External validation of postoperative nomograms for prediction of all-cause mortality, cancer-specific mortality, and recurrence in patients with urothelial carcinoma of the bladder

Abstract

Background: The Bladder Cancer Research Consortium (BCRC) created nomograms to predict all-cause mortality (ACM), cancer-specific mortality (CSM), and recurrence after radical cystectomy (RC) for urothelial carcinoma of the bladder (UCB).

Objective: To perform a formal validation of the BCRC nomograms in a large multi-institutional patient cohort from Europe.

Design, setting, and participants: Records of 2501 patients who underwent RC for UCB at eight European centers were reviewed. Complete information for external validation was available in 2404 patients for the ACM and CSM nomograms and in 2243 patients for the recurrence nomogram.

Measurements: For the purpose of external validation, model discrimination was measured using the receiver operating characteristics derived area under the curve. Calibration plots examined the relationship between predicted and observed probabilities at 2 yr, 5 yr, and 8 yr. Decision curve analyses were applied to assess the net benefit derived from the three models.

Results and limitations: The discrimination accuracies of the BCRC nomograms for ACM and CSM at 2 yr, 5 yr, and 8 yr after RC were 71.0%, 69.1%, and 68.2%, and 74.9%, 73.1%, and 72.4%, respectively. The accuracy of discrimination for the recurrence nomogram at the same time points was 76.5%, 75.3%, and 74.9%, respectively. Calibration plots revealed slight underestimations from ideal predictions. Decision curve analyses showed an increased net benefit for the use of the BCRC nomograms in this cohort. Limitations include the retrospective study design, potential surgeon bias, and lack of a central pathologic review.

Conclusions: The ACM, CSM, and recurrence nomograms showed acceptable predictive accuracies and could thus be adopted into clinical practice in UCB patients treated in Europe.