Introduction

The COVID-19 pandemic caused by SARS-CoV-2 is a public health crisis of global proportions (Worldometer, 2020; Center for Systems Science and Engineering (CSSE), 2020; European Centre for Disease Prevention and Control (ECDC), 2020). In most cases, the infection causes mild to moderate symptoms but can result in significant morbidity and mortality in those with pre-existing conditions and in the elderly; others may become infected and have no symptoms (Mizumoto et al., 2020; Wu & McGoogan, 2020). The number of confirmed cases identified by health care systems is one indicator of the progression of the pandemic. However, the true burden of infection can be more precisely estimated by SARS-CoV-2 antibody seroprevalence in the general population. Seroprevalence studies have the potential of evaluating the proportion of infections that are missed by the health care system, including those that are asymptomatic. Several studies have been published, with varying but often low seroprevalence rates, usually below ten percent (Chen et al., 2020; Lai et al., 2020). In addition, these studies often show rates of infection that are higher than those diagnosed by PCR, confirming that a substantial proportion of infected individuals remain undiagnosed. Few studies in a general population setting have correlated the presence or absence of anti-SARS-CoV-2 antibodies with the occurrence of COVID-19 symptoms (Qu et al., 2020; Zhao et al., 2020).

Québec was the province most severely hit by COVID-19 during the first wave of the pandemic, with more than 50% of all cases in the country, while representing only 23% of its population (i.e., 57,007 cases in the province of Québec, and a total of 108,829 cases in Canada as of July 15) (Institut national de santé publique du Québec, n.d.-a). The greater Montréal area, where more than half of the population of the province lives, was especially affected with a reported cumulative incidence of confirmed COVID-19 cases of 1350 per 100,000 between March and mid-July 2020.

The main goal of this study was to determine the seroprevalence of anti-SARS-CoV-2 antibodies in the Québec general population after the first wave of the pandemic, in blood samples from donors collected by Héma-Québec, the provincial blood collection agency. The dynamics of pre- and asymptomatic infections are still under scrutiny (Nogrady, 2020). Therefore, a second goal was to estimate the frequency of asymptomatic SARS-CoV-2 infections in this population.

Methods

Study participants for seroprevalence study

Residual blood samples were obtained from 7691 consenting blood donors who gave a regular blood donation at one of Héma-Québec’s permanent donation centres or mobile blood drives. Donors are residents of the province of Québec, where Héma-Québec is the sole blood component supplier for the province. Ninety-five percent of donations made at selected clinics were tested and included in the study. Sample size was stratified by region, and based on an expected regional prevalence of 3%, targeting a 95% accuracy of about ±2%. Prospective donors included in the study had to be free of COVID-19 symptoms in the preceding 14 days and otherwise eligible to donate according to standard criteria. Samples were collected between May 25 and July 9, 2020 from 12 of the 18 health regions of the province of Québec. A higher sampling rate was applied in regions where COVID-19 was more prevalent. Donors first learned about the study when they presented to donate.

At the time of this study, guidelines from Québec public health authorities recommended COVID-19 testing for: (1) people with symptoms; (2) people who have had close contact with someone who has COVID-19; and (3) people who have been asked by public health authorities to get tested.

Anti-SARS-CoV-2 Spike RBD ELISA

An in-house ELISA targeting the RBD of the SARS-CoV-2 Spike protein was adapted from an assay designed to establish seropositivity of COVID-19 convalescent plasma (Perreault et al., 2020). SARS-CoV-2 Spike RBD was chosen over other SARS-CoV-2 antigens because of its low homology with common cold human coronaviruses, thus limiting cross-reactivity (Heffron et al., 2020; Prévost et al., 2020; Shrock et al., 2020). Assay design allows for the capture of the three main classes of immunoglobulins (IgG, IgA, and IgM). Plasma samples were diluted 1:100, as this dilution enabled easily distinguishing seropositive from seronegative samples with high sensitivity (98.9%) and specificity (98.5%) (Perreault et al., 2020). As standard procedure in a blood bank context, initially reactive samples were retested twice and were considered positive if at least two out of three assay results were above the seropositivity threshold. Positive and negative controls were included in each ELISA microtiter plate to validate assay results.

