When I saw the infographic for the DAMASK trial (below), it was immediately clear to me that the trial was inconclusive (not positive, not negative – inconclusive). According to this study, masks might be totally worthless. Or masks might reduce infection by ~40%. This is an inconclusive study. It yields no answer.
The study has a myriad of major problems. Nonetheless, it was published and has garnered some scattered support from orthodox proponents of evidence-based medicine. After all, it’s a randomized study, right? Isn’t that the pinnacle of evidence?
The DAMASK trial is so fatally flawed that it doesn’t deserve further discussion (more analysis of it can be found here). A more interesting topic to explore is why it’s impossible to perform a randomized trial of masks.
How effective would masks need to be, to make wearing them worthwhile? That’s a tricky question. For the sake of argument and simplicity, I’ll argue that if masks cause a 1% relative reduction in the rate of infection, that would be worthwhile. Right now in the United States, a 1% relative reduction in infection would cut the infection rate by 1,750 people per day and save ~15 lives per day. It would probably allow dozens of additional people to avoid serious morbidity. And of course these numbers would accrue over time (e.g., 105 lives saved per week). Over the course of the entire pandemic, a 1% relative reduction in transmission could save hundreds of thousands of lives globally.
For most medical interventions, a 1% relative reduction might not be worth pursuing. However, masks are 100% safe and pretty cheap. So when we weigh the costs versus benefits, it’s not difficult for the lives saved to tip the balance.
Now, let’s say that a 1% or greater relative reduction in transmission would be clinically worthwhile. How many people would we need to study in order to determine if masks work? This depends on the baseline rate of COVID acquisition during the study. In the DAMASK trial, 2% of people contracted COVID. Let’s imagine that we study a population with a much higher rate of COVID – let’s say 15%. In that case, if masks caused a 1% relative reduction in COVID acquisition, then the rate of COVID acquisition would be 15% in the control group versus 14.85% in the mask group. To detect this difference, the study would need to include 1.8 million people:
1.8 million is probably a low estimate, for several reasons. First, mask compliance and technique are often poor – so this would increase the necessary sample size. A 15% acquisition rate of COVID is enormous (5% or 10% might be more realistic, even in high incidence areas). So the study size might need to be many millions of people. Performing this study is pragmatically impossible (and potentially unethical to boot).
So that’s frustrating, right? Shouldn’t randomized controlled trials be able to answer all questions for us? Well, not really, no. To further examine this phenomenon let’s consider another issue: hand-washing after using the bathroom.
Hand-washing prevents the transmission of pathogens via a fecal-oral route (mostly diarrheal diseases such as norovirus, cholera, and hepatitis A). In developed countries, the baseline rate of these diseases is quite low. Let’s imagine that we performed a DAMASK-style study of hand-washing. We randomize 3,000 people to wash their hands after using the bathroom, and 3,000 people to not wash their hands. There would likely be no measurable difference in the rate of diarrheal illness observed (the base rate of these diseases is low, and hand washing mostly protects others rather than ourselves). So the study would conclude that we should stop washing our hands after using the bathroom.
Randomized controlled trials are terrific, but we shouldn’t try to apply them in places where they are doomed to failure. We may never have a robust randomized controlled trial that proves that we should wear a mask during a pandemic, or wash our hands, or wear a seatbelt in a car, or obey the speed limit. But these are simple, common-sense interventions which help keep us and our communities safe.
Masks have become a politically divisive issue, leading to different camps whose beliefs are gradually driven to extremes. Some argue that masks have zero efficacy. Others will argue that masks are enormously effective (e.g., >50% reduction in infections, the initial hypothesis tested in the DAMASK trial). The truth probably lies somewhere in-between (e.g., 10-30% effective alone, more effective when combined with adequate distancing). But even if masks are only moderately effective, they could still save thousands of lives. So please, wear a mask. And wash your hands too.
more on the DAMAST trial
- Additional reasons that a study of masks is impossible (Zeynep Tufekci, New York Times). This post explores only one of the many reasons that an RCT would be impossible – more reasons explored here by Dr. Tufecki.
- Study in Annals of Internal Medicine (and accompanying editorial)
- The DANMASK-19 Trial: Masks Not Effective to Prevent COVID-19? Not So Fast!!! (RebelEM, Salim Rezaie)
- Power, the forgotten error, and inconclusive trials (PulmCrit)
- Pulmcrit wee: The cutoff razor - April 15, 2024
- PulmCrit Blogitorial – Use of ECGs for management of (sub)massive PE - March 24, 2024
- PulmCrit Wee: Propofol induced eyelid opening apraxia – the struggle is real - March 20, 2024
Thanks for the post Josh. The main issue here seems to be the research question : “Does recommendation to wear a mask while outside the home reduce 1 month incidence of SARS-CoV-2 infection by >= 50%?” As you say, a much smaller risk reduction would still make a recommendation to wear masks worthwhile. So confirmation of the null would be unlikely to change public health practice, especially in the presence of observational evidence to support this practice. Perhaps “50%” was motivated more by the practicalities of recruitment than by careful design but we shouldn’t be too harsh on the authors.… Read more »
Agreed the study is grossly underpowered and thus cannot yield definitive conclusion. However, the point is there is no conclusive evidence (from this paper or others) regarding the effectiveness of masks. I’d also argue they do have a potential to cause harm from both an environmental point of view, and making people feel less vulnerable and perhaps less likely to follow other guidelines (ie social distancing…..have you tried to talk to someone from 2m away whilst wearing a mask….?)
So, with the flaws evident in the DANMASK study known, is there a study in progress that fixes those flaws?
If so, I can’t find it.
if not, why not? (No, I don’t have to ask, it’s obvious)