Prevalence of SARS-Cov-2 antibodies and living conditions: the French national random population-based EPICOV cohort

The COVID-19 pandemic has highlighted the paramount importance of public health surveys including assessments of seroprevalence for estimating the cumulative incidence of SARS-CoV-2 infection at population level. Evaluations limited to data for confirmed cases or deaths greatly underestimate disease propagation, due to the large proportion of mildly affected or asymptomatic individuals and the lack of RT-PCR screening tests at the start of the pandemic [1]. Nationwide-representative population antibody studies have been conducted in few countries to assess SARS-CoV-2 circulation, but rarely on random sample from general population [2].

France has been severely affected by COVID-19, but disease burden has been uneven across the country. Concerns about the contributions of social inequalities to spatial variations of COVID-19 exposure or severity have been raised [3], but most of the available data are based on deaths, hospitalization or reported cases [4].

EpiCOV is a large French national random population-based public health study including serological testing and longitudinal follow-up, aiming at both analysing the impact of living conditions on the dynamics of the epidemic, and the impact of the epidemic on health and living conditions[5].

Here, we aimed to provide a national estimate of SARS-Cov2 seroprevalence in France in May 2020, at the end of the first lockdown, and to identify the most exposed populations in terms of living and socio-economic conditions.

We selected 371,000 people aged 15 years or over at random, 134,391 of whom completed the questionnaire from May 2th to June 2th 2020. Within the random subsample of 17,123 people living in mainland France eligible for home testing, 14,995 agreed to receive the kit, 12,423 sent the DBS sample to the biobank and 12,114 samples could be analyzed (Fig. 1). The median date for blood sampling was May 21st 2020 (IQR 18th–28th May).

Epicov, designed in March 2020, just before the first national lockdown in France, aimed to estimate the proportion of the population aged 15 years or over exposed to SARS-Cov2, and to identify the subpopulations most exposed during the first epidemic wave. Overall seroprevalence was 4.5% [3.9–5.0], according to the cut-offs recommended by the manufacturer for the Euroimmun ELISA-S test that was applied on home self-sampled dried blood spots.

Only two other national serological studies based on random general population samples were performed at the same period, in Spain [7] and England [8]. They reported a prevalence of seropositivity for IgG antibodies of a similar magnitude to that in France, with a similar range of geographic disparities.

EpiCov was designed to study the effects of contextual living conditions. It showed that these conditions played a major role in the initial spread of the virus, accounting for a large proportion of exposure heterogeneity. Population density at the place of residence and household size were strongly associated with ELISA-S seropositivity, independently of individual socio-demographic and occupational characteristics. The availability of masks and tests was extremely limited until May 2020. People living in the most populous areas had little opportunity for physical distancing in current life activities outside home, particularly before lockdown.

Adjustment for both residential population density and household size accounted for much of the higher seroprevalence in immigrants outside Europe, which was twice that of the native population, whereas seroprevalence was similar in immigrants from European countries and the native population. These findings highlight the role of the spatial segregation of populations originating from low-and middle-income countries [9, 10]. Higher levels of exposure may account for part of the higher burden of COVID-19 mortality in these populations [4].

Poor socio-economic status was associated with severe COVID-19 infection [11, 12]. We found a more complex pattern for relationships with seroprevalence, which was highest in the two highest and lowest deciles of family income per capita, and lowest in the individuals with the lowest level of education. This probably reflects the combination of both high exposure to COVID-19 in qualified individuals working in care professions or having multiple social activities before lockdown, and high exposure of more deprived people living in overcrowded housing in densely populated areas, with less opportunity to telework during lockdown [13]. Seroprevalence in healthcare professionals was twice that in individuals with other occupations. Healthcare workers were highly exposed to the infection during the first wave, given the shortage of surgical masks and their proximity with patients [7, 8, 14].

Seroprevalence did not differ significantly between women and men, after adjustment for contextual and individual characteristics, including professional activity, consistent with most studies conducted in France [15, 16] and elsewhere [2, 7, 8]. Men seem to be more susceptible to develop severe forms of the infection than women [17], but there is no evidence of any difference in the probability of being infected [18].

Seroprevalence was highest at intermediate ages. Most population-based serological studies have reported a lower seroprevalence in the elderly [7, 8, 14]. Older people, at least those not living in care homes, are likely to have had fewer social interactions since being told to stay at home at the start of the outbreak. Lower rates in adolescents and young adults than in mid-age range adults have been reported in some studies [7, 19] including ours, but not in others [8, 20], and may be partly explained by school closures at the start of lockdown in France. Seropositivity was strongly associated with possible cases of infection in the same household, regardless of local population density, household size and composition. This finding is consistent with the higher risk of secondary infections among people living with others [7, 8, 21]. After adjustment for household size, seropositivity was not associated with living with a child or an adolescent under the age of 18 years. Similar results were reported in the English national seroprevalence study [8]. This finding is also consistent with smaller studies showing that the mean household secondary attack rate from adults is not significantly different from that from children, as reported in a meta-analysis [21]. By contrast, a study conducted during the same period in population cohorts in three regions of France with similar home self-sampling reported a higher seroprevalence for individuals living in households containing a young below 18 years [20]. It remains unclear whether children play a major role in intra-household transmission, which is a crucial issue, because the benefits of school closure for preventing disease spread have to be weighed up against potential psychological effects and increases in educational inequalities.

We found a strong inverse association between the presence of SARS-Cov-2 antibodies and smoking at the time of the EpiCov study, as in other studies [8, 20]. The possibility of biological mechanisms preventing infection in some smokers must be counterbalanced by evidence for higher rates of severe forms of COVID-19 in infected smokers [22].

The Epicov cohort is one of the largest national representative population-based seroprevalence surveys of individuals aged 15 years and over. It revealed a major role for contextual living conditions in the initial spread of COVID-19 in France, during a period of very limited access to prevention strategies before lockdown. It provides an exceptional tool for evaluating subsequent changes in exposure risk and, particularly, for identifying the most vulnerable populations, with changes in the epidemiological context and increases in access to testing, masks, and vaccines.