Bioequivalence revisited

The FDA permits marketing of a generic formulation of a drug G for the same indications as a standard preparation S if one can show that G is bioequivalent to S. Present implementation requires convincing evidence that the population mean difference in bioavailability (drug exposure) between the two preparations lies within specified bounds. The basis for this standard does not appear to involve a comprehensive model for the dose-response relationship, or consideration of clinical issues, notably (i) whether a patient is to commence on the drug or to switch from an established regimen to a new one; or (ii) that the risk of inequivalence relates to uncertainty of outcome. In this paper, I propose a comprehensive model for dose response and a tentative model for risk that addresses these issues. Specifically, I propose two new measures of bioequivalence which are based on these models, which differ in the two clinical circumstances above, and which respond to both bias and variance of outcome. I present two examples, and some simulations of the application of the new measures.