038: Development & Qualification of a Population Model of Gene Therapy for the Treatment of Hemophilia A & Comparison With FVIII Therapies

Statement of Purpose: Hemophilia A is an inherited X-linked hemorrhagic disorder resulting in the absence of or deficiency in coagulation Factor VIII (FVIII). The current prophylaxis therapies for haemophilia A are regular (every 2-3 days) infusion of exogeneous FVIII concentrates due to the relatively short half-life of these products. Gene therapy (GT) targeting haemophilia A involves intravenous administration of the FVIII transgene within an Adeno-Associated Virus (AAV) capsid. Sufficient FVIII expression in patients with severe haemophilia leading to significantly reduced bleeding episodes has been reported. We propose a novel PopPK model for GT to evaluate GT against current therapies utilizing clinical data from a recent trial.

Description of Methods & Materials: FVIII concentrations were graphically extracted from 43 participants as reported by Ozelo et al., N. Engl. J. Med.,386, 11, 2022. A semi-mechanistic model with Ordinary Differential Equations (ODEs) was proposed for GT, explaining AAVs (in vector genomes [vg]) encoded with FVIII concentration in plasma (in IU/dL). The AAVs are depleted as the cells become transfected via nuclear translation. FVIII is upregulated at a quadratic rate proportional to the AAVs and has a linear (intrinsic) clearance with a saturable growth proportional to AAVs. The PopPK model consists of ODE predictions (of FVIII and AAV), intersubject variability and a mixed residual error. Dose (vg/kg) was determined via imputed body weights. Posterior distributions of model parameters were obtained via Markov Chain Monte Carlo (MCMC) simulations with Uniform (population) and Lognormal (individual) priors. Convergence measures (visual and Gelman and Rubin Factor), observations vs predictions and individual predictions vs observations plots were obtained with 10,000 posterior samples from three independent chains. Model has been qualified via comparison with observed weekly statistics using simulations for 132 individuals. Predictions from GT were compared against predictions for five other prophylaxis products using summary statistics of FVIII concentrations (Figure 1) for 500 subjects.

Data & Results: Three MCMC chains showed satisfactory convergence. Goodness of Fit measures showed linearity between predicted vs observed FVIII concentrations. The overall standard error of the fit for the average FVIII concentration (over time) is 6 IU/dL. Credible intervals for Maximum a Posteriori (MAP) estimates of parameters were within 11% lower and 60% higher than the MAP estimates on average. The optimized model has adequately captured the median and individual FVIII predicted concentrations. The predicted Interquartile Range (IQR) was close to the observed at the end of one-year period (46.0 Predicted vs 45.8 Observed) (Table 1). Comparison of week 49-52 median FVIII concentrations (n=500) of gene therapy with predicted steady state values for other products was within comparable range for Efm.alfa and Dam.alfa (Figure 1).

Interpretation, Conclusion or Significance: The present model fills the gap of lack of PopPK model for Gene Therapy for Hemophilia A. Our overall results indicate that the proposed model can explain the central tendency and the high variability seen in clinical data and will be useful in predicting long-term outcomes of gene therapy, determining future dosing regimens for various patient groups and predicting other outcome measures such as annualized bleeding rates, given that individual covariates are readily available.

Disclosures: The authors serve as consultants for the pharmaceutical industry.