Risk of recurrence after nephrectomy: Comparison of predictive ability of validated risk models

Objective: To assess accuracy of University of California Los Angeles Integrated Staging System (UISS), Stage, Size, Grade and Necrosis (SSIGN) score, Leibovich score and GRade, Age, Nodes and Tumor (GRANT) score, the ASSURE (Adjuvant Sunitinib or Sorafenib vs. placebo in resected Unfavorable REnal cell carcinoma) score models and seventh American Joint Committee on Cancer (AJCC)/TNM staging system in predicting recurrence-free survival (RFS) in surgically-treated non-metastatic clear cell renal cell carcinoma (ccRCC) patients. Materials and methods: Kaplan-Meier curves and the log-rank test tested RFS according to risk groups among the UISS, SSIGN, Leibovich and GRANT models and the AJCC/TNM system. The Heagerty's C-index for survival tested for discrimination of each model at different time points after nephrectomy. Results: Three hundred and fifty-eight M0 ccRCC patients were included. RFS significantly differed among each risk category for all models (P < 0.001). SSIGN showed the highest c-index over time (from 0.89 at 6-month to 0.82 at 60-month), followed by Leibovich (from 0.89–0.82), AJCC/TNM stage (from 0.82–0.77), ASSURE (from 0.81 to 0.76), GRANT (from 0.83–0.73) and UISS (from 0.76–0.72). For all models, peak discriminatory ability was reached before 12 months. The most prominent decline occurred within 24 months and reaches the lowest discriminatory ability at 60 months. Conclusions: Predictive models, with preference for SSIGN and Leibovich scores, are reliable to predict recurrence after nephrectomy and should be recommended to tailor postoperative surveillance protocols.