ABSTRACT
BACKGROUND Debate about the level of asymptomatic severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection continues. The amount of evidence is increasing and study designs have changed over time. We updated a living systematic review to address three questions: (1) Amongst people who become infected with SARS-CoV-2, what proportion does not experience symptoms at all during their infection? (2) What is the infectiousness of asymptomatic and presymptomatic, compared with symptomatic, SARS-CoV-2 infection? (3) What proportion of SARS-CoV-2 transmission in a population is accounted for by people who are asymptomatic or presymptomatic?
METHODS AND FINDINGS The protocol was first published on 1 April 2020 and last updated on 18 June 2021. We searched PubMed, Embase, bioRxiv and medRxiv, aggregated in a database of SARS-CoV-2 literature, most recently on 6 July 2021. Studies of people with PCR-diagnosed SARS-CoV-2, which documented symptom status at the beginning and end of follow-up, or mathematical modelling studies were included. Studies restricted to people already diagnosed, of single individuals or families, or without sufficient follow-up were excluded. One reviewer extracted data and a second verified the extraction, with disagreement resolved by discussion or a third reviewer. Risk of bias in empirical studies was assessed with a bespoke checklist and modelling studies with a published checklist. All data syntheses were done using random effects models. Review question (1): We included 130 studies.
Heterogeneity was high so we did not estimate a mean proportion of asymptomatic infections overall (interquartile range 14-50%, prediction interval 2-90%), or in 84 studies based on screening of defined populations (interquartile range 20-65%, prediction interval 4-94%). In 46 studies based on contact or outbreak investigations, the summary proportion asymptomatic was 19% (95% CI 15-25%, prediction interval 2-70%). (2) The secondary attack rate in contacts of people with asymptomatic infection compared with symptomatic infection was 0.32 (95% CI 0.16-0.64, prediction interval 0.11-0-95, 8 studies). (3) In 13 modelling studies fit to data, the proportion of all SARS-CoV-2 transmission from presymptomatic individuals was higher than from asymptomatic individuals. Limitations of the evidence include high heterogeneity and high risks of selection and information bias in studies that were not designed to measure persistently asymptomatic infection, and limited information about variants of concern or in people who have been vaccinated.
CONCLUSIONS Based on studies published up to July 2021, most SARS-CoV-2 infections were not persistently asymptomatic and asymptomatic infections were less infectious than symptomatic infections. Summary estimates from meta-analysis may be misleading when variability between studies is extreme and prediction intervals should be presented. Future studies should determine the asymptomatic proportion of SARS-CoV-2 infections caused by variants of concern and in people with immunity following vaccination or previous infection. Without prospective longitudinal studies with methods that minimise selection and measurement biases, further updates with the study types included in this living systematic review are unlikely to be able to provide a reliable summary estimate of the proportion of asymptomatic infections caused by SARS-CoV-2.
REVIEW PROTOCOL Open Science Framework (https://osf.io/9ewys/)
Why was this study done?
▪ The proportion of people who will remain asymptomatic throughout the course of infection with severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), the cause of coronavirus disease 2019 (covid-19), is debated.
▪ Studies that assess people at just one time point overestimate the proportion of true asymptomatic infection because those who go on to develop covid-19 symptoms will be wrongly classified as asymptomatic, but other types of study might underestimate the proportion if, for example, people with symptoms are more likely to be included in a study population.
▪ The number of published studies about SARS-CoV-2 is increasing continuously, types of studies are changing and, since 2021, vaccines have become available, and variants of concern have emerged.
What did the researchers do and find?
▪ We updated a living systematic review through 6 July 2021, using automated workflows that speed up the review processes, and allow the review to be updated when relevant new evidence becomes available.
▪ In 130 studies, we found an interquartile range of 14-50% (prediction interval 2-90%) of people with SARS-CoV-2 infection that was persistently asymptomatic; owing to heterogeneity, we did not estimate a summary proportion.
▪ Contacts of people with asymptomatic SARS-CoV-2 infection are less likely to become infected than contacts of people with symptomatic infection (risk ratio 0.38, 95% CI 0.16-0.64, prediction interval 0.11-0.95, 8 studies).
What do these findings mean?
▪ Up to mid-2021, most people with SARS-CoV-2 were not persistently asymptomatic and asymptomatic infection was less infectious than symptomatic infection.
▪ In the presence of high between-study variability, summary estimates from meta-analysis may be misleading and prediction intervals should be presented.
▪ Future studies about asymptomatic SARS-CoV-2 infections caused by variants of concern and in people with immunity following vaccination or previous infection should be specifically designed, using methods to minimise biases in the selection of study participants and in ascertainment, classification and follow-up of symptom status.
Introduction
There is ongoing debate about the true proportion of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection that remains asymptomatic [1]. A well-recognised source of overestimation arises when people without symptoms at the time of testing are reported as having asymptomatic infection, with such cross-sectional studies often reporting percentages of 80% or more [2,3]. These studies overestimate the proportion of persistently asymptomatic infection because they misclassify people with so-called presymptomatic infection, who will develop symptoms of coronavirus disease 2019 (COVID-19) if reassessed after an adequate follow-up period [1]. Other sources of bias can result in over- or underestimation of the proportion with persistent asymptomatic infections, even when participants are adequately followed up [1]. For example, studies that assess a limited range of symptoms could overestimate the proportion asymptomatic through misclassification if they do not ask participants about all possible symptoms. Since COVID-19 was first identified as a viral pneumonia, the spectrum of symptoms has grown to include gastro-intestinal symptoms and disturbances of smell and taste [1]. On the other hand, selection bias would be expected to underestimate the proportion with asymptomatic SARS-CoV-2 if people with symptoms are more likely to be tested for SARS-CoV-2 infection than those without symptoms [4].
