In an interview in March about In Covid’s Wake, Stephen Macedo and Frances Lee were asked what they did to “stress test” their “analyses” and who they spoke to. Lee explained that their “approach to doing the research was not to interview.” Instead, she and Macedo “attended carefully to the whole body of research as best [they] could reconstruct it.” Several aspects of their essay illustrate the problems with this approach to truth-seeking and the responsible deliberation they call for.

Start with their figures. The authors claim that because overall mortality up to April 2021 was similar across U.S. states, the more stringent COVID-19 policies implemented by Democratic-governed states had little effect. This is a flawed methodology for evaluating the impact of infection control measures, something that any infectious disease epidemiologist would have pointed out if consulted.

Cumulative mortality depends on both the effectiveness of control measures in reducing transmission and on the time when those measures were introduced, relative to the underlying epidemic. In spring 2020, places like New York City got hit early. So even if later measures had been very effective at reducing transmission, these places would still have substantial total mortality. Given this context, Macedo and Lee’s figures actually offer evidence in favor of the opposite conclusion to the one the authors draw: Democratic states like New York and Massachusetts got hit hard early on, then transmission came down as measures were implemented, while Republican states like Florida and Texas got hit later, and transmission then accelerated under less stringent measures.

Macedo and Lee are also wrong to infer that non-pharmaceutical interventions (NPIs) had little effect from the fact that “reopening did not trigger immediate viral resurgence.” Because COVID-19 transmission involved “superspreading,” with some individuals infecting many and most infecting few or none, we would not expect an immediate, clockwork-like increase in transmission after measures were lifted. Instead, relaxing restrictions would take time to translate into epidemic growth—which is indeed what we saw across Europe in the second half of 2020.

Some previously published analysis has assessed how NPIs affected transmission, rather than just looking at cumulative mortality. In 2021, two studies examined the relationship between different NPIs and transmission changes by comparing the timing of measures across Europe in the spring and autumn waves in 2020. These studies estimated that no single intervention reduced transmission by 50 percent or more, but an epidemic wave could be suppressed with a combination of measures. Closing businesses reduced transmission by an estimated 25 to 35 percent, while strict limits on gathering size reduced transmission by 25 to 40 percent. Once venues had been closed and gatherings restricted, analysis estimated that stay-at-home orders reduced transmission by an additional 10 to 15 percent.

Another 2021 study, focusing on U.S. counties, looked at seven NPIs: stay-at-home orders, leisure activity closures (gyms, restaurants, and movie theaters), school closures, daycare closures, mask mandates, nursing home visitation bans, and the suspension of certain medical services. Researchers found that all NPIs but the last were associated with a reduction in transmission. They also noted that there may have been indirect explanations for such reductions: “School closures, for example, may have had substantial impacts on the social interactions of nonschool-aged individuals as parents and workplaces adapt to accommodate changes in children’s schedules.”

On the matter of pre-COVID pandemic planning, Macedo and Lee are correct that the 1918 influenza pandemic loomed large in shaping scientific attitudes about the feasibility of society-wide infection control. Even if an epidemic could in theory be slowed by reducing social interactions, these pandemic plans generally deemed it impossible, in practice, to reduce interactions to such a degree that a growing influenza-like pandemic would be quelled. In short, NPIs were assumed to be able to slow a pandemic wave but not stop one.

This assumption turned out to be woefully incorrect about COVID-19, however. The possibility of suppressing a COVID-19 wave drew skepticism throughout February 2020, but a lot had changed between 1918 and 2020. First China, then other countries in southeast Asia, and then countries in Europe showed that it was in fact possible to prevent an uncontrolled epidemic using transmission-reducing NPIs. In Wuhan, social interactions dropped 87 percent during lockdown; in the UK they would later decline by 74 percent. This marked a paradigm shift in pandemic response and a new dilemma for policymaking. My colleagues and I at the Centre for Mathematical Modelling of Infectious Diseases at the London School of Hygiene & Tropical Medicine were among the first groups to estimate the impact of drastic control measures in China in late January 2020. When we announced our results, I noted that these were “unprecedented interventions that will have had a huge social and psychological toll.”

Lockdowns are indeed a terrible policy option. They are blunt, costly, and harmful, including in ways that Macedo and Lee note—ways that many people tracked and discussed in 2020, including in these pages. At best lockdowns are a holding strategy, because most of the population remains susceptible to a second wave (at least before the development of a vaccine). But a massive epidemic that kills and hospitalizes millions, collapsing health systems in the process, is also a terrible, costly, damaging policy option.

