The RISC nomogram (RN) to predict overall survival (OS) of patients (pts) with metastatic urothelial carcinoma (mUC) receiving first-line platinum-based combination chemotherapy (CT).

Abstract

e16026 Background: The available prognostic models (PM) for OS of pts with mUC were derived from clinical trial populations of cisplatin (cis)-treated pts. We sought to validate and improve these PM in light of the upcoming 1st-line trials pooling cis and carboplatin (carbo)-treated pts. Methods: Individual pt-level data from 29 centers was collected by electronic data capture. Selection criteria included mUC, 1st-line CT with a cis- or carbo-based combination CT. Prior peri-operative (po)-CT was allowed. The overall sample was randomly split into a training (Tr) and testing (Te) set. Generalized Boosted Regression Modeling was used first to exclude variables not associated with OS. Backward variable selection was then undertaken based on the p values obtained from Cox regression models. Platinum-type was incorporated in the analysis and RN as a stratification factor. A RN was built for estimating 1, 2 and 5-year OS probability. Distribution of nomogram scores (NS, from the Cox model) according to objective response (OR) was also analyzed. Results: 1,020 pts were analyzed (Tr: 687; Te: 333). Median OS was 14.3 and 13.7 months (m) in the Tr and Te set. Significant differences in OS were found between cis (N = 639) and carbo-CT (N = 381, p < .001). In the platinum-stratified Cox model, significant variables were: Performance Status (p < .001), body mass index (p = 0.002), ethnicity (p = 0.008), visceral metastases (p < .001), and prior po-CT (p = 0.018). The c-index of this model was more optimal than those of previous models that were tested in cis-treated pts from this cohort (Table). The NS were associated with OR (p < .001) and incorporating OR in the model further improved the c-index (0.683, 4m landmark analysis). Conclusions: We validated and extended prior PM with a 5-factor PM for OS using a large dataset of real world generally off-trial pts receiving 1st-line platinum based CT. The RN is applicable to pts regardless of cis or carbo-CT, it accounts for novel prognostic factors and predicts for OR. Future trials should employ this optimized RN. Model Tr set Te set Bajorin 1999 0.622 0.619 Apolo 2013 0.642 0.641 Galsky 2013 0.657 0.644 Present 0.666 0.649