Virtual Bioequivalence with Clinical Variability: Simulation Methodology, Success Criteria and Interpretation

As we race towards integrating AI/ML approaches into the drug development process with objectives of accelerating development, improve quality and reduce cost, it becomes crucial to establish a framework for decision making and develop decision criteria. This is essential if we intend to rely on AI/ML to make critical decisions that go beyond the obvious applications or “low hanging fruit”.
Drug formulation development involves numerous clinical studies for evaluating clinical performance of various formulations. Clinical bioavailability/bioequivalence (BA/BE) studies are required to demonstrate lack of differences between enabling formulations to commercial formulations at various strengths if applicable. To understand all the elements of clinical performance of formulations, PBPK models have risen to the occasion to potentially avoid a large number of clinical studies. Advances in PBPK platform development, proprietary and open-source software coupled with the core philosophy of model building which incorporates physiological and physical processes impacting formulation performance makes it a potent partner for adapting to AI/ML driven frameworks. AI/ML based clinical trial design and data analysis will help establish/verify mechanism based PBPK models. Overall better clinical trial design and more mechanism based PBPK modeling can facilitate reducing clinical trials and eliminate exposure to drug in healthy study participants who derive no medical benefit. However, to fully leverage the capability of AI/ML, it is necessary to establish simulation methodology and decision criteria.
This presentation will focus on the utilizing PBPK models for conducting virtual bioequivalence (VBE) trials outlining the simulation strategy, sample size assessment and risk-tolerance based decision criteria. PBPK models offer a solid platform for simulating large virtual populations and the respective PK profiles approximating the inter-subject variability in the population based on the variance of the physiological parameters. However, the ability of PBPK models to replicate the observed clinical variability within subjects or intrasubject coefficient of variation (ICV) is not fully explored and often not adequately predicted. In this presentation, a simulation methodology with PBPK models incorporating clinically estimated ICV will be described. Different case studies with examples of using VBE to establish BE between 1) different strengths (with differing dissolution profiles) of an immediate release (IR) formulation, 2) different formulation platforms (IR Vs modified release) and 3) adult vs pediatric formulations will be described. A limitation for developing pediatric formulations is testing clinical performance of age-appropriate formulations in target populations when bridging enabling and commercial formulations. An example of using VBE to establish BE between different formulations in pediatrics will also be described. The VBE framework described here will be an essential component for adaptation to AI/ML approaches to gain efficiency, accelerate formulation development and enable risk-tolerance based decisions.