Skeletal metastasis is seen in most CRPC patients and is a common source of morbidity. We previously reported that bone biomarkers in CRPC patients are independently prognostic for OS, and that in men with highly elevated markers there is benefit from bone-targeted therapy (Lara, et al. JNCI 2014). Here we report our prospectively planned classification & regression tree (CART) analysis of clinical covariates and bone biomarkers as part of the phase III S0421 trial of docetaxel +/- atrasentan.
Markers for bone resorption [N-telopeptide (NTx) & pyridinoline (PYD)] & formation [C-terminal collagen propeptide (CICP) & bone alkaline phosphatase (BAP)] were measured in pre-treatment sera from men on S0421. Bone biomarkers & baseline clinical covariates were included in a Cox model for OS; significant variables were allowed to compete in a CART analysis to identify prognostic groups using CTREE in the R package "party". Hazard ratios (HR) were calculated by comparing OS in each of the terminal nodes to a reference group in a Cox PH model. Kaplan Meier curves for each prognostic group defined by the tree were constructed.
750 men with evaluable bone marker & clinical data were included. Each bone marker significantly contributed to the final Cox model, with higher levels associated with worse OS. Cox stepwise selection identified BAP, CICP, PYD, hemoglobin (Hgb), pain score, age, black race, PSA, and visceral disease as associated with OS. CART analysis selected CICP, BAP, Hgb, & pain score for the final model, & identified 5 prognostic groups: