Handling missing data in a duloxetine population pharmacokinetic/pharmacodynamic model - imputation methods and...

Area of Science:

• Pharmacometrics
• Clinical Pharmacology
• Biostatistics

Background:

• Pharmacokinetic/pharmacodynamic (PK/PD) modeling and simulations (M&S) are crucial for drug development.
• Omitting patient dropouts can lead to inaccurate parameter estimation and clinical trial simulations (CTS).

Purpose of the Study:

• To assess the impact of various missing data imputation methods on PK/PD model results.
• To develop and evaluate a selection model for jointly modeling dropout and efficacy in CTS.

Main Methods:

• Applied single, multiple, and pattern mixture imputation methods to a duloxetine PK/PD model.
• Developed a selection model incorporating dropout probability.
• Conducted CTS with a hypothetical drug to simulate dropout effects.

Main Results:

• Study completion rate was 75%, with non-random dropouts.
• Imputed model parameters generally remained within 40% of non-dropout model parameters.
• CTS including dropout showed slightly lower median pain scores and coefficient of variation.

Conclusions:

• Missing data imputation had minimal impact on the original population PK/PD analyses.
• Sensitivity analyses for dropouts are recommended for M&S.
• Selection models demonstrate utility in CTS for accounting for dropout.