Janet Alvarado, Senior Specialist for Quantitative Sciences, Merck
Nelson Lee Afanador, PhD, Associate Director for Quantitative Sciences, Merck
It is a regulatory requirement to demonstrate product comparability before and after process changes in vaccines, biologics and pharmaceutical manufacturing occur. The demonstration of comparability, as described in ICH Q5E, does not mean that pre- and post-change products have to be identical, but rather highly similar. In this work, we explore the use of a novel approach for the evaluation of comparability. This multivariate approach makes use of the dissimilarity matrix from a random forest model as the input to a principal coordinate analysis as a way to examine similarity of observations from pre- and post- process changes. Confidence ellipses are used to assist in a visual assessment of comparability, where highly overlapped ellipses would suggest similarity. We assess variable importance to shed light on what variables are contributing the most to the separation between groups, which subsequently helps determine if a more focused approach is needed. Several case studies are presented, including a live demonstration in JMP.
