Research posted online yesterday that is still unreviewed backs up widespread reports on social media of people developing COVID symptoms, but not testing positive with antigen tests, the most common form of at-home rapid test.

The study identified 30 people—all vaccinated and boosted—who were given both PCR and antigen tests during routine testing at five US workplaces in December. Based on the PCR results, all but one appeared to be infected with the Omicron variant. But for the first two days after that positive result, every single individual was negative on an antigen test (either by Abbott or Quiddel). The majority also had false negative antigen tests on the third day. During that window, many of the individuals had high levels of the coronavirus, and the researchers documented four confirmed transmissions.

Those standard-issue antigen tests missed every COVID case in the first days of an infection, at a time when people were demonstrably infectious. “It was alarming,” says Anne Wyllie, who studies disease testing technology at Yale’s School of Public Health and was a co-author on the research. “What do we do when an entire country is relying on so many of these tests for travel, seeing family, having vacations, and seeing grandparents?”

It’s a small study, but fits into an emerging body of evidence that suggests that people need to rethink how they’re using at-home tests. Last week, the Food and Drug Administration (FDA) reported that antigen tests are less sensitive to Omicron infections, though it didn’t publish any of its supporting data.

Another early study from researchers at the University of Cape Town found that PCR tests using mouth swabs were significantly more sensitive at picking up Omicron infections than those using nasal swabs. (They were less sensitive toward Delta.) And unreviewed studies hint that Omicron moves through the body differently, and might be better at replicating in the throat.

If true, that could help explain why standard antigen tests didn’t pick up early Omicron infections. It’s possible the variant had started to grow elsewhere in the body, so those tests, which used a nose swab, were looking for the coronavirus in the wrong place.

But does that mean you should be swabbing your throat in addition to your nose? Not necessarily.

“There’s a lot we don’t know about Omicron,” says Susan Butler-Wu, who directs clinical testing at the University of Southern California’s School of Medicine. “We know it’s more infectious; we know it’s got a shorter doubling time. What we don’t know is: At what point do you reach peak viral loads in the different specimen types? … Where am I most likely to detect the virus?”

[Related: Why it’s so hard to make quick, accurate estimates on COVID variants]

It’s also important to keep in mind that most comparisons between different sample types—including the Cape Town study—used PCR, not antigen testing. In most antigen tests, SARS-CoV-2 proteins set off a chain of biochemical reactions that ends with them stuck to a strip of paper. The entire process is designed with a nasal swab in mind. And because your spit is chemically different from your snot—with a different pH, for one thing—you don’t know if an oral sample will return more false positives or false negatives. (To see this in action, check out the videos of people getting a positive result after putting tap water on a antigen test.)

That’s why, when you’re tested with an oral sample, you’re asked not to eat, drink, smoke, or chew gum for 30 minutes beforehand. Those substances can dilute the virus or change the pH of your mouth, making results unpredictable.

The quality of results also depends on how well the sample is collected. And Jonathon Campbell, who studies disease testing at McGill University, notes that people might balk at swabbing the back of their throats. “I am not aware of the data on this, but I suspect it may be more difficult than collecting an adequate shallow nasal specimen,” he explains in an email.

“People are having anecdotal success [with mouth swabs],” says Butler-Wu. “But as a scientist, I want to see a properly done study.” Some people, she points out, have tested negative for two days using the nasal swab, and test positive later after swabbing elsewhere. Was it the change of sample? Or just that they’d built up more viral particles over time?

“In the absence of data, there’s a vacuum—and in a vacuum is when wild-Westy stuff starts to happen,” Butler-Wu says. “People who have tried something that worked are very vocal about the fact that it worked. But we don’t know how often it’s not working.”

It’s plausible that antigen tests perform well when the sampling methods are tweaked. In the UK, for example, the government-delivered antigen test involves a throat swab, followed by a nose swab. (It’s based on the same underlying detection technology as antigen tests in the US, but isn’t identical.) A small lab study of that test found no change in sensitivity to Omicron, without any real-world data.

But for the moment, no one knows the best way to make those tweaks. There are a lot of options for gathering an oral sample. The Cape Town researchers had participants swab their gums, tongue, cheeks, and palate—and then compared the results to a shallow nose swab. Meanwhile, the UK government tells people to swab the back of their throat.

“I really hope that [companies like] Abbott and Quiddel do what they have to to really quickly validate throat swabs,” says Wyllie. “We’re in such a surge that it’s actually an easy time to find positive and negative cases. They should be able to do this.”

Until that data is available though, the solution isn’t to fiddle with the test on your own. Instead, don’t treat a negative antigen test result as definitive. “The big message is that if people are using antigen tests, they cannot ignore symptoms,” says Wyllie. “They cannot ignore a high-risk exposure.”

[Related: What to do if you test positive for COVID, were exposed, or have symptoms]

If studies continue to find that antigen tests miss early Omicron infections, they will indicate that “antigen tests are a RED LIGHT test (treat the positive as if you’re infected) and not a GREEN LIGHT test (do not treat a negative as if you are non-infectious)—which is always how they should be treated given their characteristics,” Campbell says.

So if you’ve got symptoms, assume for now that you’ve got COVID until you can get a PCR test. If you’ve been exposed to someone, quarantine and take a PCR, rather than an at-home test. (Or, since testing shortages make that challenging, don’t end your precautions on the basis of an antigen test alone.) In a couple weeks, the FDA might announce that you’ll get more definitive results if you swab your mouth or throat. Just remember, nothing is a silver bullet.