Researchers have known since early in the coronavirus pandemic that the true number of infected people is higher than what is reflected in daily case totals. That’s because many people infected with the virus develop mild symptoms or none at all, and therefore may not get tested.
A new study suggests that in the Philadelphia area, the actual number of infections was about seven times higher than the official number of confirmed cases as of April 25.
The higher number came from analyzing blood samples for antibodies — proteins that the immune system makes in response to infection. More than 3% of the samples were found to contain antibodies, whereas less than half of 1% of the region’s population had tested positive in the usual way, with the nasal swabs that detect an active infection.
But hold off before drawing firm conclusions about what those numbers mean. The study, published Tuesday in JAMA Internal Medicine, has so many limitations that “seven times” could be way off in either direction, say scientists who were not involved. Even the study authors warned against using the results to determine how close we are to herd immunity, the percentage of the population that must be immune to a virus to keep its spread to a bare minimum.
The study’s primary drawback is that the blood samples were not taken randomly, but from commercial labs that happened to be testing people for various medical conditions. Those people may have been exposed to the virus at different rates than those who were not getting their blood tested, for whatever reason, but it is hard to say for sure. The authors did not include key demographic details that can affect infection rates, including the race and ethnicity of those sampled.
The authors — among them researchers from the U.S. Centers for Disease Control and Prevention and local and state health departments, including Pennsylvania’s — were upfront about that issue. “These persons may not be representative of the general population,” the authors wrote.
Some scientists not involved with the study were more blunt. Among them was A. Marm Kilpatrick, a professor of ecology and evolutionary biology at the University of California, Santa Cruz, who wrote on Twitter: “There’s only one way to safely use these data: toss them in the circular filing bin. (for clarity: trash).”
The Pennsylvania counties in the study included Bucks, Chester, Cumberland, Delaware, Lancaster, Montgomery, and Philadelphia — collectively home to 4.9 million people. The authors also analyzed data for nine other geographic regions, including greater New York City.
In addition to the lack of a random sample, another limitation was the difficulty of timing the data. This was April, when nasal-swab results in the state typically were delayed for days. Likewise, the amount of time needed to obtain antibody results can vary widely, as does the amount of time the immune system needs to make antibodies. An apples-to-apples comparison is tricky.
For the Pennsylvania results, it is not even clear that the researchers were comparing totals for the same day.
The antibody measurements came from blood samples that had been collected as of April 25. The corresponding number of confirmed cases, as determined by the nasal swabs, came from what was publicized that same day by the Pennsylvania Department of Health.
But the state’s case totals on any given day reflect only those that have been reported as of midnight the night before — apparently meaning the researchers were comparing one set of numbers through April 24 and the other through April 25.
The state Health Department declined to comment.
Being off by a day would make a significant difference. In Philadelphia and the other six Pennsylvania counties in the study, the number of confirmed cases jumped 3.9% from April 24 to 25, from 22,125 to 22,987.
Previous studies suggest that generally, the number of asymptomatic unidentified people represents about one-third of the total who have been infected. In other words, if 6% of people in a given area have been confirmed with nasal swabs, an additional 3% have been infected at some point without realizing it — not seven times as many, as the new study would suggest.
Better data may come soon, as the National Institutes of Health is conducting an antibody study with a more representative sample of the population. It, too, is not random, but the authors are collecting ample demographic data from more than 10,000 participants, including age, geographic region, race or ethnicity, occupation, employment status, and education.