SARS‐CoV‐2 infections among pregnant women, 2020, Finland—Cross‐testing of neutralization assays

We studied the development of the severe acute respiratory syndrome‐related coronavirus (SARS‐CoV‐2) pandemic in southern Finland in 2020 and evaluated the performance of two surrogate immunoassays for the detection of neutralizing antibodies (NAbs). The data set consisted of 12 000 retrospectively collected samples from pregnant women in their first trimester throughout 2020. All the samples were initially screened for immunoglobulin G (IgG) with SARS‐CoV‐2 spike antibody assay (EIM‐S1, Euroimmun) followed by confirmation with nucleocapsid antibody assay (Architect SARS‐CoV‐2, Abbott). Samples that were reactive (positive or borderline) with both assays were subjected to testing with commercial surrogate immunoassays of NeutraLISA (EIM) and cPassTM (GenScript Biotech Corporation) by using pseudoneutralization assay (PNAbA) as a golden standard. No seropositive cases were detected between January and March. Between April and December, IgG (EIM‐S1 and Abbott positive) and NAb (PNAbA positive) seroprevalences were between 0.4% and 1.4%. NeutraLISA showed 90% and cPass 55% concordant results with PNAbA among PNAbA negative samples and 49% and 92% among PNAbA positive samples giving NeutraLISA better specificity but lower sensitivity than cPass. To conclude, seroprevalence in pregnant women reflected that of the general population but the variability of the performance of serological protocols needs to be taken into account in inter‐study comparison.

(HUS) between April and December 2020 ranged from 0.3% to 5.5% reaching its peak in November, while the prevalence in the rest of the country remained lower. 3The vaccination campaign began with healthcare workers at the end of December 2020 and continued during 2021 with other population groups.
Serological assays detecting immunoglobulin G (IgG) are commonly used in epidemiological surveys and population-level monitoring. 4Only part of the antibodies produced can neutralize the virus itself, and further protect the individual, however, other antibodies may help to protect the individual in vivo. 5Monitoring of neutralizing antibodies (NAbs) instead of IgG could be used, for example, to follow the level of protective immunity of immunosuppressed patients or candidates for convalescent plasma therapy or to assess the success of vaccination.
7][8] Due to this, several commercial surrogate assays for detecting NAbs have been designed.However, they have displayed discrepant performances in different laboratories. 4,6,8,9Performance of different immunoassays, including the ones detecting NAbs originating from natural infection or vaccination, should be validated carefully to be used in diagnostic settings for, for example, revaccination purposes.As postvaccination antibodies can be high and affect the results in assays detecting NAbs, for example, by increasing sensitivity, 7 we wanted to test these assays after natural infection without vaccination history.Surrogate assays that show optimal performance with convalescent sera should then be further validated with individuals with vaccination or hybrid immunity.
Here, our first goal was to screen antibodies against SARS-CoV-2 in 12 000 pregnant women sampled in routine maternity screening between January and December 2020 to follow the development of the pandemic in southern Finland.Another goal was to evaluate the infection risk to expectant mothers as well as the performance of commercial surrogate assays designed for the detection of NAbs.
Initial screening was carried out with a sensitive commercial immunoassay detecting SARS-CoV-2 spike antibodies (EIM), followed by a more specific immunoassay detecting antibodies against SARS-CoV-2 nucleocapsid (Abbott).The reactive samples were tested for NAbs using a pseudoneutralization assay (PNAbA). 10 Services in Kymenlaakso (Kymsote) (Figure 1).Together, these areas inhabit 36% of Finnish population (Table 1 and Supporting F I G U R E 1 Location of the hospital districts (A) and sample amounts and population densities in the municipalities in the region (B).
Information S1: Table S1), and the mean population density is 106 inhabitants/km 2 . 11During 2020, the first coronavirus disease 2019 (COVID-19) cases were reported January 29th and February 26th (travelers) and by the end of the year, 1.14% of the population in study areas had a laboratory-confirmed disease, the highest number being in the most densely populated HUS region. 2

| Screening of SARS-CoV-2 spike and N IgG antibodies
All 12 000 serum samples were screened for IgG antibodies with EUROLabworkstation (EIM) and SARS-CoV-2 IgG kit (S1-based antigen, EIM; later EIM-S1-assay) according to manufacturer's instructions.Samples that were reactive with the EIM-S1-assay (result positive or borderline, ≥0.8 index) were subjected to a second screening using Architect Analyzer (Abbott) and SARS-COV-2 IgG (nucleoprotein-based antigen, Abbott; later Abbott-N-assay) according to manufacturers' instructions (Supporting Information S1: Table S2).All the samples that were ≥0.8 index (manufacturer reported limit for borderline result) with EIM-S1-assay and ≥0.3 index with Abbott-N-assay (borderline limit determined based on earlier experience (data not shown) and this study (Supporting Information S2: Figure S1) were considered reactive and were further selected for NAb screening.Also, individuals who had been previously positive in SARS-CoV-2 nucleic acid test were selected for NAb testing (Supporting Information S1: Table S2).

