As production of the antiviral Paxlovid, made by Pfizer, ramped up this spring, the drug became a cornerstone of the US’s COVID response. The White House has invested in test-to-treat infrastructure that connects people with COVID to Paxlovid instantly, and the government is now distributing close to half a million doses of the drug each week. And the pill, which attacks a key protein involved in SARS-CoV-2 reproduction, is genuinely life-saving. In clinical trials, a five-day treatment reduced the rate of hospitalization and death by 90 percent. 

But for some people taking the pill, COVID still comes back. The phenomenon, called Paxlovid or COVID rebound, is relatively new and poorly understood. For a smattering of patients, rebound means a flare-up of fever, cough, or other symptoms. For others, the rebound is asymptomatic, but they begin testing positive after testing negative at the end of treatment.

Kami Kim, director of the Division of Infectious Disease and International Medicine at the University of South Florida Health Morsani College of Medicine, spoke to Popular Science while in quarantine with a rebound COVID case. Last month, Kim tested positive for the virus for first time during the pandemic. Although she’s not at high risk, “I got reasonably sick and my boss insisted that I take Paxlovid, because you have to give it early,” she says. Her symptoms—congestion and GI issues—cleared up soon after she started taking it.

But two weeks after her first infection started, Kim started to feel like she was allergic to something. She didn’t think much of it, until she was exposed to COVID during a trip a few days later. She took a rapid test, and it was “screamingly positive,” she says. So, she went back into quarantine. If not for that exposure, she explains, she would have just thought the symptoms were allergies. “That’s sort of like the classic story of Paxlovid rebound. You become symptomatic, but not as seriously as the first found.”

[Related: What to consider when taking Pfizer’s COVID-19 pill Paxlovid]

Reports of widespread Paxlovid rebound in the US go back to early spring 2022. On May 24, the Centers for Disease Control and Prevention (CDC) released revised guidance for people experiencing a relapse. According to the agency, patients should treat the rebound like a second round of the disease, and begin their isolation over again.

Clinical trials for Pfizer’s antiviral, which were conducted in people who smoked, were immunocompromised, or otherwise at high risk of severe COVID, saw rebound in only about 1 percent of patients. So far, there have been no comprehensive studies on the rate of Paxlovid rebound in the general public, but anecdotal reports make it more common than 1 percent. That difference could be in part because doctors are prescribing the pill to people at low risk of COVID—rather than the high-risk population included in the trial.  It “could be because we’re simply not using it in the populations in which it’s been studied,” Monica Gandhi, an infectious disease physician at the University of California at San Francisco, told STAT last month

According to the CDC, there are so far no reports of anyone experiencing severe disease during a rebound case, meaning the risk is more about continued transmission of the virus rather than danger to the patient. “It’s not like these drugs aren’t working,” Kim says. “Even with Omicron, it’s keeping people out of the hospital.”

The CDC and Food and Drug Administration (FDA) both advise against giving rebound patients a second course of the drug. The FDA even publicly contradicted Pfizer’s CEO, Albert Bourla, after he told Bloomberg that patients should repeat Paxlovid treatment after testing positive for COVID again.

Because Paxlovid was only recently developed to treat COVID, it has some unknowns. At first, researchers worried that rebound was a sign that the virus was becoming resistant to the drug. But genomic analysis of rebound cases hasn’t found any mutations to suggest that SARS-CoV-2 is adapting. Other possibilities are that the virus might survive the drug by hiding in some unknown organ reservoir, or that the body’s immune system doesn’t ramp up quickly enough to take down straggling viral particles.

[Related: An estimate of how many Americans have died from each COVID variant]

“If you think about how the drug works, it sort of mechanistically makes sense,” says Kim. Unlike monoclonal antibodies, which stop the virus from infecting cells, Paxlovid prevents SARS-CoV-2 from replicating. “But the virus is there, unless your immune system killed off all those cells that are infected. In theory, the virus could go silent,” Kim explains. “Then it wasn’t completely knocked out, and your immune system hadn’t killed off all the infected cells, in theory, it could start replicating again.”

Another reason why Paxlovid rebound has been so hard to diagnose is the unpredictability of untreated COVID infections. Some people have the live virus in their system and remain contagious well after 10 days. Early in the pandemic, here were even reports of patients recovering and relapsing repeatedly over the course of several months.

One study of those untreated relapses noted they were hard to count because they were largely self-reported. But it found that, at least in mildly ill patients, COVID rebound was unlikely to be caused by lingering virus particles. Instead, it probably had more to do with the individual’s immune response.

Immunologists are researching similar questions on long COVID, which for some people manifests as repeated flare-ups of infection symptoms. It’s likely that long COVID has a number of different causes, but one prominent hypothesis involves some kind of viral reservoir.

For individuals experiencing Paxlovid rebound, the biggest concern is that it could make them infectious once again. But at a scientific level, it’s not clear if it’s a feature of the drug, or the strangeness of COVID infections themselves.