Bioequivalence and Bioavailability Forum

Hi Kotu,

whatever you might call it, I agree with ElMaestro that this is essentially a Two-Stage Design (even if you aim at early stopping in the interim) – which would require adjusting α in order to maintain the Type I Error.
α_adj suitable for replicate designs (both for reference-scaled and conventional ABE) depends on the assumed GMR, target power, design (and the scaling method if applicable). The numerical values are currently un­known. See this thread for background information.

BTW, agencies consider analyzing only part of the data (in order to cut costs) unethical.

You estimated the sample size with 70 subjects and showed BE with only 40. Such an outcome might be pure chance. Please assess BE for the complete data set of 70.

---

Edit after reading Detlew’s post above: I would emphasize part of the EMA’s last sentence “It should be planned that all treated subjects should be included in the analysis […]”.
Four members of the forum warned you last year that such a design is not acceptable. Have you forgotten that?