Asymptomatic infections in COVID-19 revisited
In an earlier review, we examined the proportion of SARS-CoV-2 infections that may be asymptomatic– individuals infected with the virus but not displaying any symptoms. In that review, we reported that actual and estimated proportions of infections that are asymptomatic ranged from 18% to 50%. In the intervening weeks, new data has emerged adding to our understanding of this topic, but definitive answers are not yet available. A better understanding of asymptomatic cases will continue to play a central role in choosing which mitigation strategies to loosen and tighten as the first wave of the epidemic wanes in many places.
Asymptomatic infections are of interest for several reasons. They add greatly to the challenge of using public health strategies to control disease spread, even being called the “Achilles’ heel” of current measures to manage ongoing COVID-19 transmission. Multiple studies have documented that asymptomatic persons can be contagious and transmit disease to others. With persistent limitations in testing capacity, and the primary use of symptom screening to identify cases, it can be impossible to trace, isolate, and quarantine asymptomatic infections effectively to interrupt further transmission. An accurate estimate of asymptomatic infections is also important for disease epidemiology. When only more severe cases are detected, a falsely higher death rate may be calculated. Furthermore, accounting for transmission dynamics from asymptomatic cases may play a role in more accurate projections and modeling of the pandemic.
On April 6, a review of the proportion of asymptomatic cases by the Center for Evidence Based Medicine at the University of Oxford cited studies reporting between 5% and 80% of persons infected with SARS-CoV-2 as being asymptomatic. This is a wide range, and cannot meaningfully guide action, as the reviewers themselves emphasize. Some of the higher estimates came from studies where mild or pauci-symptomatic persons were included in the asymptomatic group, and some of the lower estimates came from studies where targeted testing was being performed in the setting of active case-finding. Other studies reported results from single time-point PCR testing without clinical follow up, which is problematic given that a proportion of asymptomatic persons testing positive by PCR are actually presymptomatic, and will go on to develop symptoms.
A more recent preprint systematic review focused solely on studies in which people were tested as a part of contact tracing and then monitored for symptom development. The five studies they examined from the US, Italy, and China showed that the proportion of asymptomatic cases, excluding those who went on to develop symptoms after initial testing, was between 6% and 41%. Their meta-analysis estimated the asymptomatic proportion from these studies combined to be 16%. The cases were not necessarily representative of the population, given that two of the studies were conducted at long-term care facilities and keeping in mind that older persons are more likely to have more severe symptoms. The proportion of those who are infected who have no symptoms will depend, to a great extent, on the age structure of those infected. Another systematic review seeking to answer the question of contribution of asymptomatic cases to all transmission estimated from 11 studies that the proportion of asymptomatic cases was 29%. On April 6th, the Korean CDC reported an asymptomatic infection rate of 33%, again not accounting for presymptomatic persons. A Vietnamese report examining quarantined persons with appropriate follow up after testing found that 43% of those testing positive (n=30) were asymptomatic.
Using modeling and statistical analysis, the US CDC currently projects that the most likely proportion of asymptomatic cases is 35%, based on data they received through April 29. Their modeling included five possible scenarios of the overall pandemic, using upper bounds and lower bounds of various factors affecting the dynamics of the pandemic. For asymptomatic infections, their lower bound was 20% and the upper bound was 50%, almost identical to the Resolve to Save Lives review. They also note the difficulty of measuring actual proportions of asymptomatic persons without population-level data and acknowledge that most of the information has been coming from studies in which the subjects may not be representative.
Iceland is one country where random population screening by PCR has occurred in two phases, and in each phase, about half of the patients were classified as asymptomatic. These results were covered widely by the media. The true proportion of asymptomatic cases, however, has to be lower, given that there was no follow-up to discriminate between asymptomatic and presymptomatic cases. In addition, the country had a concurrent campaign for targeted testing among high-risk individuals, including those with symptoms, those identified through contact tracing, and those with travel history. For an accurate estimate of asymptomatic cases, these persons should be accounted for in the overall analysis. Although one cannot say what the proportion of asymptomatic cases is in the country, one can reliably say that it is less than 50%.
Putting these more recent studies, models, and reviews together, with closer attention to methodology of how asymptomatic proportions are calculated, and keeping in mind that the list of possible symptoms has expanded since the prior review, the asymptomatic proportion remains similar to what was quoted in the earlier review—between 16% and 43%.
In the US, Indiana University has announced a plan to estimate asymptomatic cases in their community as part of a study called TACTIC: Tracking Asymptomatic COVID-19 Through Indianapolis. The community-based study, which has completed recruiting volunteers, aims to better understand the role of asymptomatic adults and children in COVID-19 transmission, and may also contribute to our knowledge of asymptomatic cases on a population level.
