Top 5 Papers
#1
Source: eClinicalMedicine | Authors: Ueno M, Yoo C, Ren Z, et al. (KEYNOTE-966 investigators) | Published: October 12, 2026 (published online, ahead of PubMed indexing)
Score: 10/20 — eClinicalMedicine (base 5; Other = 5, matching the two prior editions that scored it, 2026-04-24 and 2026-09-15) + randomised Phase III (+3) + overall survival benefit (+2) = 10. No new-standard-of-care bonus: this is an updated, descriptive analysis of a regimen already approved and in guidelines, and no formal testing was done. No colleague-engagement bonus: no GI oncology KOL posted this paper within the window.
KEYNOTE-966 randomised 1,069 patients with previously untreated advanced biliary tract cancer to pembrolizumab or placebo plus gemcitabine and cisplatin. This update reports a median follow-up of 36.6 months, 11 months beyond the final analysis. The OS benefit held: HR 0.86 (95% CI 0.75–0.98), median OS 12.7 vs 10.9 months, 24-month OS 25% vs 19%, and 36-month OS 13% vs 11%. PFS also favoured pembrolizumab (HR 0.85, 95% CI 0.75–0.97). ORR was identical at 29% in both arms, but responses lasted longer with pembrolizumab: median duration 8.3 vs 6.9 months, and 18% vs 7% of responders were still responding at 24 months. Grade 3–4 treatment-related adverse events were 70% vs 69%, with no new treatment-related deaths. Two caveats: the analysis is descriptive, since no alpha was allocated, and the data cutoff was November 2023.
Post angle: KEYNOTE-966 at three years: pembrolizumab + gem/cis holds an OS HR of 0.86 in advanced BTC. Response rates were identical, so the benefit is in how long responses last, not how many patients respond. The curves narrow by 36 months (13% vs 11%), so the real question is who the long-term survivors are. #GIOnc #BTC #Cholangiocarcinoma #Immunotherapy
#2
Source: Journal of Immunotherapy | Authors: Wang Y, Zhao R, Wan QY, … Chen J | Published: October 12, 2026
Score: 7/20 — Journal of Immunotherapy (base 5; not in the scoring table, Other = 5) + retrospective (+1) + biomarker-guided (+1, a dMMR/MSI-H-selected population) = 7. No survival bonus, because the endpoints are pathological response only. No colleague-engagement bonus.
This retrospective study included 39 patients with locally advanced, resectable dMMR/MSI-H gastric or GEJ adenocarcinoma who received a neoadjuvant immunotherapy-based regimen. All 39 had R0 resection, and 22 (56%) achieved pCR in both the primary tumour and the lymph nodes. Chemotherapy plus immunotherapy and immunotherapy alone each produced significantly more TRG1 responses than chemotherapy alone: RR 4.2 (95% CI 1.5–15.0, P=0.0004) and RR 4.4 (95% CI 1.7–15.7, P=0.003). Intestinal-type histology had a higher TRG1 rate than mixed-type (RR 2.5, 95% CI 1.1–8.7, P=0.0196). The cohort is small and the comparisons are non-randomised, but it adds to a growing body of evidence: perioperative pembrolizumab in IMHOTEP (covered here on 10 September) and a neoadjuvant dual-checkpoint T4 cohort with 81% pCR (covered 7 October).
Post angle: dMMR/MSI-H gastric cancer keeps telling the same story: in 39 patients, neoadjuvant IO gave 100% R0 and 56% pCR, and IO-containing regimens roughly quadrupled TRG1 response over chemotherapy alone. New here: intestinal-type histology responded best. Test MMR status before planning surgery. #GIOnc #GastricCancer #dMMR #MSI
#3
Source: Clinical Colorectal Cancer | Authors: Simmet V, Lièvre A, Samalin E, … Bennouna J (UNICANCER) | Published: September 18, 2026 (indexed in PubMed October 2026)
Score: 7/20 — Clinical Colorectal Cancer (base 5; not in the scoring table, Other = 5) + Phase II (+2) = 7. No impact bonus: the primary endpoint was not met. No colleague-engagement bonus.
This multicentre, open-label, non-comparative randomised Phase II trial assigned 75 patients with chemorefractory RAS wild-type mCRC 2:1 to afatinib 40 mg/day plus cetuximab (n=51) or to cetuximab alone (n=24) as an internal control. Crossover to afatinib was allowed at progression. The primary endpoint was not met: only 9 patients (18%, 95% CI 8–31) in the combination arm were progression-free at 6 months, against a minimum success rate of 30%. In the combination arm, ORR was 26%, median PFS 4.1 months and median OS 13.6 months. Grade ≥3 toxicity occurred in 76% vs 64% of patients, with two grade 5 events in the combination arm (sudden death and hypoglycaemia). High baseline circulating tumour cell (CTC) counts and a fall in CTCs by day 14 correlated with OS and ORR.
Post angle: Two EGFR blockers are not better than one: afatinib + cetuximab missed its primary endpoint in chemorefractory RAS-WT mCRC (6-month PFS 18% vs a 30% bar), with more grade ≥3 toxicity and two grade 5 events. The day-14 CTC drop is worth carrying forward as an early-response read. #GIOnc #CRC #mCRC #EGFR
#4
Source: eClinicalMedicine | Authors: Tranchart H, Maitre S, Girard E, et al. (six French hepatobiliary centres) | Published: October 12, 2026 (published online, ahead of PubMed indexing)
Score: 7/20 — eClinicalMedicine (base 5; Other = 5, consistent with prior editions) + Phase II (+2) = 7. No impact bonus: single-arm, with a volumetric primary endpoint. No colleague-engagement bonus.
