The Seroprevalence of SARS-CoV-2 in Europe: A Systematic Review

Background A year following the onset of the COVID-19 pandemic, new infections and deaths continue to increase in Europe. Serological studies, through providing evidence of past infection, can aid understanding of the population dynamics of SARS-CoV-2 infection. Objectives This systematic review of SARS-CoV-2 seroprevalence studies in Europe was undertaken to inform public health strategies including vaccination, that aim to accelerate population immunity. Methods We searched the databases Web of Science, MEDLINE, EMBASE, SCOPUS, Cochrane Database of Systematic Reviews and grey literature sources for studies reporting seroprevalence of SARS-CoV-2 antibodies in Europe published between 01/12/2019 - 30/09/20. We provide a narrative synthesis of included studies. Studies were categorized into subgroups including healthcare workers (HCWs), community, outbreaks, pregnancy and children/school. Due to heterogeneity in other subgroups, we only performed a random effects meta-analysis of the seroprevalence amongst HCWs stratified by their country. Results 109 studies were included spanning 17 European countries, that estimated the seroprevalence of SAR-CoV2 from samples obtained between November 2019 – August 2020. A total of 53/109 studies included HCWs with a reported seroprevalence among HCWs ranging from 0.7% to 45.3%, which did not differ significantly by country. In community studies significant heterogeneity was reported in the seroprevalence among different age groups and the majority of studies reported there was no significant difference by gender. Conclusion This review demonstrates a wide heterogeneity in reported seroprevalence of SARS-CoV-2 antibodies between populations. Continued evaluation of seroprevalence is required to understand the impact of public health measures and inform interventions including vaccination programmes.


Introduction
On 11 th March 2020 the World Health Organization (WHO) declared the spread of novel SARS-CoV-2 as a pandemic (1). SARS-CoV2 is thought to spread mainly by respiratory droplets, while some evidence also suggests spread via fomites and aerosols (2)(3)(4). SARS-CoV2 causes varying degrees of illness from mild symptoms including fatigue and myalgia to acute respiratory failure and death (5). As the pandemic unfolded evidence emerged that a large number of individuals are asymptomatic with SARS-CoV2 infection (6,7).
In order to control the spread of SARS-CoV-2 it is important to understand the extent to which different populations have already been exposed to the virus, especially as a large number of infections are asymptomatic. Many countries, organizational bodies and facilities have turned to mass testing to estimate the spread of infection and inform public health measures (8,9). One such testing strategy is by nasal and throat swabbing to detect viral RNA which has recently been piloted in England (10). A further method is mass testing of the population for antibodies against SARS-CoV-2 (11). Several tests for IgG, IgA and IgM antibodies against SARS-CoV2 have recently been developed. These broadly include enzyme linked immunosorbent assays, chemiluminescence immunoassays (CLIA) and point of care lateral flow assays (12).
Seroprevalence studies have been used in the past to help with outbreak responses. During a recent Ebola outbreak, seroprevalence studies were performed to gain further information on the immune response and immune protection (13). Seroprevalence studies have also been used for infections such rubella, mumps and measles to map resurgence and to gain further information on how public health strategies can target high risk populations (14). Moreover, seroprevalence studies provide valuable information on vaccination strategies to achieve herd immunity.
By estimating the seroprevalence in different populations we can use the results to understand transmission dynamics, herd immunity and the immune response over time.
These studies can help to guide the public health response to ultimately help prevent the spread of SARS-CoV2. Here we present the results of a systematic review on the seroprevalence of SARS-CoV2 in Europe.

