6. Oktober 2026, von Antje Wagner

Foto: RG Wicha

We are delighted to share our recent publication, “Evaluation of a model averaging algorithm for model-informed precision in the context of parameter misspecifications,” published in the Journal of Pharmacokinetics and Pharmacodynamics, by Sandra Witta and Sebastian Wicha.

One of the biggest challenges faced when integrating model-informed precision dosing (MIPD) into routine clinical care is the question: which model best describes each individual patient?
Model averaging, developed by our alumni, David Uster, aimed to resolve this gap by combining multiple models simultaneously. However, as model libraries for a given drug expand, we are presented with a new question: does it matter which models we combine, especially when some are misspecified?
In our simulation study, we looked at how three types of parameter misspecifications impact model averaging performance:

• Structural parameter (on clearance)
• Residual error
•  Interindividual variability (on clearance)

Key findings:

• Clearance misspecification: MAA showed the greatest benefit, as combining models (even poorly performing ones) with opposing bias improved performance.
• Residual error misspecification: MAA tended to down-weight models with high residual error.
• IIV misspecification: Models with higher IIV were favored.

This study highlighted how understanding model misspecifications can inform which models to combine in MAA and make MIPD more reliable in clinical practice.

For further details, please visit the link below:
https://doi.org/10.1007/s10928-026-10064-5