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

Background There is disagreement about the level of asymptomatic severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection. We conducted a living systematic review and meta-analysis 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) Amongst people with SARS-CoV-2 infection who are asymptomatic when diagnosed, what proportion will develop symptoms later? (3) What proportion of SARS-CoV-2 transmission is accounted for by people who are either asymptomatic throughout infection or presymptomatic? Methods and findings We searched PubMed, Embase, bioRxiv, and medRxiv using a database of SARS-CoV-2 literature that is updated daily, on 25 March 2020, 20 April 2020, and 10 June 2020. Studies of people with SARS-CoV-2 diagnosed by reverse transcriptase PCR (RT-PCR) that documented follow-up and symptom status at the beginning and end of follow-up or modelling studies were included. 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 an adapted checklist for case series, and the relevance and credibility of modelling studies were assessed using a published checklist. We included a total of 94 studies. The overall estimate of the proportion of people who become infected with SARS-CoV-2 and remain asymptomatic throughout infection was 20% (95% confidence interval [CI] 17–25) with a prediction interval of 3%–67% in 79 studies that addressed this review question. There was some evidence that biases in the selection of participants influence the estimate. In seven studies of defined populations screened for SARS-CoV-2 and then followed, 31% (95% CI 26%–37%, prediction interval 24%–38%) remained asymptomatic. The proportion of people that is presymptomatic could not be summarised, owing to heterogeneity. The secondary attack rate was lower in contacts of people with asymptomatic infection than those with symptomatic infection (relative risk 0.35, 95% CI 0.10–1.27). Modelling studies fit to data found a higher proportion of all SARS-CoV-2 infections resulting from transmission from presymptomatic individuals than from asymptomatic individuals. Limitations of the review include that most included studies were not designed to estimate the proportion of asymptomatic SARS-CoV-2 infections and were at risk of selection biases; we did not consider the possible impact of false negative RT-PCR results, which would underestimate the proportion of asymptomatic infections; and the database does not include all sources. Conclusions The findings of this living systematic review suggest that most people who become infected with SARS-CoV-2 will not remain asymptomatic throughout the course of the infection. The contribution of presymptomatic and asymptomatic infections to overall SARS-CoV-2 transmission means that combination prevention measures, with enhanced hand hygiene, masks, testing tracing, and isolation strategies and social distancing, will continue to be needed.


Introduction
proportion of the variability between estimates due to study differences other than chance [15]. 126 Sources of heterogeneity are often not explored in detail, however, with infrequent reporting of 127 prediction intervals [9, 10], even though they give information about all between-study variability 128 and show the range of estimates that would be expected in future studies [15]. In this fourth update 129 of our living systematic review [10] we aimed to improve and understand the changing evidence Methods 136 We conducted a living systematic review, a systematic review that provides an online summary of 137 findings and is updated when relevant new evidence becomes available [16]. The protocol, which 138 describes modifications for each update, was first published on 1 April 2020 and amended for this 139 version on 18 June 2020, (https://osf.io/9ewys/). Previous versions of the review have been posted 140 as preprints [17] and published as a peer-reviewed article [10]. We report our findings according to 141 statements on preferred reporting items for systematic reviews and meta-analyses 2020 (S1 PRISMA 142 2020 Checklist) [18] and on synthesis without meta-analysis in systematic reviews (SWiM) [19]. 143 Ethics committee review was not required for this review. Box 1 shows our definitions of symptoms, 144 asymptomatic infection and presymptomatic status. 145

Box 1. Definitions of symptoms and symptom status in a person with SARS-CoV-2 infections 146
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: . CC-BY 4.0 International license It is made available under a perpetuity.
is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted January 30, 2022. ; https://doi.org/10.1101/2022.01.20.22269581 doi: medRxiv preprint 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 147 We conducted the first search on 25 March 2020 and updated it on 20 April 2020, 10 June 2020 and 148 2 February 2021. We searched the COVID-19 living evidence database [20], which uses automated 149 workflow processes to: (1) provide daily updates of searches of four electronic databases (Medline,150 PubMed, Ovid Embase, bioRxiv and medRxiv), using medical subject headings and free-text 151 keywords for SARS-CoV-2 infection and COVID-19; (2) de-duplicate the records; (3) tag records that 152 are preprints; and (4) allow searches of titles and abstracts using Boolean operators. We used the 153 search function to identify studies of asymptomatic or presymptomatic SARS-CoV-2 infection using a 154 search string of medical subject headings and free text keywords (S1 Text). We also examined 155 articles suggested by experts and the reference lists of retrieved studies. Reports were planned to be 156 updated at 3-monthly intervals, with continuously updated searches. 157 Eligibility criteria 158 We included studies, in any language, of people with SARS-CoV-2 diagnosed by RT-PCR that 159 documented follow-up and symptom status at the beginning and end of follow-up or investigated 160 the contribution to SARS-CoV-2 transmission of asymptomatic or presymptomatic infection. We 161 included contact tracing and outbreak investigations, cohort studies, case-control studies, and 162 mathematical modelling studies. We amended eligibility criteria in the protocol for this update in 163 two ways. First, we excluded studies that only reported the proportion of presymptomatic SARS-164 CoV-2 because the settings and methods of these studies were very different and their results were 165 too heterogeneous to summarise [10]. Second, we aimed to reduce the risk of bias from studies with 166 inclusion criteria based mainly on people with symptoms, which would systematically underestimate 167 . CC-BY 4.0 International license It is made available under a perpetuity.
is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted January 30, 2022. ; https://doi.org/10.1101/2022.01.20.22269581 doi: medRxiv preprint the proportion of people with asymptomatic infection. We therefore excluded the following study 168 types: case series restricted to people already diagnosed and studies that did not report the number 169 of people tested for SARS-CoV-2, from whom the study population was derived. We also excluded 170 case reports and contact investigations of single individuals or families, and any study without 171 sufficient follow-up (Box 1). Where data from the same study population were reported in multiple 172 records, we extracted data from the most comprehensive report. 173 Study selection and data extraction 174 Reviewers worked in pairs to screen records using an application programming interface in the 175 electronic data capture system (REDCap, Vanderbilt University, Nashville, TN, USA). One reviewer 176 applied eligibility criteria to select studies and a second reviewer verified all included and excluded 177 studies. We reported the process in a flow diagram, adapted for living systematic reviews [21] (S1 178 Fig). The reviewers determined which of the three review questions each study addressed. One 179 reviewer extracted data using a pre-piloted extraction form in REDCap and a second reviewer 180 verified the extracted data. For both study selection and data extraction, a third reviewer 181 adjudicated on disagreements that could not be resolved by discussion. We contacted study authors 182 for clarification where the study description was insufficient to determine eligibility or if reported 183 data in the manuscript were internally inconsistent. The extracted variables included, study design, 184 country and/or region, study setting, population, age, sex, primary outcomes and length of follow-up 185 (full list of variables in S1 Form). We extracted raw numbers of individuals with an outcome of 186 interest and relevant denominators from empirical studies. From statistical and mathematical 187 modelling studies we extracted proportions and 95% credibility intervals. 188 The primary outcomes for each review question were (1) the proportion of people with 189 asymptomatic SARS-CoV-2 infection who did not experience symptoms at all during follow-up; (2) 190 secondary attack rate from asymptomatic or presymptomatic index cases, compared with 191 . CC-BY 4.0 International license It is made available under a perpetuity.
is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint Synthesis of the evidence 205 The data extracted from the included studies and the code used to display and synthesise the results 206 are publicly available: https://github.com/leonieheron/LSR_Asymp_v4. We used the metaprop and 207 metabin functions from the meta package (version 4.11-0) [26] and the ggplot2 package (version 208 3.3.5) in R (version 3.5.1) to display the study findings in forest plots and synthesise their results, 209 where appropriate. The 95% confidence intervals (CI) for each study were estimated using the 210 Clopper-Pearson method [27]. For review question 1, in studies that identified participants through 211 investigation of contacts or in outbreak investigations, we subtracted the index cases from the total 212 number of people with SARS-CoV-2 infection, because these people were likely to have been 213 identified because of their symptoms and their inclusion might lead to underestimation of the 214 asymptomatic proportion [14]. For all meta-analyses, we used stratified random effects models. 215 Where a meta-analysis was not done, we present the interquartile range (IQR) and describe 216 heterogeneity visually in forest plots, ordered by study sample size [19]. For statistical examination 217 . CC-BY 4.0 International license It is made available under a perpetuity.
is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted January 30, 2022. ; https://doi.org/10.1101/2022.01.20.22269581 doi: medRxiv preprint of heterogeneity, we calculated the I 2 statistic, which is the approximate proportion of between-218 study variability that is due to heterogeneity other than chance and τ 2 , the between-study variance, 219 which is used to generate the 95% prediction interval for the likely range of proportions that would 220 be obtained in subsequent studies conducted in similar settings [15]. The protocol pre-specified 221 subgroup analyses according to study design, setting and risk of bias. We did a χ 2 test to compare 222 subgroups of studies assessed as being at low risk of bias in each domain versus those of unclear or 223 high risk of bias of bias and between studies assessed as being at low risk of bias in all domains with 224 those at unclear or high risk of bias in any domain. In additional analyses, we examined studies with 225 at least ten cases of SARS-CoV-2 infection and according to publication date. To compare our 226 findings with other studies, we extracted the raw data from three large systematic reviews [12-14] 227 and calculated prediction intervals. For review question 2, as a measure of infectiousness, we 228 calculated the secondary attack rate as the number of SARS-CoV-2-infected contacts as a proportion 229 of all close contacts ascertained. For each included study, we compared the secondary attack rate 230 from asymptomatic or presymptomatic index cases with that from symptomatic cases in the same 231 study. If there were no events in a group, we added 0.5 to each cell in the 2x2 table. We did not 232 account for potential clustering of contacts because the included studies did not report the number 233 and size of infection clusters consistently. We used the Hartung-Knapp method for random effects 234 meta-analysis to estimate a summary risk ratio (with 95% CI) [28]. For review question 3, we 235 reported the findings descriptively because of large differences between study settings, methods 236 and results. We did not construct funnel plots to examine bias across studies because their utility in 237 studies reporting on proportions is not clear. 238

