A model for predicting overall survival in men with metastatic castrate-resistant prostate cancer (CRPC) for whom first-line chemotherapy failed.

24 Background: Several prognostic models for overall survival (OS) have been developed and validated in men with chemotherapy naïve mCRPC. The primary objective was to develop and validate a prognostic model that can be used to predict OS in men who have failed first-line chemotherapy. METHODS: Data was used from a phase III trial of 755 mCRPC men who had developed progressive disease following first-line chemotherapy and were randomized to cabazitaxel plus prednisone or mitoxantrone plus prednisone (TROPIC trial). The data was randomly split into training (n=507) and testing (n=248) sets. A separate data, consisting of 488 men previously treated with docetaxel who were randomly assigned to either satraplatin and prednisone or placebo and prednisone, was used as the validation set (SPARC trial). Penalized regression method was used to identify important prognostic factors. Adaptive Lasso selected nine variables of OS. A predictive score was computed from the estimated regression coefficients and used to classify patients into low (<-1.29) and high (>= -1.29) risk groups in the testing datasets. The model was assessed for its predictive accuracy using time dependent area under the curve (AUC) on the testing sets (TROPIC and SPARC trials). RESULTS: The final selected model included: ECOG performance status, time since last docetaxel use, measurable disease, presence of visceral disease, pain, duration of prior hormonal use, hemoglobin, prostate specific antigen and alkaline phosphatase. In the TROPIC testing set, the median OS in high and low risk groups were 11 and 17 months, respectively, with a hazard ratio (HR)=2.47 (p-value<0.0001). Using the SPARC set, the median OS were 11 and 20 months in the high and low risk groups, respectively, with a HR=1.94 (p<0.0001). The time dependent AUC were 0.73 and 0.70 on the testing sets. CONCLUSIONS: A prognostic model of OS in the post-docetaxel mCRPC setting was developed and validated and risk groups were identified. This model can be used to select patients based on their prognosis to participate in clinical trials. Prospective validation is needed.

Duke Scholars

Author Susan Halabi Biostatistics & Bioinformatics, Division of Biostatistics

Author Andrew John Armstrong Medicine, Medical Oncology