "doi": "10.1016/j.ejca.2024.115138", "pmid": "39579640", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Report checked. Tissue discovery used 40 cancers and eight normal samples; serum screening used 37 prospectively collected case-control pairs and validation used 36 independent pairs. miR-4533 repeated across prospective cohorts, including samples obtained years before diagnosis in some cohorts, but there were only 73 cases in serum, no locked clinical classifier, no sensitivity or specificity for an action threshold and no evidence that acting on the marker improves outcome." }, { "id": "bogumil-2024-smoking-cessation-multiethnic", "type": "paper", "title": "Excess pancreatic cancer risk due to smoking and modifying effect of quitting smoking: The Multiethnic Cohort Study", "authors": [ "Bogumil, David", "Stram, Daniel", "Preston, Dale L.", "et al." ], "publisher": "Cancer Causes & Control", "date": "2024-03", "url": "https://pubmed.ncbi.nlm.nih.gov/37924460/", "doi": "10.1007/s10552-023-01811-x", "pmid": "37924460", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Prospective Multiethnic Cohort analysis of 182,011 people with 1,831 pancreatic cancers over mean follow-up of 19.3 years. Fifty pack-years was associated with 91% excess risk and each year since quitting with 9% lower excess risk. Quitting was observed, not randomized, and the model estimates risk trajectories rather than cancers prevented by a cessation programme." }, { "id": "boileve-2024-organopredict", "type": "paper", "title": "Organoids for Functional Precision Medicine in Advanced Pancreatic Cancer", "authors": [ "Boilève, Alice", "Cartry, Jérôme", "Goudarzi, Negaar", "Jaulin, Fanny" ], "publisher": "Gastroenterology", "date": "2024-10-01", "url": "https://pubmed.ncbi.nlm.nih.gov/38866343/", "doi": "10.1053/j.gastro.2024.05.032", "pmid": "38866343", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Primary prospective ORGANOPREDICT observational study in 87 pretreated advanced-PDAC patients. Fifty-four organoids were generated, 62%; mean drug-report time was 6.8 weeks; 34 patients had complete evaluable clinical follow-up. Reported sensitivity 83.3% and specificity 92.9%, but treatment was routine rather than randomly assigned from the assay." }, { "id": "boj-2015-pdac-organoids", "type": "paper", "title": "Organoid models of human and mouse ductal pancreatic cancer", "authors": [ "Boj, Sylvia F", "Hwang, Chang-Il", "Baker, Lindsey A", "Chio, Iok In Christine", "Engle, Dannielle D", "Tuveson, David A" ], "publisher": "Cell", "date": "2015-01-15", "url": "https://pubmed.ncbi.nlm.nih.gov/25557080/", "doi": "10.1016/j.cell.2014.12.021", "pmid": "25557080", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Primary platform paper establishing normal and neoplastic mouse and human pancreatic duct organoids for disease modelling. It created a durable experimental system but did not test treatment selection in patients." }, { "id": "botta-2024-tumor-informed-ctdna", "type": "paper", "title": "Association of personalized and tumor-informed ctDNA with patient survival outcomes in pancreatic adenocarcinoma", "authors": [ "Botta, Gregory P.", "Abdelrahim, Maen", "Drengler, Ronald L.", "et al." ], "publisher": "The Oncologist", "date": "2024-10-03", "url": "https://pubmed.ncbi.nlm.nih.gov/39022993/", "doi": "10.1093/oncolo/oyae155", "pmid": "39022993", "trialId": null, "patentId": null, "grantId": "K08 CA255465; UL1 TR001863", "accessed": "2026-09-14", "notes": "Retrospective commercial-case series from multiple US institutions: 1,329 plasma samples from 298 clinically validated patients. Several authors were current or former Natera employees and held stock or options. The study tested prognosis, not whether ctDNA-guided treatment improves outcome." }, { "id": "boyd-2025-closed-loop-single-cell", "type": "preprint", "title": "Closing the loop: Teaching single-cell foundation models to learn from perturbations", "authors": [ "Boyd, Joshua R.", "et al." ], "publisher": "bioRxiv", "date": "2025-07-12", "url": "https://pubmed.ncbi.nlm.nih.gov/40672312/", "doi": "10.1101/2025.07.08.663754", "pmid": "40672312", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Primary preprint linking model predictions to CRISPR and drug experiments in blood and immune cells. Positive predictive value improved in a T-cell activation task, but no PDAC model or patient result was tested." }, { "id": "bretthauer-2022-nordicc-10y", "type": "paper", "title": "Effect of Colonoscopy Screening on Risks of Colorectal Cancer and Related Death", "authors": [ "Bretthauer, Michael", "Løberg, Magnus", "Wieszczy, Paulina", "et al." ], "publisher": "New England Journal of Medicine", "date": "2022-10-09", "url": "https://www.nejm.org/doi/10.1056/NEJMoa2208375", "doi": "10.1056/NEJMoa2208375", "pmid": "36214590", "trialId": "NCT00883792", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Multicountry population randomized trial. Of 28,220 people invited to one colonoscopy, 11,843, or 42.0%, underwent screening. At 10 years, invitation reduced colorectal-cancer incidence but did not show a significant colorectal-cancer mortality reduction. The per-protocol estimates are not protected by the original randomization." }, { "id": "bruix-2015-storm-sorafenib", "type": "paper", "title": "Adjuvant sorafenib for hepatocellular carcinoma after resection or ablation (STORM)", "authors": [ "Bruix, Jordi", "Takayama, Tadatoshi", "Mazzaferro, Vincenzo", "et al." ], "publisher": "The Lancet Oncology", "date": "2015-09-08", "url": "https://pubmed.ncbi.nlm.nih.gov/26361969/", "doi": "10.1016/S1470-2045(15)00198-9", "pmid": "26361969", "trialId": "NCT00692770", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Double-blind placebo-controlled phase 3 at 202 sites in 28 countries after complete radiological response to resection or ablation. Bayer funded the trial and employed several authors." }, { "id": "bruna-2025-diploma-long-term", "type": "paper", "title": "Minimally invasive vs open left pancreatectomy for resectable pancreatic cancer: long-term results of the randomized DIPLOMA trial", "authors": [ "Bruna, Claudia L", "van Hilst, Jony", "Korrel, Marieke", "et al." ], "publisher": "JAMA Surgery", "date": "2025-12-01", "url": "https://europepmc.org/article/MED/41060640", "doi": "10.1001/jamasurg.2025.4054", "pmid": "41060640", "trialId": "ISRCTN44897265", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Long-term DIPLOMA follow-up after median 38 months. Median overall survival was 32 versus 34 months, HR 1.02, and disease-free survival 21 versus 17 months, HR 0.96. Adjuvant therapy receipt was 70% versus 72% and median start 59 versus 56 days. No significant differences establish comparability within the trial precision; they do not prove identical effects or transfer to pancreaticoduodenectomy." }, { "id": "buscaill-2015-cyl02-gene-pdac", "type": "paper", "title": "First-in-man phase 1 clinical trial of gene therapy for advanced pancreatic cancer: safety, biodistribution, and preliminary clinical findings", "authors": [ "Buscaill and collaborators" ], "publisher": "Molecular Therapy", "date": "2015", "url": "https://pubmed.ncbi.nlm.nih.gov/25586689/", "doi": null, "pmid": "25586689", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "THERGAP phase 1 n=22; endoscopic-ultrasound delivery of nonviral CYL-02 plus gemcitabine; no objective responses." }, { "id": "cameron-2017-hera-trastuzumab", "type": "paper", "title": "11 years' follow-up of trastuzumab after adjuvant chemotherapy in HER2-positive early breast cancer: final HERA analysis", "authors": [ "Cameron, David", "Piccart-Gebhart, Martine J.", "Gelber, Richard D.", "et al." ], "publisher": "The Lancet", "date": "2017-03-25", "url": "https://pubmed.ncbi.nlm.nih.gov/28215665/", "doi": "10.1016/S0140-6736(16)32616-2", "pmid": "28215665", "trialId": "NCT00045032", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Primary 5,099-person randomized long-term report: one year of trastuzumab versus observation reduced DFS-event risk, HR 0.76, and death, HR 0.74; two years gave no added DFS benefit and cardiac toxicity increased with duration. Roche funded." }, { "id": "cancer-grand-challenges-2025-guidelines", "type": "funder-guidance", "title": "Cancer Grand Challenges 2025 expression-of-interest guidelines", "authors": [ "Cancer Grand Challenges" ], "publisher": "Cancer Grand Challenges", "date": "2025-03-05", "url": "https://www.cancergrandchallenges.org/sites/default/files/2025-03/Cancer%20Grand%20Challenges%20-%20Expression%20of%20Interest%20Guidelines%201.pdf", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Primary call guidance. Awards offered up to £20m in direct costs, with no more than 70% issued to institutions in a single country." }, { "id": "cancer-grand-challenges-2026-round", "type": "funder-guidance", "title": "Cancer Grand Challenges 2025–2026 application round", "authors": [ "Cancer Grand Challenges", "Cancer Research UK", "National Cancer Institute" ], "publisher": "Cancer Grand Challenges", "date": "2026-03-01", "url": "https://www.cancergrandchallenges.org/applications", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Primary current application page. The round is closed. It received 227 expressions of interest, shortlisted 12 teams and selected five. Teams had to address one funder-defined challenge, be international and place no more than 70% of activity and funding in one country." }, { "id": "caneque-2025-fentomycin-lysosomal-iron", "type": "paper", "title": "Activation of lysosomal iron triggers ferroptosis in cancer", "authors": [ "Caneque, Tatiana", "Baron, Leeroy", "Muller, Sebastian", "et al." ], "publisher": "Nature", "date": "2025-05-07", "url": "https://doi.org/10.1038/s41586-025-08974-4", "doi": "10.1038/s41586-025-08974-4", "pmid": "40335696", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Fentomycin-1 activates iron chemistry inside lysosomes and killed iron-rich CD44-high primary sarcoma and PDAC cells, including drug-tolerant states, in culture. Sublethal exposure selected a resistant state. In-vivo tumour reduction was shown in a breast-cancer lymph-node model, not PDAC. There is no human exposure, safety or pancreatic benefit result." }, { "id": "cao-2023-panda-ct-ai", "type": "paper", "title": "Large-scale pancreatic cancer detection via non-contrast CT and deep learning", "authors": [ "Cao, et al." ], "publisher": "Nature Medicine", "date": "2023-11-21", "url": "https://pubmed.ncbi.nlm.nih.gov/37985692/", "doi": "10.1038/s41591-023-02640-w", "pmid": "37985692", "trialId": "ChiCTR2200064645", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "PANDA was trained at one Chinese centre and externally evaluated across centres. The 20,530 consecutive-scan evaluation was retrospective and included clinical populations with much higher pancreatic-cancer prevalence than the general population. False-positive adjustment excluded some pancreatic or peripancreatic disease after multidisciplinary review." }, { "id": "capdevila-2025-seqtor-pnet", "type": "journal-article", "title": "Streptozotocin plus 5-fluorouracil followed by everolimus or the reverse sequence in patients with advanced pancreatic neuroendocrine tumors (SEQTOR-GETNE phase III study): a randomized clinical trial", "authors": [ "Capdevila, Jaume", "Tafuto, Salvatore", "Krogh, Morten", "et al." ], "publisher": "ESMO Open", "date": "2025-12", "url": "https://pubmed.ncbi.nlm.nih.gov/41337860/", "doi": "10.1016/j.esmoop.2025.105922", "pmid": "41337860", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "International open-label randomized crossover phase 3 in 141 advanced well-differentiated panNET patients. Slow accrual and longer survival led to changing the primary endpoint from 35-month PFS across two lines to first-line 12-month PFS. No significant PFS or OS sequence difference was shown; STZ/5-FU produced a higher first-line response rate." }, { "id": "caplin-2014-clarinet-lanreotide", "type": "paper", "title": "Lanreotide in metastatic enteropancreatic neuroendocrine tumors", "authors": [ "Caplin, Martyn E.", "Pavel, Marianne", "Cwikla, Jaroslaw B.", "et al." ], "publisher": "New England Journal of Medicine", "date": "2014-07-17", "url": "https://pubmed.ncbi.nlm.nih.gov/25014687/", "doi": "10.1056/NEJMoa1316158", "pmid": "25014687", "trialId": "NCT00353496", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Randomized double-blind placebo-controlled trial, n=204, in low-grade somatostatin-receptor-positive enteropancreatic NETs; primary sites were pooled. Lanreotide improved PFS, not overall survival or quality of life in the primary report. Ipsen funded." }, { "id": "carla-2024-ivospemin-retinal-atrophy", "type": "paper", "title": "Rapidly progressive central retinal pigmented epithelium atrophy after ivospemin treatment for pancreatic adenocarcinoma", "authors": [ "Carlà, Matteo Mario", "Mateo, Carlos" ], "publisher": "American Journal of Ophthalmology Case Reports", "date": "2024-10-18", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC11535749/", "doi": "10.1016/j.ajoc.2024.102200", "pmid": "39502458", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Independent case report. A 69-year-old woman had healthy retinas and 20/20 vision before ivospemin with chemotherapy. After the second monthly cycle, vision fell; at six months it was 20/5000 and 20/4000 with bilateral retinal pigment epithelium loss and no available treatment. A single case cannot establish frequency or isolate ivospemin from cancer and chemotherapy, but it shows possible severe irreversible harm despite no prior eye disease." }, { "id": "carsgen-2025-satricel-pdac-adjuvant", "type": "conference-poster", "title": "Adjuvant Therapy with Claudin18.2-specific CAR T Cells in High-Risk Pancreatic Cancer", "authors": [ "Yu, Xianjun", "Shi, Si", "Zhang, Jian", "et al." ], "publisher": "ESMO Congress / CARsgen Therapeutics", "date": "2025-10-19", "url": "https://www.carsgen.com/media/znwhl4oq/final-2025-esmo-ct041-poster.pdf", "doi": null, "pmid": null, "trialId": "NCT05911217", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Primary sponsor-hosted poster. Six selected patients were infused 6-10 months after surgery after abnormal CA19-9 but no visible recurrence; five were recurrence-free at median 6.05 months and one recurred. All six had cytokine-release syndrome after the first infusion; one had grade 3 syndrome after a second infusion, and five had severe low white-cell counts. With no control and very short follow-up, CA19-9 decline and five recurrence-free patients cannot establish adjuvant benefit." }, { "id": "carsgen-2026-interim-results", "type": "issuer-filing", "title": "Announcement of Interim Results for the Six Months Ended June 30, 2026", "authors": [ "CARsgen Therapeutics Holdings Limited" ], "publisher": "Hong Kong Exchanges and Clearing Limited", "date": "2026-08-18", "url": "https://www1.hkexnews.hk/listedco/listconews/sehk/2026/0818/2026081801887.pdf", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Primary issuer filing. CARsgen reported RMB1.400bn cash and cash equivalents at 30 June 2026, up from RMB1.123bn at year end after a placing; first-half revenue was about RMB62m. It expects at least RMB1.2bn cash at year end and says operations are funded into 2030. Finance strength does not establish pancreatic efficacy." }, { "id": "cdmrp-2026-pcarp-funding", "type": "government-data", "title": "Pancreatic Cancer Research Program: congressional appropriations and funding summary", "authors": [ "Congressionally Directed Medical Research Programs" ], "publisher": "United States Department of Defense", "date": "2026-07-16", "url": "https://cdmrp.health.mil/pcarp/default", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Primary government programme page. It reports $66m of congressional appropriations across FY2020–FY2024 and 102 awards, plus a separate $20m FY2026 appropriation. The programme began with $6m in FY2020; the FY2024 announcement states FY2020–FY2023 totalled $51m and FY2024 added $15m. FY2025 is not included in the displayed cumulative value and is not inferred. Appropriations are not identical to research expenditure." }, { "id": "cebron-2023-panect-preliminary", "type": "paper", "title": "Intraoperative electrochemotherapy of the posterior resection surface after pancreaticoduodenectomy: Preliminary results of a hybrid approach treatment of pancreatic cancer", "authors": [ "Čebron, Žan", "Djokić, Mihajlo", "Petrič, Miha", "et al." ], "publisher": "Bioelectrochemistry", "date": "2023-09-22", "url": "https://pubmed.ncbi.nlm.nih.gov/37748261/", "doi": "10.1016/j.bioelechem.2023.108576", "pmid": "37748261", "trialId": "NCT04281290", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Preliminary single-arm PanECT report. Electrochemotherapy after pancreatic-head resection was delivered in seven patients, with one grade-B pancreatic fistula and no hospital death. It cannot estimate recurrence or survival benefit." }, { "id": "chalabi-hajkarim-2025-organotropism", "type": "paper", "title": "Cellular states associated with metastatic organotropism and survival in patients with pancreatic ductal adenocarcinoma", "authors": [ "Chalabi Hajkarim, Morteza", "May, Michael", "Amin, Amit Dipak", "et al." ], "publisher": "Nature Genetics", "date": "2025-10-27", "url": "https://pubmed.ncbi.nlm.nih.gov/41145870/", "doi": "10.1038/s41588-025-02345-5", "pmid": "41145870", "trialId": null, "patentId": null, "grantId": "R38 CA231577", "accessed": "2026-09-14", "notes": "Retrospective analysis of 744 resected PDAC patients found 333 documented recurrences, including 100 initial liver-only and 31 lung-only. Archival primary-tumour single-nucleus sequencing succeeded in 21 selected cases, 8 liver and 13 lung recurrence, and found organ-linked expression states without a large genomic predictor. The result supports pre-existing route differences but cannot show that a cell was dormant, that the signature causes recurrence or that it selects treatment." }, { "id": "chan-2024-dpyd-non-european-review", "type": "paper", "title": "DPYD genetic polymorphisms in non-European patients with severe fluoropyrimidine-related toxicity", "authors": [ "Chan, T H", "et al." ], "publisher": "British Journal of Cancer", "date": "2024-08-01", "url": "https://doi.org/10.1038/s41416-024-02754-z", "doi": "10.1038/s41416-024-02754-z", "pmid": "38886557", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Review of 32 studies found 53 DPYD variants across five non-European ancestry groups. It identifies c.557A>G as a strong African-ancestry toxicity candidate absent from the four-variant UK panel. This is an equity and residual-risk warning, not prospective proof that a wider panel improves outcomes." }, { "id": "chan-2025-cabinet", "type": "paper", "title": "Phase 3 Trial of Cabozantinib to Treat Advanced Neuroendocrine Tumors", "authors": [ "Chan, Jennifer A.", "Geyer, Susan", "Zemla, Taylor", "et al." ], "publisher": "New England Journal of Medicine", "date": "2024-09-16", "url": "https://pubmed.ncbi.nlm.nih.gov/39282913/", "doi": "10.1056/NEJMoa2403991", "pmid": "39282913", "trialId": "NCT03375320", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "NCI-led double-blind randomised phase 3 with separate pancreatic and extrapancreatic cohorts. The pancreatic cohort included 95 previously treated progressive patients; overall survival was a secondary endpoint and not mature in the primary report." }, { "id": "chan-2026-control-nets", "type": "paper", "title": "Lutetium-177 DOTATATE PRRT and CAPTEM chemotherapy for pancreas and small bowel neuroendocrine tumours: the AGITG CONTROL NETS multicentre randomized trial", "authors": [ "Chan, David L.", "Sjoquist, Katrin Marie", "Ransom, David Turner", "et al." ], "publisher": "European Journal of Cancer", "date": "2026-06-18", "url": "https://pubmed.ncbi.nlm.nih.gov/42142431/", "doi": "10.1016/j.ejca.2026.116781", "pmid": "42142431", "trialId": "NCT02358356", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Open-label non-comparative randomized phase 2 trial of lutetium-177 DOTATATE plus capecitabine-temozolomide. Only 27 of 72 participants had pNET. Twelve-month PFS was 83.3% with the combination and 88.9% with chemotherapy; 27-month PFS was 61.1% and 33.3%, HR 0.41, 95% CI 0.15-1.12, p=0.08. Response was 72.2% and 33.3%. Two of 63 people exposed to radioligand therapy developed treatment-related myeloid neoplasms. The small subgroup, non-comparative phase 2 design and confidence interval crossing 1 make this a phase 3 signal, not proof of superiority." }, { "id": "chan-seng-yue-2020-evolution-subtypes", "type": "paper", "title": "Transcription phenotypes of pancreatic cancer are driven by genomic events during tumor evolution", "authors": [ "Chan-Seng-Yue, Michelle", "Kim, Julie C.", "Wilson, Gavin W.", "et al." ], "publisher": "Nature Genetics", "date": "2020-01-13", "url": "https://pubmed.ncbi.nlm.nih.gov/31932696/", "doi": "10.1038/s41588-019-0566-9", "pmid": "31932696", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Whole-genome and transcriptome analysis of purified primary and metastatic human tumour epithelium, with single-cell confirmation of mixed tumour-cell states." }, { "id": "chandrasekhara-2021-eus-verteporfin-pdt", "type": "paper", "title": "EUS-guided verteporfin photodynamic therapy for pancreatic cancer", "authors": [ "Chandrasekhara, Vinay", "Mackenzie, Trisha A.", "Carr-Locke, David L.", "et al." ], "publisher": "Gastrointestinal Endoscopy", "date": "2021-07-01", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC10434704/", "doi": "10.1016/j.gie.2021.02.027", "pmid": "33647286", "trialId": "NCT03033225", "patentId": null, "grantId": "P01CA084203", "accessed": "2026-09-15", "notes": "Single-centre feasibility report. Of 623 people screened, 54 were eligible and approached and eight were treated. Five of eight had a new dead zone on two-day CT, averaging 15.7 mm across inside tumours averaging 33.3 mm. Three had no dead zone. Seven died, with median 209 days from PDT; one was alive at 407 days. No control existed, four had metastatic disease and prior treatment varied." }, { "id": "chang-1997-hbv-vaccine-hcc", "type": "paper", "title": "Universal hepatitis B vaccination in Taiwan and the incidence of hepatocellular carcinoma in children", "authors": [ "Chang, Mei-Hwei", "Chen, Chien-Jen", "Lai, Mei-Sang", "et al." ], "publisher": "New England Journal of Medicine", "date": "1997-06-26", "url": "https://pubmed.ncbi.nlm.nih.gov/9197213/", "doi": "10.1056/NEJM199706263362602", "pmid": "9197213", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Nationwide natural experiment using Taiwan cancer-registry and medical-centre data after universal infant hepatitis B vaccination began in 1984. Childhood HCC is rare and the comparison was not randomised." }, { "id": "chari-2026-heuriskance", "type": "paper", "title": "Heuriskance: a novel paradigm for systematic earlier detection of sporadic pancreatic cancer", "authors": [ "Chari, Suresh T.", "Feng, Ziding", "Wu, Buqun", "et al." ], "publisher": "Journal of the National Cancer Institute", "date": "2026-08-01", "url": "https://pubmed.ncbi.nlm.nih.gov/41063387/", "doi": "10.1093/jnci/djaf291", "pmid": "41063387", "trialId": "NCT04662879", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Recent programme framework, not a trial outcome. It proposes one-time diagnostic work-up after a scalable early-warning sign rather than repeated general-population screening." }, { "id": "chari-2026-prospective-new-onset-diabetes", "type": "paper", "title": "Risk of Pancreatic Cancer in Glycemically Defined New-Onset Diabetes: A Prospective Cohort Study", "authors": [ "Chari, Suresh T.", "Wu, Bechien", "Lopez, Camden", "et al." ], "publisher": "Gastroenterology", "date": "2025-07-08", "url": "https://pubmed.ncbi.nlm.nih.gov/40633624/", "doi": "10.1053/j.gastro.2025.06.025", "pmid": "40633624", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Prospective active electronic-health-record surveillance of 18,838 adults aged at least 50 with glycaemically defined new-onset diabetes. Interim analysis; histology is reported as pancreatic cancer rather than PDAC only." }, { "id": "charoensuk-2024-one-health-opisthorchis", "type": "paper", "title": "Integrated One-Health approach for prevention and control of Opisthorchis viverrini infection in rural Thailand: a 3-year study", "authors": [ "Charoensuk, Lakkhana", "Chedtabud, Kittisak", "Chaipibool, Surapong", "et al." ], "publisher": "Parasitology Research", "date": "2024-06-28", "url": "https://pubmed.ncbi.nlm.nih.gov/38940830/", "doi": "10.1007/s00436-024-08275-1", "pmid": "38940830", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Three-year rural Thai One Health programme reporting data from 3,600 participants. Baseline human infection was 218, or 6.0%. Bundled interviews, education, anthelminthic treatment and environmental and animal measures were followed by no reported human reinfection in year two and 0.3% prevalence in year three. The study was not a randomized comparison and did not measure cholangiocarcinoma incidence or mortality." }, { "id": "chen-2023-element-biliary-rct", "type": "paper", "title": "ELEMENT trial: first-intent EUS-guided versus ERCP-guided metal-stent drainage in malignant distal obstruction", "authors": [ "Chen, Yen-I", "Sahai, Anand", "Donatelli, Gianfranco", "et al." ], "publisher": "Gastroenterology", "date": "2023-11-01", "url": "https://pubmed.ncbi.nlm.nih.gov/37549753/", "doi": "10.1053/j.gastro.2023.07.024", "pmid": "37549753", "trialId": "NCT03870386", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Multicentre randomized trial of 144 people with malignant distal obstruction from borderline-resectable, locally advanced or unresectable periampullary cancers. EUS-guided choledochoduodenostomy and ERCP metal stenting had technical success 90.4% versus 83.1% and one-year stent dysfunction 9.6% versus 9.9%. EUS was faster; adverse events, surgery, cancer outcomes and quality of life did not differ. The mixed cancer and treatment states and expert-centre setting limit PDAC-specific route conclusions." }, { "id": "chen-2023-fdr-cancer-risk", "type": "paper", "title": "Risk of Syndrome-Associated Cancers Among First-Degree Relatives of Patients With Pancreatic Ductal Adenocarcinoma With Pathogenic or Likely Pathogenic Germline Variants", "authors": [ "Chen, Xin", "Meyer, Mark A.", "Kemppainen, Jill L.", "et al." ], "publisher": "JAMA Oncology", "date": "2023-07-01", "url": "https://pubmed.ncbi.nlm.nih.gov/37200008/", "doi": "10.1001/jamaoncol.2023.0806", "pmid": "37200008", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Case series of 1,670 first-degree relatives reported by 234 Mayo PDAC probands with pathogenic variants. Gene-specific excesses covered pancreatic, breast, ovarian, colorectal, endometrial and melanoma risk. The probands were 94.4% White, family cancers were reported rather than systematically verified, and association does not prove that cascade testing prevents death." }, { "id": "chen-2025-tocilizumab-cachexia", "type": "paper", "title": "Gemcitabine and nab-paclitaxel with or without tocilizumab in advanced pancreatic cancer: survival and cachexia", "authors": [ "Chen, Inna M.", "Johansen, Julia S.", "Theile, Stephanie", "et al." ], "publisher": "Journal of Clinical Oncology", "date": "2025-05-12", "url": "https://pubmed.ncbi.nlm.nih.gov/40354592/", "doi": "10.1200/JCO.23.01965", "pmid": "40354592", "trialId": "NCT02767557", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Primary randomized phase 2: 147 treated including safety run-in. Six-month OS endpoint missed, median OS/PFS/ORR unchanged; muscle loss improved, while grade 3+ treatment-related events rose from 63.4% to 88.1%." }, { "id": "chen-2026-personalized-ctdna", "type": "paper", "title": "Personalized tumor-informed circulating tumor DNA monitoring for early detection of recurrence in postoperative pancreatic cancer", "authors": [ "Chen, Jingjing", "Zou, Lu", "Bai, Xinyuan", "et al." ], "publisher": "Frontiers in Oncology", "date": "2026-01-22", "url": "https://pubmed.ncbi.nlm.nih.gov/41659718/", "doi": "10.3389/fonc.2026.1745466", "pmid": "41659718", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Chinese postoperative cohort: 43 recruited patients, personalized panels successfully built for 35 and 139 plasma samples analysed. Eleven patients became ctDNA-positive before radiological relapse with median lead time 4.59 months. After adjuvant chemotherapy, positive predictive value was 75.0% and negative predictive value 82.4%. The study was small, non-randomized and two authors were employed by Geneseeq; it establishes neither a treatment rule nor survival benefit from earlier action." }, { "id": "chhoda-2026-in-cyst", "type": "paper", "title": "Clinical Outcomes in Patients With Worrisome and High-Risk Pancreatic Cysts: A Prospective Cohort Study", "authors": [ "Chhoda, Ankit", "Carlson, Erin E.", "Bamlet, William R.", "et al." ], "publisher": "Gastroenterology", "date": "2026-08-20", "url": "https://pubmed.ncbi.nlm.nih.gov/42624412/", "doi": "10.1053/j.gastro.2026.07.038", "pmid": "42624412", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Prospective IN-CYST cohort, n=230. Mayo's accompanying report gives three-year HGD or cancer estimates of about 2%, 4%, and 18% for one, two, and at least three worrisome features. Thirty-two underwent surgery within three months, leaving 198 in surveillance; the report is too small to establish mortality benefit." }, { "id": "chiorean-2023-abemaciclib-pdac", "type": "paper", "title": "Efficacy and safety of abemaciclib alone and with PI3K/mTOR inhibitor LY3023414 or galunisertib versus chemotherapy in previously treated metastatic pancreatic adenocarcinoma: a randomized controlled trial", "authors": [ "Chiorean, E. Gabriela", "Picozzi, Vincent", "Li, Chung-Pin", "et al." ], "publisher": "Cancer Medicine", "date": "2023-10-16", "url": "https://pubmed.ncbi.nlm.nih.gov/37840530/", "doi": "10.1002/cam4.6621", "pmid": "37840530", "trialId": "NCT02981342", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Primary randomized phase 2 report. Ninety-nine people were randomized equally to abemaciclib, abemaciclib plus LY3023414, or physician-choice gemcitabine/capecitabine. Disease control was 15.2%, 12.1% and 36.4%; PFS was 1.7, 1.8 and 3.3 months. OS HRs versus chemotherapy were 1.60 and 1.53. Neither experimental arm advanced. The population was not selected by a cell-cycle alteration." }, { "id": "chittenden-2021-systematic-germline", "type": "paper", "title": "Implementing Systematic Genetic Counseling and Multigene Germline Testing for Individuals With Pancreatic Cancer", "authors": [ "Chittenden, Anu", "Haraldsdottir, Sigurdis", "Ukaegbu, Chinedu", "et al." ], "publisher": "JCO Oncology Practice", "date": "2021-01-13", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC8257923/", "doi": "10.1200/OP.20.00678", "pmid": "33439686", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Single-centre implementation study. Among 1,214 new PDAC oncology consultations, 266 completed counselling and testing. Automatic offer raised testing from 16.5% to 38.0% and testing within seven days from 14.7% to 60.3%. Sixteen of 28 variant carriers had at least one relative tested; 13 of 26 carriers with incurable disease received matched treatment. Research funds helped pay testing, and 15 of 266 died within 30 days of results." }, { "id": "choi-2024-microfluidic-pdac-organoids", "type": "paper", "title": "Microfluidic Organoid Cultures Derived from Pancreatic Cancer Biopsies for Personalized Testing of Chemotherapy and Immunotherapy", "authors": [ "Choi, Daheui", "et al." ], "publisher": "Advanced Science", "date": "2024-02-01", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC10837378/", "doi": "10.1002/advs.202303088", "pmid": "38018486", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Microfluidic cultures reduced required biopsy material, standardized spheroid size and permitted chemotherapy plus NK-cell testing. The work establishes an experimental platform, not timely all-patient assay completion or improved treatment outcome." }, { "id": "chuong-2024-smart-phase2-survival", "type": "paper", "title": "Stereotactic MR-guided on-table adaptive radiation therapy for borderline resectable and locally advanced pancreatic cancer: a multi-center, open-label phase 2 study", "authors": [ "Chuong, Michael D.", "Lee, Percy", "Low, Daniel A.", "et al." ], "publisher": "Radiotherapy and Oncology", "date": "2024-02-01", "url": "https://pubmed.ncbi.nlm.nih.gov/38135187/", "doi": "10.1016/j.radonc.2023.110064", "pmid": "38135187", "trialId": "NCT03621644", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Mature report from the industry-sponsored, uncontrolled phase 2 study. It enrolled 136 people after at least three months of chemotherapy without distant progression and with CA19-9 at most 500 U/mL. Two-year survival was 53.6% from diagnosis and 40.5% from SMART; 34.6% later underwent surgery." }, { "id": "clinicaltrials-nct01077427-heat", "type": "trial-registry", "title": "HEAT: Hyperthermia European Adjuvant Trial", "authors": [ "Ludwig-Maximilians University of Munich" ], "publisher": "ClinicalTrials.gov", "date": "2017-11-17", "url": "https://clinicaltrials.gov/study/NCT01077427", "doi": null, "pmid": null, "trialId": "NCT01077427", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Primary registry checked through the ClinicalTrials.gov API. The phase 3 record remains status unknown, with 336 estimated participants, no posted results and its last update on 17 November 2017. It lists gemcitabine plus capecitabine as the comparator and gemcitabine plus cisplatin plus regional hyperthermia as the experimental arm." }, { "id": "clinicaltrials-nct01746979-maestro", "type": "trial-registry", "title": "MAESTRO: evofosfamide plus gemcitabine versus placebo plus gemcitabine in advanced pancreatic cancer", "authors": [ "Threshold Pharmaceuticals", "Merck KGaA" ], "publisher": "ClinicalTrials.gov", "date": "2025-10-30", "url": "https://clinicaltrials.gov/study/NCT01746979", "doi": null, "pmid": null, "trialId": "NCT01746979", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Completed randomized double-blind phase 3, actual enrollment 693. Evofosfamide is a prodrug intended to release its cell-killing payload under low oxygen. Registry results report median OS 8.9 versus 7.6 months and PFS 5.5 versus 3.7 months. The published main test narrowly missed its planned OS threshold, so improved control did not become proven survival benefit." }, { "id": "clinicaltrials-nct01770132-porfimer-pdt", "type": "trial-registry", "title": "Endoscopic-ultrasound-guided Photofrin photodynamic therapy with gemcitabine for locally advanced pancreatic cancer", "authors": [ "DeWitt, John M.", "Indiana University", "Pinnacle Biologics" ], "publisher": "ClinicalTrials.gov", "date": "2020-04-07", "url": "https://clinicaltrials.gov/study/NCT01770132", "doi": null, "pmid": "30222972", "trialId": "NCT01770132", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Completed single-group phase 1 record with estimated enrollment 12 and no posted results. It combines porfimer, endoscopic-ultrasound-guided light delivery and gemcitabine. Safety is primary; necrosis, response, downstaging, CA19-9, progression and survival are secondary. Pinnacle Biologics is a collaborator." }, { "id": "clinicaltrials-nct02143219-pamela70", "type": "trial-registry", "title": "NCT02143219 — PAMELA-70 posted results", "authors": [ "Institut de Cancérologie de l'Ouest" ], "publisher": "ClinicalTrials.gov", "date": "2026-04-15", "url": "https://clinicaltrials.gov/study/NCT02143219", "doi": null, "pmid": null, "trialId": "NCT02143219", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Completed single-group phase 2 of pharmacogenetically adjusted FOLFIRINOX in people aged at least 70 with metastatic PDAC. Seventy-two enrolled and 69 started treatment. Results were first posted in April 2026, more than five years after actual completion. Only one composite safety and early-efficacy outcome is posted, 65.8%, 95% CI 53.2-75.8, without separate response, autonomy or survival outcomes. Serious adverse events affected 48/69, 69.6%, and 63/69 died during the two-year adverse-event frame. The record cannot show comparative benefit or give a complete decision rule." }, { "id": "clinicaltrials-nct02170090-acticca1", "type": "trial-registry", "title": "Adjuvant Chemotherapy With Gemcitabine and Cisplatin Compared to Standard of Care After Curative Intent Resection of Biliary Tract Cancer", "authors": [ "University Medical Center Hamburg-Eppendorf" ], "publisher": "ClinicalTrials.gov", "date": "2025-03-30", "url": "https://clinicaltrials.gov/study/NCT02170090", "doi": null, "pmid": null, "trialId": "NCT02170090", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "ACTICCA-1 registry checked through the ClinicalTrials.gov API. Open-label randomized phase 3 comparison of adjuvant gemcitabine-cisplatin with capecitabine after resection, actual enrolment 789. Listed active but not recruiting, with estimated primary and study completion in December 2025, last update March 2025 and no posted results." }, { "id": "clinicaltrials-nct02336672-hybridtherm", "type": "trial-registry", "title": "EUS-Guided Cryothermal Ablation in Patients With Stage III Pancreatic Adenocarcinoma", "authors": [ "IRCCS San Raffaele" ], "publisher": "ClinicalTrials.gov", "date": "2023-09-28", "url": "https://clinicaltrials.gov/study/NCT02336672", "doi": null, "pmid": null, "trialId": "NCT02336672", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Primary registry checked through the API. It labels the completed randomized study actual enrollment 66 and posts no results, while the linked paper reports 40 randomized and 37 retained in its all-assigned set. The intervention called cryothermal ablation was an internally cooled radiofrequency probe, not tissue freezing." }, { "id": "clinicaltrials-nct02439593-heatpac", "type": "trial-registry", "title": "HEATPAC: concurrent hyperthermia and chemoradiotherapy in locally advanced pancreatic cancer", "authors": [ "Kantonsspital Aarau" ], "publisher": "ClinicalTrials.gov", "date": "2017-08-01", "url": "https://clinicaltrials.gov/study/NCT02439593", "doi": null, "pmid": null, "trialId": "NCT02439593", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Primary registry checked through the ClinicalTrials.gov API. The randomized phase 2 record remains status unknown, has 78 estimated participants, no posted results, an estimated completion date of 31 December 2021 and its last update on 1 August 2017. Its two arms use the same chemoradiotherapy and differ by weekly 41-43 degrees C regional hyperthermia for 60 minutes." }, { "id": "clinicaltrials-nct02598349-proton", "type": "trial-registry", "title": "PAN009-18: escalated-dose proton radiotherapy with capecitabine for localized pancreatic cancer", "authors": [ "Proton Collaborative Group", "University of Florida Health" ], "publisher": "ClinicalTrials.gov", "date": "2025-09-09", "url": "https://clinicaltrials.gov/study/NCT02598349", "doi": null, "pmid": null, "trialId": "NCT02598349", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Current primary registry checked through the ClinicalTrials.gov API. The non-randomized phase 2 study remains recruiting with 60 estimated participants, no posted results and estimated primary completion in October 2039. Its primary test is one-year survival of 75% versus a 50% historical rate." }, { "id": "clinicaltrials-nct02702986-imilt-verona", "type": "trial-registry", "title": "Safety and feasibility of imILT in locally advanced pancreatic cancer at Verona", "authors": [ "University of Verona", "Clinical Laserthermia Systems AB" ], "publisher": "ClinicalTrials.gov", "date": "2022-04-25", "url": "https://clinicaltrials.gov/study/NCT02702986", "doi": null, "pmid": "33278792", "trialId": "NCT02702986", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Completed single-group study planning ten participants and open surgical placement of the imILT device. Safety was primary; immune measurements, survival and progression were secondary. Clinical Laserthermia Systems was the industry collaborator. Enrollment remains estimated and no results are posted." }, { "id": "clinicaltrials-nct02912949-enrgy", "type": "trial-registry", "title": "A study of zenocutuzumab in patients with solid tumors harboring an NRG1 fusion", "authors": [ "Partner Therapeutics, Inc." ], "publisher": "ClinicalTrials.gov", "date": "2025-04-29", "url": "https://clinicaltrials.gov/study/NCT02912949", "doi": null, "pmid": null, "trialId": "NCT02912949", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Active, not recruiting phase I/II basket trial, target enrollment 250, last updated 29 April 2025. It includes a pancreatic NRG1-fusion cohort receiving zenocutuzumab 750 mg every two weeks. The registry estimates completion on 31 December 2026 and posts no tabular results, while FDA required the 50-evaluable-pancreatic report by August 2026." }, { "id": "clinicaltrials-nct02973217-imilt-marseille", "type": "trial-registry", "title": "Immunostimulating interstitial laser thermotherapy in pancreatic cancer at Marseille", "authors": [ "Clinical Laserthermia Systems AB" ], "publisher": "ClinicalTrials.gov", "date": "2020-12-10", "url": "https://clinicaltrials.gov/study/NCT02973217", "doi": null, "pmid": "33278792", "trialId": "NCT02973217", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Completed industry-sponsored single-group study with five actual participants. Safety was primary; immune response, progression, quality of life and device usability were secondary. No results are posted." }, { "id": "clinicaltrials-nct03023722-anetumab", "type": "trial-registry-results", "title": "NCT03023722 results: anetumab ravtansine in mesothelin-expressing pancreatic cancer", "authors": [ "Yale University", "National Cancer Institute" ], "publisher": "ClinicalTrials.gov", "date": "2021-10-28", "url": "https://clinicaltrials.gov/study/NCT03023722?tab=results", "doi": null, "pmid": null, "trialId": "NCT03023722", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Primary posted results from a completed single-arm phase 2. Eighteen people began treatment, 14 were response evaluable, 12 progressed and two had stable disease: zero objective responses. Median time to progression was 63.5 days, 11 of 18 had serious adverse events, and all 18 died before three years." }, { "id": "clinicaltrials-nct03033225-vertpac02", "type": "trial-registry", "title": "VERTPAC-02: ultrasound-guided verteporfin photodynamic therapy for unresectable pancreatic tumours", "authors": [ "Mayo Clinic", "National Cancer Institute" ], "publisher": "ClinicalTrials.gov", "date": "2025-06-03", "url": "https://clinicaltrials.gov/study/NCT03033225", "doi": null, "pmid": "33647286", "trialId": "NCT03033225", "patentId": null, "grantId": "P01CA084203", "accessed": "2026-09-15", "notes": "Completed single-group registry with 13 starters, eight completers and five withdrawals. The posted primary result is a two-day imaging response: five of eight had a dead zone and three did not. Serious and other adverse-event tables report zero among eight through day 14. The principal investigator is also the sponsor's employee." }, { "id": "clinicaltrials-nct03084770-aspen", "type": "trial-registry", "title": "A Prospective Evaluation of the Management of Sporadic Asymptomatic Nonfunctioning Pancreatic Neuroendocrine Neoplasms no Larger Than 2 cm", "authors": [ "IRCCS San Raffaele" ], "publisher": "ClinicalTrials.gov", "date": "2024-01-01", "url": "https://clinicaltrials.gov/study/NCT03084770", "doi": null, "pmid": null, "trialId": "NCT03084770", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Live registry API checked. The prospective observational ASPEN cohort has an estimated enrolment of 1,000, an estimated primary completion of August 2024 and study completion of December 2024. ClinicalTrials.gov now reports the status as unknown because the last verified month was January 2024. No registry results are posted. The published interim analysis contains 500 participants and 25 months median follow-up." }, { "id": "clinicaltrials-nct03187587-imilt-randomized", "type": "trial-registry", "title": "Percutaneous immunostimulating interstitial laser thermotherapy versus chemotherapy in pancreatic cancer", "authors": [ "Clinical Laserthermia Systems AB" ], "publisher": "ClinicalTrials.gov", "date": "2021-07-13", "url": "https://clinicaltrials.gov/study/NCT03187587", "doi": null, "pmid": null, "trialId": "NCT03187587", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Industry-sponsored open-label randomized study comparing imILT without concurrent chemotherapy against standard chemotherapy. It terminated with eight actual participants because recruitment was slow; no results are posted. The primary endpoint was radiographic response at 12 months, with safety and device usability secondary." }, { "id": "clinicaltrials-nct03214250-prince", "type": "trial-registry", "title": "PRINCE: sotigalimab and/or nivolumab with gemcitabine/nab-paclitaxel in untreated metastatic PDAC", "authors": [ "Parker Institute for Cancer Immunotherapy", "Apexigen" ], "publisher": "ClinicalTrials.gov", "date": "2026-05-20", "url": "https://clinicaltrials.gov/study/NCT03214250", "doi": null, "pmid": null, "trialId": "NCT03214250", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Completed phase 1b/2, actual enrollment 129, with results posted. The phase 2 arms started 37 nivolumab plus chemotherapy, 31 sotigalimab plus chemotherapy and 31 triplet participants. There was no chemotherapy-only arm, so random allocation cannot isolate the added contribution of either immune drug." }, { "id": "clinicaltrials-nct03250273-entinostat-nivolumab", "type": "trial-registry", "title": "Entinostat with nivolumab for previously treated cholangiocarcinoma and pancreatic adenocarcinoma", "authors": [ "Sidney Kimmel Comprehensive Cancer Center at Johns Hopkins" ], "publisher": "ClinicalTrials.gov", "date": "2025-05-28", "url": "https://clinicaltrials.gov/study/NCT03250273", "doi": null, "pmid": null, "trialId": "NCT03250273", "patentId": null, "grantId": "5P01CA247886", "accessed": "2026-09-15", "notes": "Completed open-label nonrandomized phase 2 registry, actual enrollment 44 across parallel pancreatic and cholangiocarcinoma cohorts, with posted summary results, protocol and analysis plan." }, { "id": "clinicaltrials-nct03257033-tigerpac", "type": "trial-registry", "title": "TIGeR-PaC: intra-arterial gemcitabine versus intravenous gemcitabine and nab-paclitaxel after induction treatment", "authors": [ "RenovoRx" ], "publisher": "ClinicalTrials.gov", "date": "2026-08-24", "url": "https://clinicaltrials.gov/study/NCT03257033", "doi": null, "pmid": null, "trialId": "NCT03257033", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Current primary registry. This open-label randomized phase 3 estimates 190 enrolled people including induction. Only people with stable or responding locally advanced disease after about four months of chemotherapy and radiotherapy, who remain non-surgical candidates, enter randomization. The experimental package gives intra-arterial gemcitabine every two weeks for eight treatments; control continues intravenous gemcitabine plus nab-paclitaxel on days 1, 8 and 15 of four 28-day cycles. Overall survival from randomization is primary. The August 2026 record still says recruiting, gives June 2027 primary completion, posts no results and contains no participant-level sharing statement." }, { "id": "clinicaltrials-nct03264404-azacitidine-pembrolizumab", "type": "trial-registry", "title": "Azacitidine and pembrolizumab in pancreatic cancer", "authors": [ "Columbia University", "Bates, Susan E." ], "publisher": "ClinicalTrials.gov", "date": "2026-01-27", "url": "https://clinicaltrials.gov/study/NCT03264404", "doi": null, "pmid": null, "trialId": "NCT03264404", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Completed open-label single-group phase 2, actual enrollment 36, with posted results. Two completed treatment, 28 stopped for progression, two died and two came off before treatment; summary results report serious adverse events in 9 of 36." }, { "id": "clinicaltrials-nct03323944-meso-cart", "type": "trial-registry", "title": "CAR T Cell Immunotherapy for Pancreatic Cancer", "authors": [ "University of Pennsylvania" ], "publisher": "ClinicalTrials.gov", "date": "2026-01-30", "url": "https://clinicaltrials.gov/study/NCT03323944", "doi": null, "pmid": null, "trialId": "NCT03323944", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Terminated for administrative reasons after actual enrollment 54 and actual completion in February 2025. The registry posts no results and gives no data-sharing statement. The 2025 main paper reports nine infused patients, leaving the status and outcomes of 45 registry participants unreconciled." }, { "id": "clinicaltrials-nct03336216-cabiralizumab", "type": "trial-registry", "title": "Cabiralizumab plus nivolumab with and without chemotherapy in advanced pancreatic cancer", "authors": [ "Bristol-Myers Squibb", "Five Prime Therapeutics" ], "publisher": "ClinicalTrials.gov", "date": "2026-05-20", "url": "https://clinicaltrials.gov/study/NCT03336216", "doi": null, "pmid": null, "trialId": "NCT03336216", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Completed randomized phase 2, actual enrollment 205, after prior chemotherapy. Investigator-choice chemotherapy n=54 was compared with cabiralizumab plus nivolumab n=54, that pair plus gemcitabine-based chemotherapy n=54, or plus 5-FU-based chemotherapy n=43. Central-review median PFS was 3.52, 1.92, 3.68 and 3.22 months; HR versus control 1.47, 1.00 and 0.78. Median OS was 6.28, 4.44, 6.72 and 5.68 months; HR 1.22, 1.04 and 0.81, all confidence intervals including no effect. Abnormal liver tests were much more common in cabiralizumab arms." }, { "id": "clinicaltrials-nct03377491-panova3", "type": "trial-registry", "title": "PANOVA-3: tumour-treating fields with gemcitabine and nab-paclitaxel in locally advanced pancreatic cancer", "authors": [ "NovoCure Ltd." ], "publisher": "ClinicalTrials.gov", "date": "2025-09-17", "url": "https://clinicaltrials.gov/study/NCT03377491", "doi": null, "pmid": null, "trialId": "NCT03377491", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Primary trial registry with posted results. The industry-sponsored open-label phase 3 trial completed with 571 actual participants randomized to gemcitabine and nab-paclitaxel with or without the NovoTTF-200T device. Overall survival was primary; progression-free survival was the key secondary endpoint. The current record was checked through the ClinicalTrials.gov API." }, { "id": "clinicaltrials-nct03412799-sbp101", "type": "trial-registry", "title": "Phase 1 SBP-101 with gemcitabine and nab-paclitaxel", "authors": [ "Panbela Therapeutics, Inc." ], "publisher": "ClinicalTrials.gov", "date": "2022-05-25", "url": "https://clinicaltrials.gov/study/NCT03412799", "doi": null, "pmid": null, "trialId": "NCT03412799", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Completed, open-label, single-group phase 1 with actual n=50. The main endpoint was recommended dose; tumour response and adverse events were secondary. No registry results are posted." }, { "id": "clinicaltrials-nct03536182-cipher", "type": "trial-registry", "title": "CIPHER: randomized carbon-ion versus photon radiotherapy for locally advanced pancreatic cancer", "authors": [ "University of Texas Southwestern Medical Center" ], "publisher": "ClinicalTrials.gov", "date": "2021-05-26", "url": "https://clinicaltrials.gov/study/NCT03536182", "doi": null, "pmid": null, "trialId": "NCT03536182", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Primary registry for the planned randomized phase 3 comparison. It was withdrawn with zero actual enrollment because of COVID and other recruitment problems. The arms also differed in fraction count and allowed concurrent chemotherapy, limiting a pure particle-type comparison even if enrollment had occurred." }, { "id": "clinicaltrials-nct03563248-losartan-localized", "type": "trial-registry", "title": "Losartan and nivolumab with FOLFIRINOX and SBRT in localized pancreatic cancer", "authors": [ "Massachusetts General Hospital" ], "publisher": "ClinicalTrials.gov", "date": "2026-04-30", "url": "https://clinicaltrials.gov/study/NCT03563248", "doi": null, "pmid": null, "trialId": "NCT03563248", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Active, not recruiting randomized phase 2, actual enrollment 168. Four groups separate FOLFIRINOX with or without losartan and SBRT with or without nivolumab before attempted surgery. The primary endpoint is R0 resection; PFS, OS, complete response and serious harm are secondary. Primary completion was February 2022, final completion is estimated April 2027, no results are posted, and individual data sharing says no." }, { "id": "clinicaltrials-nct03621644-smart", "type": "trial-registry", "title": "Phase 2 SMART for borderline or inoperable locally advanced pancreatic cancer", "authors": [ "ViewRay, Inc." ], "publisher": "ClinicalTrials.gov", "date": "2022-10-14", "url": "https://clinicaltrials.gov/study/NCT03621644", "doi": null, "pmid": null, "trialId": "NCT03621644", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Primary registry checked through the ClinicalTrials.gov API. The industry-sponsored single-arm study is active but not recruiting, lists 133 estimated participants versus 136 in the papers, has no posted results and was last updated on 14 October 2022." }, { "id": "clinicaltrials-nct03678883-elraglusib", "type": "trial-registry", "title": "NCT03678883: phase 1/2 elraglusib study including randomized first-line metastatic PDAC", "authors": [ "Actuate Therapeutics, Inc." ], "publisher": "ClinicalTrials.gov", "date": "2026-04-03", "url": "https://clinicaltrials.gov/study/NCT03678883", "doi": null, "pmid": null, "trialId": "NCT03678883", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Primary registry record. It remains active, not recruiting, with estimated total enrollment 350 and estimated completion June 2026. The registered Part 3 primary outcome remains one-year survival, no results are posted, and the IPD sharing field says no. This does not match the paper's revised survival-curve primary analysis or its data-sharing statement." }, { "id": "clinicaltrials-nct03690323-pelican", "type": "trial-registry", "title": "PELICAN: radiofrequency ablation plus chemotherapy versus chemotherapy alone in locally advanced pancreatic cancer", "authors": [ "Amsterdam University Medical Center", "Dutch Cancer Society", "Comprehensive Cancer Centre The Netherlands" ], "publisher": "ClinicalTrials.gov", "date": "2018-10-01", "url": "https://clinicaltrials.gov/study/NCT03690323", "doi": null, "pmid": null, "trialId": "NCT03690323", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Primary registry checked through the ClinicalTrials.gov API. The record remains status unknown, retains an estimated 228 participants and has not been updated since 1 October 2018. The 2026 paper reports 190 randomized and 188 analysed." }, { "id": "clinicaltrials-nct03697564-gemcan", "type": "trial-registry", "title": "GemCaN: gemcitabine plus nivolumab and cabiralizumab after first-line disease control", "authors": [ "University of Southern California", "Stand Up To Cancer" ], "publisher": "ClinicalTrials.gov", "date": "2024-04-23", "url": "https://clinicaltrials.gov/study/NCT03697564", "doi": null, "pmid": null, "trialId": "NCT03697564", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "A planned randomized phase 2 maintenance study completed with only two participants, both in one experimental group. It cannot answer efficacy or safety and is an accrual failure." }, { "id": "clinicaltrials-nct03762837-gbc-risk-cohort", "type": "trial-registry", "title": "Multicenter Prospective Cohort Study of Risk Factors for Gallbladder Cancer", "authors": [ "Tongji Hospital" ], "publisher": "ClinicalTrials.gov", "date": "2019-03-04", "url": "https://clinicaltrials.gov/study/NCT03762837", "doi": null, "pmid": null, "trialId": "NCT03762837", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Live registry checked. The Chinese cohort planned 100,000 participants split between benign gallbladder disease and healthy comparison, with a three-year cancer-incidence outcome. On 14 September 2026 it still says not yet recruiting, retains an estimated April 2019 start and December 2022 primary completion, and has not been verified since November 2018 or updated since March 2019. No results are posted." }, { "id": "clinicaltrials-nct03784417-eus-laser", "type": "trial-registry", "title": "Endoscopic-ultrasound-guided laser ablation in pancreatic adenocarcinoma", "authors": [ "Catholic University of the Sacred Heart" ], "publisher": "ClinicalTrials.gov", "date": "2022-02-08", "url": "https://clinicaltrials.gov/study/NCT03784417", "doi": null, "pmid": null, "trialId": "NCT03784417", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Single-group study of 1064 nm endoscopic-ultrasound-guided laser ablation. It terminated after five actual participants because the device was unavailable. The primary endpoint was one-month tumour necrosis; quality of life, progression and survival were secondary. No results are posted." }, { "id": "clinicaltrials-nct03816163-gleam", "type": "trial-registry", "title": "NCT03816163: GLEAM zolbetuximab plus gemcitabine and nab-paclitaxel", "authors": [ "Astellas Pharma Global Development, Inc." ], "publisher": "ClinicalTrials.gov", "date": "2026-07-14", "url": "https://clinicaltrials.gov/study/NCT03816163", "doi": null, "pmid": null, "trialId": "NCT03816163", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Primary current registry. Randomized open-label phase 2, actual n=393, active not recruiting, estimated completion 31 August 2026, no posted results. The primary endpoint is OS and the registry states that participant data, protocol, analysis plan and clinical study report may be shared after publication through reviewed requests." }, { "id": "clinicaltrials-nct03874897-ct041", "type": "trial-registry", "title": "Chimeric Antigen Receptor T Cells Targeting Claudin18.2 in Solid Tumors", "authors": [ "CARsgen Therapeutics Co., Ltd." ], "publisher": "ClinicalTrials.gov", "date": "2026-09-14", "url": "https://clinicaltrials.gov/study/NCT03874897", "doi": null, "pmid": null, "trialId": "NCT03874897", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Completed multicancer CLDN18.2 CAR-T registry with actual enrollment 134 and no posted results. It states that participant data may be shared. The pancreatic pooled paper combines selected infused patients from this study and NCT04581473 rather than reporting the entire registry population." }, { "id": "clinicaltrials-nct03891979-microbiome-pdac", "type": "registry", "title": "NCT03891979: antibiotics plus pembrolizumab before PDAC surgery", "authors": [ "NYU Langone Health" ], "publisher": "ClinicalTrials.gov", "date": "2020-06-26", "url": "https://clinicaltrials.gov/study/NCT03891979", "doi": null, "pmid": null, "trialId": "NCT03891979", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Single-arm antibiotic plus pembrolizumab study withdrawn after zero enrollment at the principal investigator's decision. Its primary endpoint was a 20% increase in selected T-cell activation markers, not survival or treatment contribution." }, { "id": "clinicaltrials-nct03899636-direct-rct", "type": "trial-registry", "title": "DIRECT randomized trial: mFOLFIRINOX plus NanoKnife IRE versus mFOLFIRINOX alone", "authors": [ "AngioDynamics, Inc." ], "publisher": "ClinicalTrials.gov", "date": "2025-02-12", "url": "https://clinicaltrials.gov/study/NCT03899636", "doi": null, "pmid": null, "trialId": "NCT03899636", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Primary registry. This open-label phase 3 trial randomizes eligible patients 1:1 after three months of mFOLFIRINOX and no progression. The experimental arm adds open or percutaneous NanoKnife irreversible electroporation; the control arm continues the same chemotherapy without the device. Overall survival from randomization is primary, with at least 24 months follow-up. The record lists six U.S. sites, 528 estimated participants, an estimated April 2025 completion, active-not-recruiting status and no posted results. The participant count is planned, not confirmed actual enrollment; the record was last updated on 12 February 2025." }, { "id": "clinicaltrials-nct03899649-direct-registry", "type": "trial-registry", "title": "DIRECT registry: NanoKnife IRE plus standard care versus standard care alone", "authors": [ "AngioDynamics, Inc." ], "publisher": "ClinicalTrials.gov", "date": "2025-02-12", "url": "https://clinicaltrials.gov/study/NCT03899649", "doi": null, "pmid": null, "trialId": "NCT03899649", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Primary registry for the observational half of DIRECT, distinct from randomized trial NCT03899636. The record lists 532 estimated participants, active-not-recruiting status, estimated April 2025 primary and study completion and no posted results. It was last updated on 12 February 2025. The published safety and survival reports cover smaller analysis sets and cannot be read as the planned enrollment having occurred." }, { "id": "clinicaltrials-nct04098237-pancax3", "type": "trial-registry", "title": "PANCAX-3: Pancreaze with standard care for PDAC cachexia and exocrine insufficiency", "authors": [ "Hendifar, Andrew", "Cedars-Sinai Medical Center" ], "publisher": "ClinicalTrials.gov", "date": "2026-01-23", "url": "https://clinicaltrials.gov/study/NCT04098237", "doi": null, "pmid": null, "trialId": "NCT04098237", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Single-arm study with 36 enrolled and 30 in posted results. Twenty-nine met the adherence endpoint. There was no untreated control; mean six-month weight change was -2.19 lb and several function measures showed no statistical comparison against a control. Active not recruiting, completion estimated January 2027; VIVUS collaborated." }, { "id": "clinicaltrials-nct04146298-kras-g12v-tcr", "type": "trial-registry", "title": "Mutant KRAS G12V-specific TCR Transduced T Cell Therapy for Advanced Pancreatic Cancer", "authors": [ "Changhai Hospital" ], "publisher": "ClinicalTrials.gov", "date": "2025-12-24", "url": "https://clinicaltrials.gov/study/NCT04146298", "doi": null, "pmid": null, "trialId": "NCT04146298", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Recruiting single-group phase 1/2 in one Shanghai hospital, estimated n=30, for advanced pancreatic cancer with KRAS or other RAS G12V plus HLA-A*11:01. Objective response and treatment-related events are primary; circulating engineered-cell fraction and survival are secondary. The registry does not state a participant-data-sharing plan. Completion is estimated December 2028." }, { "id": "clinicaltrials-nct04193904-mrx0518-pdac", "type": "registry", "title": "NCT04193904: MRx0518 with preoperative radiation in resectable pancreatic cancer", "authors": [ "4D pharma plc" ], "publisher": "ClinicalTrials.gov", "date": "2023-06-02", "url": "https://clinicaltrials.gov/study/NCT04193904", "doi": null, "pmid": null, "trialId": "NCT04193904", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Open-label phase 1 single arm of oral live bacterium MRx0518 plus 30 Gy radiation. Terminated for sponsor insolvency after 13 actual participants; no results posted. Clinical outcomes cannot isolate MRx0518 from radiation." }, { "id": "clinicaltrials-nct04194268-pack", "type": "trial-registry", "title": "PACK: pancreatic cancer and carbon-ion radiotherapy", "authors": [ "University Hospital Heidelberg" ], "publisher": "ClinicalTrials.gov", "date": "2026-03-10", "url": "https://clinicaltrials.gov/study/NCT04194268", "doi": null, "pmid": null, "trialId": "NCT04194268", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Current primary registry checked through the ClinicalTrials.gov API. The academic single-arm study completed 25 actual participants on 11 November 2025 and was updated on 10 March 2026, but has no posted results. Its primary endpoint is one-year overall survival." }, { "id": "clinicaltrials-nct04207944-3pc", "type": "trial-registry", "title": "Preventing an Incurable Disease: The Prevention of Progression to Pancreatic Cancer Trial (The 3P-C Trial)", "authors": [ "Duke University", "National Cancer Institute" ], "publisher": "ClinicalTrials.gov", "date": "2026-09-08", "url": "https://clinicaltrials.gov/study/NCT04207944", "doi": null, "pmid": null, "trialId": "NCT04207944", "patentId": null, "grantId": "1R01CA235677-01A1", "accessed": "2026-09-14", "notes": "Registry checked after its 8 September 2026 update. Triple-masked randomized phase 2 sulindac-versus-placebo study, estimated n=100, active not recruiting, primary completion recorded as 30 July 2026, no results posted. Its composite primary endpoint includes imaging growth, pancreatic resection and pancreatic adenocarcinoma, so it is not a cancer-incidence or mortality endpoint alone." }, { "id": "clinicaltrials-nct04214418-mekiauto", "type": "trial-registry", "title": "MEKiAUTO: cobimetinib and hydroxychloroquine with or without atezolizumab", "authors": [ "Columbia University" ], "publisher": "ClinicalTrials.gov", "date": "2024-12-18", "url": "https://clinicaltrials.gov/study/NCT04214418", "doi": null, "pmid": null, "trialId": "NCT04214418", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Terminated phase 1/2 for toxicity and lack of efficacy. Twenty-seven consented, 14 started treatment, and ten completed. The pancreatic, colorectal and mixed-tumour phase 2 cohorts never opened. Three of ten participants on cobimetinib plus hydroxychloroquine and one of four receiving added atezolizumab had a serious adverse event; no comparative efficacy result was posted." }, { "id": "clinicaltrials-nct04233866-giant", "type": "trial-registry", "title": "NCT04233866 — GIANT EA2186 registry", "authors": [ "ECOG-ACRIN Cancer Research Group", "National Cancer Institute" ], "publisher": "ClinicalTrials.gov", "date": "2026-09-14", "url": "https://clinicaltrials.gov/study/NCT04233866", "doi": null, "pmid": null, "trialId": "NCT04233866", "patentId": null, "grantId": "U10CA180820", "accessed": "2026-09-14", "notes": "Current registry check. Active, not recruiting; actual enrollment 176; actual primary completion 2026-03-02; estimated study completion 2026-12-31. Overall survival is primary. Eligibility required age at least 70, metastatic PDAC, a defined vulnerability in function, comorbidity, cognition or age at least 80, ECOG 0-2 and adequate organ function. The registry has no posted results, although 2024 and 2025 meeting abstracts report analyses." }, { "id": "clinicaltrials-nct04239573-ea2185", "type": "trial-registry", "title": "Comparing the Clinical Impact of Pancreatic Cyst Surveillance Programs and Associated Biomarkers", "authors": [ "ECOG-ACRIN Cancer Research Group", "National Cancer Institute" ], "publisher": "ClinicalTrials.gov", "date": "2026-09-04", "url": "https://clinicaltrials.gov/study/NCT04239573", "doi": null, "pmid": null, "trialId": "NCT04239573", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Live registry checked after its 4 September 2026 update. It now lists a prospective observational cohort, estimated n=770, recruiting, no posted results, and estimated completion 11 November 2034. The randomized low- and high-intensity arms are marked closed to accrual from 13 August 2024; the current primary outcome concerns baseline biomarkers and later worrisome or high-risk features. The registry does not state why randomization closed." }, { "id": "clinicaltrials-nct04281290-panect-results", "type": "trial-registry", "title": "PanECT: electrochemotherapy of the posterior resection surface after pancreatic cancer surgery", "authors": [ "University Medical Centre Ljubljana", "Institute of Oncology Ljubljana" ], "publisher": "ClinicalTrials.gov", "date": "2026-03-06", "url": "https://clinicaltrials.gov/study/NCT04281290", "doi": null, "pmid": null, "trialId": "NCT04281290", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Primary registry result record checked through the API. Ten single-arm participants enrolled; four had a post-procedure complication, one had a serious pancreatic fistula with later death, and seven died during follow-up. Disease-free and overall-survival outcome tables are not posted." }, { "id": "clinicaltrials-nct04340141-a021806", "type": "trial-registry", "title": "Alliance A021806: perioperative versus adjuvant mFOLFIRINOX for resectable pancreatic cancer", "authors": [ "Alliance for Clinical Trials in Oncology", "National Cancer Institute" ], "publisher": "ClinicalTrials.gov", "date": "2026-09-01", "url": "https://clinicaltrials.gov/study/NCT04340141", "doi": null, "pmid": null, "trialId": "NCT04340141", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Current registry checked 14 September 2026. Active, not recruiting, with actual enrolment 358. It compares eight cycles of preoperative plus four cycles of postoperative mFOLFIRINOX against upfront surgery plus twelve postoperative cycles. Overall survival from randomization is primary. Estimated primary completion is December 2028 and study completion November 2030; no results are posted. Alliance sponsors and NCI collaborates. This is the direct modern multicentre timing test and should report before another broad timing trial starts." }, { "id": "clinicaltrials-nct04390763-nis793-phase2", "type": "trial-registry", "title": "NIS793 with and without spartalizumab plus gemcitabine/nab-paclitaxel in first-line metastatic PDAC", "authors": [ "Novartis Pharmaceuticals" ], "publisher": "ClinicalTrials.gov", "date": "2026-05-19", "url": "https://clinicaltrials.gov/study/NCT04390763", "doi": null, "pmid": null, "trialId": "NCT04390763", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Terminated after a July 2023 NIS793 treatment halt and urgent safety measure. Actual enrollment 164: 11 run-in and randomized full-analysis groups of 50 NIS793 plus spartalizumab plus chemotherapy, 51 NIS793 plus chemotherapy, and 52 chemotherapy. Median OS was 10.7, 8.5 and 10.1 months respectively. Median PFS was 3.91, 5.52 and 4.37 months." }, { "id": "clinicaltrials-nct04524702-paricalcitol-hcq", "type": "trial-registry", "title": "Paricalcitol and hydroxychloroquine with gemcitabine and nab-paclitaxel in advanced pancreatic cancer", "authors": [ "Emory University" ], "publisher": "ClinicalTrials.gov", "date": "2024-06-25", "url": "https://clinicaltrials.gov/study/NCT04524702", "doi": null, "pmid": null, "trialId": "NCT04524702", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Completed single-group phase 2 with ten participants. Response was reported as 43% among seven evaluable people, median progression-free survival was six months among ten and median overall survival 6.8 months among seven; 3/10 had a serious adverse event. The small uncontrolled study and changing denominators cannot identify the contribution of either added drug." }, { "id": "clinicaltrials-nct04581473-ct041", "type": "trial-registry", "title": "Study to Evaluate the Efficacy, Safety and Pharmacokinetics of CT041 Autologous CAR T-cell Injection", "authors": [ "CARsgen Therapeutics Co., Ltd." ], "publisher": "ClinicalTrials.gov", "date": "2026-09-14", "url": "https://clinicaltrials.gov/study/NCT04581473", "doi": null, "pmid": null, "trialId": "NCT04581473", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Active, not recruiting, multicancer CLDN18.2 CAR-T registry with actual enrollment 192, a 2038 follow-up date and no posted results. The record says participant data will not be shared. Its early pancreatic cohort contributes to the 24-person pooled report; its later randomized gastric-cancer result cannot establish a pancreatic effect." }, { "id": "clinicaltrials-nct04632303-pdac-palliative", "type": "trial-registry", "title": "Early palliative care for patients with advanced pancreatic cancer", "authors": [ "Herlev Hospital", "Danish pancreatic palliative-care study group" ], "publisher": "ClinicalTrials.gov", "date": "2025-09-12", "url": "https://clinicaltrials.gov/study/NCT04632303", "doi": null, "pmid": null, "trialId": "NCT04632303", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Danish randomized pancreatic-only trial, active not recruiting, estimated n=250. The early-care arm starts within ten days and continues at least every four weeks, with exercise and nutrition referral. The main endpoint is adjusted 12-week quality-of-life change. Primary completion was recorded as 7 November 2024; full completion is estimated November 2026; no registry results are posted and individual data sharing is marked no." }, { "id": "clinicaltrials-nct04793633-linfu", "type": "trial-registry", "title": "Evaluating LINFU: A Noninvasive Method for Collecting Pancreatic Cells for the Early Detection of Pancreatic Cancer", "authors": [ "Adenocyte, LLC" ], "publisher": "ClinicalTrials.gov", "date": "2026-09-15", "url": "https://clinicaltrials.gov/study/NCT04793633", "doi": null, "pmid": null, "trialId": "NCT04793633", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Industry-sponsored prospective observational cohort, estimated n=500, enrolling by invitation, actual start 31 January 2024, estimated completion 30 December 2030 and no posted results. It combines non-focused ultrasound with intravenous microbubbles and secretin in elevated-risk or clinically suspected participants." }, { "id": "clinicaltrials-nct04821284-sonoporation", "type": "trial-registry", "title": "Sonoporation and Chemotherapy for the Treatment of Pancreatic Cancer", "authors": [ "Thomas Jefferson University", "National Cancer Institute" ], "publisher": "ClinicalTrials.gov", "date": "2025-11-04", "url": "https://clinicaltrials.gov/study/NCT04821284", "doi": null, "pmid": null, "trialId": "NCT04821284", "patentId": null, "grantId": "R01 CA199646", "accessed": "2026-09-15", "notes": "Randomized open-label US and Norway phase 1/2 trial of standard gemcitabine/nab-paclitaxel or FOLFIRINOX with or without Sonazoid microbubbles and contrast-enhanced ultrasound. Registry lists estimated n=120, active not recruiting, actual primary completion 2025-10-31, estimated study completion 2026-02-28 and no posted results." }, { "id": "clinicaltrials-nct04858009-hipec-pdac", "type": "trial-registry", "title": "Hyperthermic intraperitoneal chemotherapy for pancreatic cancer and peritoneal metastasis", "authors": [ "Mayo Clinic" ], "publisher": "ClinicalTrials.gov", "date": "2026-09-14", "url": "https://clinicaltrials.gov/study/NCT04858009", "doi": null, "pmid": null, "trialId": "NCT04858009", "patentId": null, "grantId": "NCI-2021-02990", "accessed": "2026-09-15", "notes": "Recruiting single-centre phase II study targeting 40 people, current through 14 September 2026 and estimated to complete July 2029. It selects peritoneal-only PDAC, Peritoneal Cancer Index no more than 7, high likelihood of complete cytoreduction and response to prior systemic therapy. Participants may receive up to five HIPEC procedures with nab-paclitaxel and cisplatin. Overall and progression-free survival start at cytoreduction plus HIPEC; there is no comparison arm." }, { "id": "clinicaltrials-nct04875325-radar-panc", "type": "trial-registry", "title": "RADAR-PANC: Recurrent Disease Detection After Resection of Pancreatic Adenocarcinoma Using a Standardized Surveillance Strategy", "authors": [ "UMC Utrecht", "Dutch Pancreatic Cancer Group", "University of Birmingham" ], "publisher": "ClinicalTrials.gov", "date": "2024-09-19", "url": "https://clinicaltrials.gov/study/NCT04875325", "doi": null, "pmid": null, "trialId": "NCT04875325", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Live registry checked. Status is enrolling by invitation, last verified September 2024, with estimated 306 participants, primary and study completion October 2026, and no posted results. It lists ten Dutch hospitals and the University of Birmingham. The stale verification date and moved completion estimate require investigator confirmation, but the study has not yet passed its registered completion month." }, { "id": "clinicaltrials-nct04888312-optimize1", "type": "trial-registry", "title": "OPTIMIZE-1: mitazalimab plus mFOLFIRINOX in metastatic PDAC", "authors": [ "Alligator Bioscience" ], "publisher": "ClinicalTrials.gov", "date": "2025-10-08", "url": "https://clinicaltrials.gov/study/NCT04888312", "doi": null, "pmid": null, "trialId": "NCT04888312", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Active, not recruiting, actual enrollment 94 after an added lower-dose cohort. Single-group phase 1b/2, no posted results, estimated completion June 2026. The registry cannot supply a controlled contribution estimate and has not posted the final 94-person results." }, { "id": "clinicaltrials-nct04892017-inlexisertib", "type": "trial-registry", "title": "Inlexisertib, a direct ULK1/2 inhibitor, in RAS/MAPK-mutant solid tumours", "authors": [ "Deciphera Pharmaceuticals, LLC" ], "publisher": "ClinicalTrials.gov", "date": "2026-04-08", "url": "https://clinicaltrials.gov/study/NCT04892017", "doi": null, "pmid": null, "trialId": "NCT04892017", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Non-randomized phase 1 study with actual enrolment 91, terminated on 30 March 2026 for a business decision stated not to be based on safety or efficacy. No results are posted. Earlier registry versions planned a pancreatic expansion of inlexisertib plus trametinib, but the expansion did not appear in the final record. Trial termination cannot establish biological failure or success." }, { "id": "clinicaltrials-nct04931381-organoid-advanced", "type": "trial-registry", "title": "Organoid-Guided Chemotherapy for Advanced Pancreatic Cancer", "authors": [ "Changhai Hospital" ], "publisher": "ClinicalTrials.gov", "date": "2022-08-02", "url": "https://clinicaltrials.gov/study/NCT04931381", "doi": null, "pmid": null, "trialId": "NCT04931381", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Randomized open-label phase 3 plan for 100 advanced-PDAC patients, comparing organoid-guided with physician-chosen chemotherapy. Six-month disease control is primary and survival is secondary. The record is now status unknown, was last verified July 2022, retains an estimated May 2025 completion and posts no results." }, { "id": "clinicaltrials-nct04931394-organoid-adjuvant", "type": "trial-registry", "title": "Organoid-Guided Adjuvant Chemotherapy for Pancreatic Cancer", "authors": [ "Changhai Hospital" ], "publisher": "ClinicalTrials.gov", "date": "2022-08-02", "url": "https://clinicaltrials.gov/study/NCT04931394", "doi": null, "pmid": null, "trialId": "NCT04931394", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Randomized open-label phase 3 plan for 200 resected PDAC patients, comparing organoid-guided with physician-chosen adjuvant chemotherapy. Disease-free time is primary and survival secondary. The record is status unknown, was last verified July 2022, retains an estimated May 2025 completion and posts no results." }, { "id": "clinicaltrials-nct04935359-danis2", "type": "trial-registry", "title": "daNIS-2: NIS793 plus gemcitabine/nab-paclitaxel versus placebo plus gemcitabine/nab-paclitaxel in first-line metastatic PDAC", "authors": [ "Novartis Pharmaceuticals" ], "publisher": "ClinicalTrials.gov", "date": "2026-05-19", "url": "https://clinicaltrials.gov/study/NCT04935359", "doi": null, "pmid": null, "trialId": "NCT04935359", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Completed randomized, double-blind phase 3 with actual enrollment 511: safety run-in n=21, then NIS793 plus chemotherapy n=245 versus placebo plus chemotherapy n=245. Median OS was 9.2 versus 11.2 months; PFS 4.6 versus 5.4 months; ORR 21.6% versus 25.3%. Serious adverse events affected 144/239=60.3% versus 110/241=45.6%, and fatal adverse events 12/239=5.0% versus 8/241=3.3%. NIS793 was halted in July 2023 for unfavorable benefit-risk." }, { "id": "clinicaltrials-nct04939610-lumiere", "type": "trial-registry", "title": "LuMIERE: 177Lu-FAP-2286 in advanced solid tumours", "authors": [ "Novartis Pharmaceuticals" ], "publisher": "ClinicalTrials.gov", "date": "2026-07-24", "url": "https://clinicaltrials.gov/study/NCT04939610", "doi": null, "pmid": null, "trialId": "NCT04939610", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Recruiting, non-randomized phase 1/2 with estimated enrolment 222 and estimated primary completion December 2027. Phase 2 includes imaging-selected PDAC monotherapy after one or two advanced-disease regimens and first-line PDAC combination treatment with modified FOLFIRINOX. Objective response is the phase 2 efficacy endpoint. No comparative control is listed." }, { "id": "clinicaltrials-nct04975217-fmt-pdac", "type": "registry", "title": "NCT04975217: fecal microbiota transfer before PDAC surgery", "authors": [ "M.D. Anderson Cancer Center" ], "publisher": "ClinicalTrials.gov", "date": "2026-05-22", "url": "https://clinicaltrials.gov/study/NCT04975217", "doi": null, "pmid": null, "trialId": "NCT04975217", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Early phase 1, open-label, single-arm transfer by colonoscopy plus weekly capsules before resection; estimated n=10. The record was suspended at investigator request in March 2026, with no results posted and an estimated December 2026 completion. Primary outcome is adverse events through 30 days after surgery." }, { "id": "clinicaltrials-nct04990609-pancardinal1", "type": "trial-registry", "title": "EUS-RFA PANCARDINAL-1 Trial", "authors": [ "University of Texas Health Science Center at Houston", "National Cancer Institute" ], "publisher": "ClinicalTrials.gov", "date": "2026-07-29", "url": "https://clinicaltrials.gov/study/NCT04990609", "doi": null, "pmid": null, "trialId": "NCT04990609", "patentId": null, "grantId": "R01CA277161-01A1", "accessed": "2026-09-15", "notes": "Primary