Abstract 2322
Background
The available prognostic models of OS for pts with metastatic UC were derived from clinical trial populations of cisplatin-treated pts only. We aimed to develop a new model based on ‘real world’ pts.
Methods
Individual pt-level data from 29 centers was collected. Pts had to be treated for metastatic UC at the participating sites between 01/2006 and 01/2011. Selection criteria included metastatic UC, and first-line cisplatin- or carboplatin-based chemotherapy. The overall sample was randomly split into a development and a validation cohort. Generalized Boosted Regression Modeling was used first to exclude variables not associated with OS. Backward variable selection was then undertaken. Platinum-type was incorporated in the analysis as a stratification factor. Two nomograms were built to estimate OS probability, the first based on baseline factors and the second incorporating objective response (OR). The performance of the present nomogram and that of the other available models was assessed for accuracy (Brier score), calibration (Hosmer-Lemeshow test), and discrimination (Harrell c-index).
Results
1,020 pts were analyzed (development: 687; validation: 333). In the platinum-stratified Cox model, significant variables for OS were: performance status (p
Conclusions
We developed and validated two nomograms that accounted for novel prognostic factors and can be used before and after completion of chemotherapy, respectively. These two nomograms may be suitable tools to enhance patient stratification and inform pts in the clinic.
Clinical trial identification
None
Legal entity responsible for the study
Fondazione IRCCS Istituto Nazionale dei Tumori
Funding
Fondazione IRCCS Istituto Nazionale dei Tumori
Disclosure
All authors have declared no conflicts of interest.