Longitudinal antibody dynamics after COVID-19 vaccine boosters based on prior infection status and booster doses … – Nature.com

Study design and participants

This prospective cohort study was conducted in a suburban area of Japan, targeting residents and workers aged18years. Bizen City is a small city located in Okayama Prefecture, in western Japan. We recruited 1972 individuals who either held registered residency or were employed in organizations situated in Bizen City. Participation was entirely voluntary. Individuals indicated their interest in study participation by responding to city-wide public announcements, receiving informational leaflets, or after finding information at medical institutions. The recruitment phase spanned from May to June 2022, with data collection between June 3, 2022 and March 27, 2023. During the study period, Japan experienced two major epidemic waves dominated by the Omicron variant, from July to September 2022 and November 2022 to January 2023. Study participants were requested to undergo antibody level measurement approximately every 2months. Participants were notified about their next antibody measurement appointment via email or telephone, or by the designated contact person within their respective organizations. Participants made an appointment, had their antibody levels measured, and completed a questionnaire survey. Throughout the study, each individual had a maximum of five opportunities for antibody measurement and survey completion. Eligibility criteria included individuals who had received a minimum of three doses of COVID-19 vaccine. Those who never underwent any measurement or lacked information on age or sex were excluded. This study comprised 1763 participants, with a collective total of 7376 antibody measurements taken (ranging from one to five measurements per participant) (Fig.3).

Flowchart of participants.

During the study period, Japan's vaccination strategy was as follows. As of June 2022, initial vaccination and the third booster dose were recommended for all individuals aged12years. Additionally, a fourth booster dose was recommended for individuals aged60years or adults with underlying medical conditions. The required interval between additional doses was at least 5months. In late July 2022, eligibility for the fourth booster dose was expanded to include health care workers in elder care facilities. Starting from September 20, 2022, all individuals aged12years who had completed their initial vaccination series could receive Omicron-compatible vaccinations. From October onward, the interval for additional doses was adjusted to a minimum of 3months. This vaccination strategy was aimed to provide comprehensive coverage and adapt to the challenges posed by emerging variants, particularly the Omicron strain, while considering the vaccination needs of specific populations such as older adults and those with underlying health conditions24.

Information regarding COVID-19 vaccination among residents of Bizen City, including details such as the number of vaccine doses administered, vaccination dates, and types of vaccines used, was obtained from official vaccination records. For non-residents and individuals lacking official vaccination records in Bizen City, we used self-reported vaccination information, updated at each antibody measurement and survey.

SARS-CoV-2 antibody levels were assessed by collecting 30L of blood using fingertip sampling with the SARS-CoV-2 IgM & IgG Quantum Dot Immunoassay (Mokobio Biotechnology R&D Center Inc., Rockville, Maryland, USA). This assay specifically targets SARS-CoV-2 spike receptor-binding domain (S-RBD immunoglobulin G [IgG]) antibodies. For samples with limited blood volume, appropriate dilutions were made prior to measurement and subsequent adjustment was made. To assess the temporal decline in antibody levels, antibody titers were logarithmically transformed.

Data regarding prior COVID-19 infection among participants, including information on infection dates, diagnosis dates, and severity of illness throughout the course, were sourced from official prefecture records. Comprehensive recording of COVID-19 infection data in Japan ceased after September 27, 2022. In cases of non-residents and individuals lacking official records in Okayama Prefecture, we used self-reported information on COVID-19 infection, which was updated at each antibody measurement and survey. Based on epidemiological surveys conducted by the National Institute of Infectious Diseases, the prevalent strains at the time of infection were classified as follows: the original strain (before March 2021), the Alpha variant (April 2021June 2021), the Delta variant (July 2021December 2021), and the Omicron variant (after January 2022)25.

It is important to note that we only considered information about the most recent infection prior to each measurement date in the analyses and did not include future infections occurring after the measurement date.

Information regarding age and sex were collected through the initial survey. In the fifth (final) survey, participants were asked about their height, weight, current medical conditions, immunosuppressive status (including use of immunosuppressive drugs), alcohol history, and smoking history. Those who reported any of the following as current medical conditions were classified as having underlying medical conditions: hypertension, obesity, dyslipidemia, chronic respiratory diseases, chronic kidney disease, diabetes, cardiovascular diseases, cerebrovascular diseases, or malignancies, and body mass index (calculated using height and weight)30kg/m326. For individuals who did not respond to the final survey, information regarding underlying medical conditions, immunosuppressive status, alcohol history, and smoking history was unavailable. However, participants who indicated any of the following current medical conditions in the first survey were similarly classified as having underlying medical conditions: hypertension, obesity, dyslipidemia, chronic obstructive pulmonary disease, angina/heart attack, stroke, or malignancy.

