key: cord-0734887-57qk34fx
authors: Stringhini, S.; Wisniak, A.; Piumatti, G.; Azman, A. S.; Lauer, S. A.; Baysson, H.; De Ridder, D.; Petrovic, D.; Schrempft, S.; Marcus, K.; Arm-Vernez, I.; Yerly, S.; Keiser, O.; Hurst, S.; Posfay-Barbe, K.; Trono, D.; Pittet, D.; Getaz, L.; Chappuis, F.; Eckerle, I.; Vuilleumier, N.; Meyer, B.; Flahault, A.; Kaiser, L.; Guessous, I.
title: Repeated seroprevalence of anti-SARS-CoV-2 IgG antibodies in a population-based sample from Geneva, Switzerland
date: 2020-05-06
journal: nan
DOI: 10.1101/2020.05.02.20088898
sha: 4ccb28268e3b1758976ea69ada6f605e5f3946e6
doc_id: 734887
cord_uid: 57qk34fx

Background: Assessing the burden of COVID-19 based on medically-attended case counts is suboptimal given its reliance on testing strategy, changing case definitions and the wide spectrum of disease presentation. Population-based serosurveys provide one avenue for estimating infection rates and monitoring the progression of the epidemic, overcoming many of these limitations. Methods: Taking advantage of a pool of adult participants from population-representative surveys conducted in Geneva, Switzerland, we implemented a study consisting of 8 weekly serosurveys among these participants and their household members older than 5 years. We tested each participant for anti-SARS-CoV-2-IgG antibodies using a commercially available enzyme-linked immunosorbent assay (Euroimmun AG, Lubeck, Germany). We estimated seroprevalence using a Bayesian regression model taking into account test performance and adjusting for the age and sex of Geneva's population. Results: In the first three weeks, we enrolled 1335 participants coming from 633 households, with 16% <20 years of age and 53.6% female, a distribution similar to that of Geneva. In the first week, we estimated a seroprevalence of 3.1% (95% CI 0.2-5.99, n=343). This increased to 6.1% (95% CI 2.6-9.33, n=416) in the second, and to 9.7% (95% CI 6.1-13.11, n=576) in the third week. We found that 5-19 year-olds (6.0%, 95% CI 2.3-10.2%) had similar seroprevalence to 20-49 year olds (8.5%, 95%CI 4.99-11.7), while significantly lower seroprevalence was observed among those 50 and older (3.7%, 95% CI 0.99-6.0, p=0.0008). Interpretation: Assuming that the presence of IgG antibodies is at least in the short-term associated with immunity, these results highlight that the epidemic is far from burning out simply due to herd immunity. Further, no differences in seroprevalence between children and middle age adults are observed. These results must be considered as Switzerland and the world look towards easing restrictions aimed at curbing transmission.

Although statistics on confirmed cases and deaths help in monitoring the dynamics of disease 69 propagation, estimates remain largely unreliable when trying to understand the proportion of the 70 population infected with SARS-CoV-2 for public health purposes 1 

European countries, including Switzerland, did not have sufficient nasopharyngeal swabs available for 72 RT-PCR screening of anyone suspected or at risk of infection with SARS-CoV-2. Generally, 73 asymptomatic individuals, including those in at-risk categories, are not screened. As a result, the 74 number of confirmed cases of SARS-CoV-2 infections is largely underestimated 2 . In this context, 75 seroprevalence surveys are of utmost importance for understanding the proportion of the population 76 that has already developed antibodies against the virus and is potentially protected towards a new 77 infection 3 . As recommended by the World Health Organization (WHO), monitoring changes of 78 seroprevalence over time is also crucial at the beginning of an epidemic to anticipate and plan an 79 adequate public health response 4 . 80

The canton of Geneva, Switzerland, reported its first confirmed COVID-19 case on February 26, with 82 5071 cases (10.15 per 1000 inhabitants) and 243 deaths as of April 30 5,6 . As in most countries, 83 changing testing strategies over the course of the epidemics made it impossible to estimate the extent 84 of the population that had been infected, while this information is crucial to plan evidence-based 85 strategies to lift confinement measures. To assess the seroprevalence of anti-SARS-CoV-2 antibodies 86 in this area, we initiated a survey in a representative sample of the population 7 . To do so, we 87 contacted subjects who already participated in the Bus Santé study (an annual health examination 88 survey of a representative sample of the population of the canton) 8 and invited them along with their 89 household members to participate to the SEROCoV-POP seroprevalence survey 7 . Here we present 90 results based on 1335 participants who took part to the study during the first 3 of 12 study weeks as 91 they are likely to inform public health policy makers in Europe and beyond. 92 93