Accurate estimates of the proportions of true asymptomatic and presymptomatic infections are needed to determine the balance and range of control measures [5]. Recognition of asymptomatic and presymptomatic infections showed the importance of control measures such as physical distancing, active case-finding through testing of asymptomatic people [6] and the need for rapid quarantine [7] in the first waves. Since late 2020, vaccines have become available [8] and several SARS-CoV-2 variants of concern have spread internationally, with varying viral characteristics [9]. The number of published studies about SARS-CoV-2 is also increasing continuously and the types of published studies are also changing [10], including the designs of studies about asymptomatic infection. In systematic reviews of studies published to April 2021, reported point estimates from random effects meta-analysis models range from 17 to 41% [11-16]. Authors of these reviews typically report values of the I2 statistic >90 [17,18], but heterogeneity is often not explored in detail and prediction intervals, which give information about sampling error and variability between studies, are recommended but rarely reported[17,19,20]. In this fifth update of our living systematic review [21] we aimed to improve and understand the changing evidence over time for three review questions: (1) Amongst people who become infected with SARS-CoV-2, what proportion does not experience symptoms at all during their infection? (2) What is the infectiousness of people with asymptomatic and presymptomatic, compared with symptomatic SARS-CoV-2 infection? (3) What proportion of SARS-CoV-2 transmission is accounted for by people who are either asymptomatic throughout infection, or presymptomatic?
Methods
We conducted an update of a living systematic review, a systematic review that provides an online summary of findings and is updated when relevant new evidence becomes available [22]. The protocol, which describes modifications for each update, was first published on 1 April 2020 and amended for this version on 18 June 2021, (https://osf.io/9ewys/). Previous versions have been posted as preprints [21,23] and published as a peer-reviewed article [12]. We report our findings according to statements on preferred reporting items for systematic reviews and meta-analyses 2020 (S1 PRISMA 2020 Checklist) [24] and on synthesis without meta-analysis in systematic reviews (SWiM) [25]. Ethics committee review was not required for this review. Box 1 shows our definitions of symptoms, asymptomatic infection and presymptomatic status.
Definitions of symptoms and symptom status in a person with SARS-CoV-2 infections
Symptoms
symptoms that a person experiences and reports. We used the authors’ definitions. We searched included manuscripts for an explicit statement that the study participant did not report symptoms that they experienced. Some authors defined ‘asymptomatic’ as an absence of self-reported symptoms. We did not include clinical signs observed or elicited on examination.
Asymptomatic infection
a person with laboratory-confirmed SARS-CoV-2 infection, who has no symptoms, according to the authors’ report, at the time of first clinical assessment and had no symptoms at the end of follow-up. The end of follow-up was defined as any of the following: virological cure, with one or more negative RT-PCR test results; follow-up for 14 days or more after the last possible exposure to an index case; follow-up for seven days or more after the first RT-PCR positive result.
Presymptomatic
a person with laboratory-confirmed SARS-CoV-2 infection, who has no symptoms, according to the authors’ report, at the time of first clinical assessment, but who developed symptoms by the end of follow-up. The end of follow-up was defined as any of the following: virological cure, with one or more negative RT-PCR test results; follow-up for 14 days or more after the last possible exposure to an index case; follow-up for seven days or more after the first RT-PCR positive result.
Information sources and search
We conducted the first search on 25 March 2020 and updated it on 20 April 2020, 10 June 2020, 2 February 2021 and, for this update, 6 July 2021. We searched the COVID-19 living evidence database [26], which uses automated workflow processes to: (1) aggregate simultaneous daily searches of four electronic databases (Medline, PubMed, Ovid Embase, bioRxiv and medRxiv), using medical subject headings and free-text keywords for SARS-CoV-2 infection and COVID-19; (2) de-duplicate the records; (3) tag records that are preprints; and (4) allow searches of titles and abstracts using Boolean operators. We used the search function to identify studies of asymptomatic or presymptomatic SARS-CoV-2 infection using a search string of medical subject headings and free text keywords (S1 Text). We also examined articles suggested by experts and the reference lists of retrieved studies. Reports were planned to be updated at 3-monthly intervals, with continuously updated searches.
Eligibility criteria
We included studies, in any language, of people with SARS-CoV-2 diagnosed by RT-PCR that documented follow-up and symptom status at the beginning and end of follow-up or investigated the contribution to SARS-CoV-2 transmission of asymptomatic or presymptomatic infection. We included contact tracing and outbreak investigations, cohort studies, case-control studies, and mathematical modelling studies. We amended eligibility criteria after the third version of the review [12] in two ways. First, we excluded studies that only reported the proportion of presymptomatic SARS-CoV-2 because the settings and methods of these studies were very different and their results were too heterogeneous to summarise [12]. Second, we aimed to reduce the risk of bias from studies with inclusion criteria based mainly on people with symptoms, which would systematically underestimate the proportion of people with asymptomatic infection. We therefore excluded the following study types: case series restricted to people already diagnosed and studies that did not report the number of people tested for SARS-CoV-2, from whom the study population was derived. We also excluded case reports and contact investigations of single individuals or families, and any study without sufficient follow-up (Box 1). Where data from the same study population were reported in multiple records, we extracted data from the most comprehensive report.
