|
Abstract:
|
The FDA classifies statistical methods for assessment of in vitro bioequivalence tests for orally inhaled products (OIPs) as either profile or non-profile analysis. Profile analysis related to the drug particle size distribution (PSD), commonly measured by cascade impactor, is one of the critical attribute tests previously and currently being considered by the FDA, in which profile comparisons between test and reference products are based on chi-square differences. An adequate assessment of profile comparison methods prompted efforts to generate realistic simulated cascade impactor profiles which take into account the mean, the variance and the inter-site correlation of mass recoveries between cascade impactor deposition sites. However, much observed profile data do not appear to follow a multivariate normal distribution, imposing significant challenges to simulating the inter-site correlations. This talk discusses considerations and approaches for methods that can be employed to transform the observed profile data into an approximately multivariate normal distribution using examples which realistically mimic the non-normal mass distribution exhibited by many OIPs.
|
