Mechanistic models of humoral kinetics following COVID-19 vaccination

AbstractIntroductionFuture COVID-19 vaccine programmes need to take into account the variable responses elicited by different vaccines and their waning protection over time. Existing descriptions of antibody response to COVID-19 vaccination convey limited information about the mechanisms of antibody production and maintenance.MethodsWe describe the antibody dynamics elicited by COVID-19 vaccination with two biologically-motivated mathematical models of antibody production by plasma cells and subsequent decay. We fit the models using Markov Chain Monte Carlo to seroprevalence data from 14,602 uninfected individuals collected via the primary care network in England between May 2020 and September 2022. We ensure our models are structurally and practically identifiable when using antibody data alone. We analyse the effect of age, vaccine type, number of doses, and the interval between doses on antibody production and longevity of response.ResultsWe find evidence that individuals over 35 years of age who received a second dose of ChAdOx1-S generate a persistent antibody response suggestive of long-lived plasma cell induction, while individuals that receive two doses of BNT162b2, or one dose of either vaccine do not. We also find that plasamblast productive capacity, the likely driver of short-term antibody responses, is greater in younger people than older people (≤ 4.5 fold change in point estimates), people vaccinated with two doses than people vaccinated with one dose (≤ 12 fold change), and people vaccinated with BNT162b2 than people vaccinated with ChAdOx1-S (≤ 440 fold change). The effect of age on antibody dynamics is more pronounced in people vaccinated with BNT162b2 than people vaccinated with ChAdOx1-S. We find the half-life of an antibody to be between 23 – 106 days.ConclusionRoutinely-collected seroprevalence data are a valuable source of information for characterising within-host mechanisms of antibody production and persistence. Extended sampling and linking seroprevalence data to outcomes would allow for powerful conclusions about how humoral kinetics protect against disease.