Risk prediction models for cancer-specific survival following cytoreductive nephrectomy in the contemporary era

Introduction: To develop a risk-stratification model for cancer-specific survival (CSS) following cytoreductive nephrectomy (CN) in the contemporary era. Materials and Methods: A retrospective review was performed of 313 patients who underwent CN for M1 renal cell carcinoma (RCC) from 1990 to 2010. To account for the introduction of targeted therapies, timing of surgery was classified as immunotherapy era (1990–2004) or contemporary era (2005–2010). Risk scores were developed to predict CSS using Cox proportional hazards regression models. Results: A total of 215 (69%) and 98 (31%) patients were treated in the immunotherapy and contemporary eras, respectively. Median follow-up among survivors was 9.6 years, during which time 291 patients died, including 279 from RCC. On multivariable analysis limited to preoperative features, age ≥ 75, (hazard ratio [HR] 1.9), female sex (HR 1.9), constitutional symptoms (HR 1.61), radiographic lymphadenopathy (HR 1.59), and IVC tumor thrombus (HR 1.65) were significantly associated with CSS. On multivariable analysis including pathologic features, the features above as well as coagulative necrosis (HR 1.51) and sarcomatoid differentiation (HR 1.44) were significantly associated with CSS (all P < 0.05). Risk scores were developed for each model and used to predict CSS according to era. Decision curve analysis revealed that the preoperative risk score conferred a net benefit over a treat-all or treat-none approach beyond a 1-year cancer-specific mortality threshold of 25%. Conclusions: We developed risk scores to predict CSS for patients treated with CN in the contemporary era. Patients with poor predicted survival may consider avoiding CN as initial management.