Highlights Thirty-three blood donor SARS-CoV-2 seroprevaelnce studies world-wide between January 2020 and January 2021 were identified. The pre-vaccine reported seroprevalence was generally less than 10%. It is difficult to distinguish true variability in seroprevalence from variability due to study designs and methodology. INTRODUCTION

As health authorities contend with the unrelenting coronavirus disease 2019 (COVID-19) pandemic, resources continue to be invested in tracking population-level exposure to severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2). Case detection can be used to monitor infection rates but may underestimate prevalence by limited testing capacity; the restricted time period SARS-CoV-2 is detectable by diagnostic tests; and a significant proportion of mild and asymptomatic cases that do not seek testing. In contrast, serological tests that identify SARS-CoV-2-specific antibodies are commonly used for surveillance studies, overcoming the limitations of relying on case detection alone.

Given the unprecedented urgency to evaluate the burden of COVID-19, SARS-CoV-2 seroprevalence studies were mobilized quickly. While in theory, random sampling from the general population (e.g., population-based seroprevalence studies) should yield the most generalizable results, this approach is both time-consuming and expensive. Additionally, time-varying response rate may lead to a complex selection bias. In contrast, populations of blood donors have increasingly been recognized as an informative and cost-effective strategy to monitor epidemics [1]. Blood services have the operational capacity to sample and test large proportions of the healthy population for surveillance purposes [2-4]. The Surveillance Risk Assessment and Policy sub-group of the Transfusion Transmitted Infectious Diseases Working Party (TTIDW) of the International Society of Blood Transfusion (ISBT) recently published that 73% (32/48) of blood operators surveyed worldwide were undertaking or planning to conduct seroprevalence studies to inform public health [5].

Methodological challenges have emerged unique to this pandemic. The validity, interpretability, and ability to pool seroprevalence studies are limited by study designs, sampling strategies, study timing, the variability in assay characteristics, and antibody kinetics. Seroprevalence studies, including those among blood donors, are compiled by on-line dashboards such as SeroTracker (https://serotracker.com) [6], and editorials and perspective articles have been published on limitations of seroprevalence studies in general [1, 7, 8], but to our knowledge there has not been an attempt to systematically bridge these two elements together to provide epidemiological guidance for future research. In this scoping review, we summarized studies conducted specifically among blood donors to characterize SARS-CoV-2 seroprevalence studies, evaluate how well subpopulations and geographic areas have been represented, and determine the diversity of methodology used to address limitations associated with these studies.

MATERIALS AND METHODS

This scoping review was conducted and supplemented with publicly available data on cumulative COVID-19 cases and social distancing policies nationally. Additionally, members of the ISBT TTIDWP (representing blood collectors from Canada, United States, Denmark, the Netherlands, and Australia) provided expert opinions. Findings were reported according to the Preferred Reporting Items for Systematic Reviews and Meta-Analysis extension for Scoping Reviews [9].

Two reviewers (S.S. and S.U.) independently searched articles using the search engines PubMed and medRxiv for SARS-CoV-2 seroprevalence studies in English among blood donor populations from 1 January 2020 to 19 January 2021. Search terms were [SARS-CoV-2] AND/OR [COVID-19] AND [seroprevalence] AND [donor] OR [blood donor]. A seroprevalence survey was defined as the serological testing of residual blood from blood donations over a restricted time period, to estimate the prevalence of SARS-CoV-2 antibodies in a specified population. Therefore, only studies that reported the sample size, sampling dates, and prevalence estimates (or the number of reactive samples) were included in this review. They excluded studies that used residual blood from convalescent plasma donors or as negative controls to evaluate assay performance. Seroprevalence estimates from the grey literature were not included in this review since methods are not routinely reported.

