Background
The SARS-CoV-2 virus affected almost all nations within a few weeks. Given the nature of the virus, a large proportion of infected individuals present only mild symptoms or no symptoms at all. Therefore, population-based sero-prevalence studies are necessary to estimate the true prevalence of the infection in the population. Starting in March 2020, such sero-prevalence studies have been conducted in many countries, mostly during or after the first wave of the pandemic [
1]. Depending on the serological test used, the type of sample drawn, the timing of the study, and the region, general population sero-prevalence ranged from < 0.1% in Brazil to well over 20% in the USA [
2]. For the German context, we reported a sero-prevalence of 1.8% in Munich, sampled towards the end of the first wave in Germany [
3].
Following the introduction of public health measures (lock-down including school closures) in March 2020 in Germany, the first wave of the pandemic was perceived as relatively mild with around 6000 cases registered in Munich during this period (Munich population ~ 1.5 Mio). Between June and October, public health measures were reduced, although physical distancing of 1.5 m between two persons, avoidance of mass events, and obligatory use of face masks, e.g. in restaurants and shops, were still required. Subsequently, officially registered monthly case numbers in Munich rose from 389 in June to 7181 in October 2020. A partial national lock-down was implemented on November 2nd, 2020. After a further rise in officially registered case numbers and COVID-19 related deaths, national lock-down measures were increased from December 16th, 2020 on, including closure of schools, shops (other than grocery and drug stores), restaurants, and hotels.
Given that asymptomatic and mildly symptomatic cases escape surveillance systems, prospective population-based cohort studies offer the chance to better understand the course of disease in the general population. They are independent of testing strategies and help to identify the population at risk over time. In addition, they provide an indication of population groups less well protected by public health measures. We therefore followed up the participants of the Munich COVID-19 cohort (KoCo19) to explore the SARS-CoV-2 antibody prevalence in the Munich general population at two time points: at the time the acute outbreaks happened and seven months later. In addition, we aimed at the identification of risk factors (demographic, social-economic, health status or individual risk behaviours factors) for acquiring SARS-CoV-2 infection defined by serology. The baseline study took place from April to June 2020, the questionnaire follow-up in summer 2020 and the 1st antibody follow-up was realised from early November 2020 to January 2021. On December 1st 2020 the KoCo19 cohort joined the ORCHESTRA (Connecting European Cohorts to Increase Common and Effective Response to SARS-CoV-2 Pandemic) project.
Discussion
Our data indicate a low SARS-CoV-2 sero-prevalence for the Munich general population living in private households eight months after the start of the pandemic. The incidence between the end of the first wave and the middle of the second wave was about as high as the related number of infections acquired during the first wave. Almost all sero-positive participants at baseline remained sero-positive at follow-up, indicating a high validity of the antibody test. Additionally, this supports previous reports suggesting that the humoral SARS-CoV-2 immune response is stable at least over the first eight months after infection [
23,
24]. We also showed a predominance in SARS-CoV-2 sero-positivity among male compared to female participants, and a reduced antibody prevalence with increasing age group.
Based on our data, the sero-prevalence for the Munich population above the age of 13 years living in private households was 3.6% (95% CI 2.9–4.3%). Until the end of November 2020, a total of 30,180 SARS-CoV-2 cases were officially registered in Munich (
https://www.muenchen.de/rathaus/Stadtinfos/Coronavirus-Fallzahlen.html#Fallzahlen; Access date: 19-April-2021) which results in a population prevalence of 1.9%. This prevalence increased to 44,377 registered cases by the end of December 2020 (population prevalence 3.0%). The data are not directly comparable, as the official data also include children and persons living in institutions. While the prevalence of infection in children was at that time considered to be smaller than in adults, it was unknown for people living in institutions (e.g., homes for the elderly). Nevertheless, the comparison gives an indication that the percentage of officially registered infections improved considerably compared to the beginning of the pandemic. In a previous publication we estimated that solely one out of four infections was registered by the official infectious diseases surveillance system [
3]. A few population-based SARS-CoV-2 sero-studies have been conducted since the beginning of the pandemic (for review see [
2]), most of them reporting sero-prevalences during or after the first wave. Up to now, only the Spanish national study reported the results of their follow-up data [
25] with a sero-prevalence at follow-up (November 2020) of 5%, and thus comparable to our results.
