057: Population Pharmacokinetic Model Development of Autologous CD30.CAR-T Cells in Relapsed or Refractory Classical Hodgkin Lymphoma (cHL) Patients

Statement of Purpose: Chimeric antigen receptor (CAR), genetically engineered receptors, in T cells, are distinct from traditional non-living therapies because of their ability to proliferate and expand post-administration. CAR-T cellular kinetics do not obey conventional pharmacokinetics, but instead, exhibit multiple phases including distribution, expansion, contraction, and persistence, with each phase a reflection of its complex physiological and pathological interactions. Our objective is to develop an initial exploratory model using empirical data from CD30.CAR-T cells to improve our understanding of how these cells behave in relapsed or refractory classical Hodgkin Lymphoma (cHL) patients. This base model will serve to further assess the impact of various intrinsic and extrinsic factors on specific phases of interest, in particular, the expansion and persistence of CAR-T cells, which are believed to strongly correlate with response to treatment.

Description of Methods & Materials: A total of 21 patients from 2 clinical trials were included in our analysis. Exploratory data analysis to obtain maximum observed CD30.CAR-T concentrations (C_max) and area under the curve (AUC) values were performed using traditional non-compartmental analysis (NCA) methods, followed by cellular kinetic modeling up to 180 days post-infusion using nonlinear mixed-effects modeling approaches with Phoenix^® NLME. First-order conditional estimation methods with extended least squares (FOCE-ELS) or Lunstrom-Bates (FOCE-LB) were used to estimate the population mean values of the PK parameters, interindividual variability (IIV) and residual variability (RV). Transient compartment, lag time and multiple compartments were also evaluated. Statistical error models were considered with or without log-transformed data. The model was validated using visual predictive checks (VPC), Q-Q plots, goodness-of-fit (GOF) plots, and objective functions values (OFVs).

Data & Results: Twenty-one (21) patients with 150 observations were included in the current analysis. All patients had lymphodepletion therapy using bendamustine and fludarabine prior to CAR-T infusion. Patients were either treated with CD30.CAR-T alone or in combination with a PD-1 inhibitor (Nivolumab), 5 days after receiving bendamustine/fludarabine lymphodepletion (LD) chemotherapy. NCA analysis resulted in a C_max of 12707.7 copies/µg DNA, AUC0-28 days of 179,124.2 copies*hr/µg DNA, and a T_max range of 7-21 days. The structural model was best described with a two-compartment model with no tlag or transient compartment. In addition, the mixed-ratio error model with log-transformed data provided the best overall objective function minimization. The estimated population clearance and volume of distribution were 0.15 L/hr and 18.32 L with a coefficient of variations (CV%) of 0.163 and 0.119 respectively.

Interpretation, Conclusion or Significance: In this model development study with cHL patients receiving CAR-T against CD30 with or without Nivolumab, a two-compartment model with log-transformed data using a mixed ratio-error model generally resulted in an adequate model using standard model evaluation techniques. Different malignancies and CAR-T constructs may differ in their kinetic profiles. This model will serve as a starting point for co-variate evaluation in cHL patients receiving targeted CD30 CAR-T therapy.

Disclosures: