Integration of genomic classification and clinical characteristics as a predictor of survival in de novo metastatic prostate cancer.

123 Background: Genomic assessment has revolutionized cancer care and individual gene alterations inform prognosis and predict benefit from specific therapies. No comprehensive DNA-based genomic classification of metastatic prostate cancer exists that accounts for both individual alterations and combinations of alterations that are frequently identified. Methods: Retrospective cross-sectional study of U.S. Veterans diagnosed with synchronous (de novo) metastatic hormone-sensitive prostate cancer (mHSPC) and DNA-based comprehensive genetic profiling (CGP) through the National Precision Oncology Program. Multivariable models with overall survival (OS) as the endpoint were used to develop a clinic-genomic prognostic risk classification. Validation was performed in metachronous mHSPC and patients in MSK-IMPACT. Results: In 2484 veterans with synchronous mHSPC and tissue (primary tumor or metastasis) CGP (median age; 72 years), baseline data including age, PSA, and Charlson comorbidity index (CCI) was collected. The cohort was divided into training and testing datasets, and 16 genes associated with survival were identified (TP53, PTEN, RB1, BRCA2, FGFR1, FGFR3-4, FGFR19, CDK12, RAD21, MYC, CCND1, LYN, AR, PRKCI, SPOP). DNA alterations associated with specific genes/gene combinations were assigned into favorable, intermediate, or unfavorable groups based upon mortality risk. In a multivariable model classification of alterations into intermediate (aHR 1.75, 95% CI 1.46-2.08) or unfavorable groups (aHR 2.71, 95% CI 2.15-3.42) was associated with increased mortality relative to the favorable/no alteration group, demonstrating a tAUC of 0.77 at 12 months. In 1236 Veterans with metachronous mHSPC intermediate (aHR 1.45, 95% CI 1.17-1.79) and an unfavorable classification (aHR 2.06, 95% CI 1.54-2.76) was associated with increased mortality with AUC of 0.69 at 12 months. In an external validation in non-veterans, intermediate (aHR 2.45, 95% CI 1.87-3.21) and unfavorable classifications (aHR 4.37, 95% CI 3.06-6.22) were associated with increased mortality with AUC of 0.75 at 12 months. Conclusions: Tumor genomic classification is prognostic for OS in patients with synchronous mHSPC. The genomic classification produced consistent, robust results in multiple validation datasets across additional clinical metastatic states, CGP analytes, and in a non-VA cohort. This prognostication approach has the potential to guide decision-making related to therapeutic intensity and duration.