Science’s COVID-19 reporting is supported by the Pulitzer Center.
After a painfully slow rollout of diagnostic testing for active coronavirus infections across the country, some 400,000 people a day in the United States may now receive such a test, estimates suggest. Yet a few public health experts say sending people back to work and school safely and identifying new outbreaks before they spread out of control could require testing much of the U.S. population of 330 million every day. Others suggest checking roughly 900,000 people per day would be enough.
Either way, nearly all the current tests to diagnose infections work by identifying the genetic material of the virus, a technology that will be difficult to scale up much further. “There will never be the ability on a nucleic acid test to do 300 million tests a day or to test everybody before they go to work or to school” Deborah Birx, White House coronavirus response coordinator, said at a press conference last month.
Birx and others have touted another option: antigen tests, which detect the presence of viral proteins in a biological sample, such as saliva or tissue swabbed from the nasal cavity. Antigen tests are typically cheap, return results in minutes, and, like the genetic tests, reveal an active infection. They already exist for strep throat, influenza, tuberculosis, HIV, and other infectious diseases. But so far, only one antigen test for SARS-CoV-2, the coronavirus that causes COVID-19, has received emergency use authorization from the U.S. Food and Drug Administration (FDA).
Can it or other antigen-based methods solve the testing problem? Some scientists are optimistic, whereas others remain skeptical, noting that such tests can be far less accurate than nucleic acid tests and may not be as easy to scale up as proponents claim. “What everyone wants is for a test to be cheap, accurate, and fast,” says Geoffrey Baird, a laboratory medicine specialist at the University of Washington, Seattle. “You can only ever have two of those.”
Developing an antigen test “is not that easy to do,” says Werner Kroll, senior vice president for research and development at Quidel, a California-based company that received the greenlight from FDA for its test earlier this month. Rather than performing all the analytical steps inside an expensive dedicated machine at a lab or a doctor’s office, as is done with tests for the DNA or RNA of virus, antigen tests build most, if not all, those steps into a paperlike strip that returns a simple yes or no answer, much like pregnancy tests.
“It’s a lab on a swab,” says Stephen Tang, president and CEO of Orasure, a diagnostics company developing its own antigen test for SARS-CoV-2. With most setups, a sample of bodily fluid is collected using a nasal swab or related procedure, then mixed with a few milliliters of a liquid, typically a sterile buffer solution. A few drops are spotted on one end of a test strip. Capillary forces pull the liquid over copies of two different antibodies specific for the same viral protein. If both antibodies spot their target—a positive test—the strip generates a signal, often a color change. This signal is generally read out by a person visually, although some setups use small readers to improve the accuracy.
What triggers the signal can differ—in some tests the antibody bindings set off a chemical reaction or expose a fluorescent marker joined to one antibody. Another test in contention for FDA approval produces an electrical readout after antibodies on an electrochemical sensor bind to their target antigen.
The challenge is finding the right antibodies, says Lee Gehrke, a virologist at the Massachusetts Institute of Technology, who has developed an antigen test for SARS-CoV-2 that E25Bio, a company he co-founded, is now evaluating. Both antibodies must bind to a single viral protein, such as the spike protein SARS-CoV-2 uses to enter cells, but at separate sites. “You have to find two antibodies that don’t interfere with each other,” Gehrke says. Those same antibodies also can’t cross react to proteins from other coronaviruses—all of which have their own spikes, for example—or anything else. “Antibodies often stick to other things nonspecifically,” Baird says.
Another challenge is weak signals. Genetic tests use the polymerase chain reaction (PCR) to amplify tagged DNA or RNA sequences, making it easy to reliably identify just a few copies of a virus. That gives PCR tests for the SARS-CoV-2 virus about a 98% sensitivity and near perfect selectivity, meaning almost every active infection is detected and only in very rare cases does someone uninfected receive a positive test. (Many false negatives, a result indicating an infected person is free of the virus, result not from the test’s deficiencies, but from poor samples, which can be difficult to collect with nasal swabs.)
