MSc Medhat Said1,2, Dr Jurjen Aman3,4, Prof. Eleonora Swart1,2, Dr. Imke Bartelink1,2, Prof Ron Mathot
1Department of Pharmacy and Clinical Pharmacology, 2Cancer Center Amsterdam, Amsterdam, The Netherlands, 3Department of Pulmonary Medicine, Amsterdam UMC, location VUmc, 4Amsterdam Cardiovascular Sciences, Amsterdam,
Introduction/Objectives: Accurate pharmacokinetic (PK) modeling of both total and unbound drug concentration is critical for understanding drug disposition, particularly when nonlinear protein binding is involved. Several modelling strategies have been used in describing total and unbound drug PK, yet there is no consensus on a standardized approach. The aim of this study was to compare PK models in describing total and unbound drug concentrations and their impact on drug exposure, PK parameters, and unbound fraction under varying PK conditions. Methods: The following three PK modeling approaches that differ in how they represent total and unbound drug concentrations: 1)A dual compartment model explicitly describing unbound and bound drug with nonlinear binding incorporated into the rate constants (M1) [1]; 2)A compartment model for total drug where unbound drug is calculated post hoc using a nonlinear binding equation but the derived unbound fraction (FU) determines elimination of unbound drug (M2) [2]; 3)A compartment model for unbound drug where total drug is calculated post hoc using a nonlinear binding equation (M3) [3]. The models were evaluated by varying key PK parameters (clearance, volume of distribution, and protein binding) and comparing their impact on total and unbound drug exposure, PK parameters, and the unbound fraction. Total drug clearance (CLtot) and volume of distribution (Vtot) were defined as CLtot = CLu * FU and Vtot = Vu * FU, where CLu and Vu represent unbound drug clearance and volume, and FU is the unbound fraction. Inter-individual variability (IIV) was set at 20% for all parameters, and overlapping PK parameters were held constant across models. The PK models were implemented using nlmixr2 in R version 4.2.3. Results: Effect on concentration profiles Changes in drug clearance had a similar effect across all three models (M1, M2, M3), with total and unbound drug concentrations showing consistent trends. However, variations in distribution volume produced different effects between M1 and M2: increasing the unbound drug volume in M1 resulted in an increase in total drug concentration, whereas in M2, it led to a decrease in total drug concentration. M3 exhibited notably different behavior, showing altered PK profiles compared to M1 and M2, even when the same PK parameters were used. Effect on drug clearance Changes in CLu led to similar effects on drug clearance (both total and unbound) across all three models. When Vu was altered, M1 showed changes CLu but no change in CLtot. In contrast, M2 did not show any changes in either CLu or CLtot with varying Vu. M3 showed changes in clearance only after a certain period, further differentiating its behavior from the other two models. Effect on drug distribution volume Modifications in CLu did not affect the unbound volume (Vu) or total volume (Vtot) in any of the models. Changes in Vu had no impact on Vtot in M1 but did affect Vtot in M2. Furthermore, when changing Kd and Bmax, all three models exhibited similar effects on drug volume. However, M3 showed notably different and less favorable results compared to M1 and M2, with decrease in drug volume over time. Effect on unbound fraction For the unbound fraction (FU), varying CLu caused different patterns in M1 and M2. In M1, higher unbound drug concentrations led to a gradual increase in FU, while in M2, FU remained constant once a certain threshold of unbound drug concentration was reached. When altering Vu, M1 showed that lower Vu led to a higher unbound fraction, with the maximum FU varying for each Vu. In M2, the maximum FU was similar across all Vu values, but the time to reach a constant FU decreased as Vu decreased. Changes in Kd and Bmax had similar effects across all models, altering FU in proportion to changes in Bmax. In M1, FU continued to increase with higher unbound drug concentrations, whereas in M2, FU reached a plateau after a certain point. Conclusions: This study demonstrated that the three PK models (M1, M2, M3) differ in how they describe drug concentration profiles, clearance, distribution volume, and unbound fraction. While M1 and M2 showed consistent trends, M3 exhibited more complex behavior, particularly in PK profiles and drug volume. These findings highlight the importance of model selection in accurately predicting drug disposition, especially with nonlinear protein binding.
(1) de Winter BC, van Gelder T, Sombogaard F, Shaw LM, van Hest RM, Mathot RA. PK role of protein binding of mycophenolic acid and its glucuronide metabolite in renal transplant recipients. J Pharmacokinet Pharmacodyn. 2009 Dec;36(6):541-64. doi: 10.1007/s10928-009-9136-6. Epub 2009 Nov 11. PMID: 19904584; PMCID: PMC2784070. (2) Standing JF, Ongas MO, Ogwang C, Kagwanja N, Murunga S, Mwaringa S, Ali R, Mturi N, Timbwa M, Manyasi C, Mwalekwa L, Bandika VL, Ogutu B, Waichungo J, Kipper K, Berkley JA; FLACSAM-PK Study Group. Dosing of Ceftriaxone and Metronidazole for Children With Severe Acute Malnutrition. Clin Pharmacol Ther. 2018 Dec;104(6):1165-1174. doi: 10.1002/cpt.1078. Epub 2018 Apr 19. PMID: 29574688; PMCID: PMC6282491. Haouala A, Widmer N, Guidi M, Montemurro M, Leyvraz S, Buclin T, Eap CB, Decosterd LA, Csajka C. Prediction of free imatinib concentrations based on total plasma concentrations in patients with gastrointestinal stromal tumours. Br J Clin Pharmacol. 2013 Apr;75(4):1007-18. doi: 10.1111/j.1365-2125.2012.04422.x. PMID: 22891806; PMCID: PMC3612719.
Reference: PAGE 33 (2025) Abstr 11337 [www.page-meeting.org/?abstract=11337]
Poster: Drug/Disease Modelling - Other Topics