In the past couple of weeks, experts and pundits have debated a big question surrounding the rollout of the Covid-19 vaccines from Pfizer/BioNTech and Moderna: Do we stick to two doses for every person, spaced a couple of weeks apart, as was originally planned? Or should we go ahead and give out single doses to more people even when we’re not sure we’ll manage the second dose as planned — to get a shot in as many arms as possible, even if there’s less certainty about the efficacy of that approach?

Britain has taken the latter approach, and much of Europe looks inclined to follow. But the US Food and Drug Administration (FDA) said Monday that the US will not follow them — at least not until we’ve collected more data. High-profile scientists including Dr. Anthony Fauci have said it would be a mistake.

To understand this debate, let’s rewind the tape to mid-December, when the FDA issued an emergency use authorization for the Pfizer/BioNTech mRNA vaccine against Covid-19. A similar vaccine from Moderna was approved shortly after. Both are highly effective at protecting people from getting sick and dying from the coronavirus. But we are just a few weeks into the rollout and we don’t have enough doses, with some 18 million shipped so far for a population of about 330 million — and lots of at-risk people aren’t yet vaccinated. Meanwhile, a more contagious virus variant is spreading, and vaccine distribution in the US is going more slowly than initially planned.

Some prominent scientists and public health researchers have proposed adapting the US’s vaccine approach in response. They suggest getting as many people the first dose as possible, even if that means there aren’t doses available to deliver the second shot on schedule (three or four weeks after the first dose, depending on the vaccine). Both Pfizer and Moderna tested the efficacy of giving people two doses of the vaccine, spaced three weeks (Pfizer) or one month (Moderna) apart. This two-dose regimen is 95 percent effective against the virus. But according to data from their trials, getting just one dose still reduces a person’s chances of getting Covid-19 by 80 to 90 percent, at least initially.

That finding is what has prompted the proposal that we delay giving people second doses until everyone has a first dose. The idea is simple: You have two people eligible for a vaccine. Does it make sense to vaccinate the one who’s already been vaccinated with a second dose, or the one who hasn’t been given a dose at all? By using a two-dose regimen, we’re choosing the one already vaccinated, despite the vaccines’ scarcity.

Britain has already moved ahead with such an approach, delaying second vaccinations with the Pfizer and AstraZeneca vaccine for as long as three months in order to get more people the first dose.

The debate over this proposal in the US gets at many of the biggest challenges of doing science in the middle of a pandemic like this one. We like to approve drugs and vaccines only once we’re highly confident in our approach — but in a pandemic, we sometimes have to guess with our best informed estimate, not with a rigorously demonstrated solution. Waiting months for additional research could cost lives, so we may have to decide what we’ll do in the interim as we wait for that work.

This debate is important in its own right: Figuring out the right way to vaccinate will mean the pandemic ends weeks or months sooner. But it’s also a debate about something deeper, a challenge that has recurred throughout the pandemic: How should we think about expertise and policy in the face of incomplete data and uncertainty?

An mRNA vaccine like the one from Pfizer and Moderna contains a strand of RNA that the body interprets as instructions to build a protein. That protein is a key one on the virus that causes Covid-19, and once the instructions are executed, the immune system sees the unfamiliar protein and learns how to mount an immune response.

In the clinical trials, Pfizer and Moderna tested the efficacy of their vaccines by giving two doses, one month apart, to people in the trial group, and giving two placebo shots to people in the control group. Their research was aimed at estimating the efficacy of the whole regimen. But they collected data on outcomes throughout the trial, which means we do know how well the first shot prevented Covid-19 by itself, during the weeks before the second shot was delivered. And the answer is that it worked reasonably well.

“We do not know for sure, but for at least a month or more, a single shot mRNA vaccines should provide ~90% protection,” Yale virologist Akiko Iwasaki wrote in a Twitter thread calling for delaying the second dose, citing an additional Moderna analysis of patients who for whatever reason missed their booster shot. Iwasaki’s estimate of 90 percent effectiveness is in the ballpark of estimates from other experts, though it’s an extrapolation from limited data (more on that below).

“Giving 100 million people — particularly those at high risk — a single shot that is 80 to 90 percent effective will save far more lives than giving 50 million people two shots that are 95 percent effective,” Brown University School of Public Health dean Ashish Jha and University of California San Francisco Department of Medicine Chair Robert Wachter argue in a Washington Post op-ed.

There’s not much data on the question of how long the immunity from the first dose will persist without the second dose. But, Wachter and Jha argue in their piece, we don’t see immunity start to wane over the weeks after the first dose, and “experts believe it is extremely unlikely immunity would somehow plummet by week eight or even week 12 following a single shot.” And hopefully by the time immunity does start to decline, there’ll be greater availability of vaccine doses, and booster shots can be provided to everyone who needs them.

What’s uncontroversial among public health officials is that giving two doses, as studied, is the ideal thing to do. We know that the two-dose regimen works. The evidence points to protection from two doses being stronger than protection from one dose. Even the strongest proponents of delaying the second dose agree that in an ideal world, the second dose would happen on time — and some of them have argued that for the highest-risk populations, such as those in nursing homes, it’s worth doing a second dose even if the second dose is being delayed for the general population.

The dispute is over what to do given the deeply non-ideal situation we face. There aren’t enough vaccines for everyone who wants them. It may be many months before there are. In the meantime, a new, more contagious variant of the virus is spreading, and hospitals are overwhelmed. Ideal solutions, argue proponents of delaying the second dose, aren’t what we have available.

But the US looks unlikely to follow Britain in delaying second doses of the vaccine. In a statement on Monday, the FDA directly addressed the debate, coming down against delaying the second dose.

