key: cord-0977927-cda94ctb
authors: Melotti, R.; Scaggiante, F.; Falciani, M.; Weichenberger, C. X.; Foco, L.; Lombardo, S.; De Grandi, A.; von Laer, D.; Mahlknecht, A.; Pramstaller, P. P.; Pagani, E.; Meier, H.; Gaertner, T.; Troi, C.; Mascalzoni, D.; Pattaro, C.; Mian, M.
title: Prevalence and determinants of serum antibodies to SARS-CoV-2 in the general population of the Gardena Valley
date: 2021-03-20
journal: nan
DOI: 10.1101/2021.03.19.21253883
sha: d97ef4f11cc8ff06f2bf749f1204c645b1524afb
doc_id: 977927
cord_uid: cda94ctb

Background. Community-based studies are essential to quantify the spread of SARS-CoV-2 infection and for unbiased characterization of its determinants and outcomes. We conducted a cross-sectional study in the Gardena valley, a major Alpine touristic destination which was struck in the expansion phase of the COVID-19 pandemic over the winter 2020. Methods. We surveyed 2244 representative study participants who underwent swab and serum antibody tests. We made multiple comparisons among the Abbott and Diasorin bioassays and serum neutralization titers. Seroprevalence accounted for the stratified design, non-response and test accuracy. Determinants and symptoms predictive of infection were analyzed by weighted multiple logistic regression. Results. SARS-CoV-2 seroprevalence was 26.9% (95% confidence interval: 25.2%, 28.6%) by June 2020. The serum antibody bioassays had modest agreement with each other. Receiver operating characteristic curve analysis on the serum neutralizing capacity showed better performance of the Abbott test at lower than the canonical threshold. Socio-demographic characteristics showed no clear evidence of association with seropositivity, which was instead associated with place of residence and economic activity. Loss of taste or smell, fever, difficulty in breathing, pain in the limbs, and weakness were the most predictive symptoms of positive antibody test results. Fever and weakness associations were age-dependent. Conclusion. The Gardena valley had one of the highest SARS-CoV-2 infection prevalence in Europe. The age-dependent risk associated with COVID-19 related symptoms implies targeted strategies for screening and prophylaxis planning.

. CC-BY 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 March 20, 2021. ;  https://doi.org/10.1101/2021.03. 19 .21253883 doi: medRxiv preprint

During the initial phase of the SARS-CoV-2 pandemic, the Gardena valley, a well-known winter tourism destination located in South Tyrol (Italy), became one of the European regions most afflicted by the coronavirus disease 2019 . While in the middle of the virus circulation vortex since February 2020, there were a multitude of holidaymakers and visitors in the valley mainly from Northern Italy and Central Europe.

Back home, tourists likely contributed to further transmission of the virus just before containment actions were endorsed by regions worldwide. [1] As expected in such an emergent phase of the pandemic, hospital-based case reports dominated the accumulation of scientific evidence on COVID-19. [2] Consequently, public awareness, ongoing knowledge of the determinants of disease and disease severity, and current prevention strategies have been profoundly influenced by clinical observations, while evidence from community studies has had limited space in context. [3] Specific knowledge of the exogenous determinants of SARS-CoV-2 infection and its related symptoms or about biological susceptibility in the general population is still incomplete, probably due to the slower pace and relative paucity of community-based studies. [4] Geographically confined regions with a relatively high incidence of infection may help characterize the spread of COVID-19, providing useful indications to policy-makers for current and future preventive efforts.

At the end of May 2020, we surveyed 2244 inhabitants of the Gardena valley representative of the local population, measured antibody test response to SARS-CoV-2 and related that response to symptoms, prior conditions and serum neutralizing capacity. The high seroprevalence qualified the in-depth analysis of determinants and COVID-19-related symptoms in a general population setting, augmenting the general understanding of the disease dynamic.

Invited to the study were 2958 of the 9424 inhabitants of Ortisei, Santa Cristina, and Selva, the main municipalities of the Gardena valley, following a one-stage random sampling design stratified by municipality, sex and age group (<6, 6-17, 18-34, 35-49, 50-64, 65+ years). Sample size was defined based on an expected 3% minimal seroprevalence with 0.25% relative standard error (SE) and accounting for finite population correction. Participants were selected with known extraction probability from the municipality registries, excluding nursing homes, using the 'surveyselect' program in SAS v9.2.