Study selection and data extraction
Reviewers, including crowdsourced trained volunteers, worked in pairs to screen records using an application programming interface in the electronic data capture system (REDCap, Vanderbilt University, Nashville, TN, USA). One reviewer applied eligibility criteria to select studies and a second reviewer verified all included and excluded studies. We reported the process in a flow diagram, adapted for living systematic reviews [27] (S1 Fig). The reviewers determined which of the three review questions each study addressed. One reviewer extracted data using a pre-piloted extraction form in REDCap and a second reviewer verified the extracted data. For both study selection and data extraction, a third reviewer adjudicated on disagreements that could not be resolved by discussion. We contacted study authors for clarifications. The extracted variables included, study design, country and/or region, study setting, population, age, sex, primary outcomes and length of follow-up (full list of variables in S1 Appendix). We extracted raw numbers of individuals with an outcome of interest and relevant denominators from empirical studies. From statistical and mathematical modelling studies we extracted proportions and 95% credibility intervals.
The primary outcomes for each review question were (1) the proportion of people with asymptomatic SARS-CoV-2 infection who did not experience symptoms at all during follow-up; (2) secondary attack rate from asymptomatic or presymptomatic index cases, compared with symptomatic cases; (3) model-estimated proportion of SARS-CoV-2 transmission accounted for by people who are asymptomatic or presymptomatic.
Risk of bias in included studies
After the third version of the review [12], we developed a new tool to assess the risk of bias because the study designs of included studies have changed. In previous versions, we used items from tools to assess case series [28] and the prevalence of mental health disorders [29]. The new tool assessed possible biases in studies of prevalence in general and COVID-19 in particular [4,30]. We developed signalling questions in the domains of selection (two items), information (three items) and selective reporting (one item) biases (S2 Text). For mathematical modelling studies, we used a checklist for assessing relevance and credibility [31]. Two authors independently assessed the risk of bias, using a customised online tool. A third reviewer resolved disagreements.
Synthesis of the evidence
The data extracted and the code used to display and synthesise the results are publicly available: https://github.com/leonieheron/LSR_Asymp_v4. We used the metaprop, metabin and metafor functions from the meta package (version 4.11-0) [32] and the ggplot2 package (version 3.3.5) in R (version 3.5.1). The 95% confidence intervals (CI) for each study were estimated using the Clopper-Pearson method [33]. For review question 1, in contact or outbreak investigations, we subtracted the index cases from the total with SARS-CoV-2 infection, because these people were likely to have been identified because of their symptoms and their inclusion might lead to underestimation of the asymptomatic proportion [16]. For all meta-analyses, we used stratified random effects models.
Where a meta-analysis was not done, we present the interquartile range (IQR) and describe heterogeneity visually in forest plots, ordered by study sample size [25]. We calculated the I2 statistic, which is the approximate proportion of between-study variability that is due to heterogeneity other than chance and τ2, the between-study variance. We calculated 95% prediction intervals for all summary estimates, to give a likely range of proportions that would have been obtained in hypothetical studies conducted in similar settings [17,19,20]. We did subgroup analyses for pre-specified characteristics of study design, setting and risk of bias and post-hoc analyses according to age group and geographic region, comparing groups using a χ2 test. We used meta-regression for post-hoc analyses examining associations with study size and publication date. To compare our findings with other studies, we extracted the raw data from five systematic reviews [11,13-16] and calculated prediction intervals [17]. For review question 2, as a measure of infectiousness, we calculated the secondary attack rate as the number of SARS-CoV-2-infected contacts as a proportion of all close contacts ascertained. For each included study, we compared the secondary attack rate from asymptomatic or presymptomatic index cases with that from symptomatic cases in the same study. If there were no events in a group, we added 0.5 to each cell in the 2×2 table. We did not account for potential clustering of contacts because the included studies did not report the number and size of infection clusters consistently. We used the Hartung-Knapp method for random effects meta-analysis to estimate a summary risk ratio (with 95% CI) [34]. For review question 3, we reported the findings descriptively because of large differences between study settings, methods and results. We did not construct funnel plots to examine bias across studies because their utility in studies reporting on proportions is not clear.
Results
The searches for studies about asymptomatic or presymptomatic SARS-CoV-2, on 25 March, 20 April and 10 June 2020 and 2 February and 6 July 2021 resulted in 89, 230, 688, 4,213 and 3,018 records for screening, respectively (S1 Fig). Owing to changes in eligible study designs, this update excludes 67 articles from earlier versions (S1 Table). We included a total of 146 studies addressing one or more review questions; 130 empirical studies that estimate the proportion of people with asymptomatic SARS-CoV-2 (summarised in Table 1 and S2 Table) [35-164], eight studies reporting on secondary attack rates [35,81,131,142,165-168], and 13 mathematical modelling studies [7,165,169-179]. At the time of the search on 6 July 2021, five records were preprints. We checked the publication status on 14 March 2022 and all were still preprints [61,62,88,169,171]. The review period from 1 January 2021 onwards includes 52 publications, three of which collected data during the period when the alpha variant of concern [85,125,133] had been described and vaccines were being introduced.
