SARS-CoV-2 Seropositivity among US Marine Recruits Attending Basic Training, United States, Spring–Fall 2020

In a study of US Marine recruits, seroprevalence of severe acute respiratory syndrome coronavirus 2 IgG was 9.0%. Hispanic and non-Hispanic Black participants and participants from states affected earlier in the pandemic had higher seropositivity rates. These results suggest the need for targeted public health strategies among young adults at increased risk for infection.

weeks. The recruits and staff were forbidden to leave and no visitors, other than persons delivering supplies and food, local essential workers, and study staff, were allowed onto the premises. All of these measures were enforced by Marines at all times. Specific public health measures have been previously described (1).
Within 48 hours of arriving at the quarantine location, ≈350-500 recruits per week were offered the opportunity to volunteer for the COVID-19 Health Action Response for Marines (CHARM) Study, which included collecting baseline SARS-CoV-2 serologic test results.
Recruits were eligible if they were >18 years and could complete follow-up encounters. Recruits 17 years of age were ineligible. Study enrollment occurred after recruits had been in-processed and had personal effects (including cell phones) secured, rooms assigned, and gear issued. The recruits attended a group consent brief of 50-100 participants using an ombudsman who explained the study, exactly what was being asked of participants, risks, benefits, and the state of COVID-19 in the recruit setting. Since recruits are a vulnerable population and at risk for coercion, special measures were undertaken including study briefers, who are active duty Navy personnel, wearing civilian clothes, not disclosing military ranks, not having members in the recruit's chain of command present, and ensuring that participation would not affect a recruit's medical care or influence the grading of a recruit's military performance.
Institutional Review Board approval was obtained from the Naval Medical Research Center (protocol no. NMRC.2020.0006) in compliance with all applicable federal regulations governing the protection of human subjects. All participants provided written informed consent for study participation.

Procedures
Recruits consented to undergo a mid-turbinate nares swab for SARS-CoV-2 qPCR testing and blood draw upon enrollment that included serum. We collected paper questionnaires (Appendix Figure) to identify demographics, risk factors, and symptoms, and assayed serum for the presence of SARS-CoV-2 IgG upon arrival at the quarantine location. Data was first recorded in Microsoft Excel spreadsheets before automated integration with the statistical programming language R 3.6.3 (2). The data collected included sex, age, ethnicity, race, place of birth, state or country of residence, medical history including smoking or vaping or exposure to secondhand smoke, and risk factors including use of masks, practicing self-quarantine before arrival, recent travel, known exposure to persons with COVID-19, and exposure to someone with flu-like or other respiratory illness.

Laboratory Methodology
Presence of SARS-CoV2 IgG in serum was evaluated using ELISA with some modifications from Amanat et al. (3), as previously described (1). Briefly, 384-well Immulon 4 HBX (Thermofisher, https://www.thermofisher.com) plates were coated overnight at 4°C with recombinant His-tagged Spike (S) receptor-binding domain (RBD) (SinoBiological, https://www.sinobiological.com) at a concentration of 2 µg/ml in phosphate-buffered saline (PBS). Plates were washed 3 times with 0.1% Tween-20 (Fisher Scientific) PBS (PBS-T) using an automated ELISA plate washer (Aquamax 4000, Molecular devices), and blocked for 1 h at room temperature (RT) with 3% milk PBS-T. Blocking solution was removed, and serum samples diluted in 1% milk PBS-T were dispensed in the wells. At least 2 positive controls (serum samples with known SARS-CoV-2 IgG presence), 8 negative controls (serum samples collected before July 2019) and 4 blanks (no serum) were included in every plate. Plates were incubated for 2 h at room temperature and washed 3 times with PBT-T. Next, peroxidase conjugated goat F(ab')2 Anti-Human IgG (abcam) were added at a dilution 1:5,000-1:10,000 dilutions (determined after optimization for each antibody lot) in 1% milk PBS-T, and plates were incubated for 1 h at RT. Plates were washed 6 times with PBS-T, developed by using SIGMAFAST OPD (Sigma-Aldrich, https://www.sigmaaldrich.com), and the reaction was stopped after 10 min with 3M HCl. Optical density (OD) at 492 nm was measured by using a Spectramax M2 microplate reader (Molecular Devices, https://www.moleculardevices.com). All serum samples were screened at a 1:50 dilution. Those samples with an OD 492 nm value higher than the average of the negative controls plus 3 times their SD in the screening underwent titration assay (6 serial 1:3 serum dilutions starting at 1:50). Serum samples were considered positive when at least 2 consecutive dilutions showed higher OD 492 nm than the average of the negative controls plus 3 times their SD at the correspondent dilution or 0.15 OD 492 nM.

