A multivariate test for population bioequivalence

Abstract

In this article, we propose a multivariate generalization of the criteria for testing univariate population bioequivalence. Recently, a number of approaches for testing multivariate equivalence have appeared in the literature. Most of them consider a multivariate equivalence region, which implies simultaneous comparison of means in each dimension. In contrast, our proposal combines a comparison of means and a comparison of variances into a single aggregate criterion, using the trace of the covariance matrix as a scalar measure of the total variability. We use a confidence interval approach to multivariate population bioequivalence testing, similar to the univariate case. Two versions of the modified large-sample confidence interval for the linearized multivariate criterion are constructed. In a simulation study, we evaluate the empirical coverage of these confidence intervals and rejection rates of the corresponding tests in finite samples. The proposed methodology is illustrated with an example of testing equivalence of the spray pattern of nasal sprays.

Construction of confidence limits about effect measures: A general approach

“…We also recognize that this procedure has previously been applied in special cases, including, for example, the construction of a confidence interval for variance components [12], the construction of limits for a difference between two normal means [13] and for a difference between two kappa statistics [14]. It has further been applied to the problem of assessing bioequivalence [15][16][17][18] and is closely related to methodology proposed for the analysis of binary data in a variety of contexts [9,[19][20][21][22][23].…”

Power Estimation in Multivariate Analysis of Variance

“…Thus, it makes no sense to use retrospective power to "enhance" the interpretation of a significance test (Baguley, 2004). In fact, for post hoc analysis, likelihood and Bayesian methods (Gelman, Carlin, Stern, & Rubin, 2004) or equivalence testing (Berger & Hsu, 1996;Chervoneva, Hyslop, & Hauck, 2007;Schuirmann, 1987) should be preferred. Therefore attention should be focused on the effect size.…”

Evaluating manufacturing process profile comparability with multivariate equivalence testing: Case study of cell‐culture small scale model transfer

“…Another general approach specified by Chervoneva et al was to convert a multivariate problem into a univariate space by using aggregate equivalence criterion (e.g., the profile approaches for cascade impaction and dissolution profiles). In dissolution profile comparison, the US FDA guideline presents a model‐independent approach using a dissimilarity factor (a measurement of the relative error between two profiles) and a similarity factor (a measurement of the similarity in the percent dissolution between two profiles) across a suitable time.…”