In our study, one predictor of change in sero-prevalence from baseline to follow-up was male sex. While a higher risk of more severe COVID-19 among men was confirmed in several studies [
26], findings on sex-differences in sero-prevalence are still inconsistent [
2]. As younger age was also related to a larger increase in SARS-CoV-2 sero-prevalence at follow-up, one might assume that differences in behaviour may contribute to these findings. We could confirm differences in health-risk taking behaviour, frequency of leisure-time activities, and number of contacts outside the own household especially by age. In the stratified analyses of the incidence of infection by age and sex, we observed a tendency that behaviour is related to higher sero-incidence of infection; although the low incidence and the reduced number of respondents to the questionnaires limited the statistical power of these analyses and result interpretation has to be done cautiously. However, the hypothesis that specific behaviour, i.e., restriction of contacts, might reduce the risk of infection is also supported by our observation that patients with autoimmune disease were at reduced risk of SARS-CoV-2 sero-positivity. This finding is in line with studies among, e.g., patients with inflammatory bowel disease [
27]. Overall, our findings support the notion that behavioural factors contribute to the spread of the pandemic, and therefore actions to increase adherence to public-health measures (such as information campaigns) are crucial especially in a time when acceptance of measures in the general population is faltering.
Our results also confirm the importance of household clustering while no indications for neighbourhood clustering were seen. The former finding is also supported by the observation that participants from higher income households were at non-significantly higher odds of SARS-CoV-2 sero-positivity. As we took into account total household income (not adjusted for number of persons in the household), single households were more likely to be in the lower income category and thus, at lower likelihood of household transmission.
We also saw a non-significant trend for lower odds of SARS-CoV-2 antibodies in smokers compared to non-smokers (OR 0.2; 95% CI 0.02–1.1), confirming results of a meta-analysis [
28]. Here, differences were mainly explained by differences in testing behaviour between smokers and non-smokers, which can be excluded in our study. One of the population-based studies published so far also indicated a lower SARS-CoV-2 sero-prevalence in smokers compared to non-smokers [
29]. Whether this is a true effect of, e.g., nicotine [
30] or vitamin D [
31] or result of some form of bias needs to be evaluated in future studies. Of note is also the tendency for higher odds of SARS-CoV-2 in participants pursuing school, however, the wide confidence interval does not permit strong conclusions.
Among the strengths of our study are its population-based, prospective nature in a large number of participants. Such population-based studies help authorities to plan public health measures based on the prevalence of exposure in the population, its spatial distribution and to further identify risk groups [
32]. With increasing availability of vaccines, this study design with further follow-ups will help public health authorities to understand the extent and duration of vaccine-induced immunity [
33]. We previously showed a high sensitivity and specificity of the Elecsys
® Anti-SARS-CoV-2 assay (Roche) used in this sero-study [
5]. For the follow-up, we developed and carefully validated a semi-automated protocol using self-sampled DBS for SARS-CoV-2 serology [
7]. This approach facilitates field work to a very considerable extent and thus, makes studies with a higher frequency of follow-ups more feasible. Acceptance was high in our study population, and the percentage of participants lost to follow-up comparably low.
However, in the analyses we had to take selective participation into account by modelling the underlying non-response mechanism and calibrating the weights. This way, we could reduce attrition bias in our prevalence and incidence estimates. It is common in prospective cohort studies that baseline participants in younger age groups, with migration background and with lower socio-economic status are less likely to participate at follow-up [
4]. While typically participants with positive outcome are also more likely to participate in follow-up studies (
34), our baseline participants who were SARS-CoV-2 antibody positive were less likely to take part at follow-up. This gives some indication that unknown sero-status motivated at least part of our baseline participants to take part in the study. Once positive sero-status was known to them, they might have lost interest. Further supported is this hypothesis by the fact that less participants were willing to complete the follow-up questionnaires than to take part in the SARS-CoV-2 antibody follow-up. As a consequence, statistical power to analyse the association between behavioural factor and SARS-CoV-2 sero-positivity was limited. Finally, the age group 14 to 19 years is of specific interest given that data on sero-prevalence in this age group is still limited. However, the number of participants in this age group was low (n = 212) and given the still low sero-prevalence at the time of the study, only 9 of them turned positive limiting the power of further analyses in this age group.
