key: cord-0713372-5kxx44kq
authors: Rostami, Ali; Sepidarkish, Mahdi; Fazlzadeh, Aylar; Mokdad, Ali H.; Sattarnezhad, Aida; Esfandyari, Sahar; Riahi, Seyed Mohammad; Mollalo, Abolfazl; Dooki, Mohammadreza Esmaeili; Bayani, Masomeh; Nazemipour, Maryam; Mansournia, Mohammad Ali; Hotez, Peter J.; Gasser, Robin B.
title: Update on SARS-CoV-2 seroprevalence: regional and worldwide
date: 2021-09-25
journal: Clin Microbiol Infect
DOI: 10.1016/j.cmi.2021.09.019
sha: 5f40342881582d88b25ef4e02309d86dd15d86ca
doc_id: 713372
cord_uid: 5kxx44kq

BACKGROUND: With limited vaccine supplies, an informed position on the status of SARS-CoV-2 infection in people can assist the prioritization of vaccine deployment. OBJECTIVES: We performed a systematic review and meta-analysis to estimate the global and regional SARS-CoV-2 seroprevalences around the world. DATA SOURCES: We systematically searched peer-reviewed databases (PubMed, Embase and Scopus), and preprint servers (medRxiv, bioRxiv and SSRN) for articles published between 1 January 2020 and 30 March 2021. STUDY ELIGIBILITY CRITERIA: Population-based studies reporting the SARS-CoV-2 seroprevalence in the general population were included. PARTICIPANTS: People of different age groups, occupations, educational levels, ethnic backgrounds and socio-economic status from the general population. INTERVENTIONS: There were no interventions. METHODS: We used the random-effects meta-analyses and empirical Bayesian method to estimate the pooled seroprevalence and conducted subgroup and meta-regression analyses to explore potential sources of heterogeneity as well as the relationship between seroprevalence and socio-demographics. RESULTS: We identified 241 eligible studies involving 6.3 million individuals from 60 countries. The global pooled seroprevalence was 9.47% (95% CI 8.99–9.95%), although the heterogeneity among studies was significant (I(2) = 99.9%). We estimated that ∼738 million people had been infected with SARS-CoV-2 (as of December 2020). Highest and lowest seroprevalences were recorded in Central and Southern Asia (22.91%, 19.11–26.72%) and Eastern and South-eastern Asia (1.62%, 1.31–1.95%), respectively. Seroprevalence estimates were higher in males, persons aged 20–50 years, in minority ethnic groups living in countries or regions with low income and human development indices. CONCLUSIONS: The present study indicates that the majority of the world's human population was still highly susceptible to SARS-CoV-2 infection in mid-2021, emphasizing the need for vaccine deployment to vulnerable groups of people, particularly in developing countries, and for the implementation of enhanced preventive measures until ‘herd immunity’ to SARS-CoV-2 has developed.

Since March 2020, the COVID-19 pandemic has been a major health challenge, devastating many communities and economies around the world [1, 2] . From the start of the pandemic to mid-August 2021,~211 million confirmed cases of COVID-19 and 4.5 million deaths were recorded worldwide [3] . However, the number of reported cases is likely substantially underestimated [4] , mainly due to a large number of asymptomatic or oligosymptomatic individuals and/or a limited availability of diagnostic testing, particularly in low-income countries [5e7] . According to a new analysis by the Institute for Health Metrics and Evaluation (IHME), COVID-19 has caused~12.2 million deathsdmore than twice the official numbers reported [4] .

Serological tests can be used to detect individuals with current or past infection with the SARS-CoV-2 virus. Such tests can be used to estimate the cumulative prevalence of SARS-CoV-2 infection and disease transmission over time [8] . Previous studies have shown that specific serum antibodies against SARS-CoV-2 can increase within 2e3 weeks following primary infection and remain detectable for 3e6 months after exposure [9e11] . Measuring the prevalence and levels of anti-SARS-CoV-2 serum antibodies in people can be helpful in prioritizing the vaccination of susceptible/ unexposed (i.e. seronegative) individuals [12] . Therefore, population-based serological screening at the national and regional levels can significantly assist health authorities to understand the toll of the epidemic, predict future spread and prioritize which people to vaccinate if/when vaccine supply is limited [12] .

