Power calculations for the sequential parallel comparison design with continuous outcomes

摘要

In the sequential parallel comparison design (SPCD) in stage 1, placebo subjects are randomized between placebo and an experimental therapy. In stage 2, stage 1 placebo non-responders are re-randomized between placebo and an experimental therapy. We give the formula for power/sample size calculations for two test statistics for the SPCD. The first one omits the correlation between the estimated treatment effects from the two stages, and the second one does not. We discuss how to construct a confidence interval for the weighted average of the treatment effects that has proper coverage.