May 4: In a recent post, we warned readers against rushing out to purchase one of the many COVID-19 antibody tests suddenly flooding the market, noting that few of these tests have been independently validated, and many are grossly inaccurate. But what about the tests the U.S. Food and Drug Administration (FDA) has approved for emergency use? How useful or accurate are those tests? And, if you were to take one, what might your individual result mean?
Antibody tests, or “serology tests,” are blood tests that look for signs of an immune response to infection — in this case, immune molecules, or antibodies, specifically targeted to fighting SARS-CoV-2, the virus that causes the COVID-19 illness. While we still don’t know if these COVID-19 antibodies can protect you from being infected with the virus again, their presence in the blood would indicate that you had previously been infected with the virus, even if you never had noticeable symptoms.
But when it comes to antibody testing, the bigger question is this: If you have a positive result on a SARS-CoV-2 antibody test, what is the probability that you actually had COVID-19? Strangely enough, the answer to this question has less to do with the accuracy of the test than with the number of people in the population who have actually been exposed to the virus.
For example, the FDA- and CE (European Union)-approved antibody test from Cellex promises 94% sensitivity (percentage of correctly identified true positives) and 96% specificity (percentage of correctly identified true negatives). In other words, it’s a pretty accurate test.
But let’s suppose we’re using this test on a random sample of 1,000 people from a population with a 5% prevalence rate, meaning that, on average, 50 out of 1,000 people will actually have antibodies to the virus.
Of these 50 people with antibodies, the Cellex test will correctly identify 94 percent of the true positives, or 47 individuals. When it comes to true negatives, the Cellex test will correctly identify 96 percent, or 912 out of 950 individuals; the other 38 will get a positive result, even though they do not have antibodies. The total number of positive results will be 85 out of 1,000. But only 47 of those positive results will be correct, meaning that if you are one of those 85 individuals with a positive result, the probability that you actually have SARS-CoV-2 antibodies is only 55 percent — a predictive value that is better than a coin toss, but not by much.
Want a complete explanation? Watch MIT Professor Michael J. Cima demonstrate the math that explains how a test’s sensitivity and specificity may combine with low population prevalence to make an individual antibody test relatively useless.
But while individual results on antibody tests may not be particularly useful at the moment, antibody tests still have a role to play as part of a larger, population-level surveillance strategy that can tell us where the virus has been and how it is spreading over time. When used for this purpose, individual false-positive results matter less, because those errors can be factored out. The hope is that this information can be used to help local and state governments plan for the resumption of normal activities across entire populations, rather than one person at a time.