Daichi Yamaguchi 1,2, Takayuki Katsube 1, Yoshifumi Nishi 2, Takahiko Aoyama 2, Yasuhiro Tsuji 2
1 Shionogi & Co., Ltd. (, Japan), 2 Nihon University (, Japan)
Introduction:
Vancomycin has been used as a first-line treatment against methicillin-resistant Staphylococcus aureus for over 60 years and is also available for paediatric patients. Therapeutic drug monitoring and individualized dosing adjustments are recommended due to vancomycin’s narrow therapeutic range. The prediction accuracies of some Bayesian dose-optimizing software depend on the single population pharmacokinetic (PK) model selected [1]. Model selection based on statistical criteria could depend on data characteristics such as richness, age ranges, and blood sampling time points, potentially leading to models with incorrect inferences. More than 50 studies of population PK modelling of vancomycin for paediatric patients have been reported, but the analysis populations, model structures, and selected covariates vary from study to study [2]. Rather than using a single population PK model to predict vancomycin concentrations with high accuracy in various patients, using a multi-model approach with prior information might be better. The Bayesian model averaging (BMA) framework [3] can obtain the posterior distribution of parameter estimation and posterior selection probability for each of multi-models. Inference by averaging multi-models based on posterior distributions and probabilities could be more reliable [4]. Few studies have used the BMA approach to develop an averaged population PK model [5], and we performed population PK modeling of vancomycin for adults based on the BMA approach using NONMEM for the first time [6]. The objective of this study was to apply BMA approach to the population PK analysis of vancomycin for paediatric patients.
Methods:
Data were collected from paediatric patients who measured at least one plasma vancomycin concentration in Kyorin University Hospital from 2016 to 2019. BMA was performed by Markov chain Monte Carlo (MCMC) Bayesian estimation algorithm implemented in NONMEM. A switch parameter, a variable (0 or 1) following a Bernoulli distribution, was adapted to each model structure and each covariate. We obtained 20,000 samples from four MCMC chains to make an inference about posterior distributions, and if the switch parameter was estimated as 1 in each sample, the model or covariate were selected in this sample. The posterior selection probability of model structure or covariate was calculated from the posterior distribution of the switch parameter. We tested two candidate model structures, one and two compartment models, and some covariates for each PK parameters as follows: body weight (WT), postnatal age (PNA) for more than two years old, maturation factor (MF) based on postmenstrual age, estimated glomerular filtration rate (eGFR), and blood urea nitrogen (BUN) for the clearance (CL), WT and PNA for the volume of distribution in central compartment (Vc) and the inter-compartmental clearance, and WT for the volume of distribution in peripheral compartment. Prior information for these model structure and covariate selections were set based on the proportions of each model structure and covariate from 51 models reported for population PK modelling of vancomycin with paediatric patients.
Results:
Total 148 observed plasma vancomycin concentrations from 46 patients aged 0 to 18 years old were available. Model structure was selected as one compartment distribution with first order elimination model from all samples (posterior selection probability = 100%). The population mean values of CL and Vc were estimated 1.91 L/hr and 7.82 L, respectively. MF was selected as a covariate for CL in all samples (100%). The posterior selection probabilities of WT, PNA, eGFR, and BUN for CL were 15.3%, 0%, 77.7%, and 0%, respectively. Those of WT and PNA for Vc were 0%, in spite of high prior selection probability of WT for Vc. It was suggested that inter-individual variability in the PK profiles of vancomycin might be explained by mainly variability of CL. The high convergence of MCMC chains was confirmed by visual diagnosis and statistical methods.
Conclusion:
We performed population PK analysis using BMA approach and developed averaged model of vancomycin for paediatric patients. Based on our averaged model, the influence of each covariate for PK parameters could be evaluated and vancomycin exposures could be estimated depend on various covariate models but not one model. This model averaging method may be applicable to other population PK analyses.
References:
[1]Turner RB. Pharmacotherapy. 2018;38:1174-1183.
[2]Chung E. Clin Pharmacokinet. 2021;60:985-1001.
[3]Hoeting JA. Stat Sci. 1999;14:382-417.
[4]Longford NT. J R Statist Soc A. 2005;168:469-472.
[5]Lunn DJ. J Pharmacokinet Pharmacodyn. 2008;35:85-100.
[6]Yamaguchi D. PAGE31. 2023;Abstr:10302.
Reference: PAGE 34 (2026) Abstr 11980 [www.page-meeting.org/?abstract=11980]
Poster: Methodology - Estimation Methods