Search strategy and selection criteria
This systematic review and meta-analysis adhere to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. The protocol was registered on the University of York database for Prospectively Registered Systematic Reviews (PROSPERO: 2020 CRD42020212149). We systematically searched electronic data sources from (01/12/2019) until (30/09/20) using search terms on seroprevalence and SARS-CoV-2.
We searched the following electronic databases: Web of Science, MEDLINE, EMBASE, SCOPUS and Cochrane Database of Systematic Reviews. We also searched database search engine EBSCO to search the following databases EBSCOhost e-book collection, biomedical reference collection, CINAHL plus and MEDLINE Complete. We conducted a secondary search by searching the reference lists of included studies for relevant articles. antibod* OR immun* 3 Seroprevalence 4 1 AND 2 AND 3 5 animals NOT humans 6 4 NOT 5 Studies were included if they were written in English and published between 1/12/19-30/09/20 and were cross-sectional or cohort studies. Vaccine evaluations and randomised controlled trials were excluded.

Data extraction
Titles and abstracts of retrieved studies were de-duplicated and screened independently by two reviewers to identify if they met the inclusion criteria. Screened references then underwent full text review by two reviewers independently. Any disagreement between them over the eligibility of studies was resolved through discussion with third reviewer.
A standardised data extraction form was used. Data were extracted on the characteristics of the study (country, date, setting, selection method, funder), antibody assay employed, specificity and sensitivity of the assay, sample characteristics (age, gender, ethnicity, comorbidities) and prevalence. Data extraction was carried out autonomously by the reviewers and consensus was sought between the team.

Assessment of the methodological quality of included studies
We assessed the risk of bias using an adapted version of the Hoy et al modified Risk of Bias Tool, as used by Nguyen et al (15). This is a tool designed to assess the risk of bias in population-based prevalence studies. It uses a scoring system to assess the external and internal validity of the study. Studies that score 0-3 points are classified as low risk, 4-6 high risk and 7-9 high risk. Two reviewers independently applied the criteria. Disagreements were resolved through discussion.

Data Analysis
We provided a narrative synthesis of the findings of all included studies, study population infection and low risk if they had no patient contact (e.g., laboratory staff and administrative staff). If studies included participants from a mixture of risk groups they were categorised as medium risk. Heterogeneity was measured using the I 2 statistic which describes the percentage of total variation due to inter-study heterogeneity. Tests of heterogeneity were undertaken within the sub-groups and for the overall meta-analysis. Sensitivity analysis was done by removing those studies with a moderate risk of bias score. This had no effect of the I 2 value, so these studies were included in the final meta-analysis.
After reviewing the relevant literature, we did not perform traditional asymmetry tests and funnel plots for assessing publication bias, as the meta-analysis we conducted was a summary of proportions and not a comparison of treatments/interventions (16,17).

Role of the Funder
This study received no external funding. Staff were supported through National Institute of Health Research (NIHR) awards and the NIHR had no role in the concept, design, analysis and interpretation of the data.

Results
The literature search yielded 1648 articles. After removing duplicates and excluding studies based on their abstract or through full text examination 109 studies were identified as eligible ( Figure 1).  ; this included government   funding, research grants and some studies received no external funding (table 1). Each study underwent a risk of bias assessment using the modified Hoy et al risk of bias tool (15). Twenty eight of the 109 studies were deemed to be at moderate risk of bias (26,29,36,40,46,48,49,58,60,64,65,79,86,88,91,97,98,101,105,107,110,111,114,116,117,121 ,123,125). This was often due to lack of information about the sampling frame, selection process and non-response bias. No studies scored high risk ( Table 2).

Moderate risk
Not enough information given about the sampling frame, selection process and nonresponse bias.
Mulchandani et al 2020 (122) Low risk Nsn et al 2020 (123) Moderate risk Not enough information given about the selection process and non-response bias.

