Are all tumor-agnostic biomarkers equally tissue-independent? Comparative cross-histology efficacy analysis of MSI-H/dMMR versus <i>BRAF</i> V600E.

Abstract

2649 Background: FDA tumor-agnostic approvals for pembrolizumab/dostarlimab (MSI-H/dMMR) and dabrafenib-trametinib (BRAF V600E) assume uniform cross-histology efficacy. However, MSI-H/dMMR depends on immune checkpoint blockade—potentially influenced by tissue-specific microenvironments—while BRAF V600E represents oncogene addiction with essential MAPK dependencies. We hypothesized these biomarkers differ in tissue independence, with MSI-H/dMMR demonstrating greater histology-dependent response variability. Methods: A systematic review and meta-analysis were conducted in accordance with PRISMA 2020 guidelines. PubMed, Embase, and the Cochrane Central Register owere searched from inception to January 2026 to identify prospective basket trials enrolling adults (≥18 years) with advanced or metastatic solid tumors harboring FDA-approved tumor-agnostic biomarkers. Eligible trials evaluated pembrolizumab or dostarlimab in MSI-H/dMMR tumors, or dabrafenib plus trametinib in BRAF V600E-mutant tumors, and reported objective response rates (ORR) stratified by tumor histology. Studies were required to include ≥3 distinct tumor histologies (not subtypes) with a minimum of 10 patients per histology cohort. Phase I trials, retrospective case series, pediatric-only populations, and studies without histology-specific efficacy data were excluded. Primary outcome was ORR heterogeneity across tumor types within each biomarker platform, quantified using the I² statistic and coefficient of variation (CV). Pooled analyses were performed using random-effects models. Effect estimates are expressed as ORR percentages with 95% confidence intervals. Results: Four trials included: KEYNOTE-158 (pembrolizumab, n=373), GARNET (dostarlimab, n=363), ROAR (dabrafenib-trametinib, n=215), VE-BASKET (vemurafenib, n=62); 1,013 patients across 12 tumor histologies. MSI-H/dMMR platform (n=736, 8 tumor types): mean ORR 35.1% (range 0-57.1%); endometrial 57.1%, gastric 45.8%, colorectal 43.5%, small intestine 42.1%, cholangiocarcinoma 40.9%, ovarian 33.3%, pancreatic 18.2%, brain 0%; CV=0.513. BRAF V600E platform (n=277, 4 tumor types): mean ORR 45.0% (range 37.1-53%); anaplastic thyroid 53%, biliary tract 47%, glioma 42.9%, NSCLC 37.1%; CV=0.149. MSI-H/dMMR demonstrated 3.4-fold greater heterogeneity than BRAF V600E (CV 0.513 vs 0.149). Conclusions: Tumor-agnostic biomarkers exhibit differential tissue independence. MSI-H/dMMR showed 3.4-fold greater response heterogeneity (CV=0.513) than BRAF V600E (CV=0.149), indicating immune checkpoint efficacy remains tissue-dependent while oncogene addiction confers uniform predictivity. These findings challenge the assumption of biomarker-driven tissue-agnostic efficacy and necessitate histology-specific outcome counseling and trial stratification.