registry checked through the API. Recruiting single-arm phase 2 study with estimated enrollment 60 and estimated completion May 2028. The primary endpoint is completing chemotherapy plus repeated EUS-RFA and later undergoing resection; disease-free survival is secondary and there is no same-care control." }, { "id": "clinicaltrials-nct05013216-kras-prevention", "type": "trial-registry", "title": "Mutant-KRAS long-peptide vaccine for people at high risk of pancreatic cancer", "authors": [ "Sidney Kimmel Comprehensive Cancer Center at Johns Hopkins" ], "publisher": "ClinicalTrials.gov", "date": "2026-06-30", "url": "https://clinicaltrials.gov/study/NCT05013216", "doi": null, "pmid": null, "trialId": "NCT05013216", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Recruiting single-centre, non-randomized prevention study, estimated n=37 with completion in May 2031. Its primary outcomes are severe toxicity and change in mutant-KRAS-specific T cells, not pancreatic-cancer incidence or death. The first published cohort included 20 participants." }, { "id": "clinicaltrials-nct05077800-elraglusib-folfirinox", "type": "trial-registry", "title": "NCT05077800: FOLFIRINOX with elraglusib and losartan in untreated metastatic PDAC", "authors": [ "Massachusetts General Hospital", "Actuate Therapeutics, Inc.", "National Cancer Institute" ], "publisher": "ClinicalTrials.gov", "date": "2026-09-11", "url": "https://clinicaltrials.gov/study/NCT05077800", "doi": null, "pmid": null, "trialId": "NCT05077800", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Primary registry record, active not recruiting, estimated n=70 and estimated completion 31 December 2026, with no posted results. It labels allocation randomized but defines the primary PFS analysis against historical controls. It states that de-identified participant data may be shared under agreement no earlier than one year after publication." }, { "id": "clinicaltrials-nct05114213-maspac", "type": "trial-registry", "title": "MASPAC: randomized MRI-guided adaptive radiation for pain control in metastatic pancreatic cancer", "authors": [ "Ludwig-Maximilians University of Munich", "Heidelberg University", "University of Zurich" ], "publisher": "ClinicalTrials.gov", "date": "2025-12-05", "url": "https://clinicaltrials.gov/study/NCT05114213", "doi": null, "pmid": null, "trialId": "NCT05114213", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Current primary registry. The randomized pain-control trial was terminated for low accrual after nine actual participants and closed in November 2025. No results are posted. Both arms continued standard combination chemotherapy." }, { "id": "clinicaltrials-nct05164458-ibi389", "type": "trial-registry", "title": "NCT05164458: IBI389 alone and with sintilimab in advanced malignancies", "authors": [ "Innovent Biologics (Suzhou) Co. Ltd." ], "publisher": "ClinicalTrials.gov", "date": "2023-12-28", "url": "https://clinicaltrials.gov/study/NCT05164458", "doi": null, "pmid": null, "trialId": "NCT05164458", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Primary registry last verified in December 2023. Status is unknown after estimated primary completion in April 2024 and study completion in September 2024; estimated n=320, non-randomized and open-label, with no posted results. The stale record cannot establish whether the programme completed or why mature PDAC efficacy was omitted from the later full paper." }, { "id": "clinicaltrials-nct05174169-circulate-us", "type": "registry", "title": "CIRCULATE-US: Colon Adjuvant Chemotherapy Based on Evaluation of Residual Disease", "authors": [ "NRG Oncology" ], "publisher": "ClinicalTrials.gov", "date": "2026-09-02", "url": "https://clinicaltrials.gov/study/NCT05174169", "doi": null, "pmid": null, "trialId": "NCT05174169", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Recruiting randomized phase 2/3 trial with estimated enrolment 1,912. It separately tests no chemotherapy with serial monitoring in ctDNA-negative resected stage II/III colon cancer and mFOLFIRINOX intensification in ctDNA-positive disease. Disease-free survival is primary, overall survival secondary, estimated primary completion March 2029 and full completion March 2030. The registry names the commercial Signatera assay. No results are posted." }, { "id": "clinicaltrials-nct05178628-impact-pro", "type": "trial-registry", "title": "imPaCT-PRO: tinzaparin plus chemotherapy versus chemotherapy in advanced pancreatic cancer", "authors": [ "Karamouzis, Michalis", "Institute of Molecular Medicine and Biomedical Research" ], "publisher": "ClinicalTrials.gov", "date": "2022-03-29", "url": "https://clinicaltrials.gov/study/NCT05178628", "doi": null, "pmid": null, "trialId": "NCT05178628", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Registry planned an open-label phase 3 randomized study of therapeutic-dose tinzaparin with gemcitabine/nab-paclitaxel versus chemotherapy alone, estimated n=450 at two Greek sites. PFS and VTE were co-primary. Status is unknown, the record has not been updated since March 2022, the December 2024 completion estimate has passed and no results are posted." }, { "id": "clinicaltrials-nct05221320-ulixertinib-hcq", "type": "trial-registry", "title": "Ulixertinib plus hydroxychloroquine in advanced gastrointestinal cancers", "authors": [ "BioMed Valley Discoveries, Inc." ], "publisher": "ClinicalTrials.gov", "date": "2025-09-25", "url": "https://clinicaltrials.gov/study/NCT05221320", "doi": null, "pmid": null, "trialId": "NCT05221320", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Non-randomized phase 2 basket trial, terminated for lack of enrolment in the remaining open baskets after 47 total participants. The pancreatic basket enrolled 18: objective response was 0/18, median progression-free survival was 1.9 months among 16, and 10/18 had at least one serious adverse event. No basket entered stage 2 expansion." }, { "id": "clinicaltrials-nct05245877-preposterous-pancreas", "type": "trial-registry", "title": "PREPOSTEROUS Pancreas: thromboprophylaxis before versus after pancreatic surgery", "authors": [ "Sallinen, Ville", "Helsinki University Central Hospital" ], "publisher": "ClinicalTrials.gov", "date": "2025-02-14", "url": "https://clinicaltrials.gov/study/NCT05245877", "doi": null, "pmid": null, "trialId": "NCT05245877", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Recruiting academic phase 4 randomized trial, estimated n=800 at five locations. It compares LMWH started 2-14 hours before pancreatic surgery with postoperative start. Thirty-day VTE is primary and post-pancreatectomy hemorrhage is a secondary endpoint. Eligibility includes operations for non-cancer indications; its estimated August 2026 primary date has passed and no results are posted." }, { "id": "clinicaltrials-nct05254171-aspire", "type": "trial-registry", "title": "ASPIRE: gemcitabine and nab-paclitaxel with or without ivospemin", "authors": [ "Panbela Therapeutics, Inc." ], "publisher": "ClinicalTrials.gov", "date": "2024-11-06", "url": "https://clinicaltrials.gov/study/NCT05254171", "doi": null, "pmid": null, "trialId": "NCT05254171", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "The U.S. registry still labels this quadruple-masked randomized phase 2/3 trial recruiting, estimates n=600, and gives overall survival as the main endpoint. Its status was verified in July 2024, its last update was posted 6 November 2024, and its estimated 29 August 2026 primary completion has passed. No results are posted and participant-level data sharing is marked no." }, { "id": "clinicaltrials-nct05262452-hifu-folfirinox", "type": "trial-registry", "title": "Concurrent FOLFIRINOX plus HIFU for locally advanced or borderline resectable pancreatic cancer", "authors": [ "Seoul National University Hospital" ], "publisher": "ClinicalTrials.gov", "date": "2022-10-25", "url": "https://clinicaltrials.gov/study/NCT05262452", "doi": null, "pmid": null, "trialId": "NCT05262452", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Primary registry checked through the API. The single-arm 60-person ALPIUS 900 study has unknown status, no results, a passed March 2023 primary completion estimate and a record last updated in October 2022. Its protocol says the sample size was based on precedent rather than an effect calculation." }, { "id": "clinicaltrials-nct05321992-echec", "type": "trial-registry", "title": "Evaluating a Community-Based Behaviour Change Communication Model to Prevent Cholangiocarcinoma in Khon Kaen, Thailand", "authors": [ "McMaster University", "Khon Kaen University" ], "publisher": "ClinicalTrials.gov", "date": "2024-11-21", "url": "https://clinicaltrials.gov/study/NCT05321992", "doi": null, "pmid": null, "trialId": "NCT05321992", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Registry record checked through the ClinicalTrials.gov API. Open-label randomized prevention study of the eCHEC programme versus usual care, estimated n=480, listed recruiting, last updated November 2024, with estimated completion April 2026 and no posted results. All primary endpoints occur at 12 months and concern self-reported diet, smoking, alcohol and praziquantel use or Opisthorchis stool-test positivity and concentration; cancer incidence and mortality are not endpoints." }, { "id": "clinicaltrials-nct05351983-organoid-screen", "type": "trial-registry", "title": "Patient-derived Organoids Drug Screen in Pancreatic Cancer", "authors": [ "Jan Schmidt", "Daniel Helbling", "Marianna Kruithof-De Julio" ], "publisher": "ClinicalTrials.gov", "date": "2025-09-25", "url": "https://clinicaltrials.gov/study/NCT05351983", "doi": null, "pmid": null, "trialId": "NCT05351983", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Completed single-group feasibility study with 60 actual participants. It defines organoid generation within 10 days and a drug hit as at least 50% lower viability. The registry was last verified December 2023, posts no results and does not assign treatment from the assay." }, { "id": "clinicaltrials-nct05371223-nab-pipac", "type": "trial-registry", "title": "Combined nab-paclitaxel PIPAC with systemic nab-paclitaxel-gemcitabine for pancreatic cancer peritoneal metastases", "authors": [ "Fondazione Policlinico Universitario Agostino Gemelli IRCCS" ], "publisher": "ClinicalTrials.gov", "date": "2024-02-26", "url": "https://clinicaltrials.gov/study/NCT05371223", "doi": null, "pmid": null, "trialId": "NCT05371223", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Single-centre, open-label, single-group phase II study with estimated enrollment 38. The plan gives three courses, each containing two cycles of intravenous nab-paclitaxel-gemcitabine and one nab-paclitaxel PIPAC. Disease-control rate is primary. Eligibility requires ECOG 0-1, at least three months expected survival, no extra-abdominal spread, no bowel obstruction, no clinical deterioration and no progression after the first two months of systemic nab-paclitaxel-gemcitabine. The record was last verified February 2024, retains last-known recruiting status but is now labelled unknown; estimated July 2025 completion has passed and no results are posted." }, { "id": "clinicaltrials-nct05432193-frontier", "type": "trial-registry", "title": "FRONTIER: 177Lu-PNT6555 dose escalation in FAP-positive solid tumours", "authors": [ "POINT Biopharma, a wholly owned subsidiary of Eli Lilly and Company" ], "publisher": "ClinicalTrials.gov", "date": "2025-06-25", "url": "https://clinicaltrials.gov/study/NCT05432193", "doi": null, "pmid": null, "trialId": "NCT05432193", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Terminated phase 1 after 20 of a planned 30 people. The registry gives the reason only as business decision. Eligible cancers included PDAC and cholangiocarcinoma. Patients needed at least half of lesions to have scan uptake at least 1.5 times mean liver uptake. No results are posted, so neither scientific failure nor success can be inferred from termination." }, { "id": "clinicaltrials-nct05458219-ibi343", "type": "trial-registry", "title": "NCT05458219: first-in-human IBI343 in advanced solid tumours", "authors": [ "Innovent Biologics (Suzhou) Co. Ltd." ], "publisher": "ClinicalTrials.gov", "date": "2026-05-04", "url": "https://clinicaltrials.gov/study/NCT05458219", "doi": null, "pmid": null, "trialId": "NCT05458219", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Primary current registry. Recruiting phase 1 platform with estimated n=470 and completion in December 2027; no posted results. It contains multiple solid-tumour dose and combination groups, so the 83-person PDAC report is a subset rather than the trial denominator." }, { "id": "clinicaltrials-nct05462496-antibiotic-pembro-pdac", "type": "registry", "title": "NCT05462496: antibiotics and pembrolizumab after chemotherapy in resectable PDAC", "authors": [ "Icahn School of Medicine at Mount Sinai" ], "publisher": "ClinicalTrials.gov", "date": "2024-10-15", "url": "https://clinicaltrials.gov/study/NCT05462496", "doi": null, "pmid": null, "trialId": "NCT05462496", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Recruiting single-arm phase 2 pilot, estimated n=25. Five FOLFIRINOX cycles are followed by ciprofloxacin, metronidazole, one pembrolizumab dose and surgery. The primary endpoint is a 20% increase in any listed T-cell activation marker; estimated primary completion August 2027. No component control is present." }, { "id": "clinicaltrials-nct05466799-tripp-ffx", "type": "trial-registry", "title": "TRIPP-FFX: FOLFIRINOX versus OncoSil in addition to FOLFIRINOX in locally advanced pancreatic adenocarcinoma", "authors": [ "OncoSil Medical Limited" ], "publisher": "ClinicalTrials.gov", "date": "2026-02-13", "url": "https://clinicaltrials.gov/study/NCT05466799", "doi": null, "pmid": null, "trialId": "NCT05466799", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Primary registry checked through the API. Industry-sponsored open-label randomized phase 2 study with actual enrollment 88. Primary completion was January 2026; final completion is estimated July 2027. It fixes FOLFIRINOX in both arms. Safety and 16-week local disease control are primary; progression, survival, symptoms, quality of life and resection are secondary. No results are posted." }, { "id": "clinicaltrials-nct05486572-premium", "type": "trial-registry", "title": "Preventing Liver Cancer Mortality Through Imaging With Ultrasound vs. MRI", "authors": [ "VA Office of Research and Development" ], "publisher": "ClinicalTrials.gov", "date": "2026-06-17", "url": "https://clinicaltrials.gov/study/NCT05486572", "doi": null, "pmid": null, "trialId": "NCT05486572", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Recruiting randomized VA trial, estimated n=4,700. Abbreviated contrast MRI plus AFP is compared with ultrasound plus AFP every six months. HCC mortality is the eight-year primary endpoint; estimated primary completion is September 2030 and full completion September 2031. No results are posted." }, { "id": "clinicaltrials-nct05546411-nis793-neoadjuvant", "type": "trial-registry", "title": "Neoadjuvant NIS793 plus mFOLFIRINOX in resectable and borderline-resectable PDAC", "authors": [ "Dana-Farber Cancer Institute", "Novartis Pharmaceuticals" ], "publisher": "ClinicalTrials.gov", "date": "2025-03-06", "url": "https://clinicaltrials.gov/study/NCT05546411", "doi": null, "pmid": null, "trialId": "NCT05546411", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Terminated after eight participants because Novartis stopped all clinical development of NIS793 in pancreatic cancer. Six received NIS793 plus mFOLFIRINOX and two received mFOLFIRINOX. The small record cannot estimate efficacy." }, { "id": "clinicaltrials-nct05549960-laser-model", "type": "trial-registry", "title": "Mathematical treatment planning for endoscopic-ultrasound-guided laser ablation of pancreatic lesions", "authors": [ "Campus Bio-Medico University", "Politecnico di Milano" ], "publisher": "ClinicalTrials.gov", "date": "2023-06-26", "url": "https://clinicaltrials.gov/study/NCT05549960", "doi": null, "pmid": null, "trialId": "NCT05549960", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Single-group study planned to test model-selected power and time for a 980 nm laser in ten mixed pancreatic lesions. Registry status is unknown, last verified June 2023, estimated completion December 2023 and no results posted. It mixes pancreatic adenocarcinoma, neuroendocrine tumours and renal-cell metastases." }, { "id": "clinicaltrials-nct05585554-lap-ablate", "type": "trial-registry", "title": "LAP-ABLATE: chemotherapy alone versus chemotherapy plus ablative MRI-guided adaptive radiation", "authors": [ "ViewRay, Inc." ], "publisher": "ClinicalTrials.gov", "date": "2022-10-19", "url": "https://clinicaltrials.gov/study/NCT05585554", "doi": null, "pmid": null, "trialId": "NCT05585554", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Primary registry checked through the ClinicalTrials.gov API. The planned 267-person randomized phase 3 trial remains not yet recruiting, has no results, an estimated July 2023 start and was last updated on 19 October 2022. It would compare induction chemotherapy alone with the same route plus 50 Gy in five MRIdian fractions." }, { "id": "clinicaltrials-nct05601323-suizenji", "type": "trial-registry", "title": "Randomized study of Suizenji HIFU plus chemotherapy versus chemotherapy in unresectable pancreatic cancer", "authors": [ "SONIRE Therapeutics Inc." ], "publisher": "ClinicalTrials.gov", "date": "2026-03-12", "url": "https://clinicaltrials.gov/study/NCT05601323", "doi": null, "pmid": null, "trialId": "NCT05601323", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Industry-sponsored, open-label randomized study with estimated enrollment of 90 and overall survival as primary endpoint. It is active but not recruiting, has no results and estimates completion in May 2027. Chemotherapy is physician choice in both groups." }, { "id": "clinicaltrials-nct05642962-pertzye-dose", "type": "trial-registry", "title": "Pertzye dose comparison in untreated stage-IV PDAC receiving FOLFIRINOX", "authors": [ "Memorial Sloan Kettering Cancer Center", "O'Reilly, Eileen" ], "publisher": "ClinicalTrials.gov", "date": "2026-02-25", "url": "https://clinicaltrials.gov/study/NCT05642962", "doi": null, "pmid": null, "trialId": "NCT05642962", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Randomized double-blind comparison of standard versus higher-dose pancrelipase, not PERT versus no PERT. Active not recruiting with 17 actual participants; four-month BMI is the sole listed primary outcome, completion estimated November 2026 and no results posted." }, { "id": "clinicaltrials-nct05673811-virage", "type": "trial-registry", "title": "Study of VCN-01 in Combination With Chemotherapy in Metastatic Pancreatic Cancer", "authors": [ "Theriva Biologics, Inc." ], "publisher": "ClinicalTrials.gov", "date": "2025-04-16", "url": "https://clinicaltrials.gov/study/NCT05673811", "doi": null, "pmid": null, "trialId": "NCT05673811", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Completed randomized open-label phase 2b trial. Registry records actual enrollment of 112, completion on 28 March 2025 and no posted results. It compares gemcitabine/nab-paclitaxel with or without up to two intravenous VCN-01 doses and names overall survival plus safety as primary endpoints." }, { "id": "clinicaltrials-nct05726864-amplify7p", "type": "trial-registry", "title": "AMPLIFY-7P: ELI-002 7P in KRAS/NRAS-mutated solid tumours", "authors": [ "Elicio Therapeutics" ], "publisher": "ClinicalTrials.gov", "date": "2026-07-10", "url": "https://clinicaltrials.gov/study/NCT05726864", "doi": null, "pmid": null, "trialId": "NCT05726864", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Current registry record: active but not recruiting phase 1/2, 158 actual participants across all stages and 28 sites. The randomized PDAC phase compares ELI-002 7P with observation and uses disease-free survival through week 150 as its primary outcome. Estimated study completion is November 2026." }, { "id": "clinicaltrials-nct05732831-vopimetostat", "type": "trial-registry", "title": "Vopimetostat in patients with MTAP-deleted solid tumours", "authors": [ "Tango Therapeutics, Inc." ], "publisher": "ClinicalTrials.gov", "date": "2026-06-26", "url": "https://clinicaltrials.gov/study/NCT05732831", "doi": null, "pmid": null, "trialId": "NCT05732831", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Recruiting open-label nonrandomized phase 1/2 with estimated enrollment 225 and completion June 2028. It tests vopimetostat alone and with pembrolizumab in confirmed homozygous MTAP-deleted tumours. No registry results are posted." }, { "id": "clinicaltrials-nct05743959-btc-mrd", "type": "trial-registry", "title": "ctDNA Detection of MRD in Predicting Postoperative Recurrence in Biliary Tract Cancers: A Multicenter Prospective Trial", "authors": [ "The First Affiliated Hospital with Nanjing Medical University" ], "publisher": "ClinicalTrials.gov", "date": "2023-05-22", "url": "https://clinicaltrials.gov/study/NCT05743959", "doi": null, "pmid": null, "trialId": "NCT05743959", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "ClinicalTrials.gov API record: prospective observational cohort, estimated n=100, with no intervention. It now has unknown status, was last updated May 2023, retains estimated primary completion December 2024 and study completion December 2025, and has no posted results. Its primary endpoint is relapse-free survival; it cannot estimate the benefit of acting on ctDNA." }, { "id": "clinicaltrials-nct05788744-circpac", "type": "trial-registry", "title": "Implementing ctDNA and Circular DNA in Patients With Localized Pancreatic Cancer", "authors": [ "Aalborg University Hospital" ], "publisher": "ClinicalTrials.gov", "date": "2023-03-29", "url": "https://clinicaltrials.gov/study/NCT05788744", "doi": null, "pmid": null, "trialId": "NCT05788744", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "CIRCPAC is a recruiting, open-label randomized Danish study across five listed hospitals with estimated enrolment 1,000. The intervention arm receives three-monthly ctDNA testing; a positive result triggers CT and EUS every three months, while the control follows Danish surveillance. Three-year disease-free and overall survival are listed among primary outcomes. Status was verified March 2026; primary completion is estimated January 2029 and study completion January 2032. Registry design only, with no posted result." }, { "id": "clinicaltrials-nct05802394-map03", "type": "trial-registry", "title": "The Value of Molecular Residual Disease Monitoring Based on ctDNA in Borderline Resectable or Locally Advanced Pancreatic Cancer", "authors": [ "Peking Union Medical College Hospital" ], "publisher": "ClinicalTrials.gov", "date": "2023-04-06", "url": "https://clinicaltrials.gov/study/NCT05802394", "doi": null, "pmid": null, "trialId": "NCT05802394", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "MAP-03 is an open-label randomized Chinese study with estimated enrolment 100. Patients whose MRD rises at first imaging without radiographic progression switch to an unspecified later-line regimen or continue current therapy. Resection rate is the primary endpoint, with survival secondary. The registry still says recruiting but was last verified March 2023; its estimated February 2026 primary-completion date has passed without posted results. Status, treatment definitions and power therefore require direct confirmation." }, { "id": "clinicaltrials-nct05911217-satricel-adjuvant", "type": "trial-registry", "title": "A Study to Evaluate the Efficacy and Safety of CT041 After Adjuvant Chemotherapy for Pancreatic Cancer", "authors": [ "CARsgen Therapeutics Co., Ltd." ], "publisher": "ClinicalTrials.gov", "date": "2024-05-24", "url": "https://clinicaltrials.gov/study/NCT05911217", "doi": null, "pmid": null, "trialId": "NCT05911217", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Recruiting, open-label, single-arm phase 1b planned for 20 resected CLDN18.2-positive PDAC patients with abnormal CA19-9 after adjuvant chemotherapy but no visible recurrence. Disease-free survival is primary. The registry says data will not be shared and was last verified in May 2024 despite a 2025 six-patient report." }, { "id": "clinicaltrials-nct05924789-gaspar", "type": "trial-registry", "title": "GASPAR — Epidemiological Characteristics of Gastric and Pancreatic Cancers in Latin America", "authors": [ "Latin American Cooperative Oncology Group" ], "publisher": "ClinicalTrials.gov", "date": "2023-06-29", "url": "https://clinicaltrials.gov/study/NCT05924789", "doi": null, "pmid": null, "trialId": "NCT05924789", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Registry lists Renata D'Alpino Peixoto as principal investigator, LACOG as sponsor, Astellas and GAICO as collaborators, observational cohort design, six named sites in Argentina and Brazil, completed status and actual enrolment 120. The 2025 meeting abstract instead reports 201 people from seven sites in Brazil, Argentina and Colombia; the unexplained count difference is retained as a data-quality limit." }, { "id": "clinicaltrials-nct05927298-prosper-panc", "type": "trial-registry", "title": "Province of Ontario Strategy for Personalized Management of Pancreatic Cancer Trial", "authors": [ "University Health Network, Toronto" ], "publisher": "ClinicalTrials.gov", "date": "2026-02-20", "url": "https://clinicaltrials.gov/study/NCT05927298", "doi": null, "pmid": null, "trialId": "NCT05927298", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Active, not recruiting, prospective Ontario observational cohort with estimated enrolment 200 and completion March 2027. It combines whole-genome and RNA sequencing with organoids and measures precision-matched treatment use. It can measure delivery and correlation, not organoid-guidance benefit." }, { "id": "clinicaltrials-nct06015659-azenosertib", "type": "trial-registry", "title": "NCT06015659: azenosertib plus gemcitabine in second-line advanced pancreatic adenocarcinoma", "authors": [ "Dana-Farber Cancer Institute", "Lustgarten Foundation", "Stand Up To Cancer", "Zentalis Pharmaceuticals" ], "publisher": "ClinicalTrials.gov", "date": "2026-05-04", "url": "https://clinicaltrials.gov/study/NCT06015659", "doi": null, "pmid": null, "trialId": "NCT06015659", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Primary current registry. Active not recruiting, actual n=12, actual primary completion 6 August 2025, estimated study completion June 2027 and no public results in the current record. The study is single-arm and open-label. Its original design intended the first 12 only as a safety lead-in before a two-stage efficacy expansion." }, { "id": "clinicaltrials-nct06040801-frail-gi", "type": "trial-registry", "title": "NCT06040801 — geriatric intervention in frail gastric, biliary and pancreatic cancer", "authors": [ "Chang Gung Memorial Hospital", "Chou, Wen-Chi" ], "publisher": "ClinicalTrials.gov", "date": "2025-02-13", "url": "https://clinicaltrials.gov/study/NCT06040801", "doi": null, "pmid": null, "trialId": "NCT06040801", "patentId": null, "grantId": "CMRPG3L1611", "accessed": "2026-09-14", "notes": "Single-site Taiwan supportive-care study estimated to enroll 138 people aged at least 65 with unresectable gastric, biliary or pancreatic cancer. The registry still says recruiting after its estimated 2026-07-31 completion date. Allocation is nonrandomized: people impaired in at least two domains receive geriatric intervention, while non-frail people receive no added intervention. Three-month chemotherapy completion, toxicity and quality of life are primary. The design cannot isolate intervention effect from baseline frailty, and PDAC-specific power is not stated." }, { "id": "clinicaltrials-nct06048367-cnsi-fe", "type": "trial-registry", "title": "Carbon nanoparticle-loaded iron CNSI-Fe(II) in advanced solid tumours", "authors": [ "Sichuan Enray Pharmaceutical Sciences Company" ], "publisher": "ClinicalTrials.gov", "date": "2025-08-03", "url": "https://clinicaltrials.gov/study/NCT06048367", "doi": null, "pmid": null, "trialId": "NCT06048367", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Completed company-sponsored 3+3 dose escalation with 19 actual participants across five dose groups. Fifteen completed and four withdrew. Only seven people received a second dose. The maximum tolerated dose was not reached; 150 mg was the maximum administered dose. The registry reports safety, short local imaging and blood iron, not a randomized drug-resensitization endpoint." }, { "id": "clinicaltrials-nct06084234-tracer", "type": "trial-registry", "title": "National Liver Cancer Screening Trial", "authors": [ "University of Texas Southwestern Medical Center", "National Cancer Institute" ], "publisher": "ClinicalTrials.gov", "date": "2026-08-12", "url": "https://clinicaltrials.gov/study/NCT06084234", "doi": null, "pmid": null, "trialId": "NCT06084234", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Recruiting randomized phase 4 trial, estimated n=5,500. GALAD is compared with ultrasound with or without AFP every six months. Primary late-stage-HCC completion is estimated for December 2029; full completion for December 2034. Secondary outcomes include curative treatment and physical, financial and psychological harms. No results are posted." }, { "id": "clinicaltrials-nct06099119-pertseverance", "type": "trial-registry", "title": "PERTseverance: immediate versus delayed Creon in unresectable pancreatic cancer", "authors": [ "Domínguez-Muñoz, J. Enrique", "Hospital Clinico Universitario de Santiago" ], "publisher": "ClinicalTrials.gov", "date": "2025-06-06", "url": "https://clinicaltrials.gov/study/NCT06099119", "doi": null, "pmid": null, "trialId": "NCT06099119", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Recruiting open-label multicentre randomized trial, estimated n=100. It compares immediate Creon with delayed treatment after three months in selected unresectable head tumours with duct dilation, at least 5% weight loss and expected survival at least six months. Primary endpoint is six-month weight change; estimated primary completion June 2026 has passed without posted results." }, { "id": "clinicaltrials-nct06132087-proactive", "type": "trial-registry", "title": "PROACTIVE pilot: resection after chemotherapy for locally advanced pancreatic cancer involving visceral arteries", "authors": [ "University of Southern California", "National Cancer Institute" ], "publisher": "ClinicalTrials.gov", "date": "2026-09-15", "url": "https://clinicaltrials.gov/study/NCT06132087", "doi": null, "pmid": null, "trialId": "NCT06132087", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Registry checked 15 September 2026. Active, not recruiting, single-arm USC and NCI pilot whose estimated enrolment is two. It requires initially unresectable stage III disease, at least four chemotherapy cycles without progression and good performance status, then performs staging and resection. Complete-margin resection is primary; survival is secondary and no results are posted. It cannot estimate surgery's treatment effect." }, { "id": "clinicaltrials-nct06160323-upfront-cgn-cpn", "type": "trial-registry", "title": "Upfront EUS celiac ganglion or plexus neurolysis versus a conventional step-up approach for painful inoperable pancreatic cancer", "authors": [ "Chinese University of Hong Kong" ], "publisher": "ClinicalTrials.gov", "date": "2026-05-06", "url": "https://clinicaltrials.gov/study/NCT06160323", "doi": null, "pmid": null, "trialId": "NCT06160323", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Recruiting 94-person randomized, participant- and assessor-masked study. The intervention is direct celiac ganglion neurolysis when ganglia are visible and celiac plexus neurolysis otherwise; control is guideline pain treatment with optional rescue neurolysis after four weeks. Three-month mean pain change is primary; opioid dose, adverse effects, quality of life and 48-week survival are secondary. It excludes current opioid users and people expected to live less than three months. Completion is estimated July 2028." }, { "id": "clinicaltrials-nct06192719-eulat-gbc", "type": "trial-registry", "title": "EULAT Eradicate GBC", "authors": [ "Centre Paul Strauss", "Heidelberg University Hospital", "Latin American partner network" ], "publisher": "ClinicalTrials.gov", "date": "2026-02-18", "url": "https://clinicaltrials.gov/study/NCT06192719", "doi": null, "pmid": null, "trialId": "NCT06192719", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Live registry checked. Recruiting prospective case-control biorepository with estimated enrolment 15,000 and completion in December 2028. Groups include cancer or dysplasia and gallstone patients already scheduled for cholecystectomy across Argentina, Bolivia, Chile, France and Peru. Samples span blood, saliva, urine, bile and stool. The primary outcome is cancer or dysplasia at inclusion or pathology; this can derive markers in operated or diagnosed groups but does not by itself validate future risk or the effect of preventive surgery." }, { "id": "clinicaltrials-nct06211933-puls-hifu", "type": "trial-registry", "title": "PULS phase 1/2 study of HIFU for unresectable pancreatic tumours", "authors": [ "Centre Léon Bérard" ], "publisher": "ClinicalTrials.gov", "date": "2026-09-08", "url": "https://clinicaltrials.gov/study/NCT06211933", "doi": null, "pmid": null, "trialId": "NCT06211933", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Primary registry checked through the API. The estimated 26-person single-arm study is suspended because an industrial partner decided to close it permanently to new participants. No actual enrolled count or results are posted." }, { "id": "clinicaltrials-nct06231160-mwa-cohort", "type": "trial-registry", "title": "Clinical comparative study of systemic therapy combined with microwave ablation and systemic therapy for pancreatic cancer", "authors": [ "Chinese PLA General Hospital" ], "publisher": "ClinicalTrials.gov", "date": "2024-01-30", "url": "https://clinicaltrials.gov/study/NCT06231160", "doi": null, "pmid": null, "trialId": "NCT06231160", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Primary registry checked through the API. Prospective observational two-cohort design with non-probability sampling and estimated enrollment 185. It was last known not yet recruiting, has not been verified since January 2024, and is now classed unknown status. Overall survival is primary and no results are posted." }, { "id": "clinicaltrials-nct06276764-linfu-ipmn", "type": "trial-registry", "title": "The LINFU U.S. Registry in Patients With IPMN", "authors": [ "Adenocyte, LLC" ], "publisher": "ClinicalTrials.gov", "date": "2025-03-07", "url": "https://clinicaltrials.gov/study/NCT06276764", "doi": null, "pmid": null, "trialId": "NCT06276764", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Industry-sponsored observational IPMN registry, estimated n=500, enrolling by invitation, actual start 15 October 2024, estimated primary completion September 2034 and no posted results. It aims to follow lesions found only by LINFU, but has no randomized action comparison." }, { "id": "clinicaltrials-nct06282809-gannon", "type": "trial-registry", "title": "GANNON: HistoSonics Edison System for treatment of pancreatic adenocarcinoma using histotripsy", "authors": [ "HistoSonics, Inc." ], "publisher": "ClinicalTrials.gov", "date": "2026-07-09", "url": "https://clinicaltrials.gov/study/NCT06282809", "doi": null, "pmid": null, "trialId": "NCT06282809", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Primary registry checked through the API. Industry-sponsored, recruiting, single-arm feasibility study with estimated enrollment 50 and completion August 2028. Its only listed primary outcome is adjudicated procedure-related complications through 30 days. Entry requires a visible acoustic path and at least 1 cm between the planned treatment volume and bowel or stomach." }, { "id": "clinicaltrials-nct06381154-pdt-pembrolizumab", "type": "trial-registry", "title": "Photodynamic priming with verteporfin, pembrolizumab and FOLFIRINOX in pancreatic cancer", "authors": [ "Mayo Clinic", "National Cancer Institute" ], "publisher": "ClinicalTrials.gov", "date": "2026-06-18", "url": "https://clinicaltrials.gov/study/NCT06381154", "doi": null, "pmid": null, "trialId": "NCT06381154", "patentId": null, "grantId": "P01CA084203", "accessed": "2026-09-15", "notes": "Active, no longer recruiting, single-group study with actual enrollment 24. It combines verteporfin light treatment, pembrolizumab, FOLFIRINOX, biopsies and imaging. Overall response is primary; survival and harm are secondary. Main completion is estimated March 2028 and full completion March 2031. The design cannot isolate the contribution of light treatment." }, { "id": "clinicaltrials-nct06390059-panova4", "type": "trial-registry", "title": "PANOVA-4: tumour-treating fields, atezolizumab and chemotherapy in metastatic pancreatic cancer", "authors": [ "NovoCure Ltd." ], "publisher": "ClinicalTrials.gov", "date": "2026-03-31", "url": "https://clinicaltrials.gov/study/NCT06390059", "doi": null, "pmid": null, "trialId": "NCT06390059", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Primary registry. This completed industry-sponsored phase 2 study enrolled 84 actual participants into one arm receiving tumour-treating fields, atezolizumab, gemcitabine and nab-paclitaxel. Disease control for at least 16 weeks was primary. No ClinicalTrials.gov results were posted when checked on 15 September 2026." }, { "id": "clinicaltrials-nct06393400-chase-pdac", "type": "registry", "title": "NCT06393400: CHASe-PDAC healthy-donor stool transfer with gemcitabine and nab-paclitaxel", "authors": [ "London Health Sciences Centre Research Institute" ], "publisher": "ClinicalTrials.gov", "date": "2025-01-23", "url": "https://clinicaltrials.gov/study/NCT06393400", "doi": null, "pmid": null, "trialId": "NCT06393400", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Recruiting phase 1, open-label, single-arm safety study, estimated n=20. Participants swallow 36-40 healthy-donor stool capsules once before standard gemcitabine plus nab-paclitaxel. Estimated completion February 2028; the design cannot estimate FMT contribution to survival." }, { "id": "clinicaltrials-nct06519097-sipp-t3", "type": "trial-registry", "title": "A Randomized, Placebo-Controlled Trial of Tocotrienol in the Prevention of Pancreatic Intraductal Papillary Mucinous Neoplasm Progression", "authors": [ "H. Lee Moffitt Cancer Center and Research Institute", "National Cancer Institute" ], "publisher": "ClinicalTrials.gov", "date": "2025-11-13", "url": "https://clinicaltrials.gov/study/NCT06519097", "doi": null, "pmid": null, "trialId": "NCT06519097", "patentId": null, "grantId": "R01CA263575", "accessed": "2026-09-14", "notes": "Double-masked randomized phase 2 delta-tocotrienol-versus-placebo study. The live registry is active not recruiting with actual enrolment four and no results. The posted protocol planned 212 participants and predicted accrual within two years; actual accrual is 4/212 = 1.9% of that target. The primary composite can count cyst growth above 1 mm as progression, which is not equivalent to HGD or cancer." }, { "id": "clinicaltrials-nct06531408-pigbc", "type": "trial-registry", "title": "Predicting Incidental Gallbladder Cancer (P-iGBC)", "authors": [ "University Hospitals Plymouth NHS Trust" ], "publisher": "ClinicalTrials.gov", "date": "2025-06-03", "url": "https://clinicaltrials.gov/study/NCT06531408", "doi": null, "pmid": null, "trialId": "NCT06531408", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Live registry and protocol checked. Active, not recruiting; current feasibility stage targets up to 500 routine UK cholecystectomies and the full prospective modelling study estimates up to 30,000, with primary completion in September 2029. The study assumes incidental cancer near one in 500 operations and excludes preoperative cancer suspicion, polyps at least 5 mm, PSC and choledochal cysts. It ends at histopathology, has no longitudinal follow-up and aims to select specimens for pathology rather than select people for preventive surgery. The registry primary-outcome wording says low negative predictive value, which is inconsistent with the safety aim and needs clarification." }, { "id": "clinicaltrials-nct06608927-prism1", "type": "trial-registry", "title": "PRISM-1: quemliclustat plus gemcitabine and nab-paclitaxel versus placebo plus the same chemotherapy", "authors": [ "Arcus Biosciences, Inc." ], "publisher": "ClinicalTrials.gov", "date": "2026-05-08", "url": "https://clinicaltrials.gov/study/NCT06608927", "doi": null, "pmid": null, "trialId": "NCT06608927", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Active, not recruiting, quadruple-masked randomized phase 3 in treatment-naive metastatic PDAC. Estimated n=610 across 111 sites. The primary endpoint is all-randomized overall survival through 72 months; PFS, response, duration, disease control and harm are secondary. The registry says individual-level data, protocol, analysis plan and clinical study report may be requested." }, { "id": "clinicaltrials-nct06714604-duration", "type": "trial-registry", "title": "Standard versus prolonged conversion chemotherapy before surgery for borderline and locally advanced pancreatic cancer", "authors": [ "Sahlgrenska University Hospital", "Karolinska University Hospital" ], "publisher": "ClinicalTrials.gov", "date": "2026-09-15", "url": "https://clinicaltrials.gov/study/NCT06714604", "doi": null, "pmid": null, "trialId": "NCT06714604", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Recruiting open-label Swedish randomized trial with estimated enrolment 432. It compares eight versus twelve modified-FOLFIRINOX cycles, or four versus six gemcitabine-nab-paclitaxel cycles, before intended surgery in fit borderline or locally advanced PDAC. Overall survival is primary; resection, complications, quality of life and treatment completion are included. Estimated primary completion is December 2032. It can answer treatment duration, not the causal value of surgery because both arms follow the surgical route." }, { "id": "clinicaltrials-nct06813079-adopt", "type": "trial-registry", "title": "ADOPT: Using Tumor Models to Determine Treatments", "authors": [ "University Health Network, Toronto" ], "publisher": "ClinicalTrials.gov", "date": "2026-06-23", "url": "https://clinicaltrials.gov/study/NCT06813079", "doi": null, "pmid": null, "trialId": "NCT06813079", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Recruiting non-randomized phase 2 at one Canadian site, estimated n=25 and completion February 2028. It assigns one of ten approved drugs from organoid screening in progressive or maintenance states and uses response as primary. It can show delivery and signal, not comparative utility." }, { "id": "clinicaltrials-nct06861088-catiq-p3", "type": "trial-registry", "title": "CATIQ phase 3: Kinisoquin versus placebo for thromboembolic events in advanced pancreatic cancer", "authors": [ "Quercis Pharma AG" ], "publisher": "ClinicalTrials.gov", "date": "2026-02-02", "url": "https://clinicaltrials.gov/study/NCT06861088", "doi": null, "pmid": null, "trialId": "NCT06861088", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Recruiting industry phase 3, estimated n=480 at three listed sites. Two Kinisoquin doses and placebo are assigned for 16 weeks; one dose is selected after 26 adjudicated events. The primary endpoint is first venous or arterial thromboembolic event. Major bleeding, PFS and OS are secondary. It excludes current anticoagulation and high bleeding risk. Primary completion is estimated October 2027; no results are posted." }, { "id": "clinicaltrials-nct06922591-vopimetostat-ras", "type": "trial-registry", "title": "Vopimetostat combinations in MTAP-deleted RAS-mutant pancreatic and lung cancer", "authors": [ "Tango Therapeutics, Inc." ], "publisher": "ClinicalTrials.gov", "date": "2026-05-14", "url": "https://clinicaltrials.gov/study/NCT06922591", "doi": null, "pmid": null, "trialId": "NCT06922591", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Recruiting open-label nonrandomized sequential phase 1/2, estimated n=183, completion December 2027. Eight listed arms test vopimetostat with daraxonrasib, zoldonrasib, mFOLFIRINOX or gemcitabine/nab-paclitaxel. No registry results and no single-agent control inside this protocol." }, { "id": "clinicaltrials-nct06989437-ponsegromab-pdac", "type": "trial-registry", "title": "Ponsegromab or placebo with first-line chemotherapy in metastatic PDAC with cachexia", "authors": [ "Pfizer" ], "publisher": "ClinicalTrials.gov", "date": "2026-08-24", "url": "https://clinicaltrials.gov/study/NCT06989437", "doi": null, "pmid": null, "trialId": "NCT06989437", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Recruiting industry phase 2b/3, estimated n=982 at 220 locations in 21 countries. Two ponsegromab doses and placebo are added to first-line chemotherapy. Week-12 weight change and a five-item anorexia score are primary; OS, PFS, response, function and chemotherapy dose changes are secondary. Eligibility requires ECOG 0-1, life expectancy at least four months and completion of an initial chemotherapy cycle, excluding some of the frailest patients. Estimated primary completion 2028-01-17 and study completion 2029-12-10; no results posted." }, { "id": "clinicaltrials-nct07028424-panoramix", "type": "registry", "title": "NCT07028424: PANORAMIX NALIRIFOX maintenance and ciprofloxacin study", "authors": [ "GERCOR - Multidisciplinary Oncology Cooperative Group" ], "publisher": "ClinicalTrials.gov", "date": "2025-07-01", "url": "https://clinicaltrials.gov/study/NCT07028424", "doi": null, "pmid": null, "trialId": "NCT07028424", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Stale US registry record says not yet recruiting and estimates 206 across the two-step phase 2 design, conflicting with the current CTIS record of 142 planned and active recruitment. Step 2 is quadruple-masked 1:1 ciprofloxacin versus placebo, but the ClinicalTrials.gov outcome list records only Step 1 endpoints. The protocol denominator and full endpoint list need reconciliation before analysis." }, { "id": "clinicaltrials-nct07033689-sunrise2", "type": "trial-registry", "title": "SUNRISE-II feasibility study of Suizenji focused ultrasound in unresectable pancreatic ductal adenocarcinoma", "authors": [ "SONIRE Therapeutics Inc.", "Stanford University" ], "publisher": "ClinicalTrials.gov", "date": "2026-04-03", "url": "https://clinicaltrials.gov/study/NCT07033689", "doi": null, "pmid": null, "trialId": "NCT07033689", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Recruiting industry-sponsored single-group U.S. feasibility study targeting ten untreated people with unresectable pancreatic ductal adenocarcinoma. More than 50% treated-volume ablation at three months is primary. It lists several standard chemotherapy options but has no control. The first participant was treated at Stanford in March 2026. This tests thermal Suizenji HIFU without K-912." }, { "id": "clinicaltrials-nct07066098-ibi343-phase3", "type": "trial-registry", "title": "NCT07066098: G-HOPE-002 IBI343 versus placebo in previously treated CLDN18.2-positive pancreatic cancer", "authors": [ "Innovent Biologics (Suzhou) Co. Ltd." ], "publisher": "ClinicalTrials.gov", "date": "2025-08-15", "url": "https://clinicaltrials.gov/study/NCT07066098", "doi": null, "pmid": null, "trialId": "NCT07066098", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Primary current registry. Recruiting randomized quadruple-masked phase 3, planned n=201 in a 2:1 IBI343-plus-supportive-care versus placebo-plus-supportive-care comparison after at least two systemic regimens. OS is primary, completion is estimated in June 2028, one recruiting Chinese site is listed, no result is posted, the CLDN18.2 cutoff is not stated and the eligibility text says ECOG 0 or 2 rather than 0 through 2." }, { "id": "clinicaltrials-nct07196020-pinpoint-vte", "type": "trial-registry", "title": "PINPOINT: serial haemostasis biomarkers for VTE prediction during chemotherapy", "authors": [ "Norris, Lucy", "University of Dublin, Trinity College" ], "publisher": "ClinicalTrials.gov", "date": "2026-05-15", "url": "https://clinicaltrials.gov/study/NCT07196020", "doi": null, "pmid": null, "trialId": "NCT07196020", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Recruiting observational study, estimated n=380 across three Irish sites, including ovarian, lung, gastric and pancreatic cancer. It measures serial markers and VTE through 12 weeks after the final sample. It does not assign prophylaxis or test clinical utility; completion is estimated September 2029." }, { "id": "clinicaltrials-nct07229768-ctr-fapi", "type": "trial-registry", "title": "177Lu-CTR-FAPI in advanced metastatic solid tumours", "authors": [ "Yan Xing" ], "publisher": "ClinicalTrials.gov", "date": "2025-11-17", "url": "https://clinicaltrials.gov/study/NCT07229768", "doi": null, "pmid": null, "trialId": "NCT07229768", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Recruiting single-centre, single-arm early phase 1 with estimated enrolment four. It gives 7.4 GBq every 6 plus or minus 2 weeks for up to four doses to FAPI-positive advanced cancers, with pancreatic cancer listed as a preferred tumour. Estimated completion is December 2026. Four people can inform immediate delivery and toxicity, not efficacy." }, { "id": "clinicaltrials-nct07306286-cryo-immunotherapy", "type": "trial-registry", "title": "Cryoablation combined with PD-1/CTLA-4 antibody after first-line treatment failure in locally advanced pancreatic cancer", "authors": [ "Ruijin Hospital" ], "publisher": "ClinicalTrials.gov", "date": "2025-12-29", "url": "https://clinicaltrials.gov/study/NCT07306286", "doi": null, "pmid": null, "trialId": "NCT07306286", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Primary registry checked through the API. Recruiting open-label single-arm phase 2 study with estimated enrollment 20 and completion August 2027. It combines cryoablation with a dual PD-1/CTLA-4 antibody and makes overall survival primary, so it cannot isolate either component from selection or subsequent care." }, { "id": "clinicaltrials-nct07325214-sonopan3", "type": "trial-registry", "title": "SonoPANIII: chemotherapy with or without non-thermal focused ultrasound in borderline resectable or locally advanced pancreatic cancer", "authors": [ "IMGT Co., Ltd." ], "publisher": "ClinicalTrials.gov", "date": "2026-07-13", "url": "https://clinicaltrials.gov/study/NCT07325214", "doi": null, "pmid": null, "trialId": "NCT07325214", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Primary registry checked through the API. The planned 360-person randomized pivotal trial uses a non-thermal, non-tissue-destructive focused-ultrasound device with identical chemotherapy and overall survival as primary endpoint. It is not yet recruiting and estimates completion in January 2030. This is a drug-delivery route, not thermal HIFU ablation." }, { "id": "clinicaltrials-nct07433283-cnsi-fe-expansion", "type": "trial-registry", "title": "Repeated intratumoral carbon nanoparticle-loaded iron in advanced solid tumours", "authors": [ "Sichuan Enray Pharmaceutical Sciences Company" ], "publisher": "ClinicalTrials.gov", "date": "2026-02-25", "url": "https://clinicaltrials.gov/study/NCT07433283", "doi": null, "pmid": null, "trialId": "NCT07433283", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Recruiting company-sponsored single-group phase Ib/IIa study targeting 54 advanced solid-tumour participants, including but not reserving a cohort for pancreatic cancer. Directly injectable lesions, ECOG 0-1 fitness and at least three months expected survival are required. Twelve-week adverse events are the sole listed primary endpoint. Estimated completion is December 2026; no results are posted." }, { "id": "clinicaltrials-nct07470268-hypanc", "type": "trial-registry", "title": "HY-PANC hypertrophy-oriented resistance training in unresectable pancreatic cancer", "authors": [ "Hospital Universitario de Fuenlabrada" ], "publisher": "ClinicalTrials.gov", "date": "2026-09-15", "url": "https://clinicaltrials.gov/study/NCT07470268", "doi": null, "pmid": "42461827", "trialId": "NCT07470268", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Not-yet-recruiting two-arm study estimates ten unresectable pancreatic patients, with September 2026 start and September 2027 completion. The main outcome is body composition after 12 weeks. At n=10 it can test route feasibility and measurement, not survival, treatment completion or a stable subgroup effect." }, { "id": "clinicaltrials-nct07591571-pdac-community-worker", "type": "trial-registry", "title": "Digitally enhanced community-health-worker intervention in newly diagnosed pancreatic ductal adenocarcinoma", "authors": [ "Loma Linda University" ], "publisher": "ClinicalTrials.gov", "date": "2026-05-15", "url": "https://clinicaltrials.gov/study/NCT07591571", "doi": null, "pmid": null, "trialId": "NCT07591571", "patentId": null, "grantId": "PA230343", "accessed": "2026-09-15", "notes": "Recruiting pilot randomized trial of 60 newly diagnosed PDAC patients and caregivers, testing community-health-worker navigation plus digital contact against attention control. The registry still said recruiting after its estimated 1 September 2026 completion date, with no results. The primary endpoint is a 12-week hepatobiliary symptom score." }, { "id": "clinicaltrials-nct07671339-eli002-tislelizumab", "type": "trial-registry", "title": "ELI-002 7P with or without tislelizumab before PDAC surgery", "authors": [ "Memorial Sloan Kettering Cancer Center" ], "publisher": "ClinicalTrials.gov", "date": "2026-06-26", "url": "https://clinicaltrials.gov/study/NCT07671339", "doi": null, "pmid": null, "trialId": "NCT07671339", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Recruiting randomized open-label pilot at seven sites, estimated n=20. Both groups receive mFOLFIRINOX and ELI-002 7P; one also receives tislelizumab. The primary outcome is grade 3 or worse drug-related adverse events through two years, so the study can test feasibility and checkpoint-added safety but is not powered to establish survival benefit." }, { "id": "clinicaltrials-nct07689955-celiac-radiotherapy", "type": "trial-registry", "title": "Radiotherapy to the coeliac plexus for pain in advanced pancreatic adenocarcinoma", "authors": [ "Royal Marsden NHS Foundation Trust" ], "publisher": "ClinicalTrials.gov", "date": "2026-07-08", "url": "https://clinicaltrials.gov/study/NCT07689955", "doi": null, "pmid": null, "trialId": "NCT07689955", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Not-yet-recruiting UK single-arm phase II study targeting 52 people with inoperable advanced PDAC and opioid-resistant pain. The main endpoint combines at least 30% pain reduction with stable or lower opioid dose. Systemic treatment must stop for at least seven days before and after radiation. Estimated start September 2026 and completion September 2031. It adds no randomized control after the earlier 125-person single-arm radiosurgery study." }, { "id": "clinicaltrials-nct07698613-asp1929-pancreas", "type": "trial-registry", "title": "EGFR-targeted ASP-1929 photoimmunotherapy for locally advanced pancreatic cancer", "authors": [ "Brown University", "Rakuten Medical", "Rhode Island Hospital", "The Miriam Hospital" ], "publisher": "ClinicalTrials.gov", "date": "2026-07-13", "url": "https://clinicaltrials.gov/study/NCT07698613", "doi": null, "pmid": null, "trialId": "NCT07698613", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Not-yet-recruiting single-group study with target enrollment 30. All receive ASP-1929 before surgical exploration; people found resectable undergo surgery without illumination, while people found inoperable receive five minutes of 690 nm light. The first ten form a safety run-in; six-month progression-free survival is primary for the expansion. Start was estimated September 2026, main completion March 2029 and full completion September 2033." }, { "id": "cls-2026-investment-case", "type": "company-report", "title": "Clinical Laserthermia Systems investment case", "authors": [ "Clinical Laserthermia Systems AB" ], "publisher": "Clinical Laserthermia Systems AB", "date": "2026-09-15", "url": "https://clinicallaser.se/investor-relations/", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Current company page describes the TRANBERG thermotherapy platform and a business model with system sales and repeat disposable revenue above 60% gross margin. It says the company is focused first on neurosurgery and then adjacent areas. Pancreatic cancer is not identified as a current commercial focus." }, { "id": "cnio-experimental-oncology-2026", "type": "institution-profile", "title": "Experimental Oncology Group", "authors": [ "Spanish National Cancer Research Centre" ], "publisher": "Centro Nacional de Investigaciones Oncológicas", "date": null, "url": "https://www.cnio.es/en/research-innovation/scientific-programmes/tumour-biology-programme/experimental-oncology-group/", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Live institutional group page checked for current leader, team, KRAS focus, pancreatic tumour models and recent publications. The page labels the original 2025 version of the triple-combination paper as retracted and lists the June 2026 replacement." }, { "id": "cnio-malats-profile-2026", "type": "institution-profile", "title": "Genetic and Molecular Epidemiology Group", "authors": [ "Spanish National Cancer Research Centre" ], "publisher": "Spanish National Cancer Research Centre", "date": null, "url": "https://www.cnio.es/en/research-innovation/scientific-programmes/cancer-genomics-programme/genetic-molecular-epidemiology-group/", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Official profile lists Núria Malats as group leader and states that the group focuses on pancreatic and bladder cancer, joining risk epidemiology with multi-source clinical and biological data for clinical and public-health use." }, { "id": "cohen-2018-cancerseek", "type": "paper", "title": "Detection and localization of surgically resectable cancers with a multi-analyte blood test", "authors": [ "Cohen, Joshua D.", "Li, Lu", "Wang, Yuxuan", "et al." ], "publisher": "Science", "date": "2018-01-18", "url": "https://pubmed.ncbi.nlm.nih.gov/29348365/", "doi": "10.1126/science.aar3247", "pmid": "29348365", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Case-control study of 1,005 people with known nonmetastatic cancer and 812 healthy controls. It measured discrimination after clinical diagnosis, not performance in an asymptomatic screening population. Several authors had patent and company interests in the technology." }, { "id": "conroy-2011-folfirinox-metastatic", "type": "paper", "title": "FOLFIRINOX versus Gemcitabine for Metastatic Pancreatic Cancer", "authors": [ "Conroy, Thierry", "Desseigne, Françoise", "Ychou, Marc", "et al." ], "publisher": "New England Journal of Medicine", "date": "2011-05-12", "url": "https://pubmed.ncbi.nlm.nih.gov/21561347/", "doi": "10.1056/NEJMoa1011923", "pmid": "21561347", "trialId": "NCT00112658", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Randomised phase 3 trial in 342 patients with metastatic pancreatic adenocarcinoma and good performance status. Median overall survival was 11.1 versus 6.8 months; toxicity was greater with FOLFIRINOX." }, { "id": "conroy-2022-prodige24-five-year", "type": "paper", "title": "Five-Year Outcomes of FOLFIRINOX vs Gemcitabine as Adjuvant Therapy for Pancreatic Cancer: A Randomized Clinical Trial", "authors": [ "Conroy, Thierry", "Castan, Fabienne", "Lopez, Anthony", "et al." ], "publisher": "JAMA Oncology", "date": "2022-09-08", "url": "https://jamanetwork.com/journals/jamaoncology/fullarticle/2795978", "doi": "10.1001/jamaoncol.2022.3829", "pmid": null, "trialId": "NCT01526135", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Mature results from the 493-patient PRODIGE 24/CCTG PA6 randomised trial after resection. Selected patients had adequate recovery and fitness for adjuvant chemotherapy." }, { "id": "cordis-h2020-target-cancer-title-query-2026", "type": "government-data", "title": "CORDIS Horizon 2020 project file: target-cancer title query", "authors": [ "Publications Office of the European Union" ], "publisher": "European Commission CORDIS", "date": "2026-08-06", "url": "https://data.europa.eu/data/datasets/cordish2020projects?locale=en", "doi": "10.2906/112117098108/12", "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Primary EU open-data project file. Reproducible calculation used project.csv from the 6 August 2026 cordis-h2020projects-csv.zip snapshot. Retain unique projects with startDate from 2015-01-01 through 2026-09-14 and a title containing one of these rules, case-insensitive: pancreas or pancreatic within 20 characters before cancer, carcinoma, adenocarcinoma, tumor or tumour; pancreatic neuroendocrine; cholangiocarcinoma; biliary tract plus cancer, carcinoma, tumor or tumour; gallbladder plus cancer, carcinoma, tumor or tumour; hepatocellular carcinoma; liver plus cancer, carcinoma, tumor or tumour; colorectal plus cancer, carcinoma, tumor or tumour; or colorectal and liver metastasis within 30 characters in either order. Sum ecMaxContribution after decimal-comma conversion. Ninety-four unique projects matched and totalled EUR 125,968,482.13. This is a conservative title-visible commitment ledger, not all relevant EU funding and not expenditure. It includes full project values when the title names more than one cancer." }, { "id": "cordis-horizon-target-cancer-title-query-2026", "type": "government-data", "title": "CORDIS Horizon Europe project file: target-cancer title query", "authors": [ "Publications Office of the European Union" ], "publisher": "European Commission CORDIS", "date": "2026-08-06", "url": "https://data.europa.eu/data/datasets/cordis-eu-research-projects-under-horizon-europe-2021-2027?locale=en", "doi": "10.2906/112117098108/20", "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Primary EU open-data project file. Reproducible calculation used project.csv from the 6 August 2026 cordis-HORIZONprojects-csv.zip snapshot. Apply the same date and title rule recorded in cordis-h2020-target-cancer-title-query-2026, then sum ecMaxContribution after decimal-comma conversion. Sixty-five unique projects matched and totalled EUR 186,751,466.36. Combined with the non-overlapping Horizon 2020 result, the strict title-visible ledger contains 159 projects and EUR 312,719,948.49. This is a lower-bound commitment ledger, not expenditure, and full values of multi-cancer projects remain included." }, { "id": "cordis-pancaid-101096309", "type": "grant-record", "title": "PANcreatic CAncer Initial Detection via liquid biopsy (PANCAID)", "authors": [ "European Commission" ], "publisher": "CORDIS", "date": "2023-01-01", "url": "https://cordis.europa.eu/project/id/101096309", "doi": "10.3030/101096309", "pmid": null, "trialId": null, "patentId": null, "grantId": "101096309", "accessed": "2026-09-14", "notes": "EU project from 2023-2027 with total cost €9,845,090.26 and EU contribution €9,838,840. It plans a multi-marker liquid-biopsy panel using biobanked samples and AI, followed by the design of a future prospective international validation study. It has no reported clinical-utility result." }, { "id": "corty-2020-antibiotic-gemcitabine-toxicity", "type": "paper", "title": "Antibacterial use is associated with an increased risk of hematologic and gastrointestinal adverse events in patients treated with gemcitabine for stage IV pancreatic cancer", "authors": [ "Corty, Robert W.", "Langworthy, Benjamin W.", "Fine, Jason P.", "et al." ], "publisher": "The Oncologist", "date": "2020-07-01", "url": "https://pubmed.ncbi.nlm.nih.gov/32181968/", "doi": "10.1634/theoncologist.2019-0570", "pmid": "32181968", "trialId": "NCT01442974", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Time-varying reanalysis of 430 gemcitabine-only participants from MPACT. Antibiotic exposure was associated with grade 3+ adverse events HR 1.77, blood-count events HR 1.64 and gastrointestinal events HR 2.14. This was not randomized antibiotic use and cannot separate antibiotic toxicity, infection and increased active gemcitabine exposure." }, { "id": "cosomil-2026-series-a1", "type": "company-announcement", "title": "Cosomil secures Series A1 financing for blood-based cancer screening", "authors": [ "Cosomil, Inc." ], "publisher": "Cosomil", "date": "2026-02-06", "url": "https://cosomil.com/en/260206e/", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Company announcement; amount undisclosed. Led by Delight Ventures with Animal Spirits, Gogin Capital, ANRI and Green Core." }, { "id": "costa-silva-2015-liver-niche", "type": "paper", "title": "Pancreatic cancer exosomes initiate pre-metastatic niche formation in the liver", "authors": [ "Costa-Silva, Bruno", "Aiello, Nils M.", "Ocean, Allyson J.", "et al." ], "publisher": "Nature Cell Biology", "date": "2015-05-18", "url": "https://pubmed.ncbi.nlm.nih.gov/25985394/", "doi": "10.1038/ncb3169", "pmid": "25985394", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Mouse transfer and blockade experiments proposed an exosomal MIF to Kupffer-cell to fibronectin pathway. A small patient comparison linked exosomal MIF with later liver metastasis; independent prospective human validation is required." }, { "id": "costamagna-2026-sems-early-surgery-rct", "type": "paper", "title": "Preoperative biliary drainage with metal stent versus early surgery in patients with pancreatic cancer: a randomized clinical trial", "authors": [ "Costamagna, Guido", "Reddy, D Nageshwar", "Chia, Nam-Hung", "et al." ], "publisher": "Annals of Surgical Oncology", "date": "2026-04-08", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC13242453/", "doi": "10.1245/s10434-026-19546-9", "pmid": "41951900", "trialId": "NCT01774019", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Open-label noninferiority trial at 11 expert centres in nine countries, run 2013-2021 and reported in 2026. It randomized 284 resectable pancreatic or periampullary cancers with bilirubin at least 5.8 mg/dL to ERCP metal-stent drainage or surgery within four weeks. Serious adverse events occurred in 40/138, 29.0%, versus 36/136, 26.5%; the 2.5-point difference had a one-sided upper 95% limit of 11.7%, within a wide 15-point margin. Pre-surgery serious events were 15.2% versus 5.1%. The sponsor, Boston Scientific, supplied the device and participated in design, execution, data systems, monitoring, analysis, interpretation and writing. Most patients did not require drainage; centres had high ERCP expertise; only 120 had histologic type recorded." }, { "id": "coupland-2016-england-resection-volume", "type": "paper", "title": "Resection rate, hospital procedure volume and survival in pancreatic cancer patients in England: population-based study, 2005-2009", "authors": [ "Coupland, Victoria H", "Konfortion, Julie", "Jack, Ruth H", "et al." ], "publisher": "European Journal of Surgical Oncology", "date": "2015-11-17", "url": "https://europepmc.org/article/MED/26705143", "doi": "10.1016/j.ejso.2015.11.003", "pmid": "26705143", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Linked English registry and hospital study of 31,973 pancreatic-cancer patients, 2,580 of whom had surgery. Higher population resection-rate areas had lower adjusted mortality, HR 0.82 for highest versus lowest quintile. Among operated patients, the hospital-volume association was imprecise and not statistically clear: HR 0.88 for at least 30 versus fewer than 15 operations per year, P for trend 0.177. Observational association does not prove that raising operation rates alone improves survival." }, { "id": "coveler-2024-oleclumab-pdac", "type": "paper", "title": "A Phase Ib/II Randomized Clinical Trial of Oleclumab with or without Durvalumab plus Chemotherapy in Patients with Metastatic Pancreatic Ductal Adenocarcinoma", "authors": [ "Coveler, Andrew L.", "Reilley, Matthew J.", "Zalupski, Mark", "et al." ], "publisher": "Clinical Cancer Research", "date": "2024-08-06", "url": "https://pubmed.ncbi.nlm.nih.gov/39106081/", "doi": "10.1158/1078-0432.CCR-24-0499", "pmid": "39106081", "trialId": "NCT03611556", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Randomized dose expansion n=170: chemotherapy n=62, oleclumab plus chemotherapy n=38, and oleclumab plus durvalumab plus chemotherapy n=70. The response endpoint was missed. Oleclumab alone had OS HR 1.26, 95% CI 0.79-1.98; with durvalumab HR 0.75, 95% CI 0.50-1.13. A favourable CD73-high subgroup was exploratory and its control underperformed. AstraZeneca employees were authors." }, { "id": "cptac-pdac-pdc000272", "type": "dataset", "title": "CPTAC pancreatic ductal adenocarcinoma discovery study — intact glycoproteome", "authors": [ "National Cancer Institute Clinical Proteomic Tumor Analysis Consortium" ], "publisher": "NCI Proteomic Data Commons", "date": "2021-01-01", "url": "https://pdc.cancer.gov/pdc/study/PDC000272", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": "phs001287", "accessed": "2026-09-14", "notes": "166 cases and 250 aliquots in this glycoproteome study; linked genomic and imaging data. The underlying programme includes 140 cancers, 67 adjacent tissues and nine normal pancreatic duct tissues." }, { "id": "cristiano-2019-fragmentomics", "type": "paper", "title": "Genome-wide cell-free DNA fragmentation in patients with cancer", "authors": [ "Cristiano, Stephen", "Leal, Adelaida", "Phallen, Jillian", "et al." ], "publisher": "Nature", "date": "2019-05-29", "url": "https://pubmed.ncbi.nlm.nih.gov/31142840/", "doi": "10.1038/s41586-019-1272-6", "pmid": "31142840", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Proof-of-principle case-control study of 236 patients across seven cancers and 245 healthy people. Several authors were inventors or founders connected to DELFI Diagnostics." }, { "id": "cruk-scotland-morton-lab-2026", "type": "institution-profile", "title": "Jen Morton — Preclinical Pancreatic Cancer", "authors": [ "Cancer Research UK Scotland Institute" ], "publisher": "Cancer Research UK Scotland Institute", "date": null, "url": "https://www.crukscotlandinstitute.ac.uk/index.php?Itemid=482&id=1352&option=com_content&view=article", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Live institutional laboratory page checked for current group membership and work on pancreatic models, fibroblasts, early detection, homologous-recombination deficiency, radiotherapy and the tumour environment." }, { "id": "cruk-therapeutic-catalyst-2026", "type": "funder-guidance", "title": "Cancer Research UK Therapeutic Catalyst", "authors": [ "Cancer Research UK", "Cancer Research Horizons" ], "publisher": "Cancer Research UK", "date": "2026-09-14", "url": "https://www.cancerresearchuk.org/for-researchers/apply-for-and-manage-your-funding/our-funding-schemes/therapeutic-catalyst", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Primary scheme page. Remaining 2026 submission deadline is 25 September. Awards provide up to £250,000 or £350,000 over 12-24 months and join an academic laboratory to Cancer Research Horizons drug-discovery staff for antibody or small-molecule work. Fits a bounded translation package, not a clinical platform." }, { "id": "cshl-tuveson-duke-transition-2026", "type": "institution", "title": "Tuveson named executive director of Duke Cancer Institute", "authors": [ "Cold Spring Harbor Laboratory" ], "publisher": "Cold Spring Harbor Laboratory", "date": "2026-09-08", "url": "https://www.cshl.edu/tuveson-named-executive-director-of-duke-cancer-institute/", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Official announcement. Appointment at Duke is future-dated to 31 October 2026; the atlas therefore retains CSHL as his current institution on 14 September 2026." }, { "id": "cshl-tuveson-profile-2026", "type": "institution", "title": "David Tuveson — Cold Spring Harbor Laboratory faculty profile and laboratory", "authors": [ "Cold Spring Harbor Laboratory" ], "publisher": "Cold Spring Harbor Laboratory", "date": null, "url": "https://facultyprofiles.cshl.edu/david.tuveson", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Live institutional profile checked for current titles, research fields and laboratory links. A separate CSHL announcement says he will move to Duke effective 31 October 2026." }, { "id": "ctis-2024-514714-12-00-aspire", "type": "trial-registry", "title": "European ASPIRE record: ivospemin plus chemotherapy versus placebo plus chemotherapy", "authors": [ "Panbela Therapeutics, Inc." ], "publisher": "Clinical Trials Information System", "date": "2025-05-02", "url": "https://euclinicaltrials.eu/search-for-clinical-trials/?lang=en&EUCT=2024-514714-12-00", "doi": null, "pmid": null, "trialId": "2024-514714-12-00", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "The European record for the same protocol labels the trial ended on 28 March 2025. It lists 547 planned participants: 407 across six EU countries and 140 elsewhere. That conflicts with the U.S. registry's recruiting label and estimated n=600; neither record supplies the actual randomized denominator or results." }, { "id": "ctis-2026-sil204-phase23", "type": "trial-registry", "title": "SIL204-SL Treatment for Locally Advanced Pancreatic Cancer", "authors": [ "Silexion Therapeutics Ltd." ], "publisher": "Clinical Trials Information System", "date": "2026-06-23", "url": "https://ctis.eu/trial/2025-525157-39-00", "doi": null, "pmid": null, "trialId": "2025-525157-39-00", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Authorised integrated phase 2/3 protocol SIL204SL-P001 with 403 planned participants. German recruitment remains pending. Segment 1 measures safety and plasma pharmacokinetics; later segments compare combined intratumour and subcutaneous SIL204 plus chemotherapy with chemotherapy. Overall survival is primary in the G12D/V subset. The registry says individual participant data will not be shared." }, { "id": "ctis-2026-virage2", "type": "trial-registry", "title": "VIRAGE2 increased-frequency VCN-01 with gemcitabine and nab-paclitaxel", "authors": [ "Theriva Biologics, S.L." ], "publisher": "Clinical Trials Information System", "date": "2026-07-20", "url": "https://www.ctis.eu/trial/2026-525566-21-00", "doi": null, "pmid": null, "trialId": "2026-525566-21-00", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Recruiting single-centre, single-arm phase 2a in Spain with six planned participants. It tests three VCN-01 doses 56 days apart with gemcitabine/nab-paclitaxel. Safety and blood viral genomes are main objectives. The record says participant-level data will not be shared. It can test repeated-dose feasibility, not comparative survival." }, { "id": "ctis-panoramix-2024-517766-41", "type": "registry", "title": "2024-517766-41-00: PANORAMIX randomized ciprofloxacin test in second-line metastatic PDAC", "authors": [ "Association GERCOR" ], "publisher": "Clinical Trials Information System", "date": "2026-02-17", "url": "https://ctis.eu/trial/2024-517766-41-00", "doi": null, "pmid": null, "trialId": "2024-517766-41-00", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Current EU record says ongoing and recruiting from 17 February 2026 at 27 French sites, 142 planned across two steps. Step 2 randomizes ciprofloxacin versus placebo with gemcitabine-based second-line treatment. Its primary endpoint is six-month survival in the ciprofloxacin arm; secondary outcomes include OS, PFS and response in both arms, resistant gut bacteria and gemcitabine/dFdU pharmacokinetics." }, { "id": "cui-2026-hrs4642-pdac", "type": "paper", "title": "KRAS-G12D inhibitor HRS-4642 plus chemotherapy in advanced KRASG12D-mutant pancreatic cancer: a phase 1b/2 trial", "authors": [ "Cui, Jiujie", "Jiang, Kuirong", "Li, Wei", "et al." ], "publisher": "Nature Medicine", "date": "2026-07-09", "url": "https://pubmed.ncbi.nlm.nih.gov/42426224/", "doi": "10.1038/s41591-026-04538-9", "pmid": "42426224", "trialId": "NCT06520488", "patentId": null, "grantId": "82541003; 82272906", "accessed": "2026-09-14", "notes": "Chinese multicentre company-linked phase 1b/2 trial; 31 treated, including 30 treatment-naive phase 2 patients." }, { "id": "cuneo-2019-adavosertib-pdac", "type": "paper", "title": "Dose escalation trial of the WEE1 inhibitor adavosertib with gemcitabine and radiation for locally advanced pancreatic cancer", "authors": [ "Cuneo, Kyle C.", "Morgan, Meredith A.", "Sahai, Vaibhav", "et al." ], "publisher": "Journal of Clinical Oncology", "date": "2019-10-10", "url": "https://pubmed.ncbi.nlm.nih.gov/31398082/", "doi": "10.1200/JCO.19.00730", "pmid": "31398082", "trialId": "NCT02037230", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Primary uncontrolled dose-escalation study, n=34. Median OS was 21.7 months and PFS 9.4 months; eight people had a dose-limiting toxicity. Reduced phospho-CDK1 in hair follicles showed WEE1 inhibition in surrogate tissue, not pancreatic tumour. Historical comparison cannot separate drug, radiation, selection or later care." }, { "id": "da-fonseca-2024-latam-cca", "type": "paper", "title": "Cholangiocarcinoma in Latin America: a multicentre observational study alerts on ethnic disparities in tumour presentation and outcomes", "authors": [ "da Fonseca, Leandro G.", "Izquierdo-Sanchez, Laura", "Hashizume, Pedro H.", "et al." ], "publisher": "The Lancet Regional Health — Americas", "date": "2024-12-01", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC11626722/", "doi": "10.1016/j.lana.2024.100952", "pmid": "39655285", "trialId": null, "patentId": null, "grantId": "Horizon 2020 grant 825510; EASL registry grants; Incyte awards 2020 and 2023", "accessed": "2026-09-14", "notes": "Retrospective LATAM-CCA Registry cohort of 309 eligible patients from five referral centres in Argentina, Brazil, Chile, Ecuador and Peru, 2010-2023. Of 276 with treatment data, 29.0% had surgery, 39.9% chemotherapy and 31.2% best supportive care. Median post-treatment survival was 33.7 months after resection, 8.3 months after chemotherapy and 2.3 months with supportive care, but treatment selection prevents causal comparison. The cohort excludes patients with missing mandatory data, is not population based, lacks central review, and cannot represent remote or undiagnosed cases. Incyte helped fund the registry and one senior author disclosed company grants, speaking fees and consulting relationships." }, { "id": "daamen-2024-radar-panc-protocol", "type": "paper", "title": "Recurrent disease detection after resection of pancreatic ductal adenocarcinoma using a recurrence-focused surveillance strategy (RADAR-PANC): protocol of an international randomized controlled trial according to the Trials within Cohorts design", "authors": [ "Daamen, Lois A", "van Goor, Iris W J M", "Groot, Vincent P", "et al." ], "publisher": "Trials", "date": "2024-06-20", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC11188210/", "doi": "10.1186/s13063-024-08223-5", "pmid": "38902836", "trialId": "NCT04875325", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Protocol for 306 randomized Dutch and UK patients after R0-R1 PDAC resection. The intervention is clinical review, CA19-9 and contrast CT chest/abdomen every three months for two years; control is non-standardized, mainly symptom-led follow-up. Primary endpoint is overall survival; quality of life, worry, treatment, compliance and cost are secondary. The power model assumes median survival 29.4 versus 19.7 months after allowing for 20% intervention refusal and 5% control imaging, a very large effect derived from observational comparisons. The protocol reported 219 enrolled by January 2024 and had expected April 2024 recruitment completion." }, { "id": "daher-2019-xct-pdac-ferroptosis", "type": "paper", "title": "Genetic Ablation of the Cystine Transporter xCT in PDAC Cells Inhibits mTORC1, Growth, Survival, and Tumor Formation via Nutrient and Oxidative Stresses", "authors": [ "Daher, Boutaina", "Parks, Scott K", "Durivault, Jerome", "et al." ], "publisher": "Cancer Research", "date": "2019-06-07", "url": "https://doi.org/10.1158/0008-5472.CAN-18-3855", "doi": "10.1158/0008-5472.CAN-18-3855", "pmid": "31175120", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Removing the SLC7A11/xCT cystine transporter from two human PDAC cell lines depleted glutathione, caused nutrient and oxidative stress and ferroptotic features, and delayed but did not eliminate xenograft growth. Other cysteine-supply routes remained in living animals. Low-dose erastin reproduced some cell effects and increased gemcitabine and cisplatin killing in culture; there was no human treatment test." }, { "id": "dai-2020-ferroptotic-damage-pdac", "type": "paper", "title": "Ferroptotic damage promotes pancreatic tumorigenesis through a TMEM173/STING-dependent DNA sensor pathway", "authors": [ "Dai, Enyong", "Han, Leng", "Liu, Jiao", "et al." ], "publisher": "Nature Communications", "date": "2020-12-11", "url": "https://doi.org/10.1038/s41467-020-20154-8", "doi": "10.1038/s41467-020-20154-8", "pmid": "33311482", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "In pancreatitis and KRAS-driven mouse models, a high-iron diet or pancreas-wide GPX4 loss increased oxidized-DNA release, activated STING and recruited macrophages, accelerating pancreatic tumour formation. Ferroptosis inhibition, macrophage depletion or blocking the oxidized-DNA-STING route reduced tumour formation. This warns that cell type, timing and immune context can reverse the intended effect; it is not evidence that blocking ferroptosis treats established human PDAC." }, { "id": "danafarber-aguirre-profile-2026", "type": "institution", "title": "Andrew Aguirre, MD, PhD", "authors": [ "Dana-Farber Cancer