Applying population mechanistic modelling to find determinants of CAR-T dynamics in month-one lymphoma patients

Chimeric Antigen Receptor (CAR) T-cells have been utilised for the treatment of several malignancies, including Non-Hodgkin Lymphomas (NHL). A myriad of product- and patient-specific factors determine the extent of patient response, and determining which are most impactful requires analysis of clinical data. We used population-level ordinary differential equation models to fit clinical flow cytometry and tumour biopsy data from the TRANSCEND-NHL-001 (NCT02631044) study [1]. We analysed the impact of lymphodepletion, CAR T-cell phenotypes and other factors on CAR T-cell dynamics for thirty days after infusion. We quantified the relative contribution of antigen-dependent and independent sources of proliferation on CAR T-cell dynamics, finding that both make a large contribution, and that antigen-independent proliferation was highly correlated with patient IL-15 and IL-7 blood concentrations. The proportion of CAR T-cells in na¨ıve, memory or effector cells was found to have limited impact on CAR T-cell dynamics, compared to lymphodepletion and tumour burden. This study shows how models can be used to link endogenous T-cells, CAR T-cells and their phenotypes, and may be useful for determining whether a given patient may be responding poorly to treatment, by observing the dynamics of their endogenous T-cells. The framework we developed can be utilised for other CAR T constructs and indications, to test product alterations or biological hypotheses at the population-level.