239
The searches for studies about asymptomatic or presymptomatic SARS-CoV-2, on 25 March, 20 April is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint tracing studies or outbreak investigations, 39 screening studies, and four mathematical models (S1 246   Table). This review version included a total of 107 studies addressing one or more objectives; 94 247 empirical studies that estimate the proportion of people with asymptomatic SARS-CoV-2 248 (summarised in Table 1 and S2 Table) (Table  255 1). Thirty-two studies, including 9,121 infected people, were done in the United States (S3 Table) Table). is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted January 30, 2022. ; https://doi.org/10.1101/2022.01.20.22269581 doi: medRxiv preprint main types of study design generated the study populations of people with SARS-CoV-2: contact 270 tracing or outbreak investigation methods were used to identify and test potentially infected 271 contacts (40 studies, referred to as contact and outbreak investigations); and studies that involved 272 screening of a defined group of people in settings in the community, institutions, such as long-term 273 care facilities, or occupational groups (54 studies, referred to as screening studies). 274 . CC-BY 4.0 International license It is made available under a perpetuity.
is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint screening studies (Fig 1). The IQR of estimates for all 94 included studies was 13-45% and the 279 prediction interval from random-effects meta-analysis was 2-89% (S2 Fig). In studies enrolling people 280 found through contact or outbreak investigations, for example in long-term care facilities, in 281 aeroplanes, or on cruise ships, we estimated a summary estimate for the proportion asymptomatic 282 (18%, 95% CI 14-24%, prediction interval 3-64%,

295
There were risks of bias in all types of empirical studies (S4 Fig). In pre-specified subgroup analyses 296 according to risk of bias domains (Table 2), statistical heterogeneity remained very high (I 2 ≥ 84%) 297 and the prediction intervals remained wide. In contact and outbreak investigations, the estimated 298 proportion of asymptomatic individuals was associated with the risk of selection bias. In studies 299 judged to be at low risk, 25% (95% CI 18-33%, prediction interval 5-66%) and 13% (95% CI 8-20%, 300 . CC-BY 4.0 International license It is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted January 30, 2022. ; https://doi.org/10.1101/2022.01.20.22269581 doi: medRxiv preprint prediction interval 1-61%) in studies at unclear or high risk of bias (p=0.02 from χ 2 test for subgroup 301 differences). In screening studies, heterogeneity was lower in studies judged to be at low risk of 302 information bias in the assessment of symptoms (p>0.01, test for subgroup differences), with a 303 summary estimate of the proportion asymptomatic of 23% (95% CI 14-35%, prediction interval 4-304 69%). Only nine studies were judged to be at low risk of bias in all domains, with some evidence of 305 reduced heterogeneity (p=0.05, test for subgroup differences). For all other domains, estimates of 306 heterogeneity were not associated with the assessment of the risk of bias. 307

