Prospective comparison of Bayesian and frequentist adaptive clinical trials : The SHADOW - SHINE project

Importance: Bayesian adaptive trial design has the potential to create more efficient clinical trials. However, one of the barriers to the uptake of Bayesian adaptive designs for confirmatory trials is limited experience with how they may perform compared to a frequentist design. Objective: Compare the performance of a Bayesian and a frequentist adaptive clinical trial design. Design: Prospective observational study comparing two trial designs using individual patient level data from a completed stroke trial, including the timing and order of enrollments and outcome ascertainment. The implemented frequentist design had group sequential boundaries for efficacy and futility interim analyses when 90-days post-randomization was met for 500, 700, 900, and 1,100 patients. The Bayesian alternative utilized predictive probability of trial success to govern early termination for efficacy and futility with a first interim analysis at 500 randomized patients, and subsequent interims after every 100 randomizations. Setting: Multi-center, acute stroke study conducted within a National Institutes of Health neurological emergencies clinical trials network. Participants: Patient level data from 1,151 patients randomized in a clinical trial comparing intensive insulin therapy to standard in acute stroke patients with hyperglycemia. Main Outcome(s) and Measure(s): Sample size at end of study. This was defined as the sample size at which each of the studies stopped accrual of patients. Results: As conducted, the frequentist design passed the futility boundary after 936 participants were randomized. Using the same sequence and timing of randomization and outcome data, the Bayesian alternative crossed the futility boundary about 3 months earlier after 800 participants were randomized. Conclusions and Relevance: Both trial designs stopped for futility prior to reaching the planned maximum sample size. In both cases, the clinical community and patients would benefit from learning the answer to the trial9s primary question earlier. The common feature across the two designs was frequent interim analyses to stop early for efficacy or for futility. Differences between how this was implemented between the two trials resulted in the differences in early stopping.