Participants were invited via letter including: the planned participation date; a link to the online questionnaire (with telephone support) covering demographic, clinical and socio-. CC-BY 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 March 20, 2021 Two hundred ninety-nine serum samples were selected for plaque reduction neutralization test (PRNT), [6] ensuring the coverage of the whole SARS-CoV-2 IgG assay AAI distribution while maximizing sample heterogeneity in terms of sex, age, and symptoms manifestation as well as previous diagnosis and hospitalization (additional details in Suppl. Mat. page 7 and Suppl. Fig. 1A were fixed for 5 minutes with 96% ethanol and subsequently stained using the serum from a SARS-CoV-2 recovered patient and a horse radish peroxidase-conjugated antihuman secondary antibody (Dianova). Plates were developed using 3-amino-9ethylcarbazole substrate. Infected cells were counted via microscope and 50% neutralization titers were calculated as the highest dilution where the mean infection of duplicate samples was reduced by >50% of the mean of control wells lacking the virus.

The presence of spike (S) protein antibodies in the 299 samples was assessed with the Diasorin LIAISON ® SARS-CoV-2 S1/S2 IgG chemiluminescent assay (Saluggia, Italy), designed to detect the number of arbitrary units (AUs) of specific IgG class antibodies directed against the S1 and S2 viral proteins, at the Microbiology and Virology laboratory of the South Tyrolean Healthcare System (Bolzano/Bozen, Italy). Results were classified as negative (values<12 AU/ml), inconclusive (values≥12 and <15 AU/ml), or positive (values≥15 AU/ml) according to manufacturer's recommendations. This assay had declared 97.4% sensitivity >15 days after symptoms onset, and 98.9% specificity. [7] Statistical analyses

We assessed pairwise agreement between quantitative variables with the Lin's concordance correlation coefficient (CCC) [8] and Bland-Altman plot, [9] and agreement between categorical variables using the Cohen's kappa statistic, [10] using the 'epiR' v1.0-15, 'BlandAltmanLeh' v0.3.1, and 'psych' v1.9.11 packages in the R software v3.3.6.

To investigate the discrimination accuracy of the 1.4 AAI threshold on thawed serum against a PRNT value≥4, which was considered as a gold-standard for prior exposure to SARS-CoV-2, we conducted receiver operating characteristic (ROC) curve analysis identifying the optimal cutoff as the Youden's index (J). [11] Overfitting was prevented by performing both 10-fold cross-validation and repeated random sub-sampling validation. In the latter approach, we randomly split the sample set into 80% and 20% training and test sets, respectively, corresponding to 239 and 60 observations, repeatedly 5000 times. The two validation procedures resulted in sets of 10 and 5000 optimal cutoffs, respectively. We reported the median optimal cutoff of each respective validation procedure.

Descriptive tables and multiple logistic regression models display observed counts for each relevant category, while accounting for the study design: stratification by sex, municipality and age group, post-stratification (citizenship by municipality) and finite population correction, for efficient proportion estimations. To correct for possible selection biases, we also adjusted the sampling weights based on proportional allocation within strata by non-response (balanced to the population strata distribution), which were then calibrated dynamically by post-stratification in each analysis. Prevalence of serum antibodies to SARS-CoV-2 (seroprevalence) in the overall sample was also . CC-BY 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) (18.1%) showed neutralization capacity. In contrast, neutralization capacity was always . CC-BY 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 March 20, 2021. ; https://doi.org/10.1101/2021.03.19.21253883 doi: medRxiv preprint confirmed when both the Abbott and Diasorin tests were positive, and almost always when at least one of them was positive (Fig. 1B) .

The optimal Abbott antibody test threshold identified in the ROC curve analysis was 1.16 ( Fig. 1C) , consistently in both the 10-fold cross-validation and 5000 repeated random sub-sampling validation, both returning 1.16 median AAI. At this threshold, the classifier performed with 89.0% sensitivity and 97.2% specificity. At the recommended 1.4 threshold, sensitivity was lower (83.3%) but specificity was the same. Sex and age stratification improved the assay performance with relatively lower thresholds but at the cost of reliability (Suppl. Fig. 3 ; Suppl. Tab. 3).

Among all sociodemographic and lifestyle characteristics considered ( (Table 2 ). Females did appear at lower risk than males, however, evidence was weak in the mutually adjusted analysis.

Focusing on symptoms as possible seroprevalence predictors, we found strong evidence for any single symptom to predict antibody positivity, both in unadjusted and mutually adjusted analyses, which considered each symptom at one time ( (Figure 2A , 2B, and 2C). In multiple logistic regression we fitted interaction terms of fever and weakness with age ( Table 5 ). The probability of seropositivity was higher in older participants who also reported either fever or weakness (Fig. 2D ). However, age was a mild predictor of infection in the absence of fever and weakness and independent of any other symptoms (OR=0.96; 95%CI: 0.93, 1.00, p=0.046; Table 5 ).

