As the number of coronavirus cases in the United States exceeds 9.2 million, experts continue to call for an enormous scale-up of testing among both the healthy and the sick — a necessary measure, they have said, to curb the spread of an infection that can move swiftly and silently through the population.

One strategy has involved the widespread use of rapid tests, which forgo sophisticated equipment and can return results in minutes. Purchased in bulk by the federal government and shipped nationwide, millions of these products have already found their way into clinics, nursing homes, schools, athletic teams’ facilities and more, buoying hopes that the tests might hasten a return to normalcy.

But a new study casts doubt on whether rapid tests perform as promised under real-world conditions, especially when used in people without symptoms.

In a head-to-head comparison, researchers at the University of Arizona found that, in symptomatic people, a rapid test made by Quidel could detect more than 80 percent of coronavirus infections found by a slower, lab-based P.C.R. test. But when the rapid test was used instead to randomly screen students and staff members who did not feel sick, it detected only 32 percent of the positive cases identified by the P.C.R. test.

Quidel’s tests are authorized for use only in people with symptoms, but their use in those who are asymptomatic has been strongly encouraged by the federal government. In September, the White House announced detailed plans to purchase and distribute 150 million rapid antigen tests manufactured by another company, Abbott.

“This is really valuable data that has been hard to come by,” said Dr. Benjamin Mazer, a pathologist at Johns Hopkins University who was not involved in the study. “But 32 percent is a very low sensitivity. I’m surprised by how low that is.”

The university’s data, which have not yet been published in a peer-reviewed scientific journal, were evaluated by Dr. Mazer and several other outside experts at the request of The New York Times. Rapid coronavirus tests like Quidel’s, they said, should be rolled out with the utmost care — and come with clear-cut explanations about how the results should be interpreted.

Much of coronavirus testing has so far relied heavily on a laboratory technique called polymerase chain reaction, or P.C.R., which can amplify very small amounts of genetic material from the virus, and which is sensitive enough to detect the pathogen even when it is scarce. But these tests are slow and expensive, sometimes taking days to return results. At the University of Arizona, which hosts nearly 50,000 students and staff in a typical year, there are enough laboratory resources to run only about 1,500 P.C.R. tests each week, said David Harris, a stem cell researcher and an author on the study.

Quidel’s test, called the Sofia, looks for bits of coronavirus proteins, or antigens, instead of genetic material. It is less accurate than P.C.R.-based tests. But it takes only 15 to 30 minutes to run, using a toaster-size instrument, and costs about $23, compared with $50 or more for a typical P.C.R. test.

Since the spring, the university has purchased and administered more than 40,000 Sofia tests, Dr. Harris said, and now has enough Quidel devices and personnel to process approximately 2,000 rapid tests each day.

Image
Credit…Quidel, via Associated Press

Dr. Harris and his colleagues evaluated nearly 2,500 people from June to August. Among 885 people who had experienced Covid-19-like symptoms or had recently been exposed to the coronavirus, 305 tested positive by P.C.R. — 54 of whom were missed by the antigen test, or 18 percent.

That number might have been lower if the test group had included only people with symptoms, Doug Bryant, Quidel’s president and chief executive, said in an interview. The Sofia’s instruction manual claims that the product can detect 96.7 of the infections that P.C.R. tests can when used on people in the first five days of their illness. (Dr. Harris said that he did not know what proportion of the people in the group of 885 were symptomatic.)

The researchers also screened for the virus in another 1,551 randomly selected people who had no symptoms. Nineteen of them tested positive by P.C.R.; of those who did, only six were caught by the Sofia.

“The data for the symptomatic group is decent,” said Jennifer Dien Bard, the director of the clinical microbiology and virology laboratory at Children’s Hospital Los Angeles, who was not involved in the study. “But to get less than 50 percent in the asymptomatic group? That’s worse than flipping a coin.”

Among people without symptoms, the Sofia also produced more false positives than P.C.R.-confirmed positives, mistakenly identifying seven participants as infected when they were not actually carrying the coronavirus.

When asked about the Sofia’s poorer performance in asymptomatic people, Dr. Harris said, “I think everybody expected that.”

But he and his colleagues argued in their manuscript that some of the asymptomatic people who tested positive with P.C.R., but negative with the rapid test, might have been missed for good reason: They were carrying too little of the coronavirus to spread it to others.

Some P.C.R.-based tests can be configured to produce a number called the cycle threshold, or C.T., which increases as the amount of virus in the body decreases.

Of the 13 asymptomatic people who were not identified by the Sofia test, 12 had C.T. values in the 30s.

“At these high values, we don’t even know if a P.C.R. test is ‘correct,’” said Werner Kroll, senior vice president of research and development at Quidel. As with any diagnostic tool, false positives can appear in P.C.R.-based tests. These tests can also pick up on harmless genetic fragments of the coronavirus that linger in the body for weeks after people have recovered.

The researchers were also unable to grow the coronavirus out of samples from volunteers whose P.C.R. tests had C.T. values above 27. “If I don’t have live virus, I am not infectious at all,” Dr. Harris said.

When Dr. Harris and his colleagues discarded all P.C.R. positives with C.T. values above 30 from their data set, the rapid test detected more than 85 percent of the coronavirus infections detected by the lab test, regardless of whether people were symptomatic.

Mr. Bryant, of Quidel, who received an early copy of the University of Arizona study, praised the results as “very, very good,” citing the Sofia’s ability to root out “people who are infectious.”

Quidel, which is running several studies of its own, does not yet have definitive results that show the Sofia works in people without Covid-19 symptoms. But “based on the data so far, it seems to be applicable to that population,” Mr. Bryant said of his company’s results.

Other experts advised caution.

Although C.T. values do tend to increase as virus levels diminish, exceptions to this trend exist — and there is no universal “magic-number cutoff” for infectiousness, Dr. Dien Bard said.

Failing to grow the coronavirus out of a person’s sample also does not guarantee that individual is not contagious to others, said Omai Garner, the associate director of clinical microbiology in the UCLA Health System, who was not involved in the study.

Several experts noted that the University of Arizona study did not track transmission among its participants, making it impossible to draw conclusions about how, and from whom, the virus spread.

Dr. Harris said that some of the concerns about the Sofia’s accuracy could be overcome with repeat testing. Screened frequently enough with a rapid test, infected people missed by one Sofia would probably be detected with the next, especially if the levels of virus in their bodies were rising, Dr. Harris said. He and his colleagues are now gathering data on University of Arizona athletes, who are tested daily, to investigate this possibility. Quidel has also partnered with the Pac-12 and Big Ten Conferences to conduct daily tests.

People with symptoms or known exposures to the coronavirus should still get the most precise and reliable tests available — those that use P.C.R., said Susan Butler-Wu, a clinical microbiologist at the University of Southern California who was not involved in the study. More data, she added, would be needed to figure out how rapid tests fit into the larger diagnostic landscape.

Still, with tens of thousands of completed coronavirus tests under its belt and a relatively low number of cases, “it does look like the University of Arizona is doing something right,” said Linoj Samuel, a clinical microbiologist at Henry Ford Health System in Detroit.

“You have to accept that you will miss some people,” Dr. Samuel said. But in a low-risk population of mostly younger students, “if that’s the price you pay for having testing versus no testing at all, that might be OK,” he added, especially if P.C.R. is available to verify positive results.

Dr. Harris said he remained confident in the ability of rapid tests to keep the student body safe. “Until someone can figure out a better way to do it with a rapid turnaround,” he said, “this seems to be the best way to go.”

[Like the Science Times page on Facebook. | Sign up for the Science Times newsletter.]