Abstract

The determination of bioequivalence is very important in the pharmaceutical industries because the regulatory agencies like the Food and Drug Administration (FDA) allow a generic drug to be marketed only if its manufacturer can demonstrate that the generic product is bioequivalent to the brand-name product. Up to date, there is a lack of widely accepted statistical procedure for assessing population bioequivalence. We propose multilevel models (MLMs) for evaluating and estimating parameters to assess population bioequivalence (PBE), and compare statistical properties of PBE estimators between MLMs and current approaches recommended by the FDA-the method of moment (MOM) and REML. The approach developed is illustrated using a real data set from the FDA. Statistical properties of MLM estimators are further explored using simulation studies as compared with MOM and restricted maximum likelihood (REML) estimators. The performance of MLM appeared to be much comparable to the existing REML procedure. The results suggest that MLM estimators that are fully comparable with REML estimators can be an adequate approach for assessing PBE. The MLMs approach proposed in the study provides an alternative and yet more flexible and powerful method than existing methods in assessing bioequivalence (BE) for complex study designs and data structures.

Key words: Population bioequivalence, multilevel models, simulation, estimation procedure, restricted iterative generalized least square (RIGLS), restricted maximum likelihood (REML), method of moments (MOM), food and drug administration (FDA).