Molecularly defined subgroups within RAS-WT mCRC: Treatment selection and overall survival in Project GENIE BPC CRC Database.

Abstract

e15626 Background: RAS wild-type (WT) metastatic colorectal cancer (mCRC) remains clinically heterogeneous despite shared eligibility for anti-EGFR–based therapy. We evaluated whether clinically actionable molecular subsets within RAS-WT mCRC (BRAF V600E, MSI-H, and ERBB2 amplification) are associated with real-world treatment selection and overall survival (OS). Methods: Using the Project GENIE Biopharma Collaborative (BPC) CRC v2.0-public database cohort, we assembled a real-world RAS-WT mCRC population with regimen level survival fields, MSI testing, somatic mutations, and copy-number data. Molecular subsets were analyzed as mutually exclusive groups (BRAF-only, MSI-only, ERBB2-amplified–only), with overlaps described when present. ERBB2 amplification was derived from copy-number segment (.seg) data using the segment mean overlapping the ERBB2 locus (high-level amplification prespecified as log2 ≥0.9). Treatment class was categorized as anti-EGFR based (cetuximab/ panitumumab) or bevacizumab-based regimens. OS from regimen start was analyzed with penalized Cox models using robust variance clustered by patient, adjusting for age at sequencing, mutation count, and fraction genome altered. Because explicit primary tumor sidedness was unavailable, an OncoTree-based anatomic proxy (COAD vs READ vs other) was used in sensitivity analyses. Due to very small molecular subgroup counts within the anti-EGFR arm, interaction testing was underpowered and not interpretable. Results: We identified 471 RAS-WT episodes and 283 deaths. Groups: triple-neg 412; BRAF-only 49; MSI-only 4; ERBB2 amp-only 5 and overlap 1. BRAF-only episodes more often received bevacizumab than anti-EGFR (42 vs 8). Adjusted OS vs triple-neg: BRAF-only HR 1.61 (95%CI 1.05-2.46), p = 0.029; MSI-only HR 5.12 (3.02-8.66), p = 1.2×10⁻9 (likely reflecting poor outcomes on chemotherapy prior to immunotherapy access); ERBB2amp-only HR 1.21 (0.79-1.87), p = 0.38. MSI-only/ERBB2amp-only estimates are exploratory given very small N and incomplete capture of modern ICI/HER2-targeted sequencing. Treatment×molecular interaction was underpowered and not interpretable. Conclusions: In real-world RAS-WT mCRC, BRAF V600E identifies a clinically meaningful high-risk molecular subgroup associated with worse OS and preferential use of bevacizumab over anti-EGFR therapy. MSI-H and ERBB2-amplified subsets were rare in this cohort; larger line-of-therapy annotated datasets are needed to contextualize outcomes in the contemporary ICI/HER2-targeted era. Group Episodes Deaths Anti-EGFR Bev Adj. HR vs TN (95% CI); p Triple-neg 412 240 155 257 Ref BRAF-only 49 34 8 41 1.61 (1.05-2.46); 0.029 MSI-only† 4 4 0 4 5.12 (3.02-8.66); 1.2e-9 ERBB2amp-only† 5 4 2 3 1.21 (0.79-1.87); 0.38 †Estimates for MSI-only and ERBB2amp-only are exploratory due to limited sample size.