German federal-state-wide seroprevalence study of 1st SARS-CoV-2 pandemic wave shows importance of long-term antibody test performance.

Background: Reliable data on the adult SARS-CoV-2 infection fatality rate in Germany are still scarce. We performed a federal state-wide cross-sectional seroprevalence study named SaarCoPS, that is representative for the adult population including elderly individuals and nursing home residents in the Saarland.
Methods: Serum was collected from 2940 adults via stationary or mobile teams during the 1st pandemic wave steady state period. We selected an antibody test system with maximal specificity, also excluding seroreversion effects due to a high longitudinal test performance. For the calculations of infection and fatality rates, we accounted for the delays of seroconversion and death after infection.
Results: Using a highly specific total antibody test detecting anti-SARS-CoV-2 responses over more than 180 days, we estimate an adult infection rate of 1.02% (95% CI: [0.64; 1.44]), an underreporting rate of 2.68-fold (95% CI: [1.68; 3.79]) and infection fatality rates of 2.09% (95% CI: (1.48; 3.32]) or 0.36% (95% CI: [0.25; 0.59]) in all adults including elderly individuals, or adults younger than 70 years, respectively.
Conclusion: The study highlights the importance of study design and test performance for seroprevalence studies, particularly when seroprevalences are low. Our results provide a valuable baseline for evaluation of future pandemic dynamics and impact of public health measures on virus spread and human health in comparison to neighbouring countries such as Luxembourg or France.

Introduction

The pandemic due to SARS-CoV-2 (severe acute respiratory syndrome coronavirus 2) dominates human life, health systems and global economies for more than one year. On November 5, 2021, the WHO reported more than 247 million infections and 5 million deaths from Corona Virus Disease 2019 (COVID-19) worldwide (https://covid19.who.int/). To implement adequate measures, authorities need reliable data to evaluate the dynamics of viral spread, the effects of upcoming viral variants on human health, as well as the impact of vaccination. Severe cases requiring medical care and hospitalization, or fatal cases are recorded on a daily basis in developed countries. It can be assumed, however, that reported numbers of cases with mild or asymptomatic infections can vary grossly in populations depending on implemented test strategies and test frequencies. Accordingly, fatality rates calculated based on reported cases may vary widely depending on the number of unrecorded cases[1-5].

Seroepidemiology providing information on infections retrospectively can help to overcome this gap. To uncover previously unrecorded infections as accurately as possible in seroepidemiological studies, a long-term longitudinal antibody detection capability along with a high test performance (specificity and sensitivity) is of utmost importance. Particularly after oligo- or asymptomatic SARS-CoV-2 infection, seroconversion is detected in an assay-dependent manner[6]. Moreover, several studies have reported a substantial decay of humoral immune responses and neutralizing capacity within 8 months and a rapid decline within the first 3 months pointing to a continuous loss of humoral immunity to the virus[7-10].

Although studies have been performed to determine SARS-CoV-2 seroprevalences during the 1st pandemic wave in subpopulations[2,3], so far few reliable data are available on the adult SARS-CoV-2 infection rate in Germany. A local seroprevalence study in a German hot-spot setting of a provincial town conducted after a superspreading event during carnival 2020 estimated an adult infection fatality rate (IFR) as low as 0.36% (95% CI: [0.29; 0.45][2]. Another study conducted from April to June 2020 in the urban environment of Munich representative for individuals 14 years or older living in private households reported an infection fatality rate of 0.86% (95% CI: [0.67; 1.23]) if all deaths were counted, or 0.47% (95% CI: [0.36; 0.67]) if only 54% of deaths were counted assumed to occur in households[3]. Based on data from international meta-analyses[11-13] with the assumption of an equal distribution of infections among all age groups the infection fatality rate in Germany has been estimated to be 1.14% (95% CI: [0.76; 1.51]) by German authorities[14]. Thus, apart from the reliability of the applied antibody test, estimates for infection rates can vary largely depending on the target population, sampling bias and period during which the study was conducted.