Among participants with any number of symptoms (n=1002), seroprevalence peaked in those with symptoms onset in the first half of March and returned the second highest figure for onset in the following fortnight (Table 4 , all p<0.001). Seroprevalence was higher also among those reporting symptoms in the second half of February, compared to other periods (p<0.001). This curvilinear trend was even more apparent when . CC-BY 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 present study shows that nearly 30% of the Gardena valley general population was infected by SARS-CoV-2 during the first pandemic wave. Neutralizing antibody capacity was measurably higher than estimated seropositivity. Our in-depth analysis of SARS-CoV-2 determinants of infection as well as COVID-19-related symptoms identified specific seroprevalence risk factors and extensively characterized the post-infection symptoms.

The main strength of the study is its representativeness that guarantees generalizability of the findings to the whole population of the valley. The study was concluded in a short time span of two weeks of a nearly three-month-long strict national lockdown. The whole framework of recruitment, sample-handling, and storage procedures allowed the remarkable participation response, while preventing the possible influence of additional external factors through the period.

. CC-BY 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) Our study also comes with several limitations. First, antibody testing methods have imperfect accuracy. We corrected seroprevalence analyses for the test sensitivity and specificity to overcome this limitation. Second, certain social groups such as, for instance, non-Italian residents might have been underrepresented. While we corrected all analyses for differential participation in known groups, we could not prevent participation bias of unknown sources such as, for example, COVID-19-related mortality and possible self-selection of severely ill individuals. A third limitation is the questionnaire self-administration: while this was the only way to collect essential information, response bias might affect some analyses. For instance, symptom onset estimation might not be totally accurate as the question was not specific to each possible symptom. Similarly, we cannot exclude that symptoms were due to competitive seasonal diseases such as flu or allergies.

A nearly 30% seroprevalence is a large figure compared to other studies. [4] This estimate aligns with those of nearby Italian regions. [13, 14] However, the Manaus case shows that it might still result in a non-negligible underestimation of the true infection rate. [15] Several reasons suggest that our estimate should be considered as a lower bound of the real seroprevalence. First, the survey had missed people who already died.

In 2020, the raw excess all-cause mortality rate over the previous five years was between +32.8% and +72.4% in the three municipalities. [16] Second, antibodies can persist until five months or more, but we cannot exclude that IgG levels had already contributing to prevalence underestimation. Third, it is debated whether previous exposure to other coronaviruses causing the common cold could generate long-lasting SARS-CoV-2-targeting antibodies. Cross-reactive antibodies were recently identified in few adults and more frequently in children and adolescents supporting preexisting immunity. [22] Additionally, multiple studies reported cross-reactive T cell memory in 28%-50% people, [23] increasing the possibility that some preexisting immunity is already present. Fourth, adding to the extant evidence, our comparison of the Abbott SARS-CoV-2 assay against the Diasorin assay and the PRNT reflects that a relatively higher proportion of individuals may have been in contact with the SARS-CoV-2.

The imperfect agreement between the Abbott and Diasorin tests can be explained by the two assays being directed against different viral antigens with different kinetics: the anti-nucleocapsid protein IgG for Abbott and the anti-S1/S2 portions of spike protein IgG for Diasorin.

[18] Furthermore, the comparison against a neutralization test identified . CC-BY 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 March 20, 2021. ; https://doi.org/10.1101/2021.03.19.21253883 doi: medRxiv preprint several samples with neutralization capacity that had negative antibody test results. Part of their misclassification against the neutralization test was due to the initial definition of precautionary thresholds set by companies, where unclassifiable/dubious indices were set to negative. The recent revision of diagnostic antibody test thresholds will limit this misclassification. [24] Standing by reported symptoms, infections have peaked during the first half of March 2020, as in neighboring regions. [3, 13, 14] Spoken language, nationality and educational level were not associated with SARS-CoV-2 seropositivity, supporting absence of social stratification in the exposure to the virus. Sex was not a major determinant of infections even though females had ~20% lower risk than males, in line with a population study conducted in the nearby province of Trento. [14] This might reflect a higher prevention-prone behavior of females or a higher rate of neutralizing autoantibodies against type I interferon in COVID-19 severely affected males. [25] In contrast to similar studies, [13, 14] while seropositivity had no general evidence of positive association with age, it was inversely associated with age in the absence of fever and weakness in our study. This is perhaps due to a socially diverse population of positive individuals, which were younger and arguably linked to winter-sporting activities in our study. Moreover, the association of seroprevalence with municipality and the accommodation and catering services suggests that infections might have been mainly driven by unavoidable occupational circumstances. The apparent protective effect of current smoking on SARS-CoV-2 seroprevalence is not novel. [26, 27] A collider bias effect may supersede sample representativeness. [28] COVID-19 increases mortality risk just as smoking does. Current smoking is causally related to more severe COVID-19 disease.

