Bioequivalence and Bioavailability Forum

Hi Lindsay!

❝ Could post-hoc power analyses help in a pilot study with two test formulations compared to 1 reference in deciding which formulation to go forward with in a pivotal study?

No – what would you expect to gain? In any higher-order Xover (e.g., 3×3 Latin Squares, 6×3 Williams’ design) you get only a common estimate of the intra-subject variance. So it boils down to assessing the point estimates T₁/R and T₂/R. Select the one closer to 100%. Theoretically [sic] you could run a replicate design to assess the intra-subject variances of formulations as well. If the test/reference ratios are similar (or one is the reciprocal of the other) you could select the one with lower variability. But with two tests you will end up with a f_**ing complicated replicate design with many periods, even if you a apply a modified Balaam’s design. Haven’t seen any such a study in 30+ years…