Development and validation of a survival nomogram for muscle-invasive bladder cancer in the elderly: using competing risk models and propensity matching to apply the prediction tool

Abstract

Aim: Patients with muscle-invasive bladder cancer (MIBC) have a low survival rate, with a 5-year survival of approximately 45%, regardless of the treatment received. The risk of death within 5 years after radical cystectomy in patients with MIBC remains as high as 60%. Over 80% of patients with bladder cancer are over 65. Therefore, identifying prognostic correlates associated with radical cystectomy in older patients with MIBC could improve survival rates. In addition, radiotherapy and chemotherapy are particularly important as adjuvant treatments for MIBC patients undergoing radical cystectomy. Therefore, this study aimed to find risk factors for cancer-specific survival (CSS) and overall survival (OS) after radical cystectomy in elderly MIBC patients. The difference in survival between radiotherapy and chemotherapy was analyzed by Kaplan-Meier (K-M) curves to provide theoretical support for whether radiotherapy is recommended for such patients. Methods: Patients 65 or older diagnosed with MIBC with radical cystectomy between 2004-2018 were obtained from the Surveillance, Epidemiology, and End Results (SEER) database. 2004-2015 patients were subjected to column line plot production and internal validation, and 2016-2018 patients were subjected to external temporal validation. A single-factor COX regression model was first used to screen for prognostic correlates. Then a multi-factor COX regression model was used to screen for independent risk factors. A nomogram was constructed by using independent risk factors. The accuracy and reliability of the nomogram were examined using calibration curves, consistency index (C-index), and area under subjects (AUC) as operational characteristic curves. Decision curve analysis (DCA) was performed to evaluate the clinical value of the prediction model. Results: A total of 11,557 patients were included in this study, divided into training set (N = 4,712), validation set (N = 4,810) and external validation set (N = 2,035). Multivariate COX regression models showed that chemotherapy, radiotherapy, TNM stage, race, and age were independent risk factors for CSS and OS patients. We constructed a nomogram to predict CSS and OS in elderly MIBC patients undergoing radical cystectomy. The C-indexes were 0.692 (95%CI: 0.680-0.704) and 0.690 (95%CI: 0.678-0.702) for the CSS training and validation sets, respectively, and 0.674 for the OS training and validation sets (95%CI: 0.664-0.684) and 0.672 (95%CI: 0.662-0.682) for the OS training and validation sets, respectively. The C-index of the external validation set CSS was 0.731 (95%CI: 709-0.753), and that of OS was 0.721 (95%CI: 0.701-0.741), indicating that the nomogram prediction model has good discriminative power. The calibration curves and AUC also suggested that the nomogram had good accuracy and discrimination. In addition, the KM curves of propensity-matched pre- and post-radiotherapy showed that radiotherapy was detrimental to patient survival. Meanwhile, chemotherapy favored OS and short-term CSS but not long-term CSS. Conclusions: We established a nomogram to predict the CSS and OS in elderly MIBC patients undergoing radical cystectomy. After internal cross-validation and external validation, the nomogram prediction model showed good accuracy and reliability, and the DCA results showed that the nomogram has good clinical value. In addition, this study gave good suggestions on whether radiotherapy or chemotherapy is necessary for radical cystectomy in elderly MIBC patients.