Bayesian re-analysis: Comparing radical cystectomy with trimodality therapy in patients with muscle invasive bladder cancer.

Abstract

737 Background: Randomized trials comparing radical cystectomy (RC) and trimodality therapy (TMT) in muscle invasive bladder cancer (MIBC) have not been feasible. In 2023, Zlotta et al. published a retrospective multi-institutional analysis evaluating metastasis free survival (MFS) and overall survival (OS) outcomes in patients treated with RC or TMT. Zlotta et al. is generally considered the highest quality data available for this comparison, but the results were presented using traditional frequentist statistics which can be challenging to interpret. The purpose of this study was to use Bayesian re-analysis of Zlotta et al. to convert its frequentist summary statistics into outcome probabilities, enabling a more intuitive treatment comparison between RC and TMT. Methods: The Zlotta et al. study results comparing RC and TMT were described by estimating the probability of any benefit (HR<1) or a minimal clinically meaningful (MCR) benefit (HR<0.8). An informative neutral prior distribution was defined by surveying bladder cancer experts on the comparison of RC and TMT use in bladder cancer patients included in Zlotta et al. (cT2–T4N0M0, <7cm, no or unilateral hydronephrosis, no CIS, average age 70). Using a neutral prior of no advantage to either treatment and methods adapted from Wijeysundera et al., we calculated probabilities of a benefit in MFS and OS. We performed sensitivity analyses by varying the strength of the study-specific priors’ neutrality assumption and by using an alternative informative neutral prior developed for oncology comparisons but not specifically for MIBC. Results: The expert-informed informative neutral prior assumption assumed a 70% chance that any differences with respect to any cancer control or overall survival outcome would be less than the minimal clinically significant thresholds. The posterior probabilities of the primary analysis are presented in the table. Sensitivity analyses of the estimates of any benefit for MFS and OS varied by up to 3%, while estimates of a MCR benefit for MFS and OS varied up to 15%. Conclusions: The best available evidence suggests that in selected MIBC patients there is a >96% chance that TMT is associated with any OS advantage over RC but a <59% chance that the benefit reaches a minimal clinically meaningful benefit threshold. There is >65% chance that TMT is associated with any MFS advantage over RC but <13% chance of a minimal clinically meaningful benefit. Bayesian re-analysis of outcome estimate from Zlotta et al. 2023. Probability of any benefit of TMT over RC (HR<1) Probability of any benefit of RC over TMT (HR>1) Probability of a MCR benefit of TMT over RC (HR<0.8) Adjusted MFS Zlotta et al. Outcomes IPTS SHR 0.89 (95% CI 0·67–1·20) 73% 27% 12% PSM SHR 0.93 (95% CI 0.71–1.24) 66% 34% 7% Adjusted OS IPTS SHR 0.70(95% CI 0.53–0.92) 98% 2% 58% PSM SHR 0.75 (95% CI 0.58–0.97) 97% 3% 44%