In the early stages of the COVID-19 pandemic in 2020, some studies estimated the seroprevalences in different countries; however, only a few investigated seroprevalence across the globe (from early to mid-2020) [5, 13, 14] . More than 1 year on, it is now critical to re-assess the situation to be in an informed position about the global and regional seroprevalences, so that there is some understanding of the SARS-CoV-2 immune status at a time when people are being vaccinated. An informed position should enable the prioritization of vaccine deployment to communities and age/ risk groups [13] . Here, we extend our previous study [5] to provide a detailed update on global and regional SARS-CoV-2 seroprevalences around the world.

We conducted an updated systematic review and meta-analysis under the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [15] . Our protocol is registered (CRD42021238432) in PROSPERO. We searched three peerreviewed databases (i.e. PubMed, Embase and Scopus) and preprint servers (i.e. medRxiv, bioRxiv and SSRN) using predefined search terms for SARS-CoV-2 and seroprevalence (Fig. S1) . We also sourced studies from Google Scholar and the bibliographies in published works. Studies published between 1 January 2020 and 30 March 2021, without language or geographical restriction, were included. We only included population-based studies of SARS-CoV-2 seroprevalence in the general population. In addition to the exclusion criteria (Table S1 ), we did not consider studies of groups of people at a high risk of acquiring infection, including the 'homeless', those with household exposure to family members with confirmed COVID-19 and healthcare and migrant workers. We also excluded studies reporting the kinetics of anti-SARS-CoV-2 antibodies.

Two independent experts extracted data on study and sample characteristics and seroprevalence data from all of the eligible studies using a predefined form (cf. [5] ). The primary focus was the seroprevalence of SARS-CoV-2 in the 'general population'dwhich we defined as randomly selected people of different age groups, occupations, educational levels, ethnic backgrounds and socioeconomic status. The samples originated from people from households, communities, blood donors, living in defined geographical regions, whose COVID-19 status was unknown [5, 13] . Seroprevalence was defined as the number of people with specific anti-SARS-CoV-2 antibodies (IgG, IgM and/or IgA) at, or above, a designated threshold value divided by the total number of people screened for serum antibodies. We employed the cut-off point and seropositivity values defined by authors in peer-reviewed publications. We recorded the numbers of people who tested seropositive for IgG and/or IgM (as these were the antibody classes tested for in most eligible studies). If seropositivity for distinct antibody isotypes was reported, we extracted the numbers of people seropositive for specific IgG antibody only, as anti-SARS-CoV-2 IgG serum antibody persists for a longer period in serum than IgM or IgA [14, 16, 17] . To avoid repeated inclusion of sequential crosssectional studies, data for the total number of participants and seropositive people tested during the whole study period were extracted. For longitudinal studies, data were extracted only for the first blood collection. If a study used multiple serological assays, we extracted results for the assay with the highest diagnostic specificity and sensitivity.

When available, seroprevalence data, stratified according to age group, gender and ethnicity, were extracted from each individual study. Most studies categorized participants into groups of 19, 20e49, 50e64 and !65 years of age (model 1) or groups of 0e9, 10e19, 20e29, 30e39, 40e49,50e59, 60e69,70e79 and !80 years of age (model 2). Therefore, we extracted data for each of these categories for two distinct subgroup analyses. Countries and territories for which seroprevalence data were available were classified according to 'Sustainable Development Goal' (SDG) regions or subregions [18] , gross national income [19] and human development index (HDI) [20] .

To determine whether there was an association between seroprevalence rate and confirmed COVID-19 cases or deaths in a country, we extracted data on the total numbers of confirmed cases and deaths on the last date of the sampling period reported in each study [21] . We estimated the total numbers of people (i.e. females and males) exposed to SARS-CoV-2 in 2020 in particular geographical regions, as defined by the United Nations Population Division (UNPD) [22] , and worldwide. The risk of bias of studies included in the meta-analysis was assessed using the modified Joanna Briggs Institute (JBI) critical appraisal tool [23] .

