Top 5 Papers
#1
Source: Journal of Hepatology | Authors: Rimini M, Ikeda M, Abidoye O, … Fornaro L, Rimassa L, Casadei-Gardini A (55 centres, 12 countries) | Published: September 26, 2026
Score: 10/20 — Base 6 (Journal of Hepatology — the corpus value applied on 2026-06-15, 2026-06-29 and 2026-07-26; the single 7 on 2026-06-17 is the outlier) + retrospective / real-world (+1) + survival benefit (+2 — OS HR 0.36 crude, 0.35 adjusted, holding across IPTW and landmark analyses) + biomarker-guided / precision (+1 — ESCAT tier-I matched therapy) = 10. No colleague-engagement bonus: this run's X sweep produced engagement almost entirely on papers that turned out to be duplicates.
This retrospective study pooled 1,358 patients with advanced biliary tract cancer treated with first-line cisplatin, gemcitabine and durvalumab across 55 centres in 12 countries. Molecular testing was performed in 1,072 patients and identified ESCAT tier-I actionable alterations in 238 (22.2%) — but only 76 of those 238 (32.0%) ever received a matched targeted agent. Among the 125 patients who received any active treatment after first-line progression, matched targeted therapy was associated with median OS of 24.8 vs 16.6 months (HR 0.36, 95% CI 0.20–0.63, p=0.0004), with an adjusted HR of 0.35 (95% CI 0.22–0.56, p<0.0001). The benefit was concentrated in second line (n=127: OS HR 0.50, 95% CI 0.30–0.84, p=0.0009; PFS 5.9 vs 3.2 months, HR 0.60, 95% CI 0.38–0.95, p=0.028; ORR 27.1% vs 9.1%, p=0.010), with no significant advantage demonstrated in third line. The design is retrospective and the authors themselves frame the results as practice-informing and hypothesis-generating rather than definitive.
Post angle: The hazard ratio is not the headline here — the access gap is. Two-thirds of patients with an ESCAT tier-I alteration never received the drug that matched it, and the survival advantage was front-loaded into second line, where it was largest, and had vanished by third. That combination argues for molecular profiling before first-line progression rather than after it, while patients are still fit enough to be treated. #GIOnc #Cholangiocarcinoma #PrecisionMedicine
#2
Source: The Lancet Oncology (Volume 27, Issue 10, pages 1237–1248) | Authors: Rumgay H, Georges D, Huang Y, Hirabayashi M, Shah R, de Martel C, … Clifford GM (IARC, Lyon) | Published: October 2026
Score: 10/20 — Base 8 (Lancet Oncology — the settled corpus value, applied on 2026-07-09, 2026-08-25 and 2026-08-31, the last explicitly citing the one before it; the 9 used earlier in 2026 has not been used since August) + population-based real-world incidence analysis (+1) + global prevention and health-policy impact (+1, on the 2026-05-05 Lancet Oncology precedent for a global-health policy bonus) = 10. Placed in the Top 5 despite being pan-tumor rather than GI-exclusive: the single largest attributable agent is a gastric carcinogen, and gastric, hepatocellular and biliary sites carry the bulk of the burden described. This is also the only Tier-1 paper in the run, so the Tier-1 safety-net rule applies regardless of placement.
Using GLOBOCAN 2024 incidence data, IARC estimated population-attributable fractions for the 12 infectious agents classified as Group 1 carcinogens. An estimated 2.3 million new cancer cases were attributable to infection globally in 2024 — 12% of all cancer cases. Helicobacter pylori is now the largest single agent at 760,000 cases (PAF 4%), ahead of HPV (750,000), HBV (360,000), EBV (260,000) and HCV (160,000). The GI detail matters: 700,000 of the H. pylori cases are non-cardia gastric cancer, 85% of that cancer worldwide, with a further 42,000 cardia gastric cancers and 18,000 gastric non-Hodgkin lymphomas; HBV accounts for 330,000 hepatocellular carcinomas (PAF 52%) and 27,000 intrahepatic cholangiocarcinomas (17%); liver flukes account for 3,800 intrahepatic cholangiocarcinomas. EBV-attributable gastric cancer is newly counted at 78,000 cases, part of why EBV has moved ahead of HCV. Eastern Asia carries 990,000 cases, 42% of the global total, at an ASIR of 31.9 per 100,000 against a global average of 22.7.
