Exploring the impact of obesity on drug clearance using PBPK modelling approaches: influential variables affecting scaling from non-obese to obese subjects for drugs metabolized in the liver
Tan Zhang (1), Elke H.J. Krekels (1), Elisa A.M.Calvier (2), Catherijne A.J. Knibbe (1,3)
(1) Division of Systems Pharmacology and Pharmacy, Leiden Academic Centre for Drug Research, Leiden University, Leiden, The Netherlands; (2) Pharmacokinetics-Dynamics and Metabolism (PKDM), Translational Medicine and Early Development, Sanofi R&D, Montpellier, France (3) Department of Clinical Pharmacy, St. Antonius Hospital, Nieuwegein, The Netherlands
Objectives:
Determining drug clearance (CL) is important to establish drug dosing. In the obese, some drugs do not show any changes in CL with bodyweight, while other drugs show large differences in the correlation between CL and bodyweight [1]. However, predictors for the identification of appropriate scaling methods have not been identified. To address this gap, we use a physiologically-based pharmacokinetic (PBPK) approach to investigate patterns of the “true” exponent that can accurately scale hepatic plasma clearance (CLp) with bodyweight and find out correlations with drug or patient characteristics. Moreover, we used the PBPK workflow to assess the systematic accuracy of commonly used bodyweight-based scaling methods.
Methods:
For 22,000 hypothetical drugs, “true” CLp for a typical non-obese subject (BMI 20 kg/m2) and five typical overweight and obese subjects (BMI from 25-60 kg/m2) were simulated using the dispersion model and previously reported physiological values [2,3]. The hypothetical drugs were assumed to either bind to human serum albumin (HSA) or α-1 acid glycoprotein (AAG) and had a wide range in fraction unbound (fu, 1-100%), blood-to-plasma partition coefficient (Kp, 0.35-40), and total unbound intrinsic microsomal clearance (CLint,mic, 0.56*10-6-0.209*10-3 ml∙min-1∙μg-1 microsomal protein). Each drug was metabolized by isoenzymes with varying hypothetical changes in obesity-induced activity (reduction of up to 75% or increase of up to 150% of normal activity [4]). All typical subjects were assumed to have a height of 172 cm. In a bodyweight-based exponential equation, the “true” exponent that would yield perfect scaling of “true” CLp was calculated. Correlations between the “true” exponent and drug or patient characteristics were investigated graphically using R. Besides this, the systematic accuracy of the following scaling methods was investigated: (1) no scaling (e.g., exponent of 0, assuming same CLp in obese and non-obese), (2) exponential bodyweight-based scaling with a fixed exponent of 0.75 (AS0.75), and (3) linear bodyweight-based scaling (e.g., exponent of 1). For this, the prediction error (PE) between “true” CLp and CLp scaled from non-obese to obese subjects was calculated and systematic accuracy was defined as PE values of all hypothetical drugs beginning within ±30%.
Results:
The “true” exponent that yields perfect scaling ranged from -5.94 to 4.81 and was most considerably impacted by changes in enzyme activity. Without changes in enzyme activity, the exponent ranges between 0.266 and 0.851, and it decreases when enzyme activity is reduced and increases with increasing activity. The “true” exponent value also depends on BMI, with the exponent of a drug increasing with increasing BMI in scenarios with the same change in enzyme activity. Drugs binding to AAG generally have a lower “true” exponent compared to drugs bound to HSA and those with a lower extraction ratio generally have a lower exponent compared to those with a high extraction ratio.
The systematic accuracy of the scaling methods was impacted by the same variables as the “true” exponent. We found that AS0.75 performs on average more accurate than linear scaling and no-scaling methods for different drug properties, however, the range in PE was still large (-64.2 – 593). The linear scaling method tends to over-scale the CLp and performs accurately only for subjects with a BMI lower than 40 kg/m2 and up to a 50% increase in isoenzyme activity. On the contrary, the no-scaling method tends to under-scale the CLp, but this method tends to accurately predict CLp for subjects with reduced enzyme activity.
Conclusions:
The “true” exponent value of a bodyweight-based covariate function has a wide range depending on changes in enzyme activity, drug properties, and BMI. AS0.75 is on average closer to the “true” exponent and therefore scales more systematically accurate compared to no scaling or linear scaling. However, no single scaling method is systematically accurate. To achieve accurate scaling drug properties and obesity-related changes in enzyme activity need to be taken into account.
References:
[1] Zhang T et al.. Expert Opin Drug Metab Toxicol. 2022 Oct;18(10):657-674.
[2] Calvier EAM, et al.. AAPS J. 2019 Mar 8;21(3):38.
[3] Berton, M, et al. Clin Pharmacokinet. 61.9 (2022): 1251-1270.
[4]. Ghobadi, C, et al. Clin. pharmacokinet. 50 (2011): 809-822.