Sequential bioequivalence trial designs with increased power and controlled type I error rates

Methods to implement two-stage designs in two-treatment, two-sequence, and two-period crossover bioequivalence studies have only recently been developed. The two-stage methods have so far only been described for a targeted study power of 80%. Since it is sometimes desirable to increase the targeted power to 90%, this study identifies sets of alphas that work for the recently developed two-stage methods while controlling type I error rates around 5% for assumed geometric mean test/reference ratios of 0.90 and 0.95 at targeted power of 90%, and provides a characterization of the methods in terms of the resulting sample sizes and power. Depending on the actual variability and the chosen sample size at stage 1, the actual power will be between 83% and 100%. The previously characterised methods at target power 80% as well as 90% result in trivial inflation of type 1 error, but the type 1 error inflation at 90% target powers with decreased alpha at the second stage result in slightly less inflation. These results may be useful for applicant wishing to achieve increased power in bioequivalence trials without a penalty for type I error rates.