BACKGROUND: Plasma tumor DNA fraction is prognostic in metastatic cancers. This could improve risk stratification before commencing a new treatment. We hypothesized that a second sample collected after one cycle of treatment could refine outcome prediction of patients identified as poor prognosis based on plasma DNA collected pre-treatment. PATIENTS AND METHODS: Plasma DNA [128 pre-treatment, 134 cycle 2 day 1 (C2D1), and 49 progression] from 151 chemotherapy-naive metastatic castration-resistant prostate cancer (mCRPC) patients in a phase II study of abiraterone acetate (NCT01867710) were subjected to custom targeted next-generation sequencing covering exons of these genes: TP53, AR, RB1, PTEN, PIK3CA, BRCA1, BRCA2, ATM, CDK12, CHEK2, FANCA HDAC2 and PALB2. We also captured 1500 pan-genome regions enriched for single nucleotide polymorphisms to allow detection of tumor DNA using the rolling B-allele method. We tested associations with overall survival (OS) and progression-free survival (PFS). RESULTS: Plasma tumor DNA detection was associated with shorter OS [hazard ratio (HR): 2.89, 95% confidence intervals (CI): 1.77-4.73, P ≤ 0.0001] and PFS (HR: 2.05; 95% CI: 1.36-3.11, P < 0.001). Using a multivariable model including plasma tumor DNA, patients who had a TP53, RB1 or PTEN gene alteration pre-treatment and at C2D1 had a significantly shorter OS than patients with no alteration at either time point (TP53: HR 7.13, 95% CI 2.37-21.47, P < 0.001; RB1: HR 6.24, 95% CI 1.97-19.73, P = 0.002; PTEN: HR 11.9, 95% CI 3.6-39.34, P < 0.001). Patients who were positive pre-treatment and converted to undetectable had no evidence of a difference in survival compared with those who were undetectable pre-treatment (P = 0.48, P = 0.43, P = 0.5, respectively). Progression samples harbored AR gain in all patients who had gain pre-treatment (9/49) and de novo AR somatic point mutations were detected in 8/49 patients. CONCLUSIONS: Plasma gene testing after one cycle treatment refines prognostication and could provide an early indication of treatment benefit.