Proportion of people with asymptomatic SARS-CoV-2 infection
The 130 studies reported on 28,426 people with SARS-CoV-2 infection (11,923 defined as having asymptomatic infection) in 42 countries [35-164] (Table 1, S3Table). Amongst all 130 included studies, 88 studies used more than one method of follow-up to ascertain asymptomatic status (Table 1, S2 Table). Only 22 of 130 studies reported the median or mean age [38,47,70,76,77,83,85,95,99,119-121,124,126,128,133,134,139,143,146,152,164] and only five studies included children only [65,67,110,115,118]. Only 31 studies reported the sex of people with asymptomatic SARS-CoV-2 (Table 1, S2 Table) [38,47,51,53,70,71,75,76,83,85,95,99,107,119-122,124,126,128,133,134,139,143,146,147,150,153,158,162,164]. The types of included studies changed across the five versions of the review. In the first version [23], six of nine studies were contact tracing investigations of single-family clusters. In this version, two main types of study design generated the study populations of people with SARS-CoV-2: contact tracing or outbreak investigation methods were used to identify and test potentially infected contacts (46 studies, referred to as contact and outbreak investigations); and studies that involved screening of a defined group of people in settings in the community, institutions, such as long-term care facilities, or occupational groups (84 studies, referred to as screening studies).
Between-study heterogeneity was considerable, so we did not estimate a mean proportion of asymptomatic infections overall, or for screening studies (Fig 1). The IQR of estimates for all 130 included studies (141 clusters) was 14-50% (prediction interval 2-90%). In 46 studies enrolling people found through contact or outbreak investigations, for example in long-term care facilities, in aeroplanes, or on cruise ships, we estimated a summary estimate for the proportion asymptomatic (19%, 95% CI 15-25%, prediction interval 2-70%, IQR 8-37%). The estimated proportions of asymptomatic SARS-CoV-2 infections were similar in studies of contact investigations (16%, 95% CI 9-27%, IQR 8-38%, 13 studies) and outbreak investigations (20%, 95% CI 15-28%, IQR 8-38%, 33 studies) (S2 Fig).
Forest plot of proportion of people with asymptomatic SARS-CoV-2 infection, stratified by study design. In contact and outbreak investigations, the summary estimate for meta-analysis was 19% (15-25%) and the interquartile range (IQR) was 8-37%. In screening studies, the IQR was 20-65%. Across all studies, the IQR was 14-50%. The x-axis displays proportions. The red bar shows the prediction interval below each group of studies, with the prediction interval for all studies at the bottom of the forest plot. Where more than one cluster was reported, clusters are annotated with [cluster identity].
In 84 screening studies, the IQR for estimates from individual studies was 20-65% and the prediction interval was 4-94% [41-48,50-54,56,58-63,67,69-71,73,74,79,83,86,88-97,99-101,103-105,107,109,110,114-117,119,120,122,124-126,128,130,133-138,143-145,147,150-156,158-164]. The ranges of estimates were similar in three settings in which screening studies were conducted; people in a community setting (23 studies, IQR 20-53%, prediction interval 2-96%), institutional settings such as nursing homes (43 studies, IQR 26-67%, prediction interval 5-93%), and occupational settings such as amongst groups of healthcare workers (18 studies, IQR 17-64%, prediction interval 3-95%) (S2 Fig).
Three studies had data collection periods from 1 January 2021. In two nursing home outbreaks, 3/4 SARS-CoV-2 infections in partially vaccinated residents [133] and 13/14 infections (with the alpha variant) [85] in full vaccinated residents were asymptomatic. One study among healthcare workers did not report on symptom status according to vaccination or variant of concern but found that 76/155 (49%) with reinfection compared with 273/1704 (17%) with primary infection were asymptomatic [125].
Risk of bias in individual studies
There were risks of bias in all types of empirical studies (S3 Fig). In pre-specified subgroup analyses according to risk of bias domains (S4 Table2), statistical heterogeneity remained very high (I2 ≥ 84%) and the prediction intervals remained wide. In screening studies, the summary proportion in was lower in studies judged to be at low risk of information bias in the assessment of symptoms (29%, 95% CI 20-42%) than in studies at unclear or high risk of bias (47%, 95% CI 37-57%) (p=0.03, test for subgroup differences). For all other domains, estimates of the proportion asymptomatic were not associated with the assessment of the risk of bias.
Factors associated with proportion of asymptomatic SARS-CoV-2
Table 2 shows analyses for factors pre-specified in the protocol and post-hoc (Table 2). Study design and setting was pre-specified and explained 16% of the variance in heterogeneity; estimates ranged from 16% (8-29%) for contact investigations to 45% (35-56%) for screening studies in institutional settings (S2 Fig). The date of publication was associated with the estimate of the proportion asymptomatic (S4 Fig), increasing in more recent publications (p 0.03), although this only explained 4% of the variance in heterogeneity. There was some evidence of variability in different world regions (p=0.06), explaining 9% of the heterogeneity. Sample size and age range of the study participants did not appear to influence the estimated asymptomatic proportion. In three systematic reviews that we re-analysed, prediction intervals were: 1-83% (241 studies [14]); 4-97% (95 studies [15]); and 3-89% (170 studies [16]). I2 values were between 94% and 99% (S2 Appendix).
