Risk-Based Similarity Testing for Potency Assays Using MCMC Simulations

In potency assays, the relative potency (RP) is the measure of the biological activity of a test product relative to a reference standard. It is calculated as the horizontal difference between the log(concentration)–response curves of each product. Parallelism (or similarity) between those curves is required for the RP to be meaningful. The literature proposes equivalence tests for similarity, based on the ratio of the curve parameters. Deriving equivalence margins (or a zone of declared similarity) to avoid rejecting good curves (lab risk) or accepting bad curves (consumer risk) often requires historical data. In this article, we propose a three-step derivation of the similarity zone to control for both consumer and lab risk, even when historical data are not available. In addition, we propose a way to test for parallelism that accounts for the correlations among the ratios of curve parameter estimates.