Articles were screened on titles and abstracts by the same two authors. The full-text assessment and data extraction were performed by one reviewer per article and subsequently checked for accuracy by a second reviewer. Consensus was reached between the authors when discrepancies arose. Data were entered into Microsoft Excel using a standardized form, which included: The full reference, region, data of sample collection, data necessary to calculate unadjusted SARS-CoV-2 seroprevalence (the number of samples tested and reactivity). When available, the adjusted seroprevalence estimate(s) and 95% confidence intervals or a range of estimates were extracted. The authors a priori identified specific challenges of seroprevalence studies (graphically represented as Figure 1): Population sampling: What was the scope of the study, national or regional? Were blood donor populations characterized? Were seroprevalence rates stratified by age, sex, socioeconomic status (SES) or by specific regions? Was the seroprevalence estimate standardized to population-level characteristics? Dynamic epidemic: What was the type of study design (single or serial cross-sectional) to evaluate temporal trends? Assay characteristics: Was the assay reported? Was the assay commercial or an in-house assay? Were the seroprevalence estimates adjusted for imperfect assay characteristics? And how? Antibody kinetics: Were estimates adjusted for waning antibody titres? Overlapping determinants of estimating population-level seroprevalence. (a) Population: The WHO endorses seroprevalence studies of blood donors because they can represent a convenient sample of a healthy adult population, but care should be taken when generalizing the results beyond the target population. Summarizing prevalence rates may miss significant differences since infection rates are likely differential by geographic regions, socioeconomic status, age and racialized populations within a country. There may also be a potential for selection bias as donors are a self-selected group of individuals. (b) Shifting public policy: Seroprevalence rates can be influenced by changing population-level trends, defined as surges of new cases followed by a downturn (epidemic waves). (c) Antibody kinetics: From the time of infection, on average, it takes 10–28 days to develop specific immunoglobulin G (IgG) antibodies to severe acute respiratory syndrome coronavirus 2. And by approximately 100 days, the level of these antibodies detectable in the blood begins to decrease (wane). If sampling occurs outside this window of detection, people both early and later in their infection will be missed. (d) Accuracy of tests: Assay performance is measured by the proportion of test results that accurately identified those having antibodies ‘sensitivity’ and not having antibodies ‘specificity’, at a given threshold (signal-to-cut-off ratio). Vaccines: Distinguishing between natural infections and vaccine-induced immunity is an additional challenge when estimating seroprevalence rates [Colour figure can be viewed at wileyonlinelibrary.com]

We extracted cumulative case counts from the World Health Organization Coronavirus Disease (COVID-19 Dashboard; https://covid19.who.int/). We estimated the ratio of reported to expected infections. Since seroprevalence data reflects infections that occurred prior to the date of measurement, case detection was extracted 2 weeks from the end of the donation collection/study period for each study [10].

Public health policies were summarized using the Government Stringency Index as the average of the daily index 2 weeks before the beginning and end of the study period [11]. This composite measure was based on nine response indicators including school closures, workplace closures and travel bans, rescaled to a value from 0 to 100 (100 = strictest). If policies varied at the subnational level, the index is shown as the response level of the strictest sub-region.

We evaluated the correlation between seroprevalence and cumulative incidence by linear regression.

RESULTS

From 157 articles (32 peer-reviewed and 125 preprints), 33 studies (22 peer-reviewed and 11 preprints) were included in this review (Figure 2). The 33 studies represented 20 countries, the majority from Europe (n = 8) followed by North America (n = 4), Asia (n = 4), Africa (n = 2), South America (n = 1) and Australia (n = 1) (Figure 3). Seroprevalence studies from low- and middle-income countries were limited.

Inclusion of studies in review

Worldwide coverage of seroprevalence studies among blood donor populations. Shaded countries are represented in this review

Table 1 characterizes the data extracted from each of the 33 studies included in this review [12-44]. Of the published studies, the vast majority (91%; 20/22) were published in clinical or public health journals, two were specifically published in transfusion medicine journals. The median sample size was 1996 donations but ranged from as many as 953,926 in the United States [28] to as few as 22 in Libya [41]. The time frames for studies ranged from beginning 1 January 2020 to ending 11 December 2020. Nine (27%) studies began between January and 29 February 2020; one study began in September 2020.