Infectiousness of people with asymptomatic or presymptomatic SARS-CoV-2
Eight studies provided data to calculate and to compare secondary attack rates by symptom status of the index case (Fig 2) [35,81,131,142,165-168]. Seven studies compared the secondary attack rate from asymptomatic with symptomatic index cases (summary risk ratio 0.32 (95% CI 0.16-0.64, prediction interval 0.11-0.95) [81,131,142,165-167]. One study compared asymptomatic with presymptomatic index cases (summary risk ratio 0.19, 95% CI 0.02-1.46) [35] and four studies compared presymptomatic with symptomatic index cases (summary risk ratio 1.00 (95% CI 0.37-2.71, prediction interval 0.11-9.12) [81,142,166,167]. The risk of information bias, specifically in symptom assessment, was judged to be high or unclear in five of the eight studies included (S3 Fig).
Forest plot of the secondary attack rate of SARS-CoV-2 infections comparing infections in contacts of asymptomatic and presymptomatic index cases with infections in contacts of symptomatic cases. The RR is on a logarithmic scale. The diamonds show the summary estimate and its 95% confidence interval. The red bar shows the prediction interval. CI, confidence interval; E, number of secondary transmission events; N, number of close contacts; RR, risk ratio; Symp. symptomatic individuals.
Contribution of asymptomatic and presymptomatic infection to SARS-CoV-2 transmission
We included 13 mathematical modelling studies (Fig 3, S5 Table) [7,165,169-179].The models in nine studies were informed by analyses of data from contact investigations in China, South Korea, Singapore, and from an outbreak on the Diamond Princess cruise ship, using data to estimate the serial interval or generation time [7,165,170,172,173,176-179]. In three studies the authors did not analyse any original data sources [169,174,175].
Forest plot of proportion of SARS-CoV-2 infection resulting from asymptomatic or presymptomatic transmission. For studies that report outcomes in multiple settings or with alternative assumptions, these are annotated in brackets. CI, confidence interval; SI, serial interval.
Estimates of the contributions of both asymptomatic and presymptomatic infections SARS-CoV-2 transmission were very heterogeneous. For asymptomatic SARS-CoV-2 infection, four studies contributed five estimates [7,171,175,177]. In three studies, the estimates suggested a contribution to SARS-CoV-2 transmission of asymptomatic infection of less than 15%. One study estimated a higher proportion (69%, 95% CrI 20–85%) with a wide credibility interval [177] (Fig 3). The estimates have large uncertainty intervals, and the disparate predictions result from differences in the proportion of asymptomatic infections and relative infectiousness of asymptomatic infection.
We included 12 studies providing 16 estimates of the contribution of presymptomatic transmission [7,165,169,170,172-179]. The models examined a range of epidemic settings and used different assumptions about the durations and distributions of infection parameters such as incubation period, generation time and serial interval (S5 Table). In seven studies, point estimates for the estimated contribution of presymptomatic infection to all SARS-CoV-2 transmission in at least one reported scenario were 40% or greater [7,169,170,172,173,175,176] (Fig 3). In one study that estimated a contribution of <1% [174], the model-fitted serial interval was longer than observed in empirical studies. The credibility of most modelling studies was limited by the absence of external validation and of uncertainty intervals for the estimates cited. (S5 Fig). The estimates from studies that relied on data from different published sources that might not have been compatible were assessed as providing low quality evidence (S5 Table).
Discussion
Summary of main findings
1) For all 130 included studies, the IQR for the proportion of asymptomatic SARS-CoV-2 was 14-50%, prediction interval 2-90%) and for 84 studies based on screening of defined populations, IQR 20-65% (prediction interval 4-94%). In 46 studies that identified participants through contact tracing of index cases and outbreak investigations, the summary proportion from meta-analysis was 19% (95% CI 15-25%, prediction interval 2-70%). 2) The risk ratio for the secondary attack rate from asymptomatic compared with symptomatic infections was 0.38 (95% CI 0.16-0.64, prediction interval 0.11-0.95) and for presymptomatic infections compared with symptomatic infection was 1.00 (95% CI 0.37-2.71, prediction interval 0.11-9.12). 3) In mathematical modelling studies, estimated proportions of all SARS-CoV-2 infections that result from transmission from asymptomatic individuals were mostly below 15%, and from presymptomatic individuals mostly higher than 40%. Evidence about asymptomatic infections caused by variants of concern, or with immunity following infection or vaccination, is limited.
Strengths and weaknesses of the living systematic review methods
A strength of the methodology of this review is the transparent reporting, with openly available data and changes over different versions reported in the protocol. Our inclusion criteria attempted to reduce risks of bias and we developed a new tool to address potential biases in the studies included in this review. In contact investigations, we subtracted index cases from the total number of people with SARS-CoV-2 to avoid underestimation of the proportion asymptomatic [16]. We examined heterogeneity in detail and, as a result of the wide prediction interval, we chose not to report an overall summary estimate [18,25]. A limitation of the methods for this living systematic review is that this update only includes published studies up to 6 July 2021. This covers the period when vaccines started to be rolled out and the alpha variant of concern became dominant in high-income countries. Although we made extensive efforts to comply with the planned 3-monthly updates, with weekly searches and a continuous process of screening, data extraction and risk of bias assessment, the pace of publications about SARS-CoV-2 exceeds the capacity of our crowd of reviewers [10,26]. Our decision to include preprints compensates for some of the delay because these articles appear sooner than peer-reviewed publications. In reviews of observational epidemiological studies, search terms are broad so the number of studies that needs to be screened is high, but the yield of included studies is low. The four databases that we searched are not comprehensive, but they cover the majority of publications and we do not believe that we have missed studies that would change our conclusions. We have also not considered the possible impact of false negative RT-PCR results, which might be more likely to occur in asymptomatic infections [180] and would underestimate the proportion of asymptomatic infections [181].
