Prediction of cancer-specific survival after radical nephroureterectomy for upper tract urothelial carcinoma: development of an optimized post-operative nomogram using decision curve analysis.

Our aim was to conceive and to propose a unique and optimized nomogram that would predict cancer-specific survival (CSS) post-radical nephroureterectomy (RNU) in patients with upper tract urothelial carcinoma (UTUC) by merging the two largest multicentric datasets reported in this population. MATERIALS AND METHODS: The international and the French national collaborative groups on UTUC pooled data on 3387 patients treated with RNU for whom full data for nomogram development were available. The merged study population was randomly split into the development cohort (2371) and external validation cohort (1016). Cox regressions were used for univariable and multivariable analyses and to build different models. The ultimate reduced nomogram was assessed using Harrell's concordance index (c-index) and decision curve analysis (DCA). RESULTS: Of the 2371 patients in the nomogram development cohort, 510 (21.5%) died during follow-up as a consequence of UTUC. The actuarial CSS probability at 5 years was 73.7% (95% CI, 71.9-75.6). DCA revealed that the use of the best model was associated with benefit gains relative to prediction of CSS. The optimized nomogram included only five variables associated with CSS in multivariable analysis: age (p=0.001), T stage (p<0.001), N status (p=0.001), architecture (p=0.02) and LVI (p=0.001). The nomogram's discriminative accuracy was 0.8 (95% CI, 0.77-0.86). CONCLUSIONS: Using standard pathologic features obtained from the largest dataset of UTUCs worldwide, we have devised and validated an accurate and ultimate nomogram, superior to any single clinical variable, for predicting CSS post-RNU.