Genomic alterations predictive of poor clinical outcomes in pan-cancer.

ABSTRACT

BACKGROUND: Genomic alterations are highly frequent across cancers, but their prognostic impact is not well characterized in pan-cancer cohorts. Here, we use pan-cancer cohorts from TCGA and MSK-IMPACT to evaluate the associations of common genomic alterations with poor clinical outcome. MATERIALS AND METHODS: Genomic alterations in commonly altered genes were extracted from Pan-Cancer TCGA and MSK-IMPACT cohorts. Multivariable Cox regression analyses stratified by cancer type defined adjusted hazard ratios (AHRs) for disease-specific survival (DSS), progression-free survival (PFS) and overall survival (OS). RESULTS: Using TCGA we identified 32 mutated genes, and 15 copy number (CN) genes with frequency >= 4% in 9,104 patients across 28 cancers. On UVA, having a TP53-mutations or any mutation in the 31 genes (mut31) were associated with worse PFS (HR 1.22, p < 0.0001 and HR 1.1, p = 0.04, respectively) and DSS (HR 1.38, p < 0.0001, and HR 1.16, p = 0.03, respectively). CDKN2A, PTEN deletions, and MYC-amplifications were associated with PFS and DSS (p < 0.05 for all). On MVA, including TP53-mutations, mut31, CDKN2A-deletion, PTEN-deletion, and MYC-amplification, all five alterations were independently prognostic of poor PFS and DSS. Similar results were observed in an independent cohort from MSK-IMPACT (n = 7,051) where TP53 was associated with poor OS independent of mut31 and CN alterations in CDKN2A, PTEN, and MYC in primary tumors (p < 0.0001). CONCLUSIONS: TP53-mutations, CDKN2A-deletion, PTEN-deletion, and MYC-amplification are independent pan-cancer prognostic genomic alterations.