Health Care
Workers 53 studies included seroprevalence data among HCWs. These studies included data from 13 countries in Europe and were conducted between February 2020 and August 2020 (Table 3).  (108). Figure 2, chart A shows the seroprevalence of HCWs by country overtime. The majority of studies report a seroprevalence < 10% between March -August 2020. A few studies predominately based in the UK report a seroprevalence 20-45% among HCWs during this time period (21,87,108,109,112,117,118,122).  There was no significant difference in seroprevalence amongst HCWs when stratified by country ( Figure 3). There is a large amount of heterogeneity between the studies (I 2 value = 99.3%, p = 0.00). There was no reduction in heterogeneity when moderate risk of bias studies were removed. Similarly, when the seroprevalence amongst HCWs was stratified by their risk of exposure to SARS-Cov-2 patients there was no significant difference (supplementary figure 1).

Figure 3:
Forest plot of the seroprevalence among HCWs stratified by country.

Community Studies
In total 33 studies were set in the community, spanning 13 countries. The studies collected data between February 2020 -August 2020 ( Table 4). Ten of these studies collected samples from blood donors; four studies used residual serum samples from clinics, laboratories and hospital facilities; one study used tissues samples and the remaining studies were randomised population-based studies.  (67). The majority of studies reported an overall seroprevalence of less than 10% during the months of February -August 2020 ( Figure 2C).

Age
Many of the community studies report the seroprevalence among different age groups.
There is significant heterogeneity between the results. In general, lower seroprevalences were reported at the extremes of age. Several studies report a higher seroprevalence among the over 50 age group (34,44,65,76,77,95). In contrast some studies report a higher seroprevalence in the less than 30 years age group, these include studies from Switzerland, The Netherlands, Denmark, France, Luxembourg and the UK (38,81,82,104,120,126).

Gender
In the majority of community studies there was no significance difference identified by March -May 2020 (107). A seroprevalence of 8.5% (6.9-10%) was reported in blood donors across England in May (121).

Children/ Schools
Six studies investigated the seroprevalence in school/university settings or among children only, across 5 different countries (41,55,60,105,110,125). Four of the studies examined the seroprevalence in schools (41,55,105,125). The lowest seroprevalence was reported in Germany; 0.6% among students in grade 8-11 and their teachers in 13 secondary schools in eastern Saxony between the months of May -June 2020 (55). The highest seroprevalence of 11.7% was reported in students and teachers across schools in England between June-July 2020 (10.5-13.3%) (125).
Fontanet et al investigated the seroprevalence among pupils and teachers in primary schools exposed to a SAR-CoV2 outbreak in Paris and reported an overall seroprevalence of 10.4% (41). They noted that 41.4% of infected children had asymptomatic infection compared to 9.9% of seropositive adults (41).
One study examined at the seroprevalence among university students and staff in Greece (60). They reported an overall seroprevalence of 1%, with no significant difference by age, gender, school or position (60).

Outbreaks
Seven studies across four countries that investigated the seroprevalence following an In contrast four studies investigated the seroprevalences in the residents or blood donors in communities where there has been an outbreak (19,37,50,54). They report much lower rates of seroprevalence compared to nursing home outbreaks. Grzelak et al investigated the seroprevalence of pauci-symptomatic individuals in Crepy-en-Valois France and blood donors in the surrounding region following an outbreak; they reported a seroprevalence of 3% in blood donors and 32% in the pauci-symptomatic individuals (37). Similarly studies in Germany following community outbreaks report low rates of seroprevalence among residents. Streeck et al reported a prevalence of SARS-COV-2 antibodies of 13.6% and Weis et al reported a seroprevalence of 8.4% (50,54). Figure 2, chart D shows the seroprevalence of these outbreak studies overtime.
Two of these studies are conducted in Italy between April -June 2020 reported a prevalence of SARS-CoV-2 antibodies of 10.1% and 14.3% in pregnant women in their first trimester screening or at delivery (74,88). The third study estimated a seroprevalence of 8% among pregnant women admitted to the delivery room in France in May 2020 (42). Mattern et al found that the seroprevalence among pregnant women was similar to that of the general public (42).