Each week, ~1300 randomly selected previous participants of the Bus Santé study with an email 96 address on file (n~10'072, Figure S1 ) were invited to participate in the SEROCoV-POP study by 97 email, which provided a link to an online appointment booking system for a visit at one of two sites 98 within the proceeding seven days. Eligible participants with a non-valid email address were contacted 99 by phone to update their contact information. Since the third study week, each potential participant 100 that hadn't replied to the initial email invitation within 72h was reminded of the invitation by phone, in 101 order to increase participation rate. Potential participants then received a confirmation by email, which 102 included a link to a questionnaire and a consent form to complete at home and bring on the day of the 103 study visit. During the visit, study staff discussed the study once again with participants to ensure 104 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted May 6, 2020. . https://doi.org/10.1101/2020.05.02.20088898 doi: medRxiv preprint informed consent. Participants were invited to bring all members of their household, aged 5 years and 105 older, to join the SEROCoV-POP study. An electronic validation system allowed participants to 106 declare that they weren't in quarantine or isolation and didn't present with symptoms compatible with 107 COVID-19 at the moment of making the appointment, in which case they were encouraged to book at 108 a further date. Participants considered vulnerable according to the Swiss Federal Office of Public 109

Health criteria 9 (over 65 years old, diabetic, suffering from cardio-vascular or respiratory disease, 110 immunocompromised, suffering from active cancer or with a BMI > 35 kg/m 2 ) were asked to contact 111 us directly by phone or email to book an appointment on time slots reserved explicitly for this 112 population, in order to reduce risk of exposure to the virus. Inclusion criteria included people aged 5 113 years and older, whose primary residence was in the canton of Geneva. Over the first 3 weeks, 31% of 114 invited eligible participants took part in the study, 16% refused to participate, and 53% have a pending 115 status (waiting for booking appointment, being recalled, etc). We collected 6 mL of peripheral venous 116 blood from each adult participant and 3 mL from each child less than 14 years old. The SEROCoV-117 POP study was approved by the Cantonal Research Ethics Commission of Geneva, Switzerland 118 (CER16-363). 119 120

We assessed anti-SARS-CoV-2-IgG antibodies using a commercially available enzyme-linked 122 immunosorbent assay (Euroimmun AG, Lübeck, Germany # EI 2606-9601 G) targeting the S1-domain 123 of the spike protein of SARS-CoV-2; sera diluted 1:101 were processed on a EuroLabWorkstation 124

ELISA. An in-house validation study, using a set of sera from 176 pre-pandemic negative controls and 125 181 RT-PCR confirmed COVID-19 cases was conducted to estimate test performance and optimal 126 thresholds for definitive results 10 . Based on this validation study, which set thresholds to maximize 127 specificity and inter-assay variability in test classification, we considered those with OD/CI >= 1.5 to 128 be positive. This threshold resulted in a sensitivity of 86.2% and a specificity of 100%. For these 129 analyses we treated all indeterminate results (>0.5 OD/CI <1.5) as negatives ( Figure S1 ). 130 131

To estimate seroprevalence, we used a simple Bayesian regression model taking into account test 133 performance, in addition to the age and sex of the population. In this framework, our goal is to 134 estimate the true underlying seroprevalence, ‫‬ , within the population: 135 136

CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 6, 2020. We used naive priors on all parameters to allow for an exploration of the parameter space. We 150 implemented this model in the Stan probabilistic programming language 11 and used the rstan package 151 to run the model, analyse outputs. We ran 20,000 iterations with 4 chains and assessed convergence 152 visually and using the R-hat statistic. While we did not account for household clustering specifically in 153 this model, we developed a separate hierarchical model with a random intercept for household, 154

including the modeling of test performance but with no covariates. We used this model to estimate the 155 overall seroprevalence across the three weeks and compared to those from the model described above 156 to understand the potential impact of not accounting for clustering within households. 157

To generate seroprevalence estimates adjusted to the age and sex distribution of the population, we 159 sampled from the posterior for all age-sex strata proportional to the demographics in canton of 160

Geneva. All estimates presented represent the mean of the posterior samples with the 2.5 th and 97.5 th 161 percentiles of this distribution as the 95% confidence intervals (95%CI). 162 163

In the first three weeks of the study starting on April 6, we enrolled a total of 1335 individuals in the 165 SEROCoV-POP survey ( Figure S2 ). Included participants came from 633 different households, with 166 53.6% being female, and 16% being under 20 years of age ( In the first week, we estimated an overall seroprevalence of 3.1% (95% CI 0.2-5.9). This figure  172 increased to 6.1% (95%CI 2.6-9.33, n=416) in the second week, and to 9.7% (95% CI 6.1-13.1, 173 n=576) in the third week, mirroring the increase in confirmed cases in the weeks before the survey 174 (Figure 1) . While there were no meaningful differences in seroprevalence between men and women 175 (Table 1), we found that 5-19 year olds (6.0%, 95% CI 2.3-10.2%) had a similar seroprevalence to 20-176 49 year olds (8.5%, 95% CI 4.9-11.7), whereas seroprevalence among those 50 and older (3.7%, 95% 177 CI 0.9-6.0, p<0.01) was significantly lower. 178 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 6, 2020. . https://doi.org/10.1101/2020.05.02.20088898 doi: medRxiv preprint