Institute" ], "publisher": "Dana-Farber Cancer Institute", "date": null, "url": "https://labs.dana-farber.org/aguirrelab/people/andrew-aguirre-md-phd", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Live institutional laboratory page checked for current appointments, pancreatic biology work, RAS focus and linked financial-disclosure page." }, { "id": "datta-2017-heatpac-protocol", "type": "paper", "title": "HEATPAC: a phase II randomized study of concurrent thermochemoradiotherapy versus chemoradiotherapy alone in locally advanced pancreatic cancer", "authors": [ "Datta, Niloy Ranjan", "Pestalozzi, Bernhard", "Clavien, Pierre-Alain", "et al." ], "publisher": "Radiation Oncology", "date": "2017-11-21", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC5698973/", "doi": "10.1186/s13014-017-0923-8", "pmid": "29162142", "trialId": "NCT02439593", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Open phase 2 protocol in locally advanced pancreatic cancer. After four FOLFIRINOX cycles, it planned to randomize 86 participants to identical gemcitabine chemoradiotherapy with or without six weekly sessions of regional hyperthermia, aiming for 40-43 degrees C for 60 minutes with intraduodenal temperature measurement." }, { "id": "davidsen-2026-chronic-pancreatitis-mortality", "type": "paper", "title": "Cause-specific mortality in patients with chronic pancreatitis: a nationwide population-based cohort study", "authors": [ "Davidsen, Line", "Bruun, Niels Henrik", "Cook, Mathias E.", "et al." ], "publisher": "European Journal of Internal Medicine", "date": "2026-08", "url": "https://pubmed.ncbi.nlm.nih.gov/41966906/", "doi": "10.1016/j.ejim.2026.106874", "pmid": "41966906", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Danish national-registry cohort of 14,565 adults with incident chronic pancreatitis, median follow-up 10.2 years. Five-year pancreatic-cancer incidence was 3.4% without a lag but 0.5% after excluding the first two years, showing that early occult cancer or misclassification can inflate apparent future risk." }, { "id": "dbouk-2022-caps5", "type": "paper", "title": "The Multicenter Cancer of Pancreas Screening Study: Impact on Stage and Survival", "authors": [ "Dbouk, Mohamad", "Katona, Bryson W.", "Brand, Randall E.", "et al." ], "publisher": "Journal of Clinical Oncology", "date": "2022-06-15", "url": "https://ascopubs.org/doi/10.1200/JCO.22.00298", "doi": "10.1200/JCO.22.00298", "pmid": "35704792", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Prospective multicentre surveillance cohort in people at high inherited or familial risk. Primary endpoint was stage distribution, not a randomised mortality comparison." }, { "id": "de-savornin-lohman-2020-incidental-gallbladder", "type": "paper", "title": "Re-resection in incidental gallbladder cancer: survival and the incidence of residual disease", "authors": [ "de Savornin Lohman, Elise A.J.", "van der Geest, Lydia G.", "de Bitter, Tessa J.J.", "et al." ], "publisher": "Annals of Surgical Oncology", "date": "2020-04-01", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC7060151/", "doi": "10.1245/s10434-019-08074-4", "pmid": "31741109", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Full open-access report checked. Netherlands Cancer Registry cohort of 463 incidental gallbladder cancers; 110 underwent re-resection. Residual disease was present in 35% of re-resected patients, most often in lymph nodes. The large apparent survival difference by re-resection is vulnerable to stage, fitness, referral and immortal-time selection and is not a randomized treatment effect." }, { "id": "de-wilde-2012-centralization", "type": "paper", "title": "Impact of nationwide centralization of pancreaticoduodenectomy on hospital mortality", "authors": [ "de Wilde, Roeland F", "Besselink, Marc G", "van der Tweel, Ingeborg", "et al." ], "publisher": "British Journal of Surgery", "date": "2012-01-11", "url": "https://europepmc.org/article/MED/22237731", "doi": "10.1002/bjs.8664", "pmid": "22237731", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Dutch nationwide registry study of 2,155 pancreaticoduodenectomies from 2004-2009. Hospitals fell from 48 to 30, medium/high-volume care rose from 52.9% to 91.2%, and nationwide in-hospital mortality fell from 9.8% to 5.1%. Across the whole period mortality was 14.7%, 9.8%, 6.3% and 3.3% in centres doing fewer than 5, 5-10, 11-19 and at least 20 procedures per year. Observational time trends, mixed indications and in-hospital mortality cannot isolate centralization from other care changes." }, { "id": "deciphera-2022-dcc3116-phase1", "type": "company-announcement", "title": "Initial phase 1 single-agent data for the ULK1/2 inhibitor DCC-3116", "authors": [ "Deciphera Pharmaceuticals, Inc." ], "publisher": "Deciphera Pharmaceuticals", "date": "2022-09-10", "url": "https://www.deciphera.com/news/deciphera-pharmaceuticals-inc-presents-initial-phase-1-single-agent-dose-escalation-data-first", "doi": null, "pmid": null, "trialId": "NCT04892017", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Sponsor report at a June 2022 cutoff: 18 people across four dose groups, 28% with pancreatic cancer. Blood cells showed reduced phosphorylated ATG14, 14 people were response-evaluable, best response was stable disease and four-month disease control was 29%. No dose-limiting toxicity or treatment-related serious adverse event was reported. The record is useful for early pharmacology but is uncontrolled, sponsor-reported and not a final 91-person result." }, { "id": "dekker-2026-taps-restaging", "type": "paper", "title": "Restaging after neoadjuvant FOLFIRINOX for localized pancreatic cancer: a clinical calculator from the Trans-Atlantic Pancreatic Surgery consortium", "authors": [ "Dekker, Esther N.", "van Klaveren, David", "Verkolf, Eva M. M.", "et al." ], "publisher": "Journal of the National Cancer Institute", "date": "2026-06-01", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC13247336/", "doi": "10.1093/jnci/djag024", "pmid": "41859816", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Retrospective five-referral-centre TAPS cohort of 2,338 localized PDAC patients after initial modified FOLFIRINOX. Eight baseline and restaging fields separated four prognostic groups whose observed three-year survival ranged from 6.0% to 65.8%; leave-one-centre-out C-index ranged 0.64-0.70. Resection was descriptive only: 989 were resected, with median survival after restaging 34.3 months versus 11.2 months without resection. The authors explicitly state that selection and immortal-time bias prevent attributing this gap to surgery. The calculator predicts prognosis, not treatment benefit." }, { "id": "dewitt-2019-eus-porfimer-pdt", "type": "paper", "title": "Phase 1 study of EUS-guided photodynamic therapy for locally advanced pancreatic cancer", "authors": [ "DeWitt, John M.", "Sandrasegaran, Kumaresan", "O'Neil, Bert", "et al." ], "publisher": "Gastrointestinal Endoscopy", "date": "2019-02-01", "url": "https://pubmed.ncbi.nlm.nih.gov/30222972/", "doi": "10.1016/j.gie.2018.09.007", "pmid": "30222972", "trialId": "NCT01770132", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Prospective single-centre dose study of endoscopic-ultrasound-guided porfimer PDT followed by nab-paclitaxel and gemcitabine in 12 untreated people with locally advanced disease. Imaging-defined tumour necrosis increased in 6/12; median progression-free survival was 2.6 months and survival 11.5 months. Two operations found one complete response and one 2 mm residual tumour. Eight serious events occurred, judged unrelated to endoscopy or PDT. There was no control and chemotherapy followed PDT." }, { "id": "dhsc-2024-pert-shortage-alert", "type": "regulatory", "title": "National Patient Safety Alert: Shortage of pancreatic enzyme replacement therapy — additional actions", "authors": [ "Department of Health and Social Care" ], "publisher": "Central Alerting System / MHRA", "date": "2024-12-18", "url": "https://www.cas.mhra.gov.uk/ViewandAcknowledgment/ViewAttachment.aspx?Attachment_id=104186", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "UK national safety alert reported limited Creon 10,000 and 25,000 supply and intermittent alternatives, caused by raw-ingredient and manufacturing-capacity limits. It required local mitigation and stated that community supply was hit more heavily than secondary care. The dated alert predicted limits into 2026; current product status must be refreshed before use." }, { "id": "digiorgio-2024-nab-pipac-protocol", "type": "paper", "title": "Combined Nabpaclitaxel pressurized intraPeritoneal aerosol chemotherapy with systemic Nabpaclitaxel-Gemcitabine chemotherapy for pancreatic cancer peritoneal metastases: protocol of single-arm, open-label, phase II trial", "authors": [ "Di Giorgio, Andrea", "Ferracci, Federica", "Bagala, Cinzia", "et al." ], "publisher": "Pleura and Peritoneum", "date": "2024-09-01", "url": "https://doi.org/10.1515/pp-2024-0010", "doi": "10.1515/pp-2024-0010", "pmid": "39544430", "trialId": "NCT05371223", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Protocol for a 38-person Simon two-stage phase II study: 12 in stage one and 26 in stage two, with 80% power and one-sided alpha 0.1. Everyone receives intravenous nab-paclitaxel-gemcitabine plus nab-paclitaxel PIPAC, so the design can estimate package activity, feasibility and harm but cannot identify the aerosol contribution." }, { "id": "digiorgio-2026-pipac-pdac-registry", "type": "paper", "title": "Pressurized intraperitoneal aerosol chemotherapy (PIPAC) for pancreatic cancer peritoneal metastases: A retrospective multicentric study from the ISSPP PIPAC Database", "authors": [ "Di Giorgio, Andrea", "et al." ], "publisher": "European Journal of Surgical Oncology", "date": "2026-08-21", "url": "https://doi.org/10.1016/j.ejso.2026.112078", "doi": "10.1016/j.ejso.2026.112078", "pmid": "42648232", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Retrospective six-centre ISSPP registry series of 156 people with PDAC peritoneal metastases receiving 350 PIPAC procedures. Only 55/156, 35.2%, completed at least three procedures. Grade 3-4 adverse events occurred after 10/350 procedures, 2.9%; six people, 3.8%, died within 30 days, all attributed to disease progression. Progression stopped PIPAC in 77 and poor condition in ten. Median survival was 9 months from first PIPAC and 19 months from diagnosis. Intraperitoneal nab-paclitaxel and completion of at least three PIPAC procedures were associated with longer survival, but treatment and repetition were not randomized; a 120-day landmark analysis addressed immortal-time bias but not residual fitness and treatment-selection bias." }, { "id": "dimatteo-2018-eus-laser-ablation", "type": "paper", "title": "Feasibility of EUS-guided Nd:YAG laser ablation of unresectable pancreatic adenocarcinoma", "authors": [ "Di Matteo, Francesco Maria", "Saccomandi, Paola", "Martino, Margareth", "et al." ], "publisher": "Gastrointestinal Endoscopy", "date": "2018-02-13", "url": "https://pubmed.ncbi.nlm.nih.gov/29452076/", "doi": "10.1016/j.gie.2018.02.007", "pmid": "29452076", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Nine people with stage IIb-III pancreatic cancer unresponsive to chemoradiation received one endoscopic-ultrasound-guided 1064 nm Nd:YAG laser application. Lesions averaged 35.4 mm across. The 24-hour dead volume ranged from 0.4 to 6.4 cubic centimetres across tested power and energy settings. All nine procedures completed and no adverse event was reported. There was no control or patient-benefit endpoint." }, { "id": "dimcevski-2016-sonoporation-phase1", "type": "paper", "title": "A human clinical trial using ultrasound and microbubbles to enhance gemcitabine treatment of inoperable pancreatic cancer", "authors": [ "Georg Dimcevski", "Spiros Kotopoulis", "Tormod Bjanes", "and collaborators" ], "publisher": "Journal of Controlled Release", "date": "2016-12-10", "url": "https://pubmed.ncbi.nlm.nih.gov/27744037/", "doi": "10.1016/j.jconrel.2016.10.007", "pmid": "27744037", "trialId": "NCT01674556", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Ten-person phase 1 feasibility study compared outcomes with 63 historical controls. The reported 17.6 versus 8.9 month median-survival difference cannot identify treatment effect because allocation was not randomized or concurrent." }, { "id": "ding-2026-prs-modifiable-risk", "type": "paper", "title": "Age-Dependent Interplay of Modifiable Risk Factors and Genetic Risk in Pancreatic Cancer", "authors": [ "Ding, Jie", "et al." ], "publisher": "JAMA Oncology", "date": "2026", "url": "https://pubmed.ncbi.nlm.nih.gov/42133339/", "doi": "10.1001/jamaoncol.2026.1192", "pmid": "42133339", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Prospective UK Biobank analysis with median 11.7-year follow-up; combines polygenic and modifiable risk scores." }, { "id": "dong-2020-pnet-recurrence", "type": "paper", "title": "Resection of pancreatic neuroendocrine tumors: defining patterns and time course of recurrence", "authors": [ "Dong, Dong H.", "Zhang, Xue F.", "Lopez-Aguiar, Alexandra G.", "et al." ], "publisher": "HPB", "date": "2020-02-01", "url": "https://pubmed.ncbi.nlm.nih.gov/31235429/", "doi": "10.1016/j.hpb.2019.05.020", "pmid": "31235429", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "US Neuroendocrine Tumor Study Group retrospective series of 1,020 curative-intent resections. Recurrence occurred in 154 people, 15.1%; 49.4% of recurrences were liver-only and 22.7% pancreas-only. Curative resection of selected recurrence was associated with overall survival comparable to no observed recurrence, but that comparison is heavily selected and not randomised." }, { "id": "dong-2021-splanchnic-neurolysis-rct", "type": "paper", "title": "Neurolytic Splanchnic Nerve Block and Pain Relief, Survival, and Quality of Life in Unresectable Pancreatic Cancer: A Randomized Controlled Trial", "authors": [ "Dong, Daosong", "Zhao, Mingfang", "Zhang, Jingmei", "et al." ], "publisher": "Anesthesiology", "date": "2021-10-01", "url": "https://doi.org/10.1097/ALN.0000000000003936", "doi": "10.1097/ALN.0000000000003936", "pmid": "34398950", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Five-centre randomized double-blind comparison of alcohol versus saline injection around the splanchnic nerves in 96 analysed people with unresectable pancreatic cancer. Maximum mean pain advantage was 0.7 points on a 0-10 scale at one month; opioid use was lower and quality of life did not differ. Survival favoured saline, HR for death 1.56, 95% CI 1.03-2.35, with stage-IV HR 1.94 and no stage-III difference. Survival was secondary and the harm mechanism is unresolved." }, { "id": "dorbin-2026-allele-specific-daraxonrasib-resistance", "type": "paper", "title": "Allele-Specific Mechanisms Guide Salvage Therapy to Overcome Daraxonrasib Resistance in Pancreatic Cancer", "authors": [ "Dorbin, Daniel", "Herrera, Juliannie", "Davidson, Raven", "et al." ], "publisher": "Cancer Research", "date": "2026-08-27", "url": "https://pubmed.ncbi.nlm.nih.gov/42658041/", "doi": "10.1158/0008-5472.CAN-26-1181", "pmid": "42658041", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Laboratory and single-patient translational study comparing G12R and G12D escape. G12R models shifted toward EGFR and wild-type RAS, while G12D models retained G12D signalling with reduced cyclophilin A. One G12R patient treated after daraxonrasib progression received a trametinib-based sixth-line combination chosen with a rapid 3D model and had about five months of disease control and 40 months overall survival. The patient result is uncontrolled n=1 and cannot establish that the model or combination caused the long survival." }, { "id": "dotan-2024-giant", "type": "paper", "title": "GIANT EA2186: randomized reduced-intensity chemotherapy for vulnerable older adults with metastatic pancreatic cancer", "authors": [ "Dotan, Efrat", "Catalano, Paul J.", "Lenchik, Leon", "et al." ], "publisher": "Journal of Clinical Oncology", "date": "2024-05-29", "url": "https://ascopubs.org/doi/10.1200/JCO.2024.42.16_suppl.4003", "doi": "10.1200/JCO.2024.42.16_suppl.4003", "pmid": null, "trialId": "NCT04233866", "patentId": null, "grantId": "U10CA180820", "accessed": "2026-09-14", "notes": "Meeting abstract reporting the first randomized national treatment trial focused on vulnerable older adults with untreated metastatic PDAC. It enrolled 176 people aged 70-90 at 92 US sites, median age 77, and stopped at its futility boundary. Median overall survival was 4.7 months with fortnightly reduced-dose gemcitabine plus nab-paclitaxel and 4.4 months with reduced-dose fluorouracil, leucovorin and liposomal irinotecan, P=.72. Grade 3 or worse toxicity was 45.6% versus 58.7%, P=.10. The abstract does not establish equivalence and does not test assessment-directed supportive action." }, { "id": "dotan-2025-giant-ga", "type": "paper", "title": "Baseline geriatric and quality-of-life factors in GIANT EA2186", "authors": [ "Dotan, Efrat", "Catalano, Paul J.", "Lenchik, Leon", "et al." ], "publisher": "Journal of Clinical Oncology", "date": "2025-01-27", "url": "https://ascopubs.org/doi/10.1200/JCO.2025.43.4_suppl.676", "doi": "10.1200/JCO.2025.43.4_suppl.676", "pmid": null, "trialId": "NCT04233866", "patentId": null, "grantId": "U10CA180820", "accessed": "2026-09-14", "notes": "Secondary meeting abstract from the 176-person GIANT trial. Instrumental daily activity, nutrition, depression and all baseline quality-of-life measures correlated with survival in unadjusted models. After adjustment for age and performance status, white-cell count and depression correlated with grade 3 or worse toxicity. These are prognostic associations from a randomized trial cohort, not randomized evidence that changing the factors changes survival." }, { "id": "drks-00012367-hifu-pc", "type": "trial-registry", "title": "HIFU-PC15: HIFU plus standard chemotherapy versus standard chemotherapy in inoperable pancreatic cancer", "authors": [ "University Hospital Bonn" ], "publisher": "German Clinical Trials Register", "date": "2024-02-13", "url": "https://drks.de/search/en/trial/DRKS00012367/details", "doi": null, "pmid": null, "trialId": "DRKS00012367", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Primary registry. The single-centre randomized study started 24 February 2017, targets 40 people, still says recruiting on a record last updated 13 February 2024, posts no final count or results and states that individual participant data will not be shared." }, { "id": "duell-2017-epic-mirna-pdac", "type": "paper", "title": "Plasma microRNAs as biomarkers of pancreatic cancer risk in a prospective cohort study", "authors": [ "Duell, Eric J.", "Lujan-Barroso, Leila", "Sala, Núria", "et al." ], "publisher": "International Journal of Cancer", "date": "2017", "url": "https://pubmed.ncbi.nlm.nih.gov/28542740/", "doi": null, "pmid": "28542740", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Nested case-control analysis in EPIC: 225 later PDAC cases and 225 matched controls, including long prediagnostic intervals." }, { "id": "dufva-2026-home-palliative-pdac", "type": "paper", "title": "Early Specialized Palliative Care for Unresectable Pancreatic Cancer: A Quasi-Experimental Study", "authors": [ "Dufva, Inge", "Juhl, Gitte", "Andersen, Fahimeh", "et al." ], "publisher": "Journal of Pain and Symptom Management", "date": "2026-03-01", "url": "https://doi.org/10.1016/j.jpainsymman.2025.11.025", "doi": "10.1016/j.jpainsymman.2025.11.025", "pmid": "41319817", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Home-visit programme offered to 65 newly diagnosed unresectable pancreatic patients, compared with 60 earlier historical controls. Median survival was 8.0 versus 4.9 months, p=0.029; monthly admissions, emergency visits and hospital days were lower, and deaths in an ordinary hospital ward were 14.5% versus 30.5%. Time-period and selection confounding prevent a causal survival claim." }, { "id": "dugic-2026-pert-mortality", "type": "paper", "title": "Association between pancreatic enzyme replacement therapy and mortality in pancreatic cancer: a matched case-control study", "authors": [ "Dugic, Ana", "Hagström, Hannes", "Palmer, Isabella", "et al." ], "publisher": "Pancreatology", "date": "2026-02-01", "url": "https://pubmed.ncbi.nlm.nih.gov/41387158/", "doi": "10.1016/j.pan.2025.12.008", "pmid": "41387158", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Tertiary-centre matched case-control study of 1,247 PDAC patients, with 487 matched pairs. PERT was associated with lower one-year mortality overall, adjusted OR .49, but not significantly in the chemotherapy-only subgroup, OR .75, 95% CI .44-1.24. Authors warn that selection may make the association too optimistic; several disclosed PERT-company relationships." }, { "id": "dynamic-pancreas-registry", "type": "trial-registry", "title": "Circulating Tumour DNA Analysis Informing Adjuvant Chemotherapy in Early Stage Pancreatic Cancer: DYNAMIC-Pancreas", "authors": [ "Australasian Gastro-Intestinal Trials Group" ], "publisher": "Australian New Zealand Clinical Trials Registry", "date": "2018-03-06", "url": "https://anzctr.org.au/Trial/Registration/TrialReview.aspx?ACTRN=12618000335291", "doi": null, "pmid": null, "trialId": "ACTRN12618000335291", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Prospectively registered multicentre randomised study designed to test ctDNA-informed adjuvant chemotherapy. The registry was last updated June 2023; a 2024 conference report addressed feasibility, not definitive survival benefit." }, { "id": "eckstein-2023-proton-registry", "type": "paper", "title": "Proton Therapy for Unresectable and Medically Inoperable Locally Advanced Pancreatic Cancer: Results From a Multi-Institutional Prospective Registry", "authors": [ "Eckstein, Joseph", "Choi, James I.", "Lozano, Anthony", "et al." ], "publisher": "Advances in Radiation Oncology", "date": "2023-09-01", "url": "https://pubmed.ncbi.nlm.nih.gov/37408677/", "doi": "10.1016/j.adro.2023.101250", "pmid": "37408677", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Prospective registry series from seven institutions. Nineteen people received definitive proton treatment, 17 with adenocarcinoma. No grade 3 or worse treatment-related event was reported; median overall survival was 14.6 months and two-year freedom from local recurrence was 81.7%. There was no concurrent photon control." }, { "id": "edeline-2019-prodige12", "type": "paper", "title": "Gemcitabine and Oxaliplatin Chemotherapy or Surveillance in Resected Biliary Tract Cancer: PRODIGE 12-ACCORD 18-UNICANCER GI", "authors": [ "Edeline, Julien", "Benabdelghani, Meher", "Bertaut, Aurélie", "et al." ], "publisher": "Journal of Clinical Oncology", "date": "2019-02-01", "url": "https://pubmed.ncbi.nlm.nih.gov/30707660/", "doi": "10.1200/JCO.18.00050", "pmid": "30707660", "trialId": "NCT01313377", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Open-label randomised phase 3 at 33 French centres; 196 patients with mixed resected biliary tract cancer sites, GEMOX versus surveillance." }, { "id": "edi-nct04662879", "type": "trial-registry", "title": "Early Detection Initiative for Pancreatic Cancer", "authors": [ "Pancreatic Cancer Action Network" ], "publisher": "ClinicalTrials.gov", "date": "2026-09-08", "url": "https://clinicaltrials.gov/study/NCT04662879", "doi": null, "pmid": null, "trialId": "NCT04662879", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "ClinicalTrials.gov API checked after its 8 September 2026 update. Active, not recruiting; actual enrolment 8,869; non-randomized parallel site design; no results posted; estimated primary completion January 2028. Intervention sites use ENDPAC above 0 and up to two CT or MRI studies, while observation sites use health-record follow-up. Outcomes include model validation, lead time, stage and incidental findings; the registry does not list a randomized mortality endpoint. Registry status and dates can change." }, { "id": "effenberger-2012-pdac-bone-marrow-dtc", "type": "paper", "title": "Disseminated tumor cells in pancreatic cancer — an independent prognosticator of disease progression and survival", "authors": [ "Effenberger, Katharina E.", "Schroeder, Cornelia", "Eulenburg, Christine", "et al." ], "publisher": "International Journal of Cancer", "date": "2012-08-15", "url": "https://pubmed.ncbi.nlm.nih.gov/21932421/", "doi": "10.1002/ijc.26439", "pmid": "21932421", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Bone marrow aspirates taken at primary surgery from 175 analysed pancreatic-cancer patients found cytokeratin-positive cells in 24, 13.7%. Ninety-six underwent complete resection. Positivity associated with progression-free survival HR 2.057 and overall survival HR 2.283 in multivariable models. The one-time assay did not prove viability, quiescence, clonality, metastatic ancestry or benefit from a bone-targeted action." }, { "id": "elevar-2024-company-overview", "type": "company", "title": "Elevar Therapeutics company overview", "authors": [ "Elevar Therapeutics" ], "publisher": "Elevar Therapeutics", "date": "2024-11-01", "url": "https://elevartherapeutics.com/wp-content/uploads/2024/11/Elevar-Therapeutics-Company-Overview-Non-Confidential.pdf", "doi": null, "pmid": null, "trialId": "NCT03764293", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Company deck used only for corporate history and ownership. It reports founding as LSK BioPharma in 2005, HLB's $1.3m initial investment in 2007 and $440m acquisition in 2020. Product and market claims require independent clinical sources." }, { "id": "elicio-2026-amplify7p-results", "type": "company-filing", "title": "Elicio Therapeutics Form 10-Q for the quarter ended June 30, 2026 — AMPLIFY-7P results", "authors": [ "Elicio Therapeutics, Inc." ], "publisher": "U.S. Securities and Exchange Commission", "date": "2026-08-10", "url": "https://www.sec.gov/Archives/edgar/data/1601485/000160148526000056/eltx-20260630.htm", "doi": null, "pmid": null, "trialId": "NCT05726864", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Sponsor filing reporting topline randomized phase 2 results in 144 people at 24 U.S. sites. The prespecified intention-to-treat DFS endpoint was missed. The filing emphasizes short landmark analyses, immune-response associations and subgroup work; mature OS and a full peer-reviewed report were not available." }, { "id": "ema-2020-dpd-testing", "type": "regulator", "title": "EMA recommendations on DPD testing before systemic fluoropyrimidines", "authors": [ "European Medicines Agency" ], "publisher": "European Medicines Agency", "date": "2020-04-30", "url": "https://www.ema.europa.eu/en/medicines/human/referrals/fluorouracil-fluorouracil-related-substances-capecitabine-tegafur-flucytosine-containing-medicinal-products", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "European regulatory review recommends testing for DPD deficiency before injected fluorouracil, capecitabine or tegafur. Complete deficiency is a contraindication; partial deficiency calls for a reduced starting dose with later adjustment. The review allows genotype or plasma-uracil testing and states that reduced-dose efficacy was not established by the review." }, { "id": "ema-2026-daraxonrasib-orphan-designation", "type": "regulatory", "title": "EU/3/26/3227 orphan designation for treatment of pancreatic cancer", "authors": [ "European Medicines Agency" ], "publisher": "European Medicines Agency", "date": "2026-04-20", "url": "https://www.ema.europa.eu/nl/medicines/human/orphan-designations/eu-3-26-3227", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Official EU orphan designation for daraxonrasib in pancreatic cancer. EMA explicitly states that orphan designation does not authorize marketing or establish availability. Sponsorship transferred to Revolution Medicines Netherlands in July 2026." }, { "id": "ema-2026-daraxonrasib-phased-review", "type": "regulatory", "title": "EMA fast tracks review of a medicine for metastatic pancreatic cancer", "authors": [ "European Medicines Agency" ], "publisher": "European Medicines Agency", "date": "2026-07-01", "url": "https://www.ema.europa.eu/is/news/ema-fast-tracks-review-medicine-metastatic-pancreatic-cancer", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "EMA's medicines committee began a phased review of daraxonrasib after the randomized metastatic PDAC result. A phased review examines evidence before a complete marketing application and may shorten assessment; it is not marketing authorization and has no predictable completion date." }, { "id": "enray-2026-company", "type": "company", "title": "Sichuan Enray Pharmaceutical Sciences company and pipeline", "authors": [ "Sichuan Enray Pharmaceutical Sciences Company" ], "publisher": "Sichuan Enray Pharmaceutical Sciences Company", "date": "2026-09-15", "url": "https://www.enraypharm.com/", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Company site identifies Enray as a controlled subsidiary of listed Chongqing Lummy Pharmaceutical and as the developer of CNSI-Fe from an earlier internal innovation laboratory. It promotes ferroptosis, immune and heat mechanisms and lists pancreatic and colorectal applications. Current recruitment highlighted on the site is a 52-person CNSI-Fe plus radiation study in sarcoma and head-and-neck cancer, not a dedicated pancreatic trial." }, { "id": "esge-2018-biliary-stenting-guideline", "type": "guideline", "title": "ESGE guideline: endoscopic biliary stenting, indications and stent choice", "authors": [ "Dumonceau, Jean-Marc", "Tringali, Andrea", "Papanikolaou, Ioannis S", "et al." ], "publisher": "European Society of Gastrointestinal Endoscopy", "date": "2018-09-01", "url": "https://www.esge.com/assets/downloads/pdfs/guidelines/2018_a_0659_9864.pdf", "doi": "10.1055/a-0659-9864", "pmid": "30086596", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Primary society guideline. It recommends against routine preoperative drainage of malignant extrahepatic obstruction, reserving it for cholangitis, severe symptomatic jaundice, delayed surgery or jaundiced patients before neoadjuvant therapy. When drainage is chosen, it recommends endoscopic placement of a 10-mm self-expanding metal stent. The recommendation predates the 2026 metal-stent-versus-early-surgery trial and must be read with newer evidence." }, { "id": "esmo-2025-virage-2216mo", "type": "meeting-report", "title": "Research in pancreatic adenocarcinoma targets antibody-drug conjugates and oncolytic viruses", "authors": [ "European Society for Medical Oncology" ], "publisher": "ESMO Daily Reporter", "date": "2025-10-20", "url": "https://dailyreporter.esmo.org/homepage/pancreatic-cancer/research-in-pancreatic-adenocarcinoma-targets-antibody-drug-conjugates-and-oncolytic-viruses", "doi": null, "pmid": null, "trialId": "NCT05673811", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Official congress report of abstract 2216MO. In the 96-person FAS, ORR was 39.6% versus 31.3%; OS HR was 0.57 with Cox P=0.034 and log-rank P=0.055. Grade 3 or worse transaminase rise affected 15.1% and influenza-like illness 13.2% after the first virus dose. The report notes a liver-metastasis imbalance, 83% control versus 65% virus, that may limit PFS interpretation." }, { "id": "ettrich-2026-unitepanc", "type": "trial-protocol", "title": "Using organoids to predict efficacy of adjuvant treatment to improve outcome in resectable pancreatic cancer: UNITEPANC", "authors": [ "Ettrich, Thomas Jens", "Perkhofer, Lukas", "Kleger, Alexander", "Seufferlein, Thomas" ], "publisher": "Journal of Clinical Oncology", "date": "2026-05-27", "url": "https://ascopubs.org/doi/10.1200/JCO.2026.44.16_suppl.TPS4266", "doi": "10.1200/JCO.2026.44.16_suppl.TPS4266", "pmid": null, "trialId": "2023-510490-34-00", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Current German AIO single-arm proof-of-concept protocol. It plans 92 resected patients for organoid generation, 38 for organoid-chosen adjuvant treatment and 34 for analysis. The recommendation must allow treatment within 12 weeks; generation and expansion targets are 75% and 60%. Efficacy is descriptive against a historical 18-month disease-free-survival benchmark. Recruitment began in Q3 2025." }, { "id": "eudract-2008-004802-14-heat", "type": "trial-registry", "title": "HEAT: gemcitabine and capecitabine versus gemcitabine, cisplatin and regional hyperthermia after pancreatic cancer resection", "authors": [ "Ludwig-Maximilians University of Munich" ], "publisher": "EU Clinical Trials Register", "date": "2022-03-01", "url": "https://www.clinicaltrialsregister.eu/ctr-search/trial/2008-004802-14/DE", "doi": null, "pmid": null, "trialId": "EudraCT 2008-004802-14", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Primary European regulator registry. It marks the non-commercial phase 3 trial prematurely ended, gives 1 March 2022 as the global end date, and preserves the protocol comparison in which the control includes capecitabine while the experimental arm includes cisplatin and regional hyperthermia. No reason for the early end is stated in this record." }, { "id": "eudract-2018-003925-27-ect", "type": "trial-registry", "title": "Laparoscopic-ECT-LAPC: randomized phase 2b electrochemotherapy trial", "authors": [ "Istituto Nazionale Tumori IRCCS Fondazione Pascale" ], "publisher": "EU Clinical Trials Register", "date": "2018-12-17", "url": "https://www.clinicaltrialsregister.eu/ctr-search/trial/2018-003925-27/IT", "doi": null, "pmid": null, "trialId": "EudraCT 2018-003925-27; EU CT 2024-519067-17-00", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Primary European registry. The 90-person, 60-month enrollment plan started in April 2019 and is marked transitioned to CTIS with no public results in the old registry. The successor EU CT identifier is 2024-519067-17-00." }, { "id": "evans-2026-fapi-feasibility", "type": "paper", "title": "68Ga-FAPI-46 PET/CT Versus 18F-FDG PET/CT for Pancreatic Cancer: A Feasibility Study", "authors": [ "Evans, Nicola", "Perkins, Ebony", "Firouzmand, Sepinoud", "et al." ], "publisher": "Journal of Medical Imaging and Radiation Oncology", "date": "2026-05-25", "url": "https://pubmed.ncbi.nlm.nih.gov/42186148/", "doi": "10.1111/1754-9485.70124", "pmid": "42186148", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Ten-person feasibility study. FAPI found additional lesions and produced imaging-based upstaging in three participants, but also accumulated in presumed inflammation." }, { "id": "exact-2025-cancerguard-brochure", "type": "company-product", "title": "Cancerguard multi-cancer early detection test clinician brochure", "authors": [ "Exact Sciences Corporation" ], "publisher": "Exact Sciences Corporation", "date": "2025-07-01", "url": "https://www.exactsciences.com/cancerguardtestinfo", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Company brochure: pooled sensitivity 67.9% in 106 samples from six aggressive cancer types. The figure is not pancreatic-specific; key development and validation reports are listed as company data on file, and the brochure notes no established imaging guideline after a positive result." }, { "id": "faggiano-2020-men1-lanreotide", "type": "paper", "title": "Lanreotide Therapy vs Active Surveillance in MEN1-Related Pancreatic Neuroendocrine Tumors Under 2 Centimeters", "authors": [ "Faggiano, Antongiulio", "Modica, Roberta", "Lo Calzo, Fabio" ], "publisher": "Journal of Clinical Endocrinology and Metabolism", "date": "2020-01-01", "url": "https://pubmed.ncbi.nlm.nih.gov/31586182/", "doi": "10.1210/clinem/dgz007", "pmid": "31586182", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Prospective but non-randomized comparison in 42 MEN1 patients with 91 pNETs below 2 cm: 23 received lanreotide and 19 active surveillance. At median 73 months, progression was reported in 4 treated versus 13 observed patients and median progression-free survival was not reached versus 40 months. Selection, small sample, lesion-level multiplicity and absence of survival or surgery-avoidance proof prevent a preventive-treatment conclusion." }, { "id": "fahrmann-2021-ca19-9-trajectory", "type": "paper", "title": "Lead-Time Trajectory of CA19-9 as an Anchor Marker for Pancreatic Cancer Early Detection", "authors": [ "Fahrmann, Johannes F.", "et al." ], "publisher": "Gastroenterology", "date": "2020-12-17", "url": "https://pubmed.ncbi.nlm.nih.gov/33333055/", "doi": "10.1053/j.gastro.2020.11.052", "pmid": "33333055", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Blinded nested case-control analysis using PLCO sera from 175 future pancreatic-cancer cases and 875 matched controls, plus independent diagnosis-time sets. Samples existed before diagnosis, but this was not a deployed screening study." }, { "id": "fan-2018-oral-microbiome-pdac-risk", "type": "paper", "title": "Human oral microbiome and prospective risk for pancreatic cancer: a population-based nested case-control study", "authors": [ "Fan, Xiaozhou", "Aleksandrov, Leila B.", "Wu, Jing", "et al." ], "publisher": "Gut", "date": "2018", "url": "https://pubmed.ncbi.nlm.nih.gov/27742762/", "doi": null, "pmid": "27742762", "trialId": null, "patentId": null, "grantId": "R01CA159036; U01CA182370; R21CA183887", "accessed": "2026-09-14", "notes": "Nested case-control study of prediagnostic oral washes from 361 later pancreatic cancers and 371 controls in two US cohorts." }, { "id": "fda-2018-larotrectinib-approval", "type": "regulator", "title": "FDA accelerated approval of larotrectinib for NTRK fusion solid tumours", "authors": [ "US Food and Drug Administration" ], "publisher": "US Food and Drug Administration", "date": "2018-11-26", "url": "https://www.fda.gov/drugs/fda-approves-larotrectinib-solid-tumors-ntrk-gene-fusions-0", "doi": null, "pmid": null, "trialId": "NCT02122913; NCT02637687; NCT02576431", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Accelerated approval pooled 55 adults and children across 12 solid-tumour types and found 75% response, 95% CI 61% to 85%. Pancreatic cancer was not reported as a separate cohort, and continued approval depended on confirmatory evidence. The result supports access for a qualifying NTRK fusion, not a pancreatic response rate." }, { "id": "fda-2018-oncology-endpoints", "type": "regulatory", "title": "Clinical Trial Endpoints for the Approval of Cancer Drugs and Biologics", "authors": [ "U.S. Food and Drug Administration" ], "publisher": "U.S. Food and Drug Administration", "date": "2018-12-01", "url": "https://www.fda.gov/regulatory-information/search-fda-guidance-documents/clinical-trial-endpoints-approval-cancer-drugs-and-biologics", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "FDA guidance distinguishes overall survival, progression-free or event-free survival, tumour response and patient-reported outcomes. It states that response and progression measures can read out sooner but may not track survival, while overall survival is direct but slower and affected by later treatment and crossover." }, { "id": "fda-2019-adaptive-trial-guidance", "type": "regulatory", "title": "Adaptive Design Clinical Trials for Drugs and Biologics Guidance for Industry", "authors": [ "US Food and Drug Administration" ], "publisher": "US Food and Drug Administration", "date": "2019-12-02", "url": "https://www.fda.gov/regulatory-information/search-fda-guidance-documents/adaptive-design-clinical-trials-drugs-and-biologics-guidance-industry", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Final FDA guidance on design, conduct and reporting of adaptive trials, including Bayesian and complex designs that use computer simulation. It is a design standard, not evidence that adaptation improves a weak treatment." }, { "id": "fda-2023-externally-controlled-trials-draft", "type": "regulatory", "title": "Considerations for the Design and Conduct of Externally Controlled Trials for Drug and Biological Products", "authors": [ "US Food and Drug Administration" ], "publisher": "US Food and Drug Administration", "date": "2023-02-01", "url": "https://www.fda.gov/regulatory-information/search-fda-guidance-documents/considerations-design-and-conduct-externally-controlled-trials-drug-and-biological-products", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Official draft guidance, not for implementation and non-binding. It defines externally controlled trials and their use for safety or effectiveness evidence; it does not make an AI-generated external control equivalent to randomisation." }, { "id": "fda-2024-bizengri-approval", "type": "regulatory", "title": "FDA grants accelerated approval to zenocutuzumab-zbco for NRG1 fusion-positive pancreatic adenocarcinoma and non-small cell lung cancer", "authors": [ "U.S. Food and Drug Administration" ], "publisher": "U.S. Food and Drug Administration", "date": "2024-12-04", "url": "https://www.fda.gov/drugs/resources-information-approved-drugs/fda-disco-burst-edition-fda-approval-bizengri-zenocutuzumab-zbco-treatment-adults-advanced", "doi": null, "pmid": null, "trialId": "NCT02912949", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Accelerated approval for previously treated advanced unresectable or metastatic NRG1 fusion-positive pancreatic adenocarcinoma. In the approval set, 12 of 30 pancreatic patients responded by blinded independent review, ORR 40%, with response durations from 3.7 to 16.6 months. The label carries an embryo-fetal toxicity boxed warning." }, { "id": "fda-2024-bizengri-multidisciplinary-review", "type": "regulatory", "title": "Multidisciplinary review and evaluation: BIZENGRI, BLA 761352", "authors": [ "U.S. Food and Drug Administration" ], "publisher": "U.S. Food and Drug Administration", "date": "2024-12-04", "url": "https://www.accessdata.fda.gov/drugsatfda_docs/nda/2025/761352Orig1s000MultidisciplineR.pdf", "doi": null, "pmid": null, "trialId": "NCT02912949", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "FDA review estimates NRG1 fusions near 1.01% of PDAC, while its benefit-risk table uses below 0.5%, showing prevalence uncertainty. The efficacy set was 30 pancreatic patients: ORR 40%, 95% CI 23% to 59%; 67% of 12 responders lasted at least six months and 17% at least twelve months. At cutoff, 24 of 30 had stopped treatment, including 20 for scan or clinical progression. No approved companion diagnostic was available. FDA required at least 50 evaluable pancreatic patients with at least 12 months of response follow-up, with final report due August 2026, plus later integrated safety and diagnostic-validation work." }, { "id": "fda-2024-zanidatamab-approval", "type": "regulatory", "title": "FDA grants accelerated approval to zanidatamab-hrii for previously treated unresectable or metastatic HER2-positive biliary tract cancer", "authors": [ "U.S. Food and Drug Administration" ], "publisher": "U.S. Food and Drug Administration", "date": "2024-11-20", "url": "https://www.fda.gov/drugs/resources-information-approved-drugs/fda-grants-accelerated-approval-zanidatamab-hrii-previously-treated-unresectable-or-metastatic-her2", "doi": null, "pmid": null, "trialId": "NCT04466891", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Primary U.S. accelerated-approval source. Approval was based on response rate and duration, not randomised survival, and requires confirmatory clinical benefit." }, { "id": "fda-2025-ai-device-lifecycle-draft", "type": "regulatory", "title": "Artificial Intelligence-Enabled Device Software Functions: Lifecycle Management and Marketing Submission Recommendations", "authors": [ "US Food and Drug Administration" ], "publisher": "US Food and Drug Administration", "date": "2025-01-07", "url": "https://www.fda.gov/regulatory-information/search-fda-guidance-documents/artificial-intelligence-enabled-device-software-functions-lifecycle-management-and-marketing", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Official draft guidance, not for implementation and non-binding. It describes proposed marketing-submission and full-product-life-cycle evidence for AI-enabled device software; it is not proof that any oncology model is safe or effective." }, { "id": "fda-2026-bizengri-cholangiocarcinoma", "type": "regulatory", "title": "FDA grants seventh approval under the National Priority Voucher Pilot Program", "authors": [ "U.S. Food and Drug Administration" ], "publisher": "U.S. Food and Drug Administration", "date": "2026-05-08", "url": "https://www.fda.gov/news-events/press-announcements/fda-grants-seventh-approval-under-national-priority-voucher-pilot-program", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Official FDA notice approving zenocutuzumab-zbco for previously treated, advanced unresectable or metastatic NRG1-fusion cholangiocarcinoma. The efficacy set was single-arm and contained 19 people; 36.8% responded and response duration ranged from 2.8 to 12.9 months. Approval establishes access for this exact molecular group, not comparative survival or durable control." }, { "id": "fda-2026-ongoing-bizengri-accelerated-approval", "type": "regulatory", "title": "Ongoing cancer accelerated approvals: BIZENGRI pancreatic indication", "authors": [ "U.S. Food and Drug Administration" ], "publisher": "U.S. Food and Drug Administration", "date": "2026-09-10", "url": "https://www.fda.gov/drugs/resources-information-approved-drugs/ongoing-cancer-accelerated-approvals", "doi": null, "pmid": null, "trialId": "NCT02912949", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Current FDA list still classifies the pancreatic indication as an ongoing accelerated approval. It repeats requirement 4727-2 for at least 50 evaluable pancreatic patients followed for response and duration, with original projected completion 31 August 2026. The deadline had passed by 15 September 2026 and verified benefit was not yet shown on this FDA status page." }, { "id": "fda-2026-optune-pax-ssed", "type": "regulatory", "title": "Optune Pax: FDA summary of safety and effectiveness data", "authors": [ "U.S. Food and Drug Administration" ], "publisher": "U.S. Food and Drug Administration", "date": "2026-02-11", "url": "https://www.accessdata.fda.gov/cdrh_docs/pdf25/P250034B.pdf", "doi": null, "pmid": null, "trialId": "NCT03377491", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Primary U.S. regulator review for PMA P250034. FDA approved Optune Pax for adults with locally advanced pancreatic cancer together with gemcitabine and nab-paclitaxel. The review reports all-randomized effectiveness, treatment exposure, quality-of-life limits, adverse events and financial disclosure. Median device use was 11.2 hours per day over 27.6 weeks; 76.3% had device-related skin toxicity, 7.7% had grade 3 or worse skin toxicity and 8.4% stopped the device because of a device-related adverse event." }, { "id": "fda-2026-rasonque-approval", "type": "regulatory", "title": "FDA approves daraxonrasib for metastatic pancreatic adenocarcinoma", "authors": [ "U.S. Food and Drug Administration" ], "publisher": "U.S. Food and Drug Administration", "date": "2026-08-26", "url": "https://www.fda.gov/drugs/resources-information-approved-drugs/fda-approves-daraxonrasib-metastatic-pancreatic-adenocarcinoma", "doi": null, "pmid": null, "trialId": "NCT06625320", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Primary U.S. regulatory source. Approval covers adults with metastatic pancreatic adenocarcinoma after at least one prior systemic therapy or patients not eligible for multiagent systemic therapy. Prescribing information was also checked for toxicity and trial details." }, { "id": "fda-den220087-edison", "type": "regulatory", "title": "De Novo Classification Request for Edison System, DEN220087", "authors": [ "US Food and Drug Administration" ], "publisher": "US Food and Drug Administration", "date": "2023-10-06", "url": "https://www.accessdata.fda.gov/cdrh_docs/reviews/DEN220087.pdf", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "FDA De Novo summary for the HistoSonics Edison System. Indication is non-invasive destruction of liver tumours, not pancreas. FDA states the device had not been evaluated for a specific disease and that safety and effectiveness for cancer treatment had not been established. The supporting 44-person liver study measured technical ablation, not disease outcome." }, { "id": "feig-2013-cxcl12-immune-exclusion", "type": "paper", "title": "Targeting CXCL12 from FAP-expressing carcinoma-associated fibroblasts synergizes with anti-PD-L1 immunotherapy in pancreatic cancer", "authors": [ "Feig, Clare", "Jones, James O.", "Kraman, Matthew", "et al." ], "publisher": "Proceedings of the National Academy of Sciences", "date": "2013-11-25", "url": "https://pubmed.ncbi.nlm.nih.gov/24277834/", "doi": "10.1073/pnas.1320318110", "pmid": "24277834", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Mechanistic work in an autochthonous mouse PDAC model. Human tumours informed the model, but therapeutic synergy was not a human result." }, { "id": "fibrogen-2024-pamrevlumab-winddown", "type": "filing", "title": "FibroGen announces two negative late-stage pamrevlumab pancreatic trials and programme wind-down", "authors": [ "FibroGen, Inc." ], "publisher": "FibroGen, Inc.", "date": "2024-07-30", "url": "https://www.sec.gov/Archives/edgar/data/921299/000095017024091766/fgen-ex99_1.htm", "doi": null, "pmid": null, "trialId": "NCT03941093; NCT04229004", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Company filing confirms that both LAPIS and the PanCAN Precision Promise pamrevlumab arm missed overall survival, after which FibroGen ended pamrevlumab R&D and announced an approximately 75% U.S. workforce reduction." }, { "id": "fietkau-2025-conko007", "type": "paper", "title": "Benefit of Chemoradiotherapy Versus Chemotherapy After Induction Therapy for Conversion of Unresectable Into Resectable Pancreatic Cancer: The Randomized CONKO-007 Trial", "authors": [ "Fietkau, Rainer", "Ghadimi, Michael", "Gruetzmann, Robert", "et al." ], "publisher": "Journal of Clinical Oncology", "date": "2025-10-20", "url": "https://pubmed.ncbi.nlm.nih.gov/40802908/", "doi": "10.1200/JCO-24-01502", "pmid": "40802908", "trialId": "NCT01827553", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Open-label phase 3 trial. 525 enrolled; 336 without progression after induction chemotherapy were randomized. The primary endpoint was changed from OS to overall R0 resection because of slow recruitment. Published erratum: PMID 41505667." }, { "id": "finn-2020-imbrave150", "type": "paper", "title": "Atezolizumab plus Bevacizumab in Unresectable Hepatocellular Carcinoma", "authors": [ "Finn, Richard S.", "Qin, Shukui", "Ikeda, Masafumi", "et al." ], "publisher": "New England Journal of Medicine", "date": "2020-05-14", "url": "https://pubmed.ncbi.nlm.nih.gov/32402160/", "doi": "10.1056/NEJMoa1915745", "pmid": "32402160", "trialId": "NCT03434379", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Global open-label phase 3 first-line trial in 501 patients with unresectable HCC. Funded by Roche/Genentech; eligibility and bleeding-risk management constrain transfer to all HCC patients." }, { "id": "forjaz-2026-3d-multiomic-preprint", "type": "preprint", "title": "Deep Learning Enabled 3D Multi-Omic Analysis Reveals Molecular Signatures of Heterogeneous Response to Chemotherapy in Pancreatic Cancer", "authors": [ "Forjaz, André", "Mojdeganlou, Hengameh", "Valentin, Alens", "et al." ], "publisher": "bioRxiv", "date": "2026-03-05", "url": "https://www.biorxiv.org/content/10.64898/2026.03.03.709150v1", "doi": "10.64898/2026.03.03.709150", "pmid": "41867770", "trialId": null, "patentId": "US18/572352", "grantId": "U54CA268083; U24CA284156; T32CA193145; S10OD034407", "accessed": "2026-09-14", "notes": "Preprint using deep learning, 3D morphology and spatial multi-omics in treated human PDAC samples; associates persistent cells with EMT-like states and syndecan-linked fibroblast crosstalk. Not peer reviewed or prospectively predictive." }, { "id": "frank-2025-pandome", "type": "paper", "title": "Pancreatic Cancer Screening in New-onset and Deteriorating Diabetes: Preliminary Results From the PANDOME Study", "authors": [ "Frank, Richard C.", "Shim, Brian", "Lo, Tammy", "et al." ], "publisher": "Journal of Clinical Endocrinology & Metabolism", "date": "2025-12-18", "url": "https://pubmed.ncbi.nlm.nih.gov/40439123/", "doi": "10.1210/clinem/dgaf319", "pmid": "40439123", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Prospective MRI/MRCP study: 625 people were screened for eligibility and 109 enrolled, including 97 with new-onset and 12 with deteriorating diabetes. Four underwent pancreas biopsy and one stage-I PDAC was found in the deteriorating-diabetes group. Preliminary single-programme results with one cancer cannot establish sensitivity, net benefit or mortality effect." }, { "id": "frere-2020-pdac-thromboprophylaxis-meta", "type": "paper", "title": "Primary thromboprophylaxis in ambulatory pancreatic cancer patients receiving chemotherapy: systematic review and meta-analysis of randomized trials", "authors": [ "Frere, Corinne", "Crichi, Benjamin", "Bournet, Barbara", "et al." ], "publisher": "Cancers", "date": "2020-07-21", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC7464699/", "doi": "10.3390/cancers12082028", "pmid": "32722064", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Five randomized studies and 1,003 pancreatic-cancer patients. Prophylaxis reduced VTE with pooled RR .31 and absolute risk difference -8 points, giving an estimated NNT 11.9. Major bleeding was 3.10% versus 2.84%, pooled RR 1.08, 95% CI .47-2.52. The analysis joined different drugs, doses, durations and endpoints and did not establish a survival benefit." }, { "id": "ftc-2024-illumina-grail", "type": "regulatory", "title": "FTC case record: Illumina and GRAIL", "authors": [ "Federal Trade Commission" ], "publisher": "Federal Trade Commission", "date": "2024-08-15", "url": "https://www.ftc.gov/legal-library/browse/cases-proceedings/201-0144-illumina-inc-grail-inc-matter", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Primary regulatory case record. The FTC describes Illumina's proposed GRAIL acquisition as $7.1bn, records its challenge and divestiture order, and records the June 2024 spin-off followed by dismissal of the complaint." }, { "id": "fujii-nishimura-2018-schwann-met", "type": "paper", "title": "Mesenchymal-epithelial transition of pancreatic cancer cells at perineural invasion sites is induced by Schwann cells", "authors": [ "Fujii-Nishimura, Yoko", "Yamazaki, Ken", "Masugi, Yohei", "et al." ], "publisher": "Pathology International", "date": "2018-02-19", "url": "https://pubmed.ncbi.nlm.nih.gov/29457853/", "doi": "10.1111/pin.12641", "pmid": "29457853", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Keio study of 168 resected PDACs plus four cell-line co-cultures. Seventy-four percent showed focal differentiation at nerve invasion sites; the mechanistic induction experiment remained in culture." }, { "id": "fusco-2021-kras-wildtype-fusions", "type": "paper", "title": "Identification of Targetable Gene Fusions and Structural Rearrangements to Foster Precision Medicine in KRAS Wild-Type Pancreatic Cancer", "authors": [ "Fusco, Michael J.", "Saeed-Vafa, Daryoush", "Carballido, Estrella M.", "et al." ], "publisher": "JCO Precision Oncology", "date": "2021-01-11", "url": "https://doi.org/10.1200/PO.20.00265", "doi": "10.1200/PO.20.00265", "pmid": "34250383", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Single-centre consecutive cohort of 100 PDAC cases with two external validation cohorts. Thirteen of 100 were KRAS wild type; targetable fusions occurred in 4 of 13 KRAS-wild-type and 0 of 87 KRAS-mutant tumours. External fusion rates among KRAS-wild-type tumours were 18 of 97 and 2 of 10. One RDX-MET fusion case had a complete response to crizotinib lasting at least 12 months." }, { "id": "gannon-2026-pancreas-histotripsy-safety", "type": "paper", "title": "Non-Invasive Pancreas Ablation Using Histotripsy: Pre-clinical Safety Study in an In Vivo Porcine Model", "authors": [ "Gannon, Jessica", "Paul, Tamalika", "Imran, Khan Mohammad", "et al." ], "publisher": "Ultrasound in Medicine & Biology", "date": "2025-10-03", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC12917764/", "doi": "10.1016/j.ultrasmedbio.2025.07.026", "pmid": "41046199", "trialId": null, "patentId": null, "grantId": "R01CA269811; EB028429; CA274439; FUF61057", "accessed": "2026-09-15", "notes": "Healthy-pancreas study in nine young pigs. Pancreas targeting succeeded in six; five had visible bubble clouds. Poor ultrasound visibility caused three bowel mistargets, including septic peritonitis and gastrointestinal blockage requiring early killing. Two pigs had possible pancreatitis. HistoSonics supplied the system and funding and several authors worked for or consulted for the company." }, { "id": "garcia-carbonero-2022-vcn01-intravenous", "type": "paper", "title": "Phase I, multicenter, open-label study of intravenous VCN-01 oncolytic adenovirus with or without nab-paclitaxel plus gemcitabine in patients with advanced solid tumors", "authors": [ "Garcia-Carbonero, Rocio", "Bazan-Peregrino, Miriam", "Gil-Martin, Marta", "et al." ], "publisher": "Journal for ImmunoTherapy of Cancer", "date": "2022-03-01", "url": "https://pubmed.ncbi.nlm.nih.gov/35338084/", "doi": "10.1136/jitc-2021-003255", "pmid": "35338084", "trialId": "NCT02045602", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Primary intravenous phase 1 report: 16 solid-tumour participants and 26 PDAC participants across concurrent and prior-dose groups. High baseline Ad5-neutralising antibodies excluded 26% screened. Viral genomes were found in five of six day-eight tumour biopsies. Dose-limiting events included grade 4 AST rise, grade 4 febrile neutropenia and one grade 5 thrombocytopenia plus enterocolitis. The 50% PDAC response rates are uncontrolled." }, { "id": "geller-2017-bacterial-gemcitabine", "type": "paper", "title": "Potential role of intratumor bacteria in mediating tumor resistance to the chemotherapeutic drug gemcitabine", "authors": [ "Geller, Leore T.", "Barzily-Rokni, Michal", "Danino, Tal", "et al." ], "publisher": "Science", "date": "2017-09-15", "url": "https://www.science.org/doi/10.1126/science.aah5043", "doi": "10.1126/science.aah5043", "pmid": "28912244", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Mechanistic study. Bacterial long-form cytidine deaminase converted gemcitabine to inactive dFdU. Bacteria were detected in 86 of 113 human PDAC samples (76%), mainly Gammaproteobacteria. Ciprofloxacin reversed bacterial gemcitabine resistance in a mouse colon-cancer model; no person received randomized antibiotic treatment." }, { "id": "genmab-2026-q1-bizengri", "type": "sec-filing", "title": "Genmab interim report for the first quarter of 2026", "authors": [ "Genmab A/S" ], "publisher": "U.S. Securities and Exchange Commission", "date": "2026-05-07", "url": "https://ir.genmab.com/node/49531/html", "doi": null, "pmid": null, "trialId": "NCT02912949", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Genmab reports BIZENGRI as part of its royalty portfolio following the Merus acquisition. Partner Therapeutics commercializes the drug in the United States and pays Genmab royalties on US net sales. Product sales are not separated in this filing, so neither pancreatic revenue nor a product value can be derived." }, { "id": "ghaneh-2023-espac5", "type": "paper", "title": "Immediate surgery compared with short-course neoadjuvant gemcitabine plus capecitabine, FOLFIRINOX, or chemoradiotherapy in borderline resectable pancreatic cancer (ESPAC5)", "authors": [ "Ghaneh, Paula", "Palmer, Daniel H.", "Cicconi, Silvia", "et al." ], "publisher": "The Lancet Gastroenterology & Hepatology", "date": "2023-02-01", "url": "https://pubmed.ncbi.nlm.nih.gov/36521500/", "doi": "10.1016/S2468-1253(22)00348-X", "pmid": "36521500", "trialId": "ISRCTN89500674", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Open-label randomized phase 2 trial across 16 UK and German centres. Ninety were allocated and 86 entered ITT: 31 immediate surgery and 55 across three short neoadjuvant regimens. Resection was 68% versus 55%, P=.33; one-year OS was 39% for immediate surgery versus 78%, 84% and 60% for the three treatment-first arms, P=.0028. Only 478 screened produced 90 randomizations, follow-up was 12.2 months and the small treatment arms cannot rank regimens. Grade 3 or worse events were 7% versus 34% among assessed patients." }, { "id": "gismalla-2026-africa-pancreatic-review", "type": "paper", "title": "Clinical Presentation and Management Outcomes of Pancreatic Cancer in African Countries: A Scoping Review", "authors": [ "Gismalla, Mohamed Daffalla Awadalla", "Elhassan, Moawia Mohammed Ali", "Abd Elrahman, Ahmed Abd Elrahman Abdalla", "et al." ], "publisher": "Journal of Gastrointestinal Cancer", "date": "2026-05-13", "url": "https://pubmed.ncbi.nlm.nih.gov/42126786/", "doi": "10.1007/s12029-026-01490-2", "pmid": "42126786", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Scoping review of 20 studies published from 1995 to 2023, representing 26,850 patients. More than 70% of reported cases were stage III-IV, staging was missing for nearly 40% of the total cohort, complete treatment paths appeared in fewer than half of studies, resection was below 15% in most cohorts, and survival was rarely available. Research was concentrated in South Africa, Egypt and Morocco. The authors declared no competing interests." }, { "id": "glewis-2024-pacific-pgx", "type": "paper", "title": "PACIFIC-PGx pharmacogenetic-guided fluoropyrimidine and irinotecan dosing trial", "authors": [ "Glewis, Sharyn", "et al." ], "publisher": "Clinical and Translational Science", "date": "2024-12-01", "url": "https://doi.org/10.1111/cts.70083", "doi": "10.1111/cts.70083", "pmid": "39614408", "trialId": "ACTRN12621000251820", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Four-hospital prospective single-arm implementation study enrolled 462/487 eligible for DPYD but only 50/109 for UGT1A1; 60% in each enrolled group received the relevant drug. Results preceded cycle one for 96% in five to seven days. Among DPYD carriers, severe toxicity was 7% versus 39% in historical controls, odds ratio 0.11, 95% CI 0.01 to 0.97. UGT1A1 poor-metabolizer counts were only four." }, { "id": "gong-2025-antibiotics-adjuvant-pdac", "type": "paper", "title": "Antibiotic treatment and survival in patients with resected, early-stage pancreatic ductal adenocarcinoma receiving chemotherapy", "authors": [ "Gong, Emma", "Fulop, Daniel J.", "Serebrenik, Joyce", "et al." ], "publisher": "JNCI Cancer Spectrum", "date": "2025-03-03", "url": "https://pubmed.ncbi.nlm.nih.gov/39982394/", "doi": "10.1093/jncics/pkaf024", "pmid": "39982394", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "SEER-Medicare claims cohort n=712 with 143 propensity-matched pairs. Antibiotics within one month after adjuvant chemotherapy initiation were associated with median OS 23.9 versus 18.2 months; in the gemcitabine subgroup, 24.3 versus 20.0 months. Treatment was not assigned and the reason for antibiotic treatment was not a microbiome selector." }, { "id": "gonzalez-2025-cholecystectomy-chile", "type": "paper", "title": "Cholecystectomy and digestive cancer in Chile: Complementary results from interrupted time series and aggregated data analyses", "authors": [ "Gonzalez, Constanza", "García-Pérez, Alfonso", "Nervi, Bruno", "et al." ], "publisher": "International Journal of Cancer", "date": "2024-08-16", "url": "https://pubmed.ncbi.nlm.nih.gov/39150415/", "doi": "10.1002/ijc.35138", "pmid": "39150415", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Interrupted time-series and aggregate-data study of Chile's 2006 programme. Ecological association cannot establish individual causal benefit and differed by sex, geography and age." }, { "id": "gou-2026-shr3821", "type": "conference-paper", "title": "SHR-3821, an ADCC-enhanced CLDN18.2/4-1BB bispecific antibody, in advanced solid tumours", "authors": [ "Gou, Hongfeng", "Zheng, Li", "Xiao, Xiuying", "et al." ], "publisher": "Journal of Clinical Oncology", "date": "2026-05-27", "url": "https://ascopubs.org/doi/10.1200/JCO.2026.44.16_suppl.4037", "doi": "10.1200/JCO.2026.44.16_suppl.4037", "pmid": null, "trialId": "NCT06618651", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Primary phase 1 abstract. Forty people included 17 with pancreatic cancer. At the 15 and 20 mg/kg expansion doses, confirmed and unconfirmed PDAC ORR were both 0/11; disease control was 2/4 and 3/7. Across all tumour types, grade 3 or worse treatment-related events affected 37.5%, one person discontinued, and no treatment-related death occurred." }, { "id": "goyal-2023-futibatinib", "type": "paper", "title": "Futibatinib for FGFR2-Rearranged Intrahepatic Cholangiocarcinoma", "authors": [ "Goyal, Lipika", "Meric-Bernstam, Funda", "Hollebecque, Antoine", "et al." ], "publisher": "New England Journal of Medicine", "date": "2023-01-19", "url": "https://pubmed.ncbi.nlm.nih.gov/36652354/", "doi": "10.1056/NEJMoa2206834", "pmid": "36652354", "trialId": "NCT02052778", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Multinational single-arm phase 2 in 103 previously treated patients with FGFR2 fusion- or rearrangement-positive intrahepatic cholangiocarcinoma. Funded by Taiho." }, { "id": "grail-2025-10k-capital-history", "type": "company-filing", "title": "GRAIL Form 10-K for the year ended December 31, 2024", "authors": [ "GRAIL, Inc." ], "publisher": "U.S. Securities and Exchange Commission", "date": "2025-02-28", "url": "https://www.sec.gov/Archives/edgar/data/1699031/000169903125000041/gral-20241231.htm", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Primary company filing used to distinguish operating loss from acquisition accounting. Illumina acquired GRAIL in 2021 and spun it off in 2024. The filing reports $9.8bn post-acquisition cumulative net loss, including $6.8bn of goodwill and intangible impairment and $461.1m amortisation; these accounting charges are not cash R&D spend." }, { "id": "grail-2026-q2-10q", "type": "company-filing", "title": "GRAIL Form 10-Q for the quarter ended June 30, 2026", "authors": [ "GRAIL, Inc." ], "publisher": "U.S. Securities and Exchange Commission", "date": "2026-08-06", "url": "https://www.sec.gov/Archives/edgar/data/1699031/000162828026054307/gral-20260630.htm", "doi": null, "pmid": null, "trialId": "ISRCTN91431511", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Primary filing used for current capital, spending and the company's description of NHS-Galleri. At 30 June 2026, cash plus short-term securities was $861.6 million; first-half R&D was $95.45 million and net loss $203.4 million." }, { "id": "graversen-2017-pipac-pdac", "type": "paper", "title": "Peritoneal metastasis from pancreatic cancer treated with pressurized intraperitoneal aerosol chemotherapy (PIPAC)", "authors": [ "Graversen, Martin", "Detlefsen, Sonke", "Bjerregaard, Jon Kroll", "Pfeiffer, Per", "Mortensen, Michael Bau" ], "publisher": "Clinical & Experimental Metastasis", "date": "2017-05-17", "url": "https://doi.org/10.1007/s10585-017-9849-7", "doi": "10.1007/s10585-017-9849-7", "pmid": "28516306", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Five-person pancreatic case series with 16 cisplatin-doxorubicin PIPAC procedures. All five received more than one procedure and were therefore eligible for biopsy response analysis; four showed tissue regression and one stable disease. Median survival from peritoneal-metastasis diagnosis was 14 months. The selected, uncontrolled survivor cohort cannot estimate added life or the all-starter response rate." }, { "id": "greenhalf-2020-prss1-surveillance", "type": "guideline", "title": "International consensus guidelines on surveillance for pancreatic cancer in chronic pancreatitis", "authors": [ "Greenhalf, William", "Lévy, Philippe", "Gress, Thomas", "et al." ], "publisher": "Pancreatology", "date": "2020-07", "url": "https://pubmed.ncbi.nlm.nih.gov/32624419/", "doi": "10.1016/j.pan.2020.05.011", "pmid": "32624419", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "International GRADE consensus; surveillance supported for inherited PRSS1 pancreatitis in specialist centres, not for SPINK1 N34S alone." }, { "id": "grewal-2026-pancan-germline-survey", "type": "paper", "title": "Germline genetic testing among patients with pancreatic adenocarcinoma: A Pancreatic Cancer Action Network patient survey", "authors": [ "Grewal, Udhayvir S.", "Patel, Rishi R.", "Loeffler, Brittany T.", "et al." ], "publisher": "Cancer", "date": "2026-05-01", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC13166122/", "doi": "10.1002/cncr.70446", "pmid": "42118569", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Self-reported survey of 1,046 PanCAN respondents: 66.2% recalled an offer and 69.2% completed testing, including 63 people who found testing without an offer. Among 322 not tested, 66.8% said testing or counselling was never discussed or offered. Among 167 mutation-positive respondents, 103, 61.7%, reported testing in a first-degree relative. The 23.2% positive fraction among tested respondents shows strong selection and must not be treated as population prevalence." }, { "id": "grillo-2016-gepnet-grade-metastasis", "type": "journal-article", "title": "Grade Increases in Gastroenteropancreatic Neuroendocrine Tumor Metastases Compared to the Primary Tumor", "authors": [ "Grillo, Federica", "Albertelli, Manuela", "Brisigotti, Maria P.", "et al." ], "publisher": "Neuroendocrinology", "date": "2016", "url": "https://pubmed.ncbi.nlm.nih.gov/26337010/", "doi": "10.1159/000439434", "pmid": "26337010", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Retrospective tissue study of 60 people with multiple GEP-NET samples. Among 49 evaluable for site discordance, grade differed in 19; discordance was 10/12 with metachronous metastases. Mixed primary sites and selected tissue availability limit direct pNET estimates." }, { "id": "groarke-2024-ponsegromab-cachexia", "type": "paper", "title": "Ponsegromab for the Treatment of Cancer Cachexia", "authors": [ "Groarke, John D.", "Crawford, Jeffrey", "Collins, Susie M.", "et al." ], "publisher": "New England Journal of Medicine", "date": "2024-12-19", "url": "https://pubmed.ncbi.nlm.nih.gov/39282907/", "doi": "10.1056/NEJMoa2409515", "pmid": "39282907", "trialId": "NCT05546476", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Pfizer-funded randomized double-blind phase 2, n=187 with GDF-15 at least 1,500 pg/mL; 32% had pancreatic cancer. The 12-week primary endpoint was body weight. Placebo-adjusted gains were 1.22, 1.92 and 2.81 kg at 100, 200 and 400 mg. Appetite, symptoms and measured activity improved at 400 mg; survival and treatment delivery were not tested as primary outcomes." }, { "id": "grossman-2022-hope-organoids", "type": "paper", "title": "Organoid Sensitivity Correlates with Therapeutic Response in Patients with Pancreatic Cancer", "authors": [ "Grossman, Julie E.", "Muthuswamy, Lakshmi", "Huang, Ling", "Hidalgo, Manuel" ], "publisher": "Clinical Cancer Research", "date": "2022-02-15", "url": "https://pubmed.ncbi.nlm.nih.gov/34789479/", "doi": "10.1158/1078-0432.CCR-20-4116", "pmid": "34789479", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Primary prospective HOPE feasibility study. It enrolled 76 people, received 75 specimens and produced 13 drug-testable organoids from 12 people, 15.8% of the enrolled denominator. Initial testing took a median 96 days, range 35-265. Response correlation used only 12 organoids and did not assign treatment." }, { "id": "grotz-2022-crs-hipec-pdac-pilot", "type": "paper", "title": "Cytoreduction with Hyperthermic Intraperitoneal Chemoperfusion for Pancreatic Cancer with Low-Volume Peritoneal Metastasis: Results from a Prospective Pilot Study", "authors": [ "Grotz, Travis E", "Yonkus, Jennifer A", "Thiels, Cornelius A", "et al." ], "publisher": "Annals of Surgical Oncology", "date": "2022-08-16", "url": "https://doi.org/10.1245/s10434-022-12328-z", "doi": "10.1245/s10434-022-12328-z", "pmid": "35972667", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Prospective 18-person pilot in exceptionally selected PDAC: positive cytology or low-volume peritoneal metastasis, no disease elsewhere and an objective response to at least six months of systemic chemotherapy. Median prior chemotherapy exposure was 14 cycles and 16/18 also had chemoradiation. All underwent complete cytoreduction plus pancreatic resection, irreversible electroporation or intraoperative radiation. Eight of 18 had grade 3 or worse complications, seven were readmitted and one died within 30 days. Median progression-free and overall survival were 20 and 26 months without a control." }, { "id": "gudmundsdottir-2023-staging-laparoscopy", "type": "paper", "title": "Yield of Staging Laparoscopy for Pancreatic Cancer in the Modern Era: Analysis of More than 1,000 Consecutive Patients", "authors": [ "Gudmundsdottir, Hallbera", "Yonkus, Jennifer A", "Alva-Ruiz, Roberto", "et al." ], "publisher": "Journal of the American College of Surgeons", "date": "2023-04-07", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC10262988/", "doi": "10.1097/XCS.0000000000000704", "pmid": "37026837", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Single-centre retrospective analysis of 1,004 consecutive Mayo Clinic patients with scan-localized PDAC who underwent staging laparoscopy from 2017 to 2021. Hidden spread or malignant abdominal-wash cytology was found in 180, 18%; 140 had visible spread and 96 positive cytology, with overlap. Among 762 separate staged procedures, 9, 1.2%, had a complication and 1, 0.1%, a major complication. In 419 untreated people with wash sampling the positive yield was 23%. In people with no unclear lesion on imaging, yield rose from 1.6% with none of four risk factors to 42% with all four. Surgeon selection, one expert centre and no no-laparoscopy comparator prevent a patient-benefit estimate." }, { "id": "guggenberger-2023-neolap-ct", "type": "paper", "title": "Predictive value of computed tomography on surgical resectability in locally advanced pancreatic cancer treated with multiagent induction chemotherapy: Results from a prospective, multicentre phase 2 trial", "authors": [ "Guggenberger, K. V.", "Bley, T. A.", "Held, S.", "et al." ], "publisher": "European Journal of Radiology", "date": "2023-06-01", "url": "https://pubmed.ncbi.nlm.nih.gov/37080059/", "doi": "10.1016/j.ejrad.2023.110834", "pmid": "37080059", "trialId": "NCT02125136", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Exploratory central CT review of 83 NEOLAP participants. Formal anatomical downstaging occurred in 7.7%, while any tumour shrinkage occurred in 90.4%. Shrinkage predicted complete-margin resection weakly, AUC 0.647; the data-derived 22.5% cutoff had 65% sensitivity and 61% specificity. Persistent vessel contact after treatment therefore does not by itself prove an operation is impossible, but the analysis neither identifies survival benefit nor validates a universal cutoff." }, { "id": "guo-2026-postsurgical-ctdna", "type": "paper", "title": "Postsurgical circulating tumor DNA as a prognostic biomarker for relapse of resected pancreatic ductal adenocarcinoma", "authors": [ "Guo, Matthew Z.", "Sachidanand, Apoorvaa S.", "Nguyen, Thanh", "et al." ], "publisher": "Journal of Gastrointestinal Surgery", "date": "2026-04", "url": "https://pubmed.ncbi.nlm.nih.gov/41644010/", "doi": "10.1016/j.gassur.2026.102359", "pmid": "41644010", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Single-institution retrospective series of 51 resected PDAC patients, 28 with relapse. Postoperative-window ctDNA sensitivity was 35.7%, specificity 88.9%, positive predictive value 79.6% and negative predictive value 53.2%. Later adjuvant or surveillance testing had sensitivity 62.5%, specificity 95.5%, positive predictive value 94.4% and negative predictive value 67.7%. The authors concluded ctDNA could not replace conventional surveillance. Several authors disclosed industry relationships, but the abstract does not identify a ctDNA-test manufacturer as study sponsor." }, { "id": "gupta-2026-gaia-mil", "type": "paper", "title": "Multiple instance learning approach for automated gallbladder cancer detection using ultrasound imaging: multi-center validation of a deep learning model with the public dataset contribution", "authors": [ "Gupta, Pankaj", "Bose, Kartik", "Dutta, Usha" ], "publisher": "The Lancet Regional Health — Southeast Asia", "date": "2026-03-06", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC12992070/", "doi": "10.1016/j.lansea.2026.100737", "pmid": "41852453", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Full open-access report checked. AURORA-GB prospectively recruited consecutive patients with suspected gallbladder pathology at one Indian centre; three small external cohorts were retrospective. Pooled external validation n=122 gave sensitivity 78.2%, specificity 73.4% and AUC 0.778, with centre AUCs 0.699-0.950. External sensitivity fell to 69.4% for lesions below 10 mm. The model diagnoses selected visible pathology; it does not screen an unselected population or forecast future cancer. Code is public and AURORA-GB access is by request rather than unrestricted download." }, { "id": "guzauskas-2026-ttfields-cost", "type": "paper", "title": "Estimating the cost-effectiveness of tumour-treating fields for locally advanced pancreatic cancer patients", "authors": [ "Guzauskas, Gregory F.", "Picozzi, Vincent", "Niño de Rivera Guzman, Jorge Fernando", "Wang, Bruce C. M." ], "publisher": "Journal of Medical Economics", "date": "2026-07-28", "url": "https://pubmed.ncbi.nlm.nih.gov/42517235/", "doi": "10.1080/13696998.2026.2707046", "pmid": "42517235", "trialId": "NCT03377491", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Company-funded U.S. payer model using PANOVA-3 survival, distant progression and pain curves. It estimated $351,400 versus $220,000 lifetime cost, a $131,400 increment, 0.34 additional quality-adjusted life-years and $387,300 per added quality-adjusted life-year; scenarios ranged from $391,800 to $547,200. Two of four authors were Novocure employees and the first author disclosed paid Novocure consultancy. Modelled value is not an observed trial outcome." }, { "id": "haglund-1986-ca19-9", "type": "paper", "title": "Evaluation of CA 19-9 as a serum tumour marker in pancreatic cancer", "authors": [ "Haglund, C", "Roberts, P J", "Kuusela, P", "Scheinin, T M", "Mäkelä, O", "Jalanko, H" ], "publisher": "British Journal of Cancer", "date": "1986-02-01", "url": "https://pubmed.ncbi.nlm.nih.gov/3456787/", "doi": "10.1038/bjc.1986.35", "pmid": "3456787", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Primary early clinical study of 91 pancreatic-cancer patients and 111 patients with benign pancreatic, biliary or liver disease. CA19-9 exceeded 37 U/mL in 78% of cancers but also in 22% of benign diseases, especially cholestasis. It established utility and the specificity problem, not screening benefit." }, { "id": "haldar-2026-kras-interception", "type": "paper", "title": "First-in-Human Testing of a Mutant KRAS Vaccine for Pancreatic Cancer Interception in High-risk Cohorts", "authors": [ "Haldar, S. D.", "Huff, A. L.", "Wang, H. H.", "et al." ], "publisher": "Cancer Discovery", "date": "2026-08-26", "url": "https://aacrjournals.org/cancerdiscovery/article/doi/10.1158/2159-8290.CD-25-2245/787692/First-in-Human-Testing-of-a-Mutant-KRAS-Vaccine", "doi": "10.1158/2159-8290.CD-25-2245", "pmid": null, "trialId": "NCT05013216", "patentId": null, "grantId": "R37CA292056; K08CA248624", "accessed": "2026-09-14", "notes": "Single-arm phase 1 trial in 20 people with hereditary PDAC predisposition and a radiographic pancreatic abnormality. The endpoints were safety, immune response and T-cell persistence, not cancer incidence or mortality." }, { "id": "hammel-2016-lap07", "type": "paper", "title": "Effect of Chemoradiotherapy vs Chemotherapy on Survival in Patients With Locally Advanced Pancreatic Cancer Controlled After 4 Months of Gemcitabine With or Without Erlotinib", "authors": [ "Hammel, Pascal", "Huguet, Florence", "van Laethem, Jean-Luc", "et al." ], "publisher": "JAMA", "date": "2016-05-01", "url": "https://pubmed.ncbi.nlm.nih.gov/27139057/", "doi": "10.1001/jama.2016.4324", "pmid": "27139057", "trialId": "NCT00634725", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "International open-label phase 3 trial with two randomisations. The chemoradiotherapy comparison applied only to 269 patients whose disease was controlled after four months of induction chemotherapy." }, { "id": "hanada-2026-onomichi-method", "type": "journal-article", "title": "A community-hospital collaborative diagnostic pathway for earlier detection of pancreatic cancer: Long-term impact of the Onomichi Method", "authors": [ "Hanada, Keiji", "Shimizu, Akinori", "Tanaka, Shinji", "Sasaki, Nobutaka" ], "publisher": "Pancreatology", "date": "2026-06", "url": "https://pubmed.ncbi.nlm.nih.gov/41997755/", "doi": "10.1016/j.pan.2026.04.011", "pmid": "41997755", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Observational regional pathway from 2007-2023: 29,642 people assessed and 925 pancreatic cancers diagnosed. Stage 0/I rose from 13.9% to 22.1% between reported eras and observed five-year survival from 14.9% to 20.7%; within-stage survival did not change significantly. No concurrent control, so referral change, diagnostic intensity, lead time and era effects remain possible." }, { "id": "hao-2024-ibi389-pdac", "type": "conference-paper", "title": "IBI389 CLDN18.2/CD3 bispecific antibody in advanced pancreatic ductal adenocarcinoma", "authors": [ "Hao, Jihui", "Zheng, Li", "Dai, Ruihong", "et al." ], "publisher": "Journal of Clinical Oncology", "date": "2024-05-29", "url": "https://ascopubs.org/doi/10.1200/JCO.2024.42.16_suppl.4011", "doi": "10.1200/JCO.2024.42.16_suppl.4011", "pmid": null, "trialId": "NCT05164458", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Primary preliminary PDAC phase 1 abstract. Sixty-four previously treated patients received IBI389. At 600 micrograms/kg, 23 patients with CLDN18.2 staining of at least 2+ in 10% of cells were efficacy evaluable: seven responded, ORR 30.4% (95% CI 13.2-52.9), and DCR 69.6%. Grade 3 or worse treatment-related events affected 54.7%; cytokine-release syndrome affected 51.6%, with none grade 3 or worse in this cutoff." }, { "id": "hartlapp-2022-neolap-ca199", "type": "paper", "title": "Prognostic and predictive value of CA 19-9 in locally advanced pancreatic cancer treated with multiagent induction chemotherapy: results from a prospective, multicenter phase II trial (NEOLAP-AIO-PAK-0113)", "authors": [ "Hartlapp, I.", "Valta-Seufzer, D.", "Siveke, Jens T.", "et al." ], "publisher": "ESMO Open", "date": "2022-08-01", "url": "https://pubmed.ncbi.nlm.nih.gov/35970013/", "doi": "10.1016/j.esmoop.2022.100552", "pmid": "35970013", "trialId": "NCT02125136", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Exploratory NEOLAP analysis. Among 133 people evaluable at baseline, a fall in CA19-9 to at most 50 U/mL was associated with median survival 27.8 versus 16.5 months and with complete-margin resection. An outcome-derived threshold of 61 U/mL gave 72% sensitivity and 62% specificity for complete-margin resection; nonresponders had no such resection in this cohort. The marker predicts prognosis and resection result in selected data. It was not randomized as a surgery decision and cannot estimate benefit from surgery. Multiple authors disclosed industry relationships." }, { "id": "hatoum-2024-tamp-rr1-rr2", "type": "paper", "title": "Treatment of locally advanced pancreatic cancer using localized trans-arterial micro perfusion of gemcitabine: combined analysis of RR1 and RR2", "authors": [ "Hatoum, Hassan", "Rosemurgy, Alexander", "Bastidas, J. Augusto", "et al." ], "publisher": "The Oncologist", "date": "2024-07-24", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC11299923/", "doi": "10.1093/oncolo/oyae178", "pmid": "39049803", "trialId": "NCT02237157; NCT02591082", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Sponsor-funded pooled report of a 20-person phase 1 and 23-person observational registry. Forty-three enrolled, 38 were analysed and survival analyses used 35 stage III patients. Median survival from diagnosis was 12.6 months, 95% CI 2.1-54.2. Longer survival after prior chemoradiation was selected and uncontrolled. Safety was available for 23 people: 88 events, 17 grade 3 or worse, 19 procedure-related, sepsis in 4/23 and one ischemic pancreatobiliary-limb injury requiring surgery. No death occurred within 30 days of last treatment. RenovoRx funded the work and company-linked authors disclosed interests." }, { "id": "heersche-2026-ugt1a1-guided-dosing", "type": "paper", "title": "Real-world impact of UGT1A1 genotype-guided irinotecan dosing on severe toxicity and hospitalization", "authors": [ "Heersche, N", "et al." ], "publisher": "Journal of the National Comprehensive Cancer Network", "date": "2026-06-11", "url": "https://doi.org/10.6004/jnccn.2026.7011", "doi": "10.6004/jnccn.2026.7011", "pmid": "42276133", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Six Dutch hospitals reported 501 genotype-dosed patients, including 54 poor metabolizers started at 70% irinotecan. Severe toxicity was 29.6% in poor metabolizers versus 34.0% in normal or intermediate metabolizers. Nine poor metabolizers unintentionally given full dose had 77.8% severe toxicity versus 29.6% with reduction, P=0.009. The comparison was not randomized and was not pancreatic-specific." }, { "id": "heffley-2024-pdac-vte-uptake", "type": "paper", "title": "Venous thromboembolism in patients with pancreatic adenocarcinoma: disease burden and initiation of ambulatory thromboprophylaxis", "authors": [ "Heffley, J.", "Ganguly, E.", "Tompkins, B. J.", "et al." ], "publisher": "Pancreatology", "date": "2024-09-01", "url": "https://pubmed.ncbi.nlm.nih.gov/38960779/", "doi": "10.1016/j.pan.2024.06.008", "pmid": "38960779", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Single-centre cohort of 243 pancreatic-adenocarcinoma patients. VTE occurred in 24%; 52% of affected patients were hospitalized, 5% died from VTE and half of events occurred within two months of diagnosis. Only 30% received ambulatory prophylaxis and median start was 43 days. Major bleeding was 5.8%; intestinal tumour invasion predicted bleeding. Prescription-defined exposure and nonrandom treatment selection limit causal inference." }, { "id": "heidsma-2021-pandora-pnet", "type": "paper", "title": "Watchful waiting for small non-functional pancreatic neuroendocrine tumours: nationwide prospective cohort study (PANDORA)", "authors": [ "Heidsma, C. M.", "Engelsman, A. F.", "van Dieren, S." ], "publisher": "British Journal of Surgery", "date": "2021-08-19", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC10364894/", "doi": "10.1093/bjs/znab088", "pmid": "33783475", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Prospective Dutch nationwide cohort of 76 patients with non-functioning pNET no larger than 2 cm managed by watchful waiting. At 17 months, eight had protocol-defined progression above 0.5 cm per year, two chose resection and no metastases were reported. Quality of life was poorer than the reference population. Short follow-up and no randomized surgery comparator prevent a survival or lifetime-safety conclusion." }, { "id": "hendifar-2026-ai-chemo-pathology", "type": "paper", "title": "Development and validation of a computational histology artificial intelligence-powered predictive biomarker for selection of chemotherapy in advanced pancreatic cancer", "authors": [ "Hendifar, Andrew E.", "Krishna, Viswesh", "Krishna, Vrishab", "Zhang, Haochen", "Tarsode, Asit", "Nimgaonkar, Vivek", "Joshi, Anirudh R.", "Knox, Jennifer J." ], "publisher": "Journal of Clinical Oncology", "date": "2026-02-11", "url": "https://pubmed.ncbi.nlm.nih.gov/41671529/", "doi": "10.1200/JCO-25-02199", "pmid": "41671529", "trialId": "NCT02750657", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Primary model-development and locked external-validation paper: 178 development patients and 299 validation patients from COMPASS and Know Your Tumor. The biomarker-treatment interaction was significant for time to next treatment/death and overall survival, but treatment selection was observational, not randomized. Several authors were Valar Labs employees, shareholders or patent holders." }, { "id": "hengrui-2025-hk-prospectus", "type": "company-filing", "title": "Jiangsu Hengrui Pharmaceuticals — Hong Kong offer prospectus", "authors": [ "Jiangsu Hengrui Pharmaceuticals Co., Ltd." ], "publisher": "Jiangsu Hengrui Pharmaceuticals Co., Ltd.", "date": "2025", "url": "https://www.hengrui.com/images/investor/hk/2.