TABLE 1. Summary of studies included in this review First author [ref] Type of publication Location National versus regional Sample size Date of sample Type of study Age Seroprevalence estimate % (95% CI) Comparison of reported and seroprevalence-derived expected number of infections Population Policy indexa Infections 14 days prior to end of the study (cumulative cases) Seroprevalence-derived expected number of infections Seroprevalence-derived expected number of infections (lower range) Seroprevalence-derived expected number of infections (upper range) Ratio of reported infections to seroprevalence-derived number of infections Europe Erikstrup [12] Clinical Denmark National 20,640 6 April to 3 May 2020 Serial CS 17–69 1.90 (0.80, 2.30) 5,792,202 71.45 7384 110,052 46,338 133,221 0.067 Iversen [13] Clinical Denmark Regional 4672 15 April and 23 April 2020 Single CS 18–64 3.04 (2.58, 3.57) 5,792,202 NA 5819b 176,083 149,439 206,782 0.033b Pedersen [14] Clinical Denmark National 1201 2 June to 19 June 2020 Single CS 70+ 1.40 (0.30, 2.50) 5,792,202 59.37 11,771 81,091 17,377 144,805 0.145 Gallian [15] Clinical France Regional 998 March to April 2020 Single CS 19–64 2.70 65,273,511 NA 102,534b 1,762,385 0 0 0.058b Fischer [16] Public health Germany Regional 3186 9 March to 3 June 2020 Single CS NA 0.91 (0.58, 1.24) 83,783,942 NA 175,210b 762,434 485,947 1,038,921 0.230b Percivalle [17] Public health Italy Regional 390 18 March to 6 April 2020 Single CS 19–70 23.00 60,461,826 NA 53,578b 13,906,220 0 0 0.004b Valenti [18] medRxiv Italy Regional 789 24 February to 8 April 2020 Serial CS 18–70 5.20 (2.40, 9.00) 60,461,826 NA 74,386b 3,144,015 1,451,084 5,441,564 0.024b Fiore [19] Clinical Italy Regional 904 1 May to 31 May 2020 Single CS 18–65 0.99 60,461,826 NA 225,886b 598,572 0 0 0.377b Slot [20] Clinical The Netherlands National 7361 1 April to 20 May 2020 Serial CS 18–72 2.70 (1.60, 5.40) 17,134,872 77.65 40,841 462,642 274,158 925,283 0.088 Olariu [21] Clinical Romania Regional 2115 8 July to 1 September 2020 Single CS 18–65 1.51 (1.07, 2.13) 19,237,691 NA 71,194b 2,904,891 205,843 409,763 0.025b Thompson [22] Public health Scotland National 3500 17 March to 18 May 2020 Serial CS Median 47 3.00 (0.00, 14.00) 5,460,000 60.65 12,226 163,800 0 764,400 0.075 Dopico [23] medRxiv Sweden Regional 2100 14 March to 11 December 2020 Serial CS NA 14.80 (12.20, 18.00) 10,099,265 NA 231,554b 1,494,691 1,232,110 1,817,868 0.155b North America Saeed [24] Transfusion Canadac National 74,642 9 May to 21 July 2020 Serial CS 18–65+ 0.70 (0.63, 0.76) 29,247,654 70.82 50,169 204,734 184,260 222,282 0.245 Martinez-Acuña [25] medRxiv Mexico Regional 1968 1 January to 30 August 2020 Serial CS 18–65 3.99 128,932,753 NA 505,751b 5,144,417 0 0 0.098b Villarreal [26] medRxiv Panama Regional 255 30 April to 7 July 2020 Single CS 20–79 11.72 (8.30, 16.30) 4,314,767 NA 24,274b 505,691 358,126 703,307 0.048b Basavaraju [27] Clinical United States Regional 5477 30 December 2019 to 17 January 2020 Serial CS 16–95 0.86 82,375,418 NA 0 708,429 0 0 0 Dodd [28] Clinical United States National 953,926 15 June to 23 August 2020 Serial CS 16–55+ 1.82 (1.79, 1.84) 331,002,651 69.11 4,999,815 6,024,248 5,924,947 6,090,449 0.830 Vassallo [29] medRxiv United States National 189,656 1 June to 31 July 2020 Single CS 16–85+ 1.55 331,002,651 70.71 3,405,494 5,130,541 0 0 0.664 Kamath [30] medRxiv United States-New York Regional 1559 March to July 2020 Serial CS 17–80 11.60 (6.00, 21.20) 20,201,249 NA 428,477 2,343,345 1,212,075 4,282,665 0.183 Jin [31] medRxiv United States-New York City Regional 1000 16 June to 15 July 2020 Single CS 16–78 10.00 (9.10, 14.27) 8,230,290 NA 212,072 823,029 748,956 1,174,462 0.258 Nesbitt [32] medRxiv United States-Rhode Island Regional 1996 27 April to 11 May 2020 Single CS Median 56 0.60 (0.20, 1.10) 1,097,379 NA 7997 6584 2195 12,071 1215 Ng [33] Clinical United States-San Francisco Regional 1000 March 2020 Single CS 19–89 0.40 (0.01, 0.79) 883,255 NA 51 3533 88 6978 0.014 Asia Chang [34] medRxiv China-Wuhan, Shenzhen, Shijiazhuang Regional 17,794 January to April 2020 Serial CS 18+ 2.29 (2.08, 2.52) 1,439,323,776 NA 83,071b 32,960,514 29,937,935 36,270,959 0.003b Xu [35] Clinical China-Guangzhou Regional 2199 23 March to 2 April 2020 Single CS 18–59 0.09 1,439,323,776 NA 81,178b 1,295,391 0 0 0.063b Sughayer [36] medRxiv