External validation of the population pharmacokinetic model of meropenem in patients undergoing neonatal extracorporeal membrane oxygenation and continuous renal replacement therapy
Pavla Pokorná, Danica Michaličková, Jan Bělohlávek, Jonas BernerIntroduction
There is a lack of data supporting rationale drug use with extracorporeal membrane oxygenation (ECMO). The aim of this study was to externally validate a previously developed population pharmacokinetic model of meropenem in neonatal ECMO and continuous renal replacement therapy.
Methods
A total of twenty-eight neonates with a body weight of 3.81 (3.45–4.11) kg, median (interquartile) and a postnatal age of 3 (2–4) days were enrolled. One hundred plasma concentrations of meropenem were used for external validation by a published population pharmacokinetic analysis model using NONMEM V7.3.0 (ICON Development Solutions, Ellicott City, MD, USA) and PsN v3.4.2, both running in Pirana 2.9.0. Prediction error analyses, and NPDE diagnostics were performed to assess the model’s extrapolative capability. Bayesian forecasting was conducted to determine the impact of prior concentration information on predictive performance.
Results
Population-based predictions showed minimal overall bias but limited accuracy, with substantial variability between predicted and observed concentrations and less than half of predictions falling within an acceptable deviation range. After incorporating one prior measured concentration through Bayesian forecasting, predictive performance improved markedly. Most predictions were then close to the values observed, and overall accuracy increased substantially.
Conclusions
The meropenem population pharmacokinetic model provides unbiased but imprecise population-level predictions and is therefore not suitable for a priori dosing in neonatal ECMO patients. Including just one prior concentration greatly improves prediction, supporting the use of this model for Bayesian, therapeutic drug monitoring–guided dosing of meropenem in this vulnerable population. (The study identification number is K 2026-0116).