%20Prospectus/2.%20Offer%20for%20Subscription%20(EN).pdf", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Company filing describes HRS-4642 as Hengrui's potent, long-acting, selective KRAS-G12D inhibitor in proprietary liposomal injectable form and says the liposome is designed for targeted delivery, controlled and sustained release and reduced systemic toxicity. These are company statements. The filing does not identify the controlling patent numbers or establish comparative clinical delivery benefit." }, { "id": "henn-2023-cyst-surgery-registry", "type": "paper", "title": "Surgical treatment for pancreatic cystic lesions—implications from the multi-center and prospective German StuDoQ|Pancreas registry", "authors": [ "Henn, Jan", "Wyzlic, Patrick K.", "Esposito, Irene", "et al." ], "publisher": "Langenbeck's Archives of Surgery", "date": "2023-01-14", "url": "https://pubmed.ncbi.nlm.nih.gov/36640188/", "doi": "10.1007/s00423-022-02740-0", "pmid": "36640188", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Prospective German surgical registry, n=1,137 mixed pancreatic cystic lesions, including 689 IPMN. Histological malignancy was 12.0%, severe morbidity 28.4%, and 30-day mortality 2.6%. The malignancy count does not include every lesion whose removal may prevent cancer, so it cannot by itself label the other operations unnecessary." }, { "id": "henricks-2018-dpyd-guided-dosing", "type": "paper", "title": "DPYD genotype-guided dose individualisation of fluoropyrimidine therapy in patients with cancer", "authors": [ "Henricks, Linda M", "Lunenburg, Carin A T C", "de Man, Femke M", "et al." ], "publisher": "The Lancet Oncology", "date": "2018-11-01", "url": "https://doi.org/10.1016/S1470-2045(18)30686-7", "doi": "10.1016/S1470-2045(18)30686-7", "pmid": "30348537", "trialId": "NCT02324452", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Prospective 17-hospital safety study enrolled 1,181 mixed-cancer patients; 1,103 were evaluable and 85 carried one of four DPYD variants. Genotype-based starting reductions lowered relative severe-toxicity risk against historical full-dose carrier controls for three variants, but the c.1236G>A comparison did not improve. This was not randomized and did not report a pancreatic-specific survival estimate." }, { "id": "henry-2023-complications-adjuvant", "type": "paper", "title": "Impact of complications after resection of pancreatic cancer on disease recurrence and survival, and mediation effect of adjuvant chemotherapy: nationwide, observational cohort study", "authors": [ "Henry, Anne Claire", "van Dongen, Jelle C", "van Goor, Iris W J M", "et al." ], "publisher": "BJS Open", "date": "2023-03-24", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC10036251/", "doi": "10.1093/bjsopen/zrac174", "pmid": "36959099", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Dutch audit cohort of 1,071 resections in 2014-2017. Major complications were associated with shorter disease-free interval, HR 1.22, and overall survival, HR 1.25; organ failure HRs were 1.86 and 1.89. Mediation analysis attributed much of the association to adjuvant chemotherapy, with a remaining direct association for organ failure. The study excluded death within 90 days and neoadjuvant treatment, collected some fields retrospectively and was not registered; mediation assumptions do not establish causality." }, { "id": "heumann-2022-azacitidine-pdac", "type": "paper", "title": "Randomized phase II trial of oral azacitidine in resected pancreatic adenocarcinoma at high risk for recurrence", "authors": [ "Heumann, Thatcher R.", "Baretti, Marina", "Sugar, Elizabeth A.", "et al." ], "publisher": "Clinical Epigenetics", "date": "2022-12-03", "url": "https://pubmed.ncbi.nlm.nih.gov/36463226/", "doi": "10.1186/s13148-022-01367-8", "pmid": "36463226", "trialId": "NCT01845805", "patentId": null, "grantId": "UL1 TR001863", "accessed": "2026-09-14", "notes": "Primary randomized study, n=49: imaging-confirmed median PFS 9.2 versus 8.9 months, HR 0.94, P=.85; OS HR 0.98, P=.96. Four of 24 stopped azacitidine for adverse events." }, { "id": "hewitt-2022-algenpantucel-l", "type": "paper", "title": "A Phase 3 Randomized Clinical Trial of Chemotherapy With or Without Algenpantucel-L Immunotherapy in Borderline Resectable or Locally Advanced Pancreatic Cancer", "authors": [ "Hewitt, et al." ], "publisher": "Annals of Surgery", "date": "2021-02-23", "url": "https://pubmed.ncbi.nlm.nih.gov/33630475/", "doi": "10.1097/SLA.0000000000004669", "pmid": "33630475", "trialId": "NCT01836432", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Open-label randomised phase 3 at 32 sites; 303 patients. NewLink Genetics and Lumos Pharma affiliations appeared among authors." }, { "id": "hidalgo-2022-palbociclib-nabpaclitaxel", "type": "paper", "title": "A preclinical and phase Ib study of palbociclib plus nab-paclitaxel in patients with metastatic adenocarcinoma of the pancreas", "authors": [ "Hidalgo, Manuel", "Garcia-Carbonero, Rocio", "Lim, Kian-Huat", "et al." ], "publisher": "Cancer Research Communications", "date": "2022-11-02", "url": "https://pubmed.ncbi.nlm.nih.gov/36970055/", "doi": "10.1158/2767-9764.CRC-22-0072", "pmid": "36970055", "trialId": "NCT02501902", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Primary phase 1b report, n=76, 80% previously treated for advanced disease. At the maximum tolerated dose, n=27, 12-month survival was 50%, 95% CI 29.9-67.2, below the prespecified 65% efficacy threshold. Neutropenia affected 76.3% overall. The model advantage did not clear the human efficacy gate." }, { "id": "hingorani-2020-halo301", "type": "paper", "title": "Randomized Phase III Trial of Pegvorhyaluronidase Alfa With Nab-Paclitaxel Plus Gemcitabine for Patients With Hyaluronan-High Metastatic Pancreatic Adenocarcinoma", "authors": [ "Van Cutsem, Eric", "Tempero, Margaret A.", "Sigal, Dana", "et al." ], "publisher": "Journal of Clinical Oncology", "date": "2020-07-24", "url": "https://pubmed.ncbi.nlm.nih.gov/32706635/", "doi": "10.1200/JCO.20.00590", "pmid": "32706635", "trialId": "NCT02715804", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Randomised double-blind phase 3 trial in 494 untreated patients selected for hyaluronan-high metastatic PDAC. Halozyme developed PEGPH20 and sponsored the programme." }, { "id": "hirode-2026-hcc-biomarker-rct", "type": "paper", "title": "Utility of Serum Biomarkers in Addition to Ultrasound for Early Detection of Hepatocellular Carcinoma in High-Risk Patients: A Randomized Controlled Trial", "authors": [ "Hirode, Grishma", "Zangneh, Hooman F", "Cerocchi, Orlando", "et al." ], "publisher": "Gastroenterology", "date": "2026-03-01", "url": "https://pubmed.ncbi.nlm.nih.gov/41556856/", "doi": "10.1053/j.gastro.2025.09.010", "pmid": "41556856", "trialId": "NCT02272504", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Randomized comparison in 1,208 adults with cirrhosis or high-risk HBV. Ultrasound plus AFP, AFP-L3 and DCP did not improve early-stage HCC detection versus ultrasound alone: HR 0.81, 95% CI 0.47-1.40. Two authors were affiliated with Fujifilm companies that make biomarker assays; the registry lists Wako Life Sciences as sponsor and describes a broader 2,500-person observational study." }, { "id": "hishinuma-2006-recurrence-autopsy", "type": "paper", "title": "Patterns of recurrence after curative resection of pancreatic cancer, based on autopsy findings", "authors": [ "Hishinuma, Shigeki", "Ogata, Yoshiaki", "Tomikawa, Masahiko", "et al." ], "publisher": "Journal of Gastrointestinal Surgery", "date": "2006-04-01", "url": "https://pubmed.ncbi.nlm.nih.gov/16627216/", "doi": "10.1016/j.gassur.2005.09.016", "pmid": "16627216", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Japanese single-centre autopsy series of 27 people after apparently curative resection. Small and selected, but distinguishes presence of local recurrence from judged direct cause of death." }, { "id": "ho-2025-cxcr4-pd1-pdac", "type": "paper", "title": "A phase 2 trial of CXCR4 antagonism and PD1 inhibition in metastatic pancreatic adenocarcinoma reveals recruitment of T cells but also immunosuppressive macrophages", "authors": [ "Ho, Won Jin", "et al." ], "publisher": "OncoImmunology", "date": "2025-09-14", "url": "https://pubmed.ncbi.nlm.nih.gov/40815609/", "doi": "10.1080/2162402X.2025.2543614", "pmid": "40815609", "trialId": "NCT04177810", "patentId": null, "grantId": "5P01CA247886", "accessed": "2026-09-14", "notes": "Investigator-led, single-centre phase 2 trial of plerixafor plus cemiplimab after progression in metastatic PDAC. The paper reports 21 treated patients: no objective responses and two best responses of stable disease, despite increased CD8 T-cell infiltration. Granulocytes and macrophages also increased; parallel models linked macrophage targeting, not granulocyte depletion, to better activity. The registry reports actual enrolment 25, so the difference between enrolled and analysed patients must be kept visible. Genzyme/Sanofi collaborated and supplied relevant drug access; the paper includes grants and disclosed relationships." }, { "id": "hochhaus-2017-imatinib-iris", "type": "paper", "title": "Long-term outcomes of imatinib treatment for chronic myeloid leukemia", "authors": [ "Hochhaus, Andreas", "Larson, Richard A.", "Guilhot, Francois", "et al." ], "publisher": "New England Journal of Medicine", "date": "2017-03-09", "url": "https://pubmed.ncbi.nlm.nih.gov/28273028/", "doi": "10.1056/NEJMoa1609324", "pmid": "28273028", "trialId": "NCT00006343; NCT00333840", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Primary long-term IRIS report: imatinib arm estimated 10-year overall survival 83.3%, complete cytogenetic response 82.8%; crossover prevents a clean 10-year randomized comparison. Funded by Novartis." }, { "id": "hopewell-2025-consort", "type": "paper", "title": "CONSORT 2025 statement: updated guideline for reporting randomised trials", "authors": [ "Hopewell, Sally", "Chan, An-Wen", "Collins, Gary S.", "et al." ], "publisher": "BMJ", "date": "2025-04-14", "url": "https://www.bmj.com/content/389/bmj-2024-081123", "doi": "10.1136/bmj-2024-081123", "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Current reporting standard for randomized trials. It requires the assigned groups, participant flow, prespecified outcomes, harms and analysis population to be visible, which supports the atlas distinction between random assignment and the selected people who completed treatment or measurement." }, { "id": "hopkins-goggins-profile-2026", "type": "institution", "title": "Michael G. Goggins — Johns Hopkins profile", "authors": [ "Johns Hopkins Medicine" ], "publisher": "Johns Hopkins Medicine", "date": null, "url": "https://profiles.hopkinsmedicine.org/provider/michael-g-goggins/2701678", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Live institutional profile checked for current titles and research focus in surveillance, inherited risk and genetic, epigenetic and protein biomarkers." }, { "id": "hopkins-vogelstein-profile-2026", "type": "institution", "title": "Bert Vogelstein, MD", "authors": [ "Johns Hopkins Medicine" ], "publisher": "Johns Hopkins Medicine", "date": null, "url": "https://profiles.hopkinsmedicine.org/provider/bert-vogelstein/2777618", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Live institutional profile checked for current roles and the laboratory's stated focus on colorectal cancer genetics, prevention and treatment." }, { "id": "horizon-cancer-mission-calls-2026", "type": "funder-guidance", "title": "Horizon Europe Cancer Mission calls 2026", "authors": [ "European Health and Digital Executive Agency" ], "publisher": "European Commission", "date": "2026-02-04", "url": "https://hadea.ec.europa.eu/calls-proposals/cancer-mission-calls-2026_en", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": "HORIZON-MISS-2026-02-CANCER", "accessed": "2026-09-14", "notes": "Primary call page. Single-stage 2026 Cancer Mission call opened 10 February and closes 15 September 2026 at 17:00 CEST. Seven fixed topics include virtual human-twin models, microbiome for early prediction and pragmatic immunotherapy trials for refractory cancers. A proposal must fit a named topic; this is not a general PDAC mission cheque." }, { "id": "htan2-pancreas-hta201", "type": "dataset-programme", "title": "HTAN2 pancreatic pre-cancer atlas", "authors": [ "Human Tumor Atlas Network" ], "publisher": "National Cancer Institute Human Tumor Atlas Network", "date": "2025-01-01", "url": "https://docs.humantumoratlas.org/overview/centers/", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": "1U01CA294548-01", "accessed": "2026-09-14", "notes": "Oregon Health & Science University pancreatic pre-cancer atlas, HTA201, in HTAN phase 2." }, { "id": "hu-2018-germline-risk", "type": "paper", "title": "Association Between Inherited Germline Mutations in Cancer Predisposition Genes and Risk of Pancreatic Cancer", "authors": [ "Hu, Chunling", "Hart, Steven N.", "Polley, Eric C.", "et al." ], "publisher": "JAMA", "date": "2018-06-01", "url": "https://pubmed.ncbi.nlm.nih.gov/29922827/", "doi": "10.1001/jama.2018.6228", "pmid": "29922827", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Case-control sequencing study of 3,030 Mayo Clinic registry cases against public exome reference controls. Cases were 95.6% non-Hispanic White; population transfer requires care." }, { "id": "hu-2019-crc-early-seeding", "type": "paper", "title": "Quantitative evidence for early metastatic seeding in colorectal cancer", "authors": [ "Hu, Zheng", "Ding, Jie", "Ma, Zhongming", "et al." ], "publisher": "Nature Genetics", "date": "2019-06-17", "url": "https://www.nature.com/articles/s41588-019-0423-x", "doi": "10.1038/s41588-019-0423-x", "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Exome sequencing of 118 biopsies from 23 patients with liver or brain metastases plus spatial modelling. Early seeding was inferred in 17 of 21 evaluable patients; timing remains model-derived." }, { "id": "hu-2020-icc-resection-recurrence", "type": "paper", "title": "Redefining conditional overall and disease-free survival after curative resection for intrahepatic cholangiocarcinoma: an international study of 1,221 patients", "authors": [ "Hu, Liang-Shuo", "Zhang, Xu-Feng", "Weiss, Matthew", "et al." ], "publisher": "Journal of Gastrointestinal Surgery", "date": "2019-12-10", "url": "https://pubmed.ncbi.nlm.nih.gov/31823320/", "doi": "10.1007/s11605-019-04472-x", "pmid": "31823320", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "International multi-institutional retrospective cohort of 1,221 curative-intent intrahepatic cholangiocarcinoma resections. Median overall survival was 36.8 months and 725 patients, 59.4%, recurred. Conditional five-year survival increased among people remaining recurrence-free, but selected survivor estimates are not a treatment effect." }, { "id": "huffman-2023-ly2880070-pdac", "type": "paper", "title": "A phase I expansion cohort study evaluating the CHK1 inhibitor LY2880070 with low-dose gemcitabine in metastatic pancreatic adenocarcinoma", "authors": [ "Huffman, Brandon M.", "Feng, Hanrong", "Parmar, Kalindi", "et al." ], "publisher": "Clinical Cancer Research", "date": "2023-12-15", "url": "https://pubmed.ncbi.nlm.nih.gov/37819936/", "doi": "10.1158/1078-0432.CCR-23-2005", "pmid": "37819936", "trialId": "NCT02632448", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Primary 11-person PDAC expansion report. No objective response occurred, all participants stopped by 3.2 months and four had drug-related grade 3 events. Organoids derived from two participant biopsies were strongly sensitive to the same combination, exposing a direct model-to-patient failure." }, { "id": "huffman-2024-azenosertib-pdac-design", "type": "meeting-abstract", "title": "Targeting replication stress in pancreatic cancer: a phase II trial of azenosertib with gemcitabine", "authors": [ "Huffman, Brandon", "Horick, Nora K.", "et al." ], "publisher": "Journal of Clinical Oncology", "date": "2024-05-29", "url": "https://www.asco.org/abstracts-presentations/238424", "doi": "10.1200/JCO.2024.42.16_suppl.TPS4211", "pmid": null, "trialId": "NCT06015659", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Primary trial-design abstract. Up to 12 safety lead-in participants were planned, followed by a Simon two-stage efficacy design testing six-month PFS of at least 44% against an unacceptable 24%. Paired tumour biopsies were included. The current registry now reports actual enrollment of only 12." }, { "id": "huggett-2014-verteporfin-pdt", "type": "paper", "title": "Phase I/II study of verteporfin photodynamic therapy in locally advanced pancreatic cancer", "authors": [ "Huggett, Mark T.", "Jermyn, Michael", "Gillams, Alice", "et al." ], "publisher": "British Journal of Cancer", "date": "2014-04-02", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC3974098/", "doi": "10.1038/bjc.2014.95", "pmid": "24569464", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Single-centre dose study in 15 selected people with inoperable locally advanced pancreatic cancer. Percutaneous verteporfin PDT produced a consistent 12 mm dead zone at 40 J but highly variable volume. Multiple fibres caused inflammation outside the pancreas because light leaked for about 1 cm behind their diffuser tips and overlapping fields raised the dose. Median survival after PDT was 8.8 months; mixed treatment before and after PDT and no control prevent a benefit claim." }, { "id": "iacobuzio-donahue-2009-autopsy-patterns", "type": "paper", "title": "DPC4 gene status of the primary carcinoma correlates with patterns of failure in patients with pancreatic cancer", "authors": [ "Iacobuzio-Donahue, Christine A.", "Fu, Baiyu", "Yachida, Shinichi", "et al." ], "publisher": "Journal of Clinical Oncology", "date": "2009-03-09", "url": "https://pubmed.ncbi.nlm.nih.gov/19273710/", "doi": "10.1200/JCO.2008.17.7188", "pmid": "19273710", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Rapid-autopsy series of 76 people who died with pancreatic cancer. End-stage sampling is unusually informative but selected by consent, referral and survival to autopsy; primary-tumour SMAD4 prediction was not prospectively validated here." }, { "id": "iarc-2012-liver-flukes", "type": "government-guidance", "title": "Opisthorchis viverrini and Clonorchis sinensis", "authors": [ "International Agency for Research on Cancer Working Group" ], "publisher": "International Agency for Research on Cancer", "date": "2012-01-01", "url": "https://www.ncbi.nlm.nih.gov/books/NBK304354/", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "IARC Monographs Volume 100B evaluation. The working group concluded that chronic Opisthorchis viverrini and Clonorchis sinensis infection causes cholangiocarcinoma and classified both as carcinogenic to humans, Group 1. This establishes a preventable causal exposure, not the size or timing of benefit from any specific control programme." }, { "id": "icgc-argo-2026-release14", "type": "dataset-release", "title": "ICGC ARGO Data Release 14", "authors": [ "International Cancer Genome Consortium ARGO" ], "publisher": "ICGC ARGO", "date": "2026-03-19", "url": "https://www.icgc-argo.org/news/42/release-14", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Cumulative release includes Australian Pancreatic Genome Initiative, Canadian pancreatic harmonized omics, SCRUM-Japan MONSTAR-SCREEN ctDNA/gut microbiome and other international programmes; most molecular data require controlled access." }, { "id": "icr-christina-yap-profile-2026", "type": "institution-profile", "title": "Professor Christina Yap", "authors": [ "Institute of Cancer Research" ], "publisher": "Institute of Cancer Research", "date": null, "url": "https://www.icr.ac.uk/research-and-discoveries/find-a-researcher/detail/professor-christina-yap", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Official profile checked for Yap's role in clinical-trial biostatistics and listed experience in early-phase, adaptive, stratified, basket, umbrella and platform designs. It does not establish a direct pancreatic programme, freedom from conflicts or availability." }, { "id": "idc-cptac-pda", "type": "dataset", "title": "CPTAC-PDA imaging collection", "authors": [ "National Cancer Institute Imaging Data Commons" ], "publisher": "NCI Imaging Data Commons", "date": "2026-09-14", "url": "https://portal.imaging.datacommons.cancer.gov/collections/", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Public CPTAC-PDA collection lists 195 subjects with CT, MR, PET, ultrasound, radiotherapy structures, clinical, genomic and proteomic links; annotation set lists 103 subjects." }, { "id": "igan-2026-adenocyte-financing", "type": "funding", "title": "iGan Partners Invests in Adenocyte Ltd to Advance the Early Detection of Pancreatic Cancer", "authors": [ "iGan Partners" ], "publisher": "iGan Partners", "date": "2026-05-26", "url": "https://iganpartners.com/blog/portfolio/igan-partners-invests-in-adenocyte-ltd-to-advance-the-early-detection-of-pancreatic-cancer", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Investor announcement says iGan led the first close of a $10m financing round and reports more than 150 procedures plus four unsuspected neoplasias among 90 high-risk people. It supplies the financing amount and sponsor claim only; no peer-reviewed denominator, protocol, adverse-event table or independent confirmation is linked." }, { "id": "igea-2026-cliniporator-vitae", "type": "company", "title": "Cliniporator VITAE technical sheet", "authors": [ "IGEA S.p.A." ], "publisher": "IGEA Medical", "date": "2026-01-01", "url": "https://www.igeamedical.com/design/igea/pdf/elettrochemioterapia/prodotti/cliniporator-vitae/Mod_VGP02tsEN_Rev.1.1_Technical_Sheet.pdf", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Manufacturer technical sheet. The generator delivers 100-microsecond high-voltage pulses, measures voltage and current waveforms and stores treatment data internally. Product specifications do not establish pancreatic patient benefit." }, { "id": "ihorst-2026-geriatric-comanagement", "type": "paper", "title": "Randomized geriatric co-management for vulnerable older patients with cancer", "authors": [ "Ihorst, Gabriele", "Jentschke, Elisabeth", "Tatschner, Kathrin", "et al." ], "publisher": "Journal of Geriatric Oncology", "date": "2026-03-01", "url": "https://pubmed.ncbi.nlm.nih.gov/41643478/", "doi": "10.1016/j.jgo.2026.102896", "pmid": "41643478", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Recent randomized mixed-cancer trial in 217 adults aged at least 70 with G8 below 15; 207 entered the full analysis. Co-management did not significantly improve the two primary outcomes, 12-week independence or quality of life. Grade 3 or worse toxicity fell from 30.8% to 15.5%, risk difference -15.2 points, 95% CI -26.5 to -3.9. Unplanned hospitalization was 21.4% versus 28.8%, with a confidence interval including no effect. No financial conflicts declared. PDAC-specific results were not reported." }, { "id": "imamura-2004-locally-invasive-surgery-rct", "type": "paper", "title": "A randomized multicenter trial comparing resection and radiochemotherapy for resectable locally invasive pancreatic cancer", "authors": [ "Imamura, Masayuki", "Doi, Ryuichiro", "Imaizumi, Toshihide", "et al." ], "publisher": "Surgery", "date": "2004-11-01", "url": "https://pubmed.ncbi.nlm.nih.gov/15523393/", "doi": null, "pmid": "15523393", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Small Japanese multicentre trial randomized 42 people during laparotomy to resection or radiochemotherapy after operative confirmation of locally invasive but resectable disease without major arterial involvement or distant spread. One operative death occurred; one-year survival was 62% versus 32% and the reported death HR was 0.46. It is the rare direct surgery comparison, but predates modern imaging and multiagent induction, enrolled only 42 people and does not represent today's initially unresectable locally advanced state." }, { "id": "imperial-hanna-profile-2026", "type": "institution-profile", "title": "Hanna Group: breath diagnostics and surgical research", "authors": [ "Imperial College London" ], "publisher": "Imperial College London", "date": null, "url": "https://www.imperial.ac.uk/department-surgery-cancer/research/surgery/groups/hanna-group/", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Official group profile checked for George Hanna's Imperial and NHS roles and the group's breath volatile-organic-compound, analytical, AI and health-economic capabilities. The page reports a small pancreatic case-control signal; VAPOR2, not that early result, is the relevant prospective test." }, { "id": "innovent-2026-interim-results", "type": "company-filing", "title": "Innovent 2026 interim results and business update", "authors": [ "Innovent Biologics, Inc." ], "publisher": "Innovent Biologics", "date": "2026-08-25", "url": "https://www.prnewswire.com/news-releases/innovent-announces-2026-interim-results-and-business-updates-302859305.html", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Issuer-authored current business update. First-half 2026 revenue was RMB8.6bn and IFRS profit about RMB1.3bn; cash reserves were RMB30.2bn at 31 July 2026. Innovent is not capital constrained for the existing IBI343 programme." }, { "id": "ioannou-2026-premium-protocol", "type": "paper", "title": "Practice changing RCT design and rationale: Abbreviated MRI plus AFP vs. ultrasound plus AFP for HCC surveillance in cirrhosis (PREMIUM study)", "authors": [ "Ioannou, George N", "Taddei, Tamar H", "Planeta, Beata M", "et al." ], "publisher": "JHEP Reports", "date": "2026-02-01", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC12857331/", "doi": "10.1016/j.jhepr.2025.101666", "pmid": "41624484", "trialId": "NCT05486572", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Protocol for a 4,700-person VA randomized comparison of abbreviated contrast MRI plus AFP against ultrasound plus AFP every six months for up to eight years. The primary endpoint is HCC mortality; the design targets at least a 35% relative reduction." }, { "id": "isrctn-14986389-lappie", "type": "trial-registry", "title": "LAP-PIE feasibility trial of percutaneous irreversible electroporation after chemotherapy", "authors": [ "University College London" ], "publisher": "ISRCTN Registry", "date": "2025-03-25", "url": "https://www.isrctn.com/ISRCTN14986389", "doi": "10.1186/ISRCTN14986389", "pmid": null, "trialId": "ISRCTN14986389", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Current primary registry. The UK randomized feasibility trial is marked completed and no longer recruiting, with first enrollment on 1 March 2024 and final enrollment on 30 November 2024. The public record has a protocol but no results or participant-level data. Planned enrollment was 50. Its feasibility targets include recruitment, technical success, retention, safety, patient acceptance and cost rather than a powered survival claim." }, { "id": "issels-2023-heat-rct", "type": "paper", "title": "Regional hyperthermia with cisplatin added to gemcitabine versus gemcitabine in patients with resected pancreatic ductal adenocarcinoma: The HEAT randomised clinical trial", "authors": [ "Issels, Rolf D.", "Boeck, Stefan", "Pelzer, Uwe", "et al." ], "publisher": "European Journal of Cancer", "date": "2023-03-01", "url": "https://pubmed.ncbi.nlm.nih.gov/36657324/", "doi": "10.1016/j.ejca.2022.12.009", "pmid": "36657324", "trialId": "NCT01077427", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Open-label randomized multicentre adjuvant trial after pancreatic cancer resection. It allocated 117 eligible participants and did not improve the primary disease-free-survival endpoint. The experimental arm changed regional heating, cisplatin and the gemcitabine schedule together, so it cannot isolate the effect of heat." }, { "id": "ius-2026-sonoporation-nct04821284-abstract", "type": "meeting-abstract", "title": "Sonoporation for Augmenting Pancreatic Cancer treatment: a Clinical Trial", "authors": [ "NCT04821284 investigators" ], "publisher": "2026 IEEE International Ultrasonics Symposium submission system", "date": "2026", "url": "https://epapers2.org/ius2026/ESR/paper_details.php?paper_id=7773", "doi": null, "pmid": null, "trialId": "NCT04821284", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Accepted lecture abstract reports 97 enrolled. Overall PFS and OS were not different, both p>0.36. A gemcitabine/nab-paclitaxel subgroup had median OS 407.5 versus 234 days, but subgroup denominators, uncertainty, interaction, prespecification and adjustment for multiple comparisons are not reported on the public page. Final proceedings paper was not yet available at access." }, { "id": "izzo-2021-laparoscopic-ect-protocol", "type": "paper", "title": "A Multicenter Randomized Controlled Prospective Study to Assess Efficacy of Laparoscopic Electrochemotherapy in Locally Advanced Pancreatic Cancer", "authors": [ "Izzo, Francesco", "Granata, Vincenza", "Fusco, Roberta", "et al." ], "publisher": "Journal of Clinical Medicine", "date": "2021-09-05", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC8432461/", "doi": "10.3390/jcm10174011", "pmid": "34501459", "trialId": "EudraCT 2018-003925-27", "patentId": null, "grantId": "F/050140/01-03/X32", "accessed": "2026-09-15", "notes": "Protocol for a 90-person randomized phase 2b comparison of laparoscopic bleomycin electrochemotherapy plus FOLFOXIRI versus FOLFOXIRI. Powered for objective response rising from 40% to 70%; survival and quality of life are secondary. Two authors worked for device maker IGEA and project funding included IGEA." }, { "id": "izzo-2021-pancreatic-ect-phase12", "type": "paper", "title": "Clinical Phase I/II Study: Local Disease Control and Survival in Locally Advanced Pancreatic Cancer Treated with Electrochemotherapy", "authors": [ "Izzo, Francesco", "Granata, Vincenza", "Fusco, Roberta", "et al." ], "publisher": "Journal of Clinical Medicine", "date": "2021-03-22", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC8005134/", "doi": "10.3390/jcm10061305", "pmid": "33810058", "trialId": "research registry 60", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Prospective uncontrolled study of 25 selected locally advanced cases accrued from 2011 to 2019. Eleven of 25 lacked six-month imaging because of early death. Fixed versus variable electrode geometry was surgeon-selected; survival medians were 6 versus 12 months, p=0.18." }, { "id": "ja-onomichi-hanada-profile-2026", "type": "institution-profile", "title": "Keiji Hanada and the Pancreatic and Biliary Disease Center", "authors": [ "JA Onomichi General Hospital" ], "publisher": "JA Onomichi General Hospital", "date": null, "url": "https://onomichi-gh.jp/subject/detail-b.php?id=136", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Official Japanese hospital profile checked for Hanada's current roles. It states that the local pancreatic early-diagnosis project began with community physicians in 2007 and expanded across Hiroshima Prefecture as Hi-PEACE in November 2022." }, { "id": "jacobio-2026-annual-results", "type": "company-results", "title": "Jacobio 2025 annual results and AstraZeneca JAB-23E73 collaboration", "authors": [ "Jacobio Pharmaceuticals" ], "publisher": "Jacobio Pharmaceuticals", "date": "2026-03-10", "url": "https://www.jacobiopharma.com/en/node/717", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Company-reported $2.015 billion total potential AstraZeneca collaboration value for JAB-23E73, including $100 million upfront; milestones are contingent, not cash received." }, { "id": "jacobio-2026-jab23e73-first-patient", "type": "company-announcement", "title": "First patient dosed in Chinese phase Ib/III JAB-23E73 pancreatic trial", "authors": [ "Jacobio Pharmaceuticals" ], "publisher": "Jacobio Pharmaceuticals", "date": "2026-08-01", "url": "https://www.jacobiopharma.com/en/news/JAB23E73_ag_cn_fpi", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Company-only operational update for oral pan-KRAS inhibitor plus gemcitabine and nab-paclitaxel; no pancreatic efficacy result." }, { "id": "jameson-2026-ketogenic-diet-pdac", "type": "paper", "title": "A randomized phase II trial of gemcitabine, nab-paclitaxel, cisplatin with or without a medically supervised ketogenic diet for patients with metastatic pancreatic cancer", "authors": [ "Jameson, Gayle S.", "Roe, Denise J.", "Borazanci, Erkut", "et al." ], "publisher": "Cancer", "date": "2026-03-15", "url": "https://pubmed.ncbi.nlm.nih.gov/41817106/", "doi": "10.1002/cncr.70343", "pmid": "41817106", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Randomized phase 2 screening trial with 32 evaluable people. A supervised ketogenic diet produced nutritional ketosis on a median 39.4% of days. PFS HR was 0.53, 95% CI 0.21-1.37, one-sided p=0.096; OS HR was 0.58, 95% CI 0.25-1.37, one-sided p=0.107. The wide intervals include no effect and the exploratory microbiome changes do not establish benefit. Feasible, not confirmatory." }, { "id": "jazz-2026-q2", "type": "company", "title": "Jazz Pharmaceuticals second-quarter 2026 financial results", "authors": [ "Jazz Pharmaceuticals" ], "publisher": "Jazz Pharmaceuticals", "date": "2026-08-03", "url": "https://investor.jazzpharma.com/news-releases/news-release-details/jazz-pharmaceuticals-announces-second-quarter-2026-financial", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Primary company result. Ziihera biliary-tract-cancer sales were $15m in Q2 2026. Company-wide 2026 GAAP R&D guidance was $818m–$873m and cannot be assigned to zanidatamab or biliary cancer." }, { "id": "jazz-zymeworks-2022-license", "type": "company", "title": "Jazz Pharmaceuticals and Zymeworks announce exclusive licence for zanidatamab", "authors": [ "Jazz Pharmaceuticals", "Zymeworks" ], "publisher": "Zymeworks", "date": "2022-10-19", "url": "https://ir.zymeworks.com/news-releases/news-release-details/jazz-pharmaceuticals-and-zymeworks-announce-exclusive-license", "doi": null, "pmid": null, "trialId": "NCT04466891", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Primary licensing announcement. Terms were $50m upfront, a later $325m opt-in payment, up to $525m regulatory and $862.5m commercial milestones, with total potential value up to $1.76bn plus royalties. Contingent milestones are not deployed research capital." }, { "id": "jin-2026-portal", "type": "paper", "title": "Robotic versus Open Pancreatoduodenectomy (PORTAL): multicentre, single masked, phase 3, non-inferiority randomised controlled trial", "authors": [ "Jin, Jiabin", "Qin, Kai", "Gemenetzis, Georgios", "et al." ], "publisher": "BMJ", "date": "2026-07-01", "url": "https://europepmc.org/article/MED/42386316", "doi": "10.1136/bmj-2026-319692", "pmid": "42386316", "trialId": "NCT04400357", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Seven high-volume Chinese-centre randomized trial of 268 adults with resectable pancreatic or periampullary disease; 254 entered modified intention-to-treat. Restricted mean time to functional recovery within 40 days was 12.1 robotic versus 16.0 open days, difference -3.9. Ninety-day mortality was 1/132 versus 3/122, while admission cost was about ¥22,834 higher with the robot. Credentialled surgeons and mixed diseases limit transport; long-term cancer control was not the primary result." }, { "id": "jonas-2025-sub-saharan-hcc-pathways", "type": "journal-article", "title": "IHPBA White Paper — The improvement of management pathways and access to care in sub-Saharan Africa for patients with hepatocellular carcinoma", "authors": [ "Jonas, Eduard", "Smith, Martin", "Kassianides, Chris", "Luyirika, Emmanuel", "Spearman, C Wendy" ], "publisher": "HPB", "date": "2025-03-04", "url": "https://pubmed.ncbi.nlm.nih.gov/40037975/", "doi": "10.1016/j.hpb.2025.02.007", "pmid": "40037975", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "IHPBA pathway paper led from UCT/Groote Schuur. It describes the linked need for viral-hepatitis control, screening, awareness, curative and life-prolonging treatment, palliation, infrastructure, policy, research and capacity. It is a programme paper, not evidence that these changes have been delivered or reduced mortality." }, { "id": "jrct-2026-enzeavour-feasibility", "type": "trial-registry", "title": "Pancreatic Cancer Screening Using Single-Molecule Enzyme Activity-Based Liquid Biopsy, Enzeavour: A Single-Arm Interventional Feasibility Study Protocol in Japan", "authors": [ "Hanada, Keiji" ], "publisher": "Japan Registry of Clinical Trials", "date": "2026-06-24", "url": "https://jrct.mhlw.go.jp/latest-detail/jRCT1050260052", "doi": null, "pmid": null, "trialId": "jRCT1050260052; NCT07605819; UMIN000059647", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Recruiting nationwide Japanese single-arm feasibility study in 10,000 asymptomatic adults undergoing routine checks; positive tests trigger MRI/MRCP, EUS or contrast CT." }, { "id": "jrct-kestose-pdac", "type": "registry", "title": "jRCTs041200109: evaluation of kestose in unresectable pancreatic cancer", "authors": [ "Fujita Health University Hospital" ], "publisher": "Japan Registry of Clinical Trials", "date": "2025-01-01", "url": "https://jrct.mhlw.go.jp/latest-detail/jRCTs041200109", "doi": null, "pmid": null, "trialId": "jRCTs041200109", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Completed phase 1-2 randomized open-label active-control trial, actual n=40. Primary outcome was improvement in blood tumour and nutrition markers at weeks 4, 8 and 12; tumour shrinkage was secondary. No statistical analysis plan was posted. Itochu Sugar supplied the kestose but the registry reports no research funding from it." }, { "id": "jusakul-2017-cholangiocarcinoma", "type": "paper", "title": "Whole-Genome and Epigenomic Landscapes of Etiologically Distinct Subtypes of Cholangiocarcinoma", "authors": [ "Jusakul, Apinya", "Cutcutache, Ioana", "Yong, Chern Han", "et al." ], "publisher": "Cancer Discovery", "date": "2017-06-30", "url": "https://pubmed.ncbi.nlm.nih.gov/28667006/", "doi": "10.1158/2159-8290.CD-17-0368", "pmid": "28667006", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "International genomic and epigenomic study with substantial Asian representation. Four groups differed by liver-fluke association and alterations including ERBB2, TP53, IDH1/2, BAP1 and FGFR-related rearrangements." }, { "id": "kaminski-2026-nordicc-13y", "type": "paper", "title": "Long-term effects of colonoscopy screening on colorectal cancer incidence and mortality: a multicountry, population-based randomised controlled trial", "authors": [ "Kaminski, Michal F.", "Kalager, Mette", "Løberg, Magnus", "et al." ], "publisher": "The Lancet", "date": "2026-05-05", "url": "https://pubmed.ncbi.nlm.nih.gov/42102826/", "doi": "10.1016/S0140-6736(26)00508-8", "pmid": "42102826", "trialId": "NCT00883792", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Thirteen-year NordICC update in 84,583 randomized people. Colorectal cancer occurred in 1.46% of the invited group and 1.80% of controls, RR 0.81. Colorectal-cancer mortality was 0.41% versus 0.47%, RR 0.88, 95% CI 0.68-1.08. The report concludes that one invitation reduced incidence but not mortality. An erratum was published under PMID 42119586." }, { "id": "kamphorst-2015-protein-scavenging", "type": "paper", "title": "Human pancreatic cancer tumors are nutrient poor and tumor cells actively scavenge extracellular protein", "authors": [ "Kamphorst, Jurre J.", "Nofal, Maryam", "Commisso, Cosimo", "et al." ], "publisher": "Cancer Research", "date": "2015-02-01", "url": "https://pubmed.ncbi.nlm.nih.gov/25644265/", "doi": "10.1158/0008-5472.CAN-14-2211", "pmid": "25644265", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Metabolomics of human PDAC and adjacent tissue plus mouse and cell experiments. Human specimens showed nutrient depletion and active macropinocytosis; dependence and treatment effect were not tested clinically." }, { "id": "kanda-2012-early-panin-mutations", "type": "paper", "title": "Presence of somatic mutations in most early-stage pancreatic intraepithelial neoplasia", "authors": [ "Kanda, Mitsuro", "Matthaei, Holger", "Wu, Jun", "et al." ], "publisher": "Gastroenterology", "date": "2012-01-05", "url": "https://pubmed.ncbi.nlm.nih.gov/22226782/", "doi": "10.1053/j.gastro.2011.12.042", "pmid": "22226782", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Microdissected early low-grade PanIN specimen study. Shows that driver mutations begin early; it does not estimate lesion prevalence, natural history or which mutation-bearing lesions progress." }, { "id": "kang-2025-laparotomic-cryoablation", "type": "paper", "title": "Safety and efficacy of cryoablation in treating locally advanced pancreatic cancer", "authors": [ "Kang, Li-Min", "He, Xiao-Long", "Lang, Li", "et al." ], "publisher": "World Journal of Gastrointestinal Oncology", "date": "2025-12-15", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC12754296/", "doi": "10.4251/wjgo.v17.i12.113387", "pmid": "41480225", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Retrospective open-surgery series of 24 locally advanced cases plus 24 separately selected chemotherapy controls. Median survival was 16.8 versus 11.4 months, p=0.021, but treatment was not assigned and systemic regimens varied. Eighteen of 24 had pancreatic fluid leakage, including four grade B fistulas; other harms included bleeding, delayed gastric emptying and pseudocysts." }, { "id": "kanno-2020-eus-cpn-modern-opioid-rct", "type": "paper", "title": "Efficacy of EUS-guided celiac plexus neurolysis compared with medication alone for unresectable pancreatic cancer in the oxycodone/fentanyl era: a prospective randomized control study", "authors": [ "Kanno, Yoshihide", "Koshita, Shinsuke", "Masu, Kaori", "et al." ], "publisher": "Gastrointestinal Endoscopy", "date": "2020-01-15", "url": "https://doi.org/10.1016/j.gie.2020.01.011", "doi": "10.1016/j.gie.2020.01.011", "pmid": "31953188", "trialId": "UMIN000037172", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Randomized study of EUS-guided celiac plexus neurolysis plus oxycodone or fentanyl versus the same medication strategy alone. Forty-eight registered; 24 intervention and 22 control participants were analysed. Four-week pain was 1.3 versus 2.3 on a 0-10 scale, p=0.10, with no significant pain difference at any measured time and no opioid or quality-of-life advantage. No severe procedure event was reported." }, { "id": "karam-2019-gpc1-nonreplication", "type": "paper", "title": "Glypican-1 and glycoprotein 2 bearing extracellular vesicles do not discern pancreatic cancer from benign pancreatic diseases", "authors": [ "Karam, et al." ], "publisher": "Oncotarget", "date": "2019-02-01", "url": "https://pubmed.ncbi.nlm.nih.gov/30800217/", "doi": "10.18632/oncotarget.26620", "pmid": "30800217", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Independent benign-disease comparison using nanoscale flow cytometry. GPC1-positive extracellular vesicles did not distinguish PDAC from benign pancreatic disease." }, { "id": "karasic-2019-hydroxychloroquine-pdac", "type": "paper", "title": "Gemcitabine and nab-paclitaxel with or without hydroxychloroquine in advanced pancreatic cancer", "authors": [ "Karasic, Thomas B.", "O'Hara, Mark H.", "Loaiza-Bonilla, Arturo", "et al." ], "publisher": "JAMA Oncology", "date": "2019-07-01", "url": "https://pubmed.ncbi.nlm.nih.gov/31120501/", "doi": "10.1001/jamaoncol.2019.0684", "pmid": "31120501", "trialId": "NCT01506973", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Primary randomized phase 2, n=112: one-year OS 41% with hydroxychloroquine versus 49% without; median OS 11.1 versus 12.1 months. Response increased while survival did not, and several grade 3/4 toxicities increased." }, { "id": "katz-2022-a021501", "type": "paper", "title": "Efficacy of Preoperative mFOLFIRINOX vs mFOLFIRINOX Plus Hypofractionated Radiotherapy for Borderline Resectable Adenocarcinoma of the Pancreas", "authors": [ "Katz, Matthew H. G.", "Shi, Qian", "Meyers, Jeff", "et al." ], "publisher": "JAMA Oncology", "date": "2022-09-01", "url": "https://pubmed.ncbi.nlm.nih.gov/35834226/", "doi": "10.1001/jamaoncol.2022.2319", "pmid": "35834226", "trialId": "NCT02839343", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Randomized phase 2 study with 126 registered participants. The radiotherapy arm closed after 10 of its first 30 evaluable patients had R0 resection, below the prespecified continuation rule." }, { "id": "katz-2026-gift-cascade", "type": "paper", "title": "Results From the Genetic Information and Family Testing Study: A Cluster-Randomized Trial", "authors": [ "Katz, Steven J.", "Hofer, Timothy P.", "Abrahamse, Paul", "et al." ], "publisher": "Journal of Clinical Oncology", "date": "2026-03-24", "url": "https://pubmed.ncbi.nlm.nih.gov/41875376/", "doi": "10.1200/JCO-25-02196", "pmid": "41875376", "trialId": "NCT05552664", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Population-based family-cluster randomized trial across hereditary cancer syndromes. Of 4,946 reported adult relatives, probands invited 948, 19.2%, and 303 enrolled. Free versus $50 testing increased the testing odds 2.5-fold but raised the mean family fraction tested by only 0.04 from a 0.03 baseline; a human navigator did not show a clear gain. This is transfer evidence, not a PDAC-only trial." }, { "id": "kawaguchi-2025-surf-hcc", "type": "paper", "title": "Surgery versus ablation for hepatocellular carcinoma: a randomized controlled trial and a nonrandomized prospective observational trial", "authors": [ "Kawaguchi, Yoshikuni", "Hasegawa, Kiyoshi", "Kashiwabara, Koji", "et al." ], "publisher": "Journal of Clinical Oncology", "date": "2025-06-24", "url": "https://pubmed.ncbi.nlm.nih.gov/40554738/", "doi": "10.1200/JCO-24-02030", "pmid": "40554738", "trialId": "UMIN000001795", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Final SURF report. Japanese multicentre randomized cohort n=302 plus prospective cohort n=753. For small HCC, surgery did not beat radiofrequency ablation for 5-year OS or recurrence-free survival; conflicts were extensively disclosed." }, { "id": "kawashiro-2018-jcros-carbon-ion", "type": "paper", "title": "Multi-institutional Study of Carbon-ion Radiotherapy for Locally Advanced Pancreatic Cancer: Japan Carbon-ion Radiation Oncology Study Group Study 1403 Pancreas", "authors": [ "Kawashiro, Shinichiro", "Yamada, Shigeru", "Okamoto, Masahiko", "et al." ], "publisher": "International Journal of Radiation Oncology, Biology, Physics", "date": "2018-08-01", "url": "https://pubmed.ncbi.nlm.nih.gov/29907490/", "doi": "10.1016/j.ijrobp.2018.04.057", "pmid": "29907490", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Retrospective three-centre Japanese series of 72 people with locally advanced pancreatic cancer. One- and two-year overall survival were 73% and 46%, while freedom from distant metastasis was 41% and 28%. There was no photon or no-radiation control." }, { "id": "khuntikeo-2020-thailand-cca-screening", "type": "paper", "title": "A comparison of the proportion of early stage cholangiocarcinoma found in an ultrasound-screening program compared to walk-in patients", "authors": [ "Khuntikeo, Narong", "et al." ], "publisher": "HPB", "date": "2020-06-01", "url": "https://pubmed.ncbi.nlm.nih.gov/31662222/", "doi": "10.1016/j.hpb.2019.10.010", "pmid": "31662222", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Thailand CASCAP comparison among 762 histologically proven cholangiocarcinomas: 161 detected through an endemic-area ultrasound programme and 601 presenting to hospital. Stage 0-II disease was 84.5% in screened versus 21.6% in walk-in cases after reported adjustment for age, sex and liver-fluke infection. Nonrandomized comparison, conditional on histologic diagnosis; no population mortality effect and incomplete false-positive burden." }, { "id": "kikuchi-2026-staging-laparoscopy", "type": "paper", "title": "The role of staging laparoscopy before therapy in patients with pancreatic ductal adenocarcinoma: a real-world, single-center experience", "authors": [ "Kikuchi, K", "et al." ], "publisher": "Frontiers in Medicine", "date": "2026-03-01", "url": "https://www.frontiersin.org/journals/medicine/articles/10.3389/fmed.2026.1776139/full", "doi": "10.3389/fmed.2026.1776139", "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Single-centre retrospective report of 102 PDAC staging laparoscopies before therapy. Hidden spread was found in 24, 23.5%; 14, 13.7%, had positive wash cytology. Mean procedure time was 70.2 minutes, major morbidity was 0, and chemotherapy began after a mean 14.8 days. There was no group without laparoscopy, so the study cannot show whether the procedure caused delay or improved survival." }, { "id": "kim-2015-egfr-expression-pdac", "type": "paper", "title": "EGFR expression in pancreatic intraepithelial neoplasia and ductal adenocarcinoma", "authors": [ "Kim, Min Joo", "Kim, Hyun Joo", "Kim, Seung A", "et al." ], "publisher": "Pancreas", "date": "2015-09-01", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC4555728/", "doi": null, "pmid": "26339400", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Tissue study of 81 pancreatic ductal adenocarcinomas plus normal pancreas and precursor lesions. EGFR overexpression appeared in 52/81 cancers, 64.2%, and rose across precursor grade. The study shows a possible cell-surface address in many but not all tumours; it does not measure ASP-1929 binding, light access or treatment response." }, { "id": "kim-2023-pancreatitis-risk-korea", "type": "paper", "title": "Incidence and risk of pancreatic cancer in patients with acute or chronic pancreatitis: a population-based cohort study", "authors": [ "Kim and collaborators" ], "publisher": "Scientific Reports", "date": "2023", "url": "https://pubmed.ncbi.nlm.nih.gov/37919344/", "doi": null, "pmid": "37919344", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Korean population data: 225,811 acute and 225,685 chronic pancreatitis patients matched with 4,514,960 controls; recurrent acute and chronic-with-acute groups had highest persistent risks." }, { "id": "kim-2024-irf6-immune-plasticity", "type": "paper", "title": "Plasticity-induced repression of Irf6 underlies acquired resistance to cancer immunotherapy in pancreatic ductal adenocarcinoma", "authors": [ "Kim, Il-Kyu", "Diamond, Mark S.", "Yuan, Salina", "et al." ], "publisher": "Nature Communications", "date": "2024-02-20", "url": "https://pubmed.ncbi.nlm.nih.gov/38378697/", "doi": "10.1038/s41467-024-46048-7", "pmid": "38378697", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Mouse PDAC relapse model. ZEB1 and SNAIL drove an epithelial-to-mesenchymal change and silenced Irf6, reducing tumour-cell sensitivity to TNF-mediated killing. The mechanism is specific to the tested model and has no patient intervention evidence." }, { "id": "kindler-2010-bevacizumab-calgb80303", "type": "paper", "title": "Gemcitabine plus bevacizumab compared with gemcitabine plus placebo in advanced pancreatic cancer: CALGB 80303", "authors": [ "Kindler, Hedy L.", "Niedzwiecki, Donna", "Hollis, Donna", "et al." ], "publisher": "Journal of Clinical Oncology", "date": "2010-07-06", "url": "https://pubmed.ncbi.nlm.nih.gov/20606091/", "doi": "10.1200/JCO.2010.28.1386", "pmid": "20606091", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Double-blind phase 3: 602 enrolled and 535 treated. Bevacizumab plus gemcitabine versus placebo plus gemcitabine gave median OS 5.8 versus 5.9 months, P=.95; PFS 3.8 versus 2.9 months, P=.07; response 13% versus 10%. Grade 3-4 hypertension was 10% versus 3% and proteinuria 5% versus 1%." }, { "id": "kindler-2011-axitinib-pdac", "type": "paper", "title": "Axitinib plus gemcitabine versus placebo plus gemcitabine in advanced pancreatic adenocarcinoma: a double-blind randomised phase 3 study", "authors": [ "Kindler, Hedy L.", "Ioka, Tatsuya", "Richel, Dirk J.", "et al." ], "publisher": "Lancet Oncology", "date": "2011-03", "url": "https://pubmed.ncbi.nlm.nih.gov/21306953/", "doi": "10.1016/S1470-2045(11)70004-3", "pmid": "21306953", "trialId": "NCT00471146", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Pfizer-funded double-blind randomized phase 3 n=632; stopped after crossing futility boundary." }, { "id": "kindler-2022-polo-overall-survival", "type": "paper", "title": "Overall Survival Results From the POLO Trial: A Phase III Study of Active Maintenance Olaparib Versus Placebo for Germline BRCA-Mutated Metastatic Pancreatic Cancer", "authors": [ "Kindler, Hedy L.", "Hammel, Pascal", "Reni, Michele", "et al." ], "publisher": "Journal of Clinical Oncology", "date": "2022-08-10", "url": "https://pubmed.ncbi.nlm.nih.gov/35834777/", "doi": "10.1200/JCO.21.01604", "pmid": "35834777", "trialId": "NCT02184195", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "154 patients with germline BRCA-mutated metastatic pancreatic adenocarcinoma whose disease had not progressed after at least 16 weeks of platinum chemotherapy. Industry involvement from AstraZeneca and Merck was disclosed." }, { "id": "klein-2024-ukb-prs-lifestyle", "type": "paper", "title": "Association of genetic risk and lifestyle with pancreatic cancer and their age dependency: a large prospective cohort study in the UK Biobank", "authors": [ "Klein and collaborators" ], "publisher": "International Journal of Epidemiology", "date": "2024", "url": "https://pubmed.ncbi.nlm.nih.gov/38066552/", "doi": null, "pmid": "38066552", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "UK Biobank prospective cohort n=340,631 with 1,129 incident pancreatic cancers over median 13.05 years; observational lifestyle and polygenic-risk analysis." }, { "id": "klein-haneveld-2021-men1-young-screening", "type": "paper", "title": "Initiating Pancreatic Neuroendocrine Tumor Screening in Young MEN1 Patients: Results From the DutchMEN Study Group", "authors": [ "Klein Haneveld, Mirthe J.", "van Treijen, Mark J. C.", "Pieterman, Carolina R. C." ], "publisher": "Journal of Clinical Endocrinology and Metabolism", "date": "2021-11-19", "url": "https://pubmed.ncbi.nlm.nih.gov/34333645/", "doi": "10.1210/clinem/dgab569", "pmid": "34333645", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Population-based DutchMEN imaging data from 350 people. Five developed a clinically relevant non-functioning pNET before age 18, defined as at least 20 mm or rapid growth; two later developed lymph-node metastases. Modelled ages at 1%, 2.5% and 5% risk were 9.5, 13.5 and 17.8 years. The study supports a surveillance-start discussion but does not compare schedules, quantify lifetime harm or show mortality reduction." }, { "id": "klotz-2024-europa", "type": "paper", "title": "Robotic versus open partial pancreatoduodenectomy (EUROPA): a randomised controlled stage 2b trial", "authors": [ "Klotz, Robert", "Mihaljevic, Andre L", "Kulu, Yakup", "et al." ], "publisher": "The Lancet Regional Health - Europe", "date": "2024-02-22", "url": "https://europepmc.org/articles/PMC10899052", "doi": "10.1016/j.lanepe.2024.100864", "pmid": "38420108", "trialId": "DRKS00020407", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Exploratory single very-high-volume-centre randomized trial. Of 81 randomized, 62 entered the modified analysis. Ninety-day cumulative complication scores were similar, but grade B/C pancreas-specific complications were 58.6% robotic versus 33.3% open, a 25.3-point difference, P=0.046. Robotic surgery took longer, cost more and converted to open in 23%; 90-day mortality was 4.8% in both groups. It mixed surgical indications and was funded by the German Federal Ministry of Education and Research." }, { "id": "knox-2025-pass01", "type": "paper", "title": "PASS-01: Randomized Phase II Trial of Modified FOLFIRINOX Versus Gemcitabine/Nab-Paclitaxel and Molecular Correlatives for Previously Untreated Metastatic Pancreatic Cancer", "authors": [ "Knox, Jennifer J.", "O'Kane, Grainne", "King, Daniel", "Tuveson, David A.", "Notta, Faiyaz", "Jaffee, Elizabeth M." ], "publisher": "Journal of Clinical Oncology", "date": "2025-11-01", "url": "https://pubmed.ncbi.nlm.nih.gov/40929627/", "doi": "10.1200/JCO-25-00436", "pmid": "40929627", "trialId": "NCT04469556", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Primary randomized phase 2 first-line trial. Of 160 randomized, 140 received treatment; genomes were generated in 94%, transcriptomes in 74% and organoids in 50%. Only 75 of 140 received second-line treatment and 33 of those 75 received a correlate-guided choice. Median time on second line was 2.1 months; median survival after a correlate-guided versus standard second-line choice was 5.4 versus 4.4 months, P=.45. The trial randomized chemotherapy, not organoid guidance." }, { "id": "koay-2014-human-pdac-drug-transport", "type": "paper", "title": "Transport properties of pancreatic cancer describe gemcitabine delivery and response", "authors": [ "Koay, Eugene J.", "Truty, Mark J.", "Cristini, Vittorio", "et al." ], "publisher": "Journal of Clinical Investigation", "date": "2014-03-10", "url": "https://pubmed.ncbi.nlm.nih.gov/24614108/", "doi": "10.1172/JCI73455", "pmid": "24614108", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Human study combining transport estimates from 176 pancreatic-protocol CT scans with intraoperative gemcitabine infusion in 12 patients. Tumour delivery varied widely and greater stroma tracked lower drug incorporation; small infusion cohort and observational outcome links limit treatment claims." }, { "id": "kollbeck-2025-complications-chemo", "type": "paper", "title": "Association of chemotherapy completion rates and overall survival with postoperative complications after pancreaticoduodenectomy for pancreatic ductal adenocarcinoma", "authors": [ "Kollbeck, Sebastian L G", "Hansen, Carsten P", "Dencker, Emil E", "et al." ], "publisher": "HPB", "date": "2024-11-09", "url": "https://europepmc.org/article/MED/39603870", "doi": "10.1016/j.hpb.2024.11.002", "pmid": "39603870", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Retrospective 2008-2022 cohort of 920 PDAC pancreaticoduodenectomies. Pancreatic leak and bile leak were associated with chemotherapy omission, OR 1.97 and 1.96. The study did not find delay beyond 90 days or an overall-survival difference. Long time span, treatment changes and residual confounding limit causal use." }, { "id": "kong-2022-pdac-organoid-drug-screen", "type": "paper", "title": "Drug screening and genome editing in human pancreatic cancer organoids identifies drug-gene interactions and candidates for off-label treatment", "authors": [ "Kong, Bo", "et al." ], "publisher": "Cell Genomics", "date": "2022-02-16", "url": "https://pubmed.ncbi.nlm.nih.gov/35187519/", "doi": null, "pmid": "35187519", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Primary PDAC organoid biobank and automated screen: 31 lines and 1,172 approved compounds, yielding laboratory hits. It is proof of screening capacity, not patient benefit or reliable treatment selection." }, { "id": "kordes-2015-metformin-pdac", "type": "paper", "title": "Metformin in patients with advanced pancreatic cancer: a double-blind randomized placebo-controlled phase 2 trial", "authors": [ "Kordes, Stefan", "Pollak, Michael N.", "Zwinderman, Aeilko H.", "et al." ], "publisher": "The Lancet Oncology", "date": "2015-07-01", "url": "https://pubmed.ncbi.nlm.nih.gov/26067687/", "doi": null, "pmid": "26067687", "trialId": "NCT01210911", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Primary double-blind randomized phase 2, n=121: median OS 6.8 months with metformin versus 7.6 with placebo, HR 1.056, P=.78." }, { "id": "korrel-2023-diploma", "type": "paper", "title": "Minimally invasive versus open distal pancreatectomy for resectable pancreatic cancer (DIPLOMA): an international randomised non-inferiority trial", "authors": [ "Korrel, Marieke", "Jones, Lucy R", "van Hilst, Jony", "et al." ], "publisher": "The Lancet Regional Health - Europe", "date": "2023-07-06", "url": "https://europepmc.org/articles/PMC10339208", "doi": "10.1016/j.lanepe.2023.100673", "pmid": "37457332", "trialId": "ISRCTN44897265", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Randomized trial of 258 patients at 35 centres in 12 countries with left-sided resectable pancreatic cancer. Modified intention-to-treat included 114 minimally invasive and 110 open resections. R0 was 73% versus 69%, difference 3.7%, 90% CI -6.2 to 13.6%, passing a -7% margin. Serious adverse events were 18% versus 22%; initial overall-survival HR 0.99. It combined laparoscopic and robotic methods and was funded by Medtronic Covidien, Johnson & Johnson Medical and the Dutch Gastroenterology Society." }, { "id": "koshiol-2026-chile-bils", "type": "paper", "title": "The Chile BiLS case-control study of gallbladder cancer", "authors": [ "Koshiol, Jill", "Van De Wyngard, Vanessa", "Araya, Juan Carlos" ], "publisher": "Archives of Public Health", "date": "2026-06-01", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC13463823/", "doi": "10.1186/s13690-026-01977-1", "pmid": "42251441", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Full open-access design report checked. Through August 2025 the Chilean high-risk study recruited 196 prevalent and 117 incident cancer or HGD cases, usually matched to gallstone cholecystectomy controls. It supplies detailed exposure, pathology, image and biospecimen data in a population with substantial Mapuche ancestry. Case-control enrichment is well suited to finding candidates but cannot estimate prospective positive predictive value or the benefit of cholecystectomy." }, { "id": "kudo-2026-emerald3", "type": "paper", "title": "Durvalumab and tremelimumab, with or without lenvatinib, combined with transarterial chemoembolisation in participants with embolisation-eligible hepatocellular carcinoma (EMERALD-3)", "authors": [ "Kudo, Masatoshi", "Abou-Alfa, Ghassan K.", "Ren, Zhenggang", "et al." ], "publisher": "The Lancet Oncology", "date": "2026-09-01", "url": "https://pubmed.ncbi.nlm.nih.gov/42636832/", "doi": "10.1016/S1470-2045(26)00278-0", "pmid": "42636832", "trialId": "NCT05301842", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Global sponsor-blinded open-label phase 3 trial. The full analysis included 760 people with embolisation-eligible HCC. Adding STRIDE immunotherapy and lenvatinib to TACE increased median progression-free survival from 9.8 to 13.0 months, HR 0.70, but the reported overall-survival comparison was 39.5 versus 34.7 months, HR 0.84, 95% CI 0.65-1.09, p=0.18. Serious adverse events occurred in 64% versus 23%, and treatment-related deaths in 2% versus 1%. AstraZeneca funded the trial. The result supports a control signal, not a survival or cure claim." }, { "id": "kulemann-2025-platelet-rna-pdac", "type": "paper", "title": "Transcriptomic profiling of blood platelets identifies a diagnostic signature for pancreatic cancer", "authors": [ "Kulemann and collaborators" ], "publisher": "BMC Medicine", "date": "2025", "url": "https://pubmed.ncbi.nlm.nih.gov/40133510/", "doi": null, "pmid": "40133510", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Multicentre case-control study of 673 platelet samples; two-RNA PLA2Sig tested across in-house and public cohorts." }, { "id": "kunzmann-2021-neolap", "type": "paper", "title": "Nab-paclitaxel plus gemcitabine versus nab-paclitaxel plus gemcitabine followed by FOLFIRINOX induction chemotherapy in locally advanced pancreatic cancer (NEOLAP-AIO-PAK-0113): a multicentre, randomised, phase 2 trial", "authors": [ "Kunzmann, Volker", "Siveke, Jens T.", "Algül, Hana", "et al." ], "publisher": "The Lancet Gastroenterology & Hepatology", "date": "2021-02-01", "url": "https://pubmed.ncbi.nlm.nih.gov/33338442/", "doi": "10.1016/S2468-1253(20)30330-7", "pmid": "33338442", "trialId": "NCT02125136", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Open-label randomized phase 2 trial at 28 German centres. Of 168 registered people, 130 without early progression were randomized between continued gemcitabine plus nab-paclitaxel and a switch to FOLFIRINOX. Complete macroscopic resection was 35.9% versus 43.9%, OR 0.72, 95% CI 0.35-1.45, and median survival was 18.5 versus 20.7 months, HR 0.86, 95% CI 0.55-1.36. It compares induction regimens, not surgery with no surgery; Celgene funded it." }, { "id": "labori-2016-adjuvant-completion", "type": "paper", "title": "Impact of early disease progression and surgical complications on adjuvant chemotherapy completion rates and survival in patients undergoing the surgery first approach for resectable pancreatic ductal adenocarcinoma", "authors": [ "Labori, Knut J", "Katz, Matthew H", "Tzeng, Ching-Wei", "et al." ], "publisher": "Acta Oncologica", "date": "2015-07-25", "url": "https://europepmc.org/article/MED/26213211", "doi": "10.3109/0284186x.2015.1068445", "pmid": "26213211", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Population-based retrospective surgery-first cohort of 203 resectable PDAC cases from 2006-2012 using an older Nordic fluorouracil regimen. Only 85, 41.9%, completed adjuvant chemotherapy; 41, 20.2%, started but did not complete and 77, 37.9%, never started. Main stated causes were progression 34.7%, complications or poor performance 32.2%, and age over 75 years 24.6%. Completion-survival associations are strongly selected and cannot estimate a treatment effect." }, { "id": "labori-2024-norpact1", "type": "paper", "title": "Neoadjuvant FOLFIRINOX versus upfront surgery for resectable pancreatic head cancer (NORPACT-1)", "authors": [ "Labori, Knut Jørgen", "Bratlie, Svein Olav", "Andersson, Bodil", "et al." ], "publisher": "The Lancet Gastroenterology & Hepatology", "date": "2024-03-01", "url": "https://pubmed.ncbi.nlm.nih.gov/38237621/", "doi": "10.1016/S2468-1253(23)00405-3", "pmid": "38237621", "trialId": "NCT02919787", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Multicentre open-label randomized phase 2 trial in 12 Nordic hospitals. Seventy-seven patients were allocated to four cycles of neoadjuvant FOLFIRINOX and 63 to upfront surgery. At 18 months, 60% versus 73% were alive; median OS was 25.1 versus 38.5 months, HR 1.52, 95% CI 1.00-2.33. Only 61/77, 79%, received neoadjuvant treatment; 17/77, 22%, were excluded from that per-protocol set. Grade 3 or worse adverse events affected 58% versus 40%. Resection was 82% versus 89%. The trial changed adjuvant regimens during recruitment, was small and was not powered as definitive phase 3 evidence." }, { "id": "lai-2026-carbon-ion-risk-adapted", "type": "paper", "title": "Carbon ion therapy for pancreatic cancer with risk-adapted dose escalation: initial clinical experience", "authors": [ "Lai, I-Chun", "Kang, Yu-Mei", "Wu, Yu-Hsuan", "et al." ], "publisher": "Radiotherapy and Oncology", "date": "2026-01-01", "url": "https://pubmed.ncbi.nlm.nih.gov/41176067/", "doi": "10.1016/j.radonc.2025.111238", "pmid": "41176067", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Retrospective first-centre series from Taiwan. Eighty-four people, mainly with locally advanced or oligometastatic disease, received induction chemotherapy then 12 carbon-ion fractions. One-year local control was 94.7%, overall survival 75.3%, distant-metastasis-free survival 44.5% and progression-free survival 42.9%." }, { "id": "laquente-2017-chk1-pdac", "type": "paper", "title": "A phase II study to evaluate LY2603618 in combination with gemcitabine in pancreatic cancer patients", "authors": [ "Laquente, Berta", "Lopez-Martin, Jose", "Richards, Donald", "et al." ], "publisher": "BMC Cancer", "date": "2017-02-15", "url": "https://pubmed.ncbi.nlm.nih.gov/28202004/", "doi": "10.1186/s12885-017-3131-x", "pmid": "28202004", "trialId": "NCT00839332", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Primary randomized phase 2 CHK1 report. Ninety-nine people were randomized 65:34 to LY2603618 plus gemcitabine or gemcitabine. Median OS was 7.8 versus 8.3 months; the posterior superiority probability was 0.3 versus the 0.8 gate. PFS, response and response duration did not improve despite reaching the planned drug exposure." }, { "id": "latenstein-2021-centralization", "type": "paper", "title": "Effect of centralization and regionalization of pancreatic surgery on resection rates and survival", "authors": [ "Latenstein, Anouk E J", "Mackay, Tara M", "van der Geest, Lydia G M", "et al." ], "publisher": "British Journal of Surgery", "date": "2021-07-01", "url": "https://europepmc.org/article/MED/33738473", "doi": "10.1093/bjs/znaa146", "pmid": "33738473", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Nationwide Dutch observational cohort of 10,079 non-metastatic pancreatic-head or periampullary cancers diagnosed in 2009-2017. Resection rose from 40.0% to 47.8%; by 2015-2017 the odds of resection were similar whether diagnosis began at a pancreatic-surgery centre or elsewhere, while survival remained slightly higher when diagnosis began at a surgery centre, adjusted HR 0.92. Mixed cancers and residual referral, staging and era confounding prevent a causal estimate." }, { "id": "lattimore-2023-bd-ipmn-surveillance", "type": "paper", "title": "Long-term surveillance of branch-duct intraductal papillary mucinous neoplasms without worrisome or high-risk features", "authors": [ "Lattimore, Catherine M.", "Kane, William J.", "Subbarao, Sandhya", "et al." ], "publisher": "Journal of Surgical Oncology", "date": "2023-08-02", "url": "https://pubmed.ncbi.nlm.nih.gov/37530526/", "doi": "10.1002/jso.27414", "pmid": "37530526", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Prospective surveillance cohort of 267 branch-duct IPMN without baseline worrisome or high-risk features. Twenty-six were resected; five had adenocarcinoma and one carcinoma in situ. Whole-cohort pancreatic adenocarcinoma mortality was 1.1% over median 44.5 months." }, { "id": "lau-2024-dendritic-resected-pdac", "type": "paper", "title": "Dendritic Cell-Based Immunotherapy in Patients With Resected Pancreatic Cancer", "authors": [ "Lau and collaborators" ], "publisher": "Journal of Clinical Oncology", "date": "2024", "url": "https://pubmed.ncbi.nlm.nih.gov/38950309/", "doi": null, "pmid": "38950309", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Single-arm adjuvant dendritic-cell vaccination study n=38 after surgery and standard therapy; estimated two-year recurrence-free survival 64%." }, { "id": "lawrence-2024-celiac-radiosurgery-phase2", "type": "paper", "title": "Celiac plexus radiosurgery for pain management in advanced cancer: a multicentre, single-arm, phase 2 trial", "authors": [ "Lawrence, Yaacov R", "Miszczyk, Marcin", "Dawson, Laura A", "et al." ], "publisher": "The Lancet Oncology", "date": "2024-07-16", "url": "https://doi.org/10.1016/S1470-2045(24)00223-7", "doi": "10.1016/S1470-2045(24)00223-7", "pmid": "39029483", "trialId": "NCT03323489", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Eight-centre single-arm study of one 25 Gy radiation dose to the celiac plexus. Of 125 treated people, 90 were evaluable at three weeks; 83/90 had pancreatic cancer. Forty-eight of 90 evaluable people had at least a two-point pain response, but that is 48/125, 38.4%, in the treated denominator. Eleven serious grade 3 or worse events were at least possibly related, including two grade-5 events judged possibly related. A randomized comparison was recommended." }, { "id": "le-2015-gvax-crs207", "type": "paper", "title": "Safety and survival with GVAX pancreas prime and Listeria monocytogenes-expressing mesothelin (CRS-207) boost vaccines for metastatic pancreatic cancer", "authors": [ "Le, Dung T.", "Wang-Gillam, Andrea", "Picozzi, Vincent", "et al." ], "publisher": "Journal of Clinical Oncology", "date": "2015-01-12", "url": "https://pubmed.ncbi.nlm.nih.gov/25584002/", "doi": "10.1200/JCO.2014.57.4244", "pmid": "25584002", "trialId": "NCT01417000", "patentId": null, "grantId": "P50 CA062924", "accessed": "2026-09-15", "notes": "Randomized phase 2 in previously treated metastatic PDAC. Ninety people received treatment: 61 received cyclophosphamide/GVAX followed by CRS-207 and 29 received cyclophosphamide/GVAX. Median OS was 6.1 versus 3.9 months, HR 0.59, P=0.02. The later 303-person ECLIPSE study did not reproduce benefit against chemotherapy." }, { "id": "le-2019-eclipse", "type": "paper", "title": "Results from a Phase IIb, Randomized, Multicenter Study of GVAX Pancreas and CRS-207 Compared with Chemotherapy in Adults with Previously Treated Metastatic Pancreatic Adenocarcinoma (ECLIPSE Study)", "authors": [ "Le, Dung T.", "Picozzi, Vincent J.", "Ko, Andrew H.", "et al." ], "publisher": "Clinical Cancer Research", "date": "2019-05-24", "url": "https://pubmed.ncbi.nlm.nih.gov/31126960/", "doi": "10.1158/1078-0432.CCR-18-2992", "pmid": "31126960", "trialId": "NCT02004262", "patentId": null, "grantId": "P50 CA062924", "accessed": "2026-09-15", "notes": "Three-arm randomized phase 2b trial in 303 previously treated metastatic-PDAC patients. In the prespecified 213-person later-line cohort, median OS was 3.7 months with cyclophosphamide/GVAX plus CRS-207, 5.4 with CRS-207 alone and 4.6 with chemotherapy. The combination-versus-chemotherapy HR was 1.17, 95% CI 0.84-1.64; the primary efficacy endpoint was missed." }, { "id": "lee-2024-dynamic-pancreas-feasibility", "type": "meeting-abstract", "title": "The potential role of serial circulating tumor DNA testing after upfront surgery to guide adjuvant chemotherapy for early stage pancreatic cancer: the AGITG DYNAMIC-Pancreas trial", "authors": [ "Lee, Belinda", "Tie, Jeanne", "Wang, Yuxuan", "et al." ], "publisher": "Journal of Clinical Oncology", "date": "2024-05-29", "url": "https://ascopubs.org/doi/10.1200/JCO.2024.42.16_suppl.107", "doi": "10.1200/JCO.2024.42.16_suppl.107", "pmid": null, "trialId": "ACTRN12618000335291", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Prospective 26-centre feasibility cohort after upfront resection. Of 102 enrolled, 40 were postoperative ctDNA-positive and 54 negative; 24 of 54 negative patients were de-escalated to three months of adjuvant chemotherapy at clinician discretion. Median recurrence-free survival was 13 months for ctDNA-positive and 22 months for ctDNA-negative patients. This cohort showed feasibility and prognosis, not a randomized treatment effect. The registry's randomized comparison applies to a separate neoadjuvant-treatment cohort. Meeting abstract only; survival analyses were ongoing. Marcus Foundation sponsored the study." }, { "id": "lee-2024-prospective-fapi-resectable-pdac", "type": "paper", "title": "Prospective Comparison of 18F-FDG and 18F-AlF-FAPI-74 PET/CT in the Evaluation of Potentially Resectable Pancreatic Ductal Adenocarcinoma", "authors": [ "Lee, Sang Hyub", "et al." ], "publisher": "Molecular Imaging and Biology", "date": "2024-10-05", "url": "https://pubmed.ncbi.nlm.nih.gov/39365411/", "doi": null, "pmid": "39365411", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Prospective head-to-head imaging study in 20 people with potentially resectable PDAC; 17 underwent curative-intent surgery. FAPI lymph-node sensitivity was 50% versus 0% for FDG." }, { "id": "lee-2026-diabetes-progression-korea", "type": "paper", "title": "Impact of initial severity and progression pattern of new-onset diabetes