Conventional portal vein embolization (PVE) occludes the portal branch permanently, which can complicate later treatment for patients who never reach resection. EMBORES, a prospective single-arm Phase II trial in six French centres (2018–2021), tested selective PVE with powdered absorbable gelatin sponge. Of 34 patients enrolled, 32 underwent temporary PVE, and 18 of those 32 (56%) had colorectal liver metastases. Technical success was 100%, with one major complication (3.1%). The portal vein recanalized completely in 17 patients (53%) and partially in 9 (28%). Future liver remnant volume rose from a mean of 580 to 690 mL (mean hypertrophy ratio 9.5%, SD 10.2), but 5 patients needed repeat embolization with a permanent agent. Overall, 23 of 32 (71.8%) underwent major hepatectomy, with no postoperative mortality.
Post angle: Can you grow the liver remnant without blocking the portal vein permanently? EMBORES says partly: absorbable-gelatin PVE was safe, most veins recanalized, the remnant grew 9.5% on average and 72% reached major hepatectomy. But 5 of 32 needed permanent re-embolization, so this is an option for selected patients, not a replacement. #GIOnc #CRLM #HPB #LiverSurgery
#5
Source: International Journal of Cancer | Authors: Sun W, Liu D, Wang D, … Ji J | Published: October 12, 2026
Score: 7/20 — International Journal of Cancer (base 5; Other = 5, matching the 2026-08-12 edition) + cohort / real-world (+1) + biomarker-guided / precision (+1, polygenic-risk-stratified screening) = 7. No colleague-engagement bonus.
In 353,118 average-risk UK Biobank participants, the authors used family-history-stratified Cox models to estimate how sex and a CRC polygenic risk score (PRS) affect CRC incidence, metastasis at diagnosis and CRC-specific mortality. Men had higher risk than women both without a family history (HR 1.41, 95% CI 1.33–1.50) and with one (HR 1.42, 95% CI 1.26–1.61). Each 1-SD increase in PRS raised risk by a factor of 1.49 (95% CI 1.45–1.53) without a family history and 1.42 (95% CI 1.34–1.51) with one. Higher PRS was also linked to metastatic disease at diagnosis and to CRC-specific mortality, which the authors read as later diagnosis in high-PRS individuals. Converted into risk-adapted screening start ages, the range ran from before 40 for high-PRS men with a family history to after 60 for low-PRS women without one.
Post angle: A single screening start age for everyone ignores a lot of measurable risk. In 353K UK Biobank participants, combining polygenic risk, sex and family history moved the risk-equivalent start age from before 40 to after 60. The open question is how to deliver it: PRS is not a routine test yet. #GIOnc #CRC #EarlyOnsetCRC #Screening
Additional Papers of Interest
-
Nature Medicine — Tier-1, pan-cancer with direct GI relevance. Using longitudinal clinicogenomic data from 52,236 patients sequenced with MSK-IMPACT, the authors mapped episodes of cachectic weight loss. These episodes clustered near disease progression and were associated with worse outcomes in nearly all cancer types. They carried a consistent serologic signature (lower albumin and haemoglobin, higher liver enzymes) and were linked to tumour genotypes, including TP53 mutation. A model predicting impending cachexia in lung, colorectal and prostate cancer was validated externally at Dana-Farber (n=5,785).
-
eClinicalMedicine (published online September 22, 2026) — Patients from eight Chinese centres, analysed with clone-censor-weighting to address immortal-time bias. Deferring RT was associated with longer median OS (44.1 vs 33.4 months, adjusted HR 0.699, 95% CI 0.580–0.843) and PFS (adjusted HR 0.760), and with less severe oesophageal fistula (4.8% vs 9.0%). Caveat: the deferred group's RT dose IQR runs from 0 Gy, so some patients never received RT, and the authors call the results hypothesis-generating.
-
Annals of Surgical Oncology — Prospective single-arm study (ChiCTR2200063680) in 30 patients. 24 of 30 (80%) completed neoadjuvant therapy and underwent R0 resection. 12-month OS was 87.5% and RFS 82.8%, median RFS 28 months, and most treatment-related adverse events were grade 1–2. Proteomics nominated NT5DC2, FAM3D, CXCL5 and TMT1B as candidate resistance markers. Small and uncontrolled.
-
Endoscopy — Single-centre cohort of 331 patients with grade C Lugol-voiding lesions after ESD, median follow-up 57 months. 49.2% developed metachronous multiple ESCCs (5-year cumulative incidence 55.2%). ≥31 lesions (HR 5.34 vs 10–20) and a positive alcohol flushing response (HR 1.59) independently raised risk, which supports risk-adapted surveillance intensity.
-
Journal of Surgical Oncology — 161 patients resected at a Utrecht centre, 2016–2020, median follow-up 51 months. Median OS was 29 months with (m)FOLFIRINOX, 21 with gemcitabine monotherapy and 28 with gemcitabine combinations. Head-to-head, (m)FOLFIRINOX was not significantly better than gemcitabine monotherapy (HR 0.62, 95% CI 0.33–1.17) or gemcitabine combinations (HR 0.66, 95% CI 0.35–1.24), even in trial-eligible patients. Small cohort, wide confidence intervals.
Back to all digests