National seroepidemiological study of COVID‐19 after the initial rollout of vaccines: Before and at the peak of the Omicron‐dominant period in Japan

Abstract Background Based on routine surveillance data, Japan has been affected much less by COVID‐19 compared with other countries. To validate this, we aimed to estimate SARS‐CoV‐2 seroprevalence and examine sociodemographic factors associated with cumulative infection in Japan. Methods A population‐based serial cross‐sectional seroepidemiological investigation was conducted in five prefectures in December 2021 (pre‐Omicron) and February–March 2022 (Omicron [BA.1/BA.2]‐peak). Anti‐nucleocapsid and anti‐spike antibodies were measured to detect infection‐induced and vaccine/infection‐induced antibodies, respectively. Logistic regression was used to identify associations between various factors and past infection. Results Among 16 296 participants (median age: 53 [43–64] years), overall prevalence of infection‐induced antibodies was 2.2% (95% CI: 1.9–2.5%) in December 2021 and 3.5% (95% CI: 3.1–3.9%) in February–March 2022. Factors associated with past infection included those residing in urban prefectures (Tokyo: aOR 3.37 [95% CI: 2.31–4.91], Osaka: aOR 3.23 [95% CI: 2.17–4.80]), older age groups (60s: aOR 0.47 [95% CI 0.29–0.74], 70s: aOR 0.41 [95% CI 0.24–0.70]), being vaccinated (twice: aOR 0.41 [95% CI: 0.28–0.61], three times: aOR 0.21 [95% CI: 0.12–0.36]), individuals engaged in occupations such as long‐term care workers (aOR: 3.13 [95% CI: 1.47–6.66]), childcare workers (aOR: 3.63 [95% CI: 1.60–8.24]), food service workers (aOR: 3.09 [95% CI: 1.50–6.35]), and history of household contact (aOR: 26.4 [95% CI: 20.0–34.8]) or non‐household contact (aOR: 5.21 [95% CI:3.80–7.14]) in February–March 2022. Almost all vaccinated individuals (15 670/15 681) acquired binding antibodies with higher titers among booster dose recipients. Conclusions Before Omicron, the cumulative burden was >10 times lower in Japan (2.2%) compared with the US (33%), the UK (25%), or global estimates (45%), but most developed antibodies owing to vaccination.

Conclusions: Before Omicron, the cumulative burden was >10 times lower in Japan (2.2%) compared with the US (33%), the UK (25%), or global estimates (45%), but most developed antibodies owing to vaccination.  19 pandemic. Although these surveys were conducted partly via convenient sampling, even in Tokyo with one of the highest reported cases (both absolute numbers and per population) consistently throughout the pandemic, the seroprevalence was extremely low in June 2020 (0.1%) after the first wave, and December 2020 (1.4%) at the beginning of the third wave. 2,3 The rollout of the primary series of COVID-19 vaccines started in mid-February 2021. BNT162b2, mRNA-1273, and AZD1222 have been approved for use (with minimal use of AZD1222). 4 The primary series (doses 1 and 2) followed manufacturer-recommended intervals.

| Survey design and participants
Using the Basic Resident Register, residents of Miyagi, Tokyo, Aichi, Osaka, and Fukuoka prefectures were randomly sampled via multistage sampling ( Figure S1). Four prefectures (Tokyo, Aichi, Osaka, and Fukuoka) were chosen to represent urban areas with a high burden of SARS-CoV-2 infection, and Miyagi prefecture was selected to represent a rural area with a low burden. For each prefecture, at least one municipality from each of the following three municipality types was chosen for the surveys: Small (<100 000 population), medium (≥100 000 population), and large (ordinance-designated city/special ward). Samples were split to the relative population sizes of the municipality, and residents were randomly sampled from each of these municipalities. We planned to enroll a total of 30 000 individuals for two surveys in five prefectures (3000 individuals per prefecture per survey). Assuming a response rate of 40%, we randomly sampled 75 000 individuals aged 20 years or older and invited them via mail. Only one participant was allowed from each household. Individuals who agreed to participate in the study answered a selfadministered questionnaire and came to the designated site, where they provided written consent and had their blood drawn. Different participants were included in the two surveys without overlap. Briefly, the questionnaire included general information (age, sex, occupation, etc.), comorbidities, COVID-19 vaccination status, history of household and non-household close contact, and history of past SARS-CoV-2 infection. In planning the investigation, we referred to the "Population-based age-stratified seroepidemiological investigation protocol for COVID-19 infection" published by the World Health Organization (WHO), 5 and the study was considered "Unity-aligned" based on the criteria previously described. 6

