Empirical Population Pharmacodynamic Models for beti-cel for the Treatment of Patients with β-Thalassemia who Require Regular Transfusions.
Shaun Kumar (1), Shamshad Ali (2), Gloria Tao (2), Himal Thakar (2), David McDougall (1), Bruce Green (1)
(1) Parexel International, Australia, (2) bluebird bio, USA
Introduction: β-thalassemia is a rare hereditary genetic blood disease caused by mutations in the β-globin gene, resulting in reduced or absent production of functional hemoglobin. The standard care for patients with transfusion-dependent beta-thalassemia (TDT) requires regular (lifelong) transfusions of packed red blood cells. Recently the FDA approved an autologous transplant/ex-vivo gene therapy beti-cel (betibeglogene autotemcel; commercial brand name: Zynteglo), beti-cel is a gene therapy designed to add functional copies of a modified β-globin gene (via a lentiviral vector [LVV]encoding the βA-T87Q-globin gene) into TDT patients own hematopoietic stem cells (HSCs, specifically CD34+ cells). A lentiviral vector encoding the βA-T87Q-globin gene (producing hemoglobin containing βA-T87Q-globin, HbAT87Q) was transduced into the patients HSCs. Peripheral blood vector copy number (PB VCN, expressed as average number of vector copies per diploid genome; c/dg) is a measure of drug product (DP) vector copies that is used to monitor transgene integration. Over the course of clinical development, the manufacturing process was optimized from DP process 1 to DP process 2 (the commercial DP process), specifically the transduction step to increase the percentage of vector-positive transduced cells (DP %LVV+ cells) and the total VCN.
Objectives: To develop a pharmacodynamic (PD) model to describe the time course PB VCN and HbAT87Q post beti-cel infusion and specifically:
- Identify and quantify both patient and DP specific covariates that impacted PB VCN and HbAT87Q time to approach steady-state and thereby understand the dose-exposure-response relationship for beti-cel.
Methods: As of 09Mar2021, data (n=59) were available from three clinical studies (HGB-204, HGB-207, HGB-212) and the long-term follow-up study (LTF-303). Patients had either the non-β0/β0 genotype (n=39) or the β0/β0 genotype (n=20). A total of 566 PB VCN and 560 HbAT87Q observations were analyzed using a non-linear mixed effects modeling approach in NONMEM (Version 7.4, ICON, Hanover, MD), specifically using $PRED. A non-parametric bootstrap was performed to determine the 95% confidence interval (CI) for each model parameter. Simulations were conducted using the final model to assess the impact of covariates on the time to approach 90% of the steady-state values for both PB VCN and HbAT87Q.
Results: PB VCN data was best described by an exponential asymptotic growth model in terms of maximum VCN (VCNMAX, 0.831 c/dg, 95%CI 0.597 – 1.07) and rate of transgene appearance (KTA, 5.87/month, 95%CI 0.0081 – 0.0159). The statistically significant covariates identified were DP manufacturing process and DP %LVV+ cells. HbAT87Q data was described by an Emax model with a Hill slope. The parameters estimates were maximum HbAT87Q (HbMAX, 9.4 g/dL, 95%CI 8.4 – 12.3), time to 50% maximal HbAT87Q (ET50, 2.42/month, 95%CI 2.24 – 2.66), and Hill slope (γ, 3.85, 95%CI 3.28 – 4.39) The statistically significant covariates were the individual maximum PB VCN (derived from the PB VCN model) and DP manufacturing process.
Patients receiving DP Process 1 compared to patients that received DP Process 2 were had a lower median steady-state maximum PB VCN (0.336 c/dg vs 1.50 c/dg) and HbAT87Q (5.66 g/dL vs 8.85 g/dL) and a longer median time to achieve 90% steady-state maximum PB VCN (1.66 months vs 0.393 months) and HbAT87Q (6.44 months vs 3.91 months). Age, weight, sex, race, or genotype did not impact PB VCN or HBAT87Q time-course profiles.
Conclusions: PB VCN and HbAT87Q were fit to empirical time-based models, similar approaches have been used for other cell or gene-based therapies such as CAR-T cell therapy. Utilizing empirical models in this setting provided a fast and robust method for identifying important covariates as well as extrapolating steady-state values compared to more complex methods such as quantitative systems pharmacology (QSP).