Are We Wearing the Wrong Masks?

The results of a cluster RCT of surgical and cloth masking conducted in Bangladesh are now available as a preprint (not yet peer reviewed), and have made a huge splash. Here are my thoughts on what the authors found, and more importantly, what it means for us going forward.

The Study

The study randomized 600 villages to cloth masks, surgical masks, or usual care. More than 340,000 people were included. The primary endpoint was reporting symptoms consistent with COVID-19 followed by a positive serology test to document SARS-CoV-2 infection. The intervention included giving people a mask — a surgical mask that could be washed and reused, or a cloth mask of high quality (3 layers) — in addition to role-modeling by community leaders and mask promotion efforts. Readers of this column will know I am happy, as I have been pushing for such a cluster RCT since 2020.

The Results

The primary findings include:

• Cloth masks had no advantage over the control arm (no intervention), but surgical masks showed a modest, statistically significant benefit
• The surgical mask intervention reduced symptomatic seroprevalence by 11.2%; the endpoint — COVID-19 symptoms followed by a positive COVID-19 test — occurred in 0.76% of people in the control group compared to 0.67% for those assigned to surgical mask villages

Some readers highlight another endpoint: having symptoms consistent with SARS-CoV-2, without testing to prove the cause. I strongly disagree with relying on this endpoint.

First, wearing a mask can change how you report symptoms. In medical research, we draw a distinction between bias-resistant endpoints and bias-susceptible ones. How you feel after wearing a mask is influenced by your faith in the mask — it is bias-susceptible. Whether you have antibodies in your blood is not based on your feelings — it is bias-resistant. I prefer my endpoints resistant to bias.

There’s proof in the trial that symptoms without testing is bias-susceptible. We know this because over three quarters of people in this study who reported SARS-CoV-2 symptoms did not actually test positive for SARS-CoV-2! Some may have had other illnesses, but others may have over-reported symptoms, particularly those who did not mask.

There is one more extremely important study endpoint that is not yet reported: the rate of seroprevalence among random villagers. This is even more bias-resistant, correcting for a possible oversampling in groups that report more symptoms. We shall see what it says — the research is still ongoing. Yet, with the results to date, the following are a few broad lessons we can take from this study.

We Wore the Wrong Masks

All those selfies of people wearing fashionable cloth masks; all those videos and memes of how to make a mask from a sock or t-shirt — those were all misguided! We spent massive political capital on the wrong mask. We pushed mask mandates and guidelines that mostly got Americans to wear any type of mask at all when certain types did not work in this study.

Had we done this study a year ago, we would have been able to provide important health guidance. We would have been able to distribute surgical masks to all Americans or, at a minimum, high-risk individuals. We could have discouraged bandanas, gaiters, and cloth masks, and focused on the mask that works: surgical masks. It isn’t too late. The CDC should immediately update all their guidance, and click on find and replace: “cloth masks” to “surgical masks.”

We Can Do Cluster RCTs

The second lesson of the study is that cluster randomization is possible, feasible, doable, and useful in a global pandemic. Knowing the right mask to use is a lamp-post in a sea of darkness. The Bangladesh study shows that even in a resource-poor setting, such trials are possible. Now imagine similar trials in key settings: U.S. schools, daycares, offices, and communities.

Open Questions

There are some notable differences between the U.S. in this moment and Bangladesh at the time of the trial.

First, the obvious: we are not Bangladesh. We are a nation where masking is the norm in some regions and areas, and strongly opposed in others. Would receiving surgical masks and instructions “work” the same in the U.S.? That depends on the cultural success of the intervention in our nation.

Second, the trial compared 13.3% masking (control arm) to 42.3% masking (experimental arm), but the effect may be different if you move from 60% masking to 90%.

Third, the trial was conducted in a region where essentially 0% of people were immune to the virus. Many parts of the U.S. are different, and enjoy much higher rates of vaccination and recovered infection. 79% of people have received at least one dose of the vaccine in San Francisco, for instance. The absolute risk reduction will surely be smaller, and even the relative risk reduction will approach the null, if vaccinated people transmit less than unvaccinated people (psst: this is true). A new question then emerges: do masks work in highly immune populations?

Fourth and finally, this was a mask intervention for adults, and measured infection among symptomatic adults. More cluster RCTs are needed to fully understand how this applies to schools and children.

Little Got Them to Wear the Mask More

Besides the intervention itself — getting free masks and education about using it — which did raise mask use by 30 percentage points, one of the disappointing findings is that none of the sub-study interventions worked to improve mask adherence besides the color of the mask. Altruistic messaging, self-interested messaging, texting, and verbal promises all amount to a hill of beans. This is disappointing and suggests that there may be practical limits to masking policies, depending on the locality or country where they’re implemented and the interventions used.

They Could Not Sustain It

Just 3 months after investigators left, mask usage plummeted, with the bulk of the effect (approximately two-thirds of increase in mask use) being lost. People in these Bangladesh villages were unable to sustain masking after the study ended at the rates seen when the study was ongoing. That shows another limit of the practice. We need to understand what types of interventions lead to long-term adherence to mask-wearing.

Economists Succeeded Where Public Health and Medicine Failed

One of the big takeaways of the study is that a cluster randomized trial led by healthcare economists succeeded in providing important answers. Economists did what the CDC, public health institutions, and venerable medical organizations failed to do: run a cluster RCT of a non-pharmacologic intervention. In my opinion, their results do come a little late. Had we had this result in the fall of 2020, and had potentially more U.S.-based studies, we could have leveraged it immediately and encouraged broader use of surgical masks. But when the history books are written, we will look back on the failure of medicine and public health to study this question, and economists will get the praise.

Precautionary Principle

Throughout this pandemic some have argued that we don’t need RCTs on questions like masking because we should follow the precautionary principle and err on the side of caution, and just do it. I agree that when a situation is uncertain, you can follow the precautionary principle, but it is time limited. When you extend unprecedented restrictions and mandates into the second year, you have some obligation to know if they help, and if so, are they worth it?

The precautionary principle is like parking your car on a city street. You can get away with it for a night or two, but if you don’t move it for a year, you are going to get towed. Similarly, there are many questions that immediately follow the Bangladesh study: how can we encourage broad use of surgical masks? Should schools require surgical masks for kids and faculty? Should vaccinated people wear masks, and, if so, when? Science owes it to society to run the studies capable of adjudicating these questions. We learned a lot from Bangladesh, but there is more yet to know. We need more cluster RCTs.

Vinay Prasad, MD, MPH, is a hematologist-oncologist and associate professor of medicine at the University of California San Francisco, and author of Malignant: How Bad Policy and Bad Evidence Harm People With Cancer.