[Can we use a Bayesian method to build a pharmacokinetic population in two steps?]

Area of Science:

• Pharmacokinetics
• Pharmacometrics
• Drug Dosing

Background:

• Traditional nonlinear regression methods for pharmacokinetic analysis can be complex and computationally intensive.
• Population pharmacokinetic (PopPK) models aim to describe drug variability across patient populations.
• Bayesian estimation provides an alternative approach for parameter estimation in PopPK models.

Purpose of the Study:

• To evaluate the utility of a two-stage pharmacokinetic population model using Bayesian estimation (MAP) versus traditional nonlinear regression (MLS).
• To compare the prediction accuracy and pharmacokinetic parameter estimations between MAP and MLS methods.
• To assess the feasibility of using Bayesian methods for adaptive drug dosage regimen control.

Main Methods:

• Retrospective analysis of 32 patient files (mean age 82 years).
• Comparison of amikacin serum level predictions using MAP and MLS.
• Evaluation of pharmacokinetic parameter values (Vd, Kslope, Kel, t1/2) for one- and two-compartment models.

Main Results:

• For a one-compartment model, both MAP and MLS showed comparable prediction performance (r=0.90-0.91) and precision, with a slight difference in systematic error favoring MAP (p<0.05).
• For a two-compartment model, MAP demonstrated superior long-term prediction accuracy (4-8 days) and precision compared to MLS (p<0.01).
• Bayesian estimation did not significantly influence pharmacokinetic parameter estimation for a one-compartment model.

Conclusions:

• The Bayesian method, when applied to a two-stage pharmacokinetic population model, is a viable alternative to nonlinear regression.
• This approach enables accurate prediction of drug serum levels and facilitates adaptive control of drug dosage regimens.
• The findings support the use of Bayesian estimation for optimizing therapeutic drug monitoring, especially in elderly populations.