Acknowledgements
We gratefully thank all study participants for their trust, time, data, and specimens. This study would also not have been possible without the staff of the Division of Infectious Diseases and Tropical Medicine at the University Hospital of LMU Munich, Helmholtz Centre Munich, and Bundeswehr Institute of Microbiology, as well as all medical students involved. We thank Judith Eckstein for outstanding support regarding public relations. We thank the teams from the press offices of LMU, University Hospital of LMU Munich, and of Helmholtz Centre Munich. We thank the KoCo19 advisory board members Stefan Endres, Stephanie Jacobs, Bernhard Liebl, Michael Mihatsch, Matthias Tschöp, Manfred Wildner, and Andreas Zapf. We thank Accenture for the development of the KoCo19 web-based survey application. We are grateful to the Statistical Office of the City of Munich, Germany, for providing statistical data on the Munich general population and to Landesamt für Gesundheit und Lebensmittelsicherheit for carrying out laboratory measurements. We also thank Guillaume Chauvet for advice on the sampling design and variance estimation and Dr. Joachim Heinrich for advice on inclusion of population density. We are grateful to the Munich police for their support in the fieldwork of the baseline study. The Munich Surgical Imaging GmbH, Cisco Systems, and the graphic/photo/IT infrastructure departments at the University Hospital of LMU Munich provided support during video production and online events. For fieldwork of the baseline study, BMW Group as part of their campaign “BMW hilft Helfenden” provided free cars. Mercedes-Benz Munich provided support with Mercedes-Benz Rent in the project infrastructure. MG acknowledges the support from the Joachim Herz Foundation through the Add-on Fellowship for Interdisciplinary Science.
Members of the KoCo19 Study group: Nikolaus Ackermann, Emad Alamoudi, Jared Anderson, Maxilmilian Baumann, Marc Becker, Franziska Bednarzki, Olimbek Bemirayev, Patrick Bitzer, Rebecca Böhnlein, Friedrich Caroli, Josephine Coleman, Lorenzo Contento, Alina Czwienzek, Flora Deák, Jana Diekmannshemke, Gerhard Dobler, Jürgen Durner, Ute Eberle, Judith Eckstein, Tabea Eser, Philine Falk, Manuela Feyereisen, Volker Fingerle, Otto Geisenberger, Christof Geldmacher, Leonard Gilberg, Kristina Gillig, Philipp Girl, Elias Golschan, Elena Maria Guglielmini, Pablo Gutierrez, Anslem Haderer, Marlene Hannes, Lena Hartinger, Alejandra Hernandez, Leah Hillari, Christian Hinske, Tim Hofberger, Sacha Horn, Kristina Huber, Christian Janke, Ursula Kappl, Antonia Keßler, Zohaib Khan, Johanna Kresin, Arne Kroidl, Magdalena Lang, Clemens Lang, Silvan Lange, Michael Laxy, Reiner Leidl, Leopold Liedl, Xhovana Lucaj, Fabian Luppa, Alexandra Sophie Nafziger, Petra Mang, Alisa Markgraf, Rebecca Mayrhofer, Hannah Müller, Katharina Müller, Ivana Paunovic, Michael Plank, Claire Pleimelding, Stephan Prückner, Elba Raimúndez, Jakob Reich, Viktoria Ruci, Nicole Schäfer, Benedikt Schluse, Lara Schneider, Mirjam Schunk, Lars Schwettmann, Andreas Sing, Alba Soler, Peter Sothmann, Kathrin Strobl, Jeni Tang, Fabian Theis, Sophie Thiesbrummel, Vincent Vollmayr, Emilia von Lovenberg, Jonathan von Lovenberg, Julia Waibel, Claudia Wallrauch, Julia Wolff, Tobias Würfel, Houda Yaqine, Sabine Zange, Eleftheria Zeggini, Anna Zielke, Thorbjörn Zimmer.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.