| Serology
We utilized two widely used commercial electrochemiluminescence immunoassays to detect vaccine-induced and infection-induced antibodies; one was Elecsys Anti-SARS-CoV-2 (Roche), which detects binding antibodies against the nucleocapsid protein (anti-N antibodies), and the other was Elecsys Anti-SARS-CoV-2 S (Roche), which detects antibodies against the receptor-binding domain of the ancestral strain spike protein (anti-spike [S] antibodies). Because Security Agency). 7-9 These assays are also reported to detect antibodies for a more extended period compared with other assays. 10

| Data analysis
The main endpoint was vaccine-induced and infection-induced seroprevalence. For the main estimates as well as prefecture-and sexspecific estimates, we further calculated estimates weighted by age group. We also conducted logistic regression to identify associations between various factors and past infection, adjusting for age group, sex, city of residence, and/or vaccination status. These potential confounders were determined a priori. We used Kruskal-Wallis test with Dunn's test to compare anti-S antibody titers. Data analyses were per-   with the general population of Japan, a smaller proportion of individuals were in their 20s (6.6% vs. 12.0% in the study and general population, respectively) and ≥70 years (16.3% vs. 27.1%), whereas more individuals were between their 40s and 60s (63.9% vs. 47.8%). 11 The proportion of individuals who had received at least one dose was 96.2%, which was higher than the vaccination coverage among individuals ≥20 years in Japan (87.5% as of December 1, 2021, and 88.0% as of

| Quantitative assessment of anti-S antibody titers
We We next looked at levels of antibodies by age group and the number of vaccine doses received ( Figure 3B). There was a decreasing trend in antibody titers as the age increased. Individuals who received three doses of vaccines had higher titers than those who received two, regardless of age group. When we examined anti-S antibody titers among vaccinees with underlying diseases, individuals who received two doses of vaccine with an underlying disease had lower anti-S antibodies than those without ( Figure 3C). This was especially evident among individuals with heart disease, diabetes mellitus, cancer, and immunodeficiency, or who used immunosuppressants. In contrast, individuals who received three doses of vaccine with an underlying disease had similar antibody titers to those without ( Figure 3D).
Finally, we examined antibody titers by time since the second vaccination ( Figure 3E). We observed a slight decrease in antibody titers over time but a robust increase after the third dose.

| DISCUSSION
In this national seroepidemiological investigation conducted before Europe and much lower than those in other countries, refuting the criticism that Japan was not testing enough, resulting in lower reported case counts. 16 Interestingly, approximately 80% of    (Table S1). Because the vaccination rate was higher among the study population than among the baseline population, the infection-induced seroprevalence may be underestimated and anti-S seroprevalence may be overestimated. Second, we used self-reported questionnaires to obtain metadata, so biases such as recall bias and social desirability bias are possible. Third, we focused on five prefectures for this investigation, with four in urban prefectures; hence, the actual prevalence in the entire nation may be even lower. Fourth, we did not include individuals <20 years of age. Fifth, the serology assays used have high sensitivity/specificity, but the waning of antibodies may lead to the misclassification of serostatus. However, this concern may be minor in Japan, where the infection burden was very low before the Delta wave ( Figure 1).

| Conclusions
This national seroepidemiological investigation indicates that the overall burden of infection was much lower in Japan than in other countries. The majority of the population remained infection-naïve at the peak of the Omicron-dominant period but developed anti-SARS-CoV-2 spike antibodies owing to vaccination. Despite the low overall burden, some sociodemographic groups such as those residing in urban prefectures, younger age groups, being unvaccinated, and individuals engaged in occupations such as long-term care workers, childcare workers, and food service workers had higher odds of past infection.
Our study complemented and supported the routine surveillance activity to provide a better picture of the cumulative burden of COVID-19.

ACKNOWLEDGMENTS
We thank the Miyagi, Tokyo, Aichi, Osaka, and Fukuoka prefecture governments for their support in implementing the study. We also thank Jun Sugihara for support from the Ministry of Health, Labour and Welfare, Japan. We also thank the staff members at the Survey Research Center and Mitsubishi Research Institute for their administrative and technical assistance.

CONFLICT OF INTEREST
The authors declare no conflicts of interest.

ETHICS STATEMENT
The ethics committee of the National Institute of Infectious Diseases approved our study (NIID 1312). Written informed consent was obtained from all participants. writing-review and editing.

PEER REVIEW
The peer review history for this article is available at https://publons. com/publon/10.1111/irv.13094.

DATA AVAILABILITY STATEMENT
Individual-level data of patients included in this manuscript after deidentification are considered sensitive and will not be shared. The study methods and statistical analyses are all described in detail in Section 2 and throughout the manuscript.