Daniel Bending 1, Venkatesh Pilla Reddy
1 Eli Lilly (, )
Introduction/Objectives
Antibody-drug conjugates (ADCs) present unique pharmacokinetic challenges arising from the interplay between antibody disposition, payload deconjugation, and the fate of released payload. A substantial fraction of payload released from ADCs fails to appear in systemic circulation—a phenomenon termed “ghost payload”—likely attributable to retro-Michael addition of the drug-linker moiety to serum albumin, with subsequent albumin catabolism or slow cleavage back to free payload¹. Traditional ADC pharmacokinetic models assume complete payload bioavailability following deconjugation, potentially leading to misestimation of payload exposure and confounding efficacy-toxicity predictions². The present work develops a hierarchy of semi-mechanistic models incorporating payload bioavailability (F<1) and depot dynamics to characterize ghost payload, and systematically evaluates structural identifiability under clinically realistic observation scenarios. A minimal DAR2 system was employed as the simplest conjugate where average drug-to-antibody ratio (DAR) provides information beyond total ADC and total monoclonal antibody (mAb) concentrations, preserving the cascade structure of higher-DAR systems while remaining analytically tractable³.
Methods
Three nested models of increasing complexity were formulated: Model 1 assuming complete bioavailability (F=1), Model 2 incorporating permanent sequestration (F<1, no return), and Model 3 adding a depot compartment permitting partial payload return (characterised by release rate krel and sink rate ksink). External constraints from intravenous payload studies (payload elimination rate constant) and conjugated ADC pharmacokinetics identical to those of naked mAb (distribution and elimination rate constants) were assumed, consistent with modern ADC development programmes. Two observation tiers were evaluated: Tier 1 comprising total mAb, total ADC, and unconjugated payload concentrations (standard clinical bioanalytical package), and Tier 2 additionally incorporating average DAR measurement. Structural identifiability analysis was performed using transfer function methods exploiting the block-triangular system structure, whereby the antibody subsystem (DAR2/DAR1/DAR0 cascade) was analysed independently, followed by the payload subsystem treated as driven by known antibody inputs. This decomposition approach enabled tractable symbolic analysis of the eight-state system.
Results
Under Tier 1 observations, the deconjugation rate constant (kdec) was structurally identifiable, though estimation precision was limited by reliance on differences between declining total mAb and total ADC concentrations. The bioavailability fraction F was identifiable from payload area-under-curve relative to expected values assuming complete bioavailability. However, depot parameters krel and ksink were not separately identifiable; only their sum (kdepot) was conditionally identifiable when the terminal payload phase was distinguishable from the known payload elimination rate constant—otherwise confounded with flip-flop kinetics. Tier 2 observations substantially improved kdec identifiability, as average DAR decline is driven purely by deconjugation and is independent of elimination. Notably, average DAR was demonstrated to be algebraically non-derivable from Tier 1 observables for DAR≥2 systems, confirming its status as genuinely new information. The depot return fraction (freturn = krel/kdepot) remained unidentifiable under both tiers, necessitating assumptions or bounds for complete model specification. Practical measurement of albumin-bound payload (which would resolve depot dynamics) was deemed infeasible for routine development settings.
Conclusions
Structural identifiability analysis of a minimal DAR2 framework demonstrated that ghost payload bioavailability is identifiable from standard clinical observations, while depot internal dynamics require assumptions that cannot be verified empirically. Average DAR measurement provides substantial value for deconjugation rate estimation and should be considered where mechanistic understanding is prioritised. These findings are contingent on availability of external pharmacokinetic constraints from companion intravenous payload studies and linear deconjugation rates and ADC pharmacokinetics through the cascade. The analytical framework presented provides a foundation for rational bioanalytical investment decisions in ADC development and highlights the inherent limitations in characterising payload trafficking from feasible clinical measurements.
References:
References
1. Shen BQ et al. Conjugation site modulates the in vivo stability and therapeutic activity of antibody-drug conjugates. Nat Biotechnol. 2012;30(2):184-189. doi:10.1038/nbt.2108
2. Khot A et al. Development of a translational physiologically based pharmacokinetic model for antibody-drug conjugates. J Pharmacokinet Pharmacodyn. 2017;44(4):365-379. doi:10.1007/s10928-017-9521-y
3. Sukumaran S et al. Mechanism-based pharmacokinetic/pharmacodynamic model for THIOMAB drug conjugates. Pharm Res. 2015;32(6):1884-1893. doi:10.1007/s11095-014-1582-1
4. Ballard TE et al. Biological activity of modified signal peptides of maleimide-linked antibody-drug conjugates. Bioorg Med Chem Lett. 2017;27(22):5044-5049. doi:10.1016/j.bmcl.2017.10.007
5. Kamath AV. Translational pharmacokinetics and pharmacodynamics of monoclonal antibodies. Drug Discov Today Technol. 2016;21-22:75-83. doi:10.1016/j.ddtec.2016.09.004
Reference: PAGE 34 (2026) Abstr 12210 [www.page-meeting.org/?abstract=12210]
Poster: Oral: Drug/Disease Modelling - Oncology