Hong Su1, Yu Fu1, Liang Yang1, Peiming Ma1
1. Clinical Pharmacology, AstraZeneca, Shanghai, China.
Objectives:
Assessing target engagement is important for understanding drug action and selecting dosing regimens. The binding of the drug to its pharmacological targets and subsequent elimination of the drug-target complexes could affect drug distribution and elimination, and result in nonlinearity of PK, a phenomenon called target-mediated drug disposition (TMDD). Because of the identifiability issue of the full TMDD model, its approximations have been developed and proved to have often similar predictive performance of drug exposure.1-3 The TMDD models were at times used to predict receptor occupancy (RO) via the Michaelis-Menten (MM) parameter KM based only on drug concentration data.4,5 However, there appear lack of understanding for this and, thus a need to investigate how to use TMDD model to accurately predict RO. In this study, we aimed at studying whether RO can be predicted using KM with only drug concentration data. We also tested if incorrect assumptions of binding sites would influence the prediction of RO.
Methods:
Simulations were performed to test if the concentration profile is sensitive to KM using full TMDD model from literature.6 KM were varied in folds (by changing the receptor-ligand complex internalization rate for 10000 folds), and corresponding simulated data were plotted. RO vs time curves were plotted to compare the RO derived by KM, defined as RO% = C/(KM+C)*100%, with the true RO, defined as the ratio of receptor-ligand complex divided by total receptor.
Stochastic simulation and estimation (SSE) analyses were conducted to evaluate the estimation performance of KM using different TMDD models. The full TMDD model was defined as the true model, while quasi-steady state/quasi-equilibrium (QSS/QE) and MM models were tested as alternative models. SSE analyses were also conducted to assess the impact of receptor-ligand binding site (central or peripheral) on KM. The parameters, dose regimens and sampling time were adopted from references.4,6,7
Results:
Sensitivity analysis of KM in full TMDD model
Simulation showed the concentration profiles are indistinguishable in the raw scale with different magnitudes of KM, consequently KM cannot be calculated accurately. Thus, KM may not be a suitable parameter for RO calculation.
The simulation also showed that the time-dependent RO profiles are different using KM-derived RO and the true RO. There was not an increasing phase (due to receptor-ligand binding) for KM-calculated RO, and it was lower than the true RO at low concentrations. KM-derived RO had similar values to the true RO when drug concentration and RO were high, e.g., if KM-derived RO was larger than 70%, its difference from the true RO was less than 5%.
SSE analysis for the full TMDD model and its approximations
The SSE results showed that the results of QSS/QE or MM models as alternative models may be inferior compared with the true TMDD model, shown as increased AICs. More importantly, the precision and/or accuracy of KM were decreased compared to the full TMDD model (the largest difference of mean KM of MM model and QSS model were 9-fold, P<0.001 and 60-fold, P<0.001; t-test). This showed that the RO derived from KM may not be trusted without additional validation if the approximation models were used.
When using sparse data, one situation occurred that the MM model showed a similar AIC as the full model (ANOVA, P>0.24) and good parameter precisions (relative standard errors <37%) but the estimated KM had over 5-fold bias (t-test, P<0.001). This suggested that the MM model may also be misleading when used for RO calculation.
SSE analysis for the central vs. peripheral TMDD models
The simulations showed that different ligand-receptor binding sites will affect the kinetics of drug concentrations in the central compartment. Wrong application of the binding compartment could lead to worse accuracy/precision of the KM parameter (the largest difference of mean KM was about 22-fold, P<0.001, t-test).
Conclusions:
This research showed that the expression of true RO (rather than KM-derived RO) should be used in the TMDD model-based prediction of RO with only drug concentration data. Full TMDD models, if not overparameterized, should be considered initially. In addition, incorrect assumption of binding sites could result in inaccurate prediction of RO. This work provides tips on how to conduct model-based prediction of RO and would further support dose optimization in clinical drug development.
References:
[1] Mager DE et al., Pharm Res. 2005 Oct;22(10):1589-96.
[2] Gibiansky L et al., J Pharmacokinet Pharmacodyn. 2008 Oct;35(5):573-91.
[3] Ma P. Pharm Res. 2012 Mar;29(3):866-82.
[4] Park WS et al., Basic Clin Pharmacol Toxicol. 2017 Mar;120(3):243-249.
[5] Gibiansky L et al., Clin Pharmacokinet. 2012 Apr 1;51(4):247-60.
[6] Peletier LA et al., J Pharmacokinet Pharmacodyn. 2012 Oct;39(5):429-51.
[7] Wu N et al., J Pharm Sci. 2022 Sep;111(9):2620-2629.
Reference: PAGE 32 (2024) Abstr 10842 [www.page-meeting.org/?abstract=10842]
Poster: Drug/Disease Modelling - Other Topics