The asymptomatic fraction of infection is the proportion of infected persons who never develop, perceive, and report symptoms (
1). Among common pathogens, the asymptomatic fraction varies widely. For example, an asymptomatic carrier state has not been documented for measles virus infection (
2), whereas a significant proportion of persons with cytomegalovirus or poliovirus infection have no symptoms and are unaware of infection (
3,
4). The asymptomatic fraction of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection seems to be sizable (
5). The range of severity of illness associated with SARS-CoV-2 infection is noteworthy because it spans asymptomatic infection; mild illness; and severe, life-threatening illness.
Perhaps because of this broad spectrum of presentation, the topic of asymptomatic SARS-CoV-2 infection has generated some controversy (
6). Imprecise use of the term “asymptomatic” is partly to blame. “Asymptomatic” should be reserved for persons who never develop symptoms, whereas “presymptomatic” is a better description of those who have no symptoms when they receive a positive test result but who eventually develop symptoms. We know for certain who is asymptomatic only in retrospect. On the basis of our current knowledge of the natural history of coronavirus disease 2019 (COVID-19), after a person is infected with SARS-CoV-2, we must wait approximately 14 days to determine whether symptoms have developed (
7). Infection without symptoms, whether presymptomatic or asymptomatic, is important because infected persons can transmit the virus to others even if they have no symptoms (
8,
9).
Methods
Data Sources, Search Terms, and Study Selection
Using the keywords antibodies, asymptomatic, coronavirus, COVID-19, PCR, seroprevalence, and SARS-CoV-2, we periodically searched Google News, Google Scholar, medRxiv, and PubMed for observational, descriptive studies and reports of mass screening for SARS-CoV-2 that were either cross-sectional or longitudinal in design; were published through 17 November 2020; and involved SARS-CoV-2 nucleic acid or antibody testing of a target population, regardless of current symptomatic status, over a defined period.
Data Extraction and Quality Assessment
We recorded the total number of persons tested, the number that tested positive, the number of positive cases without symptoms, the criteria for determining symptomatic status, whether the data were cross-sectional or longitudinal in nature, whether random selection techniques were used to achieve a representative sample of a target population, and whether the testing involved polymerase chain reaction (PCR) analysis of a nasopharyngeal swab or serologic analysis of antibodies in a blood sample. For longitudinal studies that provided information on the evolution of symptomatic status, we recorded the proportion of persons who tested positive but had no symptoms at the time of testing and who then remained asymptomatic during a follow-up period. In addition, we flagged studies that required clarification of ambiguous details.
Studies or reports that are based on PCR results and include only cross-sectional data do not make it possible to distinguish between presymptomatic and asymptomatic SARS-CoV-2 infection because symptomatic status is observed on only 1 occasion, which may occur before the development of symptoms, if any. In contrast, we can distinguish between presymptomatic and asymptomatic infection with either antibody-based studies, in which an interview or questionnaire gathers information about symptoms reported at the time a blood sample is taken and during a prior period, or PCR-based studies that include longitudinal data.
In assessing quality, we put the greatest emphasis on random selection of participants to achieve a representative sample of a regional or national population, a large number of study participants (
n > 10 000), and study designs that make it possible to distinguish between presymptomatic and asymptomatic infection. Evaluated in this manner, the highest-quality evidence comes from large-scale, national studies with representative samples that include data from either antibody or longitudinal PCR testing. In Tables
1 and
2, we show in boldface the details that increase a study's likelihood of providing higher-quality evidence.
Data Synthesis and Analysis
We synthesized evidence qualitatively by evaluating study design, including whether data were collected longitudinally; testing methods; number of participants; and setting. We compared the range and consistency of estimates of the proportion of persons who tested positive but had no symptoms at the time of testing.
Role of the Funding Source
The National Institutes of Health played no role in the design, conduct, or analysis of this review or in the decision to submit the manuscript for publication.
Discussion
Symptom detection relies on the subjective reports of patients (
73). For example, anosmia has turned out to be a distinctive symptom of COVID-19 (
74), and we depend on patients to perceive and report a diminution, however slight, of their normal olfactory abilities. But such self-reports are influenced by many factors, including variability in the ability to recall symptoms and idiosyncratic awareness of bodily sensations.
Current data suggest that infected persons without symptoms—including both presymptomatic and asymptomatic persons—account for more than 40% of all SARS-CoV-2 transmission (
75–77). The proportion of new infections caused by asymptomatic persons alone is uncertain, but when researchers in Wanzhou, China, analyzed epidemiologic data for “183 confirmed COVID-19 cases and their close contacts from five generations of transmission,” they determined that the asymptomatic cases, which made up 32.8% of infected persons, caused 19.3% of infections (
78).
