In asymptomatic people, the tests will likely perform worse. The levels of virus are likely to be lower in any individual infected person, which would increase the false-negative rate. And in the general, symptom-free population, the expected levels of infection are actually quite low, so the false-positive rate could be very high.
Yet Admiral Brett Giroir, the administration’s “testing czar” and an assistant secretary at the U.S. Department of Health and Human Services, has explicitly said that the tests could be used for asymptomatic screening, at schools perhaps.
That contradiction worries Baird. “One branch of the government is saying, ‘Use this test for asymptomatic people,’ and then on the other side, they are saying, ‘Use this test for symptomatic people,’” he said.
Baird is particularly anxious that the performance of tests will deteriorate in the field and when applied to asymptomatic people. That always happens with lab tests, he told us. “They haven’t published clinical-trials data,” he said. “You foist that test on the public after collecting evidence that it would work.”
False positives worry Mina, too. Among people tested within the first seven days of showing symptoms, the Abbott test will, according to its EUA, generate a false positive from roughly one in 50 tests. Because relatively few people test positive out of the whole population, those false positives could represent a large percentage of the positive results that a batch of the tests would generate. For now, the solution is supposed to be for people who test positive to get a confirmatory PCR test. But “saying that these tests need to be confirmed with a PCR test isn’t a good answer,” Mina told us. If a “quick” positive result then forces people to wait four days for a PCR positive, the first result stops meaning much.
Mina suggests that a cornucopia approach could provide the answer: If you take an Abbott test and get a positive result, then you would take another quick test, made by a different company, that detects a different viral protein, for confirmation. He said that such procedures were common in screening for relatively rare diseases, such as HIV, where the Centers for Disease Control and Prevention issues an “algorithm” for sequencing tests.
“People are just thinking about COVID testing differently for some reason, but imperfect screens are pretty common, so I am scratching my head,” says Dan Larremore, a computer scientist and an infectious-disease modeler at the University of Colorado, who has collaborated with Mina. “The perfect has really been the enemy of the good here, in many ways—except that we also know how to embed the good within follow-up systems to make it nearly perfect.”
Mina is running a trial comparing PCR and antigen tests in both symptomatic and asymptomatic people, in order to generate real-world data about false positives and negatives. “My hope is that six weeks from now, we’ll have a pretty good set of data to reflect the performance,” he said.