Vancomycin is frequently used to treat gram-positive infections in neonates and children. The optimal marker for efficacy and toxicity is the 24-h area under the concentration-time curve/minimal inhibitory concentration, which can be estimated using population pharmacokinetic (popPK) models. These models can be integrated into model-informed precision dosing (MIPD) tools to personalize therapy. This study aimed to systematically review neonatal and paediatric vancomycin popPK models and develop a model selection tool for MIPD. A systematic literature search (December 2025; PubMed, EMBASE, WoS, Scopus, IPA) was conducted following PRISMA guidelines using the concepts 'neonates/children', 'vancomycin' and 'popPK model'. Inclusion criteria were age 0-18 years, intravenous vancomycin, and popPK model development. Descriptive statistics summarized tested vs. retained covariates. Based on literature and multidisciplinary expert input, model selection criteria were proposed and applied to the identified models. Across 7915 articles, 112 popPK models were identified (1994-2025; 35 neonatal, 6 infantile, 39 paediatric, 2 adolescent, 30 mixed-age). Models were built on data from median (range) 114 (6-7167) patients, representing 337 (26-13 372) PK samples. There were 81 one- and 30 two-compartment models. Of 180 covariates tested across models, 47 (26.1%) were retained. Body weight was most frequently tested and retained (101/106, 95.3%). Model selection criteria included matching population, external and internal model evaluation (calibration, discrimination, uncertainty), cohort size, PK sampling schedule. The Chen (2023) model scored highest for both the neonatal and paediatric populations. This review and model selection tool can guide and standardize model selection for MIPD.
Elst et al. (Tue,) studied this question.