Adam Nasim 1,2
1 Novartis (Basel, Switzerland), 2 University of Surrey (Guildford, UK)
Objectives: Target mediated drug disposition (TMDD) models are widely used to describe monoclonal antibody pharmacokinetics and to predict suppression of soluble circulating targets [1,2], but early feasibility simulations often rely on uncertain parameter assumptions, especially for target turnover and drug target complex elimination. This work aimed to (i) curate and summarise reported TMDD parameter estimates for antibodies directed against soluble ligands, to support de novo modelling and simulation, and (ii) quantify how common assumptions and data limitations bias inference of key parameters such as baseline target concentration, turnover, complex internalisation, and affinity.
Methods: A targeted literature review identified clinical and non clinical TMDD modelling case studies for antibodies against soluble targets. For each target, baseline concentration (R0), endogenous degradation (kdeg), and complex internalisation or elimination (kint) were extracted where available, together with contextual information such as target molecular weight and antibody half life. Relationships between molecular weight and turnover parameters were explored, including comparison to a published molecular weight based regression for kdeg [3]. To investigate inference sensitivity, a two compartment TMDD model with subcutaneous dosing was used to generate synthetic total antibody and total target data, reflecting recommended assay practice [4]. Data were simulated for four dose groups (2, 4, 8, 10 mg per kg), three replicates per dose group, sampled from 0 to 32 days, with 10 percent proportional residual error, and with the association rate fixed (kon = 1.8 nM^-1 h^-1), following a pragmatic assumption used in prior work [5]. Models were refit under (i) increasing lower limits of quantification (LLOQ) for total target, mimicking assay censoring, and (ii) fixing kint to the antibody elimination rate (CL over Vc), a common simplifying assumption. Additional simulations assessed potential confounding between absorption (ka) and kint for subcutaneous dosing.
Results: TMDD modelling case studies were identified for 14 soluble targets spanning cytokines, chemokines, enzymes, and extracellular proteins. Baseline circulating target concentrations were typically in the picomolar range, but varied widely by target and disease context. Target half life estimates derived from kdeg spanned a broad range and showed partial agreement with a molecular weight based regression framework [3], supporting use of molecular weight informed priors for de novo simulations. In contrast, complex elimination rates (kint) showed substantial heterogeneity and no clear relationship with target molecular weight. Literature examples demonstrate that kint can be similar to antibody elimination for some targets, but can exceed it by several fold, or be orders of magnitude higher for specific antibody target pairs, indicating that assuming kint equals antibody clearance is not universally valid.
In the simulation refitting exercise, censoring total target data by increasing LLOQ reduced late time information and induced parameter compensation. Baseline target concentration (R0) increased by approximately two fold across the LLOQ range examined, and affinity (kD) became biased, with the most truncated scenario yielding kD estimates about 2.7 fold higher than an intermediate LLOQ case. Fixing kint to CL over Vc produced pronounced bias in affinity, with an approximately 4.7 fold difference in estimated kD between scenarios where the true kint was two fold higher versus two fold lower than the fixed value. These changes occurred despite relatively stable pharmacokinetic parameters, consistent with strong structural coupling between kint and binding parameters through the total target accumulation profile. Simulations also indicated that ka and kint are theoretically confounded when time scales are similar, but kint remained identifiable under the evaluated design, with greater challenges when kint is much slower than absorption.
Conclusions: Published TMDD parameter values for soluble targets provide a practical basis for de novo feasibility simulations, especially when combined with molecular weight informed priors for kdeg [3] and explicit uncertainty propagation for baseline target. However, complex internalisation rates are highly variable and should not routinely be fixed to antibody clearance. Assay censoring that removes baseline or late time total target observations can materially bias inferred baseline and affinity, and may contribute to discrepancies between in vitro affinity and in vivo derived binding constants. Recommendations are to prioritise sensitive total target assays, include sampling through post saturation decline, avoid fixing kint unless justified, and perform sensitivity analyses over plausible kint and baseline ranges.
References:
[1] Gibiansky L, Gibiansky E, Kakkar T, Ma P. Approximations of the target mediated drug disposition model and identifiability of model parameters. J Pharmacokinet Pharmacodyn. 2008.
[2] Dua P, Hawkins E, van der Graaf PH. A tutorial on target mediated drug disposition (TMDD) models. CPT Pharmacometrics Syst Pharmacol. 2015.
[3] Muliaditan M, Sepp A. Application of quantitative protein mass spectrometric data in the early predictive analysis of target engagement by monoclonal antibodies. Clin Transl Sci. 2022.
[4] Fairman D, Tang H. Best practices in mAb and soluble target assay selection for quantitative modelling and qualitative interpretation. AAPS J. 2023.
[5] Tiwari A, Luo H, Chen X, et al. Assessing the impact of tissue target concentration data on uncertainty in in vivo target coverage predictions. CPT Pharmacometrics Syst Pharmacol. 2016.
Reference: PAGE 34 (2026) Abstr 12155 [www.page-meeting.org/?abstract=12155]
Poster: Drug/Disease Modelling - Oncology