Some Characteristics of the Varying-Stage Adaptive Phase II/III Clinical Trial Design

Conventionally, adaptive phase II/III clinical trials are carried out with a strict two-stage design. Dong (Stat Med 33(8):1272–1287) recently proposed a varying-stage adaptive phase II/III clinical trial design, in which the number of further investigational stages is determined based upon data accumulated to the interim analysis. In this design, following the first stage, an intermediate stage can be adaptively added to obtain more data, so that a more informative decision could be made. This design considers two plausible study endpoints with one of them initially designated as the primary endpoint. Based on the interim results, another endpoint can be switched as the primary endpoint. Dong (Stat Med 33(8):1272–1287) has showed relations of design parameters (e.g., thresholds and percent of alpha allocated in the two-stage setting) as well as the trial design properties under the alternative hypotheses for both plausible endpoints. Here, we explore characteristics of the design when the alternative hypothesis for only one of the two endpoints is true, and the treatment effect for another endpoint is null (an extremely worst case) or lower than what was anticipated per trial design. The simulations show that the statistical power of the varying-stage adaptive phase II/III clinical trial design (Dong, Stat Med 33(8):1272–1287) is less sensitive to a low realized treatment effect.