Head-to-head comparison of all the prognostic models recommended by the European Association of Urology Guidelines to predict oncologic outcomes in patients with renal cell carcinoma

Abstract

European Urology Association guidelines suggest the use of integrated prognostic systems to assess oncologic outcomes after surgery in patients with localized renal cell carcinoma (RCC). We performed a head-to-head comparison among all the EAU guidelines recommended prognostic models in RCC. The study included 2,014 patients treated with surgery for clinically localized RCC. Patients were classified into prognostic risk groups, based on each of the five EAU guidelines recommended prognostic model definition, namely UISS, Leibovich 2003, VENUSS, GRANT, and Leibovich 2018 score. Prognostic accuracy of each prognostic model to predict clinical progression or cancer-specific mortality (CSM) was assessed, and ROC curves were calculated, according to histological subtype, namely clear-cell, papillary, and chromophobe RCC. Of 2,014 patients, 1,575 (78%) harboured clear-cell, 312 (16%) papillary, and 127 (6%) chromophobe RCC. Median follow-up was 66 months [Interquartile range (IQR): 29-120]. In clear-cell RCC, low-risk patients rates ranged from 21% to 64%, according prognostic model. The same phenomenon was observed for papillary and chromophobe RCC. In clear-cell RCC, Leibovich 2018 resulted the most accurate model in predicting clinical progression (88.1%) and CSM (86.8%). Conversely, VENUSS or UISS prognostic models predicting oncologic outcomes represented the most accurate in papillary (88.7% and 84.8%) or chromophobe (87.8% and 89.1%) RCC, respectively. A non-negligible difference in terms of performance accuracy exists among the EAU guidelines recommended prognostic models. Thus, their adoption in RCC should be histology-specific and follow-up strategies based on prognostic risk class appear justified only if the appropriate model is used to stratify patients into prognostic risk groups.