Development and validation of a prognostic nomogram for overall and disease-specific survival in patients with sarcomatoid urothelial carcinoma.

Introduction: Sarcomatoid urothelial carcinoma (SUC) is a rare and aggressive variant of bladder cancer with limited data guiding prognosis. In this study, we present the first prognostic nomograms in the literature for 3- and 5-year overall survival (OS) and disease-specific survival (DSS), for patients with SUC derived from the surveillance, epidemiology and end results database (SEER).

Materials And Methods: Patients with SUC were identified by using the ICD-10 topography codes C67.0-C67.9 (bladder cancer), and the morphologic code 8122 (SUC). Patients were randomly divided into a training cohort (TC) and a validation cohort (VC) (7:3 ratio). Variables significantly associated with OS and DSS were identified with multivariate Cox regression and were used to build the nomograms. Harrel's C-statistic with bootstrap resampling and calibration curves were used for internal (TC) and external (VC) validation. Clinical utility of the nomograms was assessed with the decision curve analysis (DCA). Goodness of fit between the nomograms and the AJCC 8th edition staging system was compared with the likelihood ratio test.

Results: A total of 741 patients with SUC were included (507 TC, 234 VC). No statistically significant differences in baseline characteristics were identified between the 2 cohorts. Sex, SEER stage, radical cystectomy and chemotherapy were common variables for the OS and the DSS nomograms with the addition of age in the former. Optimism-corrected C-statistic for the nomograms was 0.68 and 0.67 for OS and DSS respectively. In comparison, C-statistic for AJCC was 0.59 for OS and 0.60 for DSS (P < 0.001). Calibration curves constructed for the nomograms showed appropriate consistency between predicted and actual survival. The nomograms demonstrated optimal clinical utility in the DCA, outperforming the AJCC staging system, by maintaining a higher clinical net benefits than treat all, treat none and AJCC curves, across threshold probabilities.

Conclusion: We present the first prognostic nomograms developed in patients with SUC. Our models demonstrated superior prognostic performance to the AJCC system, by utilizing a set of variables readily available in daily practice and may serve as useful tools for the individualized risk assessment of these patients.