PII-018 - CLINICAL DOSE PREDICTION FOR CHIMERIC ANTIGEN RECEPTOR T (CAR-T) CELL THERAPIES USING A MULTI-SCALE QUANTITATIVE SYSTEMS PHARMACOLOGY (QSP) PLATFORM MODEL

B. Paleja¹, S. Garai², V. Gangwar¹, S. Yadav¹, G. Nair¹, S. Pallikonda Chakravarthy¹; ¹Vantage Research, ²Indian Institute of Science.

Background: Unlike traditional therapies, factors governing the dose-exposure-response relationship of CAR-T therapies are poorly understood. Further, there are no clear guidelines in this therapy class to allow for preclinical to clinical translation prediction of first-in-human (FIH) doses (1). Understanding the effects of both drug-specific and patient-specific determinants, such as CAR-T dose and baseline tumor burden, on the variability seen in clinical outcomes is of great importance and can be addressed using mechanistic modelling.

Methods: In the current work we have integrated the models from (2) and (3) to capture relevant physiology, and to explain variability in observed clinical efficacy outcomes. This model can be used to address questions of interest for both preclinical and clinical development of CAR-T therapies.

The model consists of the following modules –
i) Pharmacodynamics module to capture interaction between tumor cell and CAR-T;
ii) PBPK module to capture kinetics of CAR-T cells including various physiological differentiation states (memory, effector and exhausted T cells);
iii) PK-PD module to capture tumor growth inhibition (TGI).

We utilized published in-vitro (cell-line cytotoxicity, antigen expression and binding), animal (xenograft efficacy) and clinical data (CAR-T PK, efficacy) for approved CAR-T therapies targeting BCMA and CD19 antigens (4, 5, 6, 7) to calibrate and validate the model. A Virtual population based approach was utilized for prediction of clinical outcomes.

Results: The developed model captured typical multiphasic behaviour of CAR-T cell kinetics, as a function of tumor burden. The model fitted both preclinical and clinical efficacy data in terms of observed TGI. Additionally, by varying the starting drug and patient specific properties, the model was able to predict clinical outcomes for anti-BCMA and anti CD19 CAR-T, which were validated using available clinical data for other CD19 CAR-T therapy

Conclusion: This QSP platform model can take into consideration both drug and patient specific properties, to determine starting doses as well as personalized doses for CAR-T therapies. . Further, the modular nature of the model allows for flexibility in re-purposing the model to other CAR-T therapies and indications.