Predicting toxicity-related docetaxel discontinuation and survival in metastatic castration-resistant prostate cancer (mCRPC) using open phase III trial data.

Abstract

75 Background: Docetaxel (D) is widely used in mCRPC, however its optimal use remains unclear in the current treatment (tx) landscape. Biomarkers to predict D toxicity may help inform tx selection. Methods: Through Project Data Sphere, we pooled patient (pt) data from the control arms of three frontline mCRPC trials: ASCENT2, VENICE, and MAINSAIL. Tx in each control arm consisted of D 75mg/m2 every 21 days + prednisone 5mg twice/day. The primary outcome was occurrence of toxicity-related D discontinuation (TRDD). Reasons for D discontinuation were recorded and demographic/clinical data were considered in a competing risks regression (CRR) to develop a model to predict TRDD. Cumulative incidence (CI) of TRDD was estimated after accounting for the occurrence of competing events (death or progression). This model was used to build a risk calculator to predict TRDD, the output of which was used in a classification and regression tree (CART) to identify 3 risk groups. Overall survival (OS) for the pooled cohort and for each risk group was assessed via the Kaplan-Meier (KM) method. Results: 1568/1600 pts had complete data and were studied. 41.4% died from any cause during the trials; median follow-up for survivors was 12.1 months (mos). Median OS for the entire cohort was 21 mos. CRR yielded the following significant factors that were included in the predictive model and risk calculator: age, ECOG performance status, AST, bilirubin, use of analgesics, and diagnoses of diabetes and chronic kidney disease. Pooled CI of TRDD was 19% after accounting for competing events (death, 474 pts; progression, 59 pts) within 12 mos of starting tx. CART analysis defined the risk groups as low (model-derived TRDD risk ≤24%), intermediate (25-64%), and high (≥65%). In each risk group, probability of TRDD during tx was 14%, 58%, and 79%, and median OS was 24 mos, 20 mos, and 13 mos, respectively (p<0.001). Conclusions: Tx selection in mCRPC remains a challenge. Our model can help clinicians balance D toxicity and efficacy in order to make better-informed decisions.