Development and external validation of a nomogram predicting cancer-specific mortality-free survival after trimodal therapy in muscle-invasive bladder cancer.

INTRODUCTION

We developed and externally validated a novel population-based model for predicting cancer-specific mortality free survival (CSM-FS) in trimodal therapy (TMT)-treated muscle-invasive bladder cancer (MIBC) patients, and we compared its performance to that of AJCC categories.

METHODS

Within the SEER database (2004-2021), we identified MIΒC patients treated with TMT (n = 1801). The population was randomly divided into development (n = 900, 50%) and external validation (n = 901, 50%) cohorts. Univariable Cox-regression models addressing CSM were fitted. Subsequently, the most parsimonious model with the best validation metrics was selected. Accuracy, calibration, and decision curve analyses (DCAs) tested the novel nomogram relative to the AJCC staging within the external validation cohort.

RESULTS

Age at diagnosis, sex, median household income, T stage, N stage, and tumor size qualified for inclusion in the nomogram predicting CSM-FS. In external validation cohort, the accuracy of the novel nomogram predictions at two vs. three vs. four years after TMT was 0.70 vs. 0.67 vs. 0.66, respectively. Conversely, AJCC staging accuracy was 0.57 vs. 0.56 vs. 0.55 for the same timepoints. The novel nomogram predictions were well calibrated relative to observed CSM-FS rates. Finally, in DCAs the novel nomogram outperformed AJCC staging at all three timepoints.

CONCLUSION

The novel nomogram for TMT-treated patients outperformed AJCC-based predictions of CSM-FS, when accuracy and DCAs represented the benchmarks of interest. However, its predictive accuracy is not ideal.