Development and validation of a novel prognostic model for disease-specific survival in patients with non-metastatic small cell carcinoma of the bladder.

Abstract

695 Background: Small cell carcinoma of the bladder (SCCB) is a rare variant of bladder cancer with limited data informing prognosis. We present a novel prognostic model for disease specific survival (DSS) for patients with non-metastatic (M0) SCCB, derived from the Surveillance, Epidemiology and End Results (SEER) database. Methods: The SEER database (18 registries/November 2020) was queried with the ICD-10 topography codes C67.0-C67.9 (bladder), and the morphologic code 8041 (SCCB). Eligible patients (adults, histologic confirmation, available follow-up/staging data, M0) were randomly divided into a training (TC) and a validation cohort (VC) (7:3 ratio). Variables significantly associated with DSS were identified with multivariate Cox regression and backwards stepwise conditional approach and inserted in the model. Points were assigned to each variable based on the formula: Probability of an event at time t=S0(t)exp(β1x1, β2x2...), (β = regression coefficients, x = observed covariate values, S0(t) = survival function). Cumulative risk points were assigned to each patient and three distinct risk categories for DSS were constructed. Harrel’s C-statistic with bootstrap resampling was used for internal (TC) and external (VC) validation. Model performance was compared to the AJCC 8th edition (likelihood ratio test -LRT). Results: A total of 1039 patients were included (739 TC, 300 VC). No statistically significant differences in baseline characteristics/survival outcomes were identified between TC and VC. Age (years), T/N stage, radical cystectomy, chemotherapy, and radiation were significant variables inserted in the multivariate model (table). Patients were stratified to three risk groups (low, intermediate, high) based on the total cumulative risk points collected. Median DSS was NR (95%CI; NR, NR), 25 (95% CI; 18-32) and 9 months (95% CI; 7-11) for the low, intermediate, and high-risk category, log-rank p<.001. Harrel's C-statistic for the model was 0.70 (95%CI;0.67-0.73) in the TC and 0.66 (95%CI; 0.61-0.71) in the VC. In comparison, C-statistic for the AJCC 8th edition was 0.56 (95% CI; 0.52-0.59) and 0.53 (95% CI; 0.51 to 0.57), respectively. The LRT demonstrated superior performance of the model compared to AJCC (p<.001). Conclusions: We present a novel prognostic model for DSS in patients with locoregional SCCB composed of variables readily available in daily practice, which may serve as a useful tool for individualized risk assessment of patients with this rare malignancy. Further validation in prospective studies/clinical trials is warranted. [Table: see text]