Occurrence of symptoms in cases and controls

We conducted a nested case-control study comparing all 173 antibody-positive individuals and 176 randomly selected antibody-negative control donors. Case-control study participants were notified in writing that Héma-Québec would contact them by phone in the weeks following their blood donation. The survey was non-blinded as donors and interviewers were aware of the seroprevalence results. The five interviewers involved in the study followed a standard script. Donors were queried on basic socio-demographic characteristics (sex, age, healthcare worker status, and area of residence), COVID-19-related symptoms, contacts with individuals with confirmed or suspected COVID-19, COVID-19 testing, and results (see questionnaire in Supplementary Material), for the period between March 1, 2020 and the date of donation. Mean number of days between testing and survey was 54.5 (SD 12.0) and 57.6 (SD 14.0) for cases and controls, respectively, with an overall delay of 56.1 days (SD 13.1).

Statistical analyses

Crude seroprevalence rates and their respective 95% Clopper-Pearson confidence intervals (CI) were calculated from the proportion of study participants who had developed antibodies against SARS-CoV-2. Seroprevalence rates were then adjusted to reflect population demographics. Age and sex distribution of seropositivity within each Québec health region was determined. Population data were derived from the 2011 Québec census. The adjustment was calculated according to the following formula:

$$ \mathrm{Individual}\ \mathrm{adjustment}\ \mathrm{factor}={Ppop}_{region, age, sex}/{Psample}_{region, age, sex} $$

where Ppopregion, age, sex and Psampleregion, age, sex correspond, respectively, to the proportion of the population of a given health region and the proportion of the sample for that region, stratified by age and sex. Seroprevalence results were compared with the cumulative incidence of COVID-19, as reported by Québec public health authorities as of July 20, 2020 (Institut national de santé publique du Québec, n.d.-b).

For the nested case-control study, associations between seropositivity and reported symptoms were expressed as odds ratios and their 95% CI derived from logistic regression analyses, adjusting for age (continuous variable) and sex. We used parsimonious stepwise multivariable logistic regression to estimate the strongest association between a given combination of symptoms and seropositivity. We initially included variables that were significantly (p ≤ 0.05) associated with seropositivity in the univariate analysis, then each variable was iteratively excluded from the multivariable model until only those significantly associated with seropositivity remained (p ≤ 0.05), further adjusting for age and sex. The same strategy was used to estimate risk factors for SARS-CoV-2 infection.

Results

Seroprevalence study

A total of 7691 blood donors consented to participate in the study and to have a blood sample tested for the presence of anti-SARS-CoV-2 antibodies; 52.8% of these were males and 39.8% were from the Montréal-Laval area. Donors from the Montréal-Laval and Montréal surrounding regions represent 64.8% of all participants.

Of 7691 samples, 173 were found to be seropositive for anti-SARS-CoV-2 Spike RBD antibodies, for a crude seroprevalence rate of 2.25% (95% CI 1.93–2.61). After adjusting for population size, age, and sex, seroprevalence was 2.23% (95% CI 1.90–2.56) (Table 1). Seroprevalence rates by health region correlated well with cumulative incidence rates of confirmed COVID-19 cases, not restricted for the age range of donors, as reported by health authorities (r = 0.80; p = 0.0017). The cumulative incidence rate in Montréal and Laval, at 1350 cases per 100,000, was about six times higher than in health regions outside of the greater Montréal area (220 cases per 100,000). However, the adjusted seroprevalence rate in Montréal and Laval at 3.05% (95% CI 2.44–3.66) was only about 2.5-fold higher than in regions outside the greater Montréal area (1.29%; 95% CI 0.86–1.72).

Table 1 Seroprevalence data (raw and adjusted for population, sex, and age) in Montréal-Laval, surrounding Montréal and Laval urban areas, and other regions
Full size table

Women had a higher adjusted seroprevalence rate (2.51%; 95% CI 2.00–3.02) than men (1.97%; 95% CI 1.54–2.41) but the difference was not statistically significant. The adjusted seroprevalence appeared to decrease with age, but the trend, using the Cochran-Armitage test, was non-significant (18–24 years: 2.57% (95% CI 1.70–3.44); 25–39 years: 2.36% (95% CI 1.69–3.02); 40–59 years: 2.29% (95% CI 1.74–2.83); 60–69 years: 1.68% (95% CI 1.03–2.33)).

