Risk factors and titers of COVID-19 infection in a longitudinal statewide seroepidemiology cohort

Of 4030 individuals who completed the first seroprevalence survey in June to August 2020 and consented to follow up, we restricted analyses to the 784 (20%) individuals who completed the follow-up study including a blood draw. Most of the other individuals in the cohort did not respond despite at least 3 contact attempts (n = 2241, 56%). A small proportion of individuals declined participation (n = 349, 9%) or were no longer eligible because they moved outside Virginia (n = 107, 3%) or died (n = 31, 1%), with 3 deaths reported from COVID-19. An additional 518 (13%) completed the survey but not the blood draw. Participation rates varied by site, with 38% (n = 299) of participants completing the study in the Northwest region of the state, while only 12% (n = 90) participants completed the study in the Central region.

The demographics of the individuals who completed the follow-up study were enumerated and compared with the first survey (Table S1). Participants were majority white (n = 616, 78.6%), non-Hispanic (n = 749, 95.5%), and 50 years of age or older (n = 477, 60.9%). A higher proportion of participants came from these categories versus the first survey (66.3%, 91.5%, and 49.6% respectively). In addition, 732 (93.4%) reported receipt of at least one dose of a COVID-19 vaccine, 727 (92.7%) were fully vaccinated, and 604 (77.0%) were boosted. Almost half of participants reported “never” spending time indoors without a mask, and reporting indoor dining was rare.

Approximately a quarter of participants (n = 224, 28.6%) were SARS-CoV-2 nucleocapsid antibody positive, while almost all participants (n = 760, 96.9%) were SARS-CoV-2 spike antibody positive (Table S2). Nearly three-quarters (n = 504, 73.4%) of those with positive spike antibody had quantities above our limit of quantification (> 2500 U/ml). Reweighted by region, age, and sex to match the Virginia census data and corrected for diagnostic test characteristics, the seroprevalence of nucleocapsid antibodies was 33.1% (95% CI: 26.6, 39.6; Table 1).

Seroprevalence varied by geographic region, was highest in Central Virginia (43.6%, 95% CI: 27.8, 59.3) and lowest in Southwest Virginia (16.0%, 95% CI: 6.3, 25.7). Seroprevalence was higher in individuals aged 30–49, among non-White and non-Asian participants, and among the uninsured and those with Medicaid insurance. Few individuals (n = 9) who were nucleocapsid seropositive at the time of the first survey were evaluated in the current study. However 7 of 9 remained nucleocapsid positive in this study.

329 individuals (42.0%) reported one or more COVID-19-like-illnesses since the prior survey and among the 229 (29.2%) who were tested for COVID-19 for at least one of these illnesses, 119 (52.0%) tested positive (Table S3). Hospitalization for COVID-19-like illnesses was reported by 18 (2.3%). Positive COVID-19 results on tests administered for other reasons (i.e., while asymptomatic) was rare (11 additional positives among 340 individuals asked about positive tests taken for non-illness reasons). Of the 130 participants that reported any prior positive COVID-19 test, 82.3% (n = 107) were nucleocapsid seropositive. Of 224 total nucleocapsid seropositive individuals, 169 reported a COVID-19-like illness since the beginning of the pandemic, such that 25% of infections could have been asymptomatic, assuming all 169 of these individuals indeed had COVID-19. On the other hand, only 105 of these 169 individuals reported a positive COVID-19 test during their illness, such that 53% of infections could have been asymptomatic, assuming all remaining COVID-19-like illnesses reported were actually COVID-negative.

Risk factors associated with nucleocapsid seropositivity were largely limited to demographic characteristics, known contact with COVID-19 positive individuals, and vaccination (Table 2).

The odds of seropositivity among individuals from Southwest Virginia were lower than that among individuals from other regions (OR: 0.36, 95% CI: 0.16, 0.76). The odds of seropositivity were higher among African American (OR: 2.31, 95% CI: 1.05, 5.05) and Hispanic (OR: 3.44, 95% CI: 1.08, 10.98) participants compared to White and non-Hispanic participants. The odds of seropositivity among the few participants who were uninsured were also higher than the insured (OR: 4.48, 95% CI: 1.18, 17.00). The odds of seropositivity increased by 41% (95% CI: 9, 83) for each additional child living in the household. The odds of seropositivity among tndividuals who had a close contact with a COVID-19 positive individual were almost 3 times the odds among those who did not report such a contact (OR: 2.93, 95% CI: 1.58, 5.43). The odds of seropositivity among individuals who received a COVID-19 vaccine booster were 57% (95% CI: 22, 76) lower compared to those who had not received a booster. None of the behavioral risk factors were statistically significantly associated with seropositivity. However, high frequency of visiting an indoor bar was associated with increased seroprevalence, while wearing a N95-equivalent mask was associated with lower seroprevalence.

