Predictive Performance of Physiologically Based Pharmacokinetic Modelling of Beta-Lactam Antibiotic Concentrations in Adipose, Bone, and Muscle Tissues

Abstract

Physiologically based pharmacokinetic (PBPK) models consist of compartments representing different tissues. As most models are only verified based on plasma concentrations, it is unclear how reliable associated tissue profiles are. This study aimed to assess the accuracy of PBPK predicted beta-lactam antibiotic concentrations in different tissues and assess the impact of using effect site concentrations for evaluation of target attainment. Adipose, bone and muscle concentrations of five beta-lactams (piperacillin, cefazolin, cefuroxime, ceftazidime and meropenem) in healthy adults were collected from literature and compared to PBPK predictions. Model performance was evaluated with average fold errors (AFEs) and absolute AFEs (AAFEs) between predicted and observed concentrations. In total, 26 studies were included, 14 of which reported total tissue concentrations and 12 unbound interstitial fluid (uISF) concentrations. Concurrent plasma concentrations, used as baseline verification of the models, were fairly accurate (AFE: 1.14, AAFE: 1.50). Predicted total tissue concentrations were less accurate (AFE: 0.68, AAFE: 1.89). A slight trend for underprediction was observed but none of the studies had AFE or AAFE values outside threefold. Similarly, predictions of microdialysis-derived uISF concentrations were less accurate than plasma concentration predictions (AFE: 1.52, AAFE: 2.32). uISF concentrations tended to be overpredicted and two studies had AFEs and AAFEs outside threefold. Pharmacodynamic simulations in our case showed only a limited impact of using uISF concentrations instead of unbound plasma concentrations on target attainment rates. The results of this study illustrate the limitations of current PBPK models to predict tissue concentrations and the associated need for more accurate models.

Physiologically‐based pharmacokinetic modelling to predict intragastric rifabutin concentrations in the treatment of Helicobacter pylori infection

“…As stated above, this work is not based on direct intragastric sampling., Rather, we used PBPK modelling which utilised information from knowledge of the human anatomy, physiology, drug metabolising enzymes and transporter expression, plasma concentrations as well as physical–chemical properties of rifabutin to estimate tissue distribution. Such estimates are based on extensive knowledge of the relationship between the drug's physical–chemical properties and its tissue to plasma distribution, further qualified against knowledge of its human plasma pharmacokinetics in humans 12–14 . The principles supporting PBPK modelling to support a drug's biodistribution are well established.…”

“…Collection of intragastric and/or intestinal concentrations of a drug is logistically challenging and invasive. However, the advent of in silico PBPK modelling enables the prediction of tissue concentrations by leveraging a drug's physical–chemical characteristics, biochemical properties, and plasma concentration versus time profiles 12–14 . In this study, we utilised this approach to examine and understand the relationship between rifabutin dosing regimens and the resulting predicted concentrations relative to the mean inhibitory concentrations (MIC 90 ) for H. pylori .…”