Survival following cytoreductive nephrectomy: a comparison of existing prognostic models
Mary E. Westerman, Daniel D. Shapiro, Nizar M. Tannir, Matthew T. Campbell, Surena F. Matin, José A. Karam, Christopher G. Wood
Abstract
OBJECTIVE: To validate models currently used to predict metastatic renal cell carcinoma (mRCC) outcomes in a cohort of patients undergoing cytoreductive nephrectomy (CN). PATIENTS AND METHODS: A total of 10 RCC prognostic models (International Metastatic RCC Database Consortium [IMDC]; Memorial Sloan Kettering Cancer Center [MSKCC]; Culp; Leibovich; University of California at Los Angeles Integrated Staging System [UISS]; Stage, Size, Grade, and Necrosis [SSIGN]; Yaycioglu; Karakiewicz; Cindolo; and Margulis) were chosen based on clinical relevance and use in clinical trial design. Model validation was performed using patients who underwent CN at a single institution between 2005 and 2017, and model discrimination (ability to select patients at risk of death) was assessed. Concordance indices (c-index) were calculated and compared with originally published c-indices. RESULTS: A total of 515 CN patients were stratified according to the prognostic models. A total of 387 (75%) died over the study period, with estimated 3-year survival of 46.1% (95% confidence interval [CI] 41.6-50.4%). All models' discriminatory capacity underperformed when compared to the originally published c-indices. The c-indices ranged from 0.53 (95% CI 0.50-0.56) for the Cindolo model to 0.61 (95% CI 0.58-0.64) for the Leibovich model. The MSKCC and IMDC models performed poorly with c-indices of 0.55 and 0.56, respectively. CONCLUSION: Currently used prognostic models have limited discriminatory capacity when applied to a modern cohort of patients undergoing CN. They are inadequate for risk stratification and randomisation in prospective clinical trials of untreated patients with mRCC. Caution should be used when using these models for clinical decision making.