Nested case-control study of association between symptoms and seroprevalence

Of the 173 antibody-positive donors, 161 (93.0%) agreed to participate in the telephone survey; for controls, 162/176 (92.0%) answered the survey. Among antibody-positive donors, 52.2% (n = 84; 95% CI 44.2–60.1) reported at least one of 11 COVID-related symptoms between March 1 and 14 days before their donation date; this proportion was 19.1% (n = 31; 95% CI 13.4–26.1) among seronegative controls (Table 2). Except for rhinorrhea, all symptoms were significantly more frequent among antibody-positive donors than among controls. Symptoms most typically associated with COVID-19 (fever ≥38°C or sensation of fever/chills, generalized pain, shortness of breath/difficulty breathing, and loss or reduction in smell or taste sensation) were more strongly associated with seropositivity, and were experienced by less than 2% of antibody-negative controls. Using stepwise multivariable logistic regression, fever or sensation of fever/chills (odds ratio 10.4, 95% CI 3.0–35.4) and loss or reduction in smell or taste sensation (odds ratio 14.5, 95% CI 1.8–114.0) were independently associated with seropositivity. The presence of both symptoms was present in 16/84 (19%) of symptomatic antibody-positive donors, and in none of the symptomatic antibody-negative donors. The presence of at least one of these symptoms was reported by 57% of symptomatic antibody-positive donors, but by only 13% (4/31) of symptomatic antibody-negative donors.

Table 2 Association between symptoms and anti-SARS-CoV-2 seroprevalence test results, based on a univariate logistic regression model
Full size table

Table 3 presents the results of additional questions that targeted symptomatic individuals, as well as those who consulted a physician for suspected or actual COVID-19 symptoms. Among antibody-positive donors, 16 (14 symptomatic and two symptom-free) consulted a physician; 70 symptomatic donors did not consult. Only three of 31 symptomatic antibody-negative control donors consulted a physician. Furthermore, 22 donors (19 antibody-positive and 3 antibody-negative) had a COVID-19 PCR test performed; 14/19 antibody-positive donors tested positive by PCR for COVID-19, but none of the three antibody-negative donors were PCR-positive. Assuming that symptoms reported by all 84 antibody-positive donors were caused by COVID-19, we infer that 83% [(84–14)/84] were not diagnosed as such at the time of infection because, as mentioned above, they did not consult a physician. If one also considers that 19.1% of antibody-negative controls had symptoms as compared with 52.2% of antibody-positive donors, a more conservative estimate could be that about one third (52.2% − 19.1% = 33.1%) of antibody-positive individuals had symptoms related to COVID-19, which means that around 53 antibody-positive donors experienced COVID-19 symptoms, including the 14 individuals who consulted a physician, and that 74% of these clinical COVID-19 cases [(53 − 14)/53] did not contact public health authorities or consult a physician for COVID-19 diagnosis.

Table 3 Association between healthcare-seeking behaviour and anti-SARS-CoV-2 seroprevalence test results, based on a univariate logistic regression model
Full size table

The vast majority of antibody-positive donors were unaware of how they were infected: 78.9% did not recall a contact with a COVID-19-positive individual, and 73.9% were unaware of a contact with an individual having cough or fever (Table 4).

Table 4 Association between demographic characteristics and anti-SARS-CoV-2 seroprevalence test results, based on a univariate logistic regression model
Full size table

The impact of self-exclusion on seroprevalence estimates was explored. Among our 7691 participants, 14 (0.18%) seropositive donors reported a diagnosis of confirmed COVID-19, assuming that none of the seronegative donors had been infected. This is lower than the 0.55% prevalence rate (31,077 laboratory-confirmed COVID-19 cases/5.6 million) reported among the 20–69-year-olds by May 25, 2020, when this study started. In the hypothetical scenario where 0.55% of donors had been diagnosed with COVID-19 and were antibody-positive, this would have added 28 positive donors (7691 × (0.55% − 0.18%)) to the 173 already detected, and the seroprevalence would increase from 2.23% to 2.61%.

