Ethambutol disposition in humans: Challenges and limitations of whole-body physiologically-based pharmacokinetic modelling in early drug development

Bayesian PBPK modeling using R/Stan/Torsten and Julia/SciML/Turing.Jl

Physiologically-based pharmacokinetic (PBPK) models are mechanistic models that are built based on an investigator's prior knowledge of the in vivo system of interest. Bayesian inference incorporates an investigator's prior knowledge of parameters while using the data to update this knowledge. As such, Bayesian tools are well-suited to infer PBPK model parameters using the strong prior knowledge available while quantifying the uncertainty on these parameters. This tutorial demonstrates a full population Bayesian PBPK analysis framework using R/Stan/Torsten and Julia/SciML/Turing.jl.

Development of physiologically-based pharmacokinetic models for standard of care and newer tuberculosis drugs

Tuberculosis (TB) remains a global health problem and there is an ongoing effort to develop more effective therapies and new combination regimes that can reduce duration of treatment. The purpose of this study was to demonstrate utility of a physiologically‐based pharmacokinetic modeling approach to predict plasma and lung concentrations of 11 compounds used or under development as TB therapies (bedaquiline [and N‐desmethyl bedaquiline], clofazimine, cycloserine, ethambutol, ethionamide, isoniazid, kanamycin, linezolid, pyrazinamide, rifampicin, and rifapentine). Model accuracy was assessed by comparison of simulated plasma pharmacokinetic parameters with healthy volunteer data for compounds administered alone or in combination. Eighty‐four percent (area under the curve [AUC]) and 91% (maximum concentration [C_max]) of simulated mean values were within 1.5‐fold of the observed data and the simulated drug‐drug interaction ratios were within 1.5‐fold (AUC) and twofold (C_max) of the observed data for nine (AUC) and eight (C_max) of the 10 cases. Following satisfactory recovery of plasma concentrations in healthy volunteers, model accuracy was assessed further (where patients’ with TB data were available) by comparing clinical data with simulated lung concentrations (9 compounds) and simulated lung: plasma concentration ratios (7 compounds). The 5th–95th percentiles for the simulated lung concentration data recovered between 13% (isoniazid and pyrazinamide) and 88% (pyrazinamide) of the observed data points (Am J Respir Crit Care Med, 198, 2018, 1208; Nat Med, 21, 2015, 1223; PLoS Med, 16, 2019, e1002773). The impact of uncertain model parameters, such as the fraction of drug unbound in lung tissue mass (fu_mass), is discussed. Additionally, the variability associated with the patient lung concentration data, which was sparse and included extensive within‐subject, interlaboratory, and experimental variability (as well interindividual variability) is reviewed. All presented models are transparently documented and are available as open‐source to aid further research.