Comparison with other reviews and interpretation
The type of studies that provide estimates of the proportion of asymptomatic SARS-CoV-2 infections and heterogeneity between them has changed over the course of the pandemic. In our living systematic review, the prediction interval has widened from 23-37% in studies published up to 25 March 2020 [23], to 3-67% up to June 2020 [12], 2-89% up to 2 February 2021 [21]and remains at 2-90% up to 6 July. We found three systematic reviews, in which authors reported restriction to studies with adequate follow-up (S2 Appendix) [11,13,16]. In two of the reviews, authors also applied criteria to reduce the risks of selection bias, with summary estimates of 18% (95% CI 9–26%, I2 84%, 9 studies) [13] and 23% (95% CI 16–30%, I2 92%, 21 studies) [11]. In both reviews, with studies published up to mid-2020, many included studies used designs that we defined as contact or outbreak investigations (Fig 1, S2 Table, S2 Fig). Sah et al. reviewed studies published up to April 2021 and their subgroup estimate from studies in long-term care facilities, which include many outbreak investigations, was 17.8%, 95% CI 9.7-30.3%, 15 studies [16]. The summary estimates from all these reviews are compatible with our estimate from 46 studies in similar settings (19%, 95% CI 15-25%, prediction interval 2-70%, I2 90%) (Fig 1).
It may not be possible to obtain a single summary estimate from published literature of the proportion of persistently asymptomatic SARS-CoV-2 infection. Estimates from meta-analysis might be precise, but are likely to be unreliable owing to unacceptably high levels of heterogeneity. In three large systematic reviews, overall estimates had narrow confidence intervals [14-16], but I2 values were 94-99% and prediction intervals, which show the extent of all between-study variability were not reported [17]. The prediction intervals that we calculated extended more or less from zero to 100% (S2 Appendix), making differences in estimates between these studies hard to interpret. We expected this update to our living systematic review to provide a more precise and less heterogeneous estimate of the proportion of people with asymptomatic SARS-CoV-2 than in the previous version [12]. In particular, we expected that studies that detect SARS-CoV-2 through screening of defined populations and follow up of those infected would be less affected by biases in study methodology [30] and would provide a more accurate estimate of persistently asymptomatic SARS-CoV-2, which should be influenced mainly by properties of the virus and the host response to infection [182]. Study design was the factor that explained the largest proportion of variability in this review (S2 Appendix). Information bias, resulting from the way in which asymptomatic status is determined, was the factor most strongly associated with the estimated proportion of asymptomatic infection in screening studies (Table 2). Studies based on contact and outbreak investigations might obtain more detailed data about symptoms, resulting in lower estimates of the proportion that is classified as asymptomatic. Symptom report also differs between different groups of study participants, even within the same study, and could also contribute to heterogeneity [183]. Age might play a role as children appear more likely than adults to have an asymptomatic course of infection, but age was poorly reported and could not be examined in detail (Table 1, Table 2).
The analysis of secondary attack rates in this update now provides strong evidence of lower infectiousness of people with asymptomatic than symptomatic infection (Fig 2). The difference in secondary attack rates between asymptomatic and symptomatic index cases in our meta-analysis is smaller and less biased than in systematic reviews that analyse groups of studies reporting asymptomatic index cases and of symptomatic cases separately [182,184]. In meta-analyses of two proportions, the direct comparison within studies reduces heterogeneity and is less biased [34]. Since SARS-CoV-2 can be transmitted a few days before the onset of symptoms [185], presymptomatic transmission likely contributes substantially to overall SARS-CoV-2 epidemics. If both the proportion and transmissibility of asymptomatic infection are relatively low, people with asymptomatic SARS-CoV-2 infection should account for a smaller proportion of overall transmission than presymptomatic individuals. This is consistent with the findings of modelling studies in our review, although the absence of descriptions of the epidemic context in many studies made it difficult to compare findings across studies.
Implications and unanswered questions
This living systematic review shows the challenges of synthesising evidence from observational epidemiological studies. Heterogeneity in systematic reviews of prevalence is a recognised challenge [34,186]. Methodological guidance to refrain from meta-analysis, and to report prediction intervals, when the variability between studies is extreme is often ignored in favour of summary estimates, which are easy to cite [18,20]. Part of the heterogeneity in our review arises from the fact that many studies were not designed to estimate the proportion of asymptomatic SARS-CoV-2 infection. The incomplete descriptions of inclusion criteria, response rates, follow-up and of definitions of symptom status [1] made it difficult to assess the risks of bias and to investigate their contribution to between-study heterogeneity.
The finding that, in studies of contact and outbreak investigations, a substantial minority of people with SARS-CoV-2 infection remains asymptomatic throughout the course of infection, and that almost half of all transmission might occur before symptoms develop has already had implications for prevention. When SARS-CoV-2 community transmission levels are high, physical distancing measures and mask-wearing need to be sustained to prevent transmission from close contact with people with asymptomatic and presymptomatic infection. Integration of evidence from epidemiological, clinical and laboratory studies will help to clarify the relative infectiousness of asymptomatic SARS-CoV-2. Studies using viral culture as well as RNA detection are needed since RT-PCR defined viral loads appear to be broadly similar in asymptomatic and symptomatic people [180,187].