In univariate analysis, age, history of contact with a COVID-19 case, contact with someone having COVID-19 symptoms, working in the healthcare sector, and living in the Montréal-Laval area were statistically significant risk factors for seropositivity (Table 4). In multivariable parsimonious analysis (adjusted for age and sex), the only risk factors significantly associated with antibody positivity were contact with a known COVID-19 case (odds ratio 5.15; 95% CI 2.04–13.0), contact with a symptomatic individual (odds ratio 2.71; 95% CI 1.39–5.28), and living in the Montréal-Laval area (odds ratio 2.28; 95% CI 1.26–4.11). Workplace (36.6%) and home (19.9%) are the two settings where contacts with a COVID-19 case were most frequently reported by antibody-positive individuals (Table 5).

Table 5 Exposure context of antibody-positive individuals and negative controls
Full size table

Discussion

In the province of Québec, as of May 25, 2020 when our study started, a total of 31,077 laboratory-confirmed COVID-19 cases had been reported among the 20–69-year-olds. This number represents 0.55% of the 5.6 million people belonging to this age group. Our seroprevalence data in this age group rather suggest that 2.23% of these adults have been infected by SARS-CoV-2, which is four times more than laboratory-confirmed COVID-19 cases.

Based on the number of seropositive donors who reported a positive diagnosis of COVID-19 in our nested case-control study, we calculated that this number was three times lower than expected according to the number of cases reported in the general population; we only had 14 such donors, whereas our calculations predicted that we should have had 28 more. The most likely explanation for this discrepancy is that some individuals diagnosed with COVID-19 self-excluded from blood donation, even beyond the mandated 14-day period. This self-exclusion bias would only apply to symptomatic, infected individuals, especially those who had severe enough symptoms to seek medical care. When we calculated the potential impact of this bias on the true seroprevalence rate, the corrected seroprevalence would only be marginally higher, 2.61% versus 2.23%.

A number of studies reporting anti-SARS-CoV-2 seroprevalence data have been published (Barzin et al., 2020; Biggs et al., 2020; Chen et al., 2020; Hallowell et al., 2020; Kshatri et al., 2020; Lai et al., 2020; Majdoubi et al., 2020; Makaronidis et al., 2020; Murhekar et al., 2020; Picon et al., 2020; Rudberg et al., 2020; Song et al., 2020; Squeri et al., 2020; Stadlbauer et al., 2020; Takita et al., 2020; Wells et al., 2020), including a few with respect to blood donors (Amorim Filho et al., 2020; Erikstrup et al., 2020; Fiore et al., 2020; Fischer et al., 2020; Gallian et al., 2020; Ng et al., 2020; Percivalle et al., 2020; Younas et al., 2020). Our results are consistent with these reports in that seroprevalence estimates are generally low, albeit often several-fold higher than cumulative COVID-19 incidence rates. In addition to providing an estimate of the burden of the first wave of the pandemic in Québec, our nested case-control study confirms that a substantial proportion of individuals infected with SARS-CoV-2 remain asymptomatic. We also found that fever or sensations of fever/chills, and loss or decline of smell or taste sensation, are most strongly associated with a serological positive test for SARS-CoV-2. Unsurprisingly, contact with a confirmed COVID-19 case or with a symptomatic individual was also predictive of the presence of anti-SARS-CoV-2 antibodies.

Nearly half of antibody-positive individuals did not report any symptom. Among symptomatic, antibody-positive individuals, it is likely that a proportion of those who experienced mild, common cold-like symptoms that are not typical of COVID-19 may have been infected by another virus, as suggested by the fact that 19.1% of antibody-negative individuals did report similar symptoms. These observations suggest that between half and two thirds of individuals who were infected by SARS-CoV-2 remained asymptomatic. A self-exclusion bias owing to symptomatic, diagnosed COVID-19 cases might have amplified this proportion to some extent. As a side note, thus far there is no evidence that SARS-CoV-2 is blood transmissible (Leblanc et al., 2020).

