Accuracy and Clinical Utility of a Tumor Grade- and Stage-based Predictive Model in Localized Upper Tract Urothelial Carcinoma

Background Among various clinicopathologic factors used to identify low-risk upper tract urothelial carcinoma (UTUC), tumor grade and stage are of utmost importance. The clinical value added by inclusion of other risk factors remains unproven. Objective To assess the performance of a tumor grade- and stage-based (GS) model to identify patients with UTUC for whom kidney-sparing surgery (KSS) could be attempted. Design, setting, and participants In this international study, we reviewed the medical records of 1240 patients with UTUC who underwent radical nephroureterectomy. Complete data needed for risk stratification according to the European Association of Urology (EAU) and National Comprehensive Cancer Network (NCCN) guidelines were available for 560 patients. Outcome measurements and statistical analysis Univariable and multivariable logistic regression analyses were performed to determine if risk factors were associated with the presence of localized UTUC. The sensitivity, specificity, positive predictive value, negative predictive value, and accuracy of the GS, EAU, and NCCN models in predicting pathologic stage were calculated. Results and limitations Overall, 198 patients (35%) had clinically low-grade, noninvasive tumors, and 283 (51%) had ≤pT1disease. On multivariable analyses, none of the EAU and NCCN risk factors were associated with the presence of non–muscle-invasive UTUC among patients with low-grade and low-stage UTUC. The GS model exhibited the highest accuracy, sensitivity, and negative predictive value among all three models. According to the GS, EAU, and NCCN models, the proportion of patients eligible for KSS was 35%, 6%, and 4%, respectively. Decision curve analysis revealed that the net benefit of the three models was similar within the clinically reasonable range of probability thresholds. Conclusions The GS model showed favorable predictive accuracy and identified a greater number of KSS-eligible patients than the EAU and NCCN models. A decision-making algorithm that weighs the benefits of avoiding unnecessary kidney loss against the risk of undertreatment in case of advanced carcinoma is necessary for individualized treatment for UTUC patients. Patient summary We assessed the ability of three models to predict low-grade, low-stage disease in patients with cancer of the upper urinary tract. No risk factors other than grade assessed on biopsy and stage assessed from scans were associated with better prediction of localized cancer. A model based on grade and stage may help to identify patients who could benefit from kidney-sparing treatment of their cancer.