Additional analyses
308 When restricted to studies with more than ten people with SARS-CoV-2 infection (S5 Fig), the 309 estimated proportions with asymptomatic infection were very similar to the overall estimates (Fig 1,  310 Table 2). The estimates of the proportion asymptomatic in the three periods of publication date 311 were similar (S6 Fig, S7 Fig). In the three systematic reviews that we re-analysed, prediction intervals between 94% and 99% (S4 Table). 314 . CC-BY 4.0 International license It is made available under a perpetuity.
is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint   Infectiousness of people with asymptomatic or presymptomatic SARS-CoV-2

317
Five of the studies that conducted detailed contact investigations provided enough data to calculate 318 a secondary attack rate according to the symptom status of the index cases and to compare the 319 secondary attack rates by symptom status (Fig 2) [119, 129, 138-140 Contribution of asymptomatic and presymptomatic infection to SARS-CoV-2 to transmission 329 We included 11 mathematical modelling studies (Fig 4)  transmission of asymptomatic infection of less than 10%. One study estimated a higher proportion 340 (69%, 95% CrI 20-85%) with a wide credibility interval [55] (Fig 4). The estimates have large 341 . CC-BY 4.0 International license It is made available under a perpetuity.
is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted January 30, 2022.  Table). is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted January 30, 2022. Comparison with other reviews and interpretation 391 The type of studies that provide estimates of the proportion of asymptomatic SARS-CoV-2 infections 392 and heterogeneity between them has changed over the course of the pandemic. In our living 393 systematic review, the prediction interval has widened from 23-37% in studies published up to 25 394 . CC-BY 4.0 International license It is made available under a perpetuity.
is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint follow-up [9,11,14]. In two reviews of studies published up to mid-2020, authors also applied 397 inclusion criteria to reduce the risks of selection bias, with summary estimates of 18% (95% CI 9-398 26%, I 2 84%, 9 studies) [11] and 23% (95% CI 16-30%, I 2 92%, 21 studies) [9]. In both reviews, many 399 included studies used designs that we defined as contact or outbreak investigations (Fig 1, S2 Table). analysis might be precise, but are likely to be unreliable owing to unacceptably high levels of 407 heterogeneity. In the three largest systematic reviews, other than ours, authors provided overall 408 estimates with narrow confidence intervals [12][13][14]. I 2 values were 94-99%, describing heterogeneity 409 other than that due to chance, but prediction intervals, which show the extent of all between-study 410 variability were not reported [15]. The prediction intervals that we calculated extended more or less 411 from zero to 100%, making the summary estimates, and any differences in estimates between these 412 studies, uninterpretable. We expected this update to our living systematic review to provide a more is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted January 30, 2022. ; https://doi.org/10.1101/2022.01.20.22269581 doi: medRxiv preprint asymptomatic infection in screening studies (Table 2). Studies in which a wide range of possible 421 COVID-19 symptoms are assessed frequently will classify more people as having symptoms than 422 studies with a limited symptom list. Studies based on contact and outbreak investigations might 423 obtain more detailed data about symptoms, resulting in lower estimates of the proportion that is 424 classified as asymptomatic. Selection bias affected studies based on contact and outbreak 425 investigations more than screening studies, however. These studies include people identified mainly 426 through contact tracing and differential inclusion of contacts with symptoms might underestimate 427 the true proportion of asymptomatic SARS-CoV-2. Age might play a role as children appear more 428 likely than adults to have an asymptomatic course of infection, but age was poorly reported in 429 studies included in this review (Table 1). 430 The analysis of secondary attack rates in this update provides some evidence of lower infectiousness 431 of people with asymptomatic than symptomatic infection, but the small number of studies and wide 432 confidence intervals are compatible with both no difference in transmissibility or higher 433 transmissibility (Fig 2)  is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted January 30, 2022. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted January 30, 2022. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted January 30, 2022.  is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted January 30, 2022. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted January 30, 2022. ; https://doi.org/10.1101/2022.01.20.22269581 doi: medRxiv preprint . CC-BY 4.0 International license It is made available under a perpetuity.
is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted January 30, 2022. ; https://doi.org/10.1101/2022.01.20.22269581 doi: medRxiv preprint . CC-BY 4.0 International license It is made available under a perpetuity.
is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted January 30, 2022. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted January 30, 2022. ; https://doi.org/10.1101/2022.01.20.22269581 doi: medRxiv preprint