Abstract
This paper introduces a two-stage approach for evaluation of bioequivalence, where, in contrast to the designs of Diane Potvin and co-workers, two stages are mandatory regardless of the data obtained at stage 1. The approach is derived from Potvin's method C. It is shown that under circumstances with relatively high variability and relatively low initial sample size, this method has an advantage over Potvin's approaches in terms of sample sizes while controlling type I error rates at or below 5% with a minute occasional trade-off in power. Ethically and economically, the method may thus be an attractive alternative to the Potvin designs. It is also shown that when using the method introduced here, average total sample sizes are rather independent of initial sample size. Finally, it is shown that when a futility rule in terms of sample size for stage 2 is incorporated into this method, i.e., when a second stage can be abolished due to sample size considerations, there is often an advantage in terms of power or sample size as compared to the previously published methods.