Frederic Y. Bois ^1* Céline Brochot ²

• ¹ Simcyp (United Kingdom), Sheffield, United Kingdom
• ² Certara UK Limited, Canterbury, Kent, United Kingdom

Introduction: In virtual bioequivalence (VBE) assessments, pharmacokinetic models informed with in vitro data and verified with small clinical trials data are used to simulate otherwise unfeasibly large trials. Simulated VBE trials are assessed in a frequentist framework as if they were real, despite the unlimited number of virtual subjects they can use. This may control consumer risk adequately but imposes unnecessary risks to producers. We propose a fully Bayesian model-integrated VBE assessment framework which circumvents those limitations.Methods: We illustrate our approach with a case study on a hypothetical paliperidone palmitate generic long-acting injectable suspension formulation, using a validated population pharmacokinetic model published for the reference formulation. BE testing, study power, type I and type II error analyses, or their Bayesian equivalent, and safe-space analyses are demonstrated.The fully Bayesian workflow is more precise than the frequentist one. Decisions about bioequivalence and safe space analyses in the two workflows can differ markedly because the Bayesian analyses are more accurate.Discussion: A Bayesian framework can both control consumer and minimize producer risk adequately. It rewards data gathering and model-integration to make the best use of prior information. The frequentist approach is less precise, but faster to compute, and can still be used as a first step to narrow down the parameter space to explore in safe-space analyses.