Bioequivalence and Bioavailability Forum

Dear ElMaestro, dear MGR!

Both of your described designs are of course possible ones.
But both are not usable if you think in terms of scaled ABE for NTID's as described in the Warfarine guidance. The algo described there relies on a full replicate design because you have to evaluate the (1 - α)100% CI for σwT/σwR, the ratio of sd's of intra-subject variabilities of Test and Reference.

For "normal" scaled ABE aimed to a FDA submission there are some (IMHO) unresolved issues. Just to name some:

• How to estimate the sample size for scaled ABE for these designs?
• How exactly to adapt the algorithm described in the FDA Progesterone guidance to both suggested designs?
• Pairwise evaluation like the EMA suggestion for 3×3 or 4×4 cross-over designs doesn't really help since for the RR evaluation all data have to be used if such designs should have any advantage over 2 separate studies.

I personally wouldn't go with any design without knowledge about how to plan and how to evaluate!