• A clinical-genetic (CG) circulating tumor DNA (ctDNA)-based prognostic model for predicting overall survival (OS) in men with metastatic castrate-resistant prostate cancer (mCRPC) treated with potent androgen receptor inhibition (Alliance).

Background: We have previously developed and validated a clinical prognostic model of OS in mCRPC men that included these variables: performance status, disease site, opioid analgesic use, lactate dehydrogenase, albumin, hemoglobin, prostate specific antigen, and alkaline phosphatase. The goal of this analysis is to improve upon the clinical model of OS by incorporating ctDNA pathogenic genetic alterations (PGAs). Methods: Data from the A031201 phase 3 trial of enzalutamide+/- abiraterone were used to develop and validate the CG model of OS. Cell-free DNA was isolated from plasma and analyzed using a 69-gene targeted DNA-sequencing assay for detection of ctDNA PGAs. Genetic features were identified based on feature importance using a random survival forest and the final CG model was trained including clinical and selected genetic factors. Model discrimination was assessed using time-dependent area under the receiver operating characteristic curve (tAUC). Results: Data were available on 776 patients. In addition to clinical variables, the model included in this order: gains in AR and the AR enhancer, MYC, RSPO2, and losses and/or PGAs of ZBTB16, PTEN, MSH6, PPP2R2A, NKX3-1, TP53, FANCA, RB1, APC, CHD1, and BRCA2, and ichorCNA tumor fraction. tAUCs in clinical and CG models were 0.72 (95% CI=0.72-0.73) and 0.77 (95% CI= 0.76-0.77). Median OS and the hazard ratios by the three- and four- prognostic risk groups are presented in the table. Conclusions: CG model identified novel ctDNA PGAs prognostic of OS and can be utilized to classify patients into risk groups useful in selecting patients in future trials of mCRPC. Clinical trial information: NCT01949337.