The transitivity of bioequivalence testing: potential for drift

During the drug development process bioequivalence studies are required as formulations are refined. At the end of this chain of innovator-conducted studies, generic manufacturer(s) conduct bioequivalence studies comparing their generic(s) to the current marketed formulation. The question we pose is: How transitive is bioequivalence? That is, if formulation B is bioequivalent to formulation A, and C to B, what can one say about the bioequivalence of C and A? We consider 1 (A-B), 2 (A-C), 3 (A-D), and 6 (A-G) bioequivalence steps and restrict attention to the current practice of average bioequivalence with an (80%, 125%) equivalence criterion. For the cases of no to small differences in bioavailability our results suggest that one can be fairly confident of the bioequivalence of formulation C to formulation A, and of D to A, but not of G to A. The transitivity of bioequivalence decreases with additional steps and with increasing power of the individual bioequivalence studies.