Which Factors Predict Overall Survival in Patients With Metastatic Castration-Resistant Prostate Cancer Treated With Abiraterone Acetate Post-Docetaxel?

Abstract

Individual patients' survival varies greatly in metastatic castration-resistant prostate cancer. Retrospective analysis of 368 patients treated with docetaxel and starting abiraterone acetate revealed 5 adverse prognostic factors: hemoglobin < 12 g/dL, > 10 metastases, ECOG performance status >= 2, radiographic progression, and time since diagnosis < 90 months. A prognostic model stratifies patients into 3 groups with different estimated survival, which can aid in patient counseling.Abiraterone acetate (AA) increases overall survival (OS) in patients with metastatic castration-resistant prostate cancer (mCRPC) previously treated with docetaxel. However, survival time varies substantially between individuals. Our goal was to identify prognostic factors that better estimate OS. Materials and Methods: This is a retrospective multicentric analysis of 368 patients with mCRPC starting AA with prednisone after docetaxel. Cox proportional hazards statistics were applied. A multivariate model was constructed based on significant univariate predictors by using a manual stepwise forward and backward selection strategy. Model performance was determined by using receiver operating characteristic (ROC) curves. Results: Univariate analysis identified 20 significant OS predictors. A multivariate model was constructed, based on 220 patients, incorporating 5 independent risk factors for decreased OS at the time of AA initiation: hemoglobin <12 g/dL (hazard ratio [HR] 2.02), > 10 metastases (HR 1.80), ECOG performance status >= 2 (HR 1.88), radiographic progression (HR 1.50), and time since diagnosis < 90 months (HR 1.66, all P <.05). Patients were stratified into 3 groups: good (0-2 risk factors, median OS 22.6 months), intermediate (3 risk factors, median OS 13.9 months), and poor prognosis (4-5 risk factors, median OS 6.2 months). The area under the ROC curve based on the event "death by the time of median OS (13.3 months)" was 0.736 (95% confidence interval 0.670-0.803). Conclusion: We identified 5 readily available risk factors independently associated with decreased OS. The resulting model may be used for patient counseling in daily clinical practice, as well as patient stratification in clinical trials. (C) 2017 Elsevier Inc. All rights reserved.