Development and validation of a model to predict the outcomes of radical cystectomy in patients with bladder cancer

Objectives This study aims to develop a prognostic model that estimates the post‐operative risk of cancer‐specific mortality in patients with bladder cancer who underwent radical cystectomy (RC). Methods We analyzed the data from patients with bladder cancer who had undergone radical cystectomy without receiving adjuvant chemotherapy across 36 institutions in the Japan Urological Oncology Group. The data were randomly split into training ( N = 1348) and validation sets ( N = 674) in a 2:1 ratio. Twenty‐five variables were analyzed, and a multivariable Cox regression model predicting cancer‐specific mortality was developed and validated. Prognostic scores were categorized into good and poor prognostic groups based on the upper tertile. The performance of the model was compared against the CheckMate 274 risk classification as a reference, which is used for determining the indication of adjuvant nivolumab therapy. Results The final model incorporated eight variables. In the validation set, it outperformed the CheckMate 274 risk classification with superior time‐dependent area under the curves (5‐year: 0.81 vs. 0.67) and was well‐calibrated. Furthermore, our model reclassified 27.8% of patients categorized as high‐risk by the CheckMate 274 risk classification into the good prognosis group. Conclusions We developed and validated a prognostic model for patients with bladder cancer who underwent RC. This model will be beneficial in identifying patients with poor prognosis and those who are potential candidates for clinical trials of adjuvant therapy.