Population PBPK modeling using parametric and nonparametric methods of the Simcyp Simulator, and Bayesian samplers.

Area of Science:

• Pharmacokinetics and Drug Metabolism
• Computational Biology and Bioinformatics
• Systems Pharmacology

Background:

• Physiologically-based pharmacokinetic (PBPK) models rely on in vitro to in vivo extrapolation for parameter estimation.
• This extrapolation can introduce uncertainty, highlighting the need for clinical data integration.
• Population pharmacokinetic (PopPK) frameworks are crucial for handling high interindividual variability and sparse data.

Purpose of the Study:

• To compare the performance of different inferential frameworks for parameter estimation in PBPK models.
• To evaluate quasi-random parametric expectation maximization (QRPEM), nonparametric adaptive grid estimation (NPAG), and Bayesian methods (Metropolis-Hastings and Hamiltonian Monte Carlo).

Main Methods:

• A minimal PBPK model was applied to a canonical theophylline dataset.
• Four distinct inferential frameworks were employed: QRPEM, NPAG, Bayesian Metropolis-Hastings (MH), and Hamiltonian Markov Chain Monte Carlo (HMCMC).
• Population and individual parameter estimates were compared across methods.

Main Results:

• QRPEM and NPAG yielded consistent population and individual parameter estimates.
• These estimates largely agreed with those obtained from Bayesian methods.
• MH simulations demonstrated faster run times compared to other methods, with similar overall performance.

Conclusions:

• QRPEM and NPAG are reliable methods for parameter estimation in PBPK modeling, offering results comparable to Bayesian approaches.
• Bayesian methods, particularly MH, can offer computational efficiency.
• Integrating clinical data through PopPK frameworks enhances PBPK model accuracy and robustness.