This study has some limitations. First, given that the level of SARS-CoV-2-specific antibodies, using RBD or nucleocapsid antigen, tends to wane in some individuals about 100–110 days post-symptom onset (Anand et al., 2020; Perreault et al., 2020; Seow et al., 2020), some cases of previous exposure to SARS-CoV-2 might have been missed; recall that Québec’s first documented COVID-19 case was identified in late February 2020. This would lead to an underestimation of the actual seroprevalence rate. However, samples were collected within 4 months of the onset of the first wave of the epidemic in Québec; thus, we would not expect to miss a significant number of cases by loss (waning) of antibodies (seroreversion). It is interesting to note that despite the temporal decline of antibody levels in blood, seropositive individuals may still have immunity, as virus-specific T- and/or memory B-cell responses seem to increase with time and to persist for 6 to 8 months after infection (Dan et al., 2021; Sherina et al., 2020). Second, since the anti-SARS-CoV-2 Spike RBD ELISA is not 100% sensitive, some antibody-positive individuals might have been missed (Burgess et al., 2020). Conversely, given the very low prevalence of antibodies and a specificity of 98.5% (with a sensitivity of 98.9%) for our in-house assay, some seropositive donors might not have been previously infected by SARS-CoV-2. We decided not to adjust our results for imperfect specificity because we believe our in-house assay likely had better precision than what was estimated, since it was only based on one false-positive event out of 67 negative samples collected before the SARS-CoV-2 outbreak. If seroprevalence is adjusted for specificity, based on the Rogan-Gladen equation (Rogan & Gladen, 1978), then an overall seroprevalence rate of 1.18% (95% CI 0.86–1.55) is obtained. Finally, participants were blood donors, who may not be perfectly representative of the general population because they are generally healthier. Additionally, a small proportion of participants included plasma donors, who are generally more educated (high school level or higher) and are likely of a higher socio-economic status (Charbonneau et al., 2015). Furthermore, some of the populations that were more greatly affected by SARS-CoV-2 are less likely to be blood donors, specifically plasma donors. Blood donors may also self-exclude following a COVID-19 infection beyond the mandated 14-day period. As this bias only applies to symptomatically infected individuals who sought medical care, it minimally underestimates the seroprevalence in the general population. However, blood donors can provide convenient samples and are reasonably representative of the healthy adult population. More importantly, donation leftover samples could be used for COVID-19 antibody testing, and large numbers of donations are collected on a daily basis, which allow routine monitoring of seroprevalence over time (O’Brien et al., 2020). Finally, one limitation is the reliance on self-reported health events. Recall bias of such events might have under- or overestimated their true prevalence. Despite these limitations, this study likely provides an accurate estimate of the order of magnitude of the rate of SARS-CoV-2 infection among 18–69-year-old adults in the province of Québec during the first pandemic wave.

Conclusion

Our results, based on a sample set collected during the first pandemic wave, indicate that despite Québec’s being the Canadian province most severely affected by COVID-19, SARS-CoV-2 seroprevalence in Québec, though much higher than the cumulative COVID-19 incidence, was still relatively low, and far from the level of herd immunity needed to limit viral spread.

Contributions to knowledge

What does this study add to existing knowledge?

  • Although the number of confirmed COVID-19 cases identified by healthcare systems is one indicator of the progression of the pandemic, the true burden of infection can be more precisely estimated by SARS-CoV-2 antibody seroprevalence in the general population.

  • Few studies in a general population setting have correlated the presence or absence of anti-SARS-CoV-2 antibodies with the occurrence of COVID-19 symptoms, and none of these studies have used a control group to estimate the frequency of asymptomatic SARS-CoV-2 infections in this population.

  • Our study advances understanding of the epidemiology of COVID-19.

What are the key implications for public health interventions, practice, or policy?

  • Based on blood donor samples collected after the first pandemic wave, this study shows that, in Québec, which had the highest rate of PCR-confirmed COVID-19 infections in Canada, less than 3% of 18–69-year-old adults have been infected, which is relatively low, and far from the level of herd immunity needed to limit viral spread.

  • The fact that three out of four infections escaped detection by public health illustrates the limitations of current control policies based on contact tracing of confirmed cases.