Statistical Evaluation of the In Vitro Release Comparison Test for Semisolid Release Profiles

Abstract

The United States Food and Drug Administration (FDA) proposed an in vitro release comparison test to assess the similarity among two semisolid drug release profiles. In this paper we investigate the statistical properties of the test and evaluate the power (or type Ⅰ error rate) which is the passing probability of the drug products while the release rate of the prechange lot and that of the postchange lot are similar or dissimilar. The type Ⅰ error rate is the consumer's risk and the power is one minus type Ⅱ error rate which is the drug factory's risk. In addition, a bootstrap-based approach is proposed. Based on simulation studies, the empirical type Ⅰ error rate and power are calculated for the FDA method and bootstrap procedure under four different distributions of log-slope data, including normal, uniform, double exponential and Cauchy distribution. Furthermore, a numerical example is used to illustrate the application of the proposed procedure. The proposed approach has a smaller type Ⅰ error rate than that of the FDA method and has similar power to that of the FDA method.