Here we performed a federal-state-wide cross-sectional seroprevalence study SaarCoPS representative for the adult population including elderly people and nursing home residents in the 1st pandemic wave in Germany. We selected one out of three commercial SARS-CoV-2 antibody assays with superior specificity and longitudinal assay performance to determine the SARS-CoV-2 seroprevalence in the study population, to estimate the infection rate, the underestimation ratio, and the infection fatality rate in a German federal state. Our results demonstrate that seroprevalence data are highly dependent on the particular assay technique used to determine SARS-CoV-2-specific antibodies, which in turn affects infection rate estmates. We estimate an adult infection rate of 1.02% (95% CI: [0.64; 1.44]), an underreporting rate of 2.68-fold (95% CI: [1.68; 3.79]) and infection fatality rates of 2.09% (95% CI: (1.48; 3.32]) or 0.36% (95% CI: [0.25; 0.59]) in all adults including elderly individuals, or adults younger than 70 years, respectively. These results provide a valuable baseline for future evaluation of pandemic dynamics including the impact of upcoming new viral variants, vaccination and public health measures on virus spread and human health.

Methods

Study design and ethical approval

The SaarCoPS study was conducted in the federal state of Saarland with an overall adult population of 845,000 inhabitants in six districts following GCP criteria. Saarland is located in southwestern Germany bordering the state of Rhineland-Palatinate and the countries France and Luxembourg. Study design, data management, access to data files and data protection issues were defined in a study protocol. The design and conduct of the study were supported by epidemiology and biometry institutions, e.g., the Saarland Cancer Registry at the Ministry of Health, Social Affairs, Women and Family and the Institute of Medical Biometry, Epidemiology and Medical Informatics at Saarland University Medical Center. On the basis of IFR data reported in a German study available at that time[2] and COVID-19-related deaths in Saarland, a sample size of 2305 individuals was calculated, necessary to estimate an overall proportion of seropositive individuals of 4% with a relative precision of 0.2 (confidence level 95%). Overall, 10,000 inhabitants aged 18 years or older representative for the Saarland population with respect to age, sex and location were sampled from population registries and invited in two steps with an assumed participation rate of 30%. The Questor Pro software (Blubbsoft, Leipzig, Germany) was used for pseudonymization, generation of online questionnaires (covering questions, i.e. on age, sex, nursing or retirement home status), and data hosting. Serum samples and questionnaires were collected in 29 general practitioners‘ (GPs) offices distributed across the Saarland or mobile teams from July 22 to October 15, 2020. The study was approved by the Ethics Committee of the state of Saarland (Ärztekammer des Saarlandes, Saarbrücken, Germany) in accordance with the Declaration of Helsinki. Written informed consent was given by the study participants.

Specimen collection and anti-SARS-CoV-2 antibody testing

In addition to sera from the study participants, serum or plasma samples from randomly selected 30 convalescent non-hospitalized individuals with previous PCR-confirmed asymptomatic or mild SARS-CoV-2 infection (mild fatigue or respiratory symptoms that did not require hospitalization, score 1–2 according to the contemporary WHO ordinal scale classification) were collected at the Institute of Virology, Saarland University Medical Center between April and October 2020 primarily from contact tracing performed as a public health service. The time interval between PCR-positivity and first blood sampling was 23.17 ± 4.81 days in samples where information on both exact time points were available (n = 18). Samples from hospitalized patients with COVID-19 disease were collected during the same time period as the samples from convalescent non-hospitalized individuals and tested (Supplementary Fig. 1) but deliberately excluded from this study to avoid a bias toward potentially stronger antibody responses in patients with more severe disease.

