Identifying high surgical risk in muscle-invasive bladder cancer (MIBC) patients undergoing radical cystectomy (RC).

Abstract

460 Background: Although RC is the standard treatment for MIBC, it is associated with non-negligible morbidity and mortality. Given the advanced age and prevalence of comorbidities in MIBC patients, many are suboptimal RC candidates. We sought to develop a prediction rule for determining high surgical risk using routinely available pre-surgical variables. Methods: We identified N=8,634 patients with MIBC (cT2-T4aN0M0) who underwent RC, without perioperative chemotherapy, within the National Cancer Database (2003-2012). Using multi-state modeling, we calculated time spent in 3 health states: hospitalized, discharged and death over 90 days post-RC. Predictors were selected in a stepwise manner. Cross validation was performed leaving one of 6 geographic regions out to assess model performance. Time spent in each state was weighted by utility (0=death, 1=full health) to determine 90d quality-adjusted survival (QAS) as a composite of morbidity and mortality. Results: Median age: 69y (IQR 62-78), mean length of stay: 10±12d, and 90d mortality: 654/8,634 (7.6%). Increasing age, cT stage, Charlson comorbidity index (CCI) and lower hospital RC volume were associated with greater 90d mortality (Table). C-statistics of 0.57-0.73 and calibration slopes of 0.54-1.37 (1=perfect) were sufficient across the regions. Our model predicted 25% of patients had a 90d mortality >10%; observed 90d mortality in this group was 14.0% (95% CI 12.5-15.5). Mean QAS was 63d (range 43-68). Conclusions: We developed a multi-state model to identify high post-RC risk that may ultimately help to identify patients for ‘prehabilitation’ strategies and/or inform standard definitions of ‘cystectomy ineligibility’ for clinical trials. [Table: see text]