It’s been clear from quite early in the COVID-19 pandemic that a substantial number of people who get infected by SARS-CoV-2 don’t experience significant symptoms. This simple fact has enormous public health consequences, as these asymptomatic individuals can still pass the infection on to others. That means that even if we were able to get everyone with symptoms to self-isolate, we may still be unable to check the spread of the pandemic. It also makes it much harder to find out the true spread of the virus, since many people won’t bother to get tested if they aren’t feeling unwell.
Most of the data on the spread of the pandemic within the US comes from tests that pick up the presence of the virus’ genome, which indicates the presence of an active infection. But you have to catch the person while the infection is happening for this to work. The alternative is to look for an indication of a past infection: the presence of antibodies against SARS-CoV-2. While the immune response to the virus is complex and isn’t present immediately after an infection, most people have at least some antibodies a few weeks after the virus is cleared.
This allows widespread antibody testing to provide a clearer picture of the virus’ past spread through a population. On Tuesday, the CDC started releasing lots of data from past antibody testing. While it was from a period where the virus was relatively rare in the US, the data provides a sharp contrast to the RNA-based tests from the same time, showing that lots of infections have gone undetected.
A growing body of data
Most of the CDC results are described in a paper published today in JAMA Internal Medicine, based on a collaboration between it and public health agencies from around the country. The data comes from a collection of samples made available by large medical testing companies, obtained from people who had routine tests run, such as checks of cholesterol levels. But the CDC is also an aggregator of data and has posted some information on its website that was gathered by a number of states. The data have been analyzed separately.
We’ll go through the data in the paper first. It comes from two large commercial testing labs, which sent in serum from a total of 16,025 blood samples that had come in for routine testing. The tests do include data on the postal code of where they originated but don’t include detailed demographic information beyond age and sex.
The samples came from a range of locations that had very different experiences early in the pandemic: the San Francisco Bay area, Connecticut, Florida, Louisiana, Minneapolis-St Paul, Missouri, New York City, Philadelphia, Utah, and Washington state. Depending on the exact location, the samples came from late March to early May, a period where the pandemic was picking up in some locations, but others remained largely untouched. Rather than rely on commercial testing, the CDC did its own in-house assay for the presence of antibodies, one that only produced a four percent false negative rate and less than one percent false positives.
Overall, the rates were still very low. Even New York City, the focus of the pandemic’s initial spread in the US, only saw antibodies in 6.9 percent of these blood samples. On the low end, California’s Bay Area saw only one percent of the samples test positive for antibodies. Age and sex didn’t seem to have any relationship to the possibility of having antibodies.
The CDC also had data on the presence of COVID-19 cases in these same locations. The lag between the experience of symptoms and sufficient antibodies picked up by this assay can be accounted for, allowing the CDC to make an estimate of how many people who now have antibodies didn’t have a case of COVID-19 that required treatment. Depending on the location, the paper estimates that there were anywhere from six to 24 times the number of infections as there were people who ended up needing treatment.
While the size of that study is great, it’s hardly a random sample. Many individuals who felt they were healthy probably postponed routine medical care during this pandemic. But there are other, more focused studies being done that provide a more representative sample of some US locations.
For example, in a study done in the Atlanta metropolitan area, researchers selected about 400 households at random, giving them samples from a total of 700 people. Also done in late spring, this study found that 2.7 percent of the participants had antibodies to SARS-CoV-2 and that about half of them had experienced symptoms of COVID-19.
Around the same time, Indiana did a random sampling of the entire state. At the time, assays based on the presence of the viral genome were indicating about 1.7 percent of the Indiana population had an active infection at the time. A bit over half of those reported no symptoms when the tests were performed. In contrast, the antibody-based testing showed that only about 1.1 percent of the population had antibodies. Thus, the virus was clearly spreading within the Indiana population at the time, since more people were actively infected than had had a prior infection.
While none of this data will tell us much about the virus’ spread through the US on its own, collectively, more studies like this will paint a picture of how things have changed over time. Those changes can then be examined in light of things like the public health policies in effect at the time of testing.
The one bit of good news here is that it confirms something that has been clear for a while: many SARS-CoV-2 infections can be shrugged off without the infected person experiencing severe symptoms. To an extent, this will lower the disease’s burden on our healthcare system at a time where the virus seems to be spreading largely uncontrolled in many areas. But the studies don’t bring us much closer to a clear answer on what percentage of people we can expect to experience an asymptomatic infection, given the wide disparity of frequencies in the different locations. It’s possible that some of this is explained by different demographics in the different locations, but again, this study was done without having access to that data.
Beyond the good news, there’s some remaining uncertainty. As the JAMA paper puts it, “At present, the relationship between detectable antibodies to SARS-CoV-2 and protective immunity against future infections is not known.” We’re not at the point that we can advise the people who have antibodies to the virus that it’s safe for them to risk further exposure.
And we’ll have to end on a couple of down notes. One is that, even assuming the worst-hit regions of the country have similar rates of undiagnosed cases as the ones seen here, they’re still nowhere close to herd immunity. Put in concrete terms, we can look at New York City, which had some of the highest rates of known infections in the country, Even if those rates are multiplied by 10 to account for the undiagnosed cases found here, New York City would fall well short of herd immunity.
The other downer is that we know people who don’t experience COVID-19 symptoms are likely to be able to spread the virus as asymptomatic carriers. In the absence of sufficient testing to regularly check random samples of the population, this will make it difficult to track its spread. And it will also make it harder to stop the spread of the virus, given that many of those infected will have no indication that they have COVID-19. This emphasizes that all of us collectively have to act responsibly by social distancing and wearing masks and other protective gear, even if we have no reason to think we have the virus. Because there is always a chance that we do.