Post angle: Read as a GI paper, this is a prevention audit. Three of the five largest infectious carcinogens on the list drive gastric, hepatocellular and biliary cancer, and the top one — H. pylori — is a treatable bacterial infection for which, as the authors note, there has been little progress toward a vaccine or toward regimens that avoid ineffective antibiotic exposure. Japan, Korea and China screen endoscopically; Europe is piloting screen-and-treat. The burden is concentrated where the screening programmes mostly are not. #GIOnc #GastricCancer #HCC
#3
Source: Journal of Hepatology | Authors: John BV, Camardo M, Singal AG, Parikh ND, … Olson MC, Garces J, Limburg PJ | Published: September 26, 2026
Score: 9/20 — Base 6 (Journal of Hepatology, as above) + prospective multicentre registered head-to-head diagnostic study with prespecified co-primary endpoints, scored at the Phase II level (+2 — the bonus reflects design quality, not the result) + biomarker-guided / precision (+1, blood-based multi-target assay) = 9. Deliberately no clinical-impact bonus beyond that: the co-primary specificity endpoint was missed, so this is not scored as a positive trial. Several authors are Abbott employees and Abbott supplied the assay.
ALTUS prospectively enrolled 3,089 adults with cirrhosis or chronic HBV eligible for routine HCC screening across US centres between July 2021 and June 2025, with CT or MRI as the reference standard; 2,293 were evaluable with paired ultrasound and multi-target HCC blood test (mt-HBT) results, yielding 37 HCCs of which 27 were early-stage by Milan criteria. Early-stage sensitivity was 66.7% (95% CI 47.8–81.4) for mt-HBT vs 22.2% (95% CI 10.6–40.8) for ultrasound, meeting non-inferiority within the prespecified 5% margin, with a paired absolute difference of 44.4% (95% CI 20.2–63.8). Any-stage sensitivity was 73.0% for mt-HBT vs 51.4% for ultrasound plus AFP, and 29.7% for ultrasound alone. The second co-primary endpoint was not met: mt-HBT specificity was 81.8% (95% CI 80.1–83.4) against a prespecified target of 82%, compared with 98.6% for ultrasound. No serious adverse events occurred during blood collection.
Post angle: Ultrasound surveillance fails in exactly the patients we now see most — obesity and steatotic liver disease — and a blood draw is a test people will actually return for. Tripling early-stage sensitivity is a genuinely important number. But specificity of 81.8% means roughly one in five patients without cancer gets flagged, against 98.6% for ultrasound, and each of those flags becomes cross-sectional imaging, cost and anxiety. This is a strong signal that missed its own bar. #GIOnc #HCC #LiquidBiopsy
#4
Source: Diseases of the Colon and Rectum | Authors: Shiomi A, Akagi T, Kajikawa T, … Kanemitsu Y, Inomata M (JCOG, 42 Japanese institutions) | Published: September 23, 2026
Score: 9/20 — Base 5 (Diseases of the Colon and Rectum — new row, pinned at 'Other'; per the scoring rule that a specialty title does not earn a tier on the strength of one good paper, the study-type bonus is the correct lever) + Phase III randomised trial (+3) + colleague engagement (+1, one GI oncology KOL posted it within 72h) = 9. No clinical-impact bonus: PubMed's own article types list this as an Equivalence Trial, and the result is noninferiority rather than superiority, so under 'score the evidence, not the headline' no survival-benefit bonus applies.