The 61 studies and reports that we have collected provide compelling evidence that the asymptomatic fraction of SARS-CoV-2 infection is sizable. These data enable us to make reasonable inferences about the proportion of SARS-CoV-2 infections that are asymptomatic.
Studies designed to achieve representative samples of large populations provide useful data because they may accurately reflect human populations in general. Four of the PCR-based studies are in this category, with target populations of England (
10–12,
14), Iceland (
16), and Indiana (
23). The proportion of persons who tested positive but had no symptoms at the time of testing ranged from 43.0% to 76.5%, with a median of 45.6% (IQR, 43.6% to 61.8%). However, these studies fall short of providing the highest-quality evidence because they collected only cross-sectional data. As a result, we cannot distinguish between presymptomatic and asymptomatic cases.
In 14 longitudinal studies that reported information on the evolution of symptomatic status, a median of 72.3% of persons who tested positive but had no symptoms at the time of testing remained asymptomatic during a follow-up period (
15,
17,
18,
20,
22,
32,
37–40,
47,
51,
53,
54). If a similar proportion remained asymptomatic in the 4 large, representative, PCR-based studies, in which the median was 45.6%, the asymptomatic fraction of SARS-CoV-2 infection would be 33.0%.
Among the data that we have assembled here, the highest-quality evidence comes from the large-scale studies using antibody testing that were designed to achieve representative samples of nationwide populations in England (
n = 365 104) (
55) and Spain (
n = 61 075) (
56). It is remarkable that these independently conducted serosurveys yielded nearly identical proportions of asymptomatic SARS-CoV-2 infections: 32.4% in England and 33.0% in Spain.
We may infer that persons who receive positive antibody test results can be classified accurately as asymptomatic because such results are likely to occur only after the onset of symptoms, if any. In a study of 222 hospitalized patients in Wuhan, China, IgM and IgG antibodies to SARS-CoV-2 were first detected 3 and 4 days, respectively, after symptomatic onset of COVID-19 (
79). In a study of 109 health care workers and 64 hospitalized patients in Zurich, Switzerland, the severity of illness seemed to affect how quickly SARS-CoV-2 antibodies appeared (
80). Patients with severe COVID-19 had detectable SARS-CoV-2 antibody titers after symptom onset, but those with mild cases “remained negative or became positive [for SARS-CoV-2 antibodies] 12 to 14 days after symptom onset” (
80). These data suggest that positive antibody test results are unlikely to occur during the period when it is uncertain whether an infected person is presymptomatic or asymptomatic.
However, serosurveys do have significant limitations for the purpose of estimating the asymptomatic fraction. Not all persons who are believed to have been infected with SARS-CoV-2 later have a positive result for SARS-CoV-2 antibodies (
81). The reasons may include a false-positive result on the initial PCR test; a false-negative result on the antibody test; or the absence of detectable antibodies, perhaps because the infection was cleared without requiring adaptive immunity. In addition, the role of mucosal immunity in clearing SARS-CoV-2 infection has not yet been fully elucidated (
82), and a nasal wash to detect the IgA antibodies active in mucosal immunity is not part of standard testing practice. Persons who clear SARS-CoV-2 infection through innate or mucosal immunity might be more likely to be asymptomatic but would not be categorized as such in a serosurvey, possibly contributing to an underestimate of the asymptomatic fraction.
Another limitation of serosurveys is the requirement that an interview or questionnaire about symptomatic status accompany the blood sample. The onus is on the study participant to accurately recall symptoms, if any, from weeks or even months earlier. In the midst of a pandemic that has transformed everyday life around the globe, it seems reasonable to hypothesize that awareness of and memory for symptoms possibly related to COVID-19 are heightened. This might result in a greater likelihood of noticing and reporting symptoms that would otherwise be missed or ignored, thereby leading to a lower estimate of the asymptomatic fraction. For these reasons, we have evaluated serosurveys in the context of other results and found them to be concordant.
When estimates from large-scale, cross-sectional, PCR-based studies with representative samples; longitudinal PCR-based studies; and nationwide serosurveys with representative samples are combined, it seems that the asymptomatic fraction of SARS-CoV-2 infection is at least one third. To confirm this estimate, large-scale longitudinal studies using PCR testing with representative samples of national populations would be useful. As SARS-CoV-2 vaccination campaigns are implemented worldwide, though, the window for such research may be closing.
