GI Oncology Daily Digest

ESMO has issued an Express Update — its out-of-cycle mechanism for data that should not wait for the next full guideline revision — writing daraxonrasib into the treatment of metastatic pancreatic cancer. The update is authored by Thierry Conroy and Michel Ducreux, and is indexed by PubMed as a Practice Guideline with the keywords KRAS inhibitors, daraxonrasib, metastatic pancreatic cancer, molecular profiling, and systemic treatment. The evidence base is RASolute 302, the first Phase 3 randomised trial of a RAS(ON) multi-selective inhibitor in any tumour, which reported median OS 13.2 versus 6.7 months against investigator's-choice chemotherapy, HR 0.40, p<0.0001 (NEJM; ASCO 2026 plenary LBA5) — covered in our June 2, 2026 edition. Note that PubMed carries no abstract for this Express Update and the full text was not retrievable, so the precise wording and strength of the recommendation, and the exact eligible population, are not reproduced here; the citation, journal, authorship and subject are verified from the PubMed record (PMID 42617733).

LiON is a contrast-enhanced-CT-based AI system for liver malignancy diagnosis, trained on 6,443 patients and validated across 22,251 patients with an AUC of 0.975 (95% CI 0.971–0.979). The prospective component is what distinguishes it from the usual retrospective AI paper: a single-arm trial of 10,333 patients in which LiON was deployed as an added reader in routine practice met its primary endpoint with an AUC of 0.952 (95% CI 0.942–0.961), clearing the prespecified lower confidence bound of 0.900. In practice-level terms, the system flagged 51 previously overlooked lesions, 15 of which were malignant, and triggered 37 amended radiology reports and 22 multidisciplinary team escalations. Registered as NCT07153783. The design limit is explicit: single-arm, with no randomised comparison against unaided radiologists, so the counterfactual detection rate is unknown.

The MHRA's Innovative Licensing and Access Pathway has awarded an Innovation Passport to BGB-B2033, a biological medicine for hepatocellular carcinoma developed by BeOne Medicines. ILAP is the UK's early-access designation, intended to compress the development-to-patient timeline for products addressing significant unmet need; the MHRA notice cites the continued poor prognosis of hepatocellular carcinoma as the rationale, and the developer states that a global, potentially registration-enabling trial is planned. This is a designation and not an approval: no efficacy or safety data have been disclosed, no trial results accompany the announcement, and the MHRA notice does not describe the product's mechanism of action. Trade coverage elsewhere ascribes a specific mechanism to BGB-B2033, but that claim is absent from the verified primary source and is therefore not repeated here.

Validated, prevalent cell-surface targets are the bottleneck holding back antibody-drug conjugates in intrahepatic cholangiocarcinoma, and gene-level analyses miss the tumour-enriched protein isoforms produced by alternative splicing. This MD Anderson group applied an isoform-resolved proteogenomic pipeline — transcriptomics, in silico translation, and cell surfaceomics — and identified a unique peptide corresponding to FGFR2b, the splicing isoform of FGFR2, showing markedly higher tumour enrichment than the alternative FGFR2c isoform. FGFR2b was the predominant FGFR2 isoform in biliary tract cancer including iCCA, at 88.6% in their institutional cohort and 88.2% in TCGA, with FGFR2 fusions in particular associated with high FGFR2b expression. Immunohistochemical validation in a subset of 20 patients confirmed membrane-bound FGFR2b in 31.6% of iCCA cases, including every patient with an FGFR2 fusion. High FGFR2b expression tracked with improved surgical outcomes, epithelial differentiation, and reduced CD8+ T-cell infiltration.

BRAF-mutant colorectal cancer arises from the serrated pathway and is aggressive, therapy-resistant, and — in its microsatellite-stable form — poorly understood mechanistically. Using multiple genetically engineered mouse models of BRAF-mutant MSS CRC plus organoid transplantation models, this Heidelberg group shows that WNT pathway activation via APC or CTNNB1 mutation was required for tumour initiation, while RNF43 loss was not sufficient — a clean negative that narrows the mechanism. WNT activation induced a molecular subtype shift and suppressed immune response pathways; mechanistically it suppressed CCL20 expression and remodelled the microenvironment by promoting immunosuppressive myeloid populations and altering T-cell states. Critically, WNT activation enhanced tumour progression specifically in immunocompetent settings, which is the evidence that immune evasion — not a cell-intrinsic growth advantage alone — is doing the work. Analyses used bulk RNA-seq, single-cell RNA-seq, and CITE-seq. Preclinical throughout: no patient outcomes.