Bayesian estimation of parameters in the pharmacokinetic model
Seongil Jo1, Eunjung Song2, Min-Gul Kim3, SeungHwan Lee4, Woojoo Lee2, Bo-Hyung Kim5
1 Department of Statistics, Chonbuk National University, Jeonju, Korea, 2 Department of Statistics, lnha University, Incheon, Korea, 3 Department of Pharmacology, College of Medicine, Chonbuk National University, Jeonju, Korea, 4 Department of Clinical Pharmacology and Therapeutics, Seoul National University College of Medicine and Hospital, Seoul, Korea, 5 Department of Clinical Pharmacology and Therapeutics, Kyung Hee University College of Medicine and Hospital, Seoul, Korea
Objectives: The pharmacokinetic (PK) estimation and dose individualization are important to obtain a favorable outcome in a clinical setting. The consultation of therapeutic drug monitoring (TDM) is especially based on only peak or trough concentrations. In this limited observation, the dose optimization for TDM can be determined using Bayesian inference, which include the pre-defined PK model with covariate-parameter relationships defined a priori. Therefore, the objective of this study was to develop a TDM package in Stan version 2.14 (Stan Development Team) with the R package Rstan (Stan Development Team).
Methods: The vancomycin, known as a typical two-compartment PK model, was selected for the example of Bayesian estimation, because the vancomycin is widely used for the practice of TDM. The trough concentrations were randomly generated in the virtual PK model based on literature PK parameters of vancomycin. Log Gaussian Priors for the PK parameters were applied, and Bayesian inference was conducted using Markov chain Monte Carlo (MCMC) simulation.
Results: In the current study, the PK parameters for the population model were estimated and then individual drug concentrations were predicted by the developed package. The package provided suitable estimates for the population model and its prediction performance was also comparable with the other alternative packages (Abott PKS system, tdm, etc).
Conclusions: This study developed an alternative TDM package based on RStan, which can be comparable predictability to the previous software packages such as the Abott PKS system. We expect that this result will provide an accurate estimation of the PK parameters and prediction of dose concentrations, unlike other competitors with the fixed parameters for the population PK model.
References:
[1] Stan Development Team. (2017). Stan Modeling Language: User’s Guide and Reference Manual.
[2] Wakefield, J. (2013). Bayesian and Frequentist Regression methods.