on pancreatic cancer risk: a 15-year longitudinal nationwide cohort study", "authors": [ "Lee, Minyoung", "Yu, Min Heui", "Chae, Wonjeong", "et al." ], "publisher": "Gut", "date": "2026-07-24", "url": "https://pubmed.ncbi.nlm.nih.gov/42498621/", "doi": "10.1136/gutjnl-2025-335863", "pmid": "42498621", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "South Korean national insurance cohort of 402,663 people. Treatment intensity and rapid escalation are proxies and may be influenced by healthcare practice; observational associations do not prove screening benefit." }, { "id": "leidner-2022-kras-tcr", "type": "paper", "title": "Neoantigen T-Cell Receptor Gene Therapy in Pancreatic Cancer", "authors": [ "Leidner, Rom", "Sanjuan Silva, Nelson", "Huang, Huayu", "et al." ], "publisher": "New England Journal of Medicine", "date": "2022-06-02", "url": "https://pubmed.ncbi.nlm.nih.gov/35648703/", "doi": "10.1056/NEJMoa2119662", "pmid": "35648703", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Detailed report of one responding patient plus discussion of a second pancreatic-cancer patient treated with the same HLA-C*08:02-restricted KRAS-G12D TCRs who progressed and died at six months." }, { "id": "lemanska-2023-opensafely-pert-audit", "type": "paper", "title": "A National Audit of Pancreatic Enzyme Prescribing in Pancreatic Cancer from 2015 to 2023 in England Using OpenSAFELY-TPP", "authors": [ "Lemanska, Agnieszka", "Andrews, Colm", "Fisher, Louis", "et al." ], "publisher": "Seminars in Oncology Nursing", "date": "2023-06-01", "url": "https://pubmed.ncbi.nlm.nih.gov/37142468/", "doi": "10.1016/j.soncn.2023.151439", "pmid": "37142468", "trialId": null, "patentId": null, "grantId": "MR/W016729/1", "accessed": "2026-09-14", "notes": "OpenSAFELY audit of 22,860 people with pancreatic cancer. Prescribing rose about one point per year, from 41% in 2015 to 48% in early 2023, below the stated 100% standard, with regional rates up to roughly 50-60%. It measures prescribing, not receipt, adherence, symptoms or survival." }, { "id": "lemmens-2011-centralization", "type": "paper", "title": "Improving outcome for patients with pancreatic cancer through centralization", "authors": [ "Lemmens, Valery E", "Bosscha, Koop", "van der Schelling, George", "et al." ], "publisher": "British Journal of Surgery", "date": "2011-06-29", "url": "https://europepmc.org/article/MED/21717423", "doi": "10.1002/bjs.7581", "pmid": "21717423", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Regional Dutch comparison of 2,129 pancreatic-head, extrahepatic-bile-duct, ampullary and duodenal cancers before and during centralization. Resection rose from 19.0% to 30.0%, operating hospitals fell from eight to three, in-hospital mortality fell from 24.4% to 3.6%, and two-year survival after surgery rose from 38.1% to 49.4%. The eras, case mix and wider care changed together, and the cohort is not PDAC-only." }, { "id": "lennon-2020-detect-a", "type": "paper", "title": "Feasibility of blood testing combined with PET-CT to screen for cancer and guide intervention", "authors": [ "Lennon, Anne Marie", "Buchanan, Amanda H.", "Kinde, Isaac", "et al." ], "publisher": "Science", "date": "2020-04-28", "url": "https://pubmed.ncbi.nlm.nih.gov/32345712/", "doi": "10.1126/science.abb9601", "pmid": "32345712", "trialId": "NCT04213326", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Prospective interventional study of 10,006 women aged 65 to 75 using a multicancer blood test followed by confirmatory PET-CT. It tested workflow and safety, not cancer-specific mortality." }, { "id": "lentz-2025-ebc129-pdac", "type": "conference-paper", "title": "Clinical activity of the CEACAM5/6 antibody-drug conjugate EBC-129 in pancreatic ductal adenocarcinoma", "authors": [ "Lentz, Robert William", "Ng, Matthew C. H.", "Yong, Wei-Peng", "et al." ], "publisher": "Journal of Clinical Oncology", "date": "2025-05-28", "url": "https://ascopubs.org/doi/10.1200/JCO.2025.43.16_suppl.4018", "doi": "10.1200/JCO.2025.43.16_suppl.4018", "pmid": null, "trialId": "NCT05701527", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Primary 21-person selected PDAC phase 1 abstract. Overall ORR was 19.0% and unconfirmed, median PFS 12 weeks, and no relation appeared between IHC score and response. At 2.2 mg/kg, grade 3 or worse neutropenia affected 81.8% and anemia 18.2%; infusion reactions affected 57% overall." }, { "id": "lesavage-2024-engineered-matrix-chemoresistance", "type": "paper", "title": "Engineered matrices reveal stiffness-mediated chemoresistance in patient-derived pancreatic cancer organoids", "authors": [ "LeSavage, Bauer L.", "Zhang, Daiyao", "Huerta-Lopez, Carla", "et al." ], "publisher": "Nature Materials", "date": "2024-08-01", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC13098013/", "doi": "10.1038/s41563-024-01908-x", "pmid": "38965405", "trialId": null, "patentId": null, "grantId": "DP1CA238296", "accessed": "2026-09-15", "notes": "Engineered matrices matched to PDAC stiffness induced reversible chemotherapy resistance in organoids from three patients through CD44-hyaluronan signalling and drug-efflux transporters. The small model study shows that mechanics can change cell state; it does not show that softening a human tumour improves drug delivery or survival." }, { "id": "levink-2025-pacyfic", "type": "paper", "title": "Small and Stable Pancreatic Cysts Are Reassuring During Surveillance: Results From the PACYFIC Trial", "authors": [ "Levink, Iris J. M.", "Sprij, Marc L. J. A.", "Koopmann, Bernadette D. M.", "et al." ], "publisher": "United European Gastroenterology Journal", "date": "2025-06-18", "url": "https://pubmed.ncbi.nlm.nih.gov/40530613/", "doi": "10.1002/ueg2.70043", "pmid": "40530613", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Nested 44-centre cohort of 975 people without baseline worrisome or high-risk features; median follow-up 45 months and 20 HGD or pancreatic-cancer events. Slow growth was protective, but the small-stable subgroup's comparison with population incidence had a very wide 95% confidence interval of 0.01 to 6.30." }, { "id": "levy-2019-celiac-ganglia-versus-plexus-rct", "type": "paper", "title": "Combined Celiac Ganglia and Plexus Neurolysis Shortens Survival, Without Benefit, vs Plexus Neurolysis Alone", "authors": [ "Levy, Michael J", "Gleeson, Ferga C", "Topazian, Mark D", "et al." ], "publisher": "Clinical Gastroenterology and Hepatology", "date": "2018-09-12", "url": "https://doi.org/10.1016/j.cgh.2018.08.040", "doi": "10.1016/j.cgh.2018.08.040", "pmid": "30217513", "trialId": "NCT01615653", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Randomized double-blind comparison in painful unresectable PDAC: 50 received direct celiac-ganglion plus plexus neurolysis and 60 plexus neurolysis alone. Twelve-week pain response was 46.2% versus 40.4%, p=0.84, and quality of life did not differ. Median survival was 5.59 versus 10.46 months, HR 1.49, 95% CI 1.02-2.19; the survival signal was larger in non-metastatic disease." }, { "id": "li-2015-hifu-s1", "type": "paper", "title": "Retrospective analysis of high intensity focused ultrasound combined with S-1 in the treatment of metastatic pancreatic cancer after failure of gemcitabine", "authors": [ "Li, Xiaoping", "Wang, Kun", "Zheng, Leizhen", "Meng, Zhiqiang" ], "publisher": "American Journal of Cancer Research", "date": "2015-12-15", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC4759399/", "doi": null, "pmid": "27073725", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Retrospective two-centre record of 120 people after gemcitabine failure. The abstract says randomly assigned, but the methods say records were collected retrospectively and describe only division into 61 HIFU plus S-1 and 59 S-1 groups. It reports median survival 10.3 versus 6.6 months and pain remission 57% versus 20%; it does not report a random sequence, concealment or trial registration." }, { "id": "li-2021-gain", "type": "paper", "title": "GAIN: geriatric assessment-driven intervention and chemotherapy toxicity", "authors": [ "Li, Daneng", "Sun, Can-Lan", "Kim, Heeyoung", "et al." ], "publisher": "JAMA Oncology", "date": "2021-11-01", "url": "https://pubmed.ncbi.nlm.nih.gov/34591080/", "doi": "10.1001/jamaoncol.2021.4158", "pmid": "34591080", "trialId": "NCT02517034", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Single-centre randomized trial of 605 analyzable adults aged at least 65 starting chemotherapy for mixed solid cancers. A multidisciplinary team acted on assessment-defined functional, nutrition, medication, social and other needs. Grade 3 or worse toxicity fell from 60.6% to 50.5%, an absolute 10.1-point reduction, P=.02. Overall survival, hospital use, dose modification and early discontinuation did not differ. Gastrointestinal cancers were 202/605, but PDAC was not reported separately." }, { "id": "li-2023-tumour-selective-gpx4-degrader", "type": "paper", "title": "Tumor-specific GPX4 degradation enhances ferroptosis-initiated antitumor immune response in mouse models of pancreatic cancer", "authors": [ "Li, Jingbo", "Liu, Jiao", "Zhou, Zhuan", "et al." ], "publisher": "Science Translational Medicine", "date": "2023-11-01", "url": "https://doi.org/10.1126/scitranslmed.adg3049", "doi": "10.1126/scitranslmed.adg3049", "pmid": "37910602", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "N6F11 caused TRIM25-dependent GPX4 destruction in cancer cells but not tested dendritic, T, natural-killer or neutrophil cells. It triggered CD8-dependent antitumour immunity and increased checkpoint-blockade activity in mouse models including KRAS/TP53-driven pancreatic cancer. The study makes tumour-cell selectivity an explicit design requirement; N6F11 has no reported human PDAC exposure." }, { "id": "li-2024-pancreatic-early-mobilization-rct", "type": "paper", "title": "Early mobilization after pancreatic surgery: randomized controlled trial", "authors": [ "Li, Zhi", "et al." ], "publisher": "Surgery", "date": "2024-10-01", "url": "https://doi.org/10.1016/j.surg.2024.06.027", "doi": "10.1016/j.surg.2024.06.027", "pmid": "39054183", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Single-centre trial randomized 135 pancreatic-surgery patients to usual care or assisted day-one walking plus family-supervised goals. The 30-day Comprehensive Complication Index did not significantly differ; mean difference -1.7, 95% CI -8.7 to 0. Mobilization and some recovery measures improved, but the main complication endpoint did not." }, { "id": "liaki-2026-triple-combination-republished", "type": "paper", "title": "A targeted combination therapy achieves effective pancreatic cancer regression and prevents tumor resistance", "authors": [ "Liaki, Vasiliki", "Barrambana, Sara", "Kostopoulou, Myrto", "Guerra, Carmen", "Barbacid, Mariano" ], "publisher": "Proceedings of the National Academy of Sciences", "date": "2026-06-09", "url": "https://pubmed.ncbi.nlm.nih.gov/42224594/", "doi": "10.1073/pnas.2610708123", "pmid": "42224594", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Primary republished mouse and patient-derived-xenograft paper. Daraxonrasib plus afatinib plus the STAT3 degrader SD36 produced complete durable regressions in small orthotopic cohorts and broad regression in 12 genetically engineered mice. The authors state that daraxonrasib was used at about five times the dose used in clinical trials, that the afatinib dose raises clinical tolerability questions and that SD36 lacks suitable drug properties for a trial. No human efficacy. The declaration states financial interests in Vega Oncotargets and multiple company research contracts." }, { "id": "liaw-2004-lamivudine-hbv", "type": "paper", "title": "Lamivudine for patients with chronic hepatitis B and advanced liver disease", "authors": [ "Liaw, Yun-Fan", "Sung, Joseph J.Y.", "Chow, Wan-Cheng", "et al." ], "publisher": "New England Journal of Medicine", "date": "2004-10-07", "url": "https://pubmed.ncbi.nlm.nih.gov/15470215/", "doi": "10.1056/NEJMoa033364", "pmid": "15470215", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Randomised 651-patient trial, 98% Asian and 85% male, stopped early for benefit on a composite liver-disease endpoint. Older lamivudine treatment developed resistance in 49%; modern antiviral drugs differ." }, { "id": "lim-2025-attention-hbv", "type": "paper", "title": "Early antiviral treatment with tenofovir alafenamide to prevent serious clinical adverse events in adults with chronic hepatitis B and moderate or high viraemia (ATTENTION): interim results from a randomised controlled trial", "authors": [ "Lim, Young-Suk", "Yu, Ming-Lung", "Choi, Jonggi", "et al." ], "publisher": "The Lancet Gastroenterology & Hepatology", "date": "2025-04-01", "url": "https://pubmed.ncbi.nlm.nih.gov/39914435/", "doi": "10.1016/S2468-1253(24)00431-X", "pmid": "39914435", "trialId": "NCT03753074", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Prespecified interim analysis of an ongoing randomized trial in 734 noncirrhotic adults with chronic HBV, moderate or high viraemia and normal or mildly raised ALT. The composite endpoint occurred in 2 of 369 on tenofovir alafenamide and 9 of 365 under observation; HCC counts were 2 versus 7. HR 0.21, 97.5% CI 0.04-1.20; the result did not cross the early-stop boundary. Funded by the South Korean government and Gilead." }, { "id": "lim-2026-p32-landmark", "type": "paper", "title": "Combined phosphorus-32 implantation and chemotherapy versus chemotherapy alone for locally advanced pancreatic cancer: a propensity score-weighted landmark analysis", "authors": [ "Lim, Amanda H.", "Nitchingham, Darshan", "Bednarz, Jana", "et al." ], "publisher": "Gastrointestinal Endoscopy", "date": "2026-01-01", "url": "https://pubmed.ncbi.nlm.nih.gov/40348036/", "doi": "10.1016/j.gie.2025.04.054", "pmid": "40348036", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Retrospective two-hospital analysis. Of 104 considered, 86 survived to a three-month landmark and met timing rules: 35 combination and 51 standard care. Weighted restricted mean survival through 30 months was estimated 189 days longer, p=0.002. Landmarking addressed guaranteed-survival time but excluded early deaths and cannot remove unknown treatment selection. OncoSil provided research support to one author." }, { "id": "lin-2013-reversible-pdac-dormancy", "type": "paper", "title": "Dormant cancer cells contribute to residual disease in a model of reversible pancreatic cancer", "authors": [ "Lin, Wen-Chin", "Rajbhandari, Nisha", "Liu, Chia", "et al." ], "publisher": "Cancer Research", "date": "2013-03-15", "url": "https://pubmed.ncbi.nlm.nih.gov/23467612/", "doi": "10.1158/0008-5472.CAN-12-2067", "pmid": "23467612", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Genetically controlled mouse model. Turning off c-Myc caused visible primary and metastatic tumours to regress, but residual cells persisted and recurrence followed c-Myc restoration. This is not evidence that the same switch controls dormant cells in patients." }, { "id": "lin-2026-vertical-kras-inhibition", "type": "paper", "title": "Overcoming Adaptive Resistance to KRASG12D Blockade in Pancreatic Cancer through Vertical Pathway Inhibition", "authors": [ "Lin, Qingxiang", "et al." ], "publisher": "Clinical Cancer Research", "date": "2026-06-01", "url": "https://pubmed.ncbi.nlm.nih.gov/41801133/", "doi": "10.1158/1078-0432.CCR-25-1788", "pmid": "41801133", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "PDAC model study: epithelial states used EGFR-driven RAS-MAPK reactivation after G12D inhibition, while mesenchymal states relied more on FGFR. Combining a G12D-selective inhibitor with a broad RAS(ON) inhibitor suppressed reactivation more consistently than EGFR combinations in the tested models. Human TCGA expression correlations supported state associations, but there was no treated human efficacy result. One senior author disclosed Revolution Medicines funding and several company relationships." }, { "id": "liu-2020-ccga-methylation", "type": "paper", "title": "Sensitive and specific multi-cancer detection and localization using methylation signatures in cell-free DNA", "authors": [ "Liu, Minetta C.", "Oxnard, Geoffrey R.", "Klein, Eric A.", "et al." ], "publisher": "Annals of Oncology", "date": "2020-03-30", "url": "https://pubmed.ncbi.nlm.nih.gov/33506766/", "doi": "10.1016/j.annonc.2020.02.011", "pmid": "33506766", "trialId": "NCT02889978; NCT03085888", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Prospective case-control substudy with 2,482 cancer and 4,207 non-cancer participants, split into training and validation. Funded by GRAIL; several authors were employees or equity holders." }, { "id": "liu-2024-robotic-open-pd", "type": "paper", "title": "Effect of robotic versus open pancreaticoduodenectomy on postoperative length of hospital stay and complications for pancreatic head or periampullary tumours: a multicentre, open-label randomised controlled trial", "authors": [ "Liu, Qiang", "Li, Ming", "Gao, Yi", "et al." ], "publisher": "The Lancet Gastroenterology & Hepatology", "date": "2024-02-28", "url": "https://europepmc.org/article/MED/38428441", "doi": "10.1016/s2468-1253(24)00005-0", "pmid": "38428441", "trialId": "ChiCTR2200056809", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Three high-volume Chinese-centre randomized trial of 164 patients with mixed benign, premalignant or malignant head/periampullary tumours; 161 had surgery. Surgeons had done at least 40 robotic and 60 open procedures. Median stay was 11.0 versus 13.5 days; one death occurred in each arm and severe-complication confidence intervals were wide. It reported no funding and called for medium- and long-term outcomes." }, { "id": "liu-2026-ttfields-cost", "type": "paper", "title": "Tumour-treating fields plus chemotherapy for locally advanced pancreatic adenocarcinoma: an international cost-effectiveness analysis", "authors": [ "Liu, Kun", "Zhu, Youwen", "Zeng, Manting", "Zhu, Hong" ], "publisher": "Scientific Reports", "date": "2026-05-25", "url": "https://pubmed.ncbi.nlm.nih.gov/42185415/", "doi": "10.1038/s41598-026-54203-x", "pmid": "42185415", "trialId": "NCT03377491", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Independent published model funded by the Changsha Natural Science Foundation; authors declared no competing interests. It estimated 0.148 extra quality-adjusted life-years and $147,766 extra lifetime cost in the United States, or $1,001,995 per quality-adjusted life-year. The China model estimated 0.142 quality-adjusted life-years, $39,470 extra cost and $278,708 per quality-adjusted life-year. Both exceeded the model's national thresholds. Modelled value is not an observed trial outcome." }, { "id": "llach-2024-universal-germline", "type": "paper", "title": "Should We Offer Universal Germline Genetic Testing to All Patients with Pancreatic Cancer? A Multicenter Study", "authors": [ "Llach, Judit", "Luzko, Ivan", "Earl, Julie", "et al." ], "publisher": "Cancers", "date": "2024-11-09", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC11592128/", "doi": "10.3390/cancers16223779", "pmid": "39594734", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Retrospective three-centre Spanish cohort of 179 newly diagnosed PDAC patients tested regardless of history. Fourteen, 7.8%, had a pathogenic or likely pathogenic variant, and seven of those 14 did not meet the study's clinical criteria. The median age was 58 and all patients passed through high-risk clinics, so the finding does not settle yield or cost in an unselected national service." }, { "id": "lodestijn-2021-environment-clonogenicity", "type": "paper", "title": "Marker-free lineage tracing reveals an environment-instructed clonogenic hierarchy in pancreatic cancer", "authors": [ "Lodestijn, Sanne C.", "Miedema, Daniël M.", "Lenos, Kristiaan J.", "et al." ], "publisher": "Cell Reports", "date": "2021-10-19", "url": "https://pubmed.ncbi.nlm.nih.gov/34686335/", "doi": "10.1016/j.celrep.2021.109852", "pmid": "34686335", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Marker-independent clonal labelling in established PDAC tissue models found broad clonogenic potential and strong control by proximity to activated fibroblasts. It argues against a fixed rare-cell hierarchy in the tested models, not against every operational stem-like state in patients." }, { "id": "lorenzo-bermejo-2025-gallbladder-specimens", "type": "paper", "title": "Gallbladder Cancer and Dysplasia in Cholecystectomy Specimens: A Large Study in High-Incidence Regions of South America", "authors": [ "Lorenzo Bermejo, Justo", "et al." ], "publisher": "Clinical Gastroenterology and Hepatology", "date": "2025-08-01", "url": "https://pubmed.ncbi.nlm.nih.gov/40015496/", "doi": "10.1016/j.cgh.2024.12.027", "pmid": "40015496", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Retrospective clinical analysis of 10,561 gallstone patients undergoing cholecystectomy in Argentina, Bolivia, Chile and Peru. Recruitment-site heterogeneity limits transportability." }, { "id": "lowenfels-1993-chronic-pancreatitis", "type": "paper", "title": "Pancreatitis and the risk of pancreatic cancer", "authors": [ "Lowenfels, Albert B.", "Maisonneuve, Patrick", "Cavallini, Giorgio", "et al." ], "publisher": "New England Journal of Medicine", "date": "1993-05-20", "url": "https://pubmed.ncbi.nlm.nih.gov/8479461/", "doi": "10.1056/NEJM199305203282001", "pmid": "8479461", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Historical cohort of 2,015 chronic-pancreatitis patients at clinical centres in six countries. Relative-risk estimates are vulnerable to early reverse causation and older diagnostic practice; analyses excluding early follow-up remained elevated." }, { "id": "lowenfels-1997-hereditary-pancreatitis", "type": "paper", "title": "Hereditary pancreatitis and the risk of pancreatic cancer", "authors": [ "Lowenfels, Albert B.", "Maisonneuve, Patrick", "DiMagno, Eugene P.", "et al." ], "publisher": "Journal of the National Cancer Institute", "date": "1997-03-19", "url": "https://pubmed.ncbi.nlm.nih.gov/9091646/", "doi": "10.1093/jnci/89.6.442", "pmid": "9091646", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "International historical cohort of 246 people thought to have hereditary pancreatitis; eight pancreatic adenocarcinomas. The rare-disease risk estimate is large but imprecise and based on older clinical criteria." }, { "id": "ludmir-2024-extend", "type": "paper", "title": "Addition of metastasis-directed therapy to systemic therapy for oligometastatic PDAC: EXTEND randomized phase II trial", "authors": [ "Ludmir, Ethan B.", "Sherry, Alexander D.", "Fellman, Bryan M.", "et al." ], "publisher": "Journal of Clinical Oncology", "date": "2024-08-05", "url": "https://pubmed.ncbi.nlm.nih.gov/39102622/", "doi": "10.1200/JCO.24.00081", "pmid": "39102622", "trialId": "NCT03599765", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Primary multicentre randomized phase 2 basket cohort: 40 eligible PDAC patients; median PFS 10.3 versus 2.5 months, HR 0.43, 95% CI 0.20-0.94. Per-protocol primary analysis, small cohort, OS not established." }, { "id": "lummy-2025-annual-report-cnsi-fe", "type": "company-filing", "title": "Chongqing Lummy Pharmaceutical 2025 annual report", "authors": [ "Chongqing Lummy Pharmaceutical Co Ltd" ], "publisher": "CNINFO", "date": "2026-03-19", "url": "https://static.cninfo.com.cn/finalpage/2026-03-19/1225016087.PDF", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "The listed parent reports that its Sichuan Enray subsidiary is advancing CNSI-Fe phase II work and a carbon-particle heat platform. The filing confirms corporate ownership and development commitment, but does not supply an independently separated CNSI-Fe valuation or a pancreatic-specific development budget." }, { "id": "lustgarten-2025-grants", "type": "charity-report", "title": "Lustgarten Foundation announces 2025 pancreatic-cancer research grants", "authors": [ "Lustgarten Foundation" ], "publisher": "Lustgarten Foundation", "date": "2025-09-15", "url": "https://lustgarten.org/the-lustgarten-foundation-announces-landmark-year-of-fueling-cutting-edge-pancreatic-cancer-research-2/", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Primary charity release. Ten new grants totalled $13.85m; 51 active grants ran across 29 institutions; claimed cumulative grant funding exceeded $301m since 1998. Cumulative and new-commitment figures are not annual cash expense." }, { "id": "lyu-2023-metal-plastic-nat-meta", "type": "paper", "title": "Comparison of metal versus plastic stent for preoperative biliary drainage during neoadjuvant therapy", "authors": [ "Lyu, Yunxiao", "Ye, Shenjian", "Wang, Bin" ], "publisher": "BMC Gastroenterology", "date": "2023-07-12", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC10337088/", "doi": "10.1186/s12876-023-02874-5", "pmid": "37438761", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Meta-analysis of studies in pancreatic cancer receiving neoadjuvant treatment. Metal versus plastic stents reduced reintervention in four studies and 333 patients, OR .04, delay of treatment across seven studies, OR .17, recurrent obstruction in nine studies and 463 patients, OR .09, and cholangitis in four studies and 218 patients, OR .16. Pooled recurrent obstruction was 19.88% versus 57.53%. Postoperative complications and R0 resection did not differ. Only two included studies were randomized; samples were small, outcome definitions and treatment regimens varied, and metal-stent types differed." }, { "id": "lyu-2026-torch-hcc", "type": "paper", "title": "Transarterial Chemoembolization Plus Thermal Ablation in Unresectable Hepatocellular Carcinoma: The Phase 3 TORCH Randomized Clinical Trial", "authors": [ "Lyu, Ning", "Yi, Jun-Zhe", "Wu, Xin-Tong", "et al." ], "publisher": "JAMA Oncology", "date": "2026-07-30", "url": "https://pubmed.ncbi.nlm.nih.gov/42530948/", "doi": "10.1001/jamaoncol.2026.2366", "pmid": "42530948", "trialId": "NCT02435953", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Open-label phase 3 trial at two Chinese tertiary centres. The all-randomized population was 241 people with liver-confined BCLC stage B HCC. Sequential TACE and selective radiofrequency ablation improved both PFS and OS versus TACE alone. The narrow setting, two-centre geography and post-TACE selection geometry limit transfer." }, { "id": "macarulla-2025-panova3", "type": "paper", "title": "Tumor Treating Fields With Gemcitabine and Nab-Paclitaxel for Locally Advanced Pancreatic Adenocarcinoma: Randomized, Open-Label, Pivotal Phase III PANOVA-3 Study", "authors": [ "Macarulla, Teresa", "Seufferlein, Thomas", "Van Cutsem, Eric", "et al." ], "publisher": "Journal of Clinical Oncology", "date": "2025-06-02", "url": "https://pubmed.ncbi.nlm.nih.gov/40448572/", "doi": "10.1200/JCO-25-00746", "pmid": "40448572", "trialId": "NCT03377491", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Global open-label randomized phase III trial in 571 patients with newly diagnosed unresectable locally advanced pancreatic adenocarcinoma." }, { "id": "maddipati-2015-polyclonal-metastasis", "type": "paper", "title": "Pancreatic Cancer Metastases Harbor Evidence of Polyclonality", "authors": [ "Maddipati, Ravi", "Stanger, Ben Z." ], "publisher": "Cancer Discovery", "date": "2015-07-24", "url": "https://pubmed.ncbi.nlm.nih.gov/26209539/", "doi": "10.1158/2159-8290.CD-15-0120", "pmid": "26209539", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Multicolour lineage tracing in a native mouse PDAC model. It shows that multiple clones can seed a metastasis in this model; it does not establish the frequency in human PDAC." }, { "id": "maekawa-2026-small-pnet-surgery-morbidity", "type": "paper", "title": "Procedure-specific morbidity and pathologic assessment after resection of small pancreatic neuroendocrine tumors", "authors": [ "Maekawa, Aya", "Perri, Giampaolo", "Partelli, Stefano" ], "publisher": "Journal of Neuroendocrinology", "date": "2026-08-01", "url": "https://pubmed.ncbi.nlm.nih.gov/42580359/", "doi": "10.1111/jne.70243", "pmid": "42580359", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Retrospective multicentre cohort of 460 resections for PanNET no larger than 20 mm. Clinically relevant postoperative pancreatic fistula occurred in 25.0%, major morbidity in 14.8% and 90-day mortality in 0.9%. Central pancreatectomy had more major morbidity than enucleation, and lymph-node assessment varied strongly by procedure. The cohort includes only operated patients and cannot estimate the benefit of surgery over observation." }, { "id": "mahalingam-2026-elraglusib-randomized", "type": "paper", "title": "Elraglusib and chemotherapy in metastatic pancreatic ductal adenocarcinoma: a randomized controlled phase 2 trial", "authors": [ "Mahalingam, Devalingam", "Shroff, Rachna T.", "Carneiro, Benedito A.", "et al." ], "publisher": "Nature Medicine", "date": "2026-05-01", "url": "https://pubmed.ncbi.nlm.nih.gov/41981308/", "doi": "10.1038/s41591-026-04327-4", "pmid": "41981308", "trialId": "NCT03678883", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Primary peer-reviewed phase 2 report. In the treated mITT set, weekly elraglusib plus gemcitabine/nab-paclitaxel gave OS 10.1 versus 7.2 months, HR 0.62, 95% CI 0.46-0.84, P=0.010. In all randomized weekly and control participants, OS was 8.9 versus 7.2 months, HR 0.68, 95% CI 0.51-0.89, P=0.022. PFS, response and disease control did not differ significantly. The study was open label, used an adaptive dose-selection stage, changed its primary survival analysis after starting, and made the treated population primary." }, { "id": "mainz-2026-10k", "type": "sec-filing", "title": "Mainz Biomed 2025 annual report", "authors": [ "Mainz Biomed N.V." ], "publisher": "US Securities and Exchange Commission", "date": "2026-04-01", "url": "https://www.sec.gov/Archives/edgar/data/1874252/000121390026037676/ea0283918-10k_mainz.htm", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Reports a March 2025 Liquid Biosciences licence for pancreatic blood-test mRNA biomarkers and algorithm, an early stool-based PancAlert candidate, roughly 13 staff and the 2026 wind-down of colorectal operations." }, { "id": "mainz-2026-private-placement", "type": "sec-filing", "title": "Mainz Biomed $6 million private placement and pancreatic-detection pivot", "authors": [ "Mainz Biomed N.V." ], "publisher": "US Securities and Exchange Commission", "date": "2026-02-17", "url": "https://www.sec.gov/Archives/edgar/data/1874252/000121390026016890/ea027724601ex99-1_mainz.htm", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "$6 million convertible preferred private placement in two $3 million tranches, described as runway for operations, liabilities and strategic alternatives." }, { "id": "maltoni-2016-systematic-palliative-end-of-life", "type": "paper", "title": "Systematic versus on-demand early palliative care: a randomised clinical trial assessing quality of care and treatment aggressiveness near the end of life", "authors": [ "Maltoni, Marco", "Scarpi, Emanuela", "Dall'Agata, Monia", "et al." ], "publisher": "European Journal of Cancer", "date": "2016-12-01", "url": "https://doi.org/10.1016/j.ejca.2016.10.004", "doi": "10.1016/j.ejca.2016.10.004", "pmid": "27821313", "trialId": "NCT01996540", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Secondary analysis of the same 207-person randomized pancreatic trial. After 20 months, 149 of 186 evaluable participants had died. Chemotherapy in the last 30 days of life occurred in 18.7% with systematic palliative care versus 27.8% with on-demand care, adjusted p=0.036, an absolute reduction of 9.1 points. Hospice duration and number of admissions increased. Other end-of-life differences were not significant." }, { "id": "maltoni-2016-systematic-palliative-qol", "type": "paper", "title": "Systematic versus on-demand early palliative care: results from a multicentre, randomised clinical trial", "authors": [ "Maltoni, Marco", "Scarpi, Emanuela", "Dall'Agata, Monia", "et al." ], "publisher": "European Journal of Cancer", "date": "2016-09-01", "url": "https://doi.org/10.1016/j.ejca.2016.06.007", "doi": "10.1016/j.ejca.2016.06.007", "pmid": "27472648", "trialId": "NCT01996540", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Multicentre randomized trial of 207 outpatients with metastatic or inoperable locally advanced pancreatic cancer: 107 systematic early palliative care and 100 cancer care with palliative care on demand. At 12 weeks, change in the FACT-Hep Trial Outcome Index differed by 3.83 points, 95% CI 0.10 to 7.57, p=0.041, favouring systematic care. There was no overall-survival difference. Only 186 were evaluable and the intervention could not be blinded." }, { "id": "mandel-2000-fobt-incidence", "type": "paper", "title": "The effect of fecal occult-blood screening on the incidence of colorectal cancer", "authors": [ "Mandel, Jack S.", "Church, Timothy R.", "Bond, John H.", "et al." ], "publisher": "New England Journal of Medicine", "date": "2000-11-30", "url": "https://pubmed.ncbi.nlm.nih.gov/11096167/", "doi": "10.1056/NEJM200011303432203", "pmid": "11096167", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Eighteen-year Minnesota Colon Cancer Control Study follow-up; 46,551 people randomised to annual, biennial or usual-care screening. Guaiac tests with colonoscopy after positive results." }, { "id": "manji-2023-mekiauto", "type": "meeting-abstract", "title": "MEKiAUTO: cobimetinib, hydroxychloroquine and atezolizumab in KRAS-mutated advanced malignancies", "authors": [ "Manji, Gulam Abbas", "et al." ], "publisher": "JCO Global Oncology", "date": "2023-08-03", "url": "https://ascopubs.org/doi/10.1200/GO.2023.9.Supplement_1.45", "doi": "10.1200/GO.2023.9.Supplement_1.45", "pmid": null, "trialId": "NCT04214418", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Phase 1 meeting abstract reporting 14 treated people with advanced KRAS-mutated pancreatic adenocarcinoma. Four of ten dose-limiting-toxicity-evaluable patients had a dose-limiting toxicity; 7/14 had grade 3 or worse treatment-related events and one possibly related grade 5 event occurred. Best response was stable disease in 5/14; median progression-free survival was 7.7 weeks and overall survival 20.7 weeks. Roche-Genentech funded the work." }, { "id": "maraveyas-2012-fragem", "type": "paper", "title": "Gemcitabine versus gemcitabine plus dalteparin thromboprophylaxis in pancreatic cancer", "authors": [ "Maraveyas, Anthony", "Waters, Jonathan", "Roy, Robin", "et al." ], "publisher": "European Journal of Cancer", "date": "2012-06-01", "url": "https://pubmed.ncbi.nlm.nih.gov/22100906/", "doi": "10.1016/j.ejca.2011.10.017", "pmid": "22100906", "trialId": "NCT00462852", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "FRAGEM randomized 123 people with advanced pancreatic cancer to gemcitabine with or without weight-adjusted dalteparin for 12 weeks. All VTE during treatment fell from 23% to 3.4%, P=.002; across follow-up it fell from 28% to 12%, P=.039. Lethal VTE before day 100 was 8.3% versus 0%, P=.057. Published secondary accounts report no survival difference. The regimen used older gemcitabine monotherapy and higher-than-standard prophylactic dosing." }, { "id": "marchegiani-2019-pnet-recurrence", "type": "paper", "title": "Patterns of recurrence after resection for pancreatic neuroendocrine tumors: who, when, and where?", "authors": [ "Marchegiani, Giovanni", "Landoni, Luca", "Andrianello, Stefano", "et al." ], "publisher": "Neuroendocrinology", "date": "2018-11-27", "url": "https://pubmed.ncbi.nlm.nih.gov/30481765/", "doi": "10.1159/000495774", "pmid": "30481765", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Single-centre retrospective series of 487 resected pancreatic neuroendocrine neoplasms with median follow-up 71 months. Recurrence occurred in 12.3%, including liver recurrence in 11.1%; 6.4% died from recurrent disease. Tumour size above 21 mm, G3 grade, nodal metastasis and vascular infiltration independently predicted recurrence. Four recurrences appeared after ten years. No untreated comparator and subgroup cure language is observational." }, { "id": "marinova-2018-hifu-prospective", "type": "paper", "title": "Clinical Effectiveness and Potential Survival Benefit of US-Guided High-Intensity Focused Ultrasound Therapy in Patients with Advanced-Stage Pancreatic Cancer", "authors": [ "Marinova, Milka", "Huxold, Hannah C.", "Henseler, Jana", "et al." ], "publisher": "Ultraschall in der Medizin", "date": "2018-04-17", "url": "https://pubmed.ncbi.nlm.nih.gov/29665583/", "doi": "10.1055/a-0591-3386", "pmid": "29665583", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Prospective uncontrolled study of 50 advanced pancreatic-cancer patients. It reports early pain relief in 84%, median survival 16.2 months from diagnosis and 8.3 months from HIFU, but cannot estimate added survival or distinguish treatment from selection and concurrent care." }, { "id": "marinova-2024-hifu-pc-protocol", "type": "paper", "title": "Study Protocol of a Randomized, Two-Arm, Phase I/II Trial Investigating Local US-Guided High-Intensity Focused Ultrasound With Palliative Chemotherapy in Inoperable Pancreatic Cancer", "authors": [ "Marinova, Milka", "Khouri, David-Alexandre", "Küppers, Jonas", "et al." ], "publisher": "Journal of Clinical Medicine", "date": "2024-06-26", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC11242276/", "doi": "10.3390/jcm13133717", "pmid": "38999283", "trialId": "DRKS00012367", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Protocol for the 40-person HIFU-PC15 randomized phase 1/2 study. It mixes stage III and IV disease, first- and second-line therapy and several chemotherapy backbones. Safety and local response are primary; pain, quality of life and survival are secondary. The article processing charge was funded by Chongqing HAIFU." }, { "id": "martin-2024-direct-registry-safety", "type": "paper", "title": "Effectiveness and Safety of Irreversible Electroporation When Used for the Ablation of Stage 3 Pancreatic Adenocarcinoma: Initial Results from the DIRECT Registry Study", "authors": [ "Martin, Robert C. G. II", "White, Rebekah Ruth", "Bilimoria, Malcolm M.", "et al." ], "publisher": "Cancers", "date": "2024-11-21", "url": "https://pubmed.ncbi.nlm.nih.gov/39682087/", "doi": "10.3390/cancers16233894", "pmid": "39682087", "trialId": "NCT03899649", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Sponsor-funded observational safety report. It enrolled 87 IRE and 27 standard-care patients after at least three months of multi-drug chemotherapy without progression. Tumours were smaller in the IRE group, 2.2 versus 3.2 cm, P=0.0066, and 42.5% of IRE patients but no controls had prior radiation. Ninety-day mortality was 5/83 versus 2/27. Two IRE patients died from procedure-related abdominal bleeding or vascular pseudoaneurysm; one control died from chemotherapy. Grade 3 or worse events were 24/87 versus 12/27. Every IRE procedure was open surgery. AngioDynamics funded and participated in design, monitoring, data collection, analysis, interpretation and manuscript work; several authors disclosed company ties." }, { "id": "martin-2026-direct-registry-survival-preprint", "type": "preprint", "title": "Irreversible electroporation associated with improved overall survival vs standard of care for stage 3 pancreatic ductal adenocarcinoma", "authors": [ "Martin, Robert C. G. II", "White, Rebekah Ruth", "Bilimoria, Malcolm M.", "et al." ], "publisher": "medRxiv", "date": "2026-05-21", "url": "https://www.medrxiv.org/content/10.64898/2026.05.19.26353144v1", "doi": "10.64898/2026.05.19.26353144", "pmid": null, "trialId": "NCT03899649", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Not peer reviewed. The observational analysis includes 99 IRE and 38 controls; the control group combines 10 prospectively enrolled patients with 28 contemporaneous records collected later. Median diagnosis-to-enrollment time was 8 versus 4 months, P<0.0001. From enrollment, median survival was 18 versus 10 months and unadjusted HR 0.4; the paper reports adjusted HR 0.3. In the ten prospective controls, median survival was 18 versus 12 months, HR 0.5 with CI 0.3-1.0 and P=0.1016. Grade 3 or worse events affected 55/99 versus 27/38; 29/99 had an IRE-related event. The authors acknowledge that the non-random design cannot remove survival filtering, treatment differences or remaining immortal-time bias. AngioDynamics supplied financial and logistical support." }, { "id": "maruoka-2018-interstitial-photoimmunotherapy", "type": "paper", "title": "Near infrared photoimmunotherapy with combined exposure of external and interstitial light sources", "authors": [ "Maruoka, Yasuhiro", "Nagaya, Tadanobu", "Sato, Kazuhide", "et al." ], "publisher": "Molecular Pharmaceutics", "date": "2018-02-21", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC7400989/", "doi": "10.1021/acs.molpharmaceut.8b00002", "pmid": "29450993", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Mouse xenograft study of EGFR antibody-IR700 treatment. The paper states that light from outside penetrates tissue by about 2 cm and found combined outside and inserted-fibre illumination worked better than outside light alone for deeper tumours. It did not use pancreatic tumours or ASP-1929 in people." }, { "id": "matsumoto-2026-mtap-pancreatic", "type": "paper", "title": "Clinical characteristics and therapeutic implications of MTAP deletion in pancreatic cancer", "authors": [ "Matsumoto, Hiroya", "Mizuno, Kazuyuki", "Kondo, Chiaki", "et al." ], "publisher": "Pancreatology", "date": "2026-07-18", "url": "https://pubmed.ncbi.nlm.nih.gov/42493328/", "doi": "10.1016/j.pan.2026.07.015", "pmid": "42493328", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Nationwide retrospective genomic cohort, n=8,834 pancreatic cancers. MTAP deletion occurred in 27.3% overall and 28.0% of PDAC. Within the CDKN2A-deleted subgroup it was not independently prognostic, OS HR 0.96 (95% CI 0.85-1.08), and standard chemotherapy response was comparable." }, { "id": "maude-2018-tisagenlecleucel", "type": "paper", "title": "Tisagenlecleucel in children and young adults with B-cell lymphoblastic leukemia", "authors": [ "Maude, Shannon L.", "Laetsch, Theodore W.", "Buechner, Jochen", "et al." ], "publisher": "New England Journal of Medicine", "date": "2018-02-01", "url": "https://pubmed.ncbi.nlm.nih.gov/29385370/", "doi": "10.1056/NEJMoa1709866", "pmid": "29385370", "trialId": "NCT02435849", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Primary global pivotal single-arm CAR-T study. Used for the blood-cancer property bundle—lineage antigen, circulating access, cell expansion—not as direct PDAC efficacy evidence." }, { "id": "mayo-2026-in-cyst-release", "type": "institution-report", "title": "Mayo Clinic study estimates cancer risk in patients with pancreatic cysts, advancing early detection", "authors": [ "Mayo Clinic News Network" ], "publisher": "Mayo Clinic", "date": "2026-08-20", "url": "https://newsnetwork.mayoclinic.org/discussion/mayo-clinic-study-estimates-cancer-risk-in-patients-with-pancreatic-cysts-advancing-early-detection/", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Official institutional report accompanying the IN-CYST paper. It provides cohort flow and three-year risk estimates not present in the PubMed record: about 2%, 4%, and 18% with one, two, and at least three worrisome features. It is not an independent publication and should be read with the paper." }, { "id": "mazzaferro-1996-hcc-transplant", "type": "paper", "title": "Liver transplantation for the treatment of small hepatocellular carcinomas in patients with cirrhosis", "authors": [ "Mazzaferro, Vincenzo", "Regalia, Enrico", "Doci, Roberto", "et al." ], "publisher": "New England Journal of Medicine", "date": "1996-03-14", "url": "https://pubmed.ncbi.nlm.nih.gov/8594428/", "doi": "10.1056/NEJM199603143341104", "pmid": "8594428", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Prospective single-centre series of 48 cirrhotic patients with small unresectable HCC. Among the 35 whose explants met the prespecified tumour limits, actuarial four-year overall and recurrence-free survival were 85% and 92%; among 13 beyond the limits they were 50% and 59%. The PubMed record and journal abstract were checked; this was not a randomized comparison and follow-up was a median 26 months." }, { "id": "mcclellan-2026-quiescent-pdac-car-t", "type": "paper", "title": "Quiescent tumor cells shape the immunosuppressive microenvironment in pancreatic cancer", "authors": [ "McClellan, B.", "Wang, Q.", "Aung, K.", "Matsui, W." ], "publisher": "Nature Communications", "date": "2026-07-21", "url": "https://pubmed.ncbi.nlm.nih.gov/42481500/", "doi": "10.1038/s41467-026-75883-z", "pmid": "42481500", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Orthotopic mouse PDAC work found a rare quiescent population enriched after CAR-T treatment. The cells expressed EREG, increased ErbB4-positive macrophages and resisted immune attack; genetic or drug targeting of EREG improved CAR-T effects and mouse survival. This is a treatment-specific model state, not a validated human residual-cell marker or intervention." }, { "id": "mcconathy-2026-lumiere-phase1", "type": "paper", "title": "Targeting Fibroblast Activation Protein with 177Lu-FAP-2286 in Patients with Advanced Solid Tumors in the Phase I LuMIERE Trial", "authors": [ "McConathy, Jonathan", "Koshkin, Vadim S.", "Menda, Yusuf", "et al." ], "publisher": "Clinical Cancer Research", "date": "2026-06-01", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC13335097/", "doi": "10.1158/1078-0432.CCR-25-4356", "pmid": "42223513", "trialId": "NCT04939610", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Prospective, open-label, non-randomized phase 1 across six US centres. Thirty-five people were scanned and 27 treated; nine had pancreatic cancer. Two dose-limiting toxicities occurred and 11 of 27 had an all-cause grade 3 or worse event. Of 25 with a scan response assessment, one appendiceal cancer had a partial response and ten had stable disease, including four pancreatic cancers. The paper did not report tumour-lesion dosimetry and states that scan uptake could not be robustly related to response. Novartis funded the study and took part in design, data analysis and publication." }, { "id": "mcdonald-2024-germline-protocol", "type": "paper", "title": "Development of a Novel Protocol for Germline Testing in Pancreatic Cancer", "authors": [ "McDonald, Hannah G.", "Kennedy, Amanda", "Solomon, Allison L.", "et al." ], "publisher": "Annals of Surgical Oncology", "date": "2024-08-12", "url": "https://pubmed.ncbi.nlm.nih.gov/39133448/", "doi": "10.1245/s10434-024-16011-3", "pmid": "39133448", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Physician-led protocol report. Three hundred ten patients, 81.5%, completed testing; physician compliance was 82.6% and patient compliance after an order was 98.7%. Forty-four, 14.2%, had a pathogenic variant and 83, 26.8%, had a variant of uncertain significance. All pathogenic-result patients were contacted, but family cascade completion was not reported." }, { "id": "mdanderson-chari-profile-2026", "type": "institution", "title": "Suresh T. Chari — faculty profile", "authors": [ "University of Texas MD Anderson Cancer Center" ], "publisher": "University of Texas MD Anderson Cancer Center", "date": null, "url": "https://faculty.mdanderson.org/profiles/suresh_chari.html", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Live institutional profile checked for current appointment and listed work in new-onset diabetes, early detection, biomarkers, electronic records and imaging AI." }, { "id": "mehdi-2024-kras-wildtype-case-series", "type": "paper", "title": "Chemotherapy-free treatment targeting fusions and driver mutations in KRAS wild-type pancreatic ductal adenocarcinoma", "authors": [ "Mehdi, M", "et al." ], "publisher": "Therapeutic Advances in Medical Oncology", "date": "2024-01-01", "url": "https://doi.org/10.1177/17588359241253113", "doi": "10.1177/17588359241253113", "pmid": "38770091", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Single-centre review found 14 advanced KRAS-wild-type cases among 236 PDAC patients. Five received matched targeted treatment and three had reported durable benefit: 17+ months for EGFR-altered disease, 11 months for RET fusion and 18+ months for a BRAF in-frame deletion. Case selection, varied treatment lines and no control prevent a survival estimate." }, { "id": "melisi-2018-galunisertib", "type": "paper", "title": "Galunisertib plus gemcitabine vs. gemcitabine for first-line treatment of patients with unresectable pancreatic cancer", "authors": [ "Melisi, Davide", "Garcia-Carbonero, Rocio", "Macarulla, Teresa", "et al." ], "publisher": "British Journal of Cancer", "date": "2018-11-01", "url": "https://pubmed.ncbi.nlm.nih.gov/30318515/", "doi": "10.1038/s41416-018-0246-z", "pmid": "30318515", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Randomized, double-blind phase 2 in first-line unresectable PDAC: galunisertib plus gemcitabine n=104 versus placebo plus gemcitabine n=52. Median OS was 8.9 versus 7.1 months, HR 0.79 with 95% credible interval 0.59-1.09 and posterior probability HR below 1 of 0.93. The interval included no effect." }, { "id": "melo-2015-gpc1-exosomes", "type": "paper", "title": "Glypican-1 identifies cancer exosomes and detects early pancreatic cancer", "authors": [ "Melo, Sonia A.", "Luecke, Linda B.", "Kahlert, Christoph", "et al." ], "publisher": "Nature", "date": "2015-06-24", "url": "https://pubmed.ncbi.nlm.nih.gov/26106858/", "doi": "10.1038/nature14581", "pmid": "26106858", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Originating-lab case-control study claimed 100% sensitivity and specificity in discovery and validation sets. Nature published a 2022 author correction covering methods, figures and calculations; the authors stated that conclusions were unchanged." }, { "id": "metrangolo-2025-upar-adc", "type": "paper", "title": "Targeting uPAR with an antibody-drug conjugate suppresses tumour growth and reshapes the immune landscape in pancreatic cancer models", "authors": [ "Metrangolo, Virginia", "Hansen Blomquist, Michaela", "Dutta, Ananya", "et al." ], "publisher": "Science Advances", "date": "2025-01-17", "url": "https://pubmed.ncbi.nlm.nih.gov/39823326/", "doi": "10.1126/sciadv.adq0513", "pmid": "39823326", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Primary laboratory and animal-model paper. Used to map a current ADC route; it provides no human PDAC safety or efficacy evidence." }, { "id": "mhra-2020-dpd-testing", "type": "regulator", "title": "DPD testing recommended before fluorouracil, capecitabine or tegafur", "authors": [ "Medicines and Healthcare products Regulatory Agency" ], "publisher": "UK Government", "date": "2020-10-22", "url": "https://www.gov.uk/drug-safety-update/5-fluorouracil-intravenous-capecitabine-tegafur-dpd-testing-recommended-before-initiation-to-identify-patients-at-increased-risk-of-severe-and-fatal-toxicity", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "UK safety instruction says to test every patient before systemic treatment, avoid the drugs in complete deficiency, consider a reduced start in partial deficiency and continue close monitoring because a negative test does not remove severe-toxicity risk. It also warns that the four common variants have limited frequency evidence outside European populations." }, { "id": "middleton-2014-telovac", "type": "paper", "title": "Gemcitabine and capecitabine with or without telomerase peptide vaccine GV1001 in patients with locally advanced or metastatic pancreatic cancer (TeloVac): an open-label, randomised, phase 3 trial", "authors": [ "Middleton, Gary", "Silcocks, Paul", "Cox, Trevor", "et al." ], "publisher": "The Lancet Oncology", "date": "2014-06-19", "url": "https://pubmed.ncbi.nlm.nih.gov/24954781/", "doi": "10.1016/S1470-2045(14)70236-0", "pmid": "24954781", "trialId": "ISRCTN4382138", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Open-label phase 3 trial at 51 UK hospitals. It randomized 1,062 untreated people with locally advanced or metastatic PDAC to gemcitabine/capecitabine alone, followed by GV1001, or concurrent GV1001. Median OS was 7.9, 6.9 and 8.4 months. Sequential versus control HR was 1.19, 98.25% CI 0.97-1.48, P=0.05; concurrent versus control HR was 1.05, 98.25% CI 0.85-1.29, P=0.64." }, { "id": "mirzaian-2026-low-risk-cysts", "type": "paper", "title": "Pancreatic Cancer Risk in Patients With Low-Risk Cystic Lesions", "authors": [ "Haj Mirzaian, Arya", "Abbasi, Nooshin", "Kambadakone, Avinash R.", "et al." ], "publisher": "JAMA Network Open", "date": "2026-05-01", "url": "https://pubmed.ncbi.nlm.nih.gov/42160052/", "doi": "10.1001/jamanetworkopen.2026.13808", "pmid": "42160052", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Health-system cohort of 6,064 low-risk pancreatic cystic lesions contributing 20,145 person-years. Thirty-eight participants developed the combined endpoint of HGD or pancreatic cancer; 26 arose at the cyst site and 12 elsewhere. Follow-up was based on routine care rather than a fixed randomized protocol." }, { "id": "mit-2026-barzilay-aacr", "type": "institution", "title": "2026 AACR Annual Meeting Recap: MIT AI Leader Regina Barzilay Tapped for Opening Keynote", "authors": [ "MIT Jameel Clinic" ], "publisher": "Massachusetts Institute of Technology", "date": "2026-04-21", "url": "https://jclinic.mit.edu/2026-aacr-annual-meeting-recap-regina-barzilay/", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Institutional recap used only to map current model families and capabilities, including SPARC pathology, risk models and drug discovery. It is not independent clinical validation." }, { "id": "mit-barzilay-profile-2026", "type": "institution", "title": "Regina Barzilay", "authors": [ "MIT Computer Science and Artificial Intelligence Laboratory" ], "publisher": "Massachusetts Institute of Technology", "date": null, "url": "https://www.regina.csail.mit.edu/", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Live faculty page checked for current title and work in clinical AI, cancer risk prediction, disease progression and machine-learning drug discovery." }, { "id": "mizuno-2024-hpb-prehab-rct", "type": "paper", "title": "Goal-directed prehabilitation before hepatobiliary and pancreatic surgery: randomized trial", "authors": [ "Mizuno, Y", "et al." ], "publisher": "Surgery", "date": "2024-08-01", "url": "https://doi.org/10.1016/j.surg.2024.04.024", "doi": "10.1016/j.surg.2024.04.024", "pmid": "38755033", "trialId": "UMIN000038791", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Randomized 180 major hepatobiliary or pancreatic surgical patients; 144 entered the main analysis. Six-minute walk distance improved 31 metres with step goals and 27 metres with conventional physical and nutritional prehabilitation, p=0.633. The comparison did not show added functional benefit and mixed operations limit pancreatic inference." }, { "id": "mo-2026-find-crc-surveillance", "type": "paper", "title": "Dynamic Circulating Tumor DNA Methylation Monitoring Guiding Postoperative Surveillance in Nonmetastatic Colorectal Cancer: A Prospective, Randomized, Phase III FIND Trial", "authors": [ "Mo, Shaobo", "Zhou, Chaoqiang", "Ma, Maoguang", "et al." ], "publisher": "Journal of Clinical Oncology", "date": "2026-07-29", "url": "https://pubmed.ncbi.nlm.nih.gov/42525894/", "doi": "10.1200/JCO-25-03009", "pmid": "42525894", "trialId": "NCT05904665", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Randomized phase III surveillance trial in 584 eligible patients. A positive methylation ctDNA result triggered immediate CT. Curative-intent treatment among recurrent patients rose from 23.6% to 48.1%, but median follow-up was 23.3 months and long-term survival benefit remains unreported. Innovation Biomed affiliations were present among authors." }, { "id": "mohile-2021-gap70", "type": "paper", "title": "GAP70+: geriatric assessment and management for treatment toxicity in older adults with advanced cancer", "authors": [ "Mohile, Supriya G.", "Mohamed, Mostafa R.", "Xu, Huiwen", "et al." ], "publisher": "The Lancet", "date": "2021-11-05", "url": "https://pubmed.ncbi.nlm.nih.gov/34741815/", "doi": "10.1016/S0140-6736(21)01789-X", "pmid": "34741815", "trialId": "NCT02054741", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "US cluster-randomized trial across 40 community oncology practice clusters and 718 people aged at least 70 with incurable mixed cancers and at least one impaired assessment domain. Tailored assessment summaries and management recommendations reduced grade 3-5 toxicity from 263/369, 71%, to 177/349, 51%; RR .74, 95% CI .64-.86. Falls fell from 21% to 12%. Cancer types were mixed, so this is strong action evidence for the care system but not a PDAC-specific survival result." }, { "id": "mollazadegan-2025-pnet-recharacterization", "type": "journal-article", "title": "The impact of re-characterizing metastatic pancreatic neuroendocrine tumors: A prospective study", "authors": [ "Mollazadegan, Katrin", "Botling, Johan", "Skogseid, Britt", "et al." ], "publisher": "Journal of Neuroendocrinology", "date": "2025-08", "url": "https://pubmed.ncbi.nlm.nih.gov/40325349/", "doi": "10.1111/jne.70040", "pmid": "40325349", "trialId": "NCT03130205", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Prospective single-centre cohort of 21 progressive metastatic panNET patients. Adequate repeat biopsy was obtained in 16; Ki-67 increased in 13/16 and 8/16 changed from grade 2 to 3. Re-characterization produced a clinically significant change in 13/21 and treatment changed in seven. There was no randomized test of whether re-characterization improved outcomes." }, { "id": "moore-2007-erlotinib-gemcitabine", "type": "paper", "title": "Erlotinib plus gemcitabine compared with gemcitabine alone in patients with advanced pancreatic cancer", "authors": [ "Moore, Malcolm J.", "Goldstein, David", "Hamm, John", "et al." ], "publisher": "Journal of Clinical Oncology", "date": "2007-04-23", "url": "https://pubmed.ncbi.nlm.nih.gov/17452677/", "doi": "10.1200/JCO.2006.07.9525", "pmid": "17452677", "trialId": "NCT00026338", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Double-blind international phase 3 trial in 569 patients. Statistically significant overall-survival result had a 0.33-month median difference. OS was a secondary endpoint in later clinical interpretation but the paper states it was primary." }, { "id": "moyer-2024-charm-followup", "type": "paper", "title": "Successful EUS-guided pancreatic cyst chemoablation safely allows reduction in the frequency of radiographic surveillance: long-term follow-up of randomized prospective data", "authors": [ "Moyer, Matthew T.", "Heinle, James Westley", "Rhoades, Sydney E.", "et al." ], "publisher": "Gastrointestinal Endoscopy", "date": "2023-12-11", "url": "https://pubmed.ncbi.nlm.nih.gov/38092127/", "doi": "10.1016/j.gie.2023.12.008", "pmid": "38092127", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Long-term follow-up of 52 eligible completers from two randomized ChARM studies. Complete cyst-volume response occurred in 36 and partial response in 11; no cyst-associated malignancy was observed. It selected trial completers with at least two years of complete follow-up and was far too small to establish cancer prevention." }, { "id": "mpmmcc-varanasi-profile-2026", "type": "institution-profile", "title": "Mahamana Pandit Madan Mohan Malaviya Cancer Centre and Homi Bhabha Cancer Hospital, Varanasi", "authors": [ "Tata Memorial Centre" ], "publisher": "Tata Memorial Centre", "date": "2026-05-01", "url": "https://mpmmcc.tmc.gov.in/about/history", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Official institutional profile. The linked Varanasi units provide solid-tumour care, training and oncology research under Tata Memorial Centre and include 531 beds combined. Capability claims are checked against the prospective gallbladder cohort." }, { "id": "msk-balachandran-profile-2026", "type": "institution", "title": "Vinod Balachandran: Research Overview", "authors": [ "Memorial Sloan Kettering Cancer Center" ], "publisher": "Memorial Sloan Kettering Cancer Center", "date": null, "url": "https://www.mskcc.org/research-areas/labs/vinod-balachandran/overview", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Live institutional laboratory page checked for current role, rare-survivor programme, personalised vaccine work, lymphoneogenesis and the Olayan Center's in-house design-to-treatment capability." }, { "id": "msk-iacobuzio-donahue-profile-2026", "type": "institution", "title": "The Christine Iacobuzio-Donahue Lab", "authors": [ "Memorial Sloan Kettering Cancer Center" ], "publisher": "Memorial Sloan Kettering Cancer Center", "date": null, "url": "https://www.mskcc.org/research-areas/labs/christine-iacobuzio-donahue", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Live institutional laboratory page checked for current leadership, focus and annual disclosures." }, { "id": "msk-oreilly-profile-2026", "type": "institution", "title": "Eileen M. O'Reilly — gastrointestinal medical oncologist profile", "authors": [ "Memorial Sloan Kettering Cancer Center" ], "publisher": "Memorial Sloan Kettering Cancer Center", "date": null, "url": "https://www.mskcc.org/cancer-care/doctors/eileen-o-reilly", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Live institutional profile checked for current leadership, disease scope and active trials. The page also provides current annual industry disclosures, which must be read alongside company-funded trial results." }, { "id": "msk-park-profile-2026", "type": "institution", "title": "Wungki Park — gastrointestinal medical oncologist profile", "authors": [ "Memorial Sloan Kettering Cancer Center" ], "publisher": "Memorial Sloan Kettering Cancer Center", "date": null, "url": "https://www.mskcc.org/cancer-care/doctors/wungki-park", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Live institutional profile checked for KRAS-targeted and immune research, current trials, grants and listed disclosures." }, { "id": "mukherjee-2026-redmod", "type": "paper", "title": "Next-generation AI for visually occult pancreatic cancer detection in a low-prevalence setting with longitudinal stability and multi-institutional generalisability", "authors": [ "Mukherjee, Sovanlal", "et al." ], "publisher": "Gut", "date": "2026-04-01", "url": "https://pubmed.ncbi.nlm.nih.gov/42049489/", "doi": "10.1136/gutjnl-2025-337266", "pmid": "42049489", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Retrospective case-control radiomics study. Its independent test prevalence was about 1 case to 6 controls, roughly 14%, which is low only relative to a balanced case-control set and is far above population pancreatic-cancer prevalence." }, { "id": "mulcahy-2021-epoch-radioembolization", "type": "paper", "title": "Radioembolization with chemotherapy for colorectal liver metastases: a randomized, open-label, international, multicenter, phase III trial", "authors": [ "Mulcahy, Mary F.", "Mahvash, Armeen", "Pracht, Marc", "et al." ], "publisher": "Journal of Clinical Oncology", "date": "2021-09-20", "url": "https://pubmed.ncbi.nlm.nih.gov/34541864/", "doi": "10.1200/JCO.21.01839", "pmid": "34541864", "trialId": "NCT01483027", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "EPOCH phase 3, n=428, second-line colorectal liver metastases. Radioembolization improved PFS and liver PFS but not OS; grade 3 adverse events increased. Sponsor ties and an employed author were disclosed." }, { "id": "muragaki-2026-k912-sonodynamic", "type": "paper", "title": "First-in-human safety and efficacy study on high-intensity focused ultrasound and micellar nanoparticle-encapsulated epirubicin K-912 for refractory abdominal cancers", "authors": [ "Muragaki, Yoshihiro", "Sofuni, Atsushi", "Okamoto, Jun", "et al." ], "publisher": "Ultrasound in Medicine & Biology", "date": "2026-03-01", "url": "https://doi.org/10.1016/j.ultrasmedbio.2025.10.011", "doi": "10.1016/j.ultrasmedbio.2025.10.011", "pmid": "41318252", "trialId": "UMIN000027283", "patentId": null, "grantId": "17hk0102028h0003", "accessed": "2026-09-15", "notes": "Single-centre 3+3 dose study of MS-2 focused ultrasound one day after K-912 nanoparticle epirubicin. Eleven of 12 participants had stage IV pancreatic cancer and one had cholangiocarcinoma; all had peritoneal metastases. At one month, 4/12 had complete and 5/12 partial scan-defined coagulation, 5/12 had tumour-size reduction, 8/12 disease control and 2/6 pain improvement. No grade 3 or worse event occurred through 30 days; one grade 1 burn and one grade 2 biliary laboratory abnormality were attributed to ultrasound. Median survival among the 11 pancreatic participants calculated from the individual table was 140 days. Sonire supported medical writing." }, { "id": "murphy-2019-losartan-lapc", "type": "paper", "title": "Total neoadjuvant FOLFIRINOX with losartan followed by chemoradiotherapy for locally advanced pancreatic cancer", "authors": [ "Murphy, Janet E.", "Wo, Jennifer Y.", "Ryan, David P.", "et al." ], "publisher": "JAMA Oncology", "date": "2019-05-23", "url": "https://pubmed.ncbi.nlm.nih.gov/31145418/", "doi": "10.1001/jamaoncol.2019.0892", "pmid": "31145418", "trialId": "NCT01821729", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Single-arm phase 2 at one academic hospital, n=49. Thirty-nine completed eight FOLFIRINOX-losartan cycles, 42 had attempted surgery, and 34/49=69% had an R0 resection. Median PFS was 17.5 months and OS 31.4 months. Losartan contribution cannot be separated from chemotherapy, radiation, surgery or selection." }, { "id": "musher-2026-multiantigen-t-cells", "type": "paper", "title": "Autologous multiantigen-targeted T cell therapy for pancreatic cancer: a phase 1/2 trial", "authors": [ "Musher, Benjamin L.", "Vasileiou, Spyridoula", "Smaglo, Brandon G.", "et al." ], "publisher": "Nature Medicine", "date": "2025-12-01", "url": "https://pubmed.ncbi.nlm.nih.gov/41482561/", "doi": "10.1038/s41591-025-04043-5", "pmid": "41482561", "trialId": "NCT03192462", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Nonrandomised three-arm phase 1/2. Of 56 procured patients, 37 were infused. Safety and feasibility were primary; efficacy was exploratory. Authors reported extensive cell-therapy company interests." }, { "id": "nakamura-2015-biliary-genomics", "type": "paper", "title": "Genomic spectra of biliary tract cancer", "authors": [ "Nakamura, Hiromi", "Arai, Yasuhito", "Totoki, Yasushi", "et al." ], "publisher": "Nature Genetics", "date": "2015-08-10", "url": "https://pubmed.ncbi.nlm.nih.gov/26258846/", "doi": "10.1038/ng.3375", "pmid": "26258846", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Exome and transcriptome analysis of 260 Japanese biliary tract cancers. The study separated intrahepatic, extrahepatic and gallbladder cancers and found site-associated alteration patterns." }, { "id": "nakamura-2024-galaxy-ctdna", "type": "paper", "title": "ctDNA-based molecular residual disease and survival in resectable colorectal cancer", "authors": [ "Nakamura, Yoshiaki", "Watanabe, Jun", "Akazawa, Naoto", "et al." ], "publisher": "Nature Medicine", "date": "2024-09-16", "url": "https://pubmed.ncbi.nlm.nih.gov/39284954/", "doi": "10.1038/s41591-024-03254-6", "pmid": "39284954", "trialId": "UMIN000039205", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Updated observational CIRCULATE-Japan GALAXY analysis in 2,240 patients. Natera employees were authors and the tumour-informed assay was commercial. Treatment-benefit analyses are observational, not randomised." }, { "id": "nakano-2020-antibiotics-gnp", "type": "paper", "title": "Association between the use of antibiotics and efficacy of gemcitabine plus nab-paclitaxel in advanced pancreatic cancer", "authors": [ "Nakano, Shintaro", "Komatsu, Yoshito", "Kawamoto, Yasuyuki", "et al." ], "publisher": "Medicine", "date": "2020-09-25", "url": "https://pubmed.ncbi.nlm.nih.gov/32991420/", "doi": "10.1097/MD.0000000000022250", "pmid": "32991420", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Retrospective single-centre cohort n=99; 37 received antibiotics during gemcitabine plus nab-paclitaxel and 62 did not. PFS was 5.8 versus 2.7 months, HR 0.602, but OS was 11.0 versus 8.4 months, HR 0.768, 95% CI 0.491-1.202. Antibiotic indication, timing and patient state were not randomized." }, { "id": "nakaoka-2024-kestose-pdac", "type": "paper", "title": "Efficacy of 1-kestose supplementation in patients with pancreatic ductal adenocarcinoma: a randomized controlled pilot study", "authors": [ "Nakaoka, Kazunori", "Ohno, Eizaburo", "Kuramitsu, Kento", "et al." ], "publisher": "Nutrients", "date": "2024-08-29", "url": "https://pubmed.ncbi.nlm.nih.gov/39275204/", "doi": "10.3390/nu16172889", "pmid": "39275204", "trialId": "jRCTs041200109", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Single-centre open-label pilot: 40 untreated unresectable PDAC patients randomized 20:20 to 9 g/day 1-kestose or no supplement with chemotherapy; 38 completed. No power calculation. Response was 14/20 (70%) versus 11/18 (61%) and disease control 19/20 (95%) versus 15/18 (83%), neither significantly different. Reported marker results largely tested change within each arm rather than the randomized difference in change." }, { "id": "nakashima-2022-stnm01-pdac", "type": "paper", "title": "STNM01 RNA oligonucleotide targeting CHST15 as second-line therapy for unresectable pancreatic cancer", "authors": [ "Nakashima and collaborators" ], "publisher": "ESMO Open", "date": "2022", "url": "https://pubmed.ncbi.nlm.nih.gov/36425867/", "doi": null, "pmid": "36425867", "trialId": "jRCT2031190055", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Japanese open-label phase 1/2a n=22; repeated EUS-guided local RNA oligonucleotide injections plus S-1." }, { "id": "napoli-2025-surgery-versus-oncology", "type": "paper", "title": "Survival Benefit of Surgery versus Oncology-Only Therapy in Artery-Involving Borderline Resectable and Locally Advanced Pancreatic Cancer", "authors": [ "Napoli, Niccolò", "Boggi, Ugo", "et al." ], "publisher": "Annals of Surgical Oncology", "date": "2025-12-01", "url": "https://pubmed.ncbi.nlm.nih.gov/40954390/", "doi": "10.1245/s10434-025-18212-w", "pmid": "40954390", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Retrospective international cohort of 312 people with artery-involving borderline or locally advanced PDAC judged potentially operable after chemotherapy. Median survival was 39.0 months after resection and 16.7 months with continued oncology treatment; propensity matching retained 75 pairs and a large gap. Treatment allocation depended on institutional practice or surgeon preference. Matching measured fields cannot remove surgical judgment, hidden disease, treatment response, fitness or centre effects, so this supports a randomized test rather than proving surgery caused the difference." }, { "id": "natera-2025-10k", "type": "filing", "title": "Natera 2025 Annual Report on Form 10-K", "authors": [ "Natera, Inc." ], "publisher": "U.S. Securities and Exchange Commission", "date": "2026-02-26", "url": "https://www.sec.gov/Archives/edgar/data/1604821/000110465926020881/ntra-20251231x10k.htm", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Audited filing used for product design, commercial status, LDT status, coverage limits, claims of more than 170 peer-reviewed Signatera papers, the Foresight acquisition and the distinction between MRD detection and therapy selection. The filing reports company-wide R&D of $320.678m in 2023, $404.138m in 2024 and $624.110m in 2025; it does not allocate those expenses to oncology, Signatera or a particular cancer." }, { "id": "natera-2026-signatera-pdac", "type": "company-product", "title": "Signatera for pancreatic cancer", "authors": [ "Natera, Inc." ], "publisher": "Natera, Inc.", "date": null, "url": "https://www.natera.com/oncology/signatera-advanced-cancer-detection/clinicians/signatera-for-pancreatic-cancer/", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Commercial product page used only to establish that Signatera is marketed for PDAC MRD monitoring. Clinical-effect claims are taken from the linked paper and separated from marketing." }, { "id": "nccs-liver-research-programme-2026", "type": "institution-profile", "title": "Programme in Translational and Clinical Liver Cancer Research", "authors": [ "National Cancer Centre Singapore" ], "publisher": "National Cancer Centre Singapore", "date": "2026-09-01", "url": "https://www.nccs.com.sg/research-innovation/research-labs/medical-sciences/programme-in-translation-and-clinical-liver-research", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Official programme profile. Pierce Chow leads a multi-institutional Asia-Pacific HCC programme spanning translational platforms and prospective studies including PLANet 2.0. Institutional description is paired with primary cohort evidence." }, { "id": "nci-2022-budget-factbook", "type": "government-data", "title": "NCI Budget Fact Book for fiscal year 2022", "authors": [ "National Cancer Institute" ], "publisher": "National Cancer Institute", "date": "2023-01-01", "url": "https://www.cancer.gov/about-nci/budget/fact-book/archive/fy22-fact-book.pdf", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Primary archived government report used for actual 2015 values: pancreatic cancer $125.3m, liver cancer $70.3m and colorectal cancer $209.3m." }, { "id": "nci-2023-budget-factbook", "type": "government-data", "title": "NCI Budget Fact Book for fiscal year 2023", "authors": [ "National Cancer Institute" ], "publisher": "National Cancer Institute", "date": "2024-01-01", "url": "https://www.cancer.gov/about-nci/budget/fact-book/archive/fy23-fact-book.pdf", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Primary archived government report used for actual 2016–2017 research-area values: pancreatic cancer $152.6m and $178.3m; liver cancer $75.7m and $72.7m; colorectal cancer $212.2m and $208.4m. The report repeats NCI's warning that categories overlap." }, { "id": "nci-2026-research-funding", "type": "government-data", "title": "NCI Budget Fact Book: funding for research areas, fiscal years 2018–2024", "authors": [ "National Cancer Institute" ], "publisher": "National Cancer Institute", "date": "2026-06-04", "url": "https://www.cancer.gov/about-nci/budget/fact-book/data/research-funding", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Primary government table. Values are dollars in millions. NCI warns that research-area categories overlap and do not sum to the NCI budget. The pancreatic-cancer row is 182.1, 187.0, 178.5, 218.1, 226.8, 246.0 and 246.0 for 2018–2024; colorectal and liver rows were also checked. The last five values remain estimates in this edition." }, { "id": "nci-2026-rpg-funding", "type": "government-data", "title": "NCI Budget Fact Book: fiscal year 2025 research project grants", "authors": [ "National Cancer Institute" ], "publisher": "National Cancer Institute", "date": "2026-06-04", "url": "https://www.cancer.gov/about-nci/budget/fact-book/extramural-programs/rpg", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Primary NCI mechanism and competition data. In FY2025, competing new, renewal and supplement RPGs had 11,711 requests and 1,098 awards, a 9.4% success rate; average competing cost was $676,000. NCI warns FY2025 is not comparable with prior years because of rescinded interim paylines and multi-year upfront funding." }, { "id": "nci-2026-spore-funding", "type": "government-data", "title": "NCI Budget Fact Book: fiscal year 2025 SPORE grants", "authors": [ "National Cancer Institute" ], "publisher": "National Cancer Institute", "date": "2026-06-04", "url": "https://www.cancer.gov/about-nci/budget/fact-book/extramural-programs/spores", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Primary mechanism table. FY2025 amounts checked: pancreas $6,947,873; liver $4,800,010; hyperactive RAS $2,300,532; gastrointestinal $2,398,107; disparities–GI $2,455,587. These rows can overlap by disease relevance and are only the SPORE mechanism." }, { "id": "nci-cancer-terms-dictionary", "type": "regulatory", "title": "NCI Dictionary of Cancer Terms", "authors": [ "National Cancer Institute" ], "publisher": "National Cancer Institute", "date": null, "url": "https://www.cancer.gov/publications/dictionaries/cancer-terms/", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "NCI's reviewed plain-language dictionary contains more than 9,400 cancer and medical terms. The atlas uses it as the definition anchor, then adds engineering and business comparisons that are explicitly teaching aids rather than biological claims." }, { "id": "nci-maitra-edrn-2025", "type": "grant", "title": "Clinical Validation Center for Early Detection of Pancreatic Cancer", "authors": [ "National Cancer Institute Division of Cancer Prevention" ], "publisher": "National Cancer Institute", "date": "2025-09-11", "url": "https://prevention.cancer.gov/funding-and-grants/funded-grants/U01CA200468", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": "U01CA200468", "accessed": "2026-09-14", "notes": "Official active grant record: New York University awardee, project through August 2027, $1,076,354 FY2025 total cost. It covers multi-institution blinded biomarker validation and high-risk groups including germline risk, cysts and new-onset diabetes." }, { "id": "nct04526106-refocus", "type": "trial", "title": "REFOCUS: lirafugratinib in intrahepatic cholangiocarcinoma and other FGFR2-driven solid tumours", "authors": [ "Elevar Therapeutics" ], "publisher": "ClinicalTrials.gov", "date": "2026-02-27", "url": "https://clinicaltrials.gov/study/NCT04526106", "doi": null, "pmid": null, "trialId": "NCT04526106", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Registry record for the completed open-label phase 1/2 study. Actual enrolment was 490, with study completion on 5 September 2025. Sponsor changed to Elevar after licensing. No registry results were posted at the 27 February 2026 update." }, { "id": "nct05669482-ramp205", "type": "trial-registry", "title": "RAMP 205: Avutometinib and defactinib with gemcitabine and nab-paclitaxel in pancreatic cancer", "authors": [ "Verastem, Inc." ], "publisher": "ClinicalTrials.gov", "date": "2025-09-02", "url": "https://clinicaltrials.gov/study/NCT05669482", "doi": null, "pmid": null, "trialId": "NCT05669482", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Current registry record: phase 1/2, estimated n=40, active but not recruiting, single experimental programme without a randomized control." }, { "id": "nct05968326-imcode003", "type": "trial-registry", "title": "IMCODE003: Autogene cevumeran plus atezolizumab and mFOLFIRINOX versus mFOLFIRINOX after PDAC resection", "authors": [ "Genentech, Inc." ], "publisher": "ClinicalTrials.gov", "date": "2026-09-09", "url": "https://clinicaltrials.gov/study/NCT05968326", "doi": null, "pmid": null, "trialId": "NCT05968326", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Current registry record: active but not recruiting open-label randomized phase 2, estimated n=260, 87 sites, Genentech sponsor and BioNTech collaborator. The primary outcome is disease-free survival through about six years; overall survival and adverse events are secondary. Primary completion is estimated January 2031 and individual-level data may be requested for eligible studies." }, { "id": "nct06698458-alpha-dart-pdac", "type": "trial-registry", "title": "Intratumoral diffusing alpha emitters with chemotherapy for pancreatic cancer", "authors": [ "Alpha Tau Medical" ], "publisher": "ClinicalTrials.gov", "date": "2025-12-02", "url": "https://clinicaltrials.gov/study/NCT06698458", "doi": null, "pmid": null, "trialId": "NCT06698458", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Open-label safety study, up to 30 patients with locally advanced or metastatic pancreatic adenocarcinoma, combining intratumoral Alpha DaRT with chemotherapy." }, { "id": "nct07232875-hrs4642-phase3", "type": "trial-registry", "title": "HRS-4642 plus chemotherapy versus placebo plus chemotherapy in first-line KRAS G12D pancreatic cancer", "authors": [ "Jiangsu Hengrui Pharmaceuticals" ], "publisher": "ClinicalTrials.gov", "date": "2025-10-24", "url": "https://clinicaltrials.gov/study/NCT07232875", "doi": null, "pmid": null, "trialId": "NCT07232875", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Randomized controlled double-blind phase III trial of intravenous KRAS G12D inhibitor HRS-4642 with chemotherapy in advanced or metastatic pancreatic cancer." }, { "id": "nct07243262-vapor2", "type": "trial-registry", "title": "Volatile Organic Compound Assessment in Pancreatic Ductal Adenocarcinoma (VAPOR2)", "authors": [ "Imperial College London" ], "publisher": "ClinicalTrials.gov", "date": "2025-11-21", "url": "https://clinicaltrials.gov/study/NCT07243262", "doi": null, "pmid": null, "trialId": "NCT07243262", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Recruiting prospective observational cohort with estimated enrolment of 6,079. Imperial College London is sponsor, George Hanna is principal investigator and the registry lists Pancreatic Cancer UK plus 22 health-system collaborators. No accuracy or utility results are posted." }, { "id": "nct07252232-daraxonrasib-resected", "type": "trial", "title": "Study of Daraxonrasib in Patients With Resected Pancreatic Ductal Adenocarcinoma", "authors": [ "Revolution Medicines, Inc." ], "publisher": "ClinicalTrials.gov", "date": "2025-11-26", "url": "https://clinicaltrials.gov/study/NCT07252232", "doi": null, "pmid": null, "trialId": "NCT07252232", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Recruiting, sponsor-run, open-label randomised phase 3. Estimated 500 resected patients after perioperative multiagent chemotherapy, daraxonrasib versus observation; primary endpoint disease-free survival; estimated primary completion May 2029. Registry status