Last month the White House issued guidelines suggesting a way to reduce the number of false positive results in antibody tests: Run two tests. But that strategy has not yet been validated for coronavirus testing. And the details matter.
Antibody tests identify people who have been previously exposed to the coronavirus. But false positives are a big concern. A test that has a 99% specificity is still wrong 1% of the time. And if it's being used to test a population where 1% of people are infected, half the time a positive result will be a false positive.
This phenomenon may help explain why researchers from Stanford University reported a high rate of coronavirus infection in California's Santa Clara County – vastly more than would be expected based on known diagnoses.
That problem concerned entrepreneur Michael Wohl, as he set about starting a business to provide quick and reliable testing for coronavirus.
"I started doing the math of what can be done and figured out that we could do a second test," says Wohl, who directs a program at the University of Rochester's business school to teach entrepreneurship.
The two-test system would cost more. Wohl says he's seen antibody tests priced from $30 to $100 or more. But theoretically at least, the approach could dramatically improve the outcome — provided, that is, that the second test is distinctly different from the first one, so it's not making the same error.
"When I came up with this idea, I put together a PowerPoint and sent it to the governor's office in New York and the health department. And it propagated between various government agencies," he says. He got a lot of positive feedback and was excited to see the concept embraced by the White House at the end of April.
But the White House testing guidelines have an important omission. The guidelines just say test twice, but they don't say to use two different tests.
"One can infer, actually if you look at the guidelines that it's the same test given twice. But it's critical that the two tests are uncorrelated," Wohl says.
When a person is infected, the immune system produces multiple antibodies, targeted to attach to various features on the coronavirus. Wohl says this double-testing method should look for two different antibodies, targeting two different regions of the virus.
And there's another shortcoming of the White House guidelines. Nobody seems to have conducted real-world tests to see if the concept actually works for the coronavirus.
So Wohl looked up a cousin at the University of North Carolina in Chapel Hill and asked him to see about gathering evidence for this idea. The cousin, physician-researcher David Wohl, in turn got his colleague John Schmitz, interested in the question.
"This is really modeling off what we do for a variety of other tests in our lab, like HIV testing," says Schmitz, who is a pathology professor and lab director at UNC.
A positive antibody test for HIV is routinely confirmed with a different kind of test, because a false diagnosis can have huge consequences. So, while the concept is sound, Schmitz wants to see evidence. He's starting to run his own tests. One hypothetical worry is that some common factor could cause two tests to both provide false results.
"In most cases, we don't know what causes a false positive, to be honest with you," Schmitz says. Sometimes people with autoimmune disease produce antibodies that trigger false positives in tests like this. Some antibodies that target similar viruses can also create a false signal. And laboratory processes can also contribute.
"Given the pervasive talk in the field — in the lay field even — you hear about false positives, false positive, I thought we ought to look at this to get around that problem as best we can," he says. He figures it will take a couple of months to get results from his study of the issue.
That thought has also occurred to a group at the University of California, San Francisco and UC Berkeley. They have been running careful comparisons of some of the many antibody tests to measure the performance of those that are already out on the market. Dr. Alex Marson at UCSF says they may be able to take the data they've already gathered and reanalyze it quickly, to see if different tests produce a different pattern of false positives, as would be needed for the two-test strategy.
He's also thinking about broader questions, such as when it would make sense to spend the money on two tests, as opposed to running a single slower but highly accurate test in a lab.
Part of the appeal of the two-test strategy is it can use handy tests that are like an early-pregnancy test. These are called lateral-flow tests and require just a drop of blood and can provide results in minutes.
Another approach would be to use a much more sophisticated test called an ELISA test to look for antibodies. (ELISA stands for enzyme-linked immunosorbent assay). These tests need to be run in a sophisticated lab, but some are highly precise. Marson says, depending on the circumstances, it might make sense to run one of these highly accurate tests than running two with lower performance – or perhaps using the disposable test to screen a poplulation and then follow up with the ELISA.
But getting accurate test results only solves one of two major questions around antibody testing. The other is knowing what a positive result really means.
"If you know there are antibodies in your blood, are you safe from future infections?" Marson asks. "That we do not know the answer to."
Antibodies do signal that someone has been infected with the coronavirus. But it's not clear yet whether that exposure means people are in fact immune, and if so for how long. Evidence is starting to accumulate that there is at least passing immunity, but scientists would like stronger evidence before they deem these tests trustworthy.
Even if antibodies don't indicate immunity, because they are markers of exposure, they are still useful for studying the spread of coronavirus throughout the population. But the results would not provide actionable information to individuals.
You can contact NPR Science Correspondent Richard Harris at rharris@npr.org.
Copyright 2020 NPR. To see more, visit https://www.npr.org. 9(MDAwMTM1NDgzMDEyMzg2MDcwMzJjODJiYQ004))