| PNAbA and commercial surrogate ELISAs for detection of SARS-CoV-2 NAbs
NAbs were analyzed with PNAbA described earlier 10,12,13 (University of Helsinki) and two commercial surrogate immunoassays: SARS-CoV-2 NeutraLISA (EIM) and cPass TM SARS-CoV-2 Neutralization Antibody Detection Kit (GenScript).Both commercial surrogate assays were carried out and interpreted according to the manufacturer's instructions.Both surrogate assays detect NAbs, which inhibit the binding of S1/RBD to ACE2-reseptors. 14cording to the manufacturers, NeutraLISA and cPass have a sensitivity of 95.9% and specificity of 99.7%, and 93.8% and 99.4%, respectively.

| Statistical analysis
To avoid false-positive results, a two-tier approach was used to determine SARS-CoV-2 IgG positivity for seroprevalence calculations: those patients who were positive with both EIM-S1 ( > 1.1 index) and Abbott-N ( > 1.4 index) according to manufacturer's guidelines, were considered positive.Borderline results were considered negative.For NAb seroprevalence, the samples were considered positive if they had a detectable titer in PNAbA (20 or above).As it was not possible to test all the 12 000 samples with PNAbA, the samples that were not reactive with both IgG enzymelinked immunosorbent assay (ELISAs) (EIM-S1 and Abbott-N-assays) were considered PNAbA negative for calculation purposes.The titers below 20 were set to 10 for calculations.The 95% confidence intervals were calculated with Wilson score method (binominal proportion).Receiver operating characteristic curve (ROC) curve analysis was performed to compare the methods to PNAbA with titer cut-offs 20 (also week neutralization included) or 80 (only strong neutralization included).Spearman's ρ was used to study correlation.Cohen's Kappa analysis was used to compare the results between assays.6][17][18][19] Maps were drawn with QGIS 3.26.2utilizing datasets by World Food Programme and Statistics Finland. 20,21| RESULTS

| IgG and NAb seroprevalence
In total, there was one recoded sample available per individual, 10 844 of which originated from HUS, 493 from Eksote, and 663 from Kymsote.The overall mean age was 31.6 years (range 14-61, SD 5.2).Mean ages in hospital districts were 31.7 (SD 5.1) in HUS, 30.7 (SD 5.1) in Eksote, and 30.6 (SD 5.4) in Kymsote region (Supporting Information S2: Table S3).Information S2: Figure S2, Table 2).Out of the NAb-positive patients, only 28% had a PCR-confirmed disease at some point before sampling with the time from the disease varying between less than 1 and 9 months.The rest had either been negative, been tested in private health care (data not available), or not been tested at all.Ten PCR-confirmed cases were PNAbA negative and the samples were

| Comparison of surrogate immunoassays and PNAbA
Two different commercial surrogate assays of NeutraLISA (EIM) and cPass TM (GenScript) were evaluated and the results were compared with PNAbA (Supporting Information S2: Figure S3).In total 120 samples were tested with PNAbA.Samples with titer of 20 or more in PNAbA (N = 91; Table 3) were considered as samples with true NAbs.All the tests showed a significant correlation with Spearman's ρ varying between 0.448 (Abbott-N and cPass TM ) and 0.884 (NeutraLISA vs. cPass TM , Supporting Information S2: Table S4, Figure 3    Abbott-N (0.758) when titer cut-off 20 was used.When cut-off 80 was used, cPass TM had the highest AUC (0.920) and Abbott-N the lowest (0.732).However, using of manufacturer-reported guidelines led to big false-positive rates in almost all the tests.Only with NeutraLISA, the sensitivity was 71% and false positivity rate only 11% when cut-off 80 was used.
Significant differences were detected between the two IgG assays.
Only 28% of EIM-S1 positive samples were positive with Abbott-N ( ≥ 1.4), which is even lower than 44% (67/240) reported in the previous study. 22A total of 91% of IgG-positive patients were also NAb positive and 67% of NAb-positive patients were also IgG positive.