All statistical analyses were performed using Stata (v.16 Stata Corp., College Station, TX, USA). To stabilize the variances, we first transformed the raw seroprevalence estimates using the FreemaneTukey double arcsine transformation [24] . Due to the intrinsic heterogeneity between epidemiological studies, we used the DerSimonian and Laird random-effects model (REM) to conservatively estimate the pooled seroprevalence of SARS-CoV-2 in the general population [25] . We calculated the pooled seroprevalences at 95% CIs using the 'metaprop' command in Stata. The heterogeneity between studies was assessed using Cochran's Q test and quantified using the I 2 statistic. An I 2 of >75% indicates substantial heterogeneity [26] . We also conducted a proportion meta-analysis with the empirical Bayes method, as it deals more adequately with heterogeneity than the classical random-effects model in situations with zero-event studies [27, 28] . We presented the pooled seroprevalence estimates with 95% credibility intervals.

Subgroup analyses, according to SDG regions and sub-regions, sex, age, ethnicity, place of residence, national income level, HDI, serological method (e.g. ELISA, lateral flow immunoassay (LFIA), chemiluminescence enzyme immunoassay (CLIA), etc.), type of assay (commercial kit or in-house assay) used and risk of bias, were conducted to explore the possible reasons for the observed heterogeneity between eligible studies. Corresponding prevalence ratios (PRs) were estimated for variables subjected to subgroup analysis. We also performed some subgroup analyses to assess the trend of SARS-CoV-2 seroprevalence over time (at intervals of 20e30 days) and at the start date of a COVID-19 epidemic within a country. To assess the effect of these variables on seroprevalence, we carried out random-effects meta-regression analyses using the 'metareg' command in STATA [29] . Further, we performed meta-regression analyses to assess whether seroprevalence was associated with the total number of confirmed cases or deaths in particular countries. The numbers of SARS-CoV-2-infected people (worldwide and in particular regions) were inferred by multiplying the pooled seroprevalence of SARS-CoV-2 by corresponding population size (in 2020)davailable via UNPD. Publication bias was assessed by logit transformation of effect size and sample size, instead of the inverse of the standard error, because the conventional funnel plot and publication bias tests for meta-analyses of proportion studies with low proportion outcomes are inaccurate [30] .

From January 2020 to March 2021, we identified 27 938 records from bibliographic databases, with 25 331 from peer-reviewed databases, 2429 from preprint servers and 178 from Google Scholar or article references. After removing duplicate records (n ¼ 4357) and irrelevant articles (n ¼ 22 701), 880 articles reporting SARS-CoV-2 seroprevalence were assessed for eligibility ( Fig. 1) . A total of 241 articles containing 275 datasets met the inclusion criteria for quantitative synthesis; these studies involved 6 367 734 people from 60 countries in seven SDG regions. Regions with the highest numbers of datasets were Europe and Northern America (n ¼ 163), Eastern and South-eastern Asia (n ¼ 32), and Latin America and the Caribbean (n ¼ 31). Detailed information on individual studies included is presented in Table S2 .

Of 6 367 734 people (represented in 275 datasets), 519 407 had specific serum antibodies to SARS-CoV-2. As the results of the Bayesian and REM analyses were similar (Table S3) , we focused on the REM analysis. The global SARS-CoV-2 seroprevalence (for 60 countries) was 9.47% (95% CI 8.99e9.95%), although heterogeneity among studies was substantial (I 2 ¼ 99.9%, p < 0.001). The extrapolation to the global population (in 2020) indicated that~738 million individuals (range: 700 752 407e775 582 474) were SARS-CoV-2 infected (up to December 2020; see Table 1 ).

According to SDG regions (Table 1) , the highest seroprevalence estimates were in Central and Southern Asia (22.91%, 19.11e26.72%), sub-Saharan Africa (18.76%, 13.09e24.42%) and Latin America and the Caribbean (18.29%, 16.59e19.99%); the lowest seroprevalence was in the Eastern and South-eastern Asia (1.62%, 1.31e1.95%). Seroprevalence estimates in Northern Africa and Western Asia and Europe and North America were 9.21% (3.72e14.68%) and 7.29% (6.58e8.01%), respectively. Only one study was available for Australia, suggesting a seroprevalence of 0.71% (0.51e0.98%).

In countries with three or more available studies, the highest seroprevalences were recorded in Pakistan (28.8%), Russia (27.4%), India (23.3%), Colombia (19.5%), Iran (16.9%), Kenya Fig. 2 shows the SARS-CoV-2 seroprevalence estimates for individual countries, and Table 2 ranks countries according to estimated total numbers of seropositive individuals. The funnel plot for pooled seroprevalence is shown in Fig. S2 ; this plot was symmetrical, indicating there was no publication bias in the studies included.