checked 2026-09-14." }, { "id": "nct07300150-pt0511", "type": "trial-registry", "title": "A Study of PT0511 in Participants With KRAS Mutated or Amplified Advanced Solid Tumors", "authors": [ "PAQ Therapeutics" ], "publisher": "ClinicalTrials.gov", "date": "2026-05-07", "url": "https://clinicaltrials.gov/study/NCT07300150", "doi": null, "pmid": null, "trialId": "NCT07300150", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Recruiting open-label phase 1, estimated n=195 across KRAS-mutant or amplified solid tumours including pancreatic and colorectal cancer; primary completion 2028." }, { "id": "nct07409272-setidegrasib-first-line", "type": "trial", "title": "Setidegrasib plus mFOLFIRINOX or NALIRIFOX as First-Line Treatment in KRAS-G12D Metastatic Pancreatic Adenocarcinoma", "authors": [ "Astellas Pharma Global Development, Inc." ], "publisher": "ClinicalTrials.gov", "date": "2026-02-13", "url": "https://clinicaltrials.gov/study/NCT07409272", "doi": null, "pmid": null, "trialId": "NCT07409272", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Recruiting sponsor-run randomised phase 3. Estimated 614 patients, setidegrasib plus investigator-selected mFOLFIRINOX or NALIRIFOX versus matching placebo plus the same chemotherapy; quadruple masked; overall survival primary; estimated completion August 2029." }, { "id": "nct07491445-daraxonrasib-first-line", "type": "trial", "title": "Study of Daraxonrasib and Daraxonrasib plus Gemcitabine and Nab-Paclitaxel as First-Line Treatment in Metastatic Pancreatic Adenocarcinoma", "authors": [ "Revolution Medicines, Inc." ], "publisher": "ClinicalTrials.gov", "date": "2026-03-24", "url": "https://clinicaltrials.gov/study/NCT07491445", "doi": null, "pmid": null, "trialId": "NCT07491445", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Recruiting, sponsor-run, open-label randomised phase 3. Estimated 900 patients, three arms: daraxonrasib, daraxonrasib plus gemcitabine/nab-paclitaxel, or gemcitabine/nab-paclitaxel. Primary PFS and OS; estimated primary completion June 2028." }, { "id": "nct07805954-dual-ras-first-line", "type": "trial", "title": "Study of Zoldonrasib plus Daraxonrasib versus Gemcitabine and Nab-Paclitaxel as First-Line Treatment in Metastatic KRAS G12D Pancreatic Adenocarcinoma", "authors": [ "Revolution Medicines, Inc." ], "publisher": "ClinicalTrials.gov", "date": "2026-08-01", "url": "https://clinicaltrials.gov/study/NCT07805954", "doi": null, "pmid": null, "trialId": "NCT07805954", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Recruiting, sponsor-run, open-label randomised phase 3. Estimated 400 first-line KRAS G12D patients, dual RAS inhibition versus gemcitabine/nab-paclitaxel; primary PFS and OS; estimated primary completion March 2029." }, { "id": "nedo-2024-cosomil-profile", "type": "government-company-profile", "title": "Cosomil, Inc. — single-molecule liquid biopsy programme", "authors": [ "New Energy and Industrial Technology Development Organization" ], "publisher": "NEDO", "date": "2024-04-01", "url": "https://www.nedo.go.jp/activities/startups/company06.html", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": "NEDO DTSU STS", "accessed": "2026-09-14", "notes": "Japanese government startup profile: founded 2022 by Yu Kagami; technology originated from Toru Komatsu at the University of Tokyo and Rikiya Watanabe at RIKEN; ANRI partner VC; valuation undisclosed. NEDO lists a ¥499m grant decision for fiscal years 2024-2025 and planned US, stage 0/I, automation and regulatory work." }, { "id": "nell-2018-men1-surgery-watchful-waiting", "type": "paper", "title": "Management of MEN1 Related Nonfunctioning Pancreatic NETs: A Shifting Paradigm: Results From the DutchMEN1 Study Group", "authors": [ "Nell, Sjoerd", "Verkooijen, Helena M.", "Pieterman, Carolina R. C." ], "publisher": "Annals of Surgery", "date": "2018-06-01", "url": "https://pubmed.ncbi.nlm.nih.gov/28257328/", "doi": "10.1097/SLA.0000000000002183", "pmid": "28257328", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "DutchMEN1 population cohort covering more than 90% of the national MEN1 population. Among 152 patients, 53 had surgery and 99 watchful waiting. After time-dependent propensity-restricted adjustment, surgery was not associated with significantly less liver metastasis or death: HR 0.73, 95% CI 0.25-2.11. The under-2-cm and 2-3-cm estimates were extremely imprecise; five of six observed tumours above 3 cm reached liver metastasis or death." }, { "id": "neuzillet-2023-apacap-rct", "type": "paper", "title": "APACaP randomized trial of adapted physical activity in advanced pancreatic cancer", "authors": [ "Neuzillet, Cindy", "et al." ], "publisher": "Journal of the National Comprehensive Cancer Network", "date": "2023-12-01", "url": "https://doi.org/10.6004/jnccn.2023.7065", "doi": "10.6004/jnccn.2023.7065", "pmid": "38081120", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Randomized 313 of 326 included advanced-PDAC patients to standard care or a 16-week home programme. The primary week-16 analysis retained 172 and missed all three quality-of-life endpoints: adjusted differences -0.98 global health, -2.08 physical function and 4.16 fatigue, all p at least 0.17. Secondary longitudinal analyses among 259 with baseline and at least one follow-up found improvements in several dimensions, making death and questionnaire missingness central." }, { "id": "ngohuang-2023-pancfit-rct", "type": "paper", "title": "PancFit randomized trial of home exercise during neoadjuvant pancreatic treatment", "authors": [ "Ngo-Huang, An T", "et al." ], "publisher": "Annals of Surgery", "date": "2023-07-01", "url": "https://doi.org/10.1097/SLA.0000000000005878", "doi": "10.1097/SLA.0000000000005878", "pmid": "37026453", "trialId": "NCT03187951", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Randomized 151 pancreatic patients during neoadjuvant treatment. Both groups increased six-minute walking distance; the between-group exercise and physical-activity amounts were similar, while strength sessions increased more in the prescribed group. Quality of life and clinical outcomes did not significantly differ. The registry reports 152 actual enrollment and estimated final completion in December 2026." }, { "id": "nguyen-2026-fap-cd40-pd1-il2v", "type": "paper", "title": "FAP-CD40 and PD1-IL2v combination therapy reprograms immunologically cold tumors through de novo intratumoral T cell-dendritic cell clusters", "authors": [ "Nguyen, Thuy Trinh", "Gómez, Harold", "Lutge, Mechthild", "et al." ], "publisher": "Journal for ImmunoTherapy of Cancer", "date": "2026-05-28", "url": "https://pubmed.ncbi.nlm.nih.gov/42208978/", "doi": "10.1136/jitc-2025-014620", "pmid": "42208978", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Roche-led mouse PDAC study combining proprietary FAP-CD40 and PD1-IL2v molecules. The combination formed intratumour T-cell and dendritic-cell clusters and produced regression in KPC tumours across repeated experiments. It has no human PDAC treatment result. Most authors were Roche employees, several were inventors on relevant patent applications and several held Roche stock." }, { "id": "nhs-england-2024-hpb-pancreatic-service", "type": "government-guidance", "title": "Hepato-pancreatic biliary pancreatic and periampullary cancers service specification 2325", "authors": [ "NHS England" ], "publisher": "NHS England", "date": "2024-09-01", "url": "https://www.england.nhs.uk/wp-content/uploads/2024/09/2325-hpb-pancreatic-service-specification.pdf", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Current specialist-service specification checked. It requires catchments of two to four million, at least 80 pancreatic surgical procedures for neoplastic or suspected neoplastic disease per team per year, specialist MDT planning, clear network roles, prompt referral, 24/7 interventional and postoperative cover, prehabilitation, enhanced recovery and outcome reporting. These are commissioned requirements, not proof that every centre achieves them." }, { "id": "nhs-england-2026-cancer-genomic-test-directory", "type": "government-guidance", "title": "National genomic test directory for cancer: non-central nervous system, version 16", "authors": [ "NHS England" ], "publisher": "NHS England", "date": "2026-07-16", "url": "https://www.england.nhs.uk/publication/national-genomic-test-directories/", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Current M219 pancreatic route includes BRCA1/2 small-variant testing, an RNA structural-variant panel whose required targets are NTRK1/2/3, DPYD testing and MSI testing. The workbook permits reporting additional oncogenic structural variants if covered by the delivering RNA panel and names ALK, BRAF, ROS1, MET, RET and FGFR1/2/3 as additional content. NRG1 is not named. This supports a coverage audit, not a claim that no NHS laboratory can detect NRG1." }, { "id": "nhs-england-2026-r367", "type": "government-guidance", "title": "National Genomic Test Directory: R367 Inherited pancreatic cancer eligibility criteria, version 9.1", "authors": [ "NHS England" ], "publisher": "NHS England", "date": "2026-05-20", "url": "https://www.england.nhs.uk/wp-content/uploads/2018/08/rare-and-inherited-disease-eligibility-criteria-v9.1.pdf", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Live NHS England criteria checked in version 9.1. R367 covers pancreatic adenocarcinoma diagnosed before age 60, or before 70 with specified personal or family cancer history. Testing outside those criteria may be considered by a specialist genetics multidisciplinary team. This is the commissioned eligibility rule, not evidence that excluded patients have zero actionable variants." }, { "id": "nhs-england-dpyd-implementation", "type": "health-system", "title": "Accelerating genomic medicine in the NHS", "authors": [ "NHS England" ], "publisher": "NHS England", "date": "2023-10-12", "url": "https://www.england.nhs.uk/long-read/accelerating-genomic-medicine-in-the-nhs/", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "NHS England reports DPYD testing introduced for about 38,000 patients receiving fluoropyrimidines and says it has prevented potentially fatal adverse reactions. The page does not publish a pancreatic-specific eligible, tested, timely-result, dose-action, toxicity or outcome denominator." }, { "id": "nice-2017-ire-pancreatic", "type": "government-guidance", "title": "Irreversible electroporation for treating pancreatic cancer, HTG437", "authors": [ "National Institute for Health and Care Excellence" ], "publisher": "NICE", "date": "2017-05-03", "url": "https://www.nice.org.uk/guidance/htg437/chapter/1-Recommendations", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Current listed NICE procedure guidance, originally issued in 2017. It states that pancreatic IRE evidence is inadequate in quantity and quality and limits use to research. It asks randomized trials to measure local control, survival, pain and quality of life. The procedure uses several needle electrodes, high-voltage direct-current pulses, general anaesthesia, full muscle blockade and cardiac synchronization." }, { "id": "nice-2018-ng85-pancreatic", "type": "government-guidance", "title": "Pancreatic cancer in adults: diagnosis and management, NG85", "authors": [ "National Institute for Health and Care Excellence" ], "publisher": "NICE", "date": "2018-02-07", "url": "https://www.nice.org.uk/guidance/ng85/chapter/Recommendations", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Current England guideline page, originally published in 2018 and checked after its September 2025 review. It requires specialist pancreatic MDT decision-making, pancreas-protocol CT, offers FDG-PET/CT for CT-localized disease receiving treatment, and says laparoscopy with laparoscopic ultrasound should be considered when small-volume peritoneal or liver spread remains suspected before possible resection. It recommends adjuvant treatment as soon as the patient can tolerate all six cycles and ongoing specialist care after resection. The 2026 national audit describes the guidance's neoadjuvant restriction as outdated. The selective staging rule is guidance, not a comparative trial or national delivery measure." }, { "id": "nih-2026-r01", "type": "government-guidance", "title": "NIH Research Project Grant R01", "authors": [ "National Institutes of Health" ], "publisher": "National Institutes of Health", "date": "2026-01-01", "url": "https://grants.nih.gov/funding/activity-codes/R01", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Primary grant guidance. R01 awards normally cover one to five annual budget periods; budgets are generally not capped but must fit the work. Domestic applications at $250,000 or less direct cost per year use the modular format; individual opportunities can impose different limits." }, { "id": "noel-2020-ccr2-pdac", "type": "paper", "title": "Phase 1b study of the CCR2 antagonist PF-04136309 with nab-paclitaxel and gemcitabine in metastatic PDAC", "authors": [ "Noel and collaborators" ], "publisher": "Investigational New Drugs", "date": "2020", "url": "https://pubmed.ncbi.nlm.nih.gov/31297636/", "doi": null, "pmid": "31297636", "trialId": "NCT02732938", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Phase 1b n=21; 24% pulmonary toxicity and no efficacy signal beyond chemotherapy." }, { "id": "noonan-2016-pelareorep-pdac", "type": "paper", "title": "Randomized phase 2 trial of the oncolytic virus pelareorep in upfront treatment of metastatic pancreatic adenocarcinoma", "authors": [ "Noonan, Anne M.", "Farren, Matthew R.", "Geyer, Stacy M.", "et al." ], "publisher": "Molecular Therapy", "date": "2016-06-01", "url": "https://pubmed.ncbi.nlm.nih.gov/27039845/", "doi": "10.1038/mt.2016.66", "pmid": "27039845", "trialId": "NCT01280058", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Primary randomized phase 2: 73 evaluable patients; pelareorep plus carboplatin/paclitaxel PFS 4.9 months versus 5.2 months with chemotherapy, P=.6. Immune changes occurred without clinical benefit." }, { "id": "noor-2026-pdac-vte-implementation", "type": "meeting-abstract", "title": "Improving venous thromboembolism prophylaxis in pancreatic adenocarcinoma", "authors": [ "Noor, Anika", "et al." ], "publisher": "European Hematology Association 2026 Congress Library", "date": "2026-05-12", "url": "https://library.ehaweb.org/eha/2026/eha-2026/4210917/anika.noor.improving.venous.thromboembolism.prophylaxis.in.pancreatic.html", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Small single-centre quality project. Among 14 surveyed staff, 36% routinely offered prophylaxis, 29% never did, 64% cited eligibility uncertainty and 43% bleeding concern. Education changed eligible prescribing from 0/14 to 1/16; an electronic prompt was followed by 9/16, 56%, receiving prophylaxis in January-February 2026. Follow-up was short and the study cannot estimate VTE or bleeding effects." }, { "id": "nordgard-2022-unresectable-dtc", "type": "paper", "title": "Prognostic value of disseminated tumor cells in unresectable pancreatic ductal adenocarcinoma", "authors": [ "Nordgård, Oddmund", "et al." ], "publisher": "BMC Cancer", "date": "2022-06-03", "url": "https://pubmed.ncbi.nlm.nih.gov/35659265/", "doi": "10.1186/s12885-022-09714-x", "pmid": "35659265", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Prospective observational study of 48 people with locally advanced, n=11, or metastatic, n=37, PDAC. Bone marrow was sampled before and after two months of chemotherapy. An mRNA panel classified 15/48, 31%, positive before treatment and 8/25, 32%, during treatment. Pretreatment positivity associated with PFS and OS HR 2.0; during-treatment positivity was not prognostic. Marker choice produced very large HRs and thresholds came from 30 healthy marrows. The study did not measure dormancy or treatment utility." }, { "id": "notta-2016-punctuated-evolution", "type": "paper", "title": "A renewed model of pancreatic cancer evolution based on genomic rearrangement patterns", "authors": [ "Notta, Faiyaz", "Chan-Seng-Yue, Michelle", "Lemire, Mathieu", "et al." ], "publisher": "Nature", "date": "2016-10-12", "url": "https://pubmed.ncbi.nlm.nih.gov/27732578/", "doi": "10.1038/nature19823", "pmid": "27732578", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Analysis of more than 100 tumour-enriched whole genomes supported punctuated genomic change in a substantial subset of PDAC, challenging a uniformly gradual progression model." }, { "id": "novelli-2026-tigerpac-pk", "type": "paper", "title": "Pharmacokinetic and pharmacodynamic sub-study of trans-arterial vs. intravenous gemcitabine in the TIGeR-PaC phase 3 clinical trial", "authors": [ "Novelli, Paula M.", "Zervos, Emmanuel E.", "Zureikat, Amer H.", "et al." ], "publisher": "Cancer Chemotherapy and Pharmacology", "date": "2026-08-17", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC13478025/", "doi": "10.1007/s00280-026-04939-0", "pmid": "42604879", "trialId": "NCT03257033", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Randomized blood pharmacology sub-study across six sites: 11 intra-arterial and five intravenous participants. Mean active-gemcitabine blood AUC was 4.9 versus 8.8 hour-micrograms/mL, P=0.018; mean inactive dFdU AUC was 46.3 versus 37.1, P=0.097. Three people with normal baseline CA19-9 were excluded from the marker analysis; among the remaining eight intra-arterial patients, dFdU AUC correlated with two-week CA19-9 change, r=-0.75, P=0.034, with a stated outlier. Blood was sampled, not tumour tissue or intracellular active metabolites. RenovoRx funded the study; an employee-shareholder and a paid consultant were authors." }, { "id": "novocure-2026-panova4-topline", "type": "company", "title": "Novocure announces topline PANOVA-4 results", "authors": [ "NovoCure Limited" ], "publisher": "NovoCure Limited", "date": "2026-03-26", "url": "https://investor.novocure.com/news-releases/news-release-details/novocure-announces-positive-topline-results-phase-2-panova-4", "doi": null, "pmid": null, "trialId": "NCT06390059", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Company topline release, not a peer-reviewed report or posted result table. Among 78 analysed participants, the four-part regimen produced 74.4% disease control versus 48% in the old MPACT trial used as a historical control, a one-sided P value below 0.001. Objective response was 34.6%, 95% CI 24.2-46.2, and median overall survival was 9.7 months, 95% CI 7.9-12.7. The single arm cannot separate the device, atezolizumab, chemotherapy or case-mix effects." }, { "id": "novocure-2026-q2-10q", "type": "financial", "title": "Novocure quarterly report for the period ended June 30, 2026", "authors": [ "NovoCure Limited" ], "publisher": "U.S. Securities and Exchange Commission", "date": "2026-07-23", "url": "https://www.sec.gov/Archives/edgar/data/1645113/000164511326000062/nvcr-20260630.htm", "doi": null, "pmid": null, "trialId": "NCT03377491", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Primary company filing. At 30 June 2026 Novocure reported $440.6m cash, cash equivalents and short-term investments, $200m principal borrowed under its secured facility, $357.639m six-month revenue, $86.796m six-month net loss and $5.385m six-month operating cash use. U.S. Optune Pax uptake was 285 active patients at quarter end and 586 prescriptions in the first half; one German prescription made the total 587. Q2 Optune Pax revenue recognized was $1.6m." }, { "id": "npaca-2026-state-of-nation", "type": "government-audit", "title": "National Pancreatic Cancer Audit State of the Nation Report 2026", "authors": [ "National Pancreatic Cancer Audit" ], "publisher": "National Cancer Audit Collaborating Centre and Healthcare Quality Improvement Partnership", "date": "2026-09-10", "url": "https://www.hqip.org.uk/wp-content/uploads/2026/09/REF753_NATCAN-NPaCA-SoN_FINAL_v1.0.pdf", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Current national audit of 17,672 people diagnosed with exocrine pancreatic cancer in England in 2022-2023 and 951 in Wales in 2023-2024. It reports 23 English HPB specialist centres, median referral-to-treatment of 74 days, 62% receipt of chemotherapy or chemoradiotherapy within 14 weeks after Whipple with centre IQR 54-68%, and 96.4% 90-day survival after Whipple. It measures treatment receipt, not completion of every planned cycle. It is observational routine-data evidence and notes missing fields and COVID-era effects." }, { "id": "nyulangone-maitra-profile-2026", "type": "institution", "title": "Anirban Maitra, MD", "authors": [ "NYU Langone Health" ], "publisher": "NYU Langone Health", "date": null, "url": "https://nyulangone.org/doctors/1225070881/anirban-maitra", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Live institutional profile checked after resolving an outdated MD Anderson affiliation. It lists Maitra as Perlmutter Cancer Center director, associate dean, pathology and medicine professor, with pancreatic early-detection, interception and treatment work." }, { "id": "oberg-2026-pcdc-biomarker-bakeoff", "type": "paper", "title": "Pancreatic Cancer Detection Consortium Biomarker Bakeoff: A Phase II Blinded Biomarker Validation and Panel Discovery Study", "authors": [ "Oberg, Ann L.", "Bamlet, William R.", "Izmirlian, Grant", "et al." ], "publisher": "Clinical Cancer Research", "date": "2026-06-15", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC13266336/", "doi": "10.1158/1078-0432.CCR-25-4061", "pmid": "41870276", "trialId": "NCT06271291", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Blinded head-to-head phase-II comparison of eight panels and CA19-9 using 140 known stage-I-to-IV PDAC cases and 140 controls from three tertiary centres. CA19-9/FUT2/3 reached AUC 0.963 versus 0.917 for CA19-9. The authors state that all-stage cases and mainly healthy controls likely inflate performance relative to asymptomatic early disease and benign pancreatic controls. A prospective-cohort biorepository, NCT06271291, is the stated next validation stage." }, { "id": "oh-2022-topaz1", "type": "paper", "title": "Durvalumab plus Gemcitabine and Cisplatin in Advanced Biliary Tract Cancer", "authors": [ "Oh, Do-Youn", "He, Aiwu Ruth", "Qin, Shukui", "et al." ], "publisher": "NEJM Evidence", "date": "2022-08-01", "url": "https://pubmed.ncbi.nlm.nih.gov/38319896/", "doi": "10.1056/EVIDoa2200015", "pmid": "38319896", "trialId": "NCT03875235", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Double-blind phase 3 in 685 patients with mixed biliary-tract primary sites. Funded by AstraZeneca; several authors were employees." }, { "id": "oh-2025-topaz1-three-year", "type": "paper", "title": "Durvalumab plus chemotherapy in advanced biliary tract cancer: 3-year overall survival update from the phase III TOPAZ-1 study", "authors": [ "Oh, Do-Youn", "He, Aiwu Ruth", "Qin, Shukui", "Chen, Li-Tzong", "Okusaka, Takuji", "Valle, Juan W." ], "publisher": "Journal of Hepatology", "date": "2025-05-15", "url": "https://pubmed.ncbi.nlm.nih.gov/40381735/", "doi": "10.1016/j.jhep.2025.05.003", "pmid": "40381735", "trialId": "NCT03875235", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Exploratory mature follow-up of the same randomized TOPAZ-1 population: 685 patients, median follow-up 41.3 months, OS HR 0.74, median OS 12.9 versus 11.3 months and 36-month OS 14.6% versus 6.9%. This is longer follow-up, not independent replication." }, { "id": "ohtsuka-2024-bd-ipmn-surveillance", "type": "paper", "title": "Prospective multicenter surveillance study of branch-duct intraductal papillary mucinous neoplasm of the pancreas; risk of dual carcinogenesis", "authors": [ "Ohtsuka, Takao", "Maguchi, Hiroyuki", "Tokunaga, Suminori", "et al." ], "publisher": "Pancreatology", "date": "2024-08-22", "url": "https://pubmed.ncbi.nlm.nih.gov/39191596/", "doi": "10.1016/j.pan.2024.08.013", "pmid": "39191596", "trialId": "UMIN000007349", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Prospective Japan Pancreas Society cohort with six-monthly surveillance. Among 2,104 non-operated patients followed for 5.17 years, cumulative HGD or invasive carcinoma within the IPMN was 1.90% and separate concomitant PDAC was 2.11%. These are distinct cancer routes with different observed predictors." }, { "id": "ohtsuka-2024-kyoto-ipmn-guideline", "type": "paper", "title": "International evidence-based Kyoto guidelines for the management of intraductal papillary mucinous neoplasm of the pancreas", "authors": [ "Ohtsuka, Takao", "Fernandez-Del Castillo, Carlos", "Furukawa, Toru", "et al." ], "publisher": "Pancreatology", "date": "2023-12-28", "url": "https://pubmed.ncbi.nlm.nih.gov/38182527/", "doi": "10.1016/j.pan.2023.12.009", "pmid": "38182527", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "International guideline based on linked systematic reviews. For small unchanged branch-duct IPMN after five years it presents both stopping surveillance and continuing surveillance for separate concomitant PDAC as options, exposing unresolved utility rather than proving either choice." }, { "id": "oncoclinicas-2026-peixoto-profile", "type": "institution-profile", "title": "Renata D'Alpino Peixoto — gastrointestinal and neuroendocrine oncology at Centro Paulista de Oncologia", "authors": [ "Oncoclinicas Group" ], "publisher": "Oncoclinicas Group", "date": null, "url": "https://grupooncoclinicas.com/en/ocjournal/checkmate-577-opens-a-new-perspective-for-the-treatment-of-esophageal-cancer-and-esophageal-gastric-transition-with-immunotherapy", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Official employer article identifies Peixoto as a clinical oncologist specializing in gastrointestinal and neuroendocrine tumours at Centro Paulista de Oncologia, Oncoclinicas Group. This supplies current role only; it is not independent evidence of research performance." }, { "id": "oreilly-2019-durvalumab-tremelimumab", "type": "paper", "title": "Durvalumab With or Without Tremelimumab for Patients With Metastatic Pancreatic Ductal Adenocarcinoma", "authors": [ "O'Reilly, Eileen M.", "Oh, Do-Youn", "Dhalluin, Clara", "et al." ], "publisher": "JAMA Oncology", "date": "2019-10-01", "url": "https://pubmed.ncbi.nlm.nih.gov/31318392/", "doi": "10.1001/jamaoncol.2019.1588", "pmid": "31318392", "trialId": "NCT02558894", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Open-label randomised phase 2 lead-in at 21 sites in six countries. Part B was contingent on at least a 10% objective response and did not open. AstraZeneca funded the trial and several authors were employees or had company ties." }, { "id": "oreilly-2019-ntrk-pdac-response", "type": "paper", "title": "Tumour response to TRK inhibition in pancreatic adenocarcinoma harbouring an NTRK fusion", "authors": [ "O'Reilly, Eileen M", "et al." ], "publisher": "Annals of Oncology", "date": "2019-11-01", "url": "https://doi.org/10.1093/annonc/mdz385", "doi": "10.1093/annonc/mdz385", "pmid": "31605106", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Case report documents response of metastatic CTRC-NTRK1 fusion-positive PDAC to larotrectinib followed by acquired resistance and treatment with a next-generation TRK inhibitor. It proves that one pancreatic fusion can be drug-sensitive, not population benefit or durable control." }, { "id": "oreilly-2026-daraxonrasib", "type": "paper", "title": "Daraxonrasib or Chemotherapy in Previously Treated Metastatic Pancreatic Cancer", "authors": [ "O'Reilly, Eileen M.", "Wainberg, Zev A.", "Hendifar, Andrew E.", "et al." ], "publisher": "New England Journal of Medicine", "date": "2026-05-31", "url": "https://pubmed.ncbi.nlm.nih.gov/42223072/", "doi": "10.1056/NEJMoa2605555", "pmid": "42223072", "trialId": "NCT06625320", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "International open-label randomised phase 3 in 500 patients with previously treated metastatic PDAC. Revolution Medicines funded the trial and company employees were authors. The trial used investigator-choice chemotherapy as the comparator." }, { "id": "overvad-2022-khorana-validation", "type": "paper", "title": "Validation of the Khorana score for predicting venous thromboembolism in 40,218 patients with cancer initiating chemotherapy", "authors": [ "Overvad, Thure Filskov", "et al." ], "publisher": "Blood Advances", "date": "2022-04-12", "url": "https://pubmed.ncbi.nlm.nih.gov/35045569/", "doi": "10.1182/bloodadvances.2021006484", "pmid": "35045569", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Registry validation of 40,218 people starting chemotherapy. At the guideline threshold Khorana score at least 2, six-month VTE risk was 3.6%. The threshold did not stratify risk within pancreatic or hepatobiliary cancers, so the automatic two site points identify a broad high-risk class but do not solve individual selection." }, { "id": "ozdemir-2014-fibroblast-depletion", "type": "paper", "title": "Depletion of carcinoma-associated fibroblasts and fibrosis induces immunosuppression and accelerates pancreas cancer with reduced survival", "authors": [ "Özdemir, Berna C.", "Pentcheva-Hoang, Tsvetelina", "Carstens, Julienne L.", "et al." ], "publisher": "Cancer Cell", "date": "2014-05-22", "url": "https://pubmed.ncbi.nlm.nih.gov/24856586/", "doi": "10.1016/j.ccr.2014.04.005", "pmid": "24856586", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Transgenic mouse depletion of alpha-SMA-positive myofibroblasts, supported by a human tumour survival correlation. The intervention and survival effect are animal evidence." }, { "id": "padilla-valverde-2024-adjuvant-hipec-rct", "type": "paper", "title": "Safety and Effectiveness of Perioperative Hyperthermic Intraperitoneal Chemotherapy with Gemcitabine in Patients with Resected Pancreatic Ductal Adenocarcinoma: Clinical Trial EudraCT 2016-004298-41", "authors": [ "Padilla-Valverde, David", "Bodoque-Villar, Raquel", "Garcia-Santos, Esther", "et al." ], "publisher": "Cancers", "date": "2024-04-28", "url": "https://doi.org/10.3390/cancers16091718", "doi": "10.3390/cancers16091718", "pmid": "38730669", "trialId": "EudraCT 2016-004298-41", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Single-centre randomized study after margin-clear PDAC resection: 21 surgery-alone and 21 surgery-plus-HIPEC participants from 63 recruited, with 21 excluded for intraoperative unresectability or another diagnosis. Gemcitabine HIPEC reduced locoregional recurrence from 11/21 to 2/21, but median overall survival was 18.0 versus 17.1 months, p=0.899, and median disease-free survival 10 versus 14 months, p=0.888. One in-hospital death occurred in each arm. Adjuvant chemotherapy was given to 16/21 controls and 12/21 HIPEC participants." }, { "id": "padron-2022-prince", "type": "paper", "title": "Sotigalimab and/or nivolumab with chemotherapy in first-line metastatic pancreatic cancer: the randomized phase 2 PRINCE trial", "authors": [ "Padron, Lacey J.", "Maurer, Deena M.", "O'Hara, Mark H.", "et al." ], "publisher": "Nature Medicine", "date": "2022-06-03", "url": "https://pubmed.ncbi.nlm.nih.gov/35662283/", "doi": "10.1038/s41591-022-01829-9", "pmid": "35662283", "trialId": "NCT03214250", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "The three treatment groups were randomized against each other but each primary test used a historical 35% one-year survival rate, not a concurrent chemotherapy-only group. Efficacy n=105 included 12 non-randomized phase 1b patients. Nivolumab plus chemotherapy met the historical test; sotigalimab plus chemotherapy and the triplet did not. Median OS was 16.7, 11.4 and 10.1 months; ORR 50%, 33% and 31%. The study was not powered to compare groups." }, { "id": "paiella-2021-imilt-pancreas", "type": "paper", "title": "Laser treatment of pancreatic cancer with immunostimulating interstitial laser thermotherapy protocol: safety and feasibility results from two phase 2a studies", "authors": [ "Paiella, Salvatore", "Casetti, Luca", "Ewald, Jacques", "et al." ], "publisher": "Journal of Surgical Research", "date": "2020-12-02", "url": "https://pubmed.ncbi.nlm.nih.gov/33278792/", "doi": "10.1016/j.jss.2020.10.027", "pmid": "33278792", "trialId": "NCT02702986; NCT02973217", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Combined report from two prospective single-group surgical studies using the TRANBERG system after systemic treatment. Fifteen people were assigned and 13 treated; one procedure was judged too difficult and one was cancelled after liver metastases were found during surgery. Four adverse events were attributed to treatment, including three late pancreatic fistulas. The authors concluded safety, feasibility and handling were not satisfactory." }, { "id": "paik-2018-eus-ercp-biliary-rct", "type": "paper", "title": "EUS-guided biliary drainage versus ERCP for primary palliation of malignant biliary obstruction", "authors": [ "Paik, Woo Hyun", "Lee, Tae Hoon", "Park, Do Hyun", "et al." ], "publisher": "American Journal of Gastroenterology", "date": "2018-07-01", "url": "https://pubmed.ncbi.nlm.nih.gov/29961772/", "doi": "10.1038/s41395-018-0122-8", "pmid": "29961772", "trialId": "KCT0001396", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Four-centre South Korean randomized trial of 125 people with unresectable malignant distal obstruction. Technical and clinical success were similar. EUS-guided drainage had adverse events 6.3% versus 19.7%, pancreatitis 0 versus 14.8%, reintervention 15.6% versus 42.6% and 12-week quality-of-life preservation. It was a tertiary-centre palliative study of malignant obstruction, not a PDAC neoadjuvant or surgical-route trial." }, { "id": "panbela-2024-q3-10q", "type": "financial", "title": "Panbela Therapeutics quarterly report for the period ended 30 September 2024", "authors": [ "Panbela Therapeutics, Inc." ], "publisher": "U.S. Securities and Exchange Commission", "date": "2024-11-14", "url": "https://www.sec.gov/Archives/edgar/data/1029125/000143774924035315/pbla20240930_10q.htm", "doi": null, "pmid": null, "trialId": "NCT05254171", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Primary company filing. Panbela reported $142,000 cash, a $15.0m working-capital deficit, a $21.4m nine-month net loss and $12.5m nine-month operating cash use. Its ASPIRE contractor began terminating the contract on 19 August 2024 after unpaid invoices. Panbela assumed direct responsibility and sought a replacement. It projected Q2 2025 enrollment completion and early-2025 interim analysis. The filing reported 89 open sites, more than 50% enrollment and a 395-person safety database, but no actual randomized count." }, { "id": "panbela-2025-note-default", "type": "financial", "title": "Panbela Therapeutics notice of default on senior convertible notes", "authors": [ "Panbela Therapeutics, Inc." ], "publisher": "U.S. Securities and Exchange Commission", "date": "2025-01-22", "url": "https://www.sec.gov/Archives/edgar/data/1029125/000143774925001557/pbla20250122_8k.htm", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Panbela disclosed that it could not meet a covenant to pay material obligations. The $12.0m note balance moved to monthly SOFR plus 12% interest while the default continued and could be declared immediately due." }, { "id": "pancan-2025-audited-financials", "type": "charity-report", "title": "Pancreatic Cancer Action Network audited financial statements, June 30 2025 and 2024", "authors": [ "Pancreatic Cancer Action Network", "Forvis Mazars, LLP" ], "publisher": "Pancreatic Cancer Action Network", "date": "2025-10-24", "url": "https://media.pancan.org/pdf/annual-reports/2025%20-%20Pancreatic%20Cancer%20Action%20Network%20-%20FS%20FINAL.pdf", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Audited primary financial statements. FY2025 Research and Scientific Affairs expense was $13,822,863, including a $6,494,139 line labelled Research. PanCAN reported cumulative research investment of $249m. The broader expense includes staff and operating costs; the direct Research line is not the whole programme." }, { "id": "pancan-2025-grants", "type": "charity-report", "title": "PanCAN 2025 research grant recipients", "authors": [ "Pancreatic Cancer Action Network" ], "publisher": "Pancreatic Cancer Action Network", "date": "2025-07-07", "url": "https://pancan.org/research/grants-program/grants-awarded/by-year/2025-research-grant-recipients/", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Primary charity award page. It reports three new 2025 grants totalling $612,000: two fellowships and one early-detection pilot in an integrated health system. This is new-grant value, not all PanCAN research spending." }, { "id": "pancreatic-cancer-action-ali-stunt-profile-2026", "type": "organisation-profile", "title": "Our founder: Ali Stunt", "authors": [ "Pancreatic Cancer Action" ], "publisher": "Pancreatic Cancer Action", "date": null, "url": "https://pancreaticcanceraction.org/about-us/meet-the-team/our-founder/", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Official charity profile. Stunt is a pancreatic cancer survivor who founded Pancreatic Cancer Action in 2010 and stepped down as chief executive in April 2023 to advocate full time. This supports lived-experience and early-diagnosis capability, not a scientific-effect claim or availability for appointment." }, { "id": "pang-2022-fapi-pet", "type": "paper", "title": "Positron emission tomography and computed tomography with 68Ga-Ga-fibroblast activation protein inhibitors improves tumor detection and staging in patients with pancreatic cancer", "authors": [ "Pang, Yizhen", "Zhao, Long", "Shang, Qihang", "et al." ], "publisher": "European Journal of Nuclear Medicine and Molecular Imaging", "date": "2022-03-01", "url": "https://pubmed.ncbi.nlm.nih.gov/34651226/", "doi": "10.1007/s00259-021-05576-w", "pmid": "34651226", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Retrospective head-to-head comparison in 36 participants, including 26 pancreatic malignancies and 10 benign lesions. Management comparisons used 23 patients." }, { "id": "pant-2024-eli002", "type": "paper", "title": "Lymph-node-targeted, mKRAS-specific amphiphile vaccine in pancreatic and colorectal cancer: the phase 1 AMPLIFY-201 trial", "authors": [ "Pant, Shubham", "Wainberg, Zev A.", "VanderWalde, Ari", "et al." ], "publisher": "Nature Medicine", "date": "2024-01-09", "url": "https://www.nature.com/articles/s41591-023-02760-3", "doi": "10.1038/s41591-023-02760-3", "pmid": null, "trialId": "NCT04853017", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "First-in-human dose-escalation trial in 25 people with KRAS G12D or G12R pancreatic or colorectal cancer and detectable molecular residual disease after local treatment. Elicio Therapeutics employees were among the authors." }, { "id": "pant-2026-zanidatamab-final", "type": "paper", "title": "Zanidatamab in HER2-Positive Metastatic Biliary Tract Cancer: Final Results From HERIZON-BTC-01", "authors": [ "Pant, Shubham", "Fan, Jia", "Oh, Do-Youn", "et al." ], "publisher": "JAMA Oncology", "date": "2025-11-20", "url": "https://pubmed.ncbi.nlm.nih.gov/41264278/", "doi": "10.1001/jamaoncol.2025.4736", "pmid": "41264278", "trialId": "NCT04466891", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Final 33.4-month follow-up of the single-arm phase 2 trial in 80 ERBB2-amplified biliary cancers. The IHC 3+ and IHC 2+ outcome split was large. The programme was company funded and authors had extensive industry disclosures." }, { "id": "paq-2025-series-b-investors", "type": "company-announcement", "title": "PAQ Therapeutics announces $39 million Series B financing and initiates phase 1 trial", "authors": [ "PAQ Therapeutics" ], "publisher": "PAQ Therapeutics", "date": "2025-05-07", "url": "https://www.paqtx.com/news/paq-therapeutics-announces-39-million-series-b-financing-and-initiates-phase-1-trial-to-advance-novel-approach-addressing-kras-driven-cancers-with-high-unmet-need/", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Company announcement. $39m Series B co-led by MRL Ventures Fund and Bayland Capital, with JJDC, LAV Fund, BioTrack Capital and existing investor Sherpa Health Partners. The later extension brought the whole Series B to $77m; investor allocations and terms are not disclosed." }, { "id": "paq-2026-seriesb-pt0511", "type": "company-announcement", "title": "PAQ Therapeutics Series B reaches $77 million and first patient receives PT0511", "authors": [ "PAQ Therapeutics" ], "publisher": "PAQ Therapeutics", "date": "2026-01-22", "url": "https://www.paqtx.com/news/paq-therapeutics-announces-series-b-extension-bringing-total-series-b-financing-to-77-million-first-patient-dosed-in-phase-1-trial-of-pt0511-a-pan-kras-degrader/", "doi": null, "pmid": null, "trialId": "NCT07300150", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Company financing and first-patient announcement; no efficacy result." }, { "id": "parikh-2023-smart-phase2-safety", "type": "paper", "title": "A Multi-Institutional Phase 2 Trial of Ablative 5-Fraction Stereotactic Magnetic Resonance-Guided On-Table Adaptive Radiation Therapy for Borderline Resectable and Locally Advanced Pancreatic Cancer", "authors": [ "Parikh, Parag J.", "Lee, Percy", "Low, Daniel A.", "et al." ], "publisher": "International Journal of Radiation Oncology, Biology, Physics", "date": "2023-11-15", "url": "https://pubmed.ncbi.nlm.nih.gov/37210048/", "doi": "10.1016/j.ijrobp.2023.05.023", "pmid": "37210048", "trialId": "NCT03621644", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Prospective multicentre single-arm phase 2 safety report of 50 Gy in five MRI-guided adaptive fractions after induction chemotherapy. It reported no acute grade 3 or worse gastrointestinal toxicity definitely attributed to treatment and one-year survival of 65% from radiation." }, { "id": "park-2026-btc-ctdna-feasibility", "type": "paper", "title": "Feasibility of tumor-informed circulating tumor DNA for detecting minimal residual disease in surgically resected biliary tract cancer", "authors": [ "Park, Younghee", "Kim, Kyung Su", "Jo, Hyunji", "et al." ], "publisher": "PLOS ONE", "date": "2026-01-21", "url": "https://pubmed.ncbi.nlm.nih.gov/41563984/", "doi": "10.1371/journal.pone.0341432", "pmid": "41563984", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Prospective feasibility study enrolled 18 people with resectable biliary tract cancer. One was inoperable and personalized panels could not be made for three, leaving 14 analysed. Postoperative samples were available for 12: three of six ctDNA-positive and one of six ctDNA-negative patients progressed after median 17.4 months. Small numbers and panel-construction loss prevent an action threshold." }, { "id": "park-2026-gleam-zolbetuximab", "type": "conference-paper", "title": "Zolbetuximab plus gemcitabine and nab-paclitaxel in CLDN18.2-positive metastatic pancreatic adenocarcinoma: phase 2 GLEAM", "authors": [ "Park, Wungki", "Shen, Lin", "Ducreux, Michel", "et al." ], "publisher": "Annals of Oncology", "date": "2026-07-01", "url": "https://doi.org/10.1016/j.annonc.2026.05.383", "doi": "10.1016/j.annonc.2026.05.383", "pmid": null, "trialId": "NCT03816163", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Primary 393-person randomized GLEAM report. Median OS was 13.70 versus 13.57 months, HR 0.999 (95% CI 0.776-1.286), P=.4987. ORR was 44.7% versus 37.4%, disease control 70.6% versus 71.8%, and response duration 7.6 versus 5.6 months. The response changes did not translate into progression or survival benefit." }, { "id": "park-2026-polar-pdac", "type": "paper", "title": "Pembrolizumab and olaparib in homologous-recombination-deficient metastatic pancreatic cancer: the phase 2 POLAR trial", "authors": [ "Park, Wungki", "O'Connor, Catherine A.", "Chou, Joanne F.", "et al." ], "publisher": "Nature Medicine", "date": "2026-03-25", "url": "https://pubmed.ncbi.nlm.nih.gov/41882405/", "doi": "10.1038/s41591-026-04299-5", "pmid": "41882405", "trialId": "NCT04666740", "patentId": null, "grantId": "1P50 CA257881-01A1; K12 CA184746; R01 CA294640; P30 CA008748; U01 CA238444-01A1", "accessed": "2026-09-14", "notes": "Single-arm, biomarker-stratified phase 2 maintenance study after platinum in 63 metastatic pancreatic-cancer patients. The core BRCA1/2 or PALB2 HRD cohort had ORR 35% among 20 RECIST-evaluable patients and six-month PFS 64% among 33, missing both prespecified thresholds of 43% and 77%. Median OS was 28 months and three-year OS 44%, a possible selected tail without a randomized comparator. Non-core HRD and HRD-wild-type cohorts had weaker results. Extensive investigator industry relationships were disclosed." }, { "id": "park-2026-setidegrasib", "type": "paper", "title": "Setidegrasib in Advanced Non-Small-Cell Lung Cancer and Pancreatic Cancer", "authors": [ "Park, Wungki", "Kasi, Anup", "Spira, Alexander I.", "et al." ], "publisher": "New England Journal of Medicine", "date": "2026-03-25", "url": "https://pubmed.ncbi.nlm.nih.gov/41879829/", "doi": "10.1056/NEJMoa2600752", "pmid": "41879829", "trialId": "NCT05382559", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Astellas-funded phase 1 dose-escalation study across advanced KRAS-G12D solid tumours. The PDAC result at the selected 600 mg dose came from 21 second- or third-line patients, 67% in third line." }, { "id": "partelli-2022-aspen-interim", "type": "paper", "title": "Management of asymptomatic sporadic non-functioning pancreatic neuroendocrine neoplasms no larger than 2 cm: interim analysis of prospective ASPEN trial", "authors": [ "Partelli, Stefano", "Massironi, Sara", "Zerbi, Alessandro" ], "publisher": "British Journal of Surgery", "date": "2022-11-22", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC10364756/", "doi": "10.1093/bjs/znac267", "pmid": "35986682", "trialId": "NCT03084770", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Full open-access report checked. Prospective international but non-randomized cohort included 500 people with asymptomatic sporadic non-functioning PanNEN no larger than 2 cm: 406 observed and 94 assigned to surgery by the treating centre. Median follow-up was 25 months. Nine observed patients later had surgery and none developed distant metastasis during follow-up. Severe complications occurred in 13% of operated patients. Nineteen operated tumours had at least one aggressive histological feature, but assignment was strongly selected by tumour and patient factors, centre practice and preference." }, { "id": "patent-us10391096-kinisoquin", "type": "patent", "title": "Method for treating thrombotic disorders using quercetin-containing compositions", "authors": [ "Lines, Thomas Christian" ], "publisher": "USPTO / Google Patents", "date": "2019-08-27", "url": "https://patents.google.com/patent/US10391096B2/en", "doi": null, "pmid": null, "trialId": null, "patentId": "US10391096B2", "grantId": null, "accessed": "2026-09-14", "notes": "US grant from a family with May 2014 priority. Claims include treating venous thrombosis with a composition containing isoquercetin, vitamins B3 and C and folic acid, including cancer-induced thrombosis and specified dose ranges. The record names Quercis Pharma AG in the ownership history. It does not establish live global rights, trial efficacy or exact Kinisoquin formulation." }, { "id": "patent-us11690915-rason", "type": "patent", "title": "RAS inhibitors", "authors": [ "Aay, Naing", "Burnett, Leslie G.", "Cregg, James", "et al." ], "publisher": "USPTO via Google Patents", "date": "2023-07-04", "url": "https://patents.google.com/patent/US11690915B2/en", "doi": null, "pmid": null, "trialId": null, "patentId": "US11690915B2", "grantId": null, "accessed": "2026-09-14", "notes": "Granted US member assigned to Revolution Medicines in the broad RAS(ON) tri-complex chemistry family, priority 2020-09-15. The family is more relevant to compound access than a later pancreatic-use application; claim scope and validity require counsel." }, { "id": "patent-us12258636-zfp30-pdac", "type": "patent", "title": "Methods for early prediction, treatment response, recurrence and prognosis monitoring of pancreatic cancer", "authors": [ "Lin, Ruo-Kai", "Shen, Hsieh-Tsung" ], "publisher": "USPTO via Google Patents", "date": "2025-03-25", "url": "https://patents.google.com/patent/US12258636B2/en", "doi": null, "pmid": null, "trialId": null, "patentId": "US12258636B2", "grantId": null, "accessed": "2026-09-14", "notes": "Active US grant on ZFP30 and related methylation markers for pancreatic cancer. Examples include 31 PDAC versus 12 healthy plasma samples and larger case-control tables; these do not establish prospective screening utility. The record lists assignment to EG BioMed." }, { "id": "patent-us20240042066-fapi", "type": "patent", "title": "Fibroblast activation protein inhibitor", "authors": [ "Patent applicants listed in record" ], "publisher": "USPTO via Google Patents", "date": "2024-02-08", "url": "https://patents.google.com/patent/US20240042066A1/en", "doi": null, "pmid": null, "trialId": null, "patentId": "US20240042066A1", "grantId": null, "accessed": "2026-09-14", "notes": "FAPI patent application includes lutetium-177 treatment experiments in pancreatic patient-derived xenograft mice; no human efficacy claim." }, { "id": "patent-us20240131366-linfu", "type": "patent", "title": "Device and method for inducing a wide area of stable cavitation and controlling for inertial cavitation", "authors": [ "Morrison, Kyle", "Adam, Dan" ], "publisher": "Google Patents mirror of USPTO record", "date": "2024-04-25", "url": "https://patents.google.com/patent/US20240131366A1/en", "doi": null, "pmid": null, "trialId": null, "patentId": "US20240131366A1", "grantId": null, "accessed": "2026-09-15", "notes": "Published application assigned to Adenocyte for low-intensity non-focused ultrasound, microbubble cavitation control and tissue sampling. Publication shows a claimed device family, not freedom to operate, validity, product performance or clinical benefit." }, { "id": "patent-us20250009888-elicio-kras-vaccine", "type": "patent", "title": "Compounds including a mutant KRAS sequence and a lipid and uses thereof", "authors": [ "Elicio Therapeutics, Inc." ], "publisher": "USPTO via Google Patents", "date": "2025-01-09", "url": "https://patents.google.com/patent/US20250009888A1/en", "doi": null, "pmid": null, "trialId": null, "patentId": "US20250009888A1", "grantId": null, "accessed": "2026-09-14", "notes": "Application lists 2018 priority and Elicio as assignee; Google Patents flags this application as abandoned, with later related applications listed. Legal status requires counsel verification." }, { "id": "patent-us20250144252a1-cnsi-fe", "type": "patent", "title": "Carbon nanoparticle suspension injection-Fe, preparation method, application and use method", "authors": [ "Sichuan Enray Pharmaceutical Technology Co Ltd" ], "publisher": "USPTO via Google Patents", "date": "2025-05-01", "url": "https://patents.google.com/patent/US20250144252A1/en", "doi": null, "pmid": null, "trialId": null, "patentId": "US20250144252A1", "grantId": null, "accessed": "2026-09-15", "notes": "Pending US application in the PCT/CN2024/082864 family, claiming carbon-particle and ferrous-sulfate composition ranges, tumour injection, MRI-guided distribution and use. It claims priority to Chinese filings from March 2023 and March 2024. Patent claims establish an access boundary, not ferroptosis, safety or clinical benefit." }, { "id": "patent-wo2019195268-grail-methylation", "type": "patent", "title": "Methylation markers and targeted methylation probe panels", "authors": [ "Gross, Samuel S.", "Amini, Hamed", "Jamshidi, Arash", "et al." ], "publisher": "WIPO via Google Patents", "date": "2019-10-10", "url": "https://patents.google.com/patent/WO2019195268A2/en", "doi": null, "pmid": null, "trialId": null, "patentId": "WO2019195268A2", "grantId": null, "accessed": "2026-09-14", "notes": "GRAIL family covering targeted methylation probe panels and cancer classification including pancreatic and hepatobiliary cancers. The PCT entry is marked ceased, with multiple granted national continuations; ownership does not establish screening benefit." }, { "id": "patent-wo2019200228-natera-mrd", "type": "patent", "title": "Methods for cancer detection and monitoring by means of personalized detection of circulating tumor DNA", "authors": [ "Zimmermann, Bernhard", "Salari, Raheleh", "Swenerton, Ryan", "et al." ], "publisher": "WIPO via Google Patents", "date": "2019-10-17", "url": "https://patents.google.com/patent/WO2019200228A1/en", "doi": null, "pmid": null, "trialId": null, "patentId": "WO2019200228A1", "grantId": null, "accessed": "2026-09-14", "notes": "Natera family on patient-specific variant panels for recurrence or metastasis monitoring. The PCT entry is marked ceased, while national members include granted rights; the examples focus on breast, bladder and colorectal cancer rather than proving PDAC utility." }, { "id": "patent-wo2020039321-anti-gdf15", "type": "patent", "title": "Anti-GDF15 antibodies, compositions and methods of use", "authors": [ "Beaumont, Kevin Charles", "Breen, Danna M.", "Chabot, Jeffrey Raymond", "et al." ], "publisher": "WIPO / Google Patents", "date": "2020-02-27", "url": "https://patents.google.com/patent/WO2020039321A2/en", "doi": null, "pmid": null, "trialId": null, "patentId": "WO2020039321A2", "grantId": null, "accessed": "2026-09-14", "notes": "Pfizer-assigned antibody family with 2018-08-20 priority. The PCT entry is marked ceased, while listed national family members include grants in several jurisdictions. This establishes an ownership lead, not efficacy, exact claim scope or freedom to operate." }, { "id": "patent-wo2020112828-isoquercetin-cancer", "type": "patent", "title": "Compositions and methods for reducing major thrombotic events in cancer patients", "authors": [ "Zwicker, Jeffrey I.", "Flaumenhaft, Robert", "et al." ], "publisher": "WIPO / Google Patents", "date": "2020-06-04", "url": "https://patents.google.com/patent/WO2020112828A1/en", "doi": null, "pmid": null, "trialId": null, "patentId": "WO2020112828A1", "grantId": null, "accessed": "2026-09-14", "notes": "Beth Israel Deaconess family with November 2018 priority, later including US11872241B2. It covers isoquercetin or quercetin compositions and methods for reducing thrombotic events and biomarkers in cancer. The public record does not establish Quercis's licence scope, remaining national claims, validity or freedom to operate." }, { "id": "patent-wo2022173032-setidegrasib", "type": "patent", "title": "Quinazoline compound for inducing degradation of G12D-mutation KRAS protein", "authors": [ "Yoshinari, Tomohiro", "Ishioka, Hiroki", "Kawaminami, Eiji", "et al." ], "publisher": "WIPO via Google Patents", "date": "2022-08-18", "url": "https://patents.google.com/patent/WO2022173032A1/en", "doi": null, "pmid": null, "trialId": null, "patentId": "WO2022173032A1", "grantId": null, "accessed": "2026-09-14", "notes": "Astellas PCT family covering bifunctional KRAS-G12D degraders and pancreatic use; the PCT entry is marked ceased while national members include granted US, European, Chinese and Japanese records. Family status is not the same as freedom to operate." }, { "id": "patent-wo2022268051-hengrui-kras-g12d", "type": "patent", "title": "Fused tetracyclic compound, preparation method therefor and application thereof in medicine", "authors": [ "Li, Xin", "Shen, Feng", "Dong, Huaide", "Chen, Yang", "He, Feng", "Tao, Weikang" ], "publisher": "WIPO via Google Patents", "date": "2022-12-29", "url": "https://patents.google.com/patent/WO2022268051A1/en", "doi": null, "pmid": null, "trialId": null, "patentId": "WO2022268051A1", "grantId": null, "accessed": "2026-09-14", "notes": "Hengrui-assigned compound family with 2021-06-21 priority. It claims fused tetracyclic KRAS-G12D inhibitors and medical use. The PCT record is marked ceased, while national applications are listed in the US, Europe, China, Japan, Korea, Australia, Brazil, Canada, Mexico and Taiwan; several are pending. The later Hengrui liposome family expressly uses the formula-I compound from this family. Public primary documents do not explicitly print the HRS-4642 code beside this publication, so the code-to-claim match remains a strong inference requiring counsel and exact structure matching." }, { "id": "patent-wo2024022507-hengrui-liposome", "type": "patent", "title": "Pharmaceutical composition comprising KRAS G12D inhibitor", "authors": [ "Sha, Ruilin", "Duan, Ziqing", "Wang, Haifeng", "Tong, Xinyong", "Wang, Pingping", "Feng, Biao", "Wang, Bin" ], "publisher": "WIPO and EPO via Google Patents", "date": "2024-02-01", "url": "https://patents.google.com/patent/EP4563151A1/en", "doi": null, "pmid": null, "trialId": null, "patentId": "WO2024022507A1", "grantId": null, "accessed": "2026-09-14", "notes": "Hengrui-assigned formulation family with 2022-07-29 priority. Claims cover a formula-I KRAS-G12D inhibitor from WO2022268051 with lipids, liposomes, loading methods, lyophilized and reconstituted products and cancer use including pancreas, liver, gallbladder, bile duct and colorectal. The PCT entry is marked ceased while European, Chinese, Japanese, Korean, Australian and Canadian national applications are listed as pending. Hengrui's prospectus calls HRS-4642 its proprietary liposomal injectable formulation, but the patent page does not state that product code directly." }, { "id": "patent-wo2025245200-ponsegromab-cachexia", "type": "patent", "title": "Ponsegromab for use in treating cachexia in cancer patients", "authors": [ "Breen, Danna Marie", "Fang, Jarjieh", "Groarke, John David", "et al." ], "publisher": "WIPO / Google Patents", "date": "2025-11-27", "url": "https://patents.google.com/patent/WO2025245200A1/en", "doi": null, "pmid": null, "trialId": null, "patentId": "WO2025245200A1", "grantId": null, "accessed": "2026-09-14", "notes": "Pfizer-assigned pending PCT application with 2024-05-24 priority, covering ponsegromab cancer-cachexia use including 200 or 400 mg every four weeks with systemic chemotherapy and pancreatic cancer. Broad desired outcome ranges in claims are not clinical proof; national status and live claim scope require counsel." }, { "id": "patent-wo2026015825-daraxonrasib-pdac", "type": "patent", "title": "Use of RAS inhibitor for treating pancreatic cancer", "authors": [ "Revolution Medicines, Inc." ], "publisher": "WIPO PatentScope via Google Patents", "date": "2026-01-22", "url": "https://patents.google.com/patent/WO2026015825A1/en", "doi": null, "pmid": null, "trialId": null, "patentId": "WO2026015825A1", "grantId": null, "accessed": "2026-09-14", "notes": "Revolution Medicines application covering daraxonrasib use in pancreatic cancer; patent publication does not establish validity, freedom to operate or clinical efficacy." }, { "id": "pathak-2026-elraglusib-folfirinox", "type": "meeting-abstract", "title": "A phase II study of FOLFIRINOX combined with elraglusib and losartan in untreated metastatic PDAC", "authors": [ "Pathak, Priyadarshini Satyajit", "Walsh, Elizabeth P.", "Lau-Min, Kelsey S.", "et al." ], "publisher": "Journal of Clinical Oncology", "date": "2026-05-27", "url": "https://ascopubs.org/doi/10.1200/JCO.2026.44.16_suppl.4212", "doi": "10.1200/JCO.2026.44.16_suppl.4212", "pmid": null, "trialId": "NCT05077800", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Primary ASCO 2026 abstract. The open-label four-arm study called itself non-comparator and enrolled 49 people: FOLFIRINOX n=4 and three combination groups n=15 each. Median PFS was 5.1, 5.9, 6.0 and 6.5 months; median OS was 7.7, 8.2, 9.8 and 9.8 months. The authors concluded outcomes did not exceed historical standards. One treatment-related sepsis death occurred in the elraglusib arm." }, { "id": "pavlick-2026-mtap-genomics", "type": "paper", "title": "Genomics of MTAP loss in more than 500,000 solid tumour specimens", "authors": [ "Pavlick, Dean C.", "Rengarajan, Shruthi", "Lee, Jessica K.", "et al." ], "publisher": "JCO Precision Oncology", "date": "2026-05-01", "url": "https://pubmed.ncbi.nlm.nih.gov/42133901/", "doi": "10.1200/PO-26-00037", "pmid": "42133901", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Foundation Medicine retrospective series: 409,755 tissue and 85,801 liquid biopsies. Tissue MTAP loss was 28.9% in pancreatic cancer. Liquid biopsy positive agreement reached 86.2% only when tumour fraction was at least 20%; 33.7% of CDKN2A-loss cases lacked MTAP loss, so CDKN2A is not a safe stand-in. Partial MTAP loss made up 16.5% and its drug sensitivity remains uncertain." }, { "id": "pawlik-2025-psc-gallbladder-polyps", "type": "paper", "title": "Risk assessment of gallbladder cancer in patients with primary sclerosing cholangitis and gallbladder polyps: a systematic review", "authors": [ "Grikyte, Ieva", "Ignatavicius, Povilas" ], "publisher": "Langenbeck's Archives of Surgery", "date": "2025-07-09", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC12241131/", "doi": "10.1007/s00423-025-03678-9", "pmid": "40632214", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Systematic review checked. Six retrospective or case-control studies included 3,415 people with PSC. Imaging found 207 polyps or 6.2%; among the selected 116 patients with both a polyp and available histology, 48 cancers were reported. The 41.4% pathology proportion is strongly selected and cannot be applied to every PSC polyp, but PSC is a distinct high-risk pathway that must not be pooled with incidental polyps in the general population." }, { "id": "pcuk-2025-annual-report", "type": "charity-report", "title": "Pancreatic Cancer UK annual report 2024/25", "authors": [ "Pancreatic Cancer UK" ], "publisher": "Charity Commission for England and Wales", "date": "2025-12-01", "url": "https://register-of-charities.charitycommission.gov.uk/en/charity-search?_uk_gov_ccew_onereg_charitydetails_web_portlet_CharityDetailsPortlet_objectiveId=A17866229&_uk_gov_ccew_onereg_charitydetails_web_portlet_CharityDetailsPortlet_priv_r_p_mvcRenderCommandName=%2Faccounts-and-annual-returns&_uk_gov_ccew_onereg_charitydetails_web_portlet_CharityDetailsPortlet_priv_r_p_organisationNumber=4020499&p_p_cacheability=cacheLevelPage&p_p_id=uk_gov_ccew_onereg_charitydetails_web_portlet_CharityDetailsPortlet&p_p_lifecycle=2&p_p_mode=view&p_p_resource_id=%2Faccounts-resource&p_p_state=maximized", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Primary filed charity report. It reports £2,551,913 of new research grants in 2024/25, £1.3m invested in research support during the year and £17.5m cumulative research investment. The report says £0.5m of the lifetime figure arose from reclassification of earlier grants." }, { "id": "pea-2026-post-surveillance-resection", "type": "paper", "title": "Cancer Risk After Initial Surveillance of Intraductal Papillary Mucinous Neoplasms (IPMNs): Are We Missing the Window for Prevention?", "authors": [ "Pea, Antonio", "Dall'Olio, Tommaso", "Solinas, Daria", "et al." ], "publisher": "Annals of Surgical Oncology", "date": "2026-06-09", "url": "https://pubmed.ncbi.nlm.nih.gov/42265517/", "doi": "10.1245/s10434-026-19898-2", "pmid": "42265517", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Single-centre retrospective cohort of 3,304 presumed IPMN across changing guideline eras. Of 304 post-surveillance resections, low-grade dysplasia fell to 17% while high-grade dysplasia reached 28% and invasive carcinoma 45% after 2017. The study cannot separate guideline effects from changing imaging, pathology, selection and time at risk." }, { "id": "pei-2025-spatial-metastatic-pdac", "type": "paper", "title": "Spatial mapping of transcriptomic plasticity in metastatic pancreatic cancer", "authors": [ "Pei, Guangsheng", "Min, Jimin", "Rajapakshe, Kimal I.", "et al." ], "publisher": "Nature", "date": "2025-06-01", "url": "https://pubmed.ncbi.nlm.nih.gov/40269162/", "doi": "10.1038/s41586-025-08927-x", "pmid": "40269162", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Spatial transcriptomic study of primary, liver, lung and peritoneal samples from rapid autopsies of 13 people with treatment-refractory PDAC. Forty-nine spatial sections yielded 134,776 spots; five patients were untreated and eight had received chemotherapy. It found multiple lineage states in the same tissue and organ-linked shifts, validated with models and organoids. End-stage tissue cannot identify a disease-free residual cell or its earlier state, and treatment and autopsy selection limit causal interpretation. Several authors reported product, patent, consulting or research relationships." }, { "id": "peixoto-2025-gaspar-results", "type": "journal-article", "title": "Epidemiological characteristics of gastric and pancreatic cancers in Latin America: GASPAR (LACOG 0222) study", "authors": [ "D'Alpino Peixoto, Renata", "Fonseca Jesus, Victor Hugo", "Diniz Gomes, Diogo Bugano", "Yossen, David", "Weschenfelder, Rui", "Felismino, Tiago", "Esteso, Federico", "Bonilla, Carlos", "Kruschewsky Coutinho, Anelisa", "Dias, Antonio", "Freile, Berenice", "Cabalheiro, Taina", "Gossling, Gustavo", "Barrios, Pablo" ], "publisher": "Journal of Clinical Oncology", "date": "2025-01-27", "url": "https://ascopubs.org/doi/10.1200/JCO.2025.43.4_suppl.369", "doi": "10.1200/JCO.2025.43.4_suppl.369", "pmid": null, "trialId": "NCT05924789", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Meeting abstract reports 201 people at seven sites in Brazil, Argentina and Colombia, including 79 with advanced pancreatic cancer. Of the pancreatic cohort, 67 received first-line, 27 second-line and 10 third-line treatment; median first-line PFS was 5.4 months and median OS 10.2 months. The combined cohort was 76% privately insured; the pancreatic cohort was reported as 90% white, 8% American Indian or Alaska Native, no Black participants and 2% unreported. Observational, selected, small and not a regional population sample." }, { "id": "pelletier-2026-nis793-biomarkers", "type": "paper", "title": "Biomarker analysis from patients with metastatic PDAC treated with TGFbeta antibody, NIS793, plus abraxane and gemcitabine versus abraxane and gemcitabine alone in a phase II randomized study", "authors": [ "Pelletier, Maryse", "Yang, Jie", "Joshi, Maulik", "et al." ], "publisher": "Clinical Cancer Research", "date": "2026-06-25", "url": "https://pubmed.ncbi.nlm.nih.gov/42347868/", "doi": "10.1158/1078-0432.CCR-26-0805", "pmid": "42347868", "trialId": "NCT04390763", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "NIS793 reduced TGF-beta signals and markers of fibroblasts and collagen, showing that the drug changed its intended tissue system. It did not improve outcomes: the NIS793 plus gemcitabine/nab-paclitaxel OS HR was 1.32, 95% CI 0.84-2.07, versus chemotherapy. Neutrophil-related genes increased after treatment." }, { "id": "pelzer-2015-conko004", "type": "paper", "title": "Efficacy of prophylactic low-molecular-weight heparin for ambulatory patients with advanced pancreatic cancer: outcomes from CONKO-004", "authors": [ "Pelzer, Uwe", "Opitz, Bernhard", "Deutschinoff, Gerd", "et al." ], "publisher": "Journal of Clinical Oncology", "date": "2015-08-01", "url": "https://pubmed.ncbi.nlm.nih.gov/25987694/", "doi": "10.1200/JCO.2014.55.1481", "pmid": "25987694", "trialId": "ISRCTN02140505", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Randomized 312 advanced-pancreatic-cancer patients to chemotherapy with or without enoxaparin. Symptomatic VTE in the first three months was 2/160 versus 15/152, HR .12, P=.001. Overall symptomatic VTE was 6.4% versus 15.1%, HR .40. Major bleeding was 7/160 versus 5/152. PFS HR 1.06 and OS HR 1.01 showed no cancer-control or survival effect." }, { "id": "penn-vonderheide-profile-2026", "type": "institution", "title": "Robert H. Vonderheide — Vonderheide Lab", "authors": [ "Perelman School of Medicine at the University of Pennsylvania" ], "publisher": "University of Pennsylvania", "date": null, "url": "https://www.med.upenn.edu/cci/vonderheidelab", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Live institutional laboratory page checked for current position and translational focus on CD40, vaccines, pancreatic-cancer immunity and mouse-to-human work." }, { "id": "perez-2026-paricalcitol-pdac", "type": "paper", "title": "Gemcitabine and nab-paclitaxel with or without the VDR agonist paricalcitol for metastatic pancreatic cancer: a randomized, multiarm, run-in phase trial", "authors": [ "Perez, Kimberly J.", "Dias Costa, Ana", "Jordan, Ashley", "et al." ], "publisher": "Nature Cancer", "date": "2026-06-01", "url": "https://pubmed.ncbi.nlm.nih.gov/42185478/", "doi": "10.1038/s43018-026-01165-8", "pmid": "42185478", "trialId": "NCT03520790", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Thirty-six-person randomized run-in comparing chemotherapy plus placebo with intravenous or oral paricalcitol. Paired biopsies showed altered fibroblast signals and closer CD8 T-cell contact with tumour cells, but the trial was designed for safety and tissue pharmacology rather than efficacy. Five of 12 oral-paricalcitol patients had grade 2-4 hypercalcaemia requiring dose reduction. The record proves that a predicted tissue change can be measured; it does not prove patient benefit." }, { "id": "pfizer-2026-ponsegromab-profile", "type": "company", "title": "Ponsegromab | PF-06946860 development profile", "authors": [ "Pfizer" ], "publisher": "Pfizer Oncology Development", "date": "2026-05-05", "url": "https://www.pfizeroncologydevelopment.com/node/1005", "doi": null, "pmid": null, "trialId": "NCT06989437", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Sponsor profile states that PF-06946860 is a humanized antibody binding circulating GDF-15 and blocking its GFRAL interaction. It describes phase 2/3 development and states that safety and efficacy are not established. Company source, not independent efficacy evidence." }, { "id": "philip-2010-swog-s0205-cetuximab", "type": "paper", "title": "Phase III study comparing gemcitabine plus cetuximab versus gemcitabine in patients with advanced pancreatic adenocarcinoma: Southwest Oncology Group-directed intergroup trial S0205", "authors": [ "Philip, Philip A.", "Benedetti, Jacqueline", "Corless, Christopher L.", "et al." ], "publisher": "Journal of Clinical Oncology", "date": "2010-08-01", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC2917315/", "doi": "10.1200/JCO.2009.25.7550", "pmid": "20606093", "trialId": "NCT00075686", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Randomized phase 3 trial with 745 eligible people. Gemcitabine plus cetuximab gave median survival 6.3 months versus 5.9 months with gemcitabine alone, hazard ratio 1.06, 95% CI 0.91-1.23, P=0.23. Response and progression were similar. Ninety percent of tested tumours expressed EGFR, with no benefit in that subset. This tests EGFR blockade, not light-activated membrane injury." }, { "id": "philip-2022-kras-wildtype-pdac", "type": "paper", "title": "Molecular Characterization of KRAS Wild-type Tumors in Patients with Pancreatic Adenocarcinoma", "authors": [ "Philip, Philip A.", "Azar, Ibrahim", "Xiu, Joanne", "et al." ], "publisher": "Clinical Cancer Research", "date": "2022-06-13", "url": "https://doi.org/10.1158/1078-0432.CCR-21-3581", "doi": "10.1158/1078-0432.CCR-21-3581", "pmid": "35302596", "trialId": null, "patentId": null, "grantId": "P30 CA022453", "accessed": "2026-09-15", "notes": "DNA and RNA profiling of 2,483 PDAC samples found 266, 10.7%, were KRAS wild type. Those tumours were enriched for possible treatment routes: BRAF, FGFR2, ALK, RET and NRG1 fusions, NTRK and ERBB2 changes, and MSI-high disease. NRG1 fusions occurred in 1.3% of the KRAS-wild-type group." }, { "id": "philip-2024-avenger500", "type": "paper", "title": "Devimistat with modified FOLFIRINOX versus FOLFIRINOX for metastatic pancreatic cancer: AVENGER 500", "authors": [ "Philip, Philip A.", "Sahai, Vaibhav", "Bahary, Nathan", "et al." ], "publisher": "Journal of Clinical Oncology", "date": "2024-11-01", "url": "https://pubmed.ncbi.nlm.nih.gov/39088774/", "doi": "10.1200/JCO.23.02659", "pmid": "39088774", "trialId": "NCT03504423", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Global open-label phase 3 at 74 sites in six countries. Five hundred twenty-eight untreated metastatic pancreatic adenocarcinoma patients were randomized: 266 to devimistat plus modified FOLFIRINOX and 262 to standard FOLFIRINOX. Median overall survival was 11.10 versus 11.73 months, HR 0.95, 95% CI 0.77-1.18, P=.655. Median progression-free survival was 7.8 versus 8.0 months, HR 0.99, 95% CI 0.76-1.29, P=.94." }, { "id": "picozzi-2025-pert-advanced-pdac", "type": "paper", "title": "Pancreatic enzyme replacement therapy in advanced adenocarcinoma of the pancreas improved overall survival: a retrospective, single institution study", "authors": [ "Picozzi, Vincent J.", "Mandelson, Margaret T.", "Najjar, Anas", "et al." ], "publisher": "The Oncologist", "date": "2025-04-04", "url": "https://pubmed.ncbi.nlm.nih.gov/40231659/", "doi": "10.1093/oncolo/oyaf014", "pmid": "40231659", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Retrospective single-centre cohort of 501 advanced-PDAC patients with exocrine insufficiency receiving first-line chemotherapy; 188 received PERT. PERT was associated with less weight loss and median OS 17.1 versus 12.5 months, adjusted HR .73. Treatment was not randomized; three authors were AbbVie employees and may hold stock." }, { "id": "picozzi-2026-lapis", "type": "paper", "title": "LAPIS: A phase III randomized, placebo-controlled, double-blind trial of neoadjuvant pamrevlumab with chemotherapy for locally advanced pancreatic cancer", "authors": [ "Picozzi, Vincent J.", "et al." ], "publisher": "JCO Oncology Advances", "date": "2026-06-01", "url": "https://ascopubs.org/doi/10.1200/OA-26-00026", "doi": "10.1200/OA-26-00026", "pmid": null, "trialId": "NCT03941093", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Primary phase 3 paper: 284 patients across 72 centres. Overall survival 17.3 versus 17.9 months, HR 1.08, 95% CI 0.83-1.41, P=.5487; event-free survival, progression-free survival and objective response also showed no improvement." }, { "id": "pietrobono-2026-angiogenin-alk5", "type": "paper", "title": "EGFR ligand Angiogenin predicts response to ALK5 inhibition in pancreatic cancer via a TNF-alpha paracrine axis in tumor-associated macrophages", "authors": [ "Pietrobono, Sara", "De Vita, Valerio", "Mangiameli, Daniele", "et al." ], "publisher": "Oncogene", "date": "2026-06-01", "url": "https://pubmed.ncbi.nlm.nih.gov/41975075/", "doi": "10.1038/s41388-026-03774-0", "pmid": "41975075", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Later analysis of the 2018 galunisertib trial plus models. In the retrospectively defined high-baseline-angiogenin group, median OS was 8.90 months with galunisertib plus gemcitabine versus 5.65 months with placebo plus gemcitabine, HR 0.57, 95% CI 0.35-0.92. The paper links the marker to an angiogenin-EGFR-TGF-beta-TNF-alpha macrophage route. The marker threshold and interaction need independent fixed validation." }, { "id": "pishvaian-2020-know-your-tumor", "type": "paper", "title": "Overall survival in patients with pancreatic cancer receiving matched therapies following molecular profiling", "authors": [ "Pishvaian, Michael J", "Blais, Edik M", "Brody, Jonathan R", "et al." ], "publisher": "The Lancet Oncology", "date": "2020-04-01", "url": "https://doi.org/10.1016/S1470-2045(20)30074-7", "doi": "10.1016/S1470-2045(20)30074-7", "pmid": "32135080", "trialId": null, "patentId": null, "grantId": "R01 CA212600", "accessed": "2026-09-15", "notes": "Know Your Tumor referred 1,856 pancreatic patients; 1,082, 58%, received a report and 282 had an alteration labelled actionable. Among 677 with outcomes, 46 received matched treatment. Their median survival was 2.58 versus 1.51 years among 143 actionable-but-unmatched patients, HR 0.42, 95% CI 0.26 to 0.68. Treatment was not randomized, immortal-time and fitness selection remain possible, and only 46/1,856=2.5% of all referrals entered the matched group." }, { "id": "pmda-2020-akalux-review", "type": "regulatory", "title": "Report on the deliberation results for Akalux cetuximab sarotalocan sodium", "authors": [ "Pharmaceuticals and Medical Devices Agency", "Ministry of Health, Labour and Welfare, Japan" ], "publisher": "Pharmaceuticals and Medical Devices Agency", "date": "2020-09-08", "url": "https://www.pmda.go.jp/files/000239770.pdf", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Japanese regulator's review supporting conditional early approval for unresectable locally advanced or recurrent head and neck cancer, not pancreatic cancer. Conditions required a risk plan, surveillance of all treated patients because Japanese study numbers were very limited, provision of ongoing phase 3 data and use only by trained physicians." }, { "id": "pnas-2026-liaki-retraction", "type": "retraction-notice", "title": "Retraction for Liaki et al., A targeted combination therapy achieves effective pancreatic cancer regression and prevents tumor resistance", "authors": [ "Proceedings of the National Academy of Sciences editors" ], "publisher": "Proceedings of the National Academy of Sciences", "date": "2026-04-27", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC13143023/", "doi": "10.1073/pnas.2612385123", "pmid": "42044351", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Primary retraction notice. The first version, DOI 10.1073/pnas.2523039122 and PMID 41329731, was retracted because Mariano Barbacid, Vasiliki Liaki and Carmen Guerra held undisclosed financial interests in Vega Oncotargets. A new direct-submission version with declarations was published in June 2026 under DOI 10.1073/pnas.2610708123." }, { "id": "poruk-2016-ctc-pdac-diagnosis", "type": "paper", "title": "Circulating tumour cells as a biomarker for diagnosis and staging in pancreatic cancer", "authors": [ "Poruk, Katherine E.", "Blackford, Amanda L.", "Weiss, Matthew J.", "et al." ], "publisher": "British Journal of Cancer", "date": "2016", "url": "https://pubmed.ncbi.nlm.nih.gov/27300108/", "doi": null, "pmid": "27300108", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Prospective presentation cohort of 100 consecutive untreated patients; 72 confirmed PDAC cases used in diagnostic evaluation." }, { "id": "preopanc2-2025", "type": "paper", "title": "Neoadjuvant FOLFIRINOX versus neoadjuvant gemcitabine-based chemoradiotherapy in resectable and borderline resectable pancreatic cancer (PREOPANC-2)", "authors": [ "Dutch Pancreatic Cancer Group", "et al." ], "publisher": "The Lancet Oncology", "date": "2025-09-01", "url": "https://pubmed.ncbi.nlm.nih.gov/40945523/", "doi": "10.1016/S1470-2045(25)00363-8", "pmid": "40945523", "trialId": "EudraCT 2017-002036-17", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Investigator-led, open-label phase 3 trial across 19 Dutch centres. 