There is no perfect way to detect asymptomatic disease. PCR testing for active infection may overrepresent asymptomatic persons by detecting presymptomatic persons, and recall bias as well as other imperfections in serologic testing will reduce the ability of antibody testing to estimate burden of infected persons. What is important is that asymptomatic persons can transmit disease, and likely account for up to half but no less than a fifth of all infected persons. Policies and programs to control the pandemic should be made with this sizable proportion in mind.
In an earlier review, we wrote about a possible association between loss of smell (anosmia) or taste and COVID-19. In late April, CDC added ‘new loss of taste or smell’ to its list of possible symptoms of COVID-19. On May 18, public health officials in the UK added anosmia to the list of symptoms that should prompt self-isolation for possible COVID-19, citing ‘emerging data and evidence on COVID-19.’ Indeed, in the months preceding these changes, there were calls by expert otolaryngologist (ear, nose, and throat, or ENT, specialist) groups in the US and the UK to include anosmia in the official list of COVID-19 symptoms.
Olfaction, the medical term for the smell function, may be lost completely (anosmia) or partially (hyposmia) through a variety of mechanisms, including direct trauma to the nose, brain injury, or infection. Olfaction informs the perception of flavor, and patients with anosmia may also note loss of taste (ageusia) or a decrease in taste (hypogeusia). The olfaction process begins with detection of odorants by nerve cells located in the mucosal lining of the nasal cavity. These receptor neurons travel up through perforations in a skull bone and transmit signals directly into the brain. Respiratory viral infection is a common cause of anosmia or hyposmia, usually through the mechanism of mucosal inflammation; patients often concurrently experience anosmia and nasal congestion. However, the unique anatomy of the olfactory system, in which part of the central nervous system is in direct contact with the external environment, presents another mechanism of olfactory disturbance if a virus is neurotropic (tends to infect nerve cells) and may directly damage olfactory nerves. Indeed, a number of viruses have been shown to cause post-infection olfactory dysfunction after mucosal inflammation subsides. Some, including the coronaviruses that cause the common cold and Severe Acute Respiratory Syndrome (SARS), have been shown to be neuro-invasive. Animal data shows that SARS-CoV-2 may enter the brain through the olfactory system and spread between neurons. This may support the observed link between anosmia and COVID-19 even when other nasal symptoms are absent.
The diagnosis of viral-associated olfactory loss can be challenging. There may be poor self-recognition of anosmia among patients if other, more distressing symptoms are present. Olfactory changes are not mentioned in many early studies of COVID-19 symptoms; many of those studies included only hospitalized patients. Under-reporting by patients and providers may occur if there is low awareness of the potential link between a symptom and a disease. Increased internet attention to the potential link between anosmia and COVID-19 may have led to greater recognition of anosmia among COVID-19 patients later in the pandemic. There has been limited use of standardized measures of olfaction, and many studies on the association between anosmia and COVID-19 have relied on self-reported symptoms.
Despite challenges in detection and objective measurement of anosmia, some studies have shown that disordered smell or taste is more strongly associated with SARS-COV-2 positivity than other symptoms. There are numerous case reports and series describing disordered smell or taste among COVID-19 patients. Here is a snapshot of some of these data:
|59 patients hospitalized with COVID-19 in Italy|| |
|202 patients hospitalized with COVID-19 in Italy|| |
|3,191 COVID-19 outpatients in South Korea|| |
|59 COVID-19 positive and 203 COVID-19 negative out-patients in the US|| |
|2,618,862 participants in the UK and US reported symptoms|| |
|103 in-patients and out-patients with COVID-19 in Switzerland|| |
|214 hospitalized COVID-19 patients in China|| |
|417 hospitalized COVID-19 patients in 12 European hospitals|| |
Models that accurately predict the likelihood of a person having COVID-19 may be useful to clinical and public health practice. The diagnostic accuracy of a symptom-based model to predict COVID-19 among healthcare workers in the Netherlands was improved when anosmia was given extra weight within the model. In another model, sensitivity was improved and diagnostic accuracy maintained when only two symptoms, anosmia and fever, were used to predict COVID-19.
Weekly Research Highlights
(Science, 20 May 2020)
- The authors developed a non-human primate model for SARS-CoV-2 infection to further study protective immunity and immune system “memory” after viral infection.
- Nine rhesus macaques were inoculated with three different doses of virus, and all showed evidence of viral infection with high viral loads from lung fluid samples as well as nasal samples. Subsequent to infection, the monkeys were tested for SARS-CoV-2 spike (S) antibodies and neutralizing antibodies, and showed evidence of both.
- After clearing the virus from the initial infection, they were challenged with a second inoculation of the virus, and all showed evidence of protective immunity with only minimal virus levels, which were undetectable after one day. Three controls who had not been infected in the first round did show evidence of infection in the second round.
