Paul Sax is an infectious-disease physician and professor at Harvard Medical School. Even so, when Sax’s kids asked him whether they should get the updated COVID vaccine this fall, he wasn’t really sure what to tell them. His two children are in their 20s, healthy, and at no special risk of complications from disease. Each had recovered from COVID, and thus, he reasoned, had extra immunity on top of what they’d gotten from their prior shots. Another injection would likely cause them a day or two of unpleasant side effects, and expose them to a very small risk of heart inflammation. Would it also meaningfully lower their chances of infection or the severity of their symptoms if they did get sick? Would an extra shot reduce their minuscule odds of death?
Sax admits he didn’t have “iron-clad data” to back up his thinking either way. The bivalent vaccines had been authorized only on the basis of how many antibodies they produced, a useful but imperfect gauge of their positive effects—and so, even now, no one knows for sure how much real-world benefit they provide for the majority of Americans who are already shielded by vaccination and prior infection. Studies have shown that, going by antibody levels, recovering from COVID is already a good booster. There’s another problem, too: The new shots are designed to target the BA.4 and BA.5 subvariants of Omicron, but these now account for fewer than one-seventh of infections in this country, according to the CDC. The FDA believes that the boosters are still “likely to help,” but studies hint that the shots are already out of date.
Sax did end up recommending the latest dose to his kids, because there was no way of knowing exactly how much immunity they’d gotten from their infections. But the simple fact that he, like other experts, has been expected to make such decisions on the basis of conflicting laboratory data is pushing doctors back into the disquieting realm of scientific uncertainty. It suggests that we’ve reached a turning point in our pandemic expertise, where some of the astounding progress made since 2020 has started to decay.
At the outset of the crisis, the world’s scientists used their grit and genius to develop new ideas with unprecedented speed; collectively, the COVID vaccines and treatments they produced saved tens of millions of lives. But their historic push for knowledge has lately slowed and sputtered in its tracks. Society spent billions of dollars to answer a single, urgent question: How do you combat a novel respiratory virus? Now, all of a sudden, we find ourselves a little baffled by the follow-up: How do you handle a respiratory virus that is familiar?
The first COVID shots were the culmination of a project that involved governments and research institutions all around the world—a human, and humanitarian, achievement on the grandest scale. At least 12 different vaccines have since been cleared for human use. They didn’t just arrive in record time; they were also of extraordinary quality. Rigorous clinical trials demonstrated that immunization could reduce the risk of symptomatic disease by up to 95 percent.
Two years later, though, those claims no longer hold. We’re now facing down millions of infections—with no clear sense of our tweaked vaccines’ effectiveness for most of the people who take them. What’s pushing COVID science backwards? Don’t blame viral evolution—or not entirely. The emergence of new subvariants does weaken the effects of our vaccines. (It may also render some monoclonal-antibody treatments obsolete.) But the bigger problem isn’t that the virus has become a stranger. It’s that we’ve come to know it all too well.
Most of the groundbreaking research that led to our current vaccines and treatments was performed in a type of human that no longer exists in any but the smallest numbers: Homo uninoculatus uninfectus, which is to say, a person who has neither gotten sick with COVID nor ever taken a vaccine against it. The original vaccine studies by Moderna and Pfizer excluded participants who were known to have caught COVID. Paxlovid was authorized based on a study of unvaccinated subjects. The other antivirals, too, were tested only in those who hadn’t gotten any shots. Yet here’s where we are right now: Seven in 10 Americans have received at least a primary vaccine series, and more than 95 percent have SARS-CoV-2 antibodies from vaccination, infection, or both. Globally, 13 billion shots have been administered, and nearly every country has suffered through widespread disease. We developed all these drugs for a world of COVID virgins, and now that world is gone.
This is very good news: Patients with preexisting immunity are at far lower risk. New infections are a lot milder, on average, than they were in 2020. Fewer patients are entering ICUs with lungs damaged by the severe pneumonia we saw at the start of the pandemic, and those hospitalized with COVID are trending older and sicker overall. Clearly we’re living through a new—and better—phase of the pandemic. But the same development also makes it harder to figure out whether vaccines are still doing what they’re meant to do.