Of the 275 datasets selected, 114 datasets allowed pooled seroprevalences to be estimated for male and female individuals. Of the 1 142 427 males and 1 260 994 females, 52 831 males (7.73%, 7.19e8.26%) and 46 972 females (7.43%, 6.99e7.88), respectively, had specific serum antibodies against SARS-CoV-2. A higher seroprevalence was observed in males than in females (PR, 1.24; 95% CI 1.22e1.25) ( Table 3) .

Seroprevalence data were available for 45 and 38 datasets for subgroup analysis of age groups using models 1 and 2, respectively. Using model 1, subgroup analyses revealed pooled seroprevalences of 9.01% (7.22e10.79%), 6.49% (5.51e7.49%), 8.58% (7.31e9.86%) and 4.49% (3.68e5.31%) for people of 19, 20e49, 50e64 and !65 years of age, respectively ( Table 2) . Using model 2, the highest and lowest seroprevalence estimates were estimated for people of 30e39 (11.94%, 10.18e13.71%) and >80 (3.46%, 2.22e4.71%) years of age, respectively (Table 3) .

A range of serological assays were used in studies linked to the 275 datasets. ELISA was linked to 104 datasets, whereas CLIA, rapid LFIA, virus neutralization assay and other serological methods (e.g. immunofluorescence assay, microsphere immunoassay, flow cytometry assay, serum epitope repertoire analysis and coronavirus antigen microarray) were linked to 86, 62, 12 and 11 datasets, respectively. Commercial kits and in-house serological methods were associated with 231 and 44 datasets, respectively (Table S2  and Table 3 ). Subgroup analysis showed that the highest and lowest seroprevalences were estimated using ELISA (12.12%, 10.78e13.46%) and virus neutralization (0.94%, 0.63e1.26%), respectively. Seroprevalences estimated using LFIA (8.42%, 7.71e9.12%), CLIA (8.45%, 7.39e9.51%) and other serological methods (8.15%, 5.24e11.07%) were almost similar. Moreover, subgroup analysis indicated pooled seroprevalences of 10.01% (9.47e10.54%) using commercial kits and 6.36% (5.56e7.17%) for inhouse assays (Table 3) .

Seroprevalence data associated with ethnicity were available from 29 datasets. Subgroup analysis of these ethnicity data revealed pooled seroprevalences of 4.05% (3.86e4.23%), 3.32% (3.21e3.44%), 2.69% (2.57e2.81%) and 1.92% (1.91e1.94%) in people of Black, Hispanic, Asian/other and White ethnic backgrounds, respectively ( Table 2 ). People of Black (PR 2.78, 2.68e2.88), Hispanic (PR 2.05, 1.99e2.11) and Asian/other minority ethnicities (PR 1.64, 1.58e1.69) showed a significantly higher risk of SARS-CoV-2 infection than White people (Table 3) .

Subgroup analysis according to income level showed that the highest and lowest seroprevalences were in countries with lower middle (21.61%, 17.57e25.65%) and high 6.54% (5.87e7.22%) income levels, respectively (Table 3 ). Subgroup analysis ( Table 2) according to HDI level indicated that countries with medium (22.56%, 18.39e26.73%) and low (18.03%, 10.04e26.02%) HDI had higher seroprevalences than countries with high (9.88%, 9.38e10.37%) and very high (7.27%, 6.61e7.93%) HDI. Random-effects meta-regression analyses showed a decreasing trend in seroprevalence with higher income levels (coefficient (C) ¼ e1.65 Â 10 À6 ; p < 0.001), and HDI (C ¼ e0.4001; p < 0.001) (Figs. 3A,B) .

Critical appraisal using the JBI showed that 86 datasets had a low risk of bias (score 7e9/9), 113 datasets had a moderate (4e6/9) and 78 studies had a high risk of bias ( 3/9) . Moreover, the seroprevalences for studies with a low, moderate, and high risk of biases were 6.56% (5.78e7.34%), 10.31% (9.59e11.02%) and 10.39% (9.17e11.62%), respectively (Table 3) .

