But in the rush to generate and interpret serologic data, it can be hard for anyone, including experts, to understand what serologic studies can and cannot teach us. Why are they so important, and what decisions will they inform?
If carrying the covid-19 antibody confers immunity, then half to two-thirds of the population must be seropositive before we can expect to control the virus without social distancing, contact tracing or other special measures.
New serosurveys appear almost daily. Their widely varying results emphasize the hyperlocal nature of the pandemic. Researchers in Santa Clara, Calif., recently found that 1.5 percent of the county’s population sampled tested positive for the antibodies (the study’s methodology has since been roundly criticized, and a new version of the study attempts to answer those criticisms), while other cities have shown much more widespread evidence of past infection: 21 percent of those tested in New York and nearly a third in Chelsea, Mass.
More studies will follow, with a variety of strategies about whom to sample, which test to use and how to analyze the data. New ways to measure antibodies are being invented even as the pandemic unfolds, and disputes about statistical methods that would otherwise seem academic now have consequences for immediate decisions on public health and the economy.
Early studies have focused on hot spots of transmission, often to answer an important, narrowly defined question. For example, researchers from Massachusetts General Hospital studied seroprevalence in Chelsea because a disproportionate number of infected patients at the hospital lived in Chelsea. The researchers found that one reason for the excess of cases was that the virus really had spread far more widely in Chelsea than other places. Recruitment in serosurveys to date has been through what epidemiologists call a “convenience sample” — testing people out shopping in Chelsea and New York and recruiting through Facebook in Santa Clara. Such surveys can give rapid answers but can’t pin down the exact amount of infection in a location, and those who participate may not be representative of the whole population.
To understand the overall pattern in the U.S. population, larger serologic surveys must cover a wide range of areas, not just hot spots, recruiting a truly representative sample.
Testing the same people for antibodies and virus week after week can help answer another question: Do antibodies to the virus signal that a person is protected against further infection, so-called seroprotection? The idea is to follow individuals with and without antibodies, who are otherwise similar (live in the same area, have similar work patterns and otherwise are likely to have similar risks of encountering an infectious person), and find out if those with antibodies have lower rates of contracting the virus than those without.
In the best case, maybe those with antibodies are completely protected; more likely, based on experience with other coronaviruses, they will be at lower but not zero risk.
As in every epidemiologic study, the challenge in these studies is to separate causal from confounding factors, by ensuring that seropositives (those with detectable antibodies) and the seronegatives (those without) have comparable exposures to viral infection.
Many factors, including lack of hand-washing, riding crowded public transportation, job-related exposures, or living in a densely populated neighborhood, could predispose someone both to having antibodies and to getting infected again. If so, this would obscure the protective effect of antibodies; other biases could exaggerate it. Nonetheless, seroprotection studies are needed to enhance scientific understanding of whether high levels of antibody in the population signal high levels of immunity, and to inform “back to work” policies favoring those with antibodies. Epidemiologists are hard at work designing studies to minimize these biases.
With careful study design and analysis involving social epidemiologists, we could also start to understand — and try to mitigate — the higher rates of disease in groups that are already disadvantaged. There is a long tradition of “social seroepidemiology”: using serosurveys to understand social, demographic and economic reasons why people get certain infections. Applying this tradition to covid-19 could help us separate out which exposures account for high infection rates among the disadvantaged. That knowledge could help tailor social distancing policies, personal protection and other interventions to reduce these rates.
The quality of serologic tests varies widely, and many studies are being set up with the goal of quick answers rather than precision. Quick answers are important, but the next chapter must be a set of larger, carefully designed serologic studies on properly selected samples, using the best available tests. As the scale and quality of studies improve, so will our understanding of this virus and immunity to it.