“[A]t this time, suggesting changes to the FDA-authorized dosing or schedules of these vaccines is premature and not rooted solidly in the available evidence,” its statement reads. “Without appropriate data supporting such changes in vaccine administration, we run a significant risk of placing public health at risk, undermining the historic vaccination efforts to protect the population from COVID-19.”

After noting that the evidence about the lasting efficacy of a single dose is very limited, the agency statement argues, “If people do not truly know how protective a vaccine is, there is the potential for harm because they may assume that they are fully protected when they are not, and accordingly, alter their behavior to take unnecessary risks.”

“We must not snatch defeat from the jaws of victory by administering vaccines in any way besides that for which they were carefully evaluated,” Georgetown virologist Angela Rasmussen and University of Alberta infectious diseases doctor Ilan Schwartz argued in the Guardian, citing worries that we don’t know how long immunity lasts and that partial immunity could invite the spread of strains that the virus doesn’t address.

Other experts told the New York Times that they worried delaying the second dose would undermine trust in vaccination and increase vaccine hesitancy, making it harder to get everybody vaccinated.

“Even the appearance of tinkering has negatives, in terms of people having trust in the process,” Natalie Dean, a biostatistician at the University of Florida, told the Times. “The longer the duration between doses, the more likely people are to forget to come back,” she said, adding, “Or people may not remember which vaccine that they got, and we don’t know what a mix and match might do.”

Dr. Moncef Slaoui, scientific adviser to Operation Warp Speed who has himself said Operation Warp Speed is considering using half-doses of the Moderna vaccines, said in an emailed statement to the Times on Sunday that “the approach some countries are taking of delaying the booster shot could backfire and could decrease confidence in the vaccines.”

Given the absence of data, some have advocated doing full-fledged studies comparing the efficacy of one dose versus two. That’s what physician Peter Bach argues at Stat News: “this study could enroll 30,000 participants in a handful of weeks and start generating insights in a few months. If we move on this question, and I can’t think of a more important one to prioritize, preliminary data could be in hand by March.”

The FDA’s statement, which mentioned half-dosing and delayed dosing as “reasonable questions to consider and evaluate in clinical trials,” does not seem to rule out such a study, but no one has yet announced plans to conduct one.

The pandemic has repeatedly challenged some assumptions of science communications, where journalists confidently report what is evidence-based and what has “no evidence.” That vocabulary has shaped the conversation about vaccine dosing; for example, Pfizer has said there is “no data” to demonstrate that the immunity from the first dose lasts more than 21 days.

It’s true that length of immunity from the first dose is not what their clinical trials studied. But scientific evidence isn’t all-or-nothing like that. In many cases, scientists (and historians, and researchers in every field) have deeply limited evidence about the topic of interest. Often they’re in a position of trying to make inferences from the results of other similar events, from a few unusual data points that arose due to accidents, or from measured results that were not the primary aim of the trial they were measured in.

Sources of evidence like these are being employed when researchers try to estimate the likely duration of immunity from a single vaccine dose. The UK’s Working Group that made the delayed dosing recommendation “concluded that vaccine efficacy will be maintained with dosing intervals longer than 21 days … based on clinical trial data that showed the vaccine was 90.5 percent effective against preventing Covid-19 after the first dose once the protection that starts at around 12 days kicks in, and there was no evidence to suggest that the effectiveness of the vaccine is declining toward the end of the 21-day period following the first dose,” a spokesperson said.

From one perspective, that study produced “no data” about effectiveness past 21 days. For another research team, it shows there’s “no evidence” of declining effectiveness.

It seems reasonable to move past claims of “no evidence” and instead acknowledge we’re dealing with limited evidence — not ideal, but not total ignorance either. We do know some things: Booster shots are common for vaccines, but usually to make immunity years-long or lifelong; in general, immunity doesn’t vanish in the space of just a few months, and immunity from Covid-19 after contracting it looks like it’s typically longer-lasting than that (though vaccines with mRNA are new and we have less evidence about their long-term effects on immune function). The evidence from the few participants in trials who missed their booster shot is limited, but it’s still evidence.

“It’s been a repeated problem throughout the pandemic that we have relied on this ‘we have the evidence/we don’t have the evidence’ binary, meaning that we have moved slowly, waiting for rock-solid confirmation; but moving faster, making decisions on imperfect information, would have saved a lot of lives,” British science journalist Tom Chivers argued in an article Wednesday about the difference between the US and UK approaches to vaccination.

While randomized controlled trials (RCT) are incredibly valuable for determining how well vaccines work, decisions about vaccination in a pandemic require many judgment calls that we unfortunately cannot assess in advance with an RCT. And to be clear, concerns like whether the change in vaccination schedules will limit public confidence in the vaccine, increase vaccine hesitancy, or prompt individuals to take risks, believing themselves fully immune when they aren’t, are very legitimate and should no doubt feature in our reasoning and planning about vaccination.

But those concerns have also not been rigorously demonstrated in an RCT. No one has conducted a controlled study on how delaying the second dose will affect vaccine hesitancy. Instead, public health officials who raise vaccine hesitancy concerns are looking at lots of factors and making their best informed prediction about what will happen — and public health officials who estimate that delayed second dosing is worth it are doing the same thing.

It’s tempting, when faced with decisions of this magnitude, when lives are at stake and the pandemic has frayed our nerves and our trust, to want to retreat to the solid, reassuring embrace of perfect knowledge. But almost all the key questions before us will never be answered to our satisfaction in an RCT, at least not immediately. What increases vaccine hesitancy? How much risk compensation will there be? Will delaying the second dose save lives by getting vaccines first to those who need them?

There’s no proven truth to retreat to at this moment, just our best guesses — and the fight to gather more information and make better guesses next time.