Determining the viral dynamics and full clinical spectrum of infection with variants of concern is important. Variants classed as omicron differ substantially from all earlier SARS-CoV-2 variants, with high infectiousness and immune evasion [188], and viral characteristics and immunity could influence the occurrence of asymptomatic infection. Studies published in early 2022 are already reporting a wide range of estimates of asymptomatic omicron infection. In India, from the date of emergence of the omicron variant, 24 November 2021 to 4 January 2022, authors reported a high proportion of asymptomatic omicron variant infections (56.7% of 291) but did not report any follow-up and >80% of participants had been vaccinated [189]. In contrast, authors of a cohort study of an outbreak of omicron SARS-CoV-2 in Norway, found only 1 of 81 infections in a highly vaccinated group was asymptomatic after 10 days of follow-up [190]. There are increasing challenges for studies relying on routine health service or surveillance data; in many jurisdictions, indications for routine testing are being reduced, which will make selection biases more likely, and mandated quarantine and isolation periods for people with diagnosed SARS-CoV-2 infection are being reduced, which will increase information biases in the ascertainment of persistent asymptomatic status. Researchers need to design studies to address this specific research question for each variant of concern, taking into account vaccination status and prior infection. There are ongoing prospective studies that collect appropriate data [125], for which improved reporting could address the requirements for assessing asymptomatic infection status fully, but ongoing funding for these studies is not secure [191]. Without prospective longitudinal studies with methods that minimise selection and measurement biases, further updates to this living systematic review are unlikely to provide a reliable summary estimate of the proportion of asymptomatic infections caused by SARS-CoV-2.
Data Availability
The data extracted from the included studies and the code used to display and synthesise the results are publicly available: https://github.com/leonieheron/LSR_Asymp_v4
Author contributions
Conceptualization: Diana Buitrago-Garcia, Aziz Mert Ipekci, Leonie Heron, Nicola Low
Data Curation: Diana Buitrago-Garcia, Aziz Mert Ipekci, Leonie Heron, Hira Imeri, Lucia Araujo-Chaveron, Ingrid Arevalo-Rodriguez, Agustín Ciapponi, Muge Cevik, Muhammad Irfanul Alam, Kaspar Meili, Eric A. Meyerowitz, Nirmala Prajapati, Xueting Qiu, Aaron Richterman, William Gildardo Robles-Rodríguez, Shabnam Thapa, Ivan Zhelyazkov.
Data extraction: Diana Buitrago-Garcia, Aziz Mert Ipekci, Leonie Heron, Eric A. Meyerowitz, Aaron Richterman, Nirmala Prajapati, William Gildardo Robles-Rodríguez, Nicola Low
Risk of Bias assessment: Diana Buitrago-Garcia, Aziz Mert Ipekci, Leonie Heron, Ingrid Arevalo-Rodriguez, Anthony Hauser, Aaron Richterman, Nirmala Prajapati, William Gildardo Robles-Rodríguez, Nicola Low
Formal Analysis: Leonie Heron, Georgia Salanti, Nicola Low
Methodology: Diana Buitrago-Garcia, Aziz Mert Ipekci, Leonie Heron, Anthony Hauser, Georgia Salanti, Nicola Low
Project Administration: Diana Buitrago-Garcia, Aziz Mert Ipekci, Leonie Heron, Hira Imeri
Supervision: Georgia Salanti, Nicola Low
Validation: Diana Buitrago-Garcia, Aziz Mert Ipekci, Leonie Heron, Hira Imeri, Lucia Araujo-Chaveron, Ingrid Arevalo-Rodriguez, Agustín Ciapponi, Muge Cevik, Muhammad Irfanul Alam, Kaspar Meili, Eric A. Meyerowitz, Nirmala Prajapati, Xueting Qiu, Aaron Richterman, William Gildardo Robles-Rodríguez, Shabnam Thapa, Ivan Zhelyazkov, Anthony Hauser, Nicola Low
Writing-Original Draft: Diana Buitrago-Garcia, Aziz Mert Ipekci, Leonie Heron, Nicola Low
Writing-review & editing: Diana Buitrago-Garcia, Aziz Mert Ipekci, Leonie Heron, Hira Imeri, Lucia Araujo-Chaveron, Ingrid Arevalo-Rodriguez, Agustín Ciapponi, Muge Cevik, Muhammad Irfanul Alam, Kaspar Meili, Eric A. Meyerowitz, Nirmala Prajapati, Xueting Qiu, Aaron Richterman, William Gildardo Robles-Rodríguez, Shabnam Thapa, Ivan Zhelyazkov, Anthony Hauser, Georgia Salanti, Nicola Low.
Funding
Funding was received from the Swiss National Science Foundation (320030_176233, to NL), http://www.snf.ch/en/Pages/default.aspx ; European Union Horizon 2020 research and innovation programme (101003688, to NL), https://ec.europa.eu/programmes/horizon2020/en;
Swiss government excellence Scholarship (2019.0774, to DBG https://www.sbfi.admin.ch/sbfi/en/home/education/scholarships-and-grants/swiss-government-excellence-scholarships.html and the Swiss School of Public Health Global P3HS stipend (to DBG), https://ssphplus.ch/en/globalp3hs/.