In contrast to the nucleocapsid results, the reweighted seroprevalence of spike antibodies (97.5%, 95% CI: 96.1, 98.9) was universally high across regions and subgroups. After adjusting for imperfect sensitivity and specificity of the diagnostic assay, seroprevalence was estimated to be above 100% for almost all subgroups of interest (Table S4). Almost all individuals who reported at least one vaccine dose (n = 723/729, 99.2%) were spike seropositive. The 6 individuals who reported vaccination but were spike seronegative were between 50 and 59 years old (n = 5) and 80 years old (n = 1). All received at least 2 vaccine doses and 5 of 6 had received a booster dose. Most individuals who were spike seropositive and did not report vaccination were also nucleocapsid positive (n = 20/24), indicating natural infection. Vaccinated spike seropositive individuals had higher spike antibody quantities than unvaccinated spike seropositive individuals (74.5% of vaccinees had spike antibody quantities > 2500 U/ml vs. 10.8% of unvaccinated individuals).

Spike antibody quantities were associated with time since last vaccine dose. A lower proportion of individuals had spike quantities above 2500 U/ml as time since last vaccine dose increased (Table 3).

For example, the prevalence of spike levels above 2500 U/ml was 46% lower (95% CI: 22, 62) among individuals vaccinated more than 12 months ago compared to those who received their last dose within the previous 3 months. This temporal association was stronger in spike positive/nucleocapsid negative (i.e., vaccinated/not infected) individuals than spike positive/nucleocapsid positive individuals (i.e., vaccinated/infected), however these differences were not statistically significant (p = 0.2). Spike antibody quantities were not statistically significantly associated with age, sex, the vaccine product received, or the number of doses received (highly colinear with time since last vaccine dose).

Among antibody positive individuals in the follow-up survey, antibody neutralization was high for all variants studied. Almost all (99%) participants had neutralizing antibodies to wild type (n = 593/599), alpha (n = 597/599), beta (n = 591/599), gamma (n = 594/599), and delta (n = 594/599) variants. Slightly fewer had neutralizing antibodies to omicron (n = 584/599; 97%). The levels of neutralizing antibodies were also very high, with at least 94% of samples yielding greater than 95% neutralization for each variant. Neutralizing levels were higher among individuals with evidence of natural infection (i.e. nucleocapsid positive/spike positive) compared to those who did not (i.e., nucleocapsid negative/spike positive; Table 4). These differences were most extreme for the omicron variant, such that the level of omicron neutralization was on average 14.8% higher among individuals with evidence of natural infection compared to those without. Neutralizing levels to omicron were also 3.8% higher among individuals with natural infection detected in this study compared to those with natural infection detected at the first survey (i.e., using 2020 sera).

Household children were invited to participate, and 279 such children aged < 18 years were identified by their parents/guardians for possible enrollment into the study. 62 (22%) of these children completed the study, while 103 (37%) were unreachable, 73 (26%) declined participation, and 41 (15%) completed the survey but not the blood draw. Most pediatric participants (n = 41, 66.1%) were from Northwest Virginia and demographic characteristics largely matched the adult participants (Table S5).

Approximately half of pediatric participants (n = 35, 56.5%) were SARS-CoV-2 IgG nucleocapsid seropositive, and the majority (n = 54, 85.5%) were also SARS-CoV-2 IgG spike seropositive (Table S6). All vaccinated children were spike seropositive (n = 28/28). The majority of unvaccinated children were also spike seropositive (n = 22/30) with most of these being nucleocapsid seropositive indicating natural infection (n = 20/22). As with adults, vaccinated spike seropositive children had higher antibody quantities than unvaccinated spike seropositive children (88.5% of vaccinated children had spike quantities > 2500 U/ml compared to 10% of unvaccinated children).

Pediatric participants frequently reported at least one COVID-19-like illness (n = 45, 72.6%). Slightly more than half (n = 23) were tested for COVID-19 because of at least one of these illnesses, and 12 (19.4%) reported a positive result from this test (Table S7). All but one child who reported a positive test result were nucleocapsid seropositive. Being unvaccinated (n = 31 compared to n = 28 vaccinated) was associated with a 33% increased risk of being nucleocapsid positive (95% CI: -14%, 118%) among the pediatric participants. Nucleocapsid positivity in the parent was highly predictive of nucleocapsid positivity in the child: of 31 children whose parents were nucleocapsid seropositive, 27 were also nucleocapsid positive (87.1%), while among 30 children whose household adult were seronegative, only 8 were positive (26.7%; OR: 18.6, 95% CI: 4.9, 69.9). All children with neutralizing antibody testing (n = 41) were antibody positive for all variants studied, with average neutralizing antibody levels > 98% for all variants. There was no difference in neutralizing antibody levels between children who had evidence of natural infection versus those who did not.