In light of the data presented here, we believe that COVID-19 control strategies must be altered, taking into account the prevalence and transmission risk of asymptomatic SARS-CoV-2 infection. Frequent, inexpensive, rapid home tests (
83) to identify and contain presymptomatic or asymptomatic cases—along with government programs that provide financial assistance and, if necessary, housing to enable infected persons to isolate themselves (
84)—may be a viable option. And as the first generation of SARS-CoV-2 vaccines is deployed, more research will be needed to determine their efficacy in preventing asymptomatic infection (
85).
Should conclusions be based on preprints that have not been peer-reviewed yet?
With great interest, the manuscript of the research article 'The Proportion of SARS-CoV-2 Infections That Are Asymptomatic: A Systematic Review' was critically evaluated. After expressing commendation for the serious effort by authors to explore the prevalence of asymptomatic SARS-CoV-2 infections, the commenter wishes to direct the attention of the Editor towards the fact that data from preprints which have not yet been peer-reviewed have been included in the synthesis of conclusions.
It is true that the COVID-19 pandemic has lead to an immense rise in the amount of literature getting published & preprint servers provide the optimal platform for accelerated dissemination of scientific research around this global health emergency1, one can not deny that no peer-review process whatsoever is adopted while screening submitted manuscripts for publication at a preprint server which has lead to studies with flawed methodologies & biased conclusions getting published by the same preprint servers.2 Thus, when conclusions are synthesized by evaluating data from a preprint alongside that taken from research articles published in peer-reviewed journals which have gone rigorous evaluation reviewers and editors, that significant status which peer-review process maintains in the scientific research publishing global standards gets unintentionally blemished. This may translate into a prediction of the scientific community to disregard the findings of such research studies that take in data from preprints to synthesize conclusions. Consequently, the authors of such studies may not achieve their sincerest objective of positively contributing to the scientific discourse.
The commenter, therefore, suggests revising the methodology of the systematic review under consideration such that only peer-reviewed research articles are included for data extraction & subsequent qualitative systematic review. Including only peer-reviewed studies shall translate into a higher quality of synthesized conclusions which shall be better received by the scientific community. The suggested alternative is to revise the systematic review once all of the included preprints have either been published in peer-reviewed journals or have been retracted secondary to any reason.
Regards.
REFERENCES
1Majumder MS, Mandl KD. Early in the epidemic: the impact of preprints on global discourse about COVID-19 transmissibility. Lancet Glob Health. 2020 May;8(5):e627-e630. doi: 10.1016/S2214-109X(20)30113-3. Epub 2020 Mar 24. PMID: 32220289; PMCID: PMC7159059.
2Silva JATD. Silently withdrawn or retracted preprints related to Covid-19 are a scholarly threat and a potential public health risk: theoretical arguments and suggested recommendations. Online Information Review. 2020;ahead-of-print(ahead-of-print).
Disclosures:
I'm a moderator at PsyArXiv Preprints and a student member of the American College of Physicians.
Asymtomatic and presymptomatic cases
Concludng remarks should be added.
Author Response to Bangash
We are in the midst of a profound transformation in how scientific knowledge is disseminated. As recently as 30 years ago, the gating factors in scientific publishing were the speed of postal mail and the high costs of printing. Today, ubiquitous access to the Internet allows for nearly instantaneous transmission at an infinitesimal price. The logical and welcome result has been the rise of preprints.
Peer review certainly serves a valuable role in scientific publishing, usually boosting the quality of the output. But in the current publishing environment, it would be a mistake to rely exclusively on peer-reviewed articles as the definitive source of scientific knowledge. Instead, we must become more discriminating consumers of scientific publications -- from both peer-reviewed journals and preprint servers -- and learn to assess for ourselves the quality and merit of the knowledge that is being shared.
In preparing our systematic review, we carefully evaluated the preprints that met our criteria for inclusion. In our opinion, they were of sufficiently high quality to include in our analysis, particularly because they provided knowledge that was not available from peer-reviewed sources. In the midst of a pandemic, in which our findings might be useful to both clinicians and policymakers, we decided that this was the most prudent course of action.
New Variants, Vaccines and The Proportion of Asymptomatic SARS-CoV-2 Infections
TO THE EDITOR:
A recent systematic review by Oran and Topol concluded that at least one third of SARS-CoV-2 infections are completely asymptomatic (1). As included large-scale representative data, the conclusion seemed more convincing than the initial narrative review.
Notably, evidence gap was still significant. Except the nationwide program in England, all the other eligible studies included participants prior to September, 2020. Moreover, the vast majority of studies were not as long as 3 months. Therefore, the review was less likely to disclose whether the proportion of asymptomatic SARS-COV-2 infections would keep stable for a long period.
