A prognostic model for predicting overall survival (OS) in patients (pts) with metastatic castration-resistant prostate cancer (mCRPC) treated with abiraterone acetate (AA) after docetaxel.

5013 Background: COU-AA-301 was a multinational, randomized, controlled, phase III trial comparing AA + prednisone (P) (n = 797) versus placebo + P (n = 398) in mCRPC pts post-docetaxel. Using data from that trial, we developed a prognostic model for predicting OS in pts treated with AA post-chemotherapy, with a focus on readily assessable clinical parameters. Methods: The analyses used data from pts treated with AA in the COU-AA-301 trial for whom relevant baseline data were available (n = 729). Baseline variables were assessed for association with OS through a univariate Cox proportional hazards regression model. High/low values for accepted normal ranges were used for laboratory parameters. The Cox proportional hazards regression was used with a stepwise procedure to identify independent prognostic factors for OS. The model was subject to sensitivity analyses and the C-index was utilized as a measure of model accuracy. Results: The following risk factors were associated with a poor prognosis: ECOG performance status (only pts with scores of ≤ 2 were eligible for this trial) of 2 (HR = 2.19, p < 0.0001), presence of liver metastases (HR = 2.00, p < 0.0001), time from start of initial LHRH agonist therapy to start of AA treatment ≤ 36 months (HR = 1.30, p = 0.0078), low albumin (HR = 1.54, p < 0.0001), high ALP (HR = 1.38, p = 0.0016), and high LDH (HR = 2.31, p < 0.0001). Patients were categorized into 3 risk groups (good prognosis, n = 369; intermediate prognosis, n = 321; poor prognosis, n = 107) based on total number of risk factors and median OS calculated for each group (table). The C-index was 0.74 (95% CI: 0.68, 0.80). Conclusions: This prognostic model uses readily available clinical parameters to conveniently assess risk for mCRPC pts previously treated with docetaxel and initiating treatment with AA + P. If validated, the model will be useful in clinical practice and clinical trials. Clinical trial information: NCT00638690. [Table: see text]