Estimation of individual pharmacokinetic parameters using maximum a posteriori Bayesian method with D-optimal sampling strategy

This study was aimed to develop a maximum a posteriori Bayesian (MAPB) estimation method to estimate individual pharmacokinetic parameters based on D-optimal sampling strategy.  Meanwhile, the performance of MAPB was compared with the multiple linear regression (MLR) method in terms of accuracy and precision.  Pharmacokinetic study of pioglitazone was employed as the example case.  The population pharmacokinetics was characterized by nonlinear mixed effects model (NONMEM).  The sparse sampling strategy (1 − 4 points) was identified by D-optimal algorithm using WinPOPT software.  The simulated data generated by Monte Carlo method were used to access the performance of MAPB and MLR.  As the number of samples per subject decreased, the accuracy and precision of MAPB method tended to get worse.  The estimation for CL and V by MAPB using D-optimal two-point design had less bias with low inter-individual variability, and had more bias and imprecision with high residue variability.  The estimation of AUC by MAPB using D-optimal 2 points design had similar accuracy and precision to MLR.  However, MAPB estimation was better than MLR while adjusting the sampling time to one hour.  Overall, the MAPB method had similar predictive performance as MLR, but MAPB could provide more pharmacokinetic information with higher sampling flexibility.