Vincent Chang 1, Curtis Johnston, Ana Ruiz
1 Gilead Sciences (Foster City, USA)
Objectives
Assessing monoclonal antibody target engagement in vivo is critical to understanding drug activity and informing the selection of a safe and effective dose [1]. However, direct measurement of free target in biological matrices is expensive and analytically challenging due to rapid target turnover, assay artifacts, and the presence of high concentrations of drug–target complexes.
Pharmacokinetic and pharmacodynamic (PK/PD) modeling techniques provide a quantitative framework to infer target engagement indirectly from drug concentration–time data and observed nonlinear pharmacokinetics arising from target mediated drug disposition (TMDD) [2]. This simplified approach enables estimation of soluble target engagement without observed receptor occupancy data while reducing model complexity and parameter identifiability issues, making it particularly attractive in early clinical development.
This study evaluates the accuracy and precision of this PK/PD approximation method under the absence of observed receptor occupancy data with various assumptions and simulation scenarios.
Methods
A two-compartment target mediated drug disposition model with quasi equilibrium approximation was used as the true data generating model [3]. Simulated single dose phase1-like pharmacokinetic and receptor occupancy (RO) data were generated for six dose cohorts, each consisting of 3–6 participants (total = 42). Sample collection times were simulated from a normal distribution with 10% coefficient of variation around nominal times of 1.5, 2, 3, 5, 9, 12, 24, 48, 72, 168, 336, 504, 672, 1008, and 1344 hours.
An empirical two compartment model with linear clearance and nonlinear Michaelis–Menten elimination was then fit to each simulated pharmacokinetic dataset. The estimated Michaelis constant (Km) values were compared with the true model’s dissociation constant (Kd), and predicted RO compared with the true observations.
The parameters Kd, baseline receptor concentration, internalization rate (Kint), and degradation rate (Kdeg) were evaluated at an arbitrarily selected baseline value and at values 10, 100, and 1000-fold higher and lower than baseline. Each simulation scenario was repeated 200 times.
Results
The baseline true Kd was set to 2204 ng/mL and the empirically estimated median Km was 1948 ng/mL (10th – 90th percentile, 1848 – 2828) corresponding to a relative bias of -3.8% (-8.7 – 39.7). Decreasing the baseline receptor concentration 10, 100, or 1000-fold reduced the median bias marginally, but increased precision: 10-fold decrease resulted in a median bias of -0.5% (-4.4 – 28.1), 100-fold decrease had a median bias of -0.4% (-3.7 – 14.7), and 1000-fold decrease resulted in a median bias of -0.5% (-3.9 – 13.7). While increasing the baseline receptor concentration resulted in increased negative bias and larger variation between simulation replicates, 10-fold increase resulted in a median bias of -9.9% (-24.4 – 324), 100-fold increase had a median bias of 31.4% (-51.8 – 309), and 1000-fold increase resulted in a median bias of -33.4% (-91.9 – 557).
Conclusions
Across a wide range of conditions, median relative bias remained small, with Km estimates for most scenarios falling within ±30% of the true value. This level of accuracy is notable given that the empirical method does not rely on observed receptor occupancy data. Certain scenarios, such as increased baseline receptor concentrations, were associated with reduced precision, reflected by substantial variability across simulation replicates despite minor changes in median bias.
This nonlinear PKPD framework enables accurate and precise empirical estimation of target engagement from pharmacokinetic data alone across diverse conditions and represents a robust approach when receptor occupancy data are unavailable.
References:
[1] Mager DE, Jusko WJ. General pharmacokinetic model for drugs exhibiting target-mediated drug disposition. Journal of Pharmacokinetics and Pharmacodynamics. 2001;28(6):507–532.
[2] Mendizabal NV. Nonlinear Pharmacokinetics, Receptor Occupancy, and
Soluble Target Engagement as Tools to Assess In Vivo Antibody Bioactivity. Global Gators Symposium 2024. Bilbao, Spain
[3] Gibiansky, L., Gibiansky, E., Kakkar, T., & Ma, P. Approximations of the Target-Mediated Drug Disposition Model and Identifiability of Model Parameters. Journal of Pharmacokinetics and Pharmacodynamics 2008, 35, 573–591.
Reference: PAGE 34 (2026) Abstr 12265 [www.page-meeting.org/?abstract=12265]
Poster: Methodology - Model Evaluation