Publication date: Sep 17, 2025
A key issue in the post-COVID-19 pandemic era is the ongoing administration of COVID-19 vaccines. Repeated vaccination is essential for preparing against currently circulating and newly emerging severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) variants. However, optimizing vaccination strategies is crucial to efficiently manage medical resources and establish an effective vaccination framework. Therefore, a strategy to identify poor responders with lower sustained antibody titers would be beneficial because these individuals should be considered high priority for revaccination. We investigated longitudinal antibody titer data in a cohort of 2526 people in Fukushima, Japan, collected between April 2021 and November 2022. Using mathematical modeling and machine learning, we stratified the time-course patterns of antibody titers after two primary doses and one booster dose of COVID-19 messenger RNA vaccines. We identified three populations, which we refer to as the durable, the vulnerable, and the rapid-decliner populations, and approximately half of the participants remained in the same population after the booster dose. The rapid-decliner population experienced earlier infections than the others. Furthermore, when comparing spike protein-specific immunoglobulin G (IgG) titers, spike protein-specific IgA titers, and SARS-CoV-2-specific T cell responses between participants who experienced subsequent infections after booster vaccination and those who did not, we found that spike protein-specific IgA titers were lower during the early stage after booster vaccination in participants who went on to become infected with SARS-CoV-2. This approach could be used to inform policy decisions on vaccine distribution to maximize population-level immunity both in future pandemics and in the post-COVID-19 pandemic era.