A Bayesian Perspective on Estimation of Variability and Uncertainty in Mechanism-Based Models

As our understanding of biological systems and pharmacology increases, the knowledge required to develop predictive mathematical models based on underlying biological/pharmacological processes becomes only a mouse-click away. This has led to an increase in the use of mechanism-based models for prediction of drug pharmacokinetics (PK) and pharmacodynamics (PD). Because of the high dimensionality of these models and the limitations of typical non-linear mixed effects (NLME) estimation algorithms, simultaneous estimation of fixed and random effect parameters has been difficult. In addition, the incorporation of prior information can be subjective, often requiring fixing of numerous model parameters. Markov-chain Monte Carlo (MC MC) Bayesian estimation algorithms, while relatively computer intensive, are not as sensitive to increases in model complexity as traditional estimation approaches, permit transparent and robust incorporation of prior information, and allow estimation of fixed and random effect parameters for mechanism-based models. The concepts of Bayesian NLME approaches will be discussed, and examples of Bayesian mechanism-based PK and PD models will be presented.