(M-057) Quantitative Systems Pharmacology Modeling to Support First-in-Human Dose Selection for a CD3 Bispecific Antibody in B-Cell Lymphoma and Renal Cell Carcinoma

Yannuo Li – Incyte Corporation; Limei Cheng – Incyte Corporation

Objectives: Bispecific antibodies targeting CD3 and tumor-associated antigens (TAA) are promising immunotherapeutic agents but pose challenges in dose selection due to a risk of cytokine release syndrome (CRS). To support the first-in-human (FIH) dose selection for a CD3 bispecific antibody, we developed an integrated Quantitative Systems Pharmacology (QSP) model to predict tolerable doses in patients with B-cell lymphoma (BCL) or renal cell carcinoma (RCC).

Methods: The QSP model incorporated data obtained from in vitro cytotoxicity assays, in vivo pharmacokinetic data, and some publicly available data. Publicly available interleukin-6 (IL-6) data for the bispecific antibody, mosunetuzumab, were included as a benchmark. This benchmark strategy was used with the aim of addressing the challenge of translating cytokine release data from animal models to humans (in cases where bispecific antibodies do not elicit a response in animal models). The model included central, peripheral, and tumor compartments to capture drug distribution and cytokine release. The IL-6 release at each compartment was modeled as a function of synapse formation, and the dynamics of immune cells and TAA cells. The model also accounted for disease-specific tumor compartments for BCL and RCC, where systemic IL-6 levels account for cytokines released from both the disease site and plasma. The model was calibrated with in vitro and publicly available clinical data for the benchmark compound. Both system-specific and drug-specific parameters were incorporated to generate a virtual patient population, enabling predictions of cytokine dynamics.

Results: The model simulation provided peak IL-6 levels across all dose levels of interest among the generated virtual patient population. Synapse formation under different doses was simulated for efficacious dose prediction. The model is capable of capturing disease-specific differences in cytokine release predictions. The peak IL-6 level for the CD3 bispecific antibody at the minimal anticipated biological effect level proposed FIH dose remained low and was well below the predefined safety threshold. The virtual clinical trial simulation provided IL-6 time-response data for different potential step-dose regimens.

Conclusions: This integrated QSP modeling approach provided quantitative justification for FIH dose selection for the CD3 bispecific antibody in the treatment of patients with BCL and RCC. By incorporating preclinical and clinical benchmarking data, the model supported predicted tolerable starting dose selection to mitigate CRS risk, while preserving potential anti-tumor efficacy. The model can help de-risk early clinical development of bispecific antibodies and streamline interactions with regulatory authorities. Future developments include clinical validation of the model predictions with emerging patient data.

Citations: N/A

Keywords: QSP, bispecific antibody, first-in-human dose selection