Recall the headline finding from the original mRNA-vaccine trials, that the shots’ ability to prevent symptomatic illness was an incredible 95 percent. That figure, as measured in real-world populations, has taken a nosedive ever since. The initial two-shot series staved off as few as one in three infections in nursing homes during the Delta wave, and offered essentially no protection against mild illness by the time Omicron arrived. That certainly doesn’t mean the shots are worthless, and that we might as well have been receiving saline injections all along. Rather, it suggests two things. First, vaccine-induced immunity weakens over time: Antibody levels dwindle and SARS-CoV-2 may get better at eluding the protections that remain. Second—and this is the source of much uncertainty today—the shots’ continuing effects may seem to go away as more people get infected and form immune responses on their own.
Here’s how that works. A vaccine’s effectiveness is measured by comparing COVID outcomes between those who get the shots and those who don’t. When many people in the unvaccinated group have natural immunity, the vaccine’s benefit could become obscured. Now that almost all Americans are estimated to have caught SARS-CoV-2 at least once, we have no good way of controlling for this effect—and it may be very large. Just look at Hong Kong: When the city suffered an Omicron BA.2 onslaught last winter—after having suppressed most of the previous outbreaks—two doses of the Pfizer vaccine still proved 89 percent effective at preventing severe disease in a largely immune-naive population. The same two doses, measured against BA.1 and BA.2 in North America, appeared to provide as little as 34 percent protection against severe disease.
This would already leave us in an awkward and uncertain spot, even if the newest doses of the vaccine hadn’t been updated in the absence of rigorous, large-scale clinical trials. Instead we have preliminary data from the CDC suggesting that the bivalent boosters offer modest protection against symptomatic illness. Many questions remain. The study doesn’t measure the shots’ effects on staving off severe disease, and the results weren’t broken down by prior infection status. The data were also collected from mid-September to mid-November, when BA.4 and BA.5 were still predominant and antibody levels were at their peak, so we don’t know how well the shots are going to hold up through the winter.
Our current, diminished state of COVID knowledge extends beyond vaccines. COVID treatments, too, have been subject to a reemerging ignorance. Like the mRNA vaccines, antiviral drugs were conjured through historic effort, and introduced to clinics with reams of data to support their use. A randomized trial performed across 343 research sites around the world taught us that Paxlovid could nearly eliminate pandemic hospitalizations. Other treatments like molnupiravir and remdesivir were also shown to help keep patients from becoming severely ill through studies spanning many different countries. But in 2022, those findings are no longer so definitive.
If you get sick with COVID now, a doctor like Harvard’s Sax might prescribe an antiviral drug, but wouldn’t know for sure how much it will reduce your risk of hospitalization or death. Todd Lee, an infectious-disease physician at McGill University, illustrated the uncertainty this way: “When President Biden gets COVID again,” he asked me—emphasizing the word when—“should he get Paxlovid?” As an 80-year-old patient who has gotten five doses of vaccine and recovered from a prior Omicron infection, Joe Biden’s case exists outside existing evidence. Biden’s doctor will probably prescribe another round of Paxlovid—but limited data support its use. Same for molnupiravir, remdesivir, or monoclonal antibodies.
The pivotal trial for Pfizer’s antiviral drug Paxlovid showed that it could prevent 89 percent of hospitalizations or deaths in high-risk adults without a history of vaccination or infection. Yet among the subset of participants in the Pfizer trial who were found to have natural COVID antibodies (perhaps implying that they’d been infected unawares), the drug seemed to prevent fewer bad outcomes. A similar finding was seen with Merck’s molnupiravir pill. Because most Americans now fall into that antibody-positive category, can we assume that our antivirals will continue to play as important a role for us as they did before? Lee doesn’t think so. People with some form of immunity are already at low risk of hospitalization, so he thinks that any treatment will offer diminishing returns; beyond that, the specifics are unclear. Even attempting to quantify treatment’s ever-diminishing benefit with larger trials is “perhaps a fool’s errand,” he said, “because we’re so successful with vaccination and because we have all these people who have survived natural infection.”
We can all rejoice at the source of this uncertainty—that most of us are at lower risk from COVID than we’ve ever been. But our scientific ignorance still has dangers of its own. “One thing you worry about is getting behind the eight ball,” Sax told me. We can’t yet make good predictions about how SARS-CoV-2 will evolve or how devastating the next wave of disease will be. And however the pandemic might develop or recede, elderly and immunocompromised patients deserve the most relevant vaccination and treatment protocols—because they’re the ones who bear the brunt of medical uncertainty.
