All Times EDT
Keywords: Response Adaptive Randomization, Bayesian methods, composite endpoint
If treatment allocation of patients during a trial is based on the observed responses and can be adapted sequentially, it could minimize the expected number of failures and maximize patients’ benefits. In response-adaptive randomization (RAR), future treatment allocation ratios are determined based on the past treatment assignments and the response of patients collected before the decision time point. In this study, we developed a Bayesian RAR design targeting the endpoint of organ support-free days (OSFD) for patients admitted to the intensive care units (ICU). The OSFD is a mixture of mortality and morbidity assessed by the number of days free of organ support. In the past, researchers treated OSFD as an ordinal outcome variable that an arbitrary low number, for example, -1 or -100, is assigned to those who died in the ICU. We propose a novel RAR design for a composite endpoint of mortality and morbidity, e.g., OSFD, by using a Bayesian mixture model with a Markov chain Monte Carlo sampling to estimate the posterior probability of OSFD and determine treatment allocation ratios at each interim. Simulation was conducted to compare the performance of our proposed design under various randomization rules and different alpha spending functions. The results show that our RAR design using Bayesian inference benefits more patients while assuring adequate power for the target trial.