AbstractBackgroundThere is a considerable need to incorporate biomarkers of resistance to new antiandrogen agents in the management of castration‐resistant prostate cancer (CRPC).MethodsWe conducted a phase II trial of enzalutamide in first‐line chemo‐naïve asymptomatic or minimally symptomatic mCRPC and analyzed the prognostic value of TMPRSS2‐ERG and other biomarkers, including circulating tumor cells (CTCs), androgen receptor splice variant (AR‐V7) in CTCs and plasma Androgen Receptor copy number gain (AR‐gain). These biomarkers were correlated with treatment response and survival outcomes and developed a clinical–molecular prognostic model using penalized cox‐proportional hazard model. This model was validated in an independent cohort.ResultsNinety‐eight patients were included. TMPRSS2‐ERG fusion gene was detected in 32 patients with no differences observed in efficacy outcomes. CTC detection was associated with worse outcome and AR‐V7 in CTCs was associated with increased rate of progression as best response. Plasma AR gain was strongly associated with an adverse outcome, with worse median prostate specific antigen (PSA)‐PFS (4.2 vs. 14.7 m; p < 0.0001), rad‐PFS (4.5 vs. 27.6 m; p < 0.0001), and OS (12.7 vs. 38.1 m; p < 0.0001). The clinical prognostic model developed in PREVAIL was validated (C‐Index 0.70) and the addition of plasma AR (C‐Index 0.79; p < 0.001) increased its prognostic ability. We generated a parsimonious model including alkaline phosphatase (ALP); PSA and AR gain (C‐index 0.78) that was validated in an independent cohort.ConclusionsTMPRSS2‐ERG detection did not correlate with differential activity of enzalutamide in first‐line mCRPC. However, we observed that CTCs and plasma AR gain were the most relevant biomarkers.