Simulation Informed Design and Performance of In Vitro Bioequivalence Trials for Particle Size Distributions

Abstract

This study used statistical simulations to investigate the performance of the population bioequivalence test applied to image-based particle size measurements (such as morphologically directed Raman spectroscopy) and methods for designing in vitro bioequivalence trials using prior information. Simulations of in vitro population bioequivalence trials were conducted across a range of representative D₅₀ (number-weighted median particle diameter from a log-normal particle size distribution) and span (which is defined as $\frac{D_{90}-{D}_{10}}{D_{50}}$ where D₉₀ and D₁₀ are the number-weighted 90th and 10th percentiles in particle diameters sampled from a log-normal particle size distribution) values respectively. The performance of the population bioequivalence test in the simulations was driven by an interplay between overall test variability and the widening or narrowing of the bioequivalence region due to variance terms in the test statistic definition. These findings were dependent upon differences in the variability of D₅₀ and span and may generalise to a wider range of in vitro metrics. Trial design optimisation using power and assurance approaches followed patterns consistent with these findings. As more novel scientific methods are applied to the development of complex generic drug products, the procedures outlined in this study may be used at the inception stage of future in vitro bioequivalence trials to reduce the risk of conducting costly trials with low probabilities of success.