With reference to the start date of a COVID-19 epidemic in a country (in months), subgroup analysis (Table S4) showed seroprevalences of 1.73% (1.33e2.14%), 8.65% (7.79e9.51%), 11.04% (10.02e12.06%) and 14.15% (12.36e15.93%) in December 2019, January 2020, February 2020 and March 2020, respectively. Subgroup analysis of data at the beginning date of sampling showed an increasing trend of seroprevalence estimates on a monthly basis (Table S4 ). Subgroup and meta-regression analyses were also conducted to explore SARS-CoV-2 seroprevalence over timedfrom the beginning of the pandemic to the first and last times of sampling/ testing in individual studies. The results indicated increasing seroprevalence estimates over time, as the highest seroprevalences were recorded 7e10 months after the epidemic commenced in a particular country (Table S4) . Random-effects meta-regression analysis showed a significant, increasing trend in seroprevalence in a country from the beginning of a COVID-19 epidemic to the first (C ¼ 0.0013; p < 0.001) and to the last (C ¼ 0.0004; p < 0.001) day of sampling (i.e. serum collection) (Figs. 4A,B) .

We counted the numbers of confirmed cases and deaths in individual countries in WHO situation reports [31] . Subgroup analyses of the data showed that the lowest seroprevalences were observed when the confirmed cases (4.66%, 3.59e5.73%) and total deaths (6.38%, 5.36e7.41%) were lower than 10 000 and 1000 cases, respectively (Table S5) . Moreover, the highest seroprevalences were observed when the confirmed cases (19.11%, 15.77e22.44%) and total deaths (14.17%, 12.28e16.06%) were between 500 000e1 000 000 and 20 000e40 000 cases, respectively (Table S5) . Metaregression analyses indicated a non-significant, increasing trend in the number of confirmed cases (C ¼ 7.09 Â 10 À9 ; p 0.08) with increasing seroprevalence. Similarly, a non-significant, increasing trend was found in relation to the total number of deaths (C ¼ 1.33 Â 10 À7 ; p ¼ 0.36) (Fig. S2A,B) .

This meta-analysis provides a comprehensive update on the SARS-CoV-2 seroprevalence regionally and internationally. The pooled global seroprevalence was estimated at 9.47% (95% CI 8.99e9.95%), equating to~738 million (700e775 million) people worldwide, which is relatively consistent with previous seroprevalence studies [13, 14] , bearing in mind that the true prevalence of infection appears to be 6e11 times greater than the number of confirmed cases reported officially by countries [32e34]. The seroprevalence estimates here varied considerably between SDG regions and sub-regions, with the highest SARS-CoV-2 seroprevalences in southern Asia, Latin America and the Caribbean and sub-Saharan Africa. Living in overcrowded conditions, higher rates of co-morbidities and an inadequate or lack of access to medical care likely increase the vulnerability of people in developing countries to SARS-CoV-2 and other respiratory infections [35] . In addition, poor infrastructure and poverty render preventive measures (including detection of people with active infection, quarantine and reducing public transport during the daytime) more difficult [35, 36] .

In accord with other studies [5, 13, 14] , the present results showed a higher SARS-CoV-2 seroprevalence in males than in females, which could be attributed to more outdoor activities in remote areas and community exposure for males, particularly in developing countries [37] . Our findings also indicate significant differences in SARS-CoV-2 seroprevalence between age groups, with seroprevalence decreasing with age for people older than 65 years. In accordance with a previous study [5] , people of <19 years (children and adolescents) had similar seroprevalences to individuals aged 20e64 years, in contrast to other meta-analyses of global SARS-CoV-2 seroprevalence [13, 14] , indicating lower seroprevalence estimates for people of <19 years of age. A possible reason for this difference could be the exclusion of high-risk populations in the present study. Children are socially active and have more physical contact with others, especially when playing with other children or families. Thus, mandating social distancing is more difficult for them. Our results suggest that children might have the same level of exposure to infection as adults, but are less likely to develop symptoms and to be admitted to hospital [21, 38, 39] . A higher SARS-CoV-2 seroprevalence rate in adults of 20e64 years of age than in older people could be explained by a greater involvement in community activities [14,40e42] .

