GI Oncology Daily Digest

The STELLAR trial was one of the first randomized studies to test whether short-course radiotherapy (SCRT) followed by consolidation chemotherapy — a total neoadjuvant therapy (TNT) strategy — could match or beat standard long-course chemoradiotherapy (CRT) in locally advanced rectal cancer (LARC), and this update reports the durable 5-year picture. Patients with distal or middle-third LARC were randomized to SCRT-based TNT or long-course CRT, and at a median follow-up of 68.7 months the survival signal that emerged early has held: 5-year overall survival was significantly higher with TNT at 78.1% versus 69.7% (HR 0.739, 95% CI 0.550–0.993). The co-primary disease-free survival endpoint, however, did not reach significance (62.0% vs 58.7%; HR 0.849, 95% CI 0.662–1.089), and distant metastasis and locoregional recurrence rates were similar between arms — so the OS advantage, not local control or DFS, is the story. The benefit concentrated in higher-risk patients: among those meeting ESMO high-risk criteria, TNT improved OS (HR 0.663, 95% CI 0.469–0.937) with a nonsignificant DFS trend (HR 0.765). Notably, even among patients who did recur, TNT was associated with better postrecurrence progression-free survival (HR 0.691) and postrecurrence survival (HR 0.698), hinting that upfront systemic intensification changes the biology of relapse. The authors position SCRT-based TNT as a durable, viable alternative to long-course CRT, particularly for high-risk disease — a practically important message given the shorter radiotherapy course and lower resource burden of the SCRT backbone.

AstraZeneca announced positive topline results from CLARITY-Gastric01, the first Phase III trial to show an overall survival benefit with an anti-CLDN18.2 antibody-drug conjugate — a milestone for a target that has been validated (via the monoclonal antibody zolbetuximab) but not previously with an ADC. Sonesitatug vedotin (Sone-Ve), which pairs an anti-CLDN18.2 antibody with a protease-cleavable linker and an MMAE cytotoxic payload, was tested against investigator's choice of therapy in second- and later-line locally advanced or metastatic gastric, gastroesophageal junction (GEJ), or esophageal adenocarcinoma with CLDN18.2 expression on at least 25% of tumor cells at any intensity. The trial met its dual primary endpoint of overall survival in the 3rd-and-later-line setting — described as a statistically significant and highly clinically meaningful improvement — and also met a key secondary endpoint of overall survival in the overall 2L+ population. Importantly, the other component of the dual primary endpoint, progression-free survival, did not reach statistical significance, so this is an OS-driven result. No quantitative data (hazard ratios, median OS) have been released yet; AstraZeneca states the full results will be presented at a forthcoming medical meeting. This is a topline, verification-limited entry: the trial name, drug, mechanism, population, and endpoints met are confirmed against the company's press release and independent reporting, but the effect size is not yet public. If the magnitude holds up when the numbers land, Sone-Ve could become a new 2L+ option in CLDN18.2-positive gastric cancer and the first of AstraZeneca's wholly owned ADCs to reach a pivotal survival readout.

The choice between adjuvant gemcitabine (GEM) and modified FOLFIRINOX (mFOLFIRINOX) after resection of pancreatic ductal adenocarcinoma (PDAC) is today driven mostly by performance status, not tumor biology — mFOLFIRINOX is more effective but more toxic, and there is no validated tool to identify who truly needs it. This study asked whether tumor morphology itself, read by deep learning on standard whole-slide histology images, could predict differential benefit. The investigators trained regimen-specific histology models on disease-free survival in a retrospective multicenter series of 231 resected patients (54 who received mFOLFIRINOX, 177 GEM), then combined them into PANCprAId, a biomarker estimating a patient's relative benefit from GEM versus mFOLFIRINOX. Crucially, they externally validated it in the tissue of the randomized PRODIGE-24/CCTG PA6 trial (n=313) — the same trial that established adjuvant mFOLFIRINOX as a standard. There, the treatment-specific histology scores stratified outcomes within each arm (GEM HR 1.69, 95% CI 1.04–2.73, p=.03; mFOLFIRINOX HR 2.02, 95% CI 1.4–3.0, p<.001), and when combined, PANCprAId identified subgroups with differential relative benefit, with statistically significant treatment interactions for DFS (p=.003) and cancer-specific survival (p=.001). Predicted sensitivity to each regimen mapped onto distinct epithelial and stromal features. This is a genuinely predictive (not merely prognostic) signal validated in randomized-trial tissue — a meaningful proof of concept that pathology-based AI could one day help choose between adjuvant regimens in PDAC. It is not clinic-ready: the development set is modest, it needs prospective and broader external validation, and the workflow must be productized. But it points at a future where the H&E slide already sitting in the pathology archive helps triage adjuvant therapy.