Statistical Analyses
Analyses, figures, and tables were generated by using R 3.6.3 (2). Associations between demographics, risk factors, and IgG-positivity variables were analyzed with logistic regression to compute the p value and the odds ratio. None of the risk factor data (Appendix Figure) was statistically significant and is not displayed. Significance was a priori established at <0.05.
The logistic regression is analyzed with 2 approaches: a) single variable approach: log 1− = 0 + and b) multivariate approach: log 1− = 0 + 1 1 + 2 2 + ⋯ + . Note that when variable is a categorical variable with possible values coded from 1 to and the code 1 is for the reference group, should be understood as ∑ ( =2 = ) , .
The collinearity for the variables in the multivariable logistic regression was assessed by using GVIF (generalized variance-inflation factors) (4).

2 *
( is the degree of freedom of the variable) is computed for all variables in this paper. All variables were less than 1.06, indicating collinearity did not impact the analysis or violate assumptions. The collinearity is also assessed by the conditioner number which is ≈12, less than the 30, also indicating weak collinearity.
The trend test for the weekly IgG-positive rate of participants of Hispanic ethnicity is based on the Cochran-Armitage test. Because of the relatively small number of participants in the first study week (May 11), the participants' weekly IgG-positive rates have been smoothed with a 3-week running mean.
Race and ethnicity were categorized as non-Hispanic White, non-Hispanic Black, non-Hispanic Other, and Hispanic. A total of 18/3,196 (0.6%) participants did not supply any information on race or ethnicity and were grouped into the non-Hispanic Other category.
The 2020 US census data was downloaded from https://www.census.gov/data/tables/2020/demo/popest/2020-demographic-analysis-tables.html on December 20, 2020. The data contain information regarding the percentage of the US population that identifies as Black or Hispanic for each age year, but subcategories of race for the non-Hispanic population are still unavailable. For this reason, we compared data for the Black category, which was available in the census data, with data for non-Hispanic Black participants within our study. Specifically, we focused on the percentage of Black and Hispanic persons 18-20 years of age in the general population compared with our non-Hispanic Black and Hispanic study participants within the same age range.
Daily COVID-19 cases confirmed by viral tests during January 22-September 7 were downloaded from the COVID-19 Data Repository of the Center for Systems Science and Engineering (CSSE) at Johns Hopkins University (5). The heatmap in Figure 2, panel A (https://wwwnc.cdc.gov/EID/article/27/4/20-4732-F2) represents the cumulative confirmed COVID-19 cases per 1 million of the state population. In the heatmap, states are clustered by temporary profiles of cumulative confirmed COVID-19 cases, as indicated by the dendrogram, which separates the states into 3 major groups. The groups were Early Spring, for states in which the first outbreak began in March; Late Spring, for states in which the outbreak began in early June; and Summer, for states in which the outbreak began in late June-July. The overall profile of the whole country (labeled US on 1 row) is in the Late Spring group and is placed in a black box. We used the aggregated data of each state group to compute the cumulated rate (dotted lines in Figure 2, panel B, right axis); the first outbreak is identified by the first local maximum slope.
Our study had 701 (21.9%) participants from Early Spring states, 1,389 (43.5%) from Late Spring states, and 994 (31.1%) from Summer states. A total of 112 (3.5%) participants were not included in the analysis since they resided in a foreign country or did not provide a residence.

Specificity and Sensitivity of SARS-CoV-2 S-RBD IgG Serologic Test
To determine the specificity of the S-RBD IgG ELISA assay, we used 70 commercial serum samples drawn before July 2019 (44 purchased from BioChemed Services and 26 provided by Dr. Russell Tracy, Larner College of Medicine, University of Vermont, Burlington, Vermont, USA). To determine the sensitivity of the assay, we used 51 serum samples from subjects that had been previously confirmed as SARS-CoV-2-positive by PCR >14 days before serum sample collection (all of them were <90 days from PCR-positive test). All samples were screened at a 1:50 dilution, and those identified as positive were titered using 6 serial 1:3 serum dilutions (starting at 1:50). Those with at least 2 positive consecutive dilutions in the titration step (titer of 1:150) were considered seropositive. This assay was shown to have a 97.14% specificity (95% CI 93.24-100.00) and 96.08 sensitivity (95% CI 90.75-100.00).