PURPOSE: Local-regional failures (LF) following radical cystectomy (RC) plus pelvic lymph node dissection (PLND) with or without chemotherapy for invasive urothelial bladder carcinoma are more common than previously reported. Adjuvant radiation therapy (RT) could reduce LF but currently has no defined role because of previously reported morbidity. Modern techniques with improved normal tissue sparing have rekindled interest in RT. We assessed the risk of LF and determined those factors that predict recurrence to facilitate patient selection for future adjuvant RT trials. METHODS AND MATERIALS: From 1990-2008, 442 patients with urothelial bladder carcinoma at the University of Pennsylvania were prospectively followed after RC plus PLND with or without chemotherapy with routine pelvic computed tomography (CT) or magnetic resonance imaging (MRI). One hundred thirty (29%) patients received chemotherapy. LF was any pelvic failure detected before or within 3 months of distant failure. Competing risk analyses identified factors predicting increased LF risk. RESULTS: On univariate analysis, pathologic stage ≥pT3, <10 nodes removed, positive margins, positive nodes, hydronephrosis, lymphovascular invasion, and mixed histology significantly predicted LF; node density was marginally predictive, but use of chemotherapy, number of positive nodes, type of surgical diversion, age, gender, race, smoking history, and body mass index were not. On multivariate analysis, only stage ≥pT3 and <10 nodes removed were significant independent LF predictors with hazard ratios of 3.17 and 2.37, respectively (P<.01). Analysis identified 3 patient subgroups with significantly different LF risks: low-risk (≤pT2), intermediate-risk (≥pT3 and ≥10 nodes removed), and high-risk (≥pT3 and <10 nodes) with 5-year LF rates of 8%, 23%, and 42%, respectively (P<.01). CONCLUSIONS: This series using routine CT and MRI surveillance to detect LF confirms that such failures are relatively common in cases of locally advanced disease and provides a rubric based on pathological stage and number of nodes removed that stratifies patients into 3 groups with significantly different LF risks to simplify patient selection for future adjuvant radiation therapy trials.