STEVE INSKEEP, HOST:
Blood tests to detect past exposure to coronavirus are starting to hit the market. There are dozens of manufacturers, and the tests do not need FDA approval. So they're going right out the door. But even tests that claim to be more than 90% accurate apparently are not. NPR science correspondent Richard Harris is on the line. Richard, good morning.
RICHARD HARRIS, BYLINE: Good morning, Steve.
INSKEEP: What are the tests supposed to do?
HARRIS: Well, the tests cannot be used to diagnose the disease. Instead, they identify antibodies that appear in your blood about a week after you've been infected. Scientists do not know, though, whether people with antibodies are definitively protected from the disease and, if so, for how long. But that hope - that prospect - is really driving a lot of this excitement. So for example, I talked to Deborah VanderGaast in Tipton, Iowa. She runs a daycare center for children with developmental and behavioral disabilities. They're a lot like little kids everywhere.
DEBORAH VANDERGAAST: We laugh about, you know, the sanitizing everything because, you know, it's reinfected just a - just two seconds later.
HARRIS: VanderGaast is eagerly awaiting the rollout of the blood test in her county. She thinks about her staff who are being hypervigilant not to spread the disease.
VANDERGAAST: If some of the people have already been established to have antibodies, they wouldn't have to go home and isolate for two weeks; they could continue working.
HARRIS: Not to mention peace of mind a test result would bring.
VANDERGAAST: What if we're already immune and we just don't know it and we don't have to be scared?
HARRIS: She says the test isn't available in her area. But it is starting to take off nationally.
Dr. Jeremy Gabrysch runs a mobile medical service in Austin, Texas. He says he got a supply of antibody tests made by a major Chinese manufacturer and he's already run a few hundred tests in the last few days.
JEREMY GABRYSCH: We, you know, offer the tests for people who may have suspected that they had coronavirus back in February or March, when testing with a nasal swab PCR was very limited.
HARRIS: Gabrysch says he only tests people when he has other evidence that they might have been exposed.
GABRYSCH: If they had an illness that sounds like it could have been coronavirus and they have a positive antibody test, then it's very likely that this is a - what we call a true positive - that they indeed had COVID-19.
HARRIS: The test he's using boasts a specificity of 99%, which means it only falsely says a blood sample has antibodies when it doesn't just 1% of the time. But despite that impressive statistic, a test like this is not 99% correct and, in fact, in some circumstances could be much, much worse. That's because of this counterintuitive fact. The validity of a test depends not only on the test itself but, oddly, on how common the disease is in the population you're sampling.
GILBERT WELCH: It is kind of a strange thing.
HARRIS: Dr. Gilbert Welch is a scientist at Brigham and Women's Hospital in Boston.
WELCH: An antibody test is much more likely to be wrong in a population with very little COVID exposure.
INSKEEP: Richard, I think we need to slow down here. Why would the...
HARRIS: OK.
INSKEEP: ...Accuracy of a test depend on how common the disease is in a population?
HARRIS: Yeah, that is surprising. But here's a simple way to look at it. Say you are running a test that gives five falsely positive results in 100 people. Sounds like pretty good odds, right? But...
INSKEEP: Yeah.
HARRIS: But consider this, Steve - if 5% of a population is infected and you run the test on 100 people, you should get five true positives, but you also have those five false positives.
INSKEEP: Oh.
HARRIS: And Welch says there's no way to know which is which.
WELCH: The tests will be wrong half the time. Half the people will be falsely reassured.
HARRIS: So it's basically a coin flip.
And it gets worse. The Food and Drug Administration does not regulate these tests, but the White House Coronavirus Task Force set an informal standard. They are supposed to have no more than 10 false positives per hundred. If you were to use a test that meets that standard in a population where only 1% of the population had been infected with the coronavirus, a positive result would be wrong a shocking amount - more than 9 times out of 10. There's a calculator on npr.org so you can run your own numbers.
And you can see that one way to limit this problem is to focus on populations where the disease is more common. Dr. Jordan Laser, a pathologist at Northwell Health on Long Island, N.Y., says it would make sense to start with health care workers.
JORDAN LASER: So it'd be wonderful for health care workers to know their immune status and give them just a peace of mind.
HARRIS: Positive results in a heavily exposed population are more likely to be true positives. Even so, Laser says it would still be a mistake to rely on these results.
LASER: Definitely don't use these tests to change your practices in terms of personal protective equipment. It really would be more of a psychological benefit.
HARRIS: Jeremy Gabrysch, the doctor in Austin, is also using these tests to satisfy the curiosity of his patients, not to provide advice about whether it's safe for people to expose themselves to the coronavirus.
GABRYSCH: We don't recommend that people interpret the result that way, and we counsel against that. I mean, it's not a Get Out of Jail Free card.
INSKEEP: Well, if it's not a Get Out of Jail Free card, is this test really useful at all?
HARRIS: Well, clearly it falls short of being the imagined passport that we would love to have. But you know, these tests can still be incredibly useful as long as individual false positive results don't matter. And one situation where that is the case is surveying broad populations. And in fact, these tests will be used to figure out just where the coronavirus is in the United States and how common it is.
INSKEEP: Richard, thanks so much.
HARRIS: Sure.
INSKEEP: NPR science correspondent Richard Harris. Transcript provided by NPR, Copyright NPR.
