Background: Predictive genomic biomarkers in mCRPC remain elusive. Prior studies suggest that tumor suppressor (TS) loss is prognostic, and may result in less benefit from NHT, but no impact on D efficacy. We assessed genomic predictors of differential benefit of androgen receptor-targeted therapy and chemotherapy for mCRPC. Methods: Patients with mCRPC and targeted exome sequencing of biopsies obtained after metastatic diagnosis were identified (n=109). Patients with pure small cell histology (n=6) were excluded. Time from NHT or D start to clinical/radiographic progression (time to treatment failure, TTTF) was estimated by Kaplan-Meier method, with censoring at next therapy or last follow-up for non-progressors. Results: 80.1% of patients had bone and/or lymph node-only metastases at mCRPC diagnosis. In total, 87/103 (84.5%) and 61/103 (59.2%) received NHT and D for mCRPC, respectively. Median overall survival was 4.5 years from first mCRPC. The frequency and predictive association of selected recurrently-altered genes are detailed in the table. PTEN alterations (alts) were associated with worse TTTF on NHT, but not D, and a similar trend was observed with BRCA2. Biallelic RB1 loss was strongly predictive, conferring significantly shorter TTTF on both NHT and D. A score based on presence of tumor PTEN alt (1) and/or biallelic RB1 alt (1) was predictive of TTTF on NHT (median TTTF of score 0 vs 1 vs 2: 14.7 vs 12 vs 3.8 months; log rank p=0.003). Conclusions: The presence of single or compound PTEN and RB1 alts predict poorer outcomes with NHT for mCRPC. Chemotherapy may be a preferred therapeutic strategy for this patient population.

Key: mo/bi, monoallelic or biallelic loss; bi, biallelic loss only.