Validation of major prognostic models for metastatic urothelial carcinoma using a multi-institutional cohort of the real world

Background: Several prognostic models predicting survival of patients with metastatic urothelial carcinoma (UC) have been developed; however, of them, the first model by Bajorin in 1999 is still the most representative and widely used, and validations of newer models are lacking. This study aimed to validate three major prognostic models for metastatic UC constructed based on clinical trials. Methods: We reviewed 200 patients with metastatic UC who received first-line chemotherapy at our five affiliate institutions between 2003 and 2011. Using this multi-institutional cohort, we validated the following models: the “Bajorin model,” a model consisting of visceral metastasis and performance status; the “Apolo model,” a nomogram including visceral metastasis, performance status, albumin and hemoglobin; and the “Galsky model,” a nomogram including leukocyte count, number of sites of visceral metastases, site of primary tumor, performance status and lymph node metastasis. Harrell’s c-index was calculated for each model. Cox proportional hazards regression model was used for multivariate analysis. Results: Among 200 patients, 171 (85.5 %) died during the follow-up, with a median survival of 12.0 months. Multivariate analysis demonstrated ECOG performance status, visceral metastasis and leukocyte count to be independent predictors of overall survival. C-index results (95 % confidence interval) were Bajorin: 0.86 (0.74–0.95); Apolo: 0.89 (0.78–0.98); and Galsky: 0.82 (0.69–0.93). Conclusions: All models were demonstrated to have high external validities in real-world patients, and of them, the “Apolo model” achieved the highest c-index in the present population. Further studies with larger populations are needed for establishment of the next standard model.