[29] With COVID-19 and smoking intertwined to affect mortality, it is likely that the apparent protective effect of smoking on seropositivity is explained by harvesting effects on mortality, or impairment to participation by health conditions or health-prone behaviors. [30] In conclusion, we confirm that the Gardena valley had one of the highest prevalence of SARS-CoV-2 infection in Europe. Comparisons between distinct antibody detection assays and between serum assays and serum antibodies neutralizing capacity, yet suggest an underestimation of actual seroprevalence in our study. The infection spread peculiarities in the area has probably mitigated the sex and age differences observed in other contexts. In contrast, all investigated flu-like symptoms were predictive of a . CC-BY 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 March 20, 2021. ; https://doi.org/10.1101/2021.03.19.21253883 doi: medRxiv preprint positive antibody test result, with highest and cumulative evidence for the loss of taste or smell, fever, difficulty in breathing, pain in the limbs, and weakness. However, some symptoms were associated with the seroprevalence in an age-dependent mode.

Overall, findings highlight that determinants of SARS-CoV-2 infection and outcomes are context dependent, as they relate to the pattern of infection, the local population composition, and the economic dynamics. Thus prevention strategies may be tailored to the social context.

. CC-BY 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 March 20, 2021 . CC-BY 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 March 20, 2021 . CC-BY 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 March 20, 2021 Rectangle sizes proportional to the frequency interval of symptoms across periods. The symptoms most predictive of seropositivity (loss of taste or smell; weakness; pain in the limbs; fever; and breathlessness) were apparently more prevalent among seropositive than seronegative participants at the time of peak incidence of the epidemic first wave in the valley, between late February, throughout March and part of April.

. CC-BY 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. CC-BY 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) . CC-BY 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 March 20, 2021. ; https://doi.org/10.1101/2021.03.19.21253883 doi: medRxiv preprint

Spread of SARS-CoV-2 in the Icelandic Population

Defining the Epidemiology of Covid-19 -Studies Needed

Suppression of a SARS-CoV-2 outbreak in the Italian municipality of Vo'

SARS-CoV-2 seroprevalence worldwide: a systematic review and meta-analysis

Performance Characteristics of the Abbott Architect SARS-CoV-2 IgG Assay and Seroprevalence in

Blocking transmission of Middle East respiratory syndrome coronavirus (MERS-CoV) in llamas by vaccination with a recombinant spike protein

Clinical and Analytical Performance of an Automated Serological Test That Identifies S1/S2-Neutralizing IgG in COVID-19 Patients Semiquantitatively

A concordance correlation coefficient to evaluate reproducibility

Statistical Methods for Assessing Agreement between Two Methods of Clinical Measurement

Statistical methods for rates and proportions

Index for rating diagnostic tests

Estimating prevalence from the results of a screening test

Seroprevalence of SARS-CoV-2 significantly varies with age: Preliminary results from a mass population screening

Prevalence of SARS-CoV-2 IgG antibodies in an area of northeastern Italy with a high incidence of COVID-19 cases: a population-based study

Three-quarters attack rate of SARS-CoV-2 in the Brazilian Amazon during a largely unmitigated epidemic

SARS-CoV-2-Specific T Cells Exhibit Phenotypic Features of Helper Function, Lack of Terminal Differentiation, and High Proliferation Potential

Functional SARS-CoV-2-Specific Immune Memory Persists after Mild COVID-19

Robust T Cell Immunity in Convalescent Individuals with Asymptomatic or Mild COVID-19

Preexisting and de novo humoral immunity to SARS-CoV-2 in humans

Adaptive immunity to SARS-CoV-2 and COVID-19

Longitudinal observation and decline of neutralizing antibody responses in the three months following SARS-CoV-2 infection in humans

Autoantibodies against type I IFNs in patients with lifethreatening COVID-19

Seroprevalence of SARS-CoV-2 among adults in three regions of France following the lockdown and associated risk factors: a multicohort study

Antibody prevalence for SARS-CoV-2 following the peak of the pandemic in England: REACT2 study in 100,000 adults

Collider bias undermines our understanding of COVID-19 disease risk and severity

Dietary supplements No Ref

The authors thank Martin Matscher (Ripartizione aziendale per l'assistenza territoriale, Azienda Sanitaria dell'Alto Adige, Bolzano, Italy), Dagmar Regele (Dipartimento di Prevenzione, Servizio Igiene e Sanità pubblica, Azienda Sanitaria dell'Alto Adige,