Abstract
BACKGROUND AND OBJECTIVE: Population pharmacokinetic (popPK) models in pediatric patients are essential to optimize dosing and ensure therapeutic efficacy. However, study designs are often not fully optimized, leaving room to improve efficiency, which is an important goal in this population, where patients are limited and resources scarce. The aim of the present work is to optimize study designs for the development of popPK models for teicoplanin, piperacillin and meropenem in pediatric patients, with or without continuous kidney replacement therapy (CKRT), to achieve greater model precision while reducing patient burden and economic cost. METHODS: Methodology based on the optimization of the Fisher information matrix (FIM) was followed, using the $DESIGN option in NONMEM 7.5. A previously developed model was selected for each of the antibiotics. The number of subjects in the optimized designs was fixed to 28 patients (14 with and 14 without CKRT). It was assumed that only plasma samples were extracted from patients without CKRT, while prefilter, postfilter, and effluent samples could be extracted simultaneously from patients undergoing CKRT. Sensitivity to different proportions of patients with and without CKRT was tested. The optimized designs were evaluated through simulation and re-estimation procedures, including the impact of covariates. RESULTS: The number of sampling times per individual needed to achieve precise parameter estimates was 3 in teicoplanin, 4 in piperacillin, and 6 in meropenem. The optimized designs reduced the total number of samples per patient by 25, 51, and 21% for teicoplanin, piperacillin, and meropenem, respectively, compared with the original studies used in the previous studies. The resulting samples were taken during 0-40 h from the beginning of the study in teicoplanin and piperacillin, while in the case of meropenem optimal sampling times went between 0-64 h. The optimized designs remained robust under different proportions of patients with and without CKRT and under different covariate values. CONCLUSIONS: This work emphasizes the importance of optimizing study designs to improve accuracy and precision in the model parameters while reducing the number of samples needed. This is a relevant advantage especially when dealing with critically ill pediatric patients.