A Comparative Study of Confidence Interval and Confidence Region Methods in Determination of Bioequivalence between Two Formulations

Abstract

The currently accepted statistical method for declaring bioequivalence (BE) is the Classical 90% Confidence Interval (Classical CI) using log transformed AUC and C_max data. This method is based on the population mean of the ratio of test and reference products. In order to be declared BE, the results after anti-log transform must be contained within the interval 0.80-1.25. In recent years, alternative methods such as Locke's Exact Confidence Interval (Locke's CI) and Chow & Shao's Confidence Region (C&S CR) methods have been proposed. The objective of this study is to compare the current reference method to the alternative CI and CR methods, including C&S modified limit (derived from C&S CR), using the 22 data sets obtained from 10 bioequivalence studies. Both the raw (original) and log-transformed data were used where applicable. The results indicated that:

1. 1.

Using log-transformed data offered little advantage over using the raw data for Classical CI,

2.

Other than minor numerical differences, the same conclusions can be reached using either the Locke's CI or Classical CI, and

3.

In contrast, the C&S CR method offered a distinct advantage in terms of percentage passing the BE criteria and number of fail reversals where the same data set failed the reference method, but passed the comparative method.