375 were randomized; modified intention-to-treat analysis included 369. Funded by the Dutch Cancer Society and ZonMw." }, { "id": "primrose-2019-bilcap", "type": "paper", "title": "Capecitabine compared with observation in resected biliary tract cancer (BILCAP)", "authors": [ "Primrose, John N.", "Fox, Richard P.", "Palmer, Daniel H.", "et al." ], "publisher": "The Lancet Oncology", "date": "2019-03-25", "url": "https://pubmed.ncbi.nlm.nih.gov/30922733/", "doi": "10.1016/S1470-2045(18)30915-X", "pmid": "30922733", "trialId": "EudraCT 2005-003318-13", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Open-label UK phase 3 in 447 resected cholangiocarcinoma or muscle-invasive gallbladder cancers. Funded by Cancer Research UK and Roche. The primary intention-to-treat OS analysis was not statistically significant; planned adjusted and per-protocol analyses were." }, { "id": "provenzano-2012-stromal-delivery-barrier", "type": "paper", "title": "Enzymatic targeting of the stroma ablates physical barriers to treatment of pancreatic ductal adenocarcinoma", "authors": [ "Provenzano, Paolo P.", "Cuevas, Carlos", "Chang, Amy E.", "et al." ], "publisher": "Cancer Cell", "date": "2012-03-20", "url": "https://pubmed.ncbi.nlm.nih.gov/22439937/", "doi": "10.1016/j.ccr.2012.01.007", "pmid": "22439937", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Autochthonous mouse PDAC work linked hyaluronan-rich matrix to high pressure, collapsed vessels and weak drug transport. Hyaluronan depletion restored perfusion and improved gemcitabine response in mice; later phase 3 PEGPH20 evidence did not improve survival." }, { "id": "pusceddu-2026-percutaneous-cryoablation", "type": "paper", "title": "Percutaneous CT-Guided Cryoablation for Pain Palliation and Local Treatment Effect in Unresectable Pancreatic Ductal Adenocarcinoma: A Pilot Single-Center Case Series", "authors": [ "Pusceddu, Claudio", "Carrubba, Claudio", "Rinaldi, Pierluigi Maria", "et al." ], "publisher": "Cancers", "date": "2026-05-25", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC13255842/", "doi": "10.3390/cancers18111724", "pmid": "42279308", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Retrospective selected series of 11 painful unresectable cases, nine metastatic. Mean 0-10 pain score fell 6.72 to 3.45 at one month; all reduced opioid use and five stopped. Analgesic dose was not standardized, no control existed, and later pain used only surviving evaluable patients. Patients with direct bowel invasion were excluded; minor events occurred in three." }, { "id": "pushalkar-2018-pdac-microbiome", "type": "paper", "title": "The pancreatic cancer microbiome promotes oncogenesis by induction of innate and adaptive immune suppression", "authors": [ "Pushalkar, Smruti", "Hundeyin, Mautin", "Daley, Donnele", "et al." ], "publisher": "Cancer Discovery", "date": "2018-04-01", "url": "https://pubmed.ncbi.nlm.nih.gov/29567829/", "doi": "10.1158/2159-8290.CD-17-1134", "pmid": "29567829", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Mouse and human tissue study. Microbe removal and transfer altered suppressive myeloid cells and T-cell activity in PDAC models. The human specimens support presence and association, not a treatment effect in patients." }, { "id": "qi-2024-ct041-pdac", "type": "paper", "title": "Safety and Efficacy of CT041 in Patients With Refractory Metastatic Pancreatic Cancer: A Pooled Analysis of Two Early-Phase Trials", "authors": [ "Qi, Changsong", "Zhang, Panpan", "Liu, Chang", "et al." ], "publisher": "Journal of Clinical Oncology", "date": "2024-05-24", "url": "https://pubmed.ncbi.nlm.nih.gov/38788174/", "doi": "10.1200/JCO.23.02314", "pmid": "38788174", "trialId": "NCT03874897; NCT04581473", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Primary pooled PDAC report from two open-label early-phase trials. Among 24 infused patients, four responded, ORR 16.7%, disease control was 70.8%, median PFS 3.3 months, OS 10.0 months and response duration 9.5 months. There was no concurrent control. Severe events were mainly blood-count effects of the lymphocyte-depleting conditioning treatment; cytokine-release syndrome and gastrointestinal events were mainly grade 1 or 2." }, { "id": "qin-2025-cares310-final", "type": "paper", "title": "Camrelizumab plus rivoceranib versus sorafenib as first-line therapy for unresectable hepatocellular carcinoma: final CARES-310 analysis", "authors": [ "Qin, Shukui", "Gu, Shanzhi", "Chan, Stephen L", "et al." ], "publisher": "The Lancet Oncology", "date": "2025-12-01", "url": "https://pubmed.ncbi.nlm.nih.gov/41308676/", "doi": "10.1016/S1470-2045(25)00543-1", "pmid": "41308676", "trialId": "NCT03764293", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Final open-label randomized phase 3 analysis in 543 untreated unresectable HCC patients at 95 sites. Median OS was 23.8 versus 15.2 months, HR 0.64; PFS 5.6 versus 3.7 months, HR 0.54. Eighty-three percent were Asian. Hengrui and Elevar funded the trial." }, { "id": "quercis-2026-kinisoquin-profile", "type": "company", "title": "Quercis Pharma company and Kinisoquin development profile", "authors": [ "Quercis Pharma AG" ], "publisher": "Quercis Pharma", "date": "2026-09-14", "url": "https://www.quercis.com/highlights", "doi": null, "pmid": null, "trialId": "NCT06861088", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Company page presents Kinisoquin as a phase 3 oral treatment, repeats zero VTE and no major bleeding among 57 phase 2 participants, claims an FDA Special Protocol Assessment and patent protection through 2045, and expects an interim result in Q2 2027. It also states that the product is investigational and unapproved. Company claims require registry, patent and trial checks." }, { "id": "rai-2022-lappie-protocol", "type": "protocol", "title": "Treatment of unresectable locally advanced pancreatic cancer with percutaneous irreversible electroporation following initial systemic chemotherapy: LAP-PIE feasibility trial protocol", "authors": [ "Rai, Zahir L.", "Ranieri, Vania", "Palmer, Daniel H.", "et al." ], "publisher": "BMJ Open", "date": "2022-05-12", "url": "https://pubmed.ncbi.nlm.nih.gov/35551086/", "doi": "10.1136/bmjopen-2021-050166", "pmid": "35551086", "trialId": "ISRCTN14986389", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Protocol for a 50-person UK randomized feasibility study at up to seven specialist centres. After first-line FOLFIRINOX, one arm receives one percutaneous IRE procedure followed by further chemotherapy if indicated; the control receives chemotherapy alone. The trial is designed to test recruitment, technical delivery, acceptability and safety, not to establish survival efficacy. Follow-up is 12 months and includes quality of life and indicative procedure cost." }, { "id": "rakuten-medical-2026-asp1929", "type": "company-report", "title": "Rakuten Medical reports ASP-1929 programme and regulatory status", "authors": [ "Rakuten Medical, Inc." ], "publisher": "Rakuten Medical, Inc.", "date": "2026-05-30", "url": "https://rakuten-med.com/us/news/press-releases/2026/05/30/7959/", "doi": null, "pmid": null, "trialId": "NCT03769506; NCT04305795; NCT06699212", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Company source says ASP-1929 links cetuximab to IRDye 700DX and is activated locally with 690 nm light. It confirms Japanese approval only for unresectable locally advanced or recurrent head and neck cancer and says the product remains unapproved outside Japan. It does not report pancreatic human treatment." }, { "id": "ramamoorthy-2025-incidental-gbc-meta", "type": "paper", "title": "Risk Factors and Clinical Predictors Associated with Incidental Gallbladder Cancer: A Systematic Review and Meta-Analysis", "authors": [ "Ramamoorthy, Thangaraj", "Kulothungan, Venkatesan", "Mathur, Prashant" ], "publisher": "Journal of Gastrointestinal Cancer", "date": "2025-12-29", "url": "https://pubmed.ncbi.nlm.nih.gov/41460407/", "doi": "10.1007/s12029-025-01369-8", "pmid": "41460407", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Systematic review checked. Eighteen observational studies covered 788,214 cholecystectomy patients and nine entered meta-analysis. Age, female sex, raised alkaline phosphatase, polyps over 10 mm and open surgery associated with incidental cancer; stones over 3 cm did not. Open surgery is partly a consequence of intraoperative difficulty or suspicion and cannot be used as a preoperative prevention cause. The negative large-stone result conflicts with the South American operated-specimen association and should block a stone-size-only rule." }, { "id": "ramaswamy-2025-afpac-losartan", "type": "paper", "title": "Losartan plus modified FOLFIRINOX versus modified FOLFIRINOX in advanced pancreatic cancers: AFPAC randomized trial", "authors": [ "Ramaswamy, Anant", "Bhargava, Prabhat", "Gota, Vikram", "et al." ], "publisher": "Cancer", "date": "2025-07-01", "url": "https://pubmed.ncbi.nlm.nih.gov/40542713/", "doi": "10.1002/cncr.35945", "pmid": "40542713", "trialId": "CTRI/2021/05/033482", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Open-label randomized phase 3 plan, stopped after an unplanned efficacy analysis at 88 of 264 planned participants. mFOLFIRINOX n=44 versus mFOLFIRINOX plus losartan n=44 had median OS 10.4 versus 9.1 months, PFS 8.44 versus 7.36 months, and response 22% versus 23%. Plasma TGF-beta trends did not differ. Nine of 41 treated with losartan paused it." }, { "id": "ramsey-2023-mainstreaming-germline", "type": "paper", "title": "Mainstreaming germline genetic testing for patients with pancreatic cancer increases uptake", "authors": [ "Ramsey, Megan L.", "Tomlinson, Julia", "Pearlman, Rachel", "et al." ], "publisher": "Familial Cancer", "date": "2022-06-17", "url": "https://pubmed.ncbi.nlm.nih.gov/35713757/", "doi": "10.1007/s10689-022-00300-5", "pmid": "35713757", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Oncology-led testing of 245 patients produced a 6.5-fold testing-volume increase. Thirty-four, 13.9%, had a pathogenic or likely pathogenic variant, but only 22 of those 34, 64.7%, attended genetic counselling; five variants, 2.0% of all tests, had treatment implications. This is an implementation report, not a controlled outcome study." }, { "id": "rapp-2022-proton-phase2", "type": "paper", "title": "Dose-escalated proton therapy with elective nodal irradiation and concomitant chemotherapy for unresectable, borderline resectable, or medically inoperable pancreatic cancer: a phase II trial", "authors": [ "Rapp, Cooper T.", "Rutenberg, Michael S.", "Morris, Christopher G.", "Nichols, Romaine C." ], "publisher": "Journal of Gastrointestinal Oncology", "date": "2022-06-01", "url": "https://pubmed.ncbi.nlm.nih.gov/35837200/", "doi": "10.21037/jgo-21-593", "pmid": "35837200", "trialId": "NCT02598349", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Single-centre single-arm phase 2 report in 15 people. It stopped accrual early after a related multicentre study opened. One-year survival was 47%, missing the protocol target of 75% against a 50% historical benchmark." }, { "id": "raymond-2011-sunitinib-pnet", "type": "paper", "title": "Sunitinib malate for the treatment of pancreatic neuroendocrine tumors", "authors": [ "Raymond, Eric", "Dahan, Laetitia", "Raoul, Jean-Luc", "et al." ], "publisher": "New England Journal of Medicine", "date": "2011-02-10", "url": "https://pubmed.ncbi.nlm.nih.gov/21306237/", "doi": "10.1056/NEJMoa1003825", "pmid": "21306237", "trialId": "NCT00428597", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Double-blind placebo-controlled phase 3 in 171 progressive, well-differentiated advanced pNET patients. Stopped early after safety-monitoring review, which can inflate effect estimates. Funded by Pfizer." }, { "id": "relay-elevar-2024-license", "type": "company", "title": "Relay Therapeutics and Elevar Therapeutics announce exclusive global licence for lirafugratinib", "authors": [ "Relay Therapeutics", "Elevar Therapeutics" ], "publisher": "Relay Therapeutics", "date": "2024-12-03", "url": "https://ir.relaytx.com/news-releases/news-release-details/relay-therapeutics-and-elevar-therapeutics-announce-exclusive", "doi": null, "pmid": null, "trialId": "NCT04526106", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Primary company announcement. Elevar received worldwide development and commercial rights; Relay became eligible for up to $75m in upfront and regulatory milestones, up to $425m in commercial milestones and royalties. Contingent value is not cash received or research spend." }, { "id": "renovorx-2023-tigerpac-interim", "type": "company-release", "title": "RenovoRx phase III open-label TIGeR-PaC interim analysis", "authors": [ "RenovoRx, Inc." ], "publisher": "RenovoRx, Inc.", "date": "2023-03-08", "url": "https://ir.renovorx.com/news-events/press-releases/detail/57/renovorx-phase-iii-open-label-tiger-pac-interim-analysis", "doi": null, "pmid": null, "trialId": "NCT03257033", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Sponsor interim release. Forty-five people had been randomized: 23 intra-arterial and 22 control, with 13 deaths in each arm. Median survival from randomization was 16 versus 10 months; the release reports P=0.051 but no hazard ratio or confidence interval. It states that the medians omit four to five months of induction time. The monitoring committee recommended continuation. This did not cross the early stopping boundary and is not a final efficacy result." }, { "id": "renovorx-2025-10k", "type": "financial", "title": "RenovoRx annual report for the year ended December 31, 2025", "authors": [ "RenovoRx, Inc." ], "publisher": "U.S. Securities and Exchange Commission", "date": "2026-03-31", "url": "https://www.sec.gov/Archives/edgar/data/1574094/000119312526131928/rnxt-20251231.htm", "doi": null, "pmid": null, "trialId": "NCT03257033", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Primary company filing. After the trial started, RenovoRx amended the protocol in December 2021 to analyse only patients receiving stereotactic body radiation during induction, added a second interim analysis, reduced the randomized target from 200 to 114 and required 86 deaths, and reduced planned power from 90% to 80%. The filing says the modified plan was submitted to FDA, records regulator comments and says agreement is not assured. A second monitoring review at 52 deaths recommended continuation, but the company did not publish that interim result. At 24 March 2026, 104 people were randomized and 72 deaths had occurred." }, { "id": "renovorx-2026-full-enrollment", "type": "company-release", "title": "RenovoRx reports full enrollment in phase III TIGeR-PaC", "authors": [ "RenovoRx, Inc." ], "publisher": "RenovoRx, Inc.", "date": "2026-08-11", "url": "https://renovorx.com/renovorx-achieves-major-milestone-with-full-enrollment-in-phase-iii-tiger-pac-trial-for-locally-advanced-pancreatic-cancer/", "doi": null, "pmid": null, "trialId": "NCT03257033", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Sponsor release says investigators were told enrollment was closing on 7 August 2026 and that 78 of the required 86 deaths had occurred by 11 August. It projects trial completion in the first half of 2027 and top-line results in the second half. The release does not state the final number randomized. It confirms monitoring-committee continuation recommendations after 26 and 52 deaths; continuation does not establish efficacy." }, { "id": "renovorx-2026-q2-10q", "type": "financial", "title": "RenovoRx quarterly report for the period ended June 30, 2026", "authors": [ "RenovoRx, Inc." ], "publisher": "U.S. Securities and Exchange Commission", "date": "2026-08-13", "url": "https://www.sec.gov/Archives/edgar/data/1574094/000119312526349190/rnxt-20260630.htm", "doi": null, "pmid": null, "trialId": "NCT03257033", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Primary company financial source. RenovoRx reported $9.479m cash at June 30, 2026, $6.576m operating cash use and a $6.432m net loss for the first six months, and $1.472m six-month revenue. Simple unchanged-use coverage is $9.479m / ($6.576m / 6) = 8.65 months; this is a rough stress indicator, not formal runway guidance, because revenue, financing, trial payments and use can change." }, { "id": "revolution-medicines-2026-q2-10q", "type": "company-filing", "title": "Revolution Medicines Form 10-Q for the quarter ended June 30, 2026", "authors": [ "Revolution Medicines, Inc." ], "publisher": "U.S. Securities and Exchange Commission", "date": "2026-08-05", "url": "https://www.sec.gov/Archives/edgar/data/1628171/000119312526335104/rvmd-20260630.htm", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Primary company filing used for pipeline, cash, marketable securities and research spending. At 30 June 2026 cash plus marketable securities was $3.938 billion. First-half R&D was $738.9 million, including $227.8 million in third-party daraxonrasib costs." }, { "id": "rhee-2025-miracle-hcc", "type": "paper", "title": "Noncontrast Magnetic Resonance Imaging vs Ultrasonography for Hepatocellular Carcinoma Surveillance: A Randomized, Single-Center Trial", "authors": [ "Rhee, Hyungjin", "Kim, Myeong-Jin", "Kim, Do Young", "et al." ], "publisher": "Gastroenterology", "date": "2025-06-01", "url": "https://pubmed.ncbi.nlm.nih.gov/39855314/", "doi": "10.1053/j.gastro.2024.12.035", "pmid": "39855314", "trialId": "NCT02514434", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Single-centre South Korean randomized trial in 414 people with Child-Pugh A cirrhosis. Noncontrast MRI plus AFP produced an earlier BCLC stage and fewer false-positive referrals than ultrasound plus AFP, but BCLC 0/A detection rates were 12% versus 7% and not statistically different. The publication reports no mortality comparison." }, { "id": "rhim-2012-early-dissemination", "type": "paper", "title": "EMT and dissemination precede pancreatic tumor formation", "authors": [ "Rhim, Andrew D.", "Mirek, Emily T.", "Aiello, Nicole M.", "et al." ], "publisher": "Cell", "date": "2012-01-20", "url": "https://pubmed.ncbi.nlm.nih.gov/22265420/", "doi": "10.1016/j.cell.2011.11.025", "pmid": "22265420", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Lineage tracing in a genetically engineered mouse model found circulating and liver-seeding labelled cells before overt tumour formation. It does not establish the timing or clinical competence of early dissemination in humans." }, { "id": "rhim-2014-stroma-restrains", "type": "paper", "title": "Stromal elements act to restrain, rather than support, pancreatic ductal adenocarcinoma", "authors": [ "Rhim, Andrew D.", "Oberstein, Paul E.", "Thomas, Dafydd H.", "et al." ], "publisher": "Cancer Cell", "date": "2014-05-22", "url": "https://pubmed.ncbi.nlm.nih.gov/24856585/", "doi": "10.1016/j.ccr.2014.04.021", "pmid": "24856585", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Genetically engineered mouse work deleting Sonic hedgehog, plus pharmacological Smoothened inhibition. It establishes context-dependent stromal restraint in the model, not a clinical treatment result." }, { "id": "richards-2006-ci994-pdac", "type": "paper", "title": "Gemcitabine plus CI-994 offers no advantage over gemcitabine alone in advanced pancreatic cancer", "authors": [ "Richards, Donald A.", "Boehm, Kathryn A.", "Waterhouse, David M.", "et al." ], "publisher": "Annals of Oncology", "date": "2006-07-01", "url": "https://pubmed.ncbi.nlm.nih.gov/16641168/", "doi": "10.1093/annonc/mdl081", "pmid": "16641168", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Primary randomized double-blind phase 2 report, n=174. CI-994 plus gemcitabine produced median survival 194 versus 214 days, P=.908; investigator response 12% versus 14%, central response 1% versus 6%, worse two-month quality of life, and more neutropenia and thrombocytopenia." }, { "id": "rimassa-2025-himalaya-five-year", "type": "paper", "title": "Five-year overall survival update from the HIMALAYA study of tremelimumab plus durvalumab in unresectable HCC", "authors": [ "Rimassa, Lorenza", "Chan, Stephen L.", "Sangro, Bruno", "Kudo, Masatoshi", "Reig, Maria", "Abou-Alfa, Ghassan K." ], "publisher": "Journal of Hepatology", "date": "2025-04-11", "url": "https://pubmed.ncbi.nlm.nih.gov/40222621/", "doi": "10.1016/j.jhep.2025.03.033", "pmid": "40222621", "trialId": "NCT03298451", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Exploratory five-year update of the same randomized HIMALAYA trial: STRIDE n=393, sorafenib n=389, OS HR 0.76, 60-month OS 19.6% versus 9.4%. Several authors were AstraZeneca staff and many disclosed company ties." }, { "id": "riquelme-2019-pdac-microbiome-survival", "type": "paper", "title": "Tumor microbiome diversity and composition influence pancreatic cancer outcomes", "authors": [ "Riquelme, Ericka", "Zhang, Yu", "Zhang, Liang", "et al." ], "publisher": "Cell", "date": "2019-08-08", "url": "https://pubmed.ncbi.nlm.nih.gov/31398337/", "doi": "10.1016/j.cell.2019.07.008", "pmid": "31398337", "trialId": null, "patentId": null, "grantId": "SU2C-AACR-DT25-17", "accessed": "2026-09-15", "notes": "Human survivor comparison and mouse transfer study. The discovery cohort compared 21 long-term and 22 short-term survivors, followed by an independent resection cohort. Higher tumour bacterial diversity and a four-taxon signature tracked survival. Stool transfer from human groups changed tumour growth and immune entry in mice, not in patients." }, { "id": "rodon-2024-amg193-prmt5", "type": "paper", "title": "First-in-human study of AMG 193, an MTA-cooperative PRMT5 inhibitor, in MTAP-deleted solid tumors", "authors": [ "Rodon, Jordi", "Prenen, Hans", "Sacher, Adrian", "et al." ], "publisher": "Annals of Oncology", "date": "2024-12-01", "url": "https://pubmed.ncbi.nlm.nih.gov/39293516/", "doi": "10.1016/j.annonc.2024.08.2339", "pmid": "39293516", "trialId": "NCT05094336", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Primary phase 1 dose-exploration report, n=80. At active tolerable doses, ORR was 21.4% in 42 evaluable patients across tumour types, with responses including pancreatic and biliary cancers but no PDAC denominator. Paired biopsies at doses at least 480 mg showed complete intratumour PRMT5 inhibition. Eight patients had dose-limiting toxicities; no clinically significant myelosuppression was reported." }, { "id": "rodriguez-2021-alcohol-abstinence-hcc", "type": "paper", "title": "Impact of alcohol abstinence on the risk of hepatocellular carcinoma in patients with alcohol-related liver cirrhosis", "authors": [ "Rodríguez, Manuel", "González-Diéguez, María Luisa", "Varela, María", "et al." ], "publisher": "American Journal of Gastroenterology", "date": "2021-12-01", "url": "https://pubmed.ncbi.nlm.nih.gov/34569986/", "doi": "10.14309/ajg.0000000000001399", "pmid": "34569986", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Prospective Spanish surveillance cohort of 727 people with alcohol-related cirrhosis, median follow-up 54 months. Abstinence was associated with lower HCC incidence among 247 people without previous decompensation, adjusted HR 0.35, but not among 480 with previous decompensation, HR 0.95. Observational exposure and subgroup analysis prevent a causal effect estimate." }, { "id": "rodriguez-2024-generate", "type": "paper", "title": "A Randomized Trial of Two Remote Health Care Delivery Models on the Uptake of Genetic Testing and Impact on Patient-Reported Psychological Outcomes in Families With Pancreatic Cancer: The GENERATE Study", "authors": [ "Rodriguez, Nicolette J.", "Furniss, C. Sloane", "Yurgelun, Matthew B.", "et al." ], "publisher": "Gastroenterology", "date": "2024-02-05", "url": "https://pubmed.ncbi.nlm.nih.gov/38320723/", "doi": "10.1053/j.gastro.2024.01.042", "pmid": "38320723", "trialId": "NCT03762590", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Family-cluster randomized trial in 601 motivated US relatives from 424 families. Testing completion was 87% with live telemedicine plus online education and 93% with online education alone; mean anxiety, depression and cancer worry did not differ. All participants had entered the study and were offered covered home testing, so the trial does not measure how many relatives are never reached or invited." }, { "id": "rojass-2025-neoantigen-followup", "type": "paper", "title": "RNA neoantigen vaccines prime long-lived CD8+ T cells in pancreatic cancer", "authors": [ "Rojas, Luis A.", "Sethna, Zachary", "Soares, Katelyn C.", "et al." ], "publisher": "Nature", "date": "2025-02-19", "url": "https://www.nature.com/articles/s41586-024-08508-4", "doi": "10.1038/s41586-024-08508-4", "pmid": null, "trialId": "NCT04161755", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Extended immune and clinical follow-up from a single-centre phase 1 combination study. Sixteen vaccinated patients were evaluable; recurrence-free survival was compared between immune responders and non-responders, not randomised treatment groups. Genentech and BioNTech were involved." }, { "id": "rompen-2026-preopanc2-staging", "type": "paper", "title": "Staging Laparoscopy After Neoadjuvant Treatment for Pancreatic Cancer to Prevent Non-Therapeutic Laparotomy: Analysis in the PREOPANC-2 Randomized Trial", "authors": [ "Rompen, Ingmar F", "Dekker, Esther N", "Stoop, Thomas F", "et al." ], "publisher": "Annals of Surgery", "date": "2026-02-12", "url": "https://pubmed.ncbi.nlm.nih.gov/41673930/", "doi": "10.1097/SLA.0000000000007025", "pmid": "41673930", "trialId": "NTR7292", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Prespecified analysis within PREOPANC-2. Of 369 randomized patients, 322 underwent exploration after treatment; 240 began with same-session staging laparoscopy and 82 with laparotomy. Hidden spread occurred in 39/322, 12.1%, and caused 90.5% of aborted operations. Non-therapeutic laparotomy was 4.5% with laparoscopy versus 17.1% without, absolute difference 12.6 points, P=.002, reported NNT 8. Staging method itself was not randomized. Baseline CA19-9 above 500 U/ml and tumour size at least 3 cm marked higher yield; the zero-factor group still had 2.8% hidden spread, 95% CI 0.9-8.0." }, { "id": "rosemurgy-2008-p32-randomized", "type": "paper", "title": "Phosphorus-32 as an adjunct to standard therapy for locally advanced unresectable pancreatic cancer: a randomized trial", "authors": [ "Rosemurgy, Alexander", "Luzardo, German", "Cooper, Jennifer", "et al." ], "publisher": "Journal of Gastrointestinal Surgery", "date": "2008-04-01", "url": "https://pubmed.ncbi.nlm.nih.gov/18266048/", "doi": "10.1007/s11605-007-0430-6", "pmid": "18266048", "trialId": "NCT00079365", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Thirty assessable people randomized to fluorouracil, external radiation and gemcitabine with or without repeated colloidal phosphorus-32. Tumour liquefaction was 78% versus 8%, but serious events per patient were 4.2 versus 1.8, p=0.03, and survival was 7.4 versus 11.5 months, p=0.16. The formulation and schedule differ from modern silicon microparticles." }, { "id": "ross-2022-panco-p32", "type": "paper", "title": "Results of a single-arm pilot study of phosphorus-32 microparticles in unresectable locally advanced pancreatic adenocarcinoma with gemcitabine/nab-paclitaxel or FOLFIRINOX chemotherapy", "authors": [ "Ross, Paul J.", "Wasan, Harpreet S.", "Croagh, Daniel", "et al." ], "publisher": "ESMO Open", "date": "2022-02-01", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC8717429/", "doi": "10.1016/j.esmoop.2021.100356", "pmid": "34953400", "trialId": "NCT03003078", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Ten-centre single-arm PanCO pilot. Fifty began chemotherapy and 42 received OncoSil implantation. Sixteen of 42 had 41 device- or procedure-related events, including three people with grade 3 or worse events. Sixteen-week local disease control was 82% in all 50; 10 underwent resection; median survival 15.2 months. OncoSil funded the work and employed or owned shares through several authors." }, { "id": "rougier-2013-aflibercept-vanilla", "type": "paper", "title": "Aflibercept with gemcitabine in first-line metastatic pancreatic cancer: randomized placebo-controlled phase III", "authors": [ "Rougier, Philippe", "Riess, Hanno", "Manges, Robert", "et al." ], "publisher": "European Journal of Cancer", "date": "2013-04-30", "url": "https://pubmed.ncbi.nlm.nih.gov/23642329/", "doi": "10.1016/j.ejca.2013.04.002", "pmid": "23642329", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Stopped for futility. At termination 546 people had been randomized: aflibercept plus gemcitabine n=271 versus placebo plus gemcitabine n=275. Median OS was 6.5 versus 7.8 months, HR 1.165, 95% CI 0.921-1.473, P=.2034; PFS was 3.7 months in both. Adverse-event discontinuation was 23% versus 12%." }, { "id": "royal-2010-ipilimumab", "type": "paper", "title": "Phase 2 trial of single agent Ipilimumab (anti-CTLA-4) for locally advanced or metastatic pancreatic adenocarcinoma", "authors": [ "Royal, Richard E.", "Levy, Carolyn", "Turner, Kevin", "et al." ], "publisher": "Journal of Immunotherapy", "date": "2010-10-01", "url": "https://pubmed.ncbi.nlm.nih.gov/20842054/", "doi": "10.1097/CJI.0b013e3181eec14c", "pmid": "20842054", "trialId": "NCT00112580", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Single-arm phase 2 in 27 fit patients. There were no RECIST responses; one delayed response after initial progression was described." }, { "id": "roybal-2016-combinatorial-t-cell-circuit", "type": "paper", "title": "Precision Tumor Recognition by T Cells With Combinatorial Antigen-Sensing Circuits", "authors": [ "Roybal, Kole T.", "Rupp, Levi J.", "Morsut, Leonardo", "et al." ], "publisher": "Cell", "date": "2016-02-11", "url": "https://escholarship.org/uc/item/4cd8c5s4", "doi": "10.1016/j.cell.2016.01.011", "pmid": "26830879", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Engineered T cells used one synthetic receptor to induce a CAR against a second antigen, creating an AND gate that killed dual-antigen tumours while sparing single-antigen bystanders in models. This is a programmable safety principle, not human PDAC efficacy; it still requires a stable pancreatic antigen pair, tumour entry and survival in the suppressive microenvironment." }, { "id": "rudno-rudzinska-2021-irec-protocol", "type": "paper", "title": "Effects of calcium electroporation, electrochemotherapy, and irreversible electroporation on quality of life and progression-free survival in patients with pancreatic cancer: IREC clinical study", "authors": [ "Rudno-Rudzińska, Julia", "Kielan, Wojciech", "Guziński, Maciej", "Kulbacka, Julita" ], "publisher": "Advances in Clinical and Experimental Medicine", "date": "2021-07-01", "url": "https://pubmed.ncbi.nlm.nih.gov/34313408/", "doi": "10.17219/acem/139917", "pmid": "34313408", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Describes a planned three-arm 1:1:1 comparison of irreversible electroporation, calcium electroporation and bleomycin electrochemotherapy. The abstract's Results section states future endpoints rather than participant outcomes and provides no registry identifier or sample size." }, { "id": "ruers-2017-clocc-crc-liver-ablation", "type": "paper", "title": "Local treatment of unresectable colorectal liver metastases: results of a randomized phase II trial", "authors": [ "Ruers, Theo", "Van Coevorden, Frits", "Punt, Cornelis J.", "et al." ], "publisher": "Journal of the National Cancer Institute", "date": "2017-09-01", "url": "https://pubmed.ncbi.nlm.nih.gov/28376151/", "doi": "10.1093/jnci/djx015", "pmid": "28376151", "trialId": "NCT00043004", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Long-term report of EORTC 40004/CLOCC, randomized phase 2, n=119. Systemic treatment plus radiofrequency ablation with or without resection improved long-term OS in patients with fewer than ten unresectable liver metastases and no extrahepatic disease." }, { "id": "safyan-2026-azacitidine-pembrolizumab-pdac", "type": "paper", "title": "Azacitidine plus pembrolizumab as second-line treatment in pancreatic ductal adenocarcinoma", "authors": [ "Safyan, Rachel A.", "White, Richard A.", "Gonda, Tamas A.", "et al." ], "publisher": "The Oncologist", "date": "2026-03-09", "url": "https://pubmed.ncbi.nlm.nih.gov/41844546/", "doi": "10.1093/oncolo/oyag091", "pmid": "41844546", "trialId": "NCT03264404", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Primary single-arm phase 2 report: 36 enrolled, 34 in safety and 31 in efficacy analysis. Three partial responses gave ORR 9.7%; disease control was 35.5%, median PFS 1.51 months and median OS 4.83 months. Baseline CD8 and Ki67 findings were exploratory." }, { "id": "saito-2018-pert-rct", "type": "paper", "title": "A Multicenter Open-Label Randomized Controlled Trial of Pancreatic Enzyme Replacement Therapy in Unresectable Pancreatic Cancer", "authors": [ "Saito, Tomotaka", "Nakai, Yousuke", "Isayama, Hiroyuki", "et al." ], "publisher": "Pancreas", "date": "2018-08-01", "url": "https://pubmed.ncbi.nlm.nih.gov/29851751/", "doi": "10.1097/MPA.0000000000001079", "pmid": "29851751", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Japanese multicentre open-label randomized trial, n=88 receiving chemotherapy; 90% had a low NBT-PABA test. Eight-week BMI and other nutrition measures did not differ. Median OS was 19.0 versus 12.0 months, P=.070, a possible but unconfirmed survival signal." }, { "id": "sanders-2024-asco-palliative-guideline", "type": "guideline", "title": "Palliative Care for Patients With Cancer: ASCO Guideline Update", "authors": [ "Sanders, Justin J.", "Temin, Sarah", "Ghoshal, Arun", "et al." ], "publisher": "Journal of Clinical Oncology", "date": "2024-07-01", "url": "https://doi.org/10.1200/JCO.24.00542", "doi": "10.1200/JCO.24.00542", "pmid": "38748941", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Evidence-based guideline drawing on 52 studies. It recommends early specialist interdisciplinary palliative care alongside active cancer treatment for advanced cancer, with priority for uncontrolled physical, psychological or spiritual distress. It notes intervention and timing variation and no single universal time threshold; trials commonly began within 8 to 12 weeks." }, { "id": "sandhya-2023-olanzapine-anorexia", "type": "paper", "title": "Randomized Double-Blind Placebo-Controlled Study of Olanzapine for Chemotherapy-Related Anorexia in Patients With Locally Advanced or Metastatic Gastric, Hepatopancreaticobiliary, and Lung Cancer", "authors": [ "Sandhya, Lakshmi", "Sreenivasan, Nirmala Devi", "Goenka, Luxitaa", "et al." ], "publisher": "Journal of Clinical Oncology", "date": "2023-05-10", "url": "https://pubmed.ncbi.nlm.nih.gov/36977285/", "doi": "10.1200/JCO.22.01997", "pmid": "36977285", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Single-centre Indian randomized double-blind trial: 124 enrolled, 112 analyzable; only 13 had hepatopancreaticobiliary cancer. Olanzapine 2.5 mg daily plus nutrition advice produced greater than 5% weight gain in 60% versus 9%, better appetite, quality of life and reported chemotherapy toxicity. It did not test survival and cannot establish a PDAC-specific effect." }, { "id": "sankaranarayanan-2009-hpv-screening-india", "type": "paper", "title": "HPV screening for cervical cancer in rural India", "authors": [ "Sankaranarayanan, Rengaswamy", "Nene, Bhagwan M.", "Shastri, Surendra S.", "et al." ], "publisher": "New England Journal of Medicine", "date": "2009-04-02", "url": "https://pubmed.ncbi.nlm.nih.gov/19339719/", "doi": "10.1056/NEJMoa0808516", "pmid": "19339719", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Primary 131,746-woman cluster-randomized trial: one HPV-testing round reduced advanced cervical cancer, HR 0.47, and cervical-cancer death, HR 0.52, versus standard care." }, { "id": "sarno-2025-avatar", "type": "paper", "title": "A Phase III Randomized Trial of Integrated Genomics and Avatar Models for Personalized Treatment of Pancreatic Cancer: The AVATAR Trial", "authors": [ "Sarno, Francesca", "Tenorio, Jair", "Perea, Sofia", "et al." ], "publisher": "Clinical Cancer Research", "date": "2025-01-17", "url": "https://pubmed.ncbi.nlm.nih.gov/39540844/", "doi": "10.1158/1078-0432.CCR-23-4026", "pmid": "39540844", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Spanish multicentre phase III trial; 125 patients randomized 1:2 to physician-choice treatment or a whole-exome, mouse-avatar, organoid and molecular-board programme." }, { "id": "sartore-bianchi-2016-heracles", "type": "paper", "title": "Dual HER2 blockade in treatment-refractory HER2-positive metastatic colorectal cancer: HERACLES", "authors": [ "Sartore-Bianchi, Andrea", "Trusolino, Livio", "Martino, Carla", "et al." ], "publisher": "The Lancet Oncology", "date": "2016-06-01", "url": "https://pubmed.ncbi.nlm.nih.gov/27108243/", "doi": "10.1016/S1470-2045(16)00150-9", "pmid": "27108243", "trialId": "EudraCT 2012-002128-33", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Four-centre Italian proof-of-concept phase 2. Of 914 KRAS-exon-2-wild-type metastatic colorectal cancers screened, 48, 5%, were HER2 positive by colorectal-specific criteria; 27 were treated with trastuzumab plus lapatinib. Objective response was 30%, 95% CI 14-50, including one complete response. No randomized control and very small selected cohort." }, { "id": "sasieni-2026-nhs-galleri", "type": "conference-paper", "title": "NHS-Galleri: Primary results from a randomised controlled trial to assess the clinical utility of a multi-cancer early detection test in population screening", "authors": [ "Sasieni, Peter", "Neal, Richard", "et al." ], "publisher": "Journal of Clinical Oncology, 2026 ASCO Annual Meeting", "date": "2026-05-30", "url": "https://ascopubs.org/doi/10.1200/JCO.2026.44.17_suppl.LBA100", "doi": "10.1200/JCO.2026.44.17_suppl.LBA100", "pmid": null, "trialId": "ISRCTN91431511", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Primary randomized NHS-Galleri conference report. The sponsor's public fact sheet provided the full main-effect estimate: stage III/IV incidence-rate ratio 1.03, 95% CI 0.92-1.14, P=0.6324. Full peer-reviewed paper and mortality follow-up were not yet available." }, { "id": "scarpa-2017-pnet", "type": "paper", "title": "Whole-genome landscape of pancreatic neuroendocrine tumours", "authors": [ "Scarpa, Aldo", "Chang, David K.", "Nones, Katia", "et al." ], "publisher": "Nature", "date": "2017-02-15", "url": "https://pubmed.ncbi.nlm.nih.gov/28199314/", "doi": "10.1038/nature21063", "pmid": "28199314", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Whole-genome sequencing of 102 primary pancreatic neuroendocrine tumours; described distinct mutational processes and a larger-than-expected germline contribution." }, { "id": "schauer-2019-bariatric-pancreatic-risk", "type": "paper", "title": "Bariatric Surgery and the Risk of Cancer in a Large Multisite Cohort", "authors": [ "Schauer, Daniel P.", "Feigelson, Heather Spencer", "Koebnick, Corinna", "et al." ], "publisher": "Annals of Surgery", "date": "2019-01", "url": "https://pubmed.ncbi.nlm.nih.gov/28938270/", "doi": "10.1097/SLA.0000000000002525", "pmid": "28938270", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Retrospective five-site cohort of 22,198 bariatric-surgery patients and 66,427 matched nonsurgical patients. Pancreatic-cancer HR was 0.46, 95% CI 0.22-0.97, after mean follow-up of 3.5 years. The rare endpoint, wide interval, treatment selection and short mean follow-up prevent a pancreatic-specific causal claim." }, { "id": "schrag-2023-pathfinder", "type": "paper", "title": "Blood-based tests for multicancer early detection (PATHFINDER): a prospective cohort study", "authors": [ "Schrag, Deborah", "Beer, Tomasz M.", "McDonnell, Charles H.", "et al." ], "publisher": "The Lancet", "date": "2023-10-01", "url": "https://pubmed.ncbi.nlm.nih.gov/37805216/", "doi": "10.1016/S0140-6736(23)01700-2", "pmid": "37805216", "trialId": "NCT04241796", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Prospective result-return study in 6,662 adults aged 50 or older at seven US health networks. Convenience sample, 91.7% White, funded by GRAIL. Primary outcome was diagnostic resolution, not mortality." }, { "id": "schrag-2023-prospect-rectal", "type": "paper", "title": "Preoperative treatment of locally advanced rectal cancer: PROSPECT", "authors": [ "Schrag, Deborah", "et al." ], "publisher": "The New England Journal of Medicine", "date": "2023-06-04", "url": "https://pubmed.ncbi.nlm.nih.gov/37272534/", "doi": "10.1056/NEJMoa2303269", "pmid": "37272534", "trialId": "NCT01515787", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "NCI-funded multicentre randomized noninferiority trial; 1,194 patients with selected locally advanced rectal cancer, 1,128 treated. Neoadjuvant FOLFOX with selective radiation was noninferior to chemoradiotherapy: five-year DFS 80.8% versus 78.6%, HR 0.92, 90.2% CI 0.74-1.14. Only 9.1% in the FOLFOX arm needed preoperative chemoradiotherapy. Applies to sphincter-sparing surgical candidates, not all rectal cancer." }, { "id": "schram-2025-zenocutuzumab-nrg1", "type": "paper", "title": "Efficacy of Zenocutuzumab in NRG1 Fusion-Positive Cancer", "authors": [ "Schram, Alison M.", "Goto, Koichi", "Kim, Dong-Wan", "et al." ], "publisher": "New England Journal of Medicine", "date": "2025-02-06", "url": "https://doi.org/10.1056/NEJMoa2405008", "doi": "10.1056/NEJMoa2405008", "pmid": "39908431", "trialId": "NCT02912949", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Sponsor-funded open-label phase II report. Among 158 efficacy-evaluable patients across cancers, ORR was 30% and median response duration 11.1 months. Fifteen of 36 pancreatic patients responded, 42%, 95% CI 25% to 59%. Median progression-free survival across tumour types was 6.8 months. This later publication set is larger than the FDA's 30-person pancreatic approval set but remains uncontrolled." }, { "id": "schram-2026-nrg1-fusion-detection", "type": "paper", "title": "NRG1 fusion-positive solid tumors: clinical detection, genomic landscape, and real-world data in pancreatic cancer", "authors": [ "Schram, Alison M.", "Yang, Soo-Ryum", "Kahn, Genessa", "et al." ], "publisher": "Journal of the National Cancer Institute", "date": "2026-05-01", "url": "https://doi.org/10.1093/jnci/djaf361", "doi": "10.1093/jnci/djaf361", "pmid": "41390931", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Retrospective institutional series of NRG1-fusion cancers. Of 48 patients, 34 were found by RNA testing, 11 by DNA and three by both; RNA was more effective because NRG1 has many partners and large possible breakpoint regions. The pancreatic cases were KRAS wild type. The small PDAC outcome subset had median age 48.5 years and cannot establish comparative prognosis or treatment benefit." }, { "id": "sec-2025-panbela-registration-abandoned", "type": "regulatory", "title": "Order declaring Panbela registration statement abandoned", "authors": [ "U.S. Securities and Exchange Commission" ], "publisher": "U.S. Securities and Exchange Commission", "date": "2025-09-29", "url": "https://www.sec.gov/Archives/edgar/data/1029125/999999999725003686/filename1.pdf", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "The SEC declared a Panbela securities registration statement abandoned after it remained ineffective for more than nine months and the company did not respond to notice. This does not determine the trial's scientific result, but adds a current governance and financing warning." }, { "id": "sec-2026-actuate-submissions", "type": "dataset", "title": "Actuate Therapeutics SEC filing submissions through 15 September 2026", "authors": [ "U.S. Securities and Exchange Commission" ], "publisher": "SEC EDGAR", "date": "2026-09-15", "url": "https://data.sec.gov/submissions/CIK0001652935.json", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Primary filing index checked for financing after the 30 June balance-sheet date. The latest company reports were the 14 August 10-Q and 17 July 8-K; no later 8-K, 10-Q or registered financing filing was present through the access date. Insider Form 4 filings do not finance the company." }, { "id": "seelen-2026-pelican", "type": "paper", "title": "Radiofrequency Ablation and Chemotherapy vs Chemotherapy Only in Locally Advanced Pancreatic Cancer: The PELICAN Randomized Clinical Trial", "authors": [ "Seelen, Leonard W. F.", "Brada, Lilly J. H.", "Walma, Marieke S.", "et al." ], "publisher": "JAMA Network Open", "date": "2026-03-02", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC13040403/", "doi": "10.1001/jamanetworkopen.2026.3861", "pmid": "41915391", "trialId": "NCT03690323", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "International randomized trial in 188 people with unresectable locally advanced PDAC that had not progressed after two months of multiagent chemotherapy. Adding radiofrequency ablation produced median survival 12.1 versus 11.6 months, HR 1.07, 95% CI 0.80-1.45, and no progression-free benefit. Grade 3 or worse serious adverse events were 27% versus 11%, and global quality-of-life scores fell by clinically important amounts in the ablation arm. This directly rejects that an invasive local procedure should be inferred useful from selected-cohort survival. It does not test tumour resection." }, { "id": "seer-colorectal-statfacts", "type": "registry", "title": "Colorectal Cancer — Cancer Stat Facts", "authors": [ "National Cancer Institute Surveillance, Epidemiology, and End Results Program" ], "publisher": "National Cancer Institute", "date": null, "url": "https://seer.cancer.gov/statfacts/html/colorect.html", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Checked the live table. Stage-specific five-year relative survival uses SEER 21, excluding Illinois, 2016–2022." }, { "id": "seer-liver-intrahepatic-statfacts", "type": "registry", "title": "Liver and Intrahepatic Bile Duct Cancer — Cancer Stat Facts", "authors": [ "National Cancer Institute Surveillance, Epidemiology, and End Results Program" ], "publisher": "National Cancer Institute", "date": null, "url": "https://seer.cancer.gov/statfacts/html/livibd.html", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Checked the live table. It combines liver and intrahepatic bile-duct primaries, so it cannot support HCC-only or cholangiocarcinoma-only estimates." }, { "id": "seer-pancreas-statfacts", "type": "registry", "title": "Pancreatic Cancer — Cancer Stat Facts", "authors": [ "National Cancer Institute Surveillance, Epidemiology, and End Results Program" ], "publisher": "National Cancer Institute", "date": null, "url": "https://seer.cancer.gov/statfacts/html/pancreas.html", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Checked the live table. Incidence uses 2019–2023 cases; overall and stage-specific five-year relative survival use SEER 21, excluding Illinois, 2016–2022. The page combines pancreatic cancer histologies." }, { "id": "seif-el-dahan-2026-early-hcc-causes-death", "type": "journal-article", "title": "Causes of mortality among patients with early-stage hepatocellular carcinoma", "authors": [ "Seif El Dahan, Karim", "Daher, Darine", "Rich, Nicole E.", "et al." ], "publisher": "Hepatology", "date": "2026-06-01", "url": "https://pubmed.ncbi.nlm.nih.gov/40694813/", "doi": "10.1097/HEP.0000000000001471", "pmid": "40694813", "trialId": null, "patentId": null, "grantId": "P30 DK127984; R01 CA256977; U01 CA271887; U01 CA283935", "accessed": "2026-09-14", "notes": "Retrospective cohort of 1,336 BCLC 0/A HCC patients across four US health systems, 2008-2022. During median 32.4-month follow-up, 598 died: 220 HCC-related, 114 non-HCC liver-related and 179 other-cause deaths. MASLD, Child-Pugh B and ALBI grade 2 predicted liver-related mortality. Cause assignment and treatment associations are observational; authors report several industry conflicts." }, { "id": "seo-2019-covered-uncovered-sems-rct", "type": "paper", "title": "Covered and uncovered biliary metal stents provide similar relief during neoadjuvant therapy in pancreatic cancer", "authors": [ "Seo, Dong Wan", "Sherman, Stuart", "Dua, Kulwinder S", "et al." ], "publisher": "Gastrointestinal Endoscopy", "date": "2019-10-01", "url": "https://pubmed.ncbi.nlm.nih.gov/31276674/", "doi": "10.1016/j.gie.2019.06.032", "pmid": "31276674", "trialId": "NCT02238847", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Randomized 119 pancreatic-cancer patients needing drainage during neoadjuvant therapy to fully covered or uncovered metal stents. Sustained drainage was 72.2% versus 72.9%, meeting noninferiority. Covered stents had no tumour ingrowth but 6.8% migration; uncovered stents had 16.7% ingrowth and no migration. Serious stent-related events were 23.7% versus 20.0%. Boston Scientific employees were authors and the study used its WallFlex system. The result supports individual design choice rather than a universal covered-stent winner." }, { "id": "sharma-2018-endpac", "type": "paper", "title": "Model to Determine Risk of Pancreatic Cancer in Patients With New-Onset Diabetes", "authors": [ "Sharma, Abhiram", "Kandlakunta, Harika", "Nagpal, Sajan Jiv Singh", "et al." ], "publisher": "Gastroenterology", "date": "2018-06-11", "url": "https://pubmed.ncbi.nlm.nih.gov/29775599/", "doi": "10.1053/j.gastro.2018.05.023", "pmid": "29775599", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Retrospective development and locked validation in Rochester Epidemiology Project cohorts. The nine validation cancers make sensitivity imprecise; prospective validation was explicitly required." }, { "id": "sharma-2019-ugt1a1-mfolfirinox", "type": "paper", "title": "A UGT1A1 genotype-guided dosing study of modified FOLFIRINOX in advanced gastrointestinal malignancies", "authors": [ "Sharma, Manish R", "Joshi, Smita S", "Karrison, Theodore G", "et al." ], "publisher": "Cancer", "date": "2019-05-15", "url": "https://doi.org/10.1002/cncr.31938", "doi": "10.1002/cncr.31938", "pmid": "30645764", "trialId": "NCT01643499", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Genotype-guided irinotecan doses produced cycle-one dose-limiting toxicity in 2/15 *1/*1, 3/16 *1/*28 and 4/10 *28/*28 participants. In tumour expansion, 6/19 pancreatic and 4/19 biliary patients had dose-limiting toxicity and the upper confidence limits exceeded the prespecified 33% tolerability bound. The pancreatic and biliary genotype-guided regimen therefore failed its declared tolerability rule." }, { "id": "sharma-2026-fmtpanc", "type": "paper", "title": "Phase 1 faecal microbiota transplantation in patients with advanced pancreatic carcinoma: FMTPanc trial", "authors": [ "Sharma, A.", "et al." ], "publisher": "Journal of Clinical Oncology", "date": "2026-06-01", "url": "https://ascopubs.org/doi/10.1200/JCO.2026.44.16_suppl.TPS4262", "doi": "10.1200/JCO.2026.44.16_suppl.TPS4262", "pmid": null, "trialId": "ACTRN12624000455561", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Trial-in-progress abstract. Phase 1 double-blind randomized FMT-versus-placebo capsules with chemotherapy in unresectable PDAC; minimum 28 and maximum 60 participants, 14-30 per arm. Twelve participants were reported enrolled as of January 2025. BiomeBank prepares the capsules. No outcome result was reported." }, { "id": "shi-2026-chemo-dormant-dtc", "type": "paper", "title": "Chemotherapy reawakens dormant disseminated pancreatic cancer cells in the liver by promoting senescent hepatocyte-induced neutrophil extracellular traps", "authors": [ "Shi, Shangheng", "Yan, Jingrui", "Zhang, Yu", "et al." ], "publisher": "Cancer Research", "date": "2026-09-04", "url": "https://doi.org/10.1158/0008-5472.CAN-26-0151", "doi": "10.1158/0008-5472.CAN-26-0151", "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "New primary study from Tianjin. Dormancy tracking and perturbation were performed in models; human material showed treatment-associated hepatocyte senescence and clinical correlations. The accessible publisher record did not expose cohort sizes, and there is no human intervention test of the proposed TGF-beta to NET to CCDC25 route." }, { "id": "shinoto-2016-carbon-ion-gemcitabine", "type": "paper", "title": "Carbon Ion Radiation Therapy With Concurrent Gemcitabine for Patients With Locally Advanced Pancreatic Cancer", "authors": [ "Shinoto, Makoto", "Yamada, Shigeru", "Terashima, Kotaro", "et al." ], "publisher": "International Journal of Radiation Oncology, Biology, Physics", "date": "2016-05-01", "url": "https://pubmed.ncbi.nlm.nih.gov/26883565/", "doi": "10.1016/j.ijrobp.2015.12.362", "pmid": "26883565", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Single-arm Japanese dose-escalation study. It enrolled 76 and treated 72 people with carbon-ion radiation and gemcitabine. Two-year freedom from local progression was 83%; two-year overall survival was 35% overall and 48% in the higher-dose stage III group." }, { "id": "shirasu-2019-ugt1a1-folfirinox", "type": "paper", "title": "Impact of UGT1A1 genetic polymorphism on toxicity in unresectable pancreatic cancer patients undergoing FOLFIRINOX", "authors": [ "Shirasu, Hiromichi", "Todaka, Akiko", "Omae, Katsuhiro", "et al." ], "publisher": "Cancer Science", "date": "2019-02-01", "url": "https://doi.org/10.1111/cas.13883", "doi": "10.1111/cas.13883", "pmid": "30447099", "trialId": "UMIN000014658", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "JASPAC 06 analysis included 199 unresectable pancreatic-cancer patients. With original FOLFIRINOX, 25 heterozygous UGT1A1*6 or *28 carriers had more diarrhoea than 54 wild-type patients; the difference was absent among 120 patients receiving modified FOLFIRINOX. The observational regimen split does not prove genotype-guided dosing utility." }, { "id": "shsmu-2026-liwei-wang-profile", "type": "institution-profile", "title": "Wang Liwei — Shanghai Jiao Tong University School of Medicine profile", "authors": [ "Shanghai Jiao Tong University School of Medicine" ], "publisher": "Shanghai Jiao Tong University School of Medicine", "date": "2026-01-01", "url": "https://www.shsmu.edu.cn/english/info/1354/4080.htm", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Institutional profile lists Wang as professor, chief physician and chair of Oncology and the Comprehensive Pancreatic Cancer Center at Renji Hospital." }, { "id": "sigel-2018-panenen-differentiation", "type": "paper", "title": "Assessment of cytologic differentiation in high-grade pancreatic neuroendocrine neoplasms: a multi-institutional study", "authors": [ "Sigel, Carlie S.", "Klimstra, David S.", "et al." ], "publisher": "Cancer Cytopathology", "date": "2018-01-01", "url": "https://pubmed.ncbi.nlm.nih.gov/29044913/", "doi": "10.1002/cncy.21934", "pmid": "29044913", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Multi-institutional assessment of 70 evaluable fine-needle aspiration samples from grade 2 or grade 3 pancreatic neuroendocrine neoplasms. Three expert cytopathologists agreed on well versus poorly differentiated status in only 38% overall. Final classification used RB, DAXX and ATRX immunohistochemistry, mutation analysis, prior histology and consensus. Diagnostic study, not a treatment comparison." }, { "id": "silexion-2026-halfyear-filing", "type": "company-filing", "title": "Silexion Therapeutics foreign issuer report and interim financial statements, 2026", "authors": [ "Silexion Therapeutics Corp." ], "publisher": "U.S. Securities and Exchange Commission", "date": "2026-08-01", "url": "https://www.sec.gov/Archives/edgar/data/2022416/000117891326004003/zk2635896.htm", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Primary filing used for corporate history, first-generation LODER interpretation, second-generation SIL204 programme, planned adaptive phase 2/3 and financing context. Company statements about efficacy are not independent clinical evidence." }, { "id": "silexion-2026-sil204-site-initiation", "type": "company-filing", "title": "Silexion Therapeutics Successfully Initiates Phase 2/3 Clinical Trial of SIL204 in Locally Advanced Pancreatic Cancer at Tel Aviv Sourasky Medical Center", "authors": [ "Silexion Therapeutics Corp." ], "publisher": "U.S. Securities and Exchange Commission", "date": "2026-07-29", "url": "https://www.sec.gov/Archives/edgar/data/2022416/000117891326003690/exhibit_99-1.htm", "doi": null, "pmid": null, "trialId": "2025-525157-39-00", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Company release states that one Israeli site was activated and cleared to begin screening; first dosing was still anticipated, not reported. It describes about 18 in the safety run-in and a randomized cohort of about 166, while CTIS and the later filing plan about 403 overall." }, { "id": "sinan-2025-pdac-germline-implementation", "type": "paper", "title": "Cancer Susceptibility Gene Testing in Patients With Pancreatic Ductal Adenocarcinoma: Implementation in a Cancer Center Oncology Clinic", "authors": [ "Sinan, Hale", "Cunningham, Danielle", "Sleiman, Elias A.", "et al." ], "publisher": "JCO Precision Oncology", "date": "2025-02-12", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC11824855/", "doi": "10.1200/PO-24-00494", "pmid": "39938008", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Johns Hopkins chart review of 992 patients. Ninety percent were offered testing and 77.6% of those offered completed it, about 69.8% of all patients. Among tested patients, 78 had a pathogenic variant; 72 were referred for counselling and 50, 69.4%, completed it. Testing changed chemotherapy in 28 patients. Older, single, African American and advanced-stage patients were less likely to proceed." }, { "id": "singal-2024-tracer-protocol", "type": "paper", "title": "National Liver Cancer Screening Trial (TRACER) study protocol", "authors": [ "Singal, Amit G", "Parikh, Neehar D", "Kanwal, Fasiha", "et al." ], "publisher": "Hepatology Communications", "date": "2024-11-04", "url": "https://pubmed.ncbi.nlm.nih.gov/39495136/", "doi": "10.1097/HC9.0000000000000565", "pmid": "39495136", "trialId": "NCT06084234", "patentId": null, "grantId": "U01 CA271887", "accessed": "2026-09-14", "notes": "Protocol for a 5,500-person pragmatic phase 4 randomized comparison of six-month ultrasound with or without AFP against GALAD. The primary endpoint is late-stage HCC at 5.5 years, not mortality. The paper reports several author relationships with biomarker companies." }, { "id": "singh-2023-kras-wildtype-drivers", "type": "paper", "title": "Oncogenic drivers and therapeutic vulnerabilities in KRAS wild-type pancreatic cancer", "authors": [ "Singh, H", "et al." ], "publisher": "Clinical Cancer Research", "date": "2023-11-14", "url": "https://doi.org/10.1158/1078-0432.CCR-22-3930", "doi": "10.1158/1078-0432.CCR-22-3930", "pmid": "37463056", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Single-institution targeted-panel cohort included 795 exocrine pancreatic cancers and found 73, 9.2%, without KRAS mutation. Alternative MAPK-pathway drivers appeared in 32/73, 43.8%, including BRAF changes and receptor fusions. The paper includes model sensitivity for BRAF in-frame deletions and durable benefit in one ROS1-fusion patient, not a prospective matched-treatment comparison." }, { "id": "singh-2026-gallbladder-liver-resection-rct", "type": "paper", "title": "Segment 4b/5 versus wedge resection for gallbladder adenocarcinoma: a randomized controlled trial", "authors": [ "Singh, Shivendra", "Aggarwal, Abhishek", "Goel, Shaifali", "Iqbal, Syed Asif", "Talwar, Vineet" ], "publisher": "Annals of Surgery", "date": "2026-05-01", "url": "https://pubmed.ncbi.nlm.nih.gov/40996213/", "doi": "10.1097/SLA.0000000000006952", "pmid": "40996213", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Journal abstract and PubMed record checked. Single-centre, open-label phase 3 trial in Delhi; 163 patients were included in the final analysis after intraoperative exclusion of metastatic or unresectable disease. Segment 4b/5 resection did not improve R0 resection, disease-free survival or overall survival over wedge resection, and required more operating time and blood loss. Registration identifier was not reported in the PubMed record." }, { "id": "singhal-2021-sbp101-phase1", "type": "paper", "title": "SBP-101 with gemcitabine and nab-paclitaxel as first-line treatment for metastatic pancreatic ductal adenocarcinoma", "authors": [ "Singhal, Nimit", "Sigal, Darren", "Tebbutt, Niall C.", "et al." ], "publisher": "Journal of Clinical Oncology", "date": "2021-05-28", "url": "https://ascopubs.org/doi/10.1200/JCO.2021.39.15_suppl.4127", "doi": "10.1200/JCO.2021.39.15_suppl.4127", "pmid": null, "trialId": "NCT03412799", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Meeting abstract from a 50-person one-arm phase 1a/1b trial. At the recommended schedule, 12 of 28 evaluable patients had partial response and 11 had stable disease at eight weeks; survival was immature. Six serious hepatic and six serious retinal toxicities were attributed to SBP-101, and dosing was changed or stopped. The abstract says progression-free survival was confounded by drug holds." }, { "id": "singhal-2024-classical-kras-resistance", "type": "paper", "title": "A classical epithelial state drives acute resistance to KRAS inhibition in pancreatic cancer", "authors": [ "Singhal, Avish", "Styers, Harrison C.", "Rub, Julia", "et al." ], "publisher": "Cancer Discovery", "date": "2024-11-01", "url": "https://pubmed.ncbi.nlm.nih.gov/38975873/", "doi": "10.1158/2159-8290.CD-24-0740", "pmid": "38975873", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Mouse models and patient-derived xenografts with reporters and lineage tracing. KRAS inhibition enriched a classical state that seeded relapse after treatment withdrawal; state-directed ablation improved model response. No human treatment result." }, { "id": "singhi-2026-pancreaseq-gc", "type": "paper", "title": "PancreaSeq Genomic Classifier (PancreaSeq GC) Improves Pancreatic Cyst Classification and Detection of Advanced Neoplasia: A Multi-institutional Validation Study", "authors": [ "Singhi, Aatur D.", "Wald, Abigail I.", "Smith, Katelyn", "et al." ], "publisher": "Annals of Surgical Oncology", "date": "2025-12-12", "url": "https://pubmed.ncbi.nlm.nih.gov/41388172/", "doi": "10.1245/s10434-025-18848-8", "pmid": "41388172", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Blinded validation in 241 people with diagnostic follow-up; 186 had mucinous cysts and 97 already had advanced neoplasia. Advanced-neoplasia sensitivity was 86.6% and specificity 97.9%. The high event prevalence and diagnostic endpoint do not establish prediction of future progression in routine surveillance. Several authors disclosed assay intellectual property, royalties or related consulting." }, { "id": "sobnach-2024-ssa-hcc-systematic-review", "type": "paper", "title": "The management and outcomes of hepatocellular carcinoma in sub-Saharan Africa: a systematic review", "authors": [ "Sobnach, Sanju", "Kotze, Urda", "Spearman, C. Wendy", "et al." ], "publisher": "HPB", "date": "2024-01-01", "url": "https://pubmed.ncbi.nlm.nih.gov/37805364/", "doi": "10.1016/j.hpb.2023.09.015", "pmid": "37805364", "trialId": "PROSPERO CRD42022363955", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Systematic review of 39 publications from only 15 of 48 sub-Saharan African countries, covering 3,989 patients. Curative treatment was offered to 6%; 84% received best supportive care alone, and few survived one year. This is a synthesis of uneven observational reports rather than a population-wide registry or intervention test." }, { "id": "sobnach-2025-south-africa-hcc-disparities", "type": "paper", "title": "Healthcare Disparities in the Treatment and Outcomes of Hepatocellular Carcinoma in South Africa", "authors": [ "Sobnach, Sanju", "Emmamally, Muhammad", "Venter, Keith", "et al." ], "publisher": "World Journal of Surgery", "date": "2025-05-01", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC12058446/", "doi": "10.1002/wjs.12559", "pmid": "40101972", "trialId": null, "patentId": null, "grantId": "Harry Crossley Clinical Fellowship", "accessed": "2026-09-14", "notes": "Retrospective comparison using prospectively maintained databases: 551 consecutive public-sector and 51 private-sector HCC patients in Cape Town from 2001 to 2024. Public-sector patients had more advanced disease, 69.7% versus 15.7% received best supportive care alone, and median survival was 68 versus 703 days. Ablation and resection survival did not differ significantly within the highly selected treated groups. The long period, small private cohort, major baseline differences and treatment selection prevent a causal treatment or insurance effect. Funded by a Harry Crossley Clinical Fellowship; no funder role was reported." }, { "id": "sobnach-2026-global-hcc-disparities", "type": "paper", "title": "Global disparities in hepatocellular carcinoma outcomes: a multicentre study", "authors": [ "Sobnach, Sanju", "Bayadsi, Haytham", "Kotze, Urda", "et al." ], "publisher": "British Journal of Surgery", "date": "2026-08-03", "url": "https://pubmed.ncbi.nlm.nih.gov/42330118/", "doi": "10.1093/bjs/znag079", "pmid": "42330118", "trialId": null, "patentId": null, "grantId": "Harry Crossley Clinical Fellowship; Harry Crossley Foundation at the University of Cape Town", "accessed": "2026-09-14", "notes": "Comparative referral-centre cohort of 455 patients in South Africa and 504 in Sweden, treated from 2012 to 2023. The South African cohort was younger, had far more HBV-related HCC, received curative-intent therapy in 9.2% versus 42.5%, and had one-, three- and five-year overall survival of 17.8%, 7.7% and 5.3% versus 59.3%, 35.3% and 27.0%. Survival was also lower within treatment groups. Large biological, stage, liver-function and health-system differences prevent attribution to one cause." }, { "id": "sohal-2020-thu-decitabine-pdac", "type": "paper", "title": "Tetrahydrouridine combined with decitabine to target DNMT1 in advanced chemorefractory pancreatic cancer", "authors": [ "Sohal, Davendra", "Krishnamurthi, Smitha", "Tohme, Rita", "et al." ], "publisher": "American Journal of Cancer Research", "date": "2020-09-01", "url": "https://pubmed.ncbi.nlm.nih.gov/33042633/", "doi": null, "pmid": "33042633", "trialId": "NCT02847000", "patentId": null, "grantId": "P30 CA043703", "accessed": "2026-09-15", "notes": "Primary single-arm pilot, n=13. Five progressed clinically before week eight; among eight scanned at week eight, one had stable disease and seven progressed. Median treatment time was 35 days and median survival 3.1 months. Plasma CDA activity was more than ten-fold higher in metastatic than resectable PDAC, showing the proposed exposure fix remained inadequate." }, { "id": "song-2014-cryo-bypass-comparison", "type": "paper", "title": "The outcome of cryoablation in treating advanced pancreatic cancer: a comparison with palliative bypass surgery alone", "authors": [ "Song, Zhen Guo", "Hao, Ji Hui", "Gao, Song", "et al." ], "publisher": "Journal of Digestive Diseases", "date": "2014-06-24", "url": "https://pubmed.ncbi.nlm.nih.gov/24958092/", "doi": "10.1111/1751-2980.12170", "pmid": "24958092", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Retrospective records of 118 advanced cases receiving bypass surgery with or without cryoablation. Tumour size and CA19-9 fell only with cryoablation, but postoperative complications and prognosis did not differ significantly. Baseline jaundice and back pain differed between groups." }, { "id": "song-2021-ctc-resectable-pdac", "type": "paper", "title": "Detection of Circulating Tumor Cells in Resectable Pancreatic Ductal Adenocarcinoma: A Prospective Evaluation as a Prognostic Marker", "authors": [ "Song, Bum-Soo", "Kwon, Won", "Kim, Hyung Sun", "et al." ], "publisher": "Frontiers in Oncology", "date": "2021", "url": "https://pubmed.ncbi.nlm.nih.gov/33680936/", "doi": null, "pmid": "33680936", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Prospective Korean surgical cohort n=32, comparing portal and peripheral blood CTC measurements; mainly stage II/III." }, { "id": "sonire-2026-nedo-award", "type": "company-report", "title": "Sonire Therapeutics reports $13 million NEDO deep-tech award", "authors": [ "SONIRE Therapeutics Inc." ], "publisher": "SONIRE Therapeutics Inc.", "date": "2026-06-18", "url": "https://www.sonire-therapeutics.com/en/category/grand/", "doi": null, "pmid": null, "trialId": "NCT05601323; NCT07033689", "patentId": null, "grantId": "NEDO-DTSU-2026", "accessed": "2026-09-15", "notes": "Company announcement of about $13 million non-dilutive Japanese government support for clinical development, regulation, manufacturing and commercialization. It says enrollment in randomized SUNRISE-I is complete and SUNRISE-II has begun in the United States. The amount is in addition to the reported $18 million financing." }, { "id": "sonire-2026-series-a", "type": "company-report", "title": "Sonire Therapeutics closes $18 million financing for pancreatic focused ultrasound", "authors": [ "SONIRE Therapeutics Inc." ], "publisher": "SONIRE Therapeutics Inc.", "date": "2026-04-16", "url": "https://www.sonire-therapeutics.com/en/series_a_finance/", "doi": null, "pmid": null, "trialId": "NCT05601323; NCT07033689", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Company announcement of $18 million financing led by Sante Ventures, with several Japanese investors. Sonire says proceeds support the randomized SUNRISE-I trial, U.S. development and regulatory work. It describes real-time imaging, cavitation visualization, robot positioning and outpatient treatment. These are company claims; financing does not validate benefit." }, { "id": "sonire-2026-sunrise2-first-patient", "type": "company-report", "title": "Sonire Therapeutics initiates first U.S. Suizenji study in pancreatic cancer", "authors": [ "SONIRE Therapeutics Inc." ], "publisher": "SONIRE Therapeutics Inc.", "date": "2026-05-01", "url": "https://www.sonire-therapeutics.com/en/sunrise2_fpi/", "doi": null, "pmid": null, "trialId": "NCT07033689", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Company announcement says the first U.S. participant was treated at Stanford in March 2026 and the feasibility trial plans about ten people. It presents the system as real-time guided, without general anesthesia and about 20 minutes. These operational descriptions require independent confirmation from study results." }, { "id": "sousa-2016-stellate-alanine", "type": "paper", "title": "Pancreatic stellate cells support tumour metabolism through autophagic alanine secretion", "authors": [ "Sousa, Cristovão M.", "Biancur, Douglas E.", "Wang, Xiaoxu", "et al." ], "publisher": "Nature", "date": "2016-08-10", "url": "https://pubmed.ncbi.nlm.nih.gov/27509858/", "doi": "10.1038/nature19084", "pmid": "27509858", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Cell, coculture and animal work describing an alanine exchange between stellate cells and PDAC. It identifies a route, not its quantitative share of nutrient use in patients or a validated target." }, { "id": "stark-2025-boltzgen", "type": "preprint", "title": "BoltzGen: Toward universal binder design", "authors": [ "Stark, Hannes", "Faltings, Felix", "Choi, MinGyu", "et al." ], "publisher": "bioRxiv", "date": "2025-11-24", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC12697729/", "doi": "10.1101/2025.11.20.689494", "pmid": null, "trialId": null, "patentId": null, "grantId": "OT2CA297463; 1918839", "accessed": "2026-09-14", "notes": "Primary open-weight preprint with wet-lab campaigns across 26 targets. It reports nanomolar binders for several held-out targets but also target-dependent diversity collapse and benchmark leakage risk. No oncology efficacy evidence." }, { "id": "steindorf-2019-pancreatic-exercise-rct", "type": "paper", "title": "Quality of life, fatigue and sleep in pancreatic cancer: randomized exercise trial", "authors": [ "Steindorf, Karen", "Clauss, Dorothea", "Tjaden, Christine", "et al." ], "publisher": "Deutsches Arzteblatt International", "date": "2019-07-08", "url": "https://doi.org/10.3238/arztebl.2019.0471", "doi": "10.3238/arztebl.2019.0471", "pmid": "31431236", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Randomized 65 pancreatic-cancer patients across home resistance training, supervised training and usual care; 47 completed six months. No exercise effect was found at six months. Exploratory three-month analyses favoured pooled training for physical function and global quality of life, both p=0.016, but attrition and timepoint selection prevent a durable effect claim." }, { "id": "stepien-2024-prediagnostic-metabolomics", "type": "paper", "title": "Metabolomics for early pancreatic cancer detection in plasma samples from a Swedish prospective population-based biobank", "authors": [ "Stepien and collaborators" ], "publisher": "Journal of Gastrointestinal Oncology", "date": "2024", "url": "https://pubmed.ncbi.nlm.nih.gov/38756646/", "doi": null, "pmid": "38756646", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Nested case-control study within a Swedish population biobank: 82 future pancreatic cancers and 82 matched controls; internal LASSO score." }, { "id": "stoop-2026-global-staging-survey", "type": "paper", "title": "Staging laparoscopy to detect occult metastases in localized pancreatic cancer: global survey among nine international societies", "authors": [ "Stoop, Thomas F", "Lutchman, Kishan R D", "Theijse, Rutger T", "et al." ], "publisher": "HPB", "date": "2026-03-01", "url": "https://pubmed.ncbi.nlm.nih.gov/41421934/", "doi": "10.1016/j.hpb.2025.12.001", "pmid": "41421934", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Survey of 617 surgeons from 76 countries and nine international groups. Eighty-two percent used staging laparoscopy, ranging from 59% in Africa to 90% in the Americas. Most used it selectively and in the same session as intended surgery; timing and selection varied widely. This measures reported practice, not actual delivery, accuracy or patient benefit." }, { "id": "stoop-2026-preopanc4-surgery", "type": "paper", "title": "Surgical outcomes from nationwide implementation of the international best practice for locally advanced pancreatic cancer (PREOPANC-4) study", "authors": [ "Stoop, Thomas F.", "Seelen, Leonard W. F.", "van 't Land, Freek R.", "et al." ], "publisher": "British Journal of Surgery", "date": "2026-08-27", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC13541110/", "doi": "10.1093/bjs/znag075", "pmid": "42329186", "trialId": "NCT05524090", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Prospective Dutch nationwide surgical-safety analysis. After at least four months of multiagent chemotherapy without scan progression, 180 selected people were explored and 155, 86%, were resected. Major complications requiring an invasive intervention or intensive care affected 44%; in-hospital or 30-day mortality was 0.6% and 90-day mortality was 3%. Extended resection occurred in 77%, and three specialist centres performed 74% of resections. Only 37% met narrower NCCN locally advanced criteria. There was no continued-treatment control and the paper did not estimate whether surgery caused longer survival." }, { "id": "storandt-2024-atezo-bev-liver-function", "type": "journal-article", "title": "Atezolizumab plus bevacizumab as first-line systemic therapy for hepatocellular carcinoma: a multi-institutional cohort study", "authors": [ "Storandt, Matthew H.", "Zemla, Trevor J.", "Patell, Kushal", "et al." ], "publisher": "The Oncologist", "date": "2024-11-04", "url": "https://pubmed.ncbi.nlm.nih.gov/38979643/", "doi": "10.1093/oncolo/oyae142", "pmid": "38979643", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Retrospective first-line cohort of 322 HCC patients across US institutions: 226 Child-Pugh A, 86 B and 10 C. Median OS was 21.6 months for A, 9.1 for B7 and 4.7 for B8-C12. Among Child-Pugh A, ALBI grade 1 versus 2 median OS was 34.9 versus 14.2 months. Treatment benefit cannot be inferred from between-group prognosis; most patients were White and treated at high-volume centres." }, { "id": "strickler-2022-sotorasib-pdac", "type": "paper", "title": "Sotorasib in KRAS p.G12C-Mutated Advanced Pancreatic Cancer", "authors": [ "Strickler, John H.", "Satake, Hironaga", "George, Thomas J.", "et al." ], "publisher": "New England Journal of Medicine", "date": "2022-12-21", "url": "https://pubmed.ncbi.nlm.nih.gov/36546651/", "doi": "10.1056/NEJMoa2208470", "pmid": "36546651", "trialId": "NCT03600883", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Single-group pooled phase 1 and 2 analysis in 38 previously treated patients. Funded by Amgen; no randomised control. KRAS G12C occurs in about 1% to 2% of pancreatic cancers." }, { "id": "stupakov-2026-schwann-mechanics", "type": "paper", "title": "Pancreatic cancer fibrosis activates protumorigenic Schwann cells through a nuclear mechanosensing mechanism", "authors": [ "Stupakov, Pavel", "Sadatrezaei, Golbahar", "Quesada, Ines Velazquez", "et al." ], "publisher": "Gut", "date": "2026-07-01", "url": "https://pubmed.ncbi.nlm.nih.gov/42336627/", "doi": "10.1136/gutjnl-2025-337316", "pmid": "42336627", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Human-tissue correlations plus mouse, live-imaging and force-application experiments; not a human intervention study." }, { "id": "subbiah-2023-rly4008", "type": "paper", "title": "RLY-4008, the First Highly Selective FGFR2 Inhibitor with Activity across FGFR2 Alterations and Resistance Mutations", "authors": [ "Subbiah, Vivek", "Sahai, Vaibhav", "Maglic, Dejan", "Goyal, Lipika", "Schram, Alison M", "Casaletto, Jessica B" ], "publisher": "Cancer Discovery", "date": "2023-09-06", "url": "https://pubmed.ncbi.nlm.nih.gov/37270847/", "doi": "10.1158/2159-8290.CD-23-0475", "pmid": "37270847", "trialId": "NCT04526106", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Relay-led primary discovery paper joining kinase selectivity, cell assays, xenografts and early clinical cases. It supports mechanism and initial activity, not a controlled survival effect. Most listed scientists were Relay employees and the compound is patented." }, { "id": "sun-2026-talentop-hcc", "type": "paper", "title": "Liver resection after atezolizumab and bevacizumab versus maintenance therapy for locally advanced hepatocellular carcinoma (TALENTOP): a multicentre, open-label, randomised, phase 3 trial", "authors": [ "Sun, Hui-Chuan", "Zhu, Xiao-Dong", "Shen, Feng", "et al." ], "publisher": "The Lancet", "date": "2026-08-22", "url": "https://pubmed.ncbi.nlm.nih.gov/42624156/", "doi": "10.1016/S0140-6736(26)01252-3", "pmid": "42624156", "trialId": "NCT04649489", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Multicentre open-label phase 3 trial in China. Of 489 people given induction atezolizumab plus bevacizumab for HCC with macrovascular invasion and no extrahepatic spread, 201 responders or people with stable disease who were judged resectable were randomized. Resection followed by the same drugs increased median time to treatment failure from 11.8 to 20.4 months, HR 0.60, 95% CI 0.39-0.91. Grade 3-4 treatment-related events were 39% versus 21%; two treatment-related deaths occurred in the surgery arm. Overall survival was not the primary reported result. Roche and China's Ministry of Science and Technology funded the trial." }, { "id": "surana-2025-erk-hcq-pdac", "type": "paper", "title": "ERK inhibition alone and with autophagy inhibition in metastatic pancreatic cancer", "authors": [ "Surana, Rohan", "Morgado, Miguel", "Somasundaram, Ashwin", "et al." ], "publisher": "JCO Precision Oncology", "date": "2025-09-02", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC13379901/", "doi": "10.1200/PO-25-00332", "pmid": "40893054", "trialId": "NCT04386057", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "After a 13-person safety run-in, 39 treated patients were randomized to the ERK inhibitor LY3214996 with hydroxychloroquine, n=20, or LY3214996 alone, n=19. Four-month disease control was 1/20 versus 1/19; median progression-free survival 1.3 versus 1.9 months; median overall survival 2.4 versus 4.6 months. Eleven patient-derived organoid lines showed no clear combination synergy. Blood-cell LC3B and NBR1 did not show consistent autophagy inhibition." }, { "id": "szpakowski-2020-gallbladder-polyps", "type": "paper", "title": "Outcomes of gallbladder polyps and their association with gallbladder cancer in a 20-year cohort", "authors": [ "Szpakowski, Jean-Luc", "Tucker, Lue-Yen" ], "publisher": "JAMA Network Open", "date": "2020-05-18", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC7235691/", "doi": "10.1001/jamanetworkopen.2020.5143", "pmid": "32421183", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Full open-access report checked. Kaiser Permanente cohort of 622,227 adults undergoing abdominal ultrasound, including 35,856 with a quantitative gallbladder polyp. Nineteen people in the polyp cohort developed gallbladder cancer. Cancer rates were low and similar with and without a polyp, while risk rose with initial polyp size. Imaging-report extraction, surveillance selection and censoring at cholecystectomy limit causal interpretation." }, { "id": "tabernero-2021-beacon-crc", "type": "paper", "title": "Encorafenib plus cetuximab for previously treated BRAF V600E-mutant metastatic colorectal cancer: updated BEACON results", "authors": [ "Tabernero, Josep", "et al." ], "publisher": "Journal of Clinical Oncology", "date": "2021-02-01", "url": "https://pubmed.ncbi.nlm.nih.gov/33503393/", "doi": "10.1200/JCO.20.02088", "pmid": "33503393", "trialId": "NCT02928224", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Open-label phase 3 in 665 previously treated BRAF V600E metastatic colorectal cancer patients. Encorafenib plus cetuximab median OS was 9.3 versus 5.9 months for control, HR 0.61, 95% CI 0.48-0.77; response 19.5% versus 1.8%. Adding binimetinib did not extend median OS beyond the doublet and increased severe toxicity. Industry-funded molecular-subgroup evidence." }, { "id": "tai-2025-pan-asian-esmo-pancreatic", "type": "guideline", "title": "Pan-Asian adapted ESMO Clinical Practice Guidelines for the diagnosis, treatment and follow-up of patients with pancreatic cancer", "authors": [ "Tai, Darren", "Conroy, Thierry", "Lee, Jin J X", "et al." ], "publisher": "ESMO Open", "date": "2025-10-09", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC12546840/", "doi": "10.1016/j.esmoop.2025.105826", "pmid": "41067164", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Consensus adaptation that suggests regular follow-up after resection at level III, grade B after reviewing the 333-patient observational imaging study. It does not convert that association into randomized utility evidence. Applicability, access and follow-up practice vary across Asian systems." }, { "id": "takeda-2025-ibi343-license", "type": "company-report", "title": "Takeda closes strategic partnership with Innovent for IBI343", "authors": [ "Takeda Pharmaceutical Company Limited" ], "publisher": "Takeda", "date": "2025-12-04", "url": "https://www.takeda.com/newsroom/newsreleases/2025/closing-partnership-innovent/", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Primary rights source. Takeda acquired exclusive IBI343 development, manufacturing and commercial rights outside Greater China; Innovent retains Greater China. The arrangement supplies a solvent global development and manufacturing partner." }, { "id": "tamura-2021-fcsems-plastic-rct", "type": "paper", "title": "Covered self-expandable metal stents versus plastic stents for preoperative biliary drainage during neoadjuvant chemotherapy", "authors": [ "Tamura, Takashi", "Itonaga, Masahiro", "Ashida, Reiko", "et al." ], "publisher": "Digestive Endoscopy", "date": "2021-11-01", "url": "https://pubmed.ncbi.nlm.nih.gov/33410564/", "doi": "10.1111/den.13926", "pmid": "33410564", "trialId": "UMIN000030473", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Single-centre randomized trial in 22 borderline-resectable pancreatic-cancer patients receiving gemcitabine plus nab-paclitaxel. Fully covered metal stents reduced dysfunction before surgery or progression from 72.8% to 18.2%, P=.015, and reinterventions from a mean 1.27 to .27, P=.001. Procedure, surgical adverse events and stated medical costs did not differ. The very small trial estimates stent function, not chemotherapy completion or survival." }, { "id": "tan-2024-psc-mrcp-surveillance", "type": "paper", "title": "Surveillance MRI is associated with improved survival in patients with primary sclerosing cholangitis", "authors": [ "Tan, Natassia", "et al." ], "publisher": "Hepatology Communications", "date": "2024-05-01", "url": "https://pubmed.ncbi.nlm.nih.gov/38696372/", "doi": "10.1097/HC9.0000000000000442", "pmid": "38696372", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Retrospective nine-centre Australian cohort of 298 people with primary sclerosing cholangitis and 2,117 person-years. In weighted analysis, regular MRCP surveillance was associated with lower all-cause mortality, HR 0.29, 95% CI 0.14-0.59, but survival after hepatobiliary cancer diagnosis did not differ, p=0.74. Residual selection and care-intensity confounding remain possible." }, { "id": "tan-2025-nk-gem-s1", "type": "paper", "title": "Dynamic transcriptional immune landscape in response to NK-cell therapy combined with gemcitabine plus S-1 in advanced