Top 5 Papers
#1
Source: Nature Communications | Authors: Yoon HH, et al. (Xu J, senior author) | Published: Published online September 3, 2026; indexed in PubMed October 6, 2026
Score: 11/20 — Base 6 (Nature Communications) + Phase III RCT (+3) + survival benefit (+2, OS HR 0.70 in all randomised patients) = 11. No standard-of-care bonus: first-line PD-1 plus chemotherapy is already standard in ESCC, and this is an exploratory long-term follow-up of a trial whose primary OS result was reported earlier. Included now because PubMed indexed it this week; it has not appeared in a prior edition.
RATIONALE-306 randomised 649 patients with unresectable advanced or metastatic esophageal squamous cell carcinoma 1:1 to tislelizumab 200 mg or placebo, each with investigator-chosen chemotherapy, in the first-line setting. The primary OS benefit was reported at a minimum follow-up of about 15 months; this exploratory analysis extends minimum follow-up to 36.0 months. Overall survival remained improved in all patients, 17.2 versus 10.6 months (HR 0.70, 95% CI 0.59–0.80). At three years 15.0% of patients on tislelizumab plus chemotherapy had not progressed, against 2.9% on placebo plus chemotherapy. No new safety signals emerged, and median time to deterioration of physical functioning was not reached with tislelizumab versus 18.8 months with placebo.
Post angle: The number that matters is the tail: 15% progression-free at three years versus 3%. That is the durable-response subgroup PD-1 blockade was supposed to create in ESCC, and it is still there at 36 months. The next question is selection — who sits in that 15%. #GIOnc #EsophagealCancer #ESCC #Immunotherapy
#2
Source: Nature Communications | Authors: Qi C, et al. (Zhang C, senior author) | Published: Published online September 3, 2026; indexed in PubMed October 6, 2026
Score: 9/20 — Base 6 (Nature Communications) + ancillary biomarker analysis of a randomised Phase III (+2, serial plasma from the GLOW zolbetuximab trial alongside the CT041-CG4006 CAR-T trial) + biomarker-guided / precision (+1) = 9. Translational, with no survival endpoint of its own, so no survival bonus.
Claudin 18.2 is now a validated target in gastric and GEJ cancer, but there is no reliable way to predict who benefits durably from CLDN18.2-directed therapy. This study profiled extracellular vesicle (EV) proteins in serial plasma samples from patients in the CT041-CG4006 trial (CLDN18.2 CAR-T cells) and the GLOW trial (zolbetuximab). EV-associated CLDN18.2, PD-1 and PD-L2 tracked therapeutic benefit and were combined into a predictive signature, the CPP-score. Mechanistically, specific EV proteins exhausted CAR-T cells, neutralised CLDN18.2 antibodies, activated immunosuppressive IL-6-expressing cancer-associated fibroblasts and shifted the TH1/TH2 balance. In experimental models, adding anti-PD-1 or a CAF inhibitor, combining CLDN18.2 CAR-T with a CLDN18.2 antibody, or depleting plasma EVs counteracted that interference.
Post angle: Soluble target in the blood acting as a decoy is an old resistance story for antibodies — here it is shown for CLDN18.2, in patient plasma from the trials that matter, with a liquid-biopsy score attached. Hypothesis-generating, but it gives a reason to monitor plasma during zolbetuximab and CAR-T, and a rationale for PD-1 combinations. #GIOnc #GastricCancer #CLDN182 #CAR_T #LiquidBiopsy
#3
Source: JAMA Internal Medicine | Authors: Chua WY, et al. (Ng KYY, senior author) | Published: October 5, 2026
Score: 9/20 — Base 7 (JAMA Internal Medicine) + meta-analysis (+2) = 9. No clinical-impact bonus: the outcome is colonoscopy completion, not cancer incidence or survival, and heterogeneity is moderate to high for every intervention type before outlier exclusion.
A positive faecal immunochemical or occult blood test only prevents colorectal cancer deaths if a colonoscopy follows, and follow-up rates remain poor. This systematic review and random-effects meta-analysis included 21 studies, randomised and non-randomised, covering 19,613 patients aged 50–75 years. Patient navigation (RR 1.24, 95% CI 1.15–1.33), patient reminders (RR 1.32, 1.04–1.67), clinician reminders (RR 1.36, 1.12–1.65) and automated referral to a gastrointestinal consult (RR 1.68, 1.09–2.60) all increased colonoscopy completion. Pooled completion was 47.8% in control arms versus 69.1% after intervention (4,755 of 10,263 versus 6,272 of 9,350). Heterogeneity was moderate to high; excluding outliers brought it down for navigation and both reminder types, but not for automated referral.
Post angle: Half of patients with an abnormal FIT never get the colonoscopy — that is a screening programme failing at its last step. Simple, cheap system fixes push completion toward 70%. Automated referral had the biggest point estimate but the shakiest evidence. #GIOnc #CRC #CRCScreening #ColorectalCancer
#4
Source: Cancer Cell | Authors: Wang T, et al. (Wang B, senior author) | Published: October 6, 2026
Score: 8/20 — Base 7 (Cancer Cell) + biomarker-guided / precision (+1) = 8. No study-type bonus: a translational multi-omic atlas rather than a trial or clinical cohort, and the vonoprazan sensitisation is preclinical.