There are ways to rebuild expired COVID knowledge. The PANORAMIC trial, with more than 26,000 participants in the United Kingdom, is an example of one approach: the large randomized, controlled study.
PANORAMIC, which includes people who have been vaccinated or previously infected, aims to identify whether antiviral treatments help today’s COVID patients recover sooner or stay out of the hospital. The results so far have been mixed: Among the 99 percent of the study’s participants who are vaccinated, researchers found zero benefit from taking molnupiravir, when measured by hospitalization and death rates, but the drug did lead to a faster improvement in symptoms. New data on Paxlovid are forthcoming.
In the meantime, doctors will remain at sea on how to treat patients who already have some immunity to COVID. Hybrid immunity—the combination of vaccination and a prior infection—seems to offer the strongest, most durable protection. But the incremental benefit of repeated boosters or treatment on top of this is less certain. Public-health agencies and professional groups aren’t giving much direction. Right now, the CDC recommends that everyone age 5 or older get the bivalent booster, regardless of prior infection status. (The agency does say that patients “may consider” delaying the latest shot until three months after they last got sick with COVID.) Treatment guidelines from the National Institutes of Health and the Infectious Diseases Society of America are no help, either. NIH merely notes that “the efficacy of these treatments in patients who have been vaccinated is unclear,” while IDSA describes the value of treatment for immune populations as a “critical unanswered question.” Yet this unanswered question now applies to nearly everyone on Earth.
Pharmaceutical companies don’t have much incentive to sort this out, given that their treatments are widely available and reaping record profits. “I don’t think they want more data because more data might show the drugs don’t work,” David Boulware, an infectious-disease physician at the University of Minnesota, told me. I asked Pfizer and Moderna whether the companies had plans to run new, large-scale clinical studies of their COVID shots and drugs in vaccinated and previously infected individuals, but neither responded to my inquiry.
If pharmaceutical companies don’t continue to evaluate their therapies, independent scientists will have to pick up the slack. The PANORAMIC study is being conducted by academic researchers with the cooperation of governments and health systems. Trials like that one can provide important patches to our eroding expertise, but they are too costly and impractical to answer every question. Studies that measure how well a person’s antibodies and immune cells inhibit SARS-CoV-2 in the lab can plug some other holes. James Crowe, a physician and vaccine scientist at Vanderbilt University, told me he’s optimistic that this form of investigation will soon pay dividends. Scientists have become adept at measuring whether our vaccines create an effective immune response, he said, and their test-tube data will be useful even in the absence of clinical information. After all, we don’t run large-scale trials for the flu shot every year.
For most healthy people, however, the stakes of having access to this knowledge are diminishing. Many recent COVID deaths occurred not because of COVID pneumonia itself, but because the virus worsened preexisting lung, heart, kidney, or liver conditions. This puts SARS-CoV-2’s behavior more in line with other routine respiratory infections. Although the disease remains a serious threat to the elderly and chronically ill, and many are still subject to long (and medium) COVID, a growing share of the population appears to be out of harm’s way. “We’re victims of our own success,” Lee told me. We developed an incredible global research apparatus overnight. Vaccines became widely available—if too slowly and inequitably—and they did what they were supposed to do: keep people out of the hospital while the population built up immune protection.
Uncertainty does not imply paralysis. People who face the highest risks from the virus can top up with the latest booster, or take Paxlovid when they’re sick, despite not knowing how effective either option will be. The early studies showed that, fundamentally, these therapies perform. But to optimize their benefit—to save as many lives as possible and prevent additional suffering while avoiding unnecessary treatment—we’ll need more and better data.
These answers will not be forthcoming if endemic COVID leads to a return of endemic science, which faces little public appetite to study existing drugs, and bureaucratic roadblocks for promising new ones. Reverting to the default state of medical knowledge means making important decisions by instinct, authority, or flawed real-world observations. The pandemic was an exception to this rule: Scientists, industry, and government banded together to generate real answers very quickly. But instead of applying this engine of discovery to every disease imaginable, society seems on the cusp of dismantling it.