GI Oncology Daily Digest

August 15, 2026 — Two-Paper Edition — When the Trial Misses but the Tissue Talks: Biliary Immunotherapy and Pancreatic Adenosquamous Biology
Curated by Dr. Allan Pereira — Moffitt Cancer Center

Top 5 Papers

#1
Source: Clinical Cancer Research  |  Authors: Kelley RK, Bracci P, Gordan JD, … Fong L, Keenan BP (senior) — University of California, San Francisco  |  Published: August 14, 2026
Score: 9/20 — Base 6 (Clinical Cancer Research) + 2 (Phase II trial) + 1 (biomarker-guided — PD-L1 stratification and paired on-treatment biopsies) = 9. No clinical-impact bonus: the trial did not meet its pre-specified response threshold and reports no survival benefit.
Checkpoint inhibitors have never done much as monotherapy in biliary tract cancer, and the rationale for adding GM-CSF came from melanoma, where combining it with ipilimumab extended survival. This single-centre phase II gave pembrolizumab 200 mg every 21 days alongside two cycles of GM-CSF 250 µg subcutaneously on days 1–14, in 42 patients who had received prior chemotherapy but no prior checkpoint inhibitor; 67% had intrahepatic cholangiocarcinoma, 90% had stage IV disease, and 24% had underlying viral hepatitis. The confirmed objective response rate was 12% (95% CI, 4–26), which fell short of the pre-specified efficacy bar, though two patients achieved complete response and 26% remained progression-free at six months. Tolerability was good, with treatment-related grade 3–4 events in 7% and treatment-related serious adverse events in 10%. The correlative work is where the paper earns its place: tumour PD-L1 expression was present in 46% of patients and tracked with a higher six-month progression-free rate, and paired biopsies taken before and after GM-CSF showed upregulation of CD8+ T-cell populations and antigen-processing pathways. The combination did not work as designed, but it demonstrably moved the tumour immune microenvironment in the intended direction.
Post angle: A negative phase II that earned its correlative science. GM-CSF did wake up CD8+ T cells and antigen processing in biliary tumours — it just did not convert that into responses in an unselected population. The PD-L1 signal is the thread worth pulling. #Cholangiocarcinoma #GIOnc #Immunotherapy
#2
Source: Clinical Cancer Research  |  Authors: Haldar SD, Wetzel M, Lu J, … Jaffee EM, Fertig EJ, Azad NS (senior) — Johns Hopkins, MD Anderson and Foundation Medicine, multicentre  |  Published: August 14, 2026
Score: 8/20 — Base 6 (Clinical Cancer Research) + 1 (retrospective / real-world clinical cohort) + 1 (biomarker-guided — genomic and spatial subtyping with direct treatment-selection implications) = 8.
Pancreatic adenosquamous carcinoma is a rare variant that most centres treat as though it were conventional pancreatic ductal adenocarcinoma, largely because nobody has characterised it well enough to argue otherwise. This study assembles three layers of evidence to test that assumption: clinicopathologic features, treatment patterns and survival for 178 patients treated at Johns Hopkins between 2013 and 2023; a genomic comparison of 244 adenosquamous tumours against 29,021 conventional PDAC cases in the FoundationCore database; and Visium spatial transcriptomics on 11 resected adenosquamous tumours from 8 patients, benchmarked against a four-sample PDAC comparator. Outcomes were poor in both resected and advanced disease. Genomically, KRAS mutation frequency was similar to PDAC, but MTAP loss was more common and the low-frequency immune biomarkers — MSI-high, PD-L1 positivity, TMB ≥10 mutations/Mb — were all enriched. Spatial profiling found broadly comparable immune cell proportions but higher regulatory T-cell abundance, and distinct co-localisation patterns: immune infiltrates sat closest to squamous regions and farthest from glandular ones. Adenosquamous tumours showed increased squamous/basal lineage expression overall, with intratumoral squamous regions enriched for epithelial-mesenchymal transition, apical junction, inflammatory and stress-response pathways relative to the glandular niche. Shared drivers, genuinely different disease.
Post angle: We have been enrolling pancreatic adenosquamous carcinoma into PDAC trials and calling it a subgroup. This says the driver is shared but the biology is not — higher MTAP loss, enriched MSI-high/PD-L1+/TMB-high, and an immune infiltrate that hugs the squamous compartment. Subtype-specific strategy, not a footnote. #PDAC #PancreaticCancer #GIOnc

Additional Papers of Interest

  1. Journal of Gastrointestinal Cancer — Thirteen studies and 2,447 patients pooled to a validation AUC of 0.82 (sensitivity 0.79, specificity 0.72) for predicting MSI status from imaging alone; adding clinical variables lifted specificity to 0.75 from 0.68. A non-invasive route to the biomarker that gates immunotherapy in gastric cancer, still awaiting prospective validation.
  2. International Journal of Hyperthermia — Seven studies and 1,469 patients (544 cryoablation, 925 RFA). Cryoablation delivered a higher initial complete ablation rate and less intraoperative pain, but progression-free survival, overall survival and complication rates were comparable. The familiar pattern where a technical advantage does not translate into a survival one.
  3. Journal of Hepatology — An editorial on why a substantial share of hepatocellular carcinomas never respond to first-line atezolizumab/bevacizumab at all, and the case for the myeloid compartment rather than T-cell exhaustion as the operative barrier.
Back to all digests