- If it were only this simple. A cartoon published in the Washington post on July 2nd.
The mask controversy rages on. From the beginning of the pandemic it has been one of the hottest debated topics. The debate has been engulfed in a fire of differing opinions from all walks of life. Research on the topic has been high publicized regardless of the quality. Even the man everyone loves to hate, Donald Trump, has weighed in extensively on the topic.
Most of our experience from mask use comes from observing their usefulness during previous pandemics. The most often sited is the Spanish flu pandemic of 1919. Surgeon’s in a hospital setting use masks to prevent bacteria from their oral and nasal cavity from infecting wounds, and to protect themselves from splatters of blood and other body fluids. Doctor’s in hospitals have used them empirically to protect against known airborne diseases like tuberculosis. The usefulness of masks in a healthcare setting is to prevent spread, yet this has not been rigorously tested. The ability to protect the uninfected wearer remains unknown.
Our understanding of the scientific method has increased substantially since 1919. Data gathering, globalization, and coordination offered by technology has opened many avenues for medical research. COVID-19 has provided an opportunity to put some of these theories to the test.
Although I find the science behind masks interesting, it’s not even close to as interesting as the behavior human beings have exhibited around the debate, which I hope to shed some light on after a discussion of the study.
The DANMASK study was a randomized open-label trial performed in the country of Denmark. The study period was April-June of 2020, which corresponded with the first spike in COVID cases across Europe. Volunteers for the study were recruited via advertisements to participate in the study. Individuals were included if they were > 18 years old, reported spending at least 3 hours outside their home per day, and did not wear a mask for work.
Participants in the experimental group were provided with 50 surgical masks. They were instructed to wear them when they left their home. If they were out of the home for more than 8 hours, they were told to change masks. All participants were advised to practice social distancing, hand washing, and avoid visiting hospitals and nursing homes.
The outcome was positive IGM, or IGG antibodies, positive Sars-Cov PCR, or reported COVID infection in follow-up surveys. The antibody tests were self-administered by the participant one month after study enrollment.
They estimated the prevalence of COVID to be 2% in Denmark at the time of the study. They estimated masks would reduce the infection rate by 50%. For their sample size calculation, they determined they would need 4600 total participants.
This is what they found.
They were able to recruit 6000 total participants. They were evenly distributed between the two groups. There were no significant differences between the two groups, which is indication that the randomization strategy was successful. The average age was 47 and the majority of participants were female (63%) in both groups. Baseline antibody testing was completed by 80% of the participants.
Evidence of infection was found in 42 participants in the mask group and 53 in the control group. This corresponds to a prevalence of 1.7% and 2% respectively, which is close to their prediction of 2%. The 95% confidence interval was (0.53-1.23). This crosses 1 which would accept the null hypothesis as no difference between the groups. The vast majority of these were antibody positive. There were 0 positive PCR and 5 positive PCR in the mask and control group respectively.
They had a significant issue with compliance and loss to follow-up. Only 80% of participants completed the study. The majority of the attrition was due to incomplete participation in the study. Regarding mask compliance, 40% reported wearing it as instructed, 47% wore it mostly as instructed, and 7% admitted to not following instructions.
They performed two important sensitivity analyses to correct for these issues. These are statistical calculations to determine how much the missing data may have affected the results. In the first analysis they included only those participants that wore the mask correctly. The results did not change. In the second they excluded those that did not provide antibody testing at the beginning of the study, the results did not change.
To correct for the missing data, they ran a multiple imputation model. This is basically a model that develops several probable datasets based on the observed event rate and averages the results. If it produces a substantially different result than the one previously observed then the missing data likely played a significant role. In this case the confidence interval was exactly the same (0.53-1.23).
They found that 52 of the participants in the mask group and 44 in the control group had someone they lived with infected with COVID during the study period. This resulted in 2 positive cases in the mask group, and 1 in the control group. Most infections occurred outside the home.
So, what does all this mean?
The authors concluded the infection rate was not reduced by greater than 50% in mask wearers. The study was by no means perfect. However, it was a good attempt at a randomized control trial, which is the gold standard investigational study design in medical research.
Let’s get the limitations out of the way, because there were several. The biggest critics of this study site that the study was designed to study infection rates in the users, not prevention of spread. The purpose of wearing masks, according to proponents, is to prevent spread to others. They had a significant amount of missing data, which they attempted to correct statistically, but this doesn’t replace actually having the data. The event numbers were low, which questions the power of the study. I have a rule of thumb, at least 50 events in each arm before you can tell a difference. They were close. And they met their sample size calculation target. So, they should’ve had enough power, but the numbers were low. Antibody tests were used for the primary endpoint. There may have been false positive and false negative issues with these. However, the manufacturer reported 90% and 98% sensitivity and specificity respectively. In an independent external cohort, the test performed with 82% and 99% sensitivity and specificity. That means, false positives were not really an issue, but false negative rates were high.
Calling the publication of this study controversial is an understatement. The first successfully executed randomized controlled trial surrounding the issue of masks was rejected by Lancet, one of Europe’s most prestigious journals. There were 64,000 tweets and retweets about it the first day of publication. Quickly intellectuals in the medical, political, and academic space dug their heels in. Those that have publicly stated masks are infective were quick to say, “I told you so.” Proponents of masks pounced on the flaws like a pride of lions on a fresh kill.
So, I can only approach this paper from an analytical perspective and conclude it does not definitively answer the question.