| DISCUSSION
We studied the development of the pandemic during 2020 in southern Finland using retrospectively collected anonymous samples from pregnant women.We also compared serological methods for clinical settings.This Finnish serosurvey covers uniquely the whole year of 2020 with an exceptionally large sample set.
In early 2020, only highly targeted diagnostics were carried out as resources were targeted for COVID-19.Nevertheless, we had access to samples of anonymous pregnant women, providing enough samples per month (1000 per month) and used these to retrospectively study the pandemic throughout the whole year 2020 in southern Finland.It should be noted that the data set represents one population group and hence the seroprevalences should be generalized to other population groups with caution.However, Mattern et al.
did not detect differences between women at delivery and general community, suggesting that seroprevalence in pregnant women may reflect that of the whole population. 23 depending on the country, timeframe, and study group. 24Reported seroprevalences varied between 0.7% and 45.3% in health care workers and between 0.24% and 23% in community studies.The majority of the studies detected no differences between genders whereas comparison between age groups gave conflicting results. 24st like community studies, studies on pregnant women have reported seroprevalences ranging from less than 0.5% (e.g., Estonia) 6][27][28][29][30] Eskil et al. noted a gradual rise from 0.5% to 5.7% in pregnant women until the end of 2020. 27Here, the seroprevalence was bigger in the HUS region most likely due to the highest population density which is consistent with national PCR-confirmed case numbers from whole population.Seroprevalence showed surprisingly small rise compared with PCR-case numbers and may be explained by waning immunity over time, and public health-guided changes in behavior, that is, masking and distance behavior, of the study group during the pandemic in autumn 2020.The same could be detected when EIM-S1 and Abbott-N were studied separately (Supporting Information S2: Figure S2).
The positive predictive value of the test (aka is the positive test result true mark of the disease) is affected by the sensitivity and specificity of the test, but also by the disease occurrence in the population in that moment.As the disease occurrence was very low in 2020, we stressed high specificity over high sensitivity.Therefore, a two-tier system, detecting both S1 and N antibodies, was selected to ensure we detect only true positives at the early stage of the pandemic when the disease was still rare.From January until the end of March, we did not detect any COVID-19 IgG or NAb-positive individuals among 3000 samples.From April until December, comprising 9000 samples from HUS, Eksote, and Kymsote regions, the numbers slowly increased from 0.7% to 1.3% (IgG) and from 0.9% to 1.4% (NAb).A community follow-up including both sexes and different age groups carried out by THL (Supporting Information S2: Figure S6) reported 0.5%-5.5% seroprevalences in HUS region between April and December 2020. 3These numbers are a bit higher than in this study with expectant mothers, but considering the different detection protocols and confidence intervals, prevalence here does not seem to significantly differ from the prevalence detected in the community.There was no elevated infection risk seen in these data.It is also possible that during pregnancy people tend to be more careful and avoid exposures, leading to slightly lower seroprevalence.Compared with other studies on pregnant women and community studies, seroprevalences are in the lower part of the spectrum, 23,[25][26][27][28][29][30] which is expected considering that diagnosed case numbers in Finland were also relatively low compared with several other countries (Figure 2C,D).
As a limitation, two-tier methods may have resulted in underestimation of the seroprevalence.However, relying solely on EIM-S1 would have resulted in severe overestimation considering that several positive results were detected even before any cases were reported in Finland.When the disease burden is low and specificity of the test (EIM-S1) as low as 86.6% (9) the false positives occur relatively often, resulting overestimation of true seroprevalence.
Previous studies have noted the lower seroprevalence given by Abbott-N assay and faster waning of antibody response. 31

