Bioequivalence and Bioavailability Forum

Dear Ram!

❝ Can you give the rationale for applying "Williams' Design" in 4 way cross

❝ over study.

See this post for an example of a 6×3 (6 sequence, 3 treatment) Williams' design (including some references), and this one as well..

Williams' designs are balanced for eventual 1^st-order carry-over effects.
Each treatment is preceeded once by each of the 3 other treatments or is given as the 1^st treatment of the study.
In a set of classical Latin square design (ABCD|BCDA|CDAB|DABC), each treatment (when not the 1^st treatment) is always preceeded by the same treatment (B by A, C by B, and D by C), which can bias the estimates in presence of differential 1^st-order carry-over.

❝ Does any guidance recommends.

To my knowledge only Brazil ANVISA's, but please use it unconditionally - otherwise you may run into statistical troubles.