JCOG1107 randomised 195 patients with symptomatic, noncurable stage IV colon cancer at 42 Japanese institutions to open (n=95) or laparoscopic (n=100) primary tumour resection, and now reports three-year follow-up at a median of 24.4 months. Three-year progression-free survival was 5.3% (95% CI 2.0–11.0) with open surgery vs 3.0% (95% CI 0.8–7.8) with laparoscopic surgery, HR 1.028 (95% CI 0.772–1.370), meeting the prespecified noninferiority criterion at p=0.02. Three-year overall survival was 31.5% (95% CI 22.5–41.0) vs 28.5% (95% CI 20.0–37.6), HR 1.048 (95% CI 0.780–1.410). Grade ≥3 late complications were 1.1% vs 2.0%. The very low PFS figures reflect a symptomatic, noncurable palliative population rather than poor trial conduct, and the authors note that chemotherapy was delivered without knowledge of MSI or RAS/BRAF status.
Post angle: This is a noninferiority result and should be read as one — laparoscopic resection did not beat open surgery, it failed to be worse. But in a population where median survival is under three years and the entire point of the operation is palliation, the operation that gets a patient out of hospital and back onto systemic therapy faster is the one that should be the default. Randomised evidence in palliative surgical oncology is rare enough to be worth reading closely. #GIOnc #CRC #ColorectalCancer
#5
Source: JHEP Reports | Authors: Chan LL, Kim HD, Cheon J, Chon HJ, Yoo C, Chan SL | Published: September 26, 2026
Score: 9/20 — Base 6 (JHEP Reports, matching the 2026-08-29 precedent and the 2026-09-20 edition that explicitly cited it) + pooled individual patient-data meta-analysis (+2) + biomarker-guided / precision (+1 — macrovascular invasion identified as the stratification variable that actually carries prognostic weight) = 9. No survival-benefit bonus: this is a prognostic pooled analysis of single-arm outcomes, not a comparison showing benefit.
This pooled individual patient-data analysis combined three prospective Asian trials of second-line cabozantinib, lenvatinib and regorafenib in hepatocellular carcinoma, totalling 117 patients, of whom 103 had received prior atezolizumab/bevacizumab-based first-line therapy. Median PFS was 4.2 months (95% CI 3.5–5.3) and median OS 10.6 months (95% CI 9.1–15.4). Macrovascular invasion was the only independent prognostic factor for overall survival, HR 3.55 (95% CI 1.89–6.67, p<0.001). Notably, prior response to immunotherapy, extrahepatic spread and baseline liver function were not associated with survival — leading the authors to argue that future second-line trials should stratify on macrovascular invasion rather than on prior immunotherapy response. The pooled sample is small and all three source trials were conducted in Asian populations.
Post angle: Second-line HCC trial design has quietly assumed that how a patient responded to first-line immunotherapy tells you something about what happens next. In this pooled dataset it told you nothing, while macrovascular invasion more than tripled the hazard of death on its own. Small and Asian-only, so not definitive — but it is a direct, testable challenge to how we stratify the next generation of second-line trials. #GIOnc #HCC #LiverCancer
Additional Papers of Interest
-
Cancer Treatment Reviews — 24 studies (15 randomised trials, 9 retrospective cohorts) comparing reduced with standard oxaliplatin exposure: every survival endpoint null, including adjuvant colorectal DFS HR 1.06 and OS HR 0.99, metastatic colorectal PFS HR 1.11 and OS HR 1.06, while grade ≥3 neuropathy fell sharply (OR 0.34, 95% CI 0.21–0.57) alongside grade 2 neuropathy (OR 0.46) and grade ≥3 adverse events overall (OR 0.67). The evidence is driven predominantly by colorectal cancer, with thinner pancreatic and gastroesophageal data
-
Endoscopy — 608 consecutive patients across 22 Japanese tertiary centres analysed with a marginal structural model using stabilised IPTW/IPCW: failure of intended lifelong drainage occurred in 130 (21.4%), and metal stent exposure was associated with higher odds of that failure (adjusted OR 1.87, 95% CI 1.28–2.73, p=0.001), consistent across sensitivity analyses and strengthening from OR 1.79 at 90 days to 2.93 at 450 days. Counterintuitive direction, retrospective design, and the authors present it as hypothesis-generating
Back to all digests