The authors built a chromatin accessibility atlas from 650 multistage samples taken from 439 patients, spanning the sequence from pre-cancerous mucosa to advanced gastric cancer. Intestinal metaplasia already carried a cancer-resembling 'epigenetic priming' state that appears to license later tumour initiation. Advanced tumours split into three epigenomic subtypes: proliferative and mesenchymal-inflammatory tumours were sensitive to neoadjuvant immunotherapy, while a 'pseudo-gastric' subtype was intrinsically resistant. That resistant subtype mimics gastric lineage, driven in part by FOSL2 up-regulating H+/K+ ATPase proton pumps to create an acidic microenvironment that suppresses CD8+ T cells. FOSL2 depletion or the clinically available acid blocker vonoprazan sensitised these tumours to immunotherapy in experimental models.
Post angle: A tumour that pumps acid to suppress T cells, and a cheap approved potassium-competitive acid blocker that undoes it in models. Preclinical — but the drug is already on the shelf, and that makes it a testable trial idea. #GIOnc #GastricCancer #Immunotherapy #Epigenetics
#5
Source: Journal of Hepatology | Authors: Chan SL, et al. (Rimassa L, senior author) | Published: October 6, 2026
Score: 8/20 — Base 6 (Journal of Hepatology) + single-arm Phase IIIb (+2, scored at the Phase II level because there is no randomisation or comparator) = 8. No survival bonus: the co-primary endpoints are safety and ORR, with no survival comparison.
Patients with poor prognostic features — Child-Pugh B liver function, ECOG performance status 2, or main-trunk portal vein thrombosis (Vp4) — are routinely excluded from phase III HCC trials, despite being, by the authors' estimate, up to 65% of unresectable HCC in practice. SIERRA (NCT05883644) gave the STRIDE regimen (tremelimumab 300 mg once plus durvalumab 1500 mg every 4 weeks) to 111 such patients: 44 Child-Pugh B7/B8, 47 ECOG PS 2 and 20 Vp4. Grade 3/4 treatment-related adverse events within 6 months occurred in 19.8% (95% CI 12.9–28.5), from 18.2% in the Child-Pugh B cohort to 25.0% in Vp4. ORR was 14.4% (8.5–22.4): 21.3% in ECOG PS 2, 15.0% in Vp4 and 6.8% in Child-Pugh B7/B8. The authors describe it as the first prospective interventional study of STRIDE in this population.
Post angle: This is the patient in clinic every week who was never in HIMALAYA. Toxicity looks manageable across all three cohorts, but the Child-Pugh B response rate of 6.8% is the sobering line — liver reserve still limits what immunotherapy can do. #GIOnc #HCC #LiverCancer #Immunotherapy
Additional Papers of Interest
-
Gut — random-forest model built on 11 routine clinical predictors in 330,767 people with type 2 diabetes (Tianjin) and externally validated in 105,537 (Xiamen). Six-month AUC 0.86 externally; the top 5% risk group captured 59.4% and 48.0% of incident cancers in the two validation cohorts. Unlike most tools it works in long-standing diabetes, not only new-onset. Retrospective; needs prospective validation.
-
International Journal of Cancer — 23 patients with T4 locally advanced dMMR/MSI-H gastric or EGJ adenocarcinoma given neoadjuvant PD-1/PD-L1 plus CTLA-4 blockade. Of 16 who had surgery, all were R0, with pCR 81.3% (95% CI 57.0–93.4%) and MPR 93.8%; five of the seven unoperated patients chose watch-and-wait after complete clinical response. EFS 95.5% at a median 22.1 months. Small and retrospective, but T4 disease is where the existing trials have least data.
-
Endoscopy — Korean nationwide cohort of 1,038 patients aged ≥75 after eCuraC-2 endoscopic resection for early gastric cancer (352 gastrectomy, 686 surveillance). Five-year OS 83.0% vs 76.5% favoured surgery (HR 1.37 for surveillance), but the surveillance group was older and frailer, and gastric-cancer deaths were rare (17 events). Recurrence was higher with surveillance (9.9% vs 1.8%). Competing mortality drives prognosis — individualise.
-
Annals of Surgical Oncology — three-centre retrospective cohort of 695 patients with weighting for baseline balance. RAMPS was associated with longer RFS (adjusted HR 0.46, 95% CI 0.38–0.55) and improved OS, most markedly with splenic artery (HR 0.24) or vein (HR 0.35) involvement. The KRAS/TP53 interactions did not survive FDR correction and are exploratory.
-
Nature Communications — the RCSA model combines preoperative CT radiomics, H&E pathomics and clinical features to flag locally advanced gastric cancer patients at risk of postoperative liver metastasis, with AUCs of 0.862–0.909 across internal, external, public and prospective-trial validation sets. Low-risk tumours carried a more active immune environment.
-
British Journal of Surgery — Dutch nationwide longitudinal study of 1,009 patients and 2,995 questionnaires. GI symptoms were present in 95% before surgery and peaked at 0–3 months after; severe symptoms fell only from 61% before surgery to 52% at 9–12 months, and two or more severe symptoms meant clinically relevant HRQoL loss. Supportive care after pancreatectomy needs to start earlier and last longer.
Back to all digests