This study was conducted with full cooperation from Bizen City, with active participation by its residents and local businesses. Participants were provided with detailed explanations of the research and provided their informed consent before initial measurements. Participants were informed of their right to withdraw from participation at any time during the study. Additionally, as a preventive measure against COVID-19, masks, hand sanitizers, and other items were distributed to participants at each antibody measurement visit. This study adhered to the ethical guidelines for research involving human subjects in the life sciences and medical fields and received approval from the Institutional Review Board of Okayama University Graduate School of Medicine, Dentistry, and Pharmaceutical Sciences (No. K2205-061).

This study targeted participants aged18years who underwent at least one antibody measurement between June 3, 2022, and March 27, 2023 and had received a minimum of three vaccine doses by the time of measurement. After describing the attributes of participants corresponding to each recent vaccination dose, using both a measurement-based and a participant-based approach, we further categorized each IgG measurement value into previously infected and uninfected groups based on the infection status of the participants at the time of measurement. Additionally, we presented the median, interquartile range, and geometric mean titer of the measured IgG, along with its 95% confidence intervals, stratified by past infection status, the number of most recent vaccinations, and the number of months elapsed since the recent vaccination. We then used box plots to depict the logarithmically transformed antibody levels after each vaccination, categorized by the number of doses and time since vaccination.

We used simple linear regression analysis to visually represent the observed data and to qualitatively assess the temporal dynamics of antibody titers across prior infection status. Subsequent inclusion of a quadratic term in the regression model did not match the decay pattern observed in the actual data, as evidenced by the trajectories plotted (Supplementary Fig. S1 online). In particular, the coefficient associated with the quadratic term was insignificantly small, indicating a negligible deviation from linearity. Therefore, we concluded that a linear regression model was more appropriate to illustrate the gradual decline observed in the empirical data over time. Its important to note that this visualization analysis was designed to elucidate temporal trends and was distinct from our main analysis, which was designed to statistically compare antibody titers between individuals with and without prior infection.

Following the manufacturers instructions for the assay kits, considering the uncertainty of measurements exceeding 30,000 antibody units per milliliter (AU/mL), we modeled the temporal decay of antibody levels post-vaccination using a Bayesian linear mixed-effects interval-censored model with noninformative prior distributions13,27. We opted for noninformative Jeffreys prior distributions to maintain objectivity in our analysis, especially considering the uncertainty associated with measurements exceeding 30,000AU/mL. We used a multivariable model, including COVID-19 history (dichotomous), prevalent strains at time of infection (categories: original, alpha, delta, and omicron), days since infection (continuous), the most recent number of vaccine doses (categorical: 3, 4, 5), sex (dichotomous), and age (categories: 1039, 4059, 6079, and80years) as covariates. To assess the temporal decline in antibody levels post-vaccination, interaction terms with time were included for prior infection, vaccine doses, sex, and age. The model incorporated population-level fixed effects, individual-level random effects for intercepts and slopes, and correlations between random effects. The results were upper-censored at 30,000AU/mL, reflecting the uncertainty of IgG values exceeding the quantification limit. Specifically, data points exceeding 30,000 AU/mL (342/955 datapoints, 26.4% in the previously infected group; 238/5841 datapoints, 3.9% in the uninfected group) were treated as probability distributions that included the upper limit, rather than their actual values.

In sensitivity analyses, we also fitted alternative models, such as the random intercept model and the random intercept and slope model with the quadratic term for time to assess potential non-linear trends. Additionally, the random intercept and slope model without censoring where values above 30,000AU/ml were replaced, was examined to assess the robustness of our findings. All models were conducted in the Bayesian framework with 2500 burn-in iterations and 10,000 iterations performed in posterior simulation. Model evaluation was performed using the Deviance Information Criterion (DIC). We reported detailed information about the models used in our study, including the rationale and codes, following the Bayesian analysis reporting guideline28. Specifically, we used Stata v.18 (StataCorp LLC, College Station, TX, USA) for all analyses and the bayes: metobit command to perform the Bayesian multilevel interval-censored analysis, and we included the code for all models in the Supplementary Table S3 online27,29.

We also conducted a supplementary analysis including underlying medical conditions, immunosuppressive status, smoking history, and alcohol consumption history as covariates to explore potential additional factors influencing antibody dynamics, and excluding 779 data points in which this information was missing.

Go here to see the original:

Longitudinal antibody dynamics after COVID-19 vaccine boosters based on prior infection status and booster doses ... - Nature.com