Stage and Age Characterize Distinct, Time-Varying Mortality Pathways in Older Adults With Bladder Cancer: A Population-Based Competing-Risk Study

Abstract Background Older adults with bladder cancer face both cancer-related and competing mortality. We examined how tumor stage and age were associated with these outcomes and how their relative hazards changed during follow-up. Methods This retrospective cohort included adults aged ≥ 75 years with bladder urothelial carcinoma diagnosed during 2010–2015 in the Surveillance, Epidemiology, and End Results 17 Registries database. We used Aalen–Johansen estimators, adjusted cause-specific Cox models and smoothed monthly hazards. A sustained crossover required other-cause hazard to exceed bladder-cancer hazard for six consecutive monthly points with adequate numbers at risk. Secondary prediction models underwent internal temporal validation. Results Among 38,852 patients, stage IV versus stage 0a was associated with a bladder-cancer death hazard ratio of 22.92 (95% confidence interval 21.24–24.72). For other-cause death, the hazard ratio for age ≥ 90 versus 75–79 years was 3.68 (3.49–3.88). Other-cause hazard predominated near diagnosis in stages 0a, 0is and I. Sustained crossovers occurred at approximately 40.5 months in stage II and 56.5 months in stage III; no stable crossover was identified in stage IV within the 84-month observation window. Bladder-cancer death nevertheless remained the more frequent cumulative outcome at 84 months in stages II and III. Temporal-validation areas under the curve at 12, 36 and 60 months were 0.855, 0.822 and 0.800 for bladder-cancer death and 0.629, 0.632 and 0.644 for other-cause death. Conclusions Stage showed a steeper gradient for bladder-cancer mortality, whereas age differentiated other-cause mortality more strongly than stage. Their changing balance may inform the timing and focus of survivorship assessment. These population-level patterns do not define treatment thresholds, and age cannot replace an individual geriatric assessment.