Consistent with some previous studies [5,13,43e45] , minority ethnic groups are at a high risk of acquiring SARS-CoV-2 infection, which is supported by findings from the REACT-2 and OpenSAFELY studies in the UK, showing higher levels of SARS-CoV-2 serum antibodies and hospitalization in minority groups than people of White ethnicity [43, 46] . Possible explanations might include discrimination or difficulties in accessing healthcare, housing, education and financial status; communication and language barriers; cultural practices; lack of health insurance; more ethnic minority groups employed in essential work settings, such as healthcare facilities, farms, factories, grocery stores and public transport; and living in large families and/or overcrowded conditions [5] .

The SARS-CoV-2 seroprevalences estimated herein may not be entirely accurate because of limitations or characteristics of the studies included in this investigation. First, a notable number of studies did not apply rigorous (e.g. multistage cluster or stratified) sampling strategies and did not always include a representative population. Second, several serological assays with differing test performances (specificities and sensitivities) and cut-off values were used to test samples. However, few studies have independently validated the specificity and sensitivity of the used diagnostic kits prior to the serological testing of large numbers of serum samples. Despite WHO recommendations, the seroprevalence estimates reported in many studies included did not adjust for the demographic structure of the target population. Finally, as it is impractical, we did not perform inverse probability weighting using population weights to adjust for unequal probability of sampling [47, 48] . These limitations can make comparisons between/ among studies challenging, and might explain heterogeneity among studies. Other limitations (including different and timevarying sensitivities and specificities of serological methods; missing studies published in un-indexed, local journals; a lack of data for two-thirds of countries of the world) may also have an effect and has been discussed elsewhere [5, 13, 14] . 

Antibody-dependent enhancement and SARS-CoV-2 vaccines and therapies

Public health interventions for COVID-19: emerging evidence and implications for an evolving public health crisis

COVID-19 Weekly Epidemiological Update-24

Institute for Health Metrics and Evaluation (IHME)

SARS-CoV-2 seroprevalence worldwide: a systematic review and metaanalysis

Prevalence of SARS-CoV-2 in Spain (ENE-COVID): a nationwide, population-based seroepidemiological study

Occurrence and transmission potential of asymptomatic and presymptomatic SARS-CoV-2 infections: a living systematic review and meta-analysis

Population-based age-stratified seroepidemiological investigation protocol for COVID-19 virus infection

Seroprevalence of anti-SARS-CoV-2 antibodies in COVID-19 patients and healthy volunteers up to 6 months post disease onset

Persistence and decay of human antibody responses to the receptor binding domain of SARS-CoV-2 spike protein in COVID-19 patients

Persistence of serum and saliva antibody responses to SARS-CoV-2 spike antigens in COVID-19 patients

Significance of SARS-CoV-2 specific antibody testing during COVID-19 vaccine allocation

Global seroprevalence of SARS-CoV-2 antibodies: a systematic review and meta-analysis

Serological evidence of human infection with SARS-CoV-2: a systematic review and meta-analysis

Preferred reporting items for systematic review and meta-analysis protocols (PRISMA-P) 2015 statement

Dynamics of anti-SARS-Cov-2 IgM and IgG antibodies among COVID-19 patients

The duration, dynamics and determinants of SARS-CoV-2 antibody responses in individual healthcare workers

The sustainable development goals (SDGs) report 2019. Regional groupings

Gross national income per capita

United Nations Development Program

COVID-19 weekly epidemiological update -29

Total population (both sexes combined) by region, subregion and country, annually for 1950e2100 (thousands)

Methodological guidance for systematic reviews of observational epidemiological studies reporting prevalence and cumulative incidence data

Transformations related to the angular and the square root

Meta-analysis in clinical trials

Quantifying heterogeneity in a meta-analysis

Predicting the extent of heterogeneity in meta-analysis, using empirical data from the Cochrane Database of Systematic Reviews

Bayesian inference in ecology

Meta-regression in Stata

In metaanalyses of proportion studies, funnel plots were found to be an inaccurate method of assessing publication bias

World Health Organization (WHO)

Population point prevalence of SARS-CoV-2 infection based on a statewide random sampledIndiana

Robust estimates of the true (population) infection rate for COVID-19: a backcasting approach

Comparison of seroprevalence of SARS-CoV-2 infections with cumulative and imputed COVID-19 cases: systematic review

Community mitigation during SARS-CoV-2 pandemic: mission impossible in developing countries