Since the SARS-COV-2 genome has thrown up numerous variants over one year evolution, it is reasonable to suspect the coronavirus might undergo phenotypic "drift". Circumstantial evidence came from almost 20 million international entrants to China. Between mid-April and mid-October 2020, the proportion of asymptomatic infections among all positive individuals increased significantly over time from 27.8% to 59.4% (2). This finding may signal an increase in asymptomatic infection globally, in which D614-to-G614 transition might have a place.
The viral variant problem became prominent at the end of 2020. Take fast-spreading B.1.1.7 as an example. The tendencies of spike gene dropout transition in the Pillar 2 sample (community testing of people with symptoms) closely matched the trends of a random sampling of the community (3). This might suggest that the proportion of asymptomatic infection in B.1.1.7 variant remained relatively stable, whereas hazard of death was estimated higher compared with previously circulating variants (3).
Further to this, the efficacy of first-generation vaccines on reduction in asymptomatic infections will soon face the imminent challenges of new variants (4), especially more worrisome lineages such as B.1.351 and P.1.
It is vital to keep track of mutations in the genome of SARS-CoV-2. An interactive mutation tracker system based on SARS-CoV-2 isolate genomes deposited to GISAID might provide an option to accrue the clinical metadata. However, with sizable missing data and lack of longitudinal follow-up on symptoms, the tool can only give a patchy understanding of disease severity (5).
Global coordination in productive expansion of sequencing efforts and robust collection of outcome data can allow us to really building the capacity to comprehend new variants and asymptomatic infections. The insight will support meaningful public health actions to choose the highest-efficacy vaccines and to make timely alterations in the existing vaccines, which could reduce selective pressure for emergence of more variants.
References
Disclosures: No conflicts of interest.
Authors' Response to Yang and Ma
As noted in our review, we included data published as of 17 November 2020. It is inevitable, then, because of typically long lead times in journal publishing, that the data were collected before the widespread circulation of the SARS-CoV-2 variants of concern mentioned by Drs. Yang and Ma. In addition, all of the studies that they cite were published after our review appeared.
As new SARS-CoV-2 variants of concern emerge and the mix of variants in widespread circulation changes, we agree that it will be important to reassess the prevalence of asymptomatic infection.
Are "asymptomatic" infections truly "asymptomatic?"
In their recent review article on the incidence of asymptomatic SARS-Co-V-2 infection, Drs. Oran and Topol conclude that at least one third of SARS-Co-V-2 infections are "asymptomatic." The conclusion is based upon a summary of 61 studies and reports.
To arrive at this conclusion, one has to have a clear definition of "asymptomatic" versus “asymptomatic” cases. The understanding of symptoms related to SARS-Co-V-2 infection has evolved since we began to identify SARS-Co-V-2 infections in March 2020. I have been involved personally, in advising camps, schools and many individuals in managing outbreak situations. Early on in the course of the pandemic, we focused on the symptoms of cough and fever. Later on, we observed that many individuals, especially children and young adults, presented with more subtle symptoms such as nasal stuffiness, headaches, G.I. symptoms or just fatigue. We would sometimes ask individuals who tested positive about their symptoms and were told that they had none. Upon more persistent questioning, we would learn that they had a runny nose or some of the other mentioned symptoms for several days, but did not relate these symptoms to their positive test result.
In addition, it is difficult to define "asymptomatic" among the elderly or debilitated patients. Such patients often lack awareness of symptoms. Frequently, they are unable able to report subjective symptoms. These patients would not be defined as being "symptomatic," unless they developed fever, cough or shortness of breath.
In order to properly report on the incidence of "asymptomatic" infection, one must know that the individuals were carefully questioned about any symptoms, some of which would be subtle. In addition, it would be reasonable to place elderly debilitated patients in a separate category, as their symptoms can be easily missed.
In this way, I believe strongly that the actual incidence of "asymptomatic" infection is much lower than what Oran and Topol estimate. Summarizing studies without a clear understanding of how a history of symptoms was obtained can lead to false conclusions, which tends to increase the level of anxiety concerning asymptomatic infections. Such anxiety might not be warranted
Authors' Response to Berman
Personal observations are often the starting point of scientific inquiry, so we appreciate that Dr. Berman has shared his experience in advising camps, schools, and others. But he makes no mention of involvement in mass SARS-CoV-2 screening programs that test all persons in a group or locale without regard to symptomatic status. It is only through the study of data from such mass screening that the actual asymptomatic fraction can be ascertained. In our review, we assembled 61 data sets in this category, including more than 1.8 million persons worldwide. Regarding Dr. Berman's concern about ambiguity in defining the asymptomatic condition, we note in our review that "thirty of the studies included a list of specific symptoms, independent of signs, used to determine symptomatic status." In preparing our review, we relied on the competence and veracity of researchers in applying these criteria and assessing the symptomatic status of study participants.