- This study is not able to identify which part of the immune response is most responsible for conferring immunity. In addition, none of the macaques developed severe illness or respiratory failure, possibly representing some differences in infection between humans and non-human primates. And, the duration of immunity cannot be assessed, nor can it be confirmed that challenge infection is equivalent to either natural or vaccine-induced exposure.
(EID, Early release 18 May 2020)
- To evaluate the plausibility of transmission of disease through feces which was postulated as one mechanism in the 2003 SARS outbreak, feces samples from an infected patient were collected on four separate occasions.
- All samples were positive for viral RNA by PCR testing, and one sample was used to isolate virus to eventually grow by culture. The authors were then able to replicate a cytopathic effect in lab cells from virus isolated in one of the feces samples. After inoculation, the virus was able to replicate in the lab cells.
- Although the authors do not show direct evidence of human to human transmission from fecal-oral or fecal-respiratory route, they establish that it is plausible.
Decline in Child Vaccination Coverage During COVID-19 Pandemic – Michigan Care Improvement Registry, May 2016 – May 2020
(MMWR Early release 18 May 2020)
- The study authors examined immunization data in the Michigan Care Improvement Registry and compared routine vaccination rates for children under two years age at various age cohorts from 2016-2019 to rates in 2020.
- Fewer than half of the children in the registry were considered up-to-date for all recommended vaccines in their age cohort in the 2020 registry, compared to more than two thirds from 2016-2019. Children who were enrolled in Medicaid were less likely to have up-to-date vaccinations than those not enrolled in Medicaid.
- By adjusting patient flow in outpatient offices, separating well and sick child visits, or providing in-person services in alternate settings, providers may be able to improve continuity in immunization services. Catch-up campaigns will be necessary to reduce the risk of outbreaks of measles and other vaccine preventable diseases, and special outreach may be necessary for more vulnerable groups.
(MMWR. Early Release 19 May 2020)
- Upon learning of a church-pastor and wife who both tested positive for COVID-19, the Arkansas Department of Health launched an investigation to identify how the two had become infected and identify contacts they could have exposed.
- There were 35 confirmed cases of COVID-19 among 92 attendees of the church’s events. Among the surrounding community, 26 additional confirmed cases reported contact to a known case from the church cluster. There were three deaths from the church cluster. Children accounted for 35% of church attendees during the period of interest, but only 8% of confirmed cases.
- The study could not account for ill persons not seeking testing, asymptomatic infections, and any persons infected through community transmission.
(EID early release, 15 May 2020)
- An outbreak investigation in Cheonan, South Korea identified 112 cases of COVID-19 linked to fitness dance classes at 12 different sports facilities in the city. Cases included 6 instructors, 59 fitness class participants, and 47 additional household and community contacts of either infected instructors or students.
- Five of the infected instructors had attended the same training workshop in Cheonan on 15 February. This workshop also included fitness instructors from Daegu, where a COVID-19 outbreak had been on-going at the time. One of the instructors at the workshop became symptomatic on 18 February and is implicated as the index case.
- The attack rate among participants in fitness dance classes led by a SARS-CoV-2-infected instructor was 26%. No cases were detected among students who participated in low intensity Pilates or yoga classes nor among those who attended fitness dance classes with fewer than five participants.
- The authors conclude that the intensity of the 50-minute fitness dance classes, conducted with 5-22 participants in confined facilities, contributed to the high attack rate.
(Pre-print, German Research network Zoonotic Infectious Diseases website)
- Researchers analyzed viral loads in samples found to be RT-PCR positive for SARS-CoV-2. Of note, most testing was initiated based on symptoms. For the analysis, participants were categorized in two ways: 1) into 10-year age brackets and 2) on the basis of age groups associated with education levels (kindergarten, grade school, high school, university, adult, and mature age). An analysis of variance between participant groups was conducted.
- From January through April, the laboratory tested 59,831 people for SARS-CoV-2 using RT-PCR, and 3,712 (6.2%) tested positive. There were relatively few children tested and there was a lower detection rate of SARS-CoV-2 among children than among adults (127 patients were 20 years of age or less). There was no significant difference in viral load between any subgroups using either categorization method.
- Limitations include lack of analysis by symptom status or underlying disease, and biases in the study population because of test referral methods. For example, if only a small proportion of infected children have symptoms, the viral load among these individuals may not be representative of the viral load among all those infected. This article has not been peer-reviewed, and epidemiological information is required to make conclusions about actual infectivity by different age groups.
Suggested citation: Cash-Goldwasser S, Kardooni S, Kachur SP, Cobb L, Bradford E and Shahpar C. Weekly COVID-19 Science Review May 16–22 2020. Resolve to Save Lives. 2020 May 26. Available from https://preventepidemics.org/coronavirus/weekly-science-review/