I will quickly summarize the key components of the paper form an evidence-based medicine perspective. The study was randomized. Randomization was successful since both groups were equal. Obviously, there was no blinding since one cannot create a placebo mask, and there was no group concealment. This certainly could have changed the behavior of the participants. Mask wearers may have avoided certain situations because they were biased towards proving the hypothesis. Also, they may have taken greater risks because of the mask. Those randomized to no mask may have felt less secure and avoided close contact with others. They had significant loss to follow-up and inappropriate mask usage. Although they attempted to correct for this statistically though modeling, it’s not as good as not having the issue to begin with. The event numbers were again low, which has plagued about every trial I’ve reviewed about COVID, and the confidence intervals were wide. The truth likely lies somewhere along this interval, but we don’t know what that number is. There was no assessment of their ability to prevent spread to others, which is their primary purpose. Overall, the results of the study cannot be trusted. But I applaud their efforts.
It continues to baffle me how low the event numbers are in these trials. This study was executed April-June of 2020 in Denmark, which may have been towards the end of Europe’s first surge. Still, there should have been plenty of virus floating around. One of the most challenging aspects of COVID research has been the behavior of participants in trials. Humans are not acting normally. We aren’t getting on planes. Traveling to foreign lands. Mingling with each-other at various establishements. We are under significant restrictions from local and federal governments. How will we apply the knowledge we gain from research to the real world? We probably can’t because our interventions are not being challenged in research settings. But they are being challenged in the real world and they appear to be falling flat as the pandemic rages on.
Personally, I find the idea of masks far more interesting than the research itself. Why does the mask debate burn with such ferocity? I have a few thoughts. The remainder of this post will be theoretical, so if you don’t care to read about that feel free to take the information you’ve gained and move on with your day.
Masks play into two significant survival instincts. First, you can see it. Only birds of prey have better vision than humans. We rely on it more than any other sense. We can quickly discern safety from danger thanks to our clear view of the world around us. Obviously, you can quickly see if someone is wearing a mask, and you feel safe. Second, it’s an action. We most certainly favor action over inaction, when it comes to our health. Masks fit this bill perfectly. They are inexpensive, “easy” to use, visible, and are an actionable demonstration of your care for others.
But they may be more of an illusion of security than security itself. Honestly, you are less likely to get COVID passing someone on the street than choosing not to go to your friend’s birthday party. Far less likely. In fact, dispersion is the mechanism by which the pandemic spreads. So called “super-spreader events.” One person (likely asymptomatic) quickly spreading it to many others at a single event.
Although you may wear your mask to the event, you likely take it off, because the above conditions are completely flipped. You feel secure around people you know, and you can visually see they are not ill. You care about the hosts of the event enough to not ruin the memory by keeping your distance and having your face and voice obscured by a mask.
Quite the paradox.
When the safest action was actually the inaction of not accepting the invitation. Inaction over action.
I’m not trying to be judgmental of partygoers. More to highlight the idea that you may think this solution is keeping you safe, instead of viewing the other side of the coin, your increased risk taking. Which you accept to return to normal life.
Finally, the mask solution is simple. Masks cut down the cognitive power required to try to solve the problem. The solution is cognitively kind. They are visual, cheap, easily dispersed, and there are anecdotes in other areas that suggest they work. So, they are a classic heuristic. You have to think hard for a reason not to promote them. You also have to think hard about other solutions to the problem. Other solutions require more physical and cognitive effort. So, they are less likely to be promoted.
Also, as the issue gained in prominence you were less likely to be reprimanded socially if you wore a mask (no doubt the largest influence on me). When the CDC director testified in front of congress that masks would end the pandemic in 8 weeks if everyone wore one, it was a simple solution. When Gov. Cuomo got on TV and said, “wearing this says I love you, I respect you.” It was an easy answer. A couple months prior to that the intellectual debate about masks was on full display for the public. Dr. Fauci and the WHO wavered back and forth on their recommendations for the public and would not commit to a mask mandate. They were later criticized for being lukewarm. But they understood that solving the pandemic was far more complex than this. And they were right. The world around you has shown they are a mildly effective tool in this pandemic
Many people have stated to me, “So, what’s the harm? It hurts no one to wear a mask. It’s a small inconvenience. It demonstrates I care about society.” Which is true. I would argue, the illusion of security could be propagating more risk taking, and therefore higher case numbers. To prove my point, a brief thought experiment… How would your behavior change if there were a federal mandate to NOT wear a mask?
Maybe masks provide just enough security, or the impression of security, to overcome the risk threshold required for human beings to interact with one another again. Maybe this is their greatest impact. And in my case, and the sake of my patients, this cheap, low cost, widely available tool is extremely valuable, even if the safety they provide is unknown.
One editorial I read applied Bayesian probabilities to this study. The author proposed that the Odd’s ratio and confidence intervals did not apply to reality because it was not discussed in probabilities. He stated that based on the statistics there is a 93% chance that masks reduce the risk of infection by 5%.
I thought this was pretty accurate description of the effect size, if you can wrap your head around it. They probably help a little bit. They probably help more in certain situations, say an emergency room, a small meeting room, or a cruise ship, more than others. For example, sitting quietly at your desk, or at a “table for two” with your spouse. It's clear, at least to me, the effect has been overestimated. Despite this trial’s effort to enlighten us, I'm sure the debate will rage on.