In comparative bioavailability trials, the focus in the determination
of bioequivalence has been to show that the average bioavailability on
a test formulation is similar to the average bioavailability on a ref
erence formulation. It has been argued that this is not adequate to in
sure interchangeability of formulations. We introduce a moment-based c
riterion which ensures interchangeability for at least a certain propo
rtion of the population. An algorithm is given for determining the con
stant in the criterion. An example is used to illustrate how the crite
rion may be applied.