Adriana Savoca (1), Christina Vasalou (2), Cesar Pichardo-Almarza (3), Pavan Vajjah (4)
(1) Discovery Translational PKPD, Clinical Pharmacology and Quantitative Pharmacology, R&D Biopharmaceuticals, AstraZeneca, Cambridge, UK, (2) Discovery Translational PKPD, Clinical Pharmacology and Quantitative Pharmacology, R&D Biopharmaceuticals, AstraZeneca, Waltham, MA, US, (3) Systems Medicine, Clinical Pharmacology and Quantitative Pharmacology, R&D Biopharmaceuticals, AstraZeneca, Cambridge, UK ,(4) Clinical Pharmacology and Quantitative Pharmacology, R&D Biopharmaceuticals, AstraZeneca, Cambridge, UK
Introduction/Objectives: Antibody-Drug Conjugates (ADCs) aim at exploiting the targeted nature of monoclonal antibodies (mAb) to selectively deliver cytotoxic payloads into tumour cells. Despite the exciting “magic bullet” concept and 24 years since the first approval of this modality, only 15 of more than 250 clinically tested ADCs [1] have been approved to this day, with several discontinued because of limited efficacy or excessive toxicity. Quantitative approaches integrating learnings from clinical data with mathematical model simulations can elucidate the effect of key design parameters on ADC and unconjugated payload disposition into tumour cells. Achieving sufficient tumour payload exposure can in fact drive a wider safety margin for ADCs as opposed to classic chemotherapy. Here, we show how we have integrated published data with mechanistic modelling to understand optimal ranges and impact of specific designable parameters, with the goal of identifying learnings on what makes a clinically successful ADC.
Methods: We have gathered available clinical data on total antibody (totAb) and ADC exposures, and affinity values for a number of approved, in-development or discontinued ADCs. In parallel, we have adapted a literature mechanistic model [2] describing the totAb, the payload-conjugated antibody and the unconjugated payload kinetics in plasma, tumour, and peripheral tissue. In tumour, the model accounts for the ADC interaction with the target antigen and target-mediated endocytosis, then release of unconjugated payload upon lysosomal degradation. The free payload can drive cell killing in tumour and influx/efflux rates regulate its diffusion from tumour cells. The model has been used to simulate plasma and tumour payload exposure and compare different scenarios of (i) mAb clearance (CL) and conjugation stability, (ii) affinity to target for different expression levels and internalization rates, and (iii) payload efflux rate and potency. Simulations have been generated in Phoenix WinNonlin (Version 8.3, Certara USA, Inc.).
Results: Analysis of ADC AUC from published clinical data for discontinued/in-development vs approved ADCs does not indicate a clear threshold for success. However, model simulations show how a slower CL (range for ADCs approved on Q3W dosing: ~5-25 mL/day/kg) has a positive effect on tumour payload exposure, with limited effect on plasma payload AUC. Therefore, in the design phase, this is still a key parameter to enhance tumour payload delivery and mediated efficacy. Conjugation stability is considered a critical ADC design parameter. Simulations confirm and clarify this, by showing that for worse stability, the tumour payload exposure is more limited and plasma Cmax is higher. Clinical data based on ADC to totAb AUC ratio show that stability of approved ADCs is quite varied. Hence, while it is an important feature, other mechanisms can compensate and drive efficacy. Simulations also suggest that improving this metric over a certain threshold (~0.9 AUC ratio) achieves minimal benefit.
Affinity range for approved ADCs in solid tumours is 0.026-~5 nM. Simulations of tumour payload exposure show that a lower Kd is beneficial for lower target expression and slower internalization rates, while for fast-internalizing targets and higher expression, sensitivity to Kd is less significant.
For ADCs approved in solid tumours, payloads range from high (10-10-10-9 M, auristatins and maytansinoids) to mid-high potency (10-9-10-8 M, TOP1 inhibitors). Simulations suggest that tumour efflux kinetics should be optimized together with potency. In fact, improved tumour exposure can be obtained by optimizing physico-chemical properties to compensate with a lower potency, therefore reducing probability of adverse effects.
Conclusions: By combining available clinical data with mechanistic model simulations, we have shed light on designable parameters ranges for approved ADCs, their interaction with population-related parameters such as target expression, and their effect on tumour and plasma payload exposure, which can be correlated to clinically observed efficacy and toxicity. Analysis of clinical competitors data and generation of targeted population-related data to inform mechanistic models can guide decision-making across ADC discovery and development programs and improve probability of clinical success, while the ranges we have reported can serve as benchmarks to build confidence in the selected candidates.
References:
[1] Maecker et al. (2023). Mabs, 15(1), 2229101
[2] Singh et al. (2016). The AAPS journal, 18, 861-875
Reference: PAGE 32 (2024) Abstr 11041 [www.page-meeting.org/?abstract=11041]
Poster: Drug/Disease Modelling - Oncology