METHODS 2 . 1 |
Finally, two different commercial surrogate immunoassays of NeutraLISA (EIM) and cPass TM (GenScript) aimed for the detection of SARS-CoV-2 NAbs were validated against PNAbA. 2 | Sampling and data set Altogether, 12 000 serum samples were retrospectively collected in 2020 (1000 samples per month).These serum samples were originally sent for routine screening of hepatitis B S-antigen, antibodies and antigen for HIV, and syphilis antibodies during the first trimester of pregnancy (Helsinki University Hospital, Diagnostic center).Samples were randomly collected and recoded.Only age, hospital district, and SARS-CoV-2 PCR info were available.Samples originated from three hospital districts in Southern Finland: Hospital District of Helsinki and Uusimaa (HUS), South Karelia Social and Health Care District (Eksote), and Social and Health Out of the 12 000 individuals, 435 were reactive (either borderline or positive) in EIM-S1-assay.Only 116 of them also had reactive results in Abbot-N-assay, and all these 116 were further tested with PNAbA for NAbs.All previously PCR-positive individuals (N = 35; PCR positivity varied from 0 to 9 months prior serum T A B L E 1 Geographical information about hospital districts and population.
Abbott-N negative but EIM-S1-positive individuals with PCRconfirmed previous infection.
taken 1 to 6 months after the infection.Out of these, four had no detectable NAbs with cPass TM or NeutraLISA either, two were positive with both of them and four only with cPass TM .Differences were detected between hospital districts, as only one NAb positive (0.2%) and no IgG-positive patients were detected in the least densely populated Eksote region whereas 0.61% were IgG positive and 0.80% NAb positive in the most populated HUS region.
and Supporting T A B L E 2 IgG and NAb prevalences by hospital district and month.
Information S2: FigureS4).Correlation with PNAbA titer was 0.819 with cPass TM and 0.757 with NeutraLISA.AUC value based on ROC curve (Supporting Information S2: FigureS5, considering PNAbA a golden standard) was highest with EIM-S1 (0.882) and lowest with Also, a study group of pregnant women allows a good comparison between time points and regions because the data set is more uniform when it comes to age and gender factors and behavioral patterns.Random selection from all women in routine pregnancy screening also allows sampling with little selection bias.Most of the seroprevalence studies use a timeframe of a few weeks to a few months and methodology has varied greatly, complicating comparison between studies.A review on seroprevalence in Europe in 2020 described a lot of heterogenicity F I G U R E 3 Comparison of EIM-S1 and Abbott-N (A) and pseudoneutralization assay (PNAbA) and cPass TM (B) and NeutraLISA (C).PNAbA titers are expressed as log2-scale and 0.3 jitter has been used for clarity.Manufacturer's cut-offs are marked with vertical and horizontal lines.
Abbott-N assay and faster waning of antibody response.31Supporting these earlier observations, the biggest differences between Abbott-N and EIM-S1 were detected between August and November (only 10%-30% of EIM-S1-positive samples being also Abbott-N positive), a few months after the first disease wave.Between April and July as well as December, 40%-50% of EIM-S1 positive samples were also Abbott-N positive.Due to the lower sensitivity of Abbott-N, a low borderline cut-off was used when selecting samples for neutralization assays to avoid leaving true positive samples out of further testing.The second aim was to evaluate the commercial surrogate immunoassays (NeutraLISA, EIM and cPass TM , GenScript) for the detection of NAbs.Several studies have indicated that surrogate neutralization assays show moderate to good correlation with titers of neutralization assays.6,8,32,33Here, we used a positive PNAbA standard) have true neutralizing capacity.Hence, surrogate tests with possibly higher sensitivity than neutralization assays should be interpreted with caution in diagnostic settings when it is important to know if the patient has some immune protection or not.Selection of the diagnostic tests depends on many factors, including the intended use.One of the major applications of NAb assays is monitoring the immunity status of patients, for example, immunosuppressed patients and expectant mothers, and evaluating the need for vaccine booster.Another application is the assessment of herd immunity.If surrogate NAb assays are used in these settings, false positives can be considered problematic.Differences in sensitivity and specificity of different tests as well as varying cutoff values, and interpretation methods between laboratories cause significant challenges for interstudy comparison, which should be addressed to form a proper picture of the global development of COVID-19 and possible future pandemics.With these issues in mind, more development and more optimizations may be needed for surrogate assays to reach the performance needed in diagnostic settings.More data are also needed to establish the real-life performance of these surrogate assays, including the performance after vaccination and hybrid immunity.
Comparison of PNAbA to cPass TM and NeutraLISA results.
Note: Numbers are reported in parenthesis and 95% confidence intervals (Wilson score method) in brackets.Abbreviations: IgG, immunoglobulin G; NAb, neutralizing antibodies.a Positive in both EIM ( ≥ 1.1 = pos) and Abbott ( ≥ 1.4 = pos); EIM ≥ 0.8-1.0Borderline.T A B L E 3 358,32,33Here, we used a positive PNAbA result as a real indication of detectable NAbs.Among the negative IgG-positive samples were NAb positive but only 2.5% of NAb-positive samples were IgG negative.They did not compare the tests to microneutralization assay.35In these dat, 60% of IgG-positive samples were positive in NeutraLISA.If we had considered borderline results positive like El-Ghitany et al., it would have been 81%.Olbrich et al. noticed a lower sensitivity with microneutralization assay (81%) than cPass TM (96%) when compared with PCR-positive results in the past and suggested that cPass TM might be more sensitive in detecting past infection than traditional microneutralization assay.
36However, more studies would be needed to determine if samples that are positive with surrogate assays but negative with neutralization assays (commonly accepted golden