As a sensitivity analysis, we estimated seroprevalence among the subset of participants who were 180 originally enrolled in our previous representative surveys ( Table 2) . Within this subset, we observed 181 similar seroprevalence as in the overall study population, with higher estimates in younger individuals 182 (20-49 years vs. 50+ years), and with increasing seroprevalence estimates from the first to the third 183

week of the study. 184 185 Discussion 186

The preliminary results of this study provide an important benchmark to assess the state of the 187 epidemic. With an estimated 48'500 people having developed antibodies (9.7% of 500'000 188 inhabitants) while 4741 cases were confirmed on April 24 th 5 , we observe that there are roughly 10 189 infections for every COVID-19 confirmed case in Geneva, reflecting the variability in disease severity, 190 testing practices and care-seeking behaviors. Further, we show that three weeks after the peak of 191 confirmed cases, only 1 in 10 people has developed antibodies against SARS-CoV-2, even in one of 192 the more heavily affected areas in Europe 13 . Thus, assuming that the presence of IgG antibodies 193 measured in this study is at least in the short-term associated with immunity, these results highlight 194 that the epidemic is far from burning out simply due to herd immunity. CoV-2 IgG antibodies where participants were selected from a representative sample of the general 212 population. It is also unique in its repeated nature allowing the study of immunity dynamics in both 213 adult and children populations. Our population-based design as well as the fact that we informed 214 participants that individual results were not going to be disclosed until the end of the study, mitigate 215 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 6, 2020. Our study also has some important limitations to acknowledge. First, the primary analyses include 220 randomly selected participants as well as of members of their households, making this sample not 221 entirely randomly selected. We attempt to adjust for some aspects of this through poststratification 222 within our statistical model. Further, in sensitivity analyses we estimated the seroprevalence for just 223 the original and found that it is similar to that of the full sample. Second, we were able to account for 224 clustering within households only for overall prevalence pooling the three weeks, as there was not 225 enough data to account for age, sex, and household clustering all together within this framework. The 226 use of a random effects model to account for household clustering did not change pooled results. 227

Finally, the recruitment of participants by email might exclude non tech-proficient individuals or 228 people without access to technology; another survey specifically targeting vulnerable populations 229 (socially and clinically) is ongoing. 230 231 Over the next weeks, we will continue monitoring weekly seroprevalence in the general population 232 and will be able to provide more refined analysis on symptomatology and other socio-demographic 233 data in relation to immunological status. Yet, a preliminary presentation of these results is deemed to 234 be necessary to timely inform global policy makers on how to adapt planning of the next phases. 235

Public sharing of our protocol can also help the global academic community to implement serosurveys 236 in their areas. For at least one year, we will follow up participants via an online digital platform and 237 repeated serological/RT-PCR testing in order to assess COVID19 incidence in the population. These 238 longitudinal results will also inform us on the dynamics of immunity which remain debated at this 239 stage. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 6, 2020. . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 6, 2020. . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 6, 2020. 

. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted May 6, 2020. Week 3 282 21 (7.4%) 9.0, 95% CI (4.5-13.7) 0.39 302 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted May 6, 2020. . https://doi.org/10.1101/2020.05.02.20088898 doi: medRxiv preprint Figure S2 . Density of ELISA ratios for all participants colored by result interpretation. All 303 indeterminates were considered negative for these analyses. 304 305 306 307 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted May 6, 2020. . https://doi.org/10.1101/2020.05.02.20088898 doi: medRxiv preprint

Why is it difficult to accurately predict the COVID-19 310 epidemic? Infectious Disease Modelling

Estimates of the severity of coronavirus disease 2019: a 312 model-based analysis. The Lancet Infectious Diseases

Defining the epidemiology of Covid-19-studies needed

World Health Organization. Population-based age-stratified seroepidemiological investigation 316 protocol for COVID-19 virus infection

Rapport sur la situation épidémiologique en Suisse et dans la 318 Principauté de Liechtenstein

COVID-19: Situation épidémiologique à Genève

Primary Care Division-Geneva University Hospitals H. Repeated seroprevalence survey of anti-327

IgG antibodies in the population of the canton of Geneva

High prevalence of forgoing healthcare for 331 economic reasons in Switzerland: a population-based study in a region with universal health 332 insurance coverage

Validation of a commercially available SARS-CoV-2 338 serological Immunoassay

A probabilistic programming language

Statistiques cantonales -République et canton de Genève-T 342 01.01.8.01: Population résidante selon l'origine, le sexe, le groupe d'âges et l'état matrimonial

Coronavirus disease 2019 (COVID-19) Situation Report -99

A systematic review of antibody mediated 349 immunity to coronaviruses: antibody kinetics, correlates of protection, and association of antibody 350 responses with severity of disease

Epidemiology and transmission of COVID-19 in 391 cases and 1286 of 352 their close contacts in Shenzhen, China: a retrospective cohort study. The Lancet Infectious 353 Diseases

COVID-19 and Kawasaki disease: novel virus and novel case

Can an effective SARS-CoV-2 vaccine be developed for the older 357 population?

Posterior predictive assessment of model fitness via realized 359 discrepancies