GI Oncology Daily Digest

Immuneering presented updated survival data from its Phase 2a trial of atebimetinib — a 'deep cyclic inhibition' MEK inhibitor — combined with modified gemcitabine/nab-paclitaxel (mGnP) as first-line therapy in 55 patients with metastatic pancreatic ductal adenocarcinoma. As of the April 24, 2026 data cutoff, median overall survival was 17.3 months, roughly double the 8.5-month mOS seen with standard gemcitabine/nab-paclitaxel in the pivotal MPACT trial; median PFS was 8.3 months and the confirmed objective response rate was 36%. Notably, no Grade ≥3 treatment-related adverse event above 10% was attributable to atebimetinib itself — the high-grade toxicities (anemia, neutropenia) were chemotherapy-related — suggesting the MEK inhibitor adds efficacy without the rash/ocular toxicity that has historically limited MEK inhibition. The data were presented in an oral session at the 2026 ASCO Annual Meeting (Abstract 4013) by Peter Vu of UC San Diego Health. These are single-arm results benchmarked against a historical control, so the confirmatory Phase 3 MAPKeeper 301 trial (~510 patients, now enrolling) will be decisive — but a near-doubling of mOS in first-line metastatic PDAC is among the more provocative signals in a tumor that has resisted nearly every targeted approach.

In this prospective study from Moffitt Cancer Center's GI oncology group, 30 patients with metastatic, poorly differentiated extrapulmonary neuroendocrine carcinoma (NEC) underwent [68Ga]Ga-DOTATATE PET, with 28 also imaged by [18F]FDG PET. Primary tumor sites included pancreas (n=7) and colon/rectum (n=6), with 9 of unknown primary; histologies were small cell (n=15), large cell (n=6), and mixed/unspecified (n=9). Only 4 patients (13%) showed uniformly avid, Krenning grade 3-4 DOTATATE uptake indicating strong somatostatin receptor (SSTR) expression — far below the up-to-40% figures reported in earlier retrospective series, which the authors attribute to selection bias toward lower-Ki-67 tumors. The practical message is that routine SSTR-targeted imaging — and, by extension, peptide receptor radionuclide therapy (PRRT) — is unlikely to benefit most patients with unselected high-grade extrapulmonary NEC. It is a useful, appropriately deflationary data point for a setting where clinicians are often tempted to extrapolate well-differentiated NET biology onto aggressive NEC.

This prospective, 14-center European trial enrolled 209 patients (mean age 73) with unresectable malignant distal biliary obstruction — typically from pancreatic cancer or cholangiocarcinoma — in whom ERCP had failed or was not feasible, comparing EUS-guided biliary drainage (EUS-BD) against percutaneous transhepatic biliary drainage (PTBD), both with primary metal stenting and 1:1 propensity matching. Non-inferiority of PTBD for the primary endpoint of technical success could not be demonstrated (91.2% vs 97.1%). EUS-BD was significantly better on clinical success (intention-to-treat 73.5% vs 55.9%, p=0.05), procedure time, length of stay, and — most strikingly — 30-day biliary reintervention rate (0.03 vs 0.31 events per patient, p<0.001), an advantage driven mainly by EUS-guided choledochoduodenostomy (EUS-CDS). Adverse events, pain scores, and overall survival were similar. For GI oncologists managing biliary obstruction after failed ERCP, the data strengthen the case for EUS-BD — particularly EUS-CDS — over percutaneous drainage where the expertise exists.

HFE-related hereditary hemochromatosis — driven in 80-90% of cases by the homozygous C282Y variant — causes progressive hepatic iron accumulation that can culminate in fibrosis, cirrhosis, and hepatocellular carcinoma, yet has no causal therapy beyond phlebotomy. Researchers used lipid nanoparticles to deliver an adenine base editor (mRNA plus guide RNA) that corrects the HFE C282Y mutation in vivo. In iron-challenged mice, editing reached up to 67% of hepatocytes with no off-target effects detected at sites with one or two mismatches; treated animals showed significantly reduced hepatic iron overload despite continued high dietary iron, and transcriptomic signatures of fibrosis and cancer fell. The approach achieved up to 63.8% correction in hepatocyte-like cells derived from C282Y-homozygous patient iPSCs. This is a preclinical proof of concept — years from the clinic and requiring long-term safety and durability data — but it reframes a common iron-overload disorder, and one of the more preventable routes to HCC, as a potential one-time gene-correction target.

This prospective study across three Japanese centers tested 'generative chromoendoscopy' — a deep-learning model that converts standard white-light endoscopy images into synthetic indigo-carmine chromoendoscopy in real time, avoiding the need for actual dye spraying. Sixty patients with early gastric cancer or adenoma were imaged sequentially with white-light, generative chromoendoscopy, and real chromoendoscopy while two endoscopists scored lesion visibility on a 7-point scale. Generative chromoendoscopy improved visibility of pre-identified gastric neoplasms versus white light (mean difference 0.16; 95% CI 0.01-0.30) and was comparable to real chromoendoscopy; differentiated-type histology, active H. pylori infection, and expert-endoscopist status were each associated with larger gains. As a proof of concept the visibility increment is modest and the lesions were already known, so the clinically important question — whether synthetic dye improves real-time detection of lesions not yet seen — remains open, but it is an elegant demonstration of AI image enhancement that needs no extra hardware or dye.