Serum or plasma samples from individuals with a pre-pandemic PCR-proven infection by endemic coronaviruses (OC43, NL63, HKU1 and 229E) were selected from the clinical sample repository of the Institute of Virology for retrospective testing of cross-reactivity in anti-SARS-CoV-2 antibody assays. All blood samples were investigated in three anti-SARS-CoV-2 antibody test systems (Table 1) suitable for the analysis of serum or plasma samples. Parallel testing of serum and plasma from the same patient yielded comparable results confirming another study using a different SARS-CoV-2 antibody assay[15]. The following assays were used.

Table 1

Anti-SARS-CoV-2 antibody assays.

Assay Name	Anti-SARS-CoV-2-ELISA	SARS-CoV-2 IgG	Elecsys Anti-SARS-CoV-2
Provider	Euroimmun	Abbott	Roche
Cat. number	EI 2606–9601 G	6R86-22	09203095190
Assay Principle	ELISA	CMIAa	ECLIAb
Antigen	SP S1 domain	Nucleocapsid protein (NCP), C-Terminus	Nucleocapsid protein (NCP)
Detected Antibody	IgG	IgG	Total Ig
Abbreviation	Euroimmun-IgG	Abbott-IgG	Roche-Ig
Cut-Off value	1.1	1.4	1.0
Sensitivity (%)	84.9 [81.5; 87.9]c	88.9 [85.8; 91.6]	90.3 [87.4; 92.8]
Specificity (%)	99.3 [98.29; 99.76]	99.4 [98.50; 99.84]	100.0 [99.46; 100]

aChemoluminescence-microparticle immunoassay.

bElectrochemoluminescence immunoassay.

c95% confidence interval.

(1) A semiquantitative, automated Euroimmun assay (IgG: Cat# EI 2606–9601 G, IgA: Cat# EI 2606–9601 A, Lübeck, Germany) detecting SARS-CoV-2-specific IgG or IgA antibodies targeting the spike protein (S1 domain) and an Euroimmun Analyzer I that measures the OD of the samples. The assay is a classical ELISA, with stripes coated with SARS-CoV-2 antigen, that is incubated with samples and probed with enzyme-labeled anti-human IgG/IgA antibodies. Results are provided as OD and ratio of OD(Sample)/OD(calibrator). A threshold of 0.8 indicates borderline result, above a threshold of 1.1 a sample was considered as positive. (2) A two-step immunoassay for the qualitative detection of IgG antibodies directed against the nucleocapsid protein (NCP) of SARS-CoV-2 (Abbott, Cat# 6R86-22, Wiesbaden, Germany). This assay is based on the chemiluminescence-microparticle immunoassay (CMIA) technique. Briefly, sample, paramagnetic beads coated with SARS-CoV-2 antigen and diluent were incubated together. SARS-CoV-2-specific IgG antibodies bind to the antigen-coated beads. After washing, the acridinium-labeled anti-human IgG conjugate was added and incubated. After washing, pre-trigger and trigger solution were added and the resulting chemiluminescence reaction measured as relative light units. The amount of detectable IgG is directly proportional to the measured signal. Results are given as index (signal probe/signal calibrator) and considered positive above a threshold of 1.4. (3) The third assay is based on the electrochemiluminescence immunoassay (ECLIA) technique for the qualitative detection of antibodies directed against the NCP of SARS-CoV-2 (Elecsys Anti-SARS-CoV-2 assay, Roche Diagnostics GmbH, Cat# 09203095190, Frankfurt, Germany). Thus, unlike the above assays, the Roche-Ig test detects all immunoglobulins (Ig), and not solely IgG or IgA antibodies. This assay is based on a sandwich principle. Briefly, samples are sandwiched between biotinylated SARS-CoV-2-specific recombinant antigen and SARS-CoV-2-specific recombinant antigen labeled with a ruthenium complex. After addition of streptavidin-coated microparticles the complex is immobilized on the solid phase. The mixture is transferred into the measuring cells, where the magnetic beads are captured on the surface of the electrodes. After washing, a voltage pulse triggers chemiluminescence, which is measured by a photomultiplier. Results are provided as cut-off index (COI), and are considered as positive from a value of 1.0.

