Performance of different population pharmacokinetic algorithms.

Area of Science:

• Pharmacokinetics
• Pharmacometrics
• Drug Development

Background:

• Increased focus on population pharmacokinetics (PK) for drug development, spurred by FDA's
• Emergence of new algorithms to meet this demand.
• Need for comparative analysis of these novel PK algorithms.

Purpose of the Study:

• Compare the performance of novel algorithms iterative-2-stage (ITS) and maximum likelihood expectation maximization (MLEM) in ADAPT 5.
• Evaluate these algorithms against established methods like standard-2-stage, Iterative-2-Stage (IT2S), and first-order conditional estimate (FOCE).

Main Methods:

• Simulation of 29 clinical trials with diverse designs.
• Application of ITS, MLEM, IT2S, and FOCE algorithms to analyze simulated data.
• Comparison of estimated parameters against true values using bias and imprecision metrics for population and individual PK parameters and variances.

Main Results:

• All nonlinear mixed-effect modeling approaches outperformed standard-2-stage analyses.
• MLEM demonstrated superior prediction of PK and variability parameters compared to IT2S and ITS.
• MLEM and FOCE provided better estimation of residual variability.
• MLEM exhibited significantly fewer shrinkage issues in variance estimation compared to FOCE.

Conclusions:

• The MLEM algorithm in ADAPT 5 is superior to IT2S and ITS for predicting PK parameters, variances, and residual variability.
• MLEM is comparable to FOCE but with a significant advantage in reducing variance estimation shrinkage.
• The number of samples in the expectation maximization step does not impact MLEM results.