Antigen tests don’t amplify their protein signal, so they are inherently less sensitive. To make matters worse, that signal gets diluted when samples are mixed with the liquid needed to enable the material to flow across test strips. As a result, most antigen tests have a sensitivity of anywhere between 50% and 90%—in other words, one in two infected people might incorrectly be told they don’t have the virus. Last month, Spanish health authorities returned thousands of SARS-CoV-2 antigen tests to the Chinese firm Shengzhen Bioeasy Biotechnology after finding the tests correctly identified infected people only 30% of the time, according to a report by the Spanish newspaper El Pais.
Quidel executives say the company’s initial SARS-CoV-2 test meets FDA’s minimum of 80% sensitivity. (That means it could still generate false negative results 20% of the time.) A revised sample preparation protocol that doesn’t require dilution of the nasal swab is expected to boost that figure to nearly 90%, but that’s still below the 98% sensitivity of state-of-the-art PCR tests.
Antigen tests, however, bring advantages to the table as well. Because they don’t require the expensive equipment and chemicals needed to perform PCR, they can be more easily used as point-of-care tests in doctor’s offices, urgent care centers, hospitals, and even at companies and schools. They also don’t require trained specialists, making them cheaper to administer—although there are a few point-of-care PCR tests, most still involve sending a sample to a lab for manual processing.
And the fast results from an antigen test mean that people who test positive can be isolated quickly, before they risk infecting others. Even if the tests have a 10% false negative rate, “people could easily be tested repeatedly, making it likely that anyone missed on the first round would be flagged on the second,” says Doug Bryant, Quidel’s president and CEO.
Another advantage is scalability. Once researchers settle on effective antibodies, the tests are easy to manufacture in bulk, and running them doesn’t require additional reagents as PCR tests do. Quidel says it expects to ship 282,000 tests this week and 1 million tests per week by early June. Ultimately, Bryant says the company should be able to produce 84 million tests per year.
That’s still well below the 300 million tests per day that would allow most every person in the United States to have a daily SARS-CoV-2 check, Birx’s ambitious hope. (One recent model from the Harvard Global Health Institute said 900,000 diagnostic tests a day in the United States would be enough to have confidence most infections were being caught before an outbreak grew big.) But other companies, including OraSure, which expects to file for FDA emergency use authorization in September, say they expect to rapidly scale up to providing tens of millions of coronavirus antigen tests as well. The demand for such tests, which could cost as little as $1 or less, could be even greater in developing countries without a broad network of centralized labs.
Taken together, the advantages of antigen tests provide real hope that they “will be very valuable for stemming this pandemic,” says Bettina Fries, chief of infectious diseases at Stony Brook University.
Baird and others are less confident. Not all antigen tests are as simple to read as a pregnancy test. Quidel’s test requires using a $1200 toaster-size reader to achieve the relatively high sensitivity it has. And even though 43,000 Quidel readers already exist for other antigen tests, most are in the United States, making the test harder to put into use overseas.
Otto Yang, an infectious disease expert at the University of California, Los Angeles, says the tests’ modest sensitivity is a bigger hurdle. Even a test with the 90% sensitivity and 100% specificity that Quidel is aiming for could misinform more than help. Assuming the virus has a prevalence of about 1% and such a test is given to 1000 people, nine people would correctly be told they are infected whereas one person would be mistakenly told they don’t have the virus. Given how readily SARS-CoV-2 spreads, “a misdiagnosis is worse than no diagnosis,” Yang says.
Fries doesn’t agree. “Even if the sensitivity [of antigen tests] is not perfect, if you test over and over you will pick up those cases,” she says. “We need to let go of the notion that all the tests have to be perfect.”
*Correction, 22 May.: An earlier version of this story mistakenly identified the number of true positive and false negative results if an antigen test was 90% sensitive and 100% selective, and the presence of the virus was 1% in a population of 1000.