A comparison of adaptive designs in clinical trials : when multiple treatments are tested in multiple stages

In recent times, there has been an increasing interest in adaptive designs for clinical trials. As opposed to conventional designs, adaptive designs allow flexible design adaptation in the middle of a trial based on accumulated data. Although various models have been developed using both frequentist and Bayesian perspectives, relative statistical performances of adaptive designs are somewhat controversial and little is known about those of Bayesian adaptive designs. Most comparison studies also focused on single experimental treatment rather than multiple experimental treatments. In this report, both frequentist and Baysian adaptive designs were compared in terms of statistical power by a simulation study, assuming the situation when multiple experimental treatments are tested in multiple stages. The designs included in the current study are group sequential design (frequentist), adaptive design based on combination tests (frequentist), and Bayesian adaptive design (Bayesian). Based upon the results under multiple scenarios, the Bayesian adaptive design showed the highest power, and the design based on combination tests performed better than group sequential designs when proper interim adaptation could be conducted to increase power.