Clinicopathologic and molecular characteristics of colorectal cancer patients with brain metastases: a multicenter cohort study

Full abstract

Read the full abstract

Abstract Brain metastases (BM) from colorectal cancer (CRC) are uncommon but clinically consequential. We characterized clinicomolecular features of CRC patients with BM and evaluated the prognostic impact of BM timing. We screened CRC patients treated/followed between 2015 and 2024 and included those with radiologically confirmed BM. Primary tumor location was classified as rectum, left-sided colon or right-sided colon. BM timing was defined as de novo (≤ 90 days from CRC diagnosis) versus interval (> 90 days). Biomarkers included MSI status, RAS(KRAS/NRAS) mutations, BRAF and HER2. Overall survival (OS) was analyzed using Kaplan-Meier/log-rank and Cox regression. We screened 5,617 CRC patients and identified 94 (1.7%) with radiologically confirmed BM. Median age was 60 years; 60.6% were male. Primary tumors were rectal in 44.7%, left colon in 33.0%, and right colon in 22.3%, with left-sided/rectal primaries comprising 77.7% of cases. dMMR/MSI-H was observed in 6.2% (5/81), RAS mutations in 45.1% (37/82), BRAF mutations in 8.1% (6/74), and HER2 positivity in 10.8% (8/74). Median OS from CRC diagnosis was 36.1 months; median time from CRC diagnosis to BM was 21.8 months (IQR, 7.3–41.1), and median post-BM OS was 5.8 months. Interval BM was associated with longer OS than de novo BM (42.5 vs. 18.7 months; log-rank p < 0.001). In multivariable analysis adjusting for age, sex and ECOG performance status, de novo BM remained independently associated with worse OS (aHR 6.86; p < 0.001). CRC brain metastases occurred predominantly in left-sided colon and rectal primaries. HER2 positivity and BRAF mutations were identified among tested patients. BM timing provides clinically meaningful, independent prognostic stratification, and outcomes after BM remain poor. Prospective studies integrating standardized molecular profiling, treatment data, and longitudinal neuroimaging are warranted to refine risk stratification and risk-adapted surveillance.