Assays
In total 45 different commercial assays and 22 in-house assays were used. The majority of studies used more than one assay. Of the commercial assays 11 were enzyme-linked immunosorbent assay (ELISA), six were chemiluminescent microparticle immunoassays, two were based on flow cytometry and 26 were point of care tests (POC). The most commonly used commercial assay was the SARS-CoV-2 (IgA/IgG) ELISA EUROIMMUN Medizinische Labordiagnostik, Lübeck, Germany (supplementary table 2).

Discussion
Our systematic review demonstrates a large variation in the seroprevalence of SARS-CoV-2 antibodies throughout Europe in the first half of 2020. Our systematic review found that in the majority of studies in Europe there was no difference in seroprevalence between female and male participants. Our findings are in keeping with a meta-analysis which showed there was no difference in the proportion of males and females with confirmed COVID-19 (131).
However, there is strong evidence that males are more likely to have more severe disease compared to females. A meta-analysis looking at 92 studies world-wide concluded that male patients have almost three times the odds of requiring intensive treatment unit (ITU) admission and have higher odds of death compared to females (131). Similarly, Castro-Dopico et al reported that severe disease was associated with virus -specific IgA and that IgA responses were lower in females compared to males (98).
Throughout the current pandemic there has been debate on the role of children in the transmission of SARS-CoV-2 and the need for school closure to slow the pandemic. In this review three studies were in schools not involved in an outbreak of SARS-CoV-2. A study in Germany reported the lowest seroprevalence among students and teachers in a school of 0.6% considered by the authors to be in keeping with local surveillance data of the surrounding community (55). Ulyte et al reported that seroprevalence is inversely related to age in their school study (105). They conclude this could be due to the lack of social distancing among young children and differences in immune response (105). In contrast Ladhani et al reported no significant difference between the seroprevalence in students compared to staff (125). However, all studies concluded that there was no major transmission in schools and that the majority of children were asymptomatic or had mild symptoms (55,105,125). This review highlights the need for more school-based studies investigating the seroprevalence among staff and students to fully understand transmission dynamics and immune responses throughout the pandemic.
In studies conducted during local outbreaks there was a noticeable difference between those conducted in care/nursing homes compared to community and school settings. Those that took place in care/ nursing homes reported a seroprevalence as high as 77.9%, whereas those in a community setting reported a seroprevalence ranging from 3% -42.4% and those in a school reported a seroprevalence of 10.4% (19,37,41,50,54,119) This large discrepancy could be attributable to the close proximity of care/nursing home residents, shared living spaces and the intimate care and handling of residents by staff.

Limitations
This meta-analysis had several limitations. Firstly, of the 109 studies included in this review, not all of them could be included in sub-analysis as complete data sets could not be retrieved from every study and data quality was heterogeneous. In addition, the majority of studies were performed either in the UK, Italy, Spain or Germany. There was a large gap in studies being performed in Eastern European countries. Those studies performed in the UK predominately took place in the South of England. Furthermore, many of the studies were pre-print articles that had not undergone peer-review.

Conclusion
This systematic review and metanalysis highlights substantial heterogeneity between countries, within countries, among professions, and among settings. This heterogeneity, in addition to indicating the general trajectory of the pandemic in different regions, will also be driven by a variety of other factors including governmental policies and restrictions, local guidelines and restrictions, availability of PPE, the time period when the study was conducted and serological test performance. Nevertheless, seroprevalence studies yield large amounts of useful, locally relevant information and should be regularly repeated as the pandemic evolves and local guidelines and restrictions change. As testing standardizes and new studies are reported they will also help identify different national experiences across Europe and provide a means to distil best pandemic control practices for the future.
Finally, as new variants of SARS-CoV-2 now emerge, and many countries prepare for future waves it is vital that regular seroprevalence studies are conducted to aid control in the context of vaccine implementation.

Author Contributions
NMV -wrote the protocol, develop search terms, conducted searches, screened studies, extract data, conducted data analysis and meta-analysis and wrote the final draft. All authors have approved and read the final manuscript.