A population pharmacokinetic model library for polymyxin B: clinical implementation of model-informed precision dosing

Vollständiger Abstract

Worum geht es in dieser Arbeit?

Introduction Polymyxin B remains a key last-resort therapy for infections caused by multidrug-resistant Gram-negative bacteria. Nevertheless, its substantial pharmacokinetic (PK) variability and narrow therapeutic window pose significant challenges to dose individualization. Despite the availability of multiple published population pharmacokinetic (popPK) models, a comprehensive repository to enable model-informed precision dosing has not yet been established. Accordingly, this study aimed to develop a comprehensive library of polymyxin B popPK models to facilitate individualized treatment in clinical practice. Methods A systematic review was performed to locate published population pharmacokinetic models of polymyxin B. Three bibliographic databases, namely, PubMed, Web of Science, and Embase, were searched through November 2025 without a lower date limit. For each eligible study, information was collected on study design, administered doses, blood-sampling schemes, structural model characteristics, parameter values, and covariate effects. Articles published in languages other than English, duplicate records, and studies based on nonparametric modeling approaches were excluded. The model database was constructed in R with the mrgsolve package, and an interactive Shiny platform was created to support exposure simulation and individualized model-informed dose prediction. Results Twenty-two studies were incorporated into the polymyxin B model repository. The study populations included patients receiving extracorporeal membrane oxygenation (two studies), patients undergoing continuous renal replacement therapy or related renal replacement modalities (four studies), and patients with severe burns, liver dysfunction, lung or renal transplantation, obesity, renal impairment, or advanced age, as well as healthy volunteers. The repository also incorporated subpopulations with clinically relevant covariates known to influence PK, such as renal function and obesity. Most polymyxin B popPK models were described by a two-compartment model with linear elimination. In 10 of the 22 studies, renal function metrics were found to significantly impact distribution or elimination PK parameters. Internal validation was reported for all included models, whereas only two studies evaluated model performance using external datasets. Exposure simulations showed marked differences across models, indicating that fixed standard dosing regimens may have limited ability to consistently achieve target therapeutic concentrations. Conclusion By harmonizing published population pharmacokinetic models in a single framework, this repository enables systematic exploration of model-based predictions across clinically relevant populations and may facilitate individualized polymyxin B therapy. However, given the complex clinical scenarios in critically ill patients, the heterogeneity of their physiological and pathological states, and the substantial interindividual PK variability of polymyxin B, further investigations are warranted to establish a more robust evidence base for precision dosing.