Estimation of potential global pandemic influenza mortality on the basis of vital registry data from the 1918e20 pandemic: a quantitative analysis

COVID-19: cultural predictors of gender differences in global prevalence patterns

COVID-19 transmission and children: the child is not to blame

Characterisation of COVID-19 pandemic in paediatric age group: a systematic review

Children are unlikely to be the main drivers of the COVID-19 pandemic e a systematic review

On determining the age distribution of COVID-19 Pandemic

Age-dependent effects in the transmission and control of COVID-19 epidemics

Ethnic differences in SARS-CoV-2 infection and COVID-19-related hospitalisation, intensive care unit admission, and death in 17 million adults in England: an observational cohort study using the OpenSAFELY platform

Racial and/or ethnic and socioeconomic disparities of SARS-CoV-2 infection among children

Racial and ethnic disparities in SARS-CoV-2 pandemic: analysis of a COVID-19 observational registry for a diverse US metropolitan population

SARS-CoV-2 antibody prevalence in England following the first peak of the pandemic

Inverse probability weighting

Reflection on modern methods: demystifying robust standard errors for epidemiologists

Immunological characteristics govern the transition of COVID-19 to endemicity

Prevalence of antibody positivity to SARS-CoV-2 following the first peakof infection in England: Serial cross-sectional studies of 365,000 adults

COVID-19 results briefing: global. Seattle, USA: IHME, University of Washington

Sincere thanks to Malihe Nourollahpour Shiadeh for critically reviewing the earlier version of the manuscript.

Supplementary data to this article can be found online at https://doi.org/10.1016/j.cmi.2021.09.019.

Although not accounted for here, sero-reversion can lead to a classification challenge (infected vs. non-infected), a distortion of epidemiological estimates and/or possible shifts in susceptibility of people to infection in subsequent 'waves' of COVID- 19 . It has been shown that infection-blocking immunity wanes rapidly, but that disease-reducing immunity is long-lived [49] . A real-time assessment of community transmission (REACT-2) study involving 365 104 people in the UK, and conducted over three phases of testing, showed that anti-SARS-CoV-2 immunity waned over time; serum antibody prevalence declined from 6% to 4.4% between 20 June and 28 September 2020 [50] . Another point is that the present study was conducted before the emergence of new SARS-CoV-2 variants/lineages, such as B.1.352, P.1, B.1.17 and B.1.617; infections with new variants are likely to have spread in recent months and require rigorous monitoring, as some (e.g. B.1.617) are markedly more transmissible (60%) than the 'original virus' [51] . Moreover, recent analysis by IHME [4] estimated that 32% of people globally were infected since 23 August 2021. If we consider that there are ten undetected people per confirmed case,~2140 million individuals (~27.5%) of the world's population have been infected since this date. Our lower estimate (27.5% vs. 32%) might be explained by a higher community transmission of new variants (delta and lambda) from December 2020 to August 2021, particularly in countries such as Brazil, India, Iran and Peru [51] .The present, updated meta-analysis reveals a higher SARS-CoV-2 seroprevalence in countries with low-and lower middle-income levels, emphasizing the need to accelerate vaccination 'roll-out' in developing countries. The high risk of SARS-CoV-2 infection in Black, Hispanic, Asian and minority ethnicities emphasizes that vaccine allocation to these groups of people needs to be a priority. For future seroprevalence investigations, we recommend improved study designs, consistent with WHO protocols [8] , which would reduce heterogeneity among investigations, and allow for enhanced seroprevalence estimates, meta-analyses, interpretations and policy decisions. Given the pace of work on COVID-19 and the rapid emergence and spread of the delta, kappa and lambda variants of SARS-CoV-2, we refer to recent seroprevalence surveys (see Table S6 ), published while this paper was under review (i.e. 30 March 2021 to 26 August 2021). Clearly, seroprevalence rates have increased markedly in countries including India (54.2%), Kenya (44.2%), Poland (35.5%), Jordan (34.2%), Greece (26.3%), Brazil (14.8%), United States (14.5%), Portugal (13.1%), Croatia (11.1%) and England (9.8%).

The authors declare no conflict of interest. This study was supported by the Health Research Institute at the Babol University of Medical Sciences, Babol, Iran (IR.MUBABOL.REC.1399.304).