On Thursday, the Food and Drug Administration made what may be the most momentous drug-approval decision in its history: It granted emergency-use authorization for Merck’s molnupiravir to treat covid-19. This approval is significant not because molnupiravir is an especially good drug, but because it is a rather ineffective and dangerous one. In particular, molnupiravir might create new variants of SARS-CoV-2 that evade immunity and prolong the pandemic.

The problem with molnupiravir lies in its mechanism of action. Unlike any previous antiviral drug, molnupiravir does only one thing: It introduces mutations into the viral genome. We are already familiar with the fact that viruses naturally mutate to evade immunity; the many mutations of the spike protein in omicron, for example, allow it to evade the antibodies created by prior infections or vaccines. Molnupiravir relies on inducing even more mutations so that eventually the virus’s proteins are damaged beyond function. That molnupiravir can mutate SARS-CoV-2 to death has been demonstrated in the controlled conditions of a petri dish and lab animal cages, leading Merck to test it in covid-19 patients in clinical trials.

But people are not petri dishes or lab animals, and while molnupiravir works to some extent, it has not worked very well in covid-19 patients. Specifically, molnupiravir reduced hospitalizations by only 30 percent. In contrast, Pfizer’s antiviral drug Paxlovid, which works by a different mechanism and was also approved this week by the FDA, reduced hospitalization by 89 percent. (My lab does research on drugs using the same mechanism as Paxlovid — inhibition of the viral protease enzyme — independently of any company affiliations.) This means that most of the time that molnupiravir was given the opportunity, it failed to inhibit viral replication enough to allow the patient to avoid hospitalization.

Merck’s own research, published Thursday, explains why. It found that viable virus can still be detected in some patients on the third day of treatment with the drug. That means that for at least several days, the drug is in the body mutating the virus — but not all virus genomes have picked up enough mutations to die off. For those initial few days, then, the patient is a breeding ground for viable mutated viruses.

The first days of molnupiravir treatment present a clear opportunity for mutant viruses to be transmitted to family members or caregivers. Viral evolution is a process of selecting for rare mutations that are beneficial to the virus. It doesn’t matter if just one out of the billions of copies of viruses in an infected individual mutates to a higher level of fitness. That single copy, either by evading existing antibodies or replicating to yet higher levels of fitness, will become amplified either in that patient or in the next person infected.

The worst-case scenario is worrisome. As long as molnupiravir is in use somewhere in the world, it could generate repeated cycles of new variants, with people desperately taking the drug to fight the new variants it spawns, creating a vicious positive feedback loop while causing more suffering and deaths.

Molnupiravir’s low efficacy may come as no surprise, because drugs that only mutate a viral genome have never been tested before in people. By contrast, the previous antiviral medication capable of mutating viruses, ribavirin, also had direct effects, including blocking the viral replication enzyme and stimulating innate immunity — and that was with much less contagious viruses. We didn’t know how well a drug whose sole function is to introduce mutations could work against a highly contagious, rapidly replicating virus. Now we know: not very well.

The FDA’s fact sheet for prescribers, also released Thursday, actually recognizes the risk that a mutated virus could escape. It says: “Completion of the full 5-day treatment course and continued isolation in accordance with public health recommendations are important to maximize viral clearance and minimize transmission of SARS-CoV-2.” But how are we going to prevent people from stopping the drug, or forgetting a dose, or merely talking and dining with family members without masks, throughout treatment? This is simply not realistic in the general population.

In addition, the fact sheet recognizes that “changes in the spike protein occurred at positions targeted by monoclonal antibodies and vaccines.” Bafflingly, however, it adds, “The clinical and public health significance of these changes are unknown.” The significance of changes to spike protein positions by antibodies and vaccines is very well known: These changes are what allowed each variant of concern — from alpha to beta to delta to omicron — to evade immunity from previous infection, vaccines or monoclonal antibody treatment.

What can be done? We can take some comfort in the FDA’s requiring Merck to report, within three months, the viral mutations induced by molnupiravir in clinical trial participants. Merck will also need to report viral mutations in immunocompromised patients, who are likely to harbor viruses longer. As this crucial information should have been supplied before approval, a responsible approach would be to limit molnupiravir use for the next three months to the best controlled settings. For example, health care providers could prescribe it only to people who live alone, or who live in managed care or nursing facilities where effective isolation can be implemented. And then it will be important for the FDA to be ready to revoke the emergency-use authorization if viable immunoevasive variants do indeed arise, even if only once.

The FDA and Merck have essentially engaged the public in a gamble without public debate. They are betting that every single mutated virus copy that will be transmitted from patients taking molnupiravir will be neutral, or hurt the virus itself and not its host — that there won’t be even one case of a lucky hit that creates a more capable or evasive virus. This seems like a bad bet, as SARS-CoV-2 has a track record in this pandemic of winning its own bets. But now that the dice have been rolled, we must take every measure we can to use the drug responsibly and quantify its risks. Our ability to end the pandemic may well depend on it.