Grading the risk-comparing existing nomograms in predicting outcomes post radical nephroureterectomy (RNU) for upper tract urothelial carcinoma (UTUC).

Abstract

Introduction & Objectives: UTUC is an aggressive disease that is often understaged on imaging and initial pathology. We aim to validate published prognostic nomograms using our cohort, to determine the most optimal model for risk stratification following RNU. Method(s): A retrospective analysis was conducted on 103 patients treated with RNU from 2011 to 2021. Survival outcomes were compared between high-grade (HG) and lowgrade (LG) tumours, with uni- and multivariable cox regression used to identify grade-specific predictors of overall survival (OS). Three nomograms (Seisen, Abdul-Muhsin and Cha; Table 1) were externally validated using Harrell's C-index and bootstrap-corrected calibration. Result(s): HG tumours had poorer 5-year OS (53.1% vs 85.8%, P < 0.01), cancer specific survival (CSS; 71.4% vs 93.3%, P = 0.056), and progression free survival (PFS; 52.2% vs 85.8%, P = 0.002), but similar recurrence free survival (RFS; 40.55% vs 44.04%, P = 0.56). Among HG cases, only lymphovascular invasion (LVI) remained significant predictor of poor OS in multivariable analysis. No significant predictors were identified in the LG cohort. The Seisen nomogram achieved excellent discrimination for 5-year CSS (C-index 0.814), with close calibration. Abdul-Muhsin showed good discrimination for OS (C-index 0.708), but poorer performance and miscalibration for CSS (C-index 0.651). Cha's model showed moderate-to-good discrimination (CSS C-index 0.770; RFS C-index 0.605), but calibration was unreliable for RFS (Fig. 1). Conclusion(s): The Seisen nomogram demonstrated superior predictive accuracy for CSS, likely due to its integration of key clinical and pathological features. Incorporating validated nomograms alongside gradespecific variables can refine postoperative risk stratification and help guide patient management.