Data on sensitivity and specificity were kindly provided by the Paul-Ehrlich-Institute (PEI), a German federal institute, and respective test performance characteristics were further used for correction of seroprevalences. At PEI, 84% of the individuals tested had a very low COVID-19 symptom score of 1–2 (out of 7); specificity was tested on 676 pre-pandemic negative blood samples (100 serum, 576 citrated plasma samples). Part of these data were recently published[16].

Data processing, statistics and reproducibility

To estimate the seroprevalence in the general population, the observed seroprevalences were adjusted for age and sex using direct standardization (weights were derived from the population of the calendar year 2018, Supplementary Table 1). Confidence intervals (CI) were calculated using R version 4.0.3 as suggested by Waller and colleagues[17]. For sex and age strata 95% Blyth-Still-Casella CI were calculated using the rbscCI package of R[18]. Specific estimates were derived for the following ages: 18–44, 45–69, 18–69, ≥70 years). Adjusted estimates of seroprevalences were further corrected using test performance characteristics (sensitivity and specificity) of each individual test (data from PEI, Table 1) by applying following equation: (prevalence + specificity − 1)/(sensitivity + specificity − 1)[19]. The 95% confidence intervals for corrected seroprevalence, underestimation ratio and infection fatality rate were estimated using percentile bootstrap confidence intervals from 50,000 bootstrap samples[20]. For this purpose, the sensitivity and specificity were estimated from bootstrapped samples of the PEI data and the adjusted seroprevalence was estimated from bootstrapped samples of the SaarCoP study data. Subsequently, the corrected seroprevalence was estimated using the Rogan-Gladen estimator[19].

Seroconversion occurs ~10–14 days after infection[21]. According to the Robert Koch-Institute, the median time from symptom to death was 11 days during the 1st pandemic wave in Germany[20], while multinational studies reported an interval of 16–18 days[22-24]. We therefore used both serological results and registered COVID-19 death numbers 14 days after the registered numbers of SARS-CoV-2 PCR-confirmed cases for both the estimation of infection and fatality rates. Case-fatality rate was calculated as the ratio of COVID-19 death cases (14 days after the PCR case reporting date) in relation to PCR-confirmed SARS-CoV-2-positive cases (Supplementary Table 2). Underestimation ratio is the underreporting rate calculated as the ratio of age- and sex-adjusted seroprevalences (reporting date October 15, 2020) that were corrected with respect to test performance characteristics, in relation to SARS-CoV-2 PCR-positive cases (reporting date October 1, 2020, Supplementary Table 3). Infection fatality rate (IFR) was calculated as the ratio of COVID-19 death cases (reporting date October 15, 2020, Supplementary Table 4) in relation to age- and sex-adjusted seroprevalences that were corrected with respect to test performance characteristics. Data were illustrated and statistical calculations performed with Graph Pad Prism 9 (Graph Pad Software, San Diego, USA) using the indicated tests. Significant differences were accepted if p ≤ 0.05. Test statistic (t) and degree of freedom (df) are indicated.

Table 2

Assay performance: calculation of the sensitivities with sera from convalescent donors and specificities based on the individual assay performances in A.

Convalescent donors				Potentially cross-reactive sera
	EI-IgG	Abbott-IgG	Roche-Ig		EI-IgG	Abbott-IgG	Roche-Ig
n	61	61	61	n	128	78	122
Positive	52	49	55	Crossreaction	2	4	0
% Negative	14.75	19.67	9.84	%	1.67	1.69	0
Sensitivity (%)	85.25	80.33	90.16	Specificity (%)	98.33	98.31	100
CI (%)	73.83;  93.02	68.16; 89.04	79.81; 96.30	CI (%)	94.47; 99.81	87.39; 98.59	97.02; 100.00

Euroimmun-IgG = EI-IgG. 95% confidence intervals (CI %) were calculated according to Clopper and Pearson[34].