The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript
Competing interest statement
NL is an academic editor at PLOS Medicine, received the grant funding from the organisations declared above, is a member of the Swiss National COVID-19 Science Task Force, and is on the scientific board of Sefunda, a start-up company that develops point-of-care diagnostics for sexually transmitted infections.
DBG receives funding from the organisations declared above.
AMI, LH, HI receive salary support from the grants to NL from the Swiss National Science Foundation 320030_176233, and/or European Union H2020 101003688.
Supporting Information
S1. PRISMA Checklist
S1 Text. Search strings
S2 Text. Risk of Bias tool
S1 Appendix. Data extraction forms
S2 Appendix. Analysis of other systematic reviews on asymptomatic SARS-CoV-2
S1 Table. Studies included in version 3.0 and excluded in version 4.0 of the living systematic review
S2 Table. Characteristics of studies reporting on proportion of asymptomatic SARS-CoV-2 infections (review question 1 and question 2)
S3 Table. Location of studies contributing data to review question 1
S4 Table. Subgroup analysis according to risk of bias
S5 Table. Characteristics of mathematical modelling studies
S1 Fig. Flowchart of identified, excluded, and included records as of 6 July 2021
S2 Fig. Forest plot of proportion of people with asymptomatic SARS-CoV-2 infection, stratified by study design.
S3 Fig. Risk of bias assessment of studies in question 1 and 2
S4 Fig. Forest plot of proportion of people with asymptomatic SARS-CoV-2 infection by date of publication.
S5 Fig. Assessment of credibility of mathematical modelling studies.
Acknowledgements
We are very grateful to Dianne Egli-Gany, Michel J. Counotte, Stefanie Hossmann who were part of our original core team of reviewers and contributed to the first three versions of this living systematic review.
Footnotes
We updated the search to cover publications from 2 Feb 2021 to 6 July 2021. This update resulted in the inclusion of 40 new studies, including three with any information about asymptomatic proportions in people with infections caused by the alpha variant, after vaccination, or with reinfection. We also extended the statistical investigation of heterogeneity in estimates of the asymptomatic proportion: in addition to subgroup analyses according to assessments of the risk of bias, we examined study design and setting and country of the study, and we used meta-regression to examine associations with sample size and publication date. Tsble 2 has been replaced and former Table 2 is now S4 Table. We removed many of the forest plots in the supplementary material because they are now summarised in the new analyses in Table 2. In the text, we added an 'author summary' and extended the discussion about the implications for infections caused by variants of concern and after vaccination.
References
- 1.↵
- 2.↵
- 3.↵
- 4.↵
- 5.↵
- 6.↵
- 7.↵
- 8.↵
- 9.↵
- 10.↵
- 11.↵
- 12.↵
- 13.↵
- 14.↵
- 15.↵
- 16.↵
- 17.↵
- 18.↵
- 19.↵
- 20.↵
- 21.↵
- 22.↵
- 23.↵
- 24.↵
- 25.↵
- 26.↵
- 27.↵
- 28.↵
- 29.↵
- 30.↵
- 31.↵
- 32.↵
- 33.↵
- 34.↵
- 35.↵
- 36.
- 37.
- 38.↵
- 39.
- 40.
- 41.↵
- 42.
- 43.
- 44.
- 45.
- 46.
- 47.↵
- 48.↵
- 49.
- 50.↵
- 51.↵
- 52.
- 53.↵
- 54.↵
- 55.
- 56.↵
- 57.
- 58.↵
- 59.
- 60.
- 61.↵
- 62.↵
- 63.↵
- 64.
- 65.↵
- 66.
- 67.↵
- 68.
- 69.↵
- 70.↵
- 71.↵
- 72.
- 73.↵
- 74.↵
- 75.↵
- 76.↵
- 77.↵
- 78.
- 79.↵
- 80.
- 81.↵
- 82.
- 83.↵
- 84.
- 85.↵
- 86.↵
- 87.
- 88.↵
- 89.
- 90.
- 91.
- 92.
- 93.
- 94.
- 95.↵
- 96.
- 97.↵
- 98.
- 99.↵
- 100.
- 101.↵
- 102.
- 103.↵
- 104.
- 105.↵
- 106.
- 107.↵
- 108.
- 109.↵
- 110.↵
- 111.
- 112.
- 113.
- 114.↵
- 115.↵
- 116.
- 117.↵
- 118.↵
- 119.↵
- 120.↵
- 121.↵
- 122.↵
- 123.
- 124.↵
- 125.↵
- 126.↵
- 127.
- 128.↵
- 129.
- 130.↵
- 131.↵
- 132.
- 133.↵
- 134.↵
- 135.
- 136.
- 137.
- 138.↵
- 139.↵
- 140.
- 141.
- 142.↵
- 143.↵
- 144.
- 145.↵
- 146.↵
- 147.↵
- 148.
- 149.
- 150.↵
- 151.
- 152.↵
- 153.↵
- 154.
- 155.
- 156.↵
- 157.
- 158.↵
- 159.
- 160.
- 161.
- 162.↵
- 163.
- 164.↵
- 165.↵
- 166.↵
- 167.↵
- 168.↵
- 169.↵
- 170.↵
- 171.↵
- 172.↵
- 173.↵
- 174.↵
- 175.↵
- 176.↵
- 177.↵
- 178.
- 179.↵
- 180.↵
- 181.↵
- 182.↵
- 183.↵
- 184.↵
- 185.↵
- 186.↵
- 187.↵
- 188.↵
- 189.↵
- 190.↵
- 191.↵