Could a policy that did not attempt to suppress transmission have worked better, as the authors of the Great Barrington Declaration implored? Already by February 2020, it was becoming clear that COVID-19 severity varied strongly by age. This prompted several research groups, including ours, to model the potential impact of strategies attempting to “shield” older groups during epidemic waves in order to reduce disease with less disruption to society at large. But around 1 in 6 people in the UK are over the age of seventy, and this demographic has a large volume of social interactions with the wider population. Even if the elderly reduced their social interactions by 75 percent, we estimated that peak demand for intensive care beds would still have exceeded available capacity many times over.

Even with restrictions, critical illness quickly overwhelmed health services in many areas throughout the world during 2020 and 2021, reducing the quality of care and making survivable conditions fatal. Analysis based on detailed testing in the UK suggests that before vaccines became available, around 1 in 3 people over the age of seventy-five who got infected were hospitalized, and around 1 in 6 died. Among people aged sixty-five to seventy-four, around 1 in 10 of those infected were hospitalized and around 1 in 30 died.

In October 2020, the Great Barrington Declaration argued that “those who are not vulnerable should immediately be allowed to resume life as normal,” which would imply a large epidemic wave to follow. By this point, we had emerging evidence that vaccines generated neutralizing antibody responses and that such responses could reduce the risk of coronavirus infection. With a vaccine on the near horizon, was having a very large epidemic while trying to shield vulnerable groups really a workable, let alone optimal, strategy?

The authors of the Declaration never outlined precisely what they thought would happen if such a policy were to be implemented, or even how exactly it should be pursued. Suppose protection were aimed at those older than sixty-five and others with health conditions that make them clinically extremely vulnerable. In the UK, this comprises just over 20 percent of the population. An equivalent percentage would translate into over 65 million people in the United States. It is very hard to imagine preventing a high volume of infections among lower-risk groups from spreading to such a large number of high-risk individuals. In 2020, there were in fact attempts to “shield” vulnerable groups in many countries. But large epidemics still spilled over because not all transmission routes were prevented, as well as the epidemic causing severe illness in those not deemed vulnerable. Even if social interactions are reduced dramatically, it’s extremely difficult to keep infection away from vulnerable groups when infection levels are high.

I suspect that the acceptability of a large epidemic wave as an option in the Great Barrington Declaration in part stemmed from an underestimation of the risk that COVID-19 posed. Two of the coauthors, Jay Bhattacharya and Sunetra Gupta, had previously suggested the average fatality risk across all ages was closer to 0.1 percent or even 0.01 percent overall, rather than the 1 percent that was in fact observed during the first waves in the UK and the United States.

Instead of engaging with the logistical and epidemiological plausibility of their proposal, the Great Barrington authors resorted to soundbites about lockdowns favoring the “laptop class”—a term also adopted by Macedo and Lee. This is a classic motte-and-bailey argument: a bold, controversial claim is made (somehow vulnerable groups can be shielded, despite everyone else returning to normal, and this was the optimal strategy to take in October 2020), but when challenged, the argument shifts to a more easily defensible claim (lockdowns had negative consequences).

There is no doubt that lockdowns had unequal impacts. But the epidemic itself did too, even in the absence of lockdown. Lower-income and working-class neighborhoods in pre-lockdown New York City and Chicago suffered most disease and death early in 2020. Among people of working age in Sweden, those in the lowest income tertile were five times more likely to die from COVID-19 than those in the highest. Impacts have also been unequal historically, such as during the 1918 and 2009 influenza pandemics.

Good debate requires good evidence. Any suggestion that a large epidemic is preferable to suppressing transmission must give a realistic assessment of disease risk. It must also engage with evidence available from other countries. Macedo and Lee notably fail to mention the experience of Vietnam, Australia, Japan, South Korea, or even Sweden. All these countries brought transmission down with a combination of behavioral change and control measures; as a result, only a small proportion of their populations were infected in 2020. All, in their own way, showed that pre-COVID pandemic planning was inadequate.

Countries like the United Kingdom and the United States do need a better strategy for the next pandemic. Relying on blanket lockdown-type measures again would betray a failure to learn and prepare, including by scaling targeted measures that can break chains of transmission without widespread disruption and investing in vaccine platforms that can provide immunity sooner. But equally, if countries avoid any efforts to curb transmission, and listen to those claiming a pathogen is much less severe than it really is, or that a large epidemic would be much less damaging that the global evidence suggests, or that interventions are ineffective because deaths occurred before they were introduced, that will also be a failure.