Dynamics of the humoral immune response to a two-dose heterologous vaccine regimen against Ebola virus.
Alexandre Marie (1), Prague Mélanie (1), Clairon Quentin (1), Van Effelterre Thierry (2), Thiébaut Rodolphe (1)
(1) INSERM U1219, INRIA SISTM, University of Bordeaux, France, (2) Janssen Pharmaceutica N.V., Beerse, Belgium
Introduction: The two-dose heterologous vaccine regimen Ad26.ZEBOV,MVA-BN-Filo against Ebola virus has been evaluated in multiple clinical trials. However, the long-term duration of the immune response to the vaccine is still unknown. Pasin et al.[1] proposed to evaluate the long-term humoral immune response using an ordinary differential equation (ODE)-based mechanistic model. Based on data from 177 participants from three phase I trials of the EBOVAC Consortium, the lower bound of the half-life of the long-lived antibody secreting cells (LL ASCs) was estimated at 6 years with [2.7 ; 13] years as 95% confidence interval. Meanwhile, phase II and III trials achieved a follow-up of more than 2 years allowing a validation of the predictions and an update of the estimations.
Objectives: We aim at validating the predictions at the individual level, as well as updating the estimation of the parameters by combining “all available” data.
Methods: Mechanistic models are defined by three elements: i) the mathematical model, ii) the statistical model, and iii) the observation model. i) The longevity of the humoral response is modelled using a system of ODEs describing the dynamics of the antibody (Ab) response and both long-lived (LL) and short-lived (SL) antibody-secreting cells. ii) Mixed-effect models allow to model the inter-individual variability. iii) The concentration of Ab was measured by ELISA and a constant error model was assumed. For the validation procedure, first individual parameters (random effects) of participants of phase II trials were estimated with Empirical Bayes Estimates (EBEs) using data restricted to the first year of follow up and without any modification of population parameters or random-effect distributions. Based on these results, the quality of the predictions of the model was evaluated by comparing the long-term model predictions and the observations measured after the first year of follow up in the phase II trials. In order to consolidate the knowledge about the durability of the long-term immune response, we re-estimated the model using all available phase I and phase II data.
Results: In addition to the 44 participants from the three phase I trials receiving Ad26.ZEBOV as first injection and MVA-BN-Filo as second injection 56 days later, we evaluated the ODE-based model on data from the 454 participants involved in three phase II trials. Among these participants, 82 Europeans and 137 Africans were followed up for one year and 235 Africans were followed for two years after the first injection. The validation of EBEs-based individual predictions demonstrated the robustness of the model estimated by Pasin et al. with more than 90% of the observations measured between the 1st and the 2nd years of follow-up falling within the individual 95% prediction intervals. Finally, the re-estimation of the model enabled to update the value of the lower bound of LL ASCs half-life to 15 years.
Conclusions: Through a 2-step method, we validated the ODE-based model as well as its quality of prediction by considering all new available data. Moreover, we extended our knowledge about the LL ASCs half-life thanks to the information provided by a longer follow-up of patients in those trials.
References:
[1] Chloé Pasin et al. “Dynamics of the humoral immune response to a prime-boost Ebola vaccine: quantification and sources of variation”. In: Journal of virology 93.18 (2019).