pancreatic cancer", "authors": [ "Tan, Qin", "Li, Yifei", "Liu, Caixia", "et al." ], "publisher": "Signal Transduction and Targeted Therapy", "date": "2025-11-21", "url": "https://pubmed.ncbi.nlm.nih.gov/41266303/", "doi": "10.1038/s41392-025-02488-1", "pmid": "41266303", "trialId": "ChiCTR1900021764", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Primary single-arm Chinese phase 1b/2: 24 treated, 19 efficacy-evaluable, ORR 31.6%, median PFS 6.6 months and OS 10.8 months. No chemotherapy-only control." }, { "id": "tan-2026-neoadjuvant-ipd-meta", "type": "paper", "title": "Neoadjuvant therapy versus upfront surgery for resectable pancreatic cancer: updated systematic review, individual-patient-data meta-analysis and trial sequential analysis", "authors": [ "Tan, Hwee Leong", "Zhao, Yun", "Chua, Darren Weiquan", "et al." ], "publisher": "Surgery", "date": "2026-05-01", "url": "https://pubmed.ncbi.nlm.nih.gov/41381268/", "doi": "10.1016/j.surg.2025.109872", "pmid": "41381268", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Updated review of nine randomized trials, 604 neoadjuvant and 527 upfront-surgery patients. Event-free survival improved, HR .77, 95% CI .65-.90; overall survival did not reach significance, HR .85, 95% CI .68-1.05. Reconstructed median OS was 29.6 versus 23.7 months and event-free survival 13.6 versus 11.4. Noncurative exploration fell, RR .90, and node-negative histology rose, RR 1.73. Trial-sequential analysis judged event-free survival conclusive but OS and R0 evidence still insufficient. Regimens, staging and trial quality differed, and reconstructed curves are not original patient data." }, { "id": "tanaka-2012-microscopic-local-recurrence", "type": "paper", "title": "An autopsy study to clarify characteristics of local recurrence after extended pancreatectomy with intraoperative radiation therapy", "authors": [ "Tanaka, H.", "et al." ], "publisher": "Langenbeck's Archives of Surgery", "date": "2012-08-01", "url": "https://pubmed.ncbi.nlm.nih.gov/22382704/", "doi": "10.1007/s00423-012-0934-8", "pmid": "22382704", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Retrospective autopsy subset from 41 people who had extended resection and intraoperative radiation; 14 had autopsy. Five of 14, 36%, had microscopic local recurrence despite no relapse on follow-up imaging or gross autopsy. Small cancer-cell groups were scattered through dense connective tissue. The selected end-stage sample proves an imaging blind spot, not dormancy, viability during an earlier disease-free interval or treatment action." }, { "id": "tanaka-2020-panec-g3", "type": "paper", "title": "Pancreatic neuroendocrine carcinoma G3 may be heterogeneous and could be classified into two distinct groups", "authors": [ "Tanaka, Hiroki", "Hijioka, Susumu", "Hosoda, Waki", "et al." ], "publisher": "Pancreatology", "date": "2020-08-08", "url": "https://pubmed.ncbi.nlm.nih.gov/32891532/", "doi": "10.1016/j.pan.2020.07.400", "pmid": "32891532", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Multicentre retrospective subgroup of 49 Japanese patients reclassified as pancreatic neuroendocrine carcinoma G3. RB loss and KRAS mutation occurred in 54.5% and 48.7%. RB loss and/or KRAS mutation was associated with platinum response of 70.0% versus 33.3%, odds ratio 9.22, 95% CI 1.26-67.3. Median overall survival was 239 versus 473 days, HR 2.11, 95% CI 0.92-4.86; not statistically significant. Small retrospective selector evidence, not a treatment rule." }, { "id": "tanaka-2026-ai-eus-reader", "type": "paper", "title": "Development and crossover evaluation of an artificial intelligence-assisted system for solid pancreatic lesion detection and pancreatic parenchyma recognition in endoscopic ultrasonography", "authors": [ "Fujisawa, Takafumi", "Kuwahara, Taiki", "Sato, Tomoki", "et al." ], "publisher": "Digestive Endoscopy", "date": "2026-09-01", "url": "https://pubmed.ncbi.nlm.nih.gov/42720264/", "doi": "10.1111/den.70282", "pmid": "42720264", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Six-centre development followed by a randomized two-period crossover reader study. Eight endosonographers interpreted selected image sets from 118 patients. Among novices, AI assistance raised solid-lesion sensitivity from 76.8% to 88.7% and accuracy from 78.7% to 86.4%, with specificity 84.2% versus 80.5%. This tests readers on image sets, not prospective procedures, management, cancer yield, complications or patient outcomes." }, { "id": "tango-2025-vopimetostat-monotherapy", "type": "company-report", "title": "Vopimetostat phase 1/2 monotherapy data in MTAP-deleted cancers", "authors": [ "Tango Therapeutics, Inc." ], "publisher": "Tango Therapeutics", "date": "2025-10-23", "url": "https://ir.tangotx.com/news-releases/news-release-details/tango-therapeutics-reports-positive-data-ongoing-phase-12-study", "doi": null, "pmid": null, "trialId": "NCT05732831", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Sponsor interim report. Sixty-four pancreatic patients were enrolled; 39 at active doses had more than six months since first dose and were included regardless of outcome. In that 39-person set, all-line ORR was 15%, disease control 71%, second-line ORR 25%, second-line median PFS 7.2 months and third-line-or-later PFS 4.1 months. No randomized comparator." }, { "id": "tango-2026-q2-10q", "type": "company-filing", "title": "Tango Therapeutics quarterly report for the period ended 30 June 2026", "authors": [ "Tango Therapeutics, Inc." ], "publisher": "U.S. Securities and Exchange Commission", "date": "2026-08-06", "url": "https://www.sec.gov/Archives/edgar/data/1819133/000119312526343424/tngx-20260630.htm", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Primary current finance source. Cash, equivalents and marketable securities were $1.0bn at 30 June 2026; six-month operating cash use was $83.226m and vopimetostat direct programme expense was $21.677m. June financing supplied $651.4m net. The company plans a randomized front-line combination trial after regulatory discussion." }, { "id": "tango-2026-vopimetostat-ras-data", "type": "company-filing", "title": "Initial vopimetostat and RAS inhibitor combination data", "authors": [ "Tango Therapeutics, Inc." ], "publisher": "U.S. Securities and Exchange Commission", "date": "2026-06-08", "url": "https://www.sec.gov/Archives/edgar/data/1819133/000119312526260679/d158334d8k.htm", "doi": null, "pmid": null, "trialId": "NCT06922591", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Sponsor filing with May 28 cutoff: 20 PDAC patients received vopimetostat plus daraxonrasib and 34 received vopimetostat plus zoldonrasib. At least 14 weeks after first dose, 11 of 12 daraxonrasib-combination patients responded, nine confirmed; 14 of 27 zoldonrasib-combination patients responded, ten confirmed. The 39 evaluable PDAC patients come from different drugs and doses, with no concurrent single-agent or standard-care control." }, { "id": "tcga-2017-hcc", "type": "paper", "title": "Comprehensive and Integrative Genomic Characterization of Hepatocellular Carcinoma", "authors": [ "Cancer Genome Atlas Research Network" ], "publisher": "Cell", "date": "2017-06-15", "url": "https://pubmed.ncbi.nlm.nih.gov/28622513/", "doi": "10.1016/j.cell.2017.05.046", "pmid": "28622513", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Multi-platform analysis: mutations and copy number in 363 patients; methylation, RNA, microRNA and protein in 196. Identified three integrated molecular groups, with one associated with poorer prognosis in three cohorts." }, { "id": "tcga-2017-pdac", "type": "paper", "title": "Integrated Genomic Characterization of Pancreatic Ductal Adenocarcinoma", "authors": [ "Cancer Genome Atlas Research Network" ], "publisher": "Cancer Cell", "date": "2017-08-14", "url": "https://pubmed.ncbi.nlm.nih.gov/28810144/", "doi": "10.1016/j.ccell.2017.07.007", "pmid": "28810144", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Integrated genomic, transcriptomic, epigenetic and protein analysis of 150 clinically annotated pancreatic cancers, with explicit attention to low tumour purity." }, { "id": "testoni-2021-hybridtherm-rct", "type": "paper", "title": "Efficacy of Endoscopic Ultrasound-Guided Ablation with the HybridTherm Probe in Locally Advanced or Borderline Resectable Pancreatic Cancer: A Phase II Randomized Controlled Trial", "authors": [ "Testoni, Sabrina Gloria Giulia", "Petrone, Maria Chiara", "Reni, Michele", "et al." ], "publisher": "Cancers", "date": "2021-09-07", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC8464946/", "doi": "10.3390/cancers13184512", "pmid": "34572743", "trialId": "NCT02336672", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Single-centre phase 2 randomized report. Forty people were assigned, but three intervention assignments were excluded after metastases were found. Six-month progression-free rates were 41.2% versus 30%, p=0.48; median survival was 13 versus 17 months, HR 1.18, p=0.79. The manufacturer held the allocation list, monitored data and withdrew the probe after slow accrual." }, { "id": "theijse-2024-nontherapeutic-laparotomy", "type": "paper", "title": "Impact of a non-therapeutic laparotomy in patients with locally advanced pancreatic cancer treated with induction (m)FOLFIRINOX: Trans-Atlantic Pancreatic Surgery Consortium study", "authors": [ "Theijse, Rutger T", "Stoop, Thomas F", "Janssen, Quisette P", "et al." ], "publisher": "British Journal of Surgery", "date": "2024-03-02", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC10921832/", "doi": "10.1093/bjs/znae033", "pmid": "38456678", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Retrospective five-centre TAPS study of 663 selected people with locally advanced PDAC after at least one cycle of modified FOLFIRINOX. Of 238 surgical explorations, 67, 28.2%, did not produce a resection; 30/67, 44.8%, were stopped for hidden metastases. Two of 67 died within 90 days, 3.0%. Overall survival and later palliative therapy were similar to selected non-operated patients, but confounding is severe. It measures the residual failure after treatment and expert selection, not the causal benefit of staging laparoscopy." }, { "id": "theriva-2023-virage-protocol", "type": "trial-protocol", "title": "P-VCNA-003 VIRAGE protocol amendment 2.1", "authors": [ "Theriva Biologics, S.L.", "Simbec-Orion" ], "publisher": "ClinicalTrials.gov document archive", "date": "2023-08-10", "url": "https://cdn.clinicaltrials.gov/large-docs/11/NCT05673811/Prot_000.pdf", "doi": null, "pmid": null, "trialId": "NCT05673811", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Public protocol planned 92 randomized participants and 57 deaths. It defines its ITT set as randomized people who received study medication and its FAS as people receiving at least one third of planned chemotherapy plus, in the virus arm, a complete first VCN-01 dose. It says one-sided alpha 0.05 for sample size, a two-sided log-rank primary test, and elsewhere a 5% two-sided ordered-testing threshold, leaving an internal alpha and analysis-set conflict that the dated final analysis plan must resolve." }, { "id": "theriva-2025-virage-topline", "type": "company-filing", "title": "Theriva announces primary endpoints for VIRAGE phase 2b", "authors": [ "Theriva Biologics, Inc." ], "publisher": "U.S. Securities and Exchange Commission", "date": "2025-05-07", "url": "https://www.sec.gov/Archives/edgar/data/894158/000110465925045309/tm2514050d1_ex99-1.htm", "doi": null, "pmid": null, "trialId": "NCT05673811", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Sponsor topline report for a 96-person full analysis set, 48 per arm. Median OS was 10.8 versus 8.6 months, HR 0.57, 95% CI 0.34-0.96, two-sided log-rank P=0.0546; median PFS was 7.0 versus 4.6 months, HR 0.55, P=0.0105. The release says the design used one-sided alpha 0.05, equivalent to two-sided 0.10. A two-dose, four-cycle subgroup is selected after randomization and cannot establish the effect of the second dose." }, { "id": "theriva-2026-q2-10q", "type": "company-filing", "title": "Theriva Biologics quarterly report for the period ended 30 June 2026", "authors": [ "Theriva Biologics, Inc." ], "publisher": "U.S. Securities and Exchange Commission", "date": "2026-08-12", "url": "https://www.sec.gov/Archives/edgar/data/894158/000110465926093786/tmb-20260630x10q.htm", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Primary current company and finance source. Cash was $11.337m on 30 June and $9.6m in early August; six-month operating cash use was $5.541m. The filing says cash reaches Q1 2027, does not fund further VCN-01 trials beyond VIRAGE2, and the company needs new capital or a partner for phase 3. It reports FDA agreement in principle with a double-blind placebo-controlled adaptive phase 3 design." }, { "id": "third-rock-2015-revolution-medicines-launch", "type": "company-announcement", "title": "Third Rock Ventures launches Revolution Medicines with $45 million Series A", "authors": [ "Third Rock Ventures", "Revolution Medicines" ], "publisher": "Revolution Medicines", "date": "2015-02-04", "url": "https://ir.revmed.com/news-releases/news-release-details/third-rock-ventures-launches-revolution-medicines-45-million", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Joint origin and financing announcement. Third Rock formed Revolution Medicines with a $45m Series A; Mark Goldsmith was founding CEO, Martin Burke scientific founder and advisory-board chair, and the company licensed University of Illinois chemistry. This 2015 event is outside the atlas's 2015-2024 public-funding window and is not added to current venture-event totals." }, { "id": "tie-2022-dynamic-colon", "type": "paper", "title": "Circulating Tumor DNA Analysis Guiding Adjuvant Therapy in Stage II Colon Cancer", "authors": [ "Tie, Jeanne", "Cohen, Joshua D.", "Lahouel, Kamel", "et al." ], "publisher": "New England Journal of Medicine", "date": "2022-06-04", "url": "https://pubmed.ncbi.nlm.nih.gov/35657320/", "doi": "10.1056/NEJMoa2200075", "pmid": "35657320", "trialId": "ACTRN12615000381583", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Australian randomised noninferiority trial in 455 stage II colon-cancer patients. It tested a treatment-decision strategy and chemotherapy use, not just ctDNA prognosis. Some assay inventors and company interests were disclosed." }, { "id": "tie-2025-dynamic3", "type": "paper", "title": "Circulating tumor DNA-guided adjuvant therapy in locally advanced colon cancer: the randomized phase 2/3 DYNAMIC-III trial", "authors": [ "Tie, Jeanne", "et al." ], "publisher": "Nature Medicine", "date": "2025-10-16", "url": "https://pubmed.ncbi.nlm.nih.gov/41115959/", "doi": "10.1038/s41591-025-04030-w", "pmid": "41115959", "trialId": "ACTRN12617001566325", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Randomized phase 2/3 stage III colon-cancer trial with 968 evaluable patients. The ctDNA-negative de-escalation rule reduced oxaliplatin and hospitalizations but failed its recurrence-free-survival noninferiority test. CtDNA-positive chemotherapy escalation did not improve two-year recurrence-free survival." }, { "id": "timmer-2024-crossfire", "type": "paper", "title": "MRI-guided stereotactic ablative body radiotherapy versus CT-guided percutaneous irreversible electroporation for locally advanced pancreatic cancer (CROSSFIRE)", "authors": [ "Timmer, Florentine E. F.", "Geboers, Bart", "Ruarus, Alette H.", "et al." ], "publisher": "The Lancet Gastroenterology & Hepatology", "date": "2024-05-01", "url": "https://pubmed.ncbi.nlm.nih.gov/38513683/", "doi": "10.1016/S2468-1253(24)00017-7", "pmid": "38513683", "trialId": "NCT02791503", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Single-centre, open-label randomized phase 2 trial of 68 patients after FOLFIRINOX. Stopped at interim analysis for futility. Funded by Adessium Foundation and AngioDynamics." }, { "id": "tiriac-2018-organoid-profiling", "type": "paper", "title": "Organoid profiling identifies common responders to chemotherapy in pancreatic cancer", "authors": [ "Tiriac, Hervé", "Belleau, Pascal", "Engle, Dannielle D.", "Plenker, Dennis", "Deschênes, Aline", "Somerville, Timothy D. D.", "Tuveson, David A." ], "publisher": "Cancer Discovery", "date": "2018-09-01", "url": "https://aacrjournals.org/cancerdiscovery/article/8/9/1112/6245/Organoid-Profiling-Identifies-Common-Responders-to", "doi": "10.1158/2159-8290.CD-18-0349", "pmid": "29853643", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Primary organoid platform paper. Molecular and drug-response profiles reproduced major PDAC features and paralleled patient outcomes in case studies. It proposed prospective treatment selection but did not test organoid-guided care in a randomized trial." }, { "id": "topal-2025-mwa-dual-ici", "type": "paper", "title": "Progression-free survival for unresectable non-metastatic locally advanced pancreatic cancer after surgical microwave ablation plus durvalumab and tremelimumab: phase-2 non-randomized prospective clinical trial", "authors": [ "Topal, Baki", "et al." ], "publisher": "Communications Medicine", "date": "2025-11-13", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC12627779/", "doi": "10.1038/s43856-025-01186-x", "pmid": "41254229", "trialId": "NCT04156087", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Single-arm trial stopped at 12 of 20 planned participants for slow recruitment and efficacy concerns. Four had metastases found at laparoscopy and only eight received the full microwave, durvalumab, tremelimumab and gemcitabine package. One died from hemorrhagic shock after pancreatic fistula following direct biopsy and ablation; seven of 12 had a treatment-related event." }, { "id": "trinchet-2011-hcc-surveillance-interval", "type": "paper", "title": "Ultrasonographic surveillance of hepatocellular carcinoma in cirrhosis: a randomized trial comparing 3- and 6-month periodicities", "authors": [ "Trinchet, Jean-Claude", "Chaffaut, Cendrine", "Bourcier, Valerie", "et al." ], "publisher": "Hepatology", "date": "2011-12-01", "url": "https://pubmed.ncbi.nlm.nih.gov/22144108/", "doi": "10.1002/hep.24545", "pmid": "22144108", "trialId": "NCT00190385", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Randomized 1,278 people with compensated cirrhosis at 43 French and Belgian sites to ultrasound every three or six months. Three-month scanning found more lesions 10 mm or smaller, 41% versus 28%, but did not improve HCC incidence or the proportion of HCC 30 mm or smaller, 79% versus 70%." }, { "id": "tsuboi-2023-adaura-os", "type": "paper", "title": "Overall survival with osimertinib in resected EGFR-mutated non-small-cell lung cancer", "authors": [ "Tsuboi, Masahiro", "Herbst, Roy S.", "John, Thomas", "et al." ], "publisher": "New England Journal of Medicine", "date": "2023-07-13", "url": "https://pubmed.ncbi.nlm.nih.gov/37272535/", "doi": "10.1056/NEJMoa2304594", "pmid": "37272535", "trialId": "NCT02511106", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Primary randomized phase 3 OS report: stage II-IIIA five-year OS 85% with osimertinib versus 73% with placebo, HR 0.49. Funded by AstraZeneca." }, { "id": "uct-2026-cancer-projects", "type": "institution-profile", "title": "University of Cape Town Cancer Research Initiative — cancer projects", "authors": [ "University of Cape Town Faculty of Health Sciences" ], "publisher": "University of Cape Town", "date": null, "url": "https://health.uct.ac.za/research-research-groupings-cancer-research-initiative/cancer-projects", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Official project directory. It names Eduard Jonas as the UCT contact for pancreatic adenocarcinoma biomarker work, a prospective longitudinal quality-of-life study in South African pancreatic adenocarcinoma, a locally advanced PDAC review, malignant hilar obstruction work and a GEP-NET natural-history study. Project listing does not prove recruitment, completion or outcome." }, { "id": "uemura-2026-elemental-diet-pdac", "type": "paper", "title": "Early elemental diet supplementation during chemotherapy improves treatment response and survival in advanced pancreatic cancer: a single-center randomized controlled trial", "authors": [ "Uemura, Shinya", "Iwashita, Takuji", "Koizumi, Takao", "et al." ], "publisher": "Clinical Nutrition ESPEN", "date": "2026-06-01", "url": "https://pubmed.ncbi.nlm.nih.gov/41794391/", "doi": "10.1016/j.clnesp.2026.103106", "pmid": "41794391", "trialId": "UMIN000028659", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Single-centre open randomized trial, 30 per arm with three intervention withdrawals. Elemental diet during chemotherapy preserved some nutrition and grip measures and reported median overall survival 21.4 versus 12.8 months, HR 0.55, 95% CI 0.31-0.98, p=0.038; PFS was not significant, 9.5 versus 6.4 months, p=0.068. The large survival difference from 60 people needs independent multicentre replication with allocation, treatment exposure and all withdrawals visible." }, { "id": "ukri-accelerated-cancer-diagnosis-2026", "type": "funder-guidance", "title": "Advancing innovation in accelerated cancer diagnosis", "authors": [ "Innovate UK", "Office for Life Sciences" ], "publisher": "UK Research and Innovation", "date": "2026-08-20", "url": "https://www.ukri.org/opportunity/advancing-innovation-in-accelerated-cancer-diagnosis/", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Primary call page. Open 1 September to 28 October 2026 with a total fund up to £25.4m for clinical validation and evaluation of early cancer diagnostic technologies and devices. Collaborative applications must be led by a UK business, academic institution or NHS organisation and contain at least one UK SME claiming grant funding. Call-level fund, not one project award." }, { "id": "ukri-gtr-pancreatic-title-query-2026", "type": "government-data", "title": "Gateway to Research project query for pancreatic cancer", "authors": [ "UK Research and Innovation" ], "publisher": "UK Research and Innovation", "date": "2026-07-06", "url": "https://gtr.ukri.org/api/search/project?term=%22pancreatic%20cancer%22&fetchSize=100&selectedSortableField=score&selectedSortOrder=DESC&page=1", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Primary UKRI award API, last refreshed 6 July 2026. Reproducible calculation: retrieve pages 1 and 2 for the quoted phrase pancreatic cancer; retain projects starting on or after 1 January 2015 whose title contains pancreas or pancreatic and cancer or tumour in either order; sum fund.valuePounds. The query returned 147 projects; 74 passed the title rule, 43 had positive displayed values and 31 displayed zero. Positive displayed commitments totalled £24,106,508. This is a conservative title-visible ledger, not all relevant UK funding. It can omit broad cancer, RAS, liver, biliary, colorectal, infrastructure and block-grant work, and includes full award values when pancreatic cancer shares a multi-cancer project. Commitments are not annual expenditure." }, { "id": "umin-000061741-staging-laparoscopy", "type": "trial-registry", "title": "SL-PC Study: staging laparoscopy before neoadjuvant chemotherapy in pancreatic cancer", "authors": [ "Nanno, Yoshihide", "Kobe University" ], "publisher": "UMIN Clinical Trials Registry", "date": "2026-05-29", "url": "https://center6.umin.ac.jp/cgi-open-bin/icdr_e/ctr_view.cgi?recptno=R000070658", "doi": null, "pmid": null, "trialId": "UMIN000061741", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Prospective single-centre Japanese study planned for 200 patients with scan-localized resectable or borderline-resectable PDAC. The registry lists a September 2026 start, nonrandom single-arm design, hidden-metastasis detection as the primary outcome, no named funding and follow-up to December 2029. Its current preinitiation state means it cannot answer the utility question, and the design still counts discovery rather than comparative patient benefit." }, { "id": "umin-2025-kobe-ai-early-pdac", "type": "trial-registry", "title": "Validation of an Artificial Intelligence-Based Imaging Algorithm for Early Pancreatic Cancer: A Multicenter Study", "authors": [ "Masuda, Atsuhiro" ], "publisher": "Japan Registry of Clinical Trials / UMIN", "date": "2025-09-18", "url": "https://rctportal.mhlw.go.jp/s/detail/um?trial_id=UMIN000059076", "doi": null, "pmid": null, "trialId": "UMIN000059076", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Japanese 400-record retrospective multicentre study of CT AI in cancers 20 mm or smaller, including carcinoma in situ/high-grade PanIN; funded by Fujifilm." }, { "id": "uspstf-2019-pancreatic-screening", "type": "guideline", "title": "Recommendation: Pancreatic Cancer: Screening", "authors": [ "US Preventive Services Task Force" ], "publisher": "US Preventive Services Task Force", "date": "2019-08-06", "url": "https://www.uspreventiveservicestaskforce.org/uspstf/recommendation/pancreatic-cancer-screening", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "The recommendation applies to asymptomatic adults not known to be at high inherited or familial risk. It is a 2019 assessment and is recorded with its date rather than treated as permanent." }, { "id": "vadhan-raj-2020-cassini-pdac", "type": "paper", "title": "Rivaroxaban thromboprophylaxis in ambulatory patients with pancreatic cancer: prespecified CASSINI subgroup", "authors": [ "Vadhan-Raj, Saroj", "McNamara, Mairéad G.", "Venerito, Marino", "et al." ], "publisher": "Cancer Medicine", "date": "2020-09-01", "url": "https://pubmed.ncbi.nlm.nih.gov/32663379/", "doi": "10.1002/cam4.3269", "pmid": "32663379", "trialId": "NCT02555878", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Prespecified pancreatic subgroup: 273 randomized after baseline screening excluded 24/362, 6.6%, with VTE. Through day 180 the primary composite was 9.6% versus 13.0%, HR .70, P=.328. While receiving treatment it was 3.7% versus 10.1%, HR .35, P=.034. Major bleeding was 1.5% versus 2.3%. Only 57% completed the double-blind period; Janssen employees were authors." }, { "id": "valar-2024-series-a", "type": "company-announcement", "title": "Valar Labs announces $22m Series A", "authors": [ "Valar Labs" ], "publisher": "Valar Labs", "date": "2024-05-30", "url": "https://valarlabs.com/news/series-a", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Primary company announcement. Series A was $22m, co-led by DCVC and Andreessen Horowitz with Pear VC participation. It names founders Anirudh Joshi, Viswesh Krishna and Damir Vrabac. Funding and origin are company claims; clinical evidence is assessed separately." }, { "id": "van-cutsem-2009-bevacizumab-erlotinib", "type": "paper", "title": "Phase III trial of bevacizumab with gemcitabine and erlotinib in metastatic pancreatic cancer", "authors": [ "Van Cutsem, Eric", "Vervenne, Willy L.", "Bennouna, Jaafar", "et al." ], "publisher": "Journal of Clinical Oncology", "date": "2009-03-23", "url": "https://pubmed.ncbi.nlm.nih.gov/19307500/", "doi": "10.1200/JCO.2008.20.0238", "pmid": "19307500", "trialId": "NCT00088894", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Double-blind phase 3: 306 bevacizumab versus 301 placebo, both with gemcitabine and erlotinib. Median OS was 7.1 versus 6.0 months, HR 0.89, 95% CI 0.74-1.07, P=.2087, so the main endpoint failed. PFS improved, HR 0.73, 95% CI 0.61-0.86, without proven survival gain." }, { "id": "van-cutsem-2016-maestro", "type": "conference-paper", "title": "MAESTRO: evofosfamide plus gemcitabine in untreated advanced PDAC", "authors": [ "Van Cutsem, Eric", "Lenz, Heinz-Josef", "Furuse, Junji", "et al." ], "publisher": "Journal of Clinical Oncology", "date": "2016-05-20", "url": "https://ascopubs.org/doi/10.1200/JCO.2016.34.15_suppl.4007", "doi": "10.1200/JCO.2016.34.15_suppl.4007", "pmid": null, "trialId": "NCT01746979", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Randomized double-blind phase 3, n=693. Median OS was 8.7 versus 7.6 months, HR 0.84, 95% CI 0.71-1.01, P=.059, so the main survival endpoint failed. PFS was 5.5 versus 3.7 months, HR 0.77, P=.004; confirmed response was 15% versus 9%, P=.009. Blood-count adverse events were more common with evofosfamide." }, { "id": "van-der-gaag-2010-pbd-rct", "type": "paper", "title": "Preoperative biliary drainage for cancer of the head of the pancreas", "authors": [ "van der Gaag, Niels A", "Rauws, Erik A J", "van Eijck, Casper H J", "et al." ], "publisher": "New England Journal of Medicine", "date": "2010-01-14", "url": "https://pubmed.ncbi.nlm.nih.gov/20071702/", "doi": "10.1056/NEJMoa0903230", "pmid": "20071702", "trialId": "ISRCTN31939699", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Multicentre randomized trial of 202 patients with pancreatic-head cancer, obstructive jaundice and bilirubin 40-250 micromol/L. Early surgery within one week produced serious complications in 37/96, 39%, versus 75/106, 74%, after four to six weeks of mainly plastic-stent drainage. Drainage itself caused complications in 47/102 analyzed patients, 46%, including 27 cholangitis cases reported in the full text. Mortality did not differ. The result rejects routine delayed plastic-stent drainage in this selected bilirubin range; it does not address neoadjuvant therapy or modern metal stents." }, { "id": "van-dijk-2025-polyp-study", "type": "paper", "title": "Revised Guidelines for the Treatment and Follow-Up of Gallbladder Polyps Do Not Reduce Unwarranted Cholecystectomies: Results of the POLYP Study", "authors": [ "van Dooren, Mike", "de Savornin Lohman, Elise A. J.", "de Reuver, Philip" ], "publisher": "United European Gastroenterology Journal", "date": "2025-07-23", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC12463707/", "doi": "10.1002/ueg2.70057", "pmid": "40697033", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Full open-access report checked. Prospective Dutch cohort across 26 centres included 302 patients already in polyp follow-up or surgery. Eighty-eight underwent cholecystectomy; 68 met the 2022 guideline indication. Six adenomas, four dysplastic adenomas, one high-grade dysplastic adenoma and no cancers were found. Recruitment through surgical services and unfinished follow-up enrich surgery and prevent population-risk inference." }, { "id": "van-dongen-2023-preopanc-staging", "type": "paper", "title": "The yield of staging laparoscopy for resectable and borderline resectable pancreatic cancer in the PREOPANC randomized controlled trial", "authors": [ "van Dongen, Jelle C", "Versteijne, Eva", "Bonsing, Bert A", "et al." ], "publisher": "European Journal of Surgical Oncology", "date": "2023-04-01", "url": "https://pubmed.ncbi.nlm.nih.gov/36585300/", "doi": "10.1016/j.ejso.2022.12.011", "pmid": "36585300", "trialId": "NTR3709", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Post-hoc analysis of PREOPANC. Among 246 randomized patients, 133 had staging laparoscopy and 106 went directly to open exploration according to their assigned treatment route, not random assignment of the staging method. Hidden spread was found in 13/133, 9.8%, at laparoscopy and 9/106, 8.5%, at laparotomy. Palliative chemotherapy was received by 76.9% versus 30.0%, P=.040, when spread was found by laparoscopy rather than laparotomy. One person had peritoneal spread at open exploration after a negative laparoscopy." }, { "id": "van-goor-2025-nationwide-imaging", "type": "paper", "title": "Routine Imaging After Resection of Pancreatic Ductal Adenocarcinoma: Nationwide Utilization and Survival Analysis", "authors": [ "van Goor, Iris W J M", "Andel, Pepijn C M", "Augustinus, Simone", "et al." ], "publisher": "Annals of Surgery", "date": "2025-06-03", "url": "https://pubmed.ncbi.nlm.nih.gov/40458871/", "doi": "10.1097/sla.0000000000006772", "pmid": "40458871", "trialId": "NCT04605237", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Nationwide Dutch observational cohort of 1,311 resected PDAC patients from 2014-2019. Only 139, 11%, had routine imaging. Median overall survival was 43 versus 22 months; routine imaging was associated with asymptomatic recurrence OR 3.79, recurrence treatment OR 2.50 and overall survival HR 0.56. Non-random allocation and the very large survival difference make healthy-patient, centre, trial-participation and treatment-selection bias serious alternatives. Authors reported no conflicts." }, { "id": "van-hilst-2019-leopard2", "type": "paper", "title": "Laparoscopic versus open pancreatoduodenectomy for pancreatic or periampullary tumours (LEOPARD-2): a multicentre, patient-blinded, randomised controlled phase 2/3 trial", "authors": [ "van Hilst, Jony", "de Rooij, Thijs", "Bosscha, Koop", "et al." ], "publisher": "The Lancet Gastroenterology & Hepatology", "date": "2019-01-24", "url": "https://europepmc.org/article/MED/30685489", "doi": "10.1016/s2468-1253(19)30004-4", "pmid": "30685489", "trialId": "NTR5689", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Four-centre randomized trial in centres doing at least 20 pancreaticoduodenectomies yearly with trained surgeons who had done at least 20 laparoscopic procedures. It stopped after 105 randomizations and 99 operations because 90-day complication-related mortality was 5/50, 10%, laparoscopic versus 1/49, 2%, open; RR 4.90 with very wide 95% CI 0.59-40.44. Functional recovery was 10 versus 8 days. It included benign, premalignant and malignant disease and was funded by an investigator-initiated Johnson & Johnson Medical grant." }, { "id": "van-laethem-2024-optimize1", "type": "paper", "title": "Combining CD40 agonist mitazalimab with mFOLFIRINOX in previously untreated metastatic PDAC: OPTIMIZE-1", "authors": [ "Van Laethem, Jean-Luc", "Borbath, Ivan", "Prenen, Hans", "et al." ], "publisher": "The Lancet Oncology", "date": "2024-06-01", "url": "https://pubmed.ncbi.nlm.nih.gov/38834087/", "doi": "10.1016/S1470-2045(24)00263-8", "pmid": "38834087", "trialId": "NCT04888312", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Single-arm phase 1b/2. Seventy enrolled, 57 were in the efficacy set after receiving at least two cycles at the chosen dose; ORR was 23/57=40%. Serious adverse events affected 29/70=41%; no treatment-related deaths were reported. Without a concurrent mFOLFIRINOX group the added drug's contribution cannot be estimated." }, { "id": "van-laethem-2025-optimize1-biomarkers", "type": "paper", "title": "CD40 agonist mitazalimab with mFOLFIRINOX in untreated metastatic pancreatic cancer: biomarkers associated with outcomes from OPTIMIZE-1", "authors": [ "Van Laethem, Jean-Luc", "Geboes, Karen", "Borbath, Ivan", "et al." ], "publisher": "Cell Reports Medicine", "date": "2025-10-07", "url": "https://pubmed.ncbi.nlm.nih.gov/41061701/", "doi": "10.1016/j.xcrm.2025.102407", "pmid": "41061701", "trialId": "NCT04888312", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Updated single-arm OPTIMIZE-1 analysis set n=57. Confirmed ORR was 42.1%, median response duration 12.6 months, PFS 7.7 months and OS 14.9 months. Baseline fibrosis-linked tumour measurements and on-treatment immune changes correlated with outcomes but were found inside one uncontrolled product cohort." }, { "id": "van-vliembergen-2026-men1-mnet-assay", "type": "paper", "title": "Circulating Gene Expression Assay as a Diagnostic and Prognostic Biomarker for Pancreatic Neuroendocrine Tumors in MEN1", "authors": [ "van Vliembergen, Eline N. M.", "van Treijen, Mark J. C.", "van Leeuwaarde, Rachel S." ], "publisher": "Journal of Clinical Endocrinology and Metabolism", "date": "2026-01-21", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC12819851/", "doi": "10.1210/clinem/dgaf374", "pmid": "40581735", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Single-centre prospective cohort of 110 eligible MEN1 patients followed for a median reported 39 months. The recalibrated circulating mRNA assay did not predict new PanNET, growth, tumour count, new metastasis or metastatic progression; reported AUCs were 0.65, 0.49, 0.54, 0.39 and 0.46 respectively. This is a useful negative biomarker result, not evidence that all blood-based risk signals will fail." }, { "id": "varghese-2025-kras-dosage", "type": "paper", "title": "Clinicogenomic landscape of pancreatic adenocarcinoma identifies KRAS mutant dosage as prognostic of overall survival", "authors": [ "Varghese, Anna M.", "Perry, Megan A.", "Chou, Joanne F.", "et al." ], "publisher": "Nature Medicine", "date": "2025-02-01", "url": "https://pubmed.ncbi.nlm.nih.gov/39753968/", "doi": "10.1038/s41591-024-03362-3", "pmid": "39753968", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Clinicogenomic study of 2,336 patients across stages. Observational survival associations and tumour sequencing; not a treatment-selection trial." }, { "id": "vehvilainen-2024-cholangitis-nat-interruption", "type": "paper", "title": "Cholangitis and interruptions of neoadjuvant chemotherapy associate with reduced survival in pancreatic cancer", "authors": [ "Vehviläinen, Sini", "Kuuliala, Antti", "Udd, Marianne", "et al." ], "publisher": "Annals of Surgical Oncology", "date": "2024-04-01", "url": "https://pubmed.ncbi.nlm.nih.gov/38153645/", "doi": "10.1245/s10434-023-14793-6", "pmid": "38153645", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Retrospective cohort of 162 pancreatic-head cancers that underwent neoadjuvant treatment, biliary drainage and later pancreaticoduodenectomy. Cholangitis affected 33, 20%; median OS was 26 versus 36 months and PFS 8 versus 17. Stent-failure interruption affected 26, 16%; median OS was 26 versus 35 months and PFS 7 versus 17. No survival difference was observed by stent type. Only eventual surgical patients were included, and infection, tumour biology, treatment tolerance and route selection can confound the survival association." }, { "id": "verastem-2026-ramp205-update", "type": "company-announcement", "title": "Updated RAMP 205 results in first-line metastatic pancreatic cancer", "authors": [ "Verastem, Inc." ], "publisher": "Verastem Oncology", "date": "2026-06-17", "url": "https://investor.verastem.com/news-releases/news-release-details/verastem-oncology-announces-positive-updated-results-ramp-205-0", "doi": null, "pmid": null, "trialId": "NCT05669482", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Sponsor update in 29 patients at the recommended phase 2 dose: confirmed response 52%, six-month progression-free survival 68%, six-month overall survival 86%, median follow-up 9.8 months. No randomized comparator and survival remained immature." }, { "id": "verkolf-2026-taps-resection-prediction", "type": "paper", "title": "Prediction of resection after preoperative FOLFIRINOX in patients with localized pancreatic adenocarcinoma: a Trans-Atlantic Pancreatic Surgery (TAPS) Consortium study", "authors": [ "Verkolf, Eva M. M.", "van Dam, Jacob L.", "Dekker, Esther N.", "et al." ], "publisher": "Journal of the National Cancer Institute", "date": "2026-06-01", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC13247335/", "doi": "10.1093/jnci/djag033", "pmid": "42108563", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Retrospective five-referral-centre TAPS cohort of 1,835 localized PDAC patients treated first with modified FOLFIRINOX. Observed resection was 17.8% for initially locally advanced disease. Stage, baseline CA19-9 above 500 U/mL, performance status and tumour size produced estimated resection probabilities from 6.7% to 81.7%, with AUC 0.767-0.803 in development, later patients from the same centres and the combined cohort. It predicts the expert-centre decision and result, not whether surgery improves survival." }, { "id": "versteijne-2022-preopanc-longterm", "type": "paper", "title": "Neoadjuvant Chemoradiotherapy Versus Upfront Surgery for Resectable and Borderline Resectable Pancreatic Cancer: Long-Term Results of the Dutch Randomized PREOPANC Trial", "authors": [ "Versteijne, Eva", "van Dam, Jacob L.", "Suker, Mustafa", "et al." ], "publisher": "Journal of Clinical Oncology", "date": "2022-04-10", "url": "https://pubmed.ncbi.nlm.nih.gov/35084987/", "doi": "10.1200/JCO.21.02233", "pmid": "35084987", "trialId": "EudraCT 2012-003181-40", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Long-term intention-to-treat result in 246 eligible randomized patients. The initial report did not reach significance for OS; the 59-month follow-up did." }, { "id": "viewray-2023-chapter11-8k", "type": "financial-filing", "title": "ViewRay Chapter 11 filing", "authors": [ "ViewRay, Inc." ], "publisher": "U.S. Securities and Exchange Commission", "date": "2023-07-16", "url": "https://www.sec.gov/Archives/edgar/data/1597313/000162828023024952/vray-20230715.htm", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Primary SEC current report. ViewRay and its operating subsidiary filed voluntary Chapter 11 petitions on 16 July 2023." }, { "id": "viewray-2023-q1-10q", "type": "financial-filing", "title": "ViewRay quarterly report for the period ended 31 March 2023", "authors": [ "ViewRay, Inc." ], "publisher": "U.S. Securities and Exchange Commission", "date": "2023-05-10", "url": "https://www.sec.gov/Archives/edgar/data/1597313/000162828023017657/vray-20230331.htm", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Primary SEC filing. ViewRay reported $81.3m cash, $4.2m restricted cash, $77.8m debt outstanding, a $29.6m quarterly net loss and $53.6m operating cash use. Management stated substantial doubt about continuing as a going concern." }, { "id": "villard-2023-psc-surveillance", "type": "paper", "title": "Prospective surveillance for cholangiocarcinoma in unselected individuals with primary sclerosing cholangitis", "authors": [ "Villard, Christina", "Friis-Liby, Ingalill", "Rorsman, Fredrik", "et al." ], "publisher": "Journal of Hepatology", "date": "2022-11-19", "url": "https://pubmed.ncbi.nlm.nih.gov/36410555/", "doi": "10.1016/j.jhep.2022.11.011", "pmid": "36410555", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Prospective nationwide Swedish surveillance cohort: 512 unselected people with primary sclerosing cholangitis from 11 hospitals, followed for five years with annual MRI/MRCP, CA19-9 and clinical review. Eleven cholangiocarcinomas and two high-grade dysplasias were diagnosed; median survival after cholangiocarcinoma was 13 months. Severe or progressive duct change occurred in 24%, with malignancy in 10% of that subgroup. The surveillance route did not detect cancer early enough for long-term survival." }, { "id": "vineet-2026-north-india-gallbladder-cohort", "type": "paper", "title": "Comprehensive real-world outcomes of multimodal treatment in gallbladder cancer: a prospective cohort study", "authors": [ "Vineet, Kumar", "Tripathi, Mayank", "Shukla, Piyush Kant", "et al." ], "publisher": "Journal of Gastrointestinal Cancer", "date": "2026-02-23", "url": "https://pubmed.ncbi.nlm.nih.gov/41724874/", "doi": "10.1007/s12029-026-01425-x", "pmid": "41724874", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Prospective observational cohort of 1,500 newly diagnosed patients at Tata Memorial Centre in Varanasi, North India. Eighty percent presented metastatic. Treatment allocation was not randomized, so surgery outcomes are heavily selected." }, { "id": "vogl-2026-tace-mwa", "type": "paper", "title": "Interventional management of unresectable pancreatic cancer using transarterial chemoembolization and microwave ablation: a single-center evaluation over 12 years", "authors": [ "Vogl, Thomas J.", "Naguib, N. N. N.", "et al." ], "publisher": "Journal of Cancer Research and Clinical Oncology", "date": "2026-02-11", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC13038777/", "doi": "10.1007/s00432-026-06463-3", "pmid": "41917374", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Retrospective single-centre comparison over 2010-2023: TACE plus microwave n=67 versus TACE n=83. The tumour board preferentially gave microwave to lower-burden cases: mean tumours 3.4 versus 4.9 cm and metastases 34.3% versus 54.2%. Median survival was 14.6 versus 9.0 months, but no causal comparison was intended. Imaging was missing for 20/67 microwave cases versus 0/83 controls." }, { "id": "wainberg-2023-napoli3", "type": "paper", "title": "NALIRIFOX versus nab-paclitaxel and gemcitabine in treatment-naive patients with metastatic pancreatic ductal adenocarcinoma (NAPOLI 3)", "authors": [ "Wainberg, Zev A.", "Melisi, Davide", "Macarulla, Teresa", "et al." ], "publisher": "The Lancet", "date": "2023-09-11", "url": "https://pubmed.ncbi.nlm.nih.gov/37708904/", "doi": "10.1016/S0140-6736(23)01366-1", "pmid": "37708904", "trialId": "NCT04083235", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "International open-label phase 3 trial at 187 sites in 18 countries. Funded by Ipsen. Both groups were restricted to ECOG performance status 0 or 1." }, { "id": "wainberg-2025-eli002-final", "type": "paper", "title": "Lymph node-targeted, mKRAS-specific amphiphile vaccine in pancreatic and colorectal cancer: phase 1 AMPLIFY-201 trial final results", "authors": [ "Wainberg, Zev A.", "Weekes, Colin D.", "Furqan, Muhammad", "et al." ], "publisher": "Nature Medicine", "date": "2025-11-01", "url": "https://pubmed.ncbi.nlm.nih.gov/40790272/", "doi": "10.1038/s41591-025-03876-4", "pmid": "40790272", "trialId": "NCT04853017", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Final phase 1 follow-up in 25 patients with pancreatic or colorectal cancer and molecular residual disease. Outcomes were compared after treatment by strength of immune response, not against an untreated or randomised control. Elicio Therapeutics employees were authors and the study had extensive industry disclosures." }, { "id": "wainberg-2026-quemliclustat-arc8", "type": "paper", "title": "Quemliclustat and chemotherapy with or without zimberelimab in metastatic pancreatic adenocarcinoma: a randomized phase 1 trial", "authors": [ "Wainberg, Zev A.", "Manji, Gulam A.", "Bahary, Nathan", "et al." ], "publisher": "Nature Medicine", "date": "2026-03-30", "url": "https://www.nature.com/articles/s41591-026-04283-z", "doi": "10.1038/s41591-026-04283-z", "pmid": null, "trialId": "NCT04104672", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "ARC-8 phase 1b. Dose expansion screened 158 and enrolled 116; 90 were randomized 2:1 between quemliclustat plus chemotherapy with or without zimberelimab, so no randomized arm omitted quemliclustat. Across 122 people given quemliclustat 100 mg, median OS was 15.7 months versus 9.8 in a post-hoc matched external control, HR 0.634, while ORR and PFS did not differ significantly. Thirty-seven paired biopsies supported target-pathway change, but outcome analyses were post-treatment associations. Arcus employees were authors." }, { "id": "walker-2021-embedded-germline-station", "type": "paper", "title": "Implementation of an Embedded In-Clinic Genetic Testing Station to Optimize Germline Testing for Patients with Pancreatic Adenocarcinoma", "authors": [ "Walker, Evan J.", "Goldberg, Dena", "Gordon, Kelly M.", "et al." ], "publisher": "The Oncologist", "date": "2021-09-20", "url": "https://pubmed.ncbi.nlm.nih.gov/34506673/", "doi": "10.1002/onco.13968", "pmid": "34506673", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Retrospective UCSF implementation comparison. Testing was offered to 209 of 223 eligible patients and completed by 158, 71%, versus 19% before the embedded station. Attrition between referral and genetics appointment fell from 36% to 3%. Younger, White and English-speaking patients were more likely to complete testing." }, { "id": "walter-2026-compete", "type": "paper", "title": "177Lu-Lu-edotreotide versus everolimus for gastroenteropancreatic neuroendocrine tumours (COMPETE)", "authors": [ "Walter, Thomas", "Jann, Henning", "Ansquer, Catherine", "et al." ], "publisher": "The Lancet", "date": "2026-07-18", "url": "https://pubmed.ncbi.nlm.nih.gov/42392118/", "doi": "10.1016/S0140-6736(26)00604-5", "pmid": "42392118", "trialId": "NCT03049189", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Open-label phase 3 across 49 specialist centres in 14 countries; 309 randomized with mixed grade 1-2 gastroenteropancreatic NET origins. Funded by ITM Solucin; company employees were authors." }, { "id": "wang-2022-oncology-germline-cascade", "type": "paper", "title": "Oncology clinic-based germline genetic testing for exocrine pancreatic cancer enables timely return of results and unveils low uptake of cascade testing", "authors": [ "Wang, Yifan", "Golesworthy, Blair", "Cuggia, Audrey", "et al." ], "publisher": "Journal of Medical Genetics", "date": "2021-09-23", "url": "https://pubmed.ncbi.nlm.nih.gov/34556502/", "doi": "10.1136/jmedgenet-2021-108054", "pmid": "34556502", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Prospective clinic cohort. Universal point-of-care testing reached 97.4% versus 58.5% under criteria-led testing and cut median turnaround from 42 to 13 days. Probands disclosed results to 84% of first-degree relatives, but only 31% of informed relatives were tested; the mean was 0.52 newly identified carriers per index case." }, { "id": "wang-2023-sonochemotherapy-pdac", "type": "paper", "title": "Clinical sonochemotherapy of inoperable pancreatic cancer using diagnostic ultrasound and microbubbles: a multicentre, open-label, randomised, controlled trial", "authors": [ "Feng Han", "Yanjie Wang", "Xiaoxiao Dong", "and collaborators" ], "publisher": "European Radiology", "date": "2024-03", "url": "https://pubmed.ncbi.nlm.nih.gov/37796294/", "doi": "10.1007/s00330-023-10210-4", "pmid": "37796294", "trialId": "ChiCTR2100044721", "patentId": null, "grantId": "2017YFC0107300; 82127804; 82102075; 82102077", "accessed": "2026-09-14", "notes": "Chinese multicentre open-label randomized trial; 82 recruited and 78 allocated, comparing chemotherapy with or without 30 minutes of diagnostic ultrasound plus microbubbles. Published online in 2023 and in the March 2024 issue." }, { "id": "wang-2024-nglycan-pdac", "type": "paper", "title": "N-glycan biosignatures as a potential diagnostic biomarker for early-stage pancreatic cancer", "authors": [ "Wang and collaborators" ], "publisher": "World Journal of Gastroenterology", "date": "2024", "url": "https://pubmed.ncbi.nlm.nih.gov/38577461/", "doi": null, "pmid": "38577461", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Case-control development set: 71 PDAC across stages including 29 stage I and 88 controls; internal model only." }, { "id": "wang-2025-glp1-pancreatic-risk", "type": "paper", "title": "Glucagon-like peptide-1 receptor agonists and pancreatic cancer risk: target trial emulation using real-world data", "authors": [ "Wang, Lindsey", "Wang, Qian", "Li, Lin", "Kaelber, David C.", "Xu, Rong" ], "publisher": "Journal of the National Cancer Institute", "date": "2025-03-01", "url": "https://pubmed.ncbi.nlm.nih.gov/39418202/", "doi": "10.1093/jnci/djae260", "pmid": "39418202", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Retrospective electronic-record cohort of 1,636,056 people with type 2 diabetes, including 167,091 prescribed GLP-1 receptor agonists. Propensity-matched comparisons against six medicine classes gave pancreatic-cancer HRs from 0.42 to 0.82 over five years. The design imitates a trial but does not randomize treatment and remains vulnerable to treatment selection, missing exposure and detection differences." }, { "id": "wang-2026-cfdna-prospective-detection", "type": "paper", "title": "Development and Prospective Validation of a Cell-free DNA-Based Model for the Early Detection of Pancreatic Cancer", "authors": [ "Wang, Xiuchao", "Wang, Hongwei", "Zhang, Meng", "et al." ], "publisher": "Cancer Discovery", "date": "2026-01-12", "url": "https://aacrjournals.org/cancerdiscovery/article/16/1/66/771234/Development-and-Prospective-Validation-of-a-Cell", "doi": "10.1158/2159-8290.CD-25-0323", "pmid": "40982573", "trialId": null, "patentId": null, "grantId": "2021YFA1201100", "accessed": "2026-09-15", "notes": "Development included 232 pancreatic-cancer cases and 235 healthy controls. The prospective diabetes/obesity cohort enrolled 1,926 people and observed only eight PDAC cases over 12 to 24 months. The locked test found six of eight, including three of three stage-0 lesions, but produced 40 false positives: sensitivity 75%, specificity 97.9% and positive predictive value 13.0%. The stage-0 label included high-grade PanIN or IPMN/carcinoma in situ. Clinical utility and mortality were not tested. Several authors were affiliated with Geneseeq, which is relevant to independence and product replication." }, { "id": "wang-2026-itih3-ceacam1-panel", "type": "paper", "title": "A Biomarker Panel for the Detection of Pancreatic Cancer", "authors": [ "Wang, Yuefan", "Huang, Yuanyu", "Lih, Tung-Shing M.", "et al." ], "publisher": "Cancers", "date": "2026-04-24", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC13162998/", "doi": "10.3390/cancers18091397", "pmid": "42122194", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Matched plasma and serum from 649 participants, 250 with known PDAC and 399 controls, were split into training and blinded validation. CA19-9 plus ITIH3 and CEACAM1 reached validation AUC 0.917. At a cutoff fixed for roughly 95% sensitivity, specificity was only 53.3%; for stage-I/II cases versus healthy or normal-pancreas controls, sensitivity was 96.6% and specificity 55.6%. This is assay validation in selected known cases, not prospective screening utility." }, { "id": "wehrle-2025-subcentimeter-ipmn", "type": "paper", "title": "Surveillance of subcentimeter side-branch intraductal papillary mucinous neoplasms: risk of invasive disease and follow-up recommendations", "authors": [ "Wehrle, Chase J.", "Walsh, R. Matthew", "Kumar, Pranav", "et al." ], "publisher": "Journal of Gastrointestinal Surgery", "date": "2025-01-08", "url": "https://pubmed.ncbi.nlm.nih.gov/39793956/", "doi": "10.1016/j.gassur.2025.101959", "pmid": "39793956", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Prospectively maintained single-system surveillance database, n=1,000. Subcentimetre cysts progressed by clinical criteria less often, but invasive cancer was 1.4% versus 1.8% in larger cysts and growth rate was not associated with high-risk pathology. This was not a randomized surveillance comparison." }, { "id": "weinberg-2020-ea2185-protocol", "type": "paper", "title": "Comparing the clinical impact of pancreatic cyst surveillance programs: A trial of the ECOG-ACRIN cancer research group (EA2185)", "authors": [ "Weinberg, David S.", "Gatsonis, Constantine", "Zeh, Herbert J.", "et al." ], "publisher": "Contemporary Clinical Trials", "date": "2020-09-10", "url": "https://pubmed.ncbi.nlm.nih.gov/32920242/", "doi": "10.1016/j.cct.2020.106144", "pmid": "32920242", "trialId": "NCT04239573", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Published protocol planned to randomize 4,606 asymptomatic people 1:1 between high- and low-intensity surveillance for five years, powered for a 30% relative reduction in a composite unfavorable outcome. The live registry no longer describes this as the current primary design." }, { "id": "wellcome-2026-discovery-awards", "type": "funder-guidance", "title": "Wellcome Discovery Awards", "authors": [ "Wellcome" ], "publisher": "Wellcome", "date": "2026-04-01", "url": "https://wellcome.org/research-funding/schemes/wellcome-discovery-awards", "doi": null, "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Primary scheme page. Wellcome reports an average Discovery Award of £3.5m and average duration of seven years, with awards spanning three to eight years. The route supports discovery research at eligible not-for-profit hosts; it is not a substitute for a commercial pivotal programme." }, { "id": "wiessner-2024-organoid-biopsy-randomized", "type": "paper", "title": "Comparison of endoscopic ultrasound-guided fine-needle aspiration and fine-needle biopsy to generate pancreatic cancer organoids: Randomized trial", "authors": [ "Wiessner, Johannes Roman", "Orben, Felix", "Schäfer, Arlett", "Schlag, Christoph" ], "publisher": "Endoscopy International Open", "date": "2024-03-07", "url": "https://pubmed.ncbi.nlm.nih.gov/38464982/", "doi": "10.1055/a-2257-3171", "pmid": "38464982", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Primary paired randomized-order sampling study in 50 suspected pancreatic cancers. Organoids were generated from 17 people, 34%: nine by fine-needle biopsy only, two by fine-needle aspiration only and six by both. Fine-needle biopsy was numerically better but not statistically significant; only 13 of 17 organoid cases were malignant on histology." }, { "id": "winter-2013-smad4-failure-patterns", "type": "paper", "title": "Failure patterns in resected pancreas adenocarcinoma: lack of predicted benefit to SMAD4 expression", "authors": [ "Winter, Jordan M.", "Tang, Laura H.", "Klimstra, David S.", "et al." ], "publisher": "Annals of Surgery", "date": "2013-08-01", "url": "https://pubmed.ncbi.nlm.nih.gov/23360922/", "doi": "10.1097/SLA.0b013e31827fe9ce", "pmid": "23360922", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Retrospective analysis of 127 resected PDACs. Did not confirm the proposed use of SMAD4 expression to predict local versus distant recurrence." }, { "id": "wode-2026-mistral", "type": "paper", "title": "Mistletoe extract in patients with advanced pancreatic cancer: health-related quality of life in a double-blind, randomized, placebo-controlled trial", "authors": [ "Wode, Karin", "Björ, Ove", "Klein, Reinhold", "et al." ], "publisher": "Palliative Medicine", "date": "2026-07-01", "url": "https://pubmed.ncbi.nlm.nih.gov/42059487/", "doi": "10.1177/02692163261437609", "pmid": "42059487", "trialId": "NCT02948309", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Investigator-led, nine-centre, double-blind phase 3 trial of 290 people with advanced pancreatic cancer. Mistletoe extract added to chemotherapy or best supportive care did not improve quality of life, weight, measured care use or biomarkers versus placebo. The authors reported no clinical reason to recommend it." }, { "id": "wolchok-2024-checkmate067-ten-year", "type": "paper", "title": "Final, 10-year outcomes with nivolumab plus ipilimumab in advanced melanoma", "authors": [ "Wolchok, Jedd D.", "Chiarion-Sileni, Vanna", "Rutkowski, Piotr", "et al." ], "publisher": "New England Journal of Medicine", "date": "2024-09-15", "url": "https://pubmed.ncbi.nlm.nih.gov/39282897/", "doi": "10.1056/NEJMoa2407417", "pmid": "39282897", "trialId": "NCT01844505", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Primary randomized 10-year report: median OS 71.9 months with nivolumab plus ipilimumab, 36.9 with nivolumab and 19.9 with ipilimumab; 37% alive at end of trial in combination arm. Funded by Bristol Myers Squibb and others." }, { "id": "wong-2025-ambrose-cholecystectomy", "type": "paper", "title": "30-day Morbidity and Mortality after Cholecystectomy for Benign Gallbladder Disease (AMBROSE): A Prospective, International Collaborative Cohort Study", "authors": [ "Wong, Geoffrey Yuet Mun", "Wadhawan, Himanshu", "Singhal, Rishi" ], "publisher": "Annals of Surgery", "date": "2025-02-01", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC11723498/", "doi": "10.1097/SLA.0000000000006236", "pmid": "38348652", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Full open-access report checked. Prospective international cohort captured 21,706 consecutive cholecystectomies for benign disease across 57 countries. Thirty-day complications occurred in 8.0%, mortality in 0.4%, bile leak in 1.3% and severe bile-duct injury in 0.2%. The cohort mixes elective, emergency and delayed surgery, so these are a visible global harm envelope rather than the expected risk of a selected asymptomatic prevention operation." }, { "id": "woo-2016-pert-rct", "type": "paper", "title": "Efficacy of pancreatic exocrine replacement therapy for patients with unresectable pancreatic cancer in a randomized trial", "authors": [ "Woo, Sang Myung", "Joo, Jungnam", "Kim, So Young", "et al." ], "publisher": "Pancreatology", "date": "2016-11-01", "url": "https://pubmed.ncbi.nlm.nih.gov/27618657/", "doi": "10.1016/j.pan.2016.09.001", "pmid": "27618657", "trialId": "NCT01587534", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Double-blind placebo-controlled phase 2, n=67 with unresectable pancreatic cancer. Eight-week weight change, nutritional score, quality of life and OS did not differ significantly. Median OS was 5.84 versus 8.13 months, P=.744. The trial may have included patients without a selected enzyme-deficient state." }, { "id": "woo-2019-pancreatobiliary-palliative-rct", "type": "paper", "title": "Effect of Early Management on Pain and Depression in Patients with Pancreatobiliary Cancer: A Randomized Clinical Trial", "authors": [ "Woo, Sang Myung", "Song, Mi Kyung", "Lee, Meeyoung", "et al." ], "publisher": "Cancers", "date": "2019-01-11", "url": "https://doi.org/10.3390/cancers11010079", "doi": "10.3390/cancers11010079", "pmid": "30641928", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Randomized 288 people with advanced pancreatic or biliary cancer and pain or depression to early structured symptom management or on-demand palliative care. The two main four-week endpoints were negative: at least 50% pain reduction 29.5% versus 25.2%, p=0.4194; depression reduction 30.8% versus 36.8%, p=0.5732. More early-care patients reached worst-pain score at most three, 51.1% versus 38.9%, but this was secondary." }, { "id": "wu-2011-gnas-ipmn", "type": "paper", "title": "Recurrent GNAS mutations define an unexpected pathway for pancreatic cyst development", "authors": [ "Wu, Jun", "Matthaei, Holger", "Maitra, Anirban", "et al." ], "publisher": "Science Translational Medicine", "date": "2011-07-01", "url": "https://pubmed.ncbi.nlm.nih.gov/21775669/", "doi": "10.1126/scitranslmed.3002543", "pmid": "21775669", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Cyst-fluid and tissue sequencing. The study shows diagnostic association and clonal continuity in a small linked subset; it does not show that mutation detection predicts which IPMN will progress." }, { "id": "wu-2022-nod-imaging-pilot", "type": "paper", "title": "Imaging of the Pancreas in New-Onset Diabetes: A Prospective Pilot Study", "authors": [ "Wu, Bechien U.", "Lustigova, Eva", "Chen, Qiaoling", "et al." ], "publisher": "Clinical and Translational Gastroenterology", "date": "2022-06-01", "url": "https://pubmed.ncbi.nlm.nih.gov/35333778/", "doi": "10.14309/ctg.0000000000000478", "pmid": "35333778", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Prospective health-system pilot in adults aged 50-85 with new glycaemic abnormality. Ninety-three completed pancreas-protocol CT scans: one stage-IV PDAC, 12 people with other pancreatic findings and 52 people with extrapancreatic findings. Twelve of 57 extrapancreatic findings prompted clinical evaluation. Small size and COVID-19 interruption prevent utility inference." }, { "id": "wu-2026-cyst-surveillance-practice", "type": "paper", "title": "Surveillance Testing and Pancreatic Cancer Incidence Among Patients With Pancreatic Cysts", "authors": [ "Wu, Bechien U.", "Chen, Qiaoling", "Luong, Tiffany Q.", "et al." ], "publisher": "JAMA Network Open", "date": "2026-08-03", "url": "https://pubmed.ncbi.nlm.nih.gov/42545702/", "doi": "10.1001/jamanetworkopen.2026.27103", "pmid": "42545702", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Retrospective Kaiser Permanente cohort, n=13,258, median follow-up 3.1 years. EUS and surgery use fell over time while stage and survival among 149 cancers did not change significantly by practice era. Calendar-era comparison is not randomization and later cohorts have shorter possible follow-up." }, { "id": "wyse-2011-early-eus-cpn-rct", "type": "paper", "title": "Randomized, double-blind, controlled trial of early endoscopic ultrasound-guided celiac plexus neurolysis to prevent pain progression in patients with newly diagnosed, painful, inoperable pancreatic cancer", "authors": [ "Wyse, Jonathan M", "Carone, Marco", "Paquin, Sarto C", "Usatii, Mariana", "Sahai, Anand V" ], "publisher": "Journal of Clinical Oncology", "date": "2011-08-15", "url": "https://doi.org/10.1200/JCO.2010.32.2750", "doi": "10.1200/JCO.2010.32.2750", "pmid": "21844506", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Randomized double-blind comparison of early EUS-guided celiac plexus alcohol neurolysis versus conventional pain management in 96 of 580 eligible patients seen. Pain change favoured neurolysis at three months, with between-group mean percent-change difference -60.7, 95% CI -86.6 to -25.5; morphine use only trended lower and quality of life and survival did not improve." }, { "id": "xia-2020-repeat-hcc-treatment", "type": "paper", "title": "Long-term effects of repeat hepatectomy vs percutaneous radiofrequency ablation among patients with recurrent hepatocellular carcinoma: a randomized clinical trial", "authors": [ "Xia, Yong", "Li, Jun", "Liu, Guanghua", "et al." ], "publisher": "JAMA Oncology", "date": "2020-02-01", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC6902111/", "doi": "10.1001/jamaoncol.2019.4477", "pmid": "31774468", "trialId": "NCT00822562", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Full open-access report checked. Open-label Chinese randomized trial of 240 patients with early recurrent HCC, Child-Pugh A liver function and disease treatable by either approach. Five-year OS and repeat RFS did not differ significantly. Ablation caused more local and early repeat recurrence; surgery caused more complications. Figure 3 was corrected after publication." }, { "id": "xie-2026-cnsi-fe-first-human", "type": "paper", "title": "First-in-human evidence of multidrug resistance reversal in solid tumors: a cohort analysis of carbon nanoparticles-Fe(II) complex trials", "authors": [ "Xie, Ping", "Huang, Yuanfang", "Wang, Yongsheng", "et al." ], "publisher": "BMC Cancer", "date": "2026-06-05", "url": "https://doi.org/10.1186/s12885-026-16216-7", "doi": "10.1186/s12885-026-16216-7", "pmid": "42249319", "trialId": "NCT06048367", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Post-trial analysis of 19 people across many cancers after direct tumour injection of carbon-particle-carried ferrous iron. Two had pancreatic cancer. One pancreatic participant had stable injected-lesion size, +12%, then responded to liposomal irinotecan and S-1 and remained alive beyond 28 months. Ferroptosis was not measured in individual people. Mean tumour-growth-rate change in 14 evaluable people was -5.85 percentage points per month, 95% CI -26.62 to 14.92, p=0.554. Fourteen of 19 had grade 3-plus events, eight had serious events, two were lost to survival follow-up and no treatment-related death was reported." }, { "id": "xu-2025-ibi389-phase1", "type": "paper", "title": "Safety and preliminary efficacy of IBI389 in solid tumours and gastric or gastro-oesophageal tumours", "authors": [ "Li, Xiaobing", "Dai, Ruihong", "Xu, Qiang", "et al." ], "publisher": "BMC Medicine", "date": "2026-01-16", "url": "https://pubmed.ncbi.nlm.nih.gov/41540424/", "doi": "10.1186/s12916-025-04597-8", "pmid": "41540424", "trialId": "NCT05164458", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Primary full phase 1 paper included 121 digestive-system cancer patients, 73 with pancreatic adenocarcinoma. It reports mature efficacy only for 27 gastric or gastro-oesophageal patients with CLDN18.2 staining of at least 2+ in 10% of cells, not the PDAC cohort. Across all patients, treatment-related events occurred in 99.2%, grade 3 or worse GGT elevation in 21.5%, and cytokine-release syndrome in 59.5%." }, { "id": "xu-2026-kras-g12v-tcr", "type": "paper", "title": "T cell receptor gene therapy targeting KRAS G12V for advanced pancreatic cancer in a single-arm phase 1/2 clinical trial", "authors": [ "Xu, Xiongfei", "Gu, Haihui", "Cha, Zhanshan", "et al." ], "publisher": "Molecular Therapy", "date": "2026-07-01", "url": "https://doi.org/10.1016/j.ymthe.2026.03.012", "doi": "10.1016/j.ymthe.2026.03.012", "pmid": null, "trialId": "NCT04146298", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Non-prespecified interim analysis of five men with KRAS G12V and HLA-A*11:01. One complete response lasted 5.5 months, two patients had short stable disease, median PFS was 14.3 weeks and OS 12.6 months. Three received repeat infusion; none responded to retreatment and two showed fast immune rejection associated with antibodies against the engineered receptor. Severe events were mainly blood-count effects of lymphocyte depletion." }, { "id": "yachida-2010-metastatic-timing", "type": "paper", "title": "Distant metastasis occurs late during the genetic evolution of pancreatic cancer", "authors": [ "Yachida, Shinichi", "Jones, Siân", "Bozic, Ivana", "et al." ], "publisher": "Nature", "date": "2010-10-27", "url": "https://www.nature.com/articles/nature09515", "doi": "10.1038/nature09515", "pmid": null, "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Human primary and metastasis sequencing from rapid-autopsy material plus a mathematical timing model. The long time estimates depend on a small selected cohort and modelling assumptions." }, { "id": "yan-2024-crs-hipec-pdac-retrospective", "type": "paper", "title": "Efficacy and safety of cytoreductive surgery combined with hyperthermic intraperitoneal chemotherapy in patients with pancreatic cancer peritoneal metastasis", "authors": [ "Yan, Guojun", "Zhang, Kai", "Yan, Lijun", "Zhang, Yanbin" ], "publisher": "World Journal of Surgical Oncology", "date": "2024-09-02", "url": "https://doi.org/10.1186/s12957-024-03464-9", "doi": "10.1186/s12957-024-03464-9", "pmid": "39218891", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Retrospective single-centre series of ten pancreatic-cancer patients with peritoneal metastases treated by cytoreductive surgery plus HIPEC from 2017 to 2023. Median survival was 24.2 months; two people required further surgery for severe complications. There was no comparator, no perioperative death and no identified prognostic factor." }, { "id": "yang-2025-gbcseeker", "type": "paper", "title": "Multimodal integration of liquid biopsy and radiology for the noninvasive diagnosis of gallbladder cancer and benign disorders", "authors": [ "Yang, Meng", "Zhao, Yuan", "Liu, Yingbin" ], "publisher": "Cancer Cell", "date": "2025-03-10", "url": "https://pubmed.ncbi.nlm.nih.gov/40068597/", "doi": "10.1016/j.ccell.2025.02.011", "pmid": "40068597", "trialId": "ChiCTR2100049249", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Full report and Table 2 checked. Prospective Chinese multicentre study enrolled 301 people with gallbladder-occupying lesions: discovery n=203 and external validation n=98. External GBCseeker performance was sensitivity 93.55%, specificity 77.78%, PPV 87.88%, NPV 87.50%, accuracy 87.76% and AUC 0.93. The external set contained 62 cancers and 36 benign lesions, so observed predictive values do not transport to low-prevalence screening or routine cholecystectomy. The model combines targeted cfDNA sequencing, CT radiomics, CA19-9 and sex and has not been tested in a randomized action study." }, { "id": "yao-2011-everolimus-pnet", "type": "paper", "title": "Everolimus for advanced pancreatic neuroendocrine tumors", "authors": [ "Yao, James C.", "Shah, Manisha H.", "Ito, Tetsuhide", "et al." ], "publisher": "New England Journal of Medicine", "date": "2011-02-10", "url": "https://pubmed.ncbi.nlm.nih.gov/21306238/", "doi": "10.1056/NEJMoa1009290", "pmid": "21306238", "trialId": "NCT00510068", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Randomised placebo-controlled phase 3 in 410 progressive low- or intermediate-grade advanced pNET patients. Crossover to open-label everolimus was allowed after progression. Funded by Novartis." }, { "id": "ygberg-2025-pancanai-validation", "type": "paper", "title": "Validation of a Pretrained Artificial Intelligence Model for Pancreatic Cancer Detection on Diagnosis and Prediagnosis Computed Tomography Scans", "authors": [ "Ygberg, et al." ], "publisher": "Investigative Radiology", "date": "2025-06-23", "url": "https://pubmed.ncbi.nlm.nih.gov/40553024/", "doi": "10.1097/RLI.0000000000001209", "pmid": "40553024", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Retrospective Danish cohort of 1,083 biopsy-confirmed pancreatic-cancer patients. Prediagnosis sensitivity was evaluated, but the cancer-only validation cannot estimate real-world specificity or positive predictive value." }, { "id": "yoo-2025-extrahepatic-cca-ctdna", "type": "paper", "title": "Circulating tumor DNA status and dynamics predict recurrence in patients with resected extrahepatic cholangiocarcinoma", "authors": [ "Yoo, Changhoon", "Jeong, Hansol", "Jeong, Joon Ho", "et al." ], "publisher": "Journal of Hepatology", "date": "2025-05-01", "url": "https://pubmed.ncbi.nlm.nih.gov/39532185/", "doi": "10.1016/j.jhep.2024.10.043", "pmid": "39532185", "trialId": "NCT03079427", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Prospectively collected biomarker cohort nested within the randomized STAMP adjuvant trial: 89 of 101 resected extrahepatic cholangiocarcinoma patients supplied 254 serial samples. Capecitabine and gemcitabine-cisplatin did not significantly differ in disease-free or overall survival. ctDNA positivity before adjuvant treatment was associated with disease-free survival, HR 1.8; at 12 weeks HR 7.72; at 24 weeks HR 5.24. The treatment comparison was randomized, but the ctDNA result did not assign treatment." }, { "id": "yoo-2026-saber", "type": "paper", "title": "Modified FOLFIRINOX with or Without Stereotactic Body Radiation in Locally Advanced Unresectable Pancreatic Cancer (SABER)", "authors": [ "Yoo, Changhoon", "Jeong, Hyehyun", "Park, Inkeun", "et al." ], "publisher": "Annals of Surgical Oncology", "date": "2026-08-01", "url": "https://pubmed.ncbi.nlm.nih.gov/42098550/", "doi": "10.1245/s10434-026-19762-3", "pmid": "42098550", "trialId": "NCT04986930", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Open-label randomized phase 2 trial. Only 37 of 92 planned patients enrolled; it stopped for slow accrual. PFS and OS comparisons were not significant." }, { "id": "yopp-2026-imbrave050-update", "type": "paper", "title": "Updated data from IMbrave050: Adjuvant atezolizumab plus bevacizumab for high-risk hepatocellular carcinoma", "authors": [ "Yopp, Adam", "Chen, Minshan", "Cheng, Ann-Lii", "et al." ], "publisher": "Journal of Hepatology", "date": "2026-01-22", "url": "https://pubmed.ncbi.nlm.nih.gov/41580093/", "doi": "10.1016/j.jhep.2026.01.006", "pmid": "41580093", "trialId": "NCT04102098", "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Longer follow-up of the 668-patient randomised IMbrave050 trial. The earlier significant recurrence-free survival result was not sustained. Roche/Genentech sponsored the programme and company employees were authors." }, { "id": "yu-2025-btc-ctdna-recurrence", "type": "paper", "title": "Detecting Early Recurrence With Circulating Tumor DNA in Stage I-III Biliary Tract Cancer After Curative Resection", "authors": [ "Yu, Jason", "He, A. Ruth", "Ouf, Mohamed", "et al." ], "publisher": "JCO Precision Oncology", "date": "2025-01-01", "url": "https://pubmed.ncbi.nlm.nih.gov/39772829/", "doi": "10.1200/PO-24-00443", "pmid": "39772829", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Retrospective multicentre real-world cohort of 56 resected stage I-III biliary tract cancers, median follow-up 12.8 months. Sixteen recurred; serial tumour-informed ctDNA identified 15 of 16, 93.8%, with average 3.7-month lead time. Positivity in the postoperative residual-disease window was associated with relapse-free survival, HR 26, 95% CI 2.6-265. The study did not randomize treatment or imaging triggered by ctDNA." }, { "id": "yu-2025-ibi343-pdac", "type": "conference-paper", "title": "CLDN18.2 expression and efficacy in pancreatic ductal adenocarcinoma: IBI343 phase 1 expansion", "authors": [ "Yu, Xianjun", "Ying, Jieer", "Li, Enxiao", "et al." ], "publisher": "Journal of Clinical Oncology", "date": "2025-05-28", "url": "https://ascopubs.org/doi/10.1200/JCO.2025.43.16_suppl.4017", "doi": "10.1200/JCO.2025.43.16_suppl.4017", "pmid": null, "trialId": "NCT05458219", "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Primary PDAC phase 1 abstract. At 6 mg/kg, 44 tumours with CLDN18.2 staining in at least 60% of cells had confirmed ORR 22.7%, PFS 5.4 months and OS 8.5 months; 12 below that expression boundary had zero responses, PFS 1.4 months and OS 6.2 months. Across all 83 treated patients, 50.6% had grade 3 or worse treatment-emergent events and 7.2% discontinued; no treatment-emergent event caused death." }, { "id": "zarate-rodriguez-2023-pancreatectomy-steps-rct", "type": "paper", "title": "Remote telephone intervention to increase preoperative activity before pancreatectomy", "authors": [ "Zarate Rodriguez, J G", "et al." ], "publisher": "Surgery in Practice and Science", "date": "2023-12-01", "url": "https://doi.org/10.1016/j.sipas.2023.100212", "doi": "10.1016/j.sipas.2023.100212", "pmid": "39844811", "trialId": null, "patentId": null, "grantId": "P50 CA196510", "accessed": "2026-09-15", "notes": "Randomized 152 patients scheduled for pancreatectomy, but 83 completed the study. Intervention completers increased from 4,568 to 5,071 daily steps, 11.0%, p=0.042. Severe complications were 22.9% versus 20.5%, p=0.807. More steps prove actuator response, not surgical benefit." }, { "id": "zhang-2004-hcc-screening", "type": "paper", "title": "Randomized controlled trial of screening for hepatocellular carcinoma", "authors": [ "Zhang, Bing-Hui", "Yang, Bing-Hui", "Tang, Zhao-You" ], "publisher": "Journal of Cancer Research and Clinical Oncology", "date": "2004-07-01", "url": "https://pubmed.ncbi.nlm.nih.gov/15042359/", "doi": "10.1007/s00432-004-0552-0", "pmid": "15042359", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Randomised screening trial in 18,816 people with hepatitis B or chronic hepatitis in urban Shanghai. Screening uptake was 58.2%; methods and applicability to modern diverse liver-disease populations require caution." }, { "id": "zhang-2017-evolutionary-adaptive-therapy", "type": "paper", "title": "Integrating evolutionary dynamics into treatment of metastatic castrate-resistant prostate cancer", "authors": [ "Zhang, Jingsong", "Cunningham, Jessica J.", "Brown, Joel S.", "Gatenby, Robert A." ], "publisher": "Nature Communications", "date": "2017-11-22", "url": "https://www.nature.com/articles/s41467-017-01968-5", "doi": "10.1038/s41467-017-01968-5", "pmid": "29180633", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Pilot of model-guided on/off abiraterone in 11 selected prostate-cancer patients compared with 16 contemporaneous controls. It reduced cumulative dose below half and prolonged observed control, but was nonrandomized, used PSA as a dense disease sensor and depended on competition between sensitive and resistant populations. Those conditions are not established in PDAC." }, { "id": "zhang-2022-uk-biobank-risk-fractions", "type": "paper", "title": "Risk Factors Associated with Pancreatic Cancer in the UK Biobank Cohort", "authors": [ "Zhang, et al." ], "publisher": "Cancers", "date": "2022-10-01", "url": "https://pubmed.ncbi.nlm.nih.gov/36291775/", "doi": "10.3390/cancers14204991", "pmid": "36291775", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Prospective UK Biobank association analysis with 728 pancreatic cancer cases. Population-attributable fractions assume causal, removable exposures and do not account fully for overlap; UK Biobank selection limits population transport." }, { "id": "zhang-2026-planet-hcc-recurrence", "type": "paper", "title": "Clonal diversity underpins distinct modes of recurrence in hepatocellular carcinoma: the PLANet cohort study", "authors": [ "Zhang, Ying", "Sekar, Karthik", "Phua, Cheryl Zi Jin", "et al." ], "publisher": "Gut", "date": "2026-07-21", "url": "https://pubmed.ncbi.nlm.nih.gov/42481381/", "doi": "10.1136/gutjnl-2026-338227", "pmid": "42481381", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-14", "notes": "Asia-Pacific prospective surgical cohort: multiregion whole-genome sequencing of 408 samples and RNA sequencing of 406 samples from 106 primary HCCs, with 24 paired recurrence cases. Recurrence-route treatment predictions remain observational." }, { "id": "zhou-2024-pancreatic-mwa", "type": "paper", "title": "Image-guided percutaneous microwave ablation for unresectable pancreatic cancers: A multicenter retrospective study", "authors": [ "Zhou, Jian", "Dong, Guang", "Jing, Xiaoyan", "et al." ], "publisher": "European Journal of Radiology", "date": "2024-09-23", "url": "https://pubmed.ncbi.nlm.nih.gov/39326234/", "doi": "10.1016/j.ejrad.2024.111720", "pmid": "39326234", "trialId": null, "patentId": null, "grantId": null, "accessed": "2026-09-15", "notes": "Four-hospital retrospective series of 72 people and 74 tumours. Complete ablation was 30/74, 40.5%; grade IIIa complications occurred in 7/72, 9.8%; median follow-up was 4.6 months and median survival 5.6 months. Pain relief was reported in 32/33 symptomatic people, with mean score 6.3 to 2.0, without a comparator." }, { "id": "zwicker-2019-isoquercetin-catiq", "type": "paper", "title": "Targeting protein disulfide isomerase with isoquercetin to improve hypercoagulability in advanced cancer", "authors": [ "Zwicker, Jeffrey I.", "Schlechter, Benjamin L.", "Stopa, Jack D.", "et al." ], "publisher": "JCI Insight", "date": "2019-02-21", "url": "https://insight.jci.org/articles/view/125851", "doi": "10.1172/jci.insight.125851", "pmid": "30652973", "trialId": "NCT02195232", "patentId": null, "grantId": "U54HL112302", "accessed": "2026-09-14", "notes": "Open-label sequential-dose phase 2 in 57 people with several advanced cancers, 28 at 500 mg and 29 at 1,000 mg for 56 days. D-dimer fell 21.9% at 1,000 mg and PDI, thrombin-generation and P-selectin measures changed. No primary VTE or major hemorrhage occurred, but there was no randomized or untreated control and the main efficacy endpoint was D-dimer. Quercegen Pharmaceuticals and NHLBI funded the study." } ] }