An ongoing dialogue on HIV/AIDS, infectious diseases,
January 4th, 2023
Medical Masks vs. N95 Respirators for Preventing COVID-19 Among Healthcare Workers
As promised, the end of 2022 saw a trio of controversial COVID-19–related publications.
First up is something that always causes a stir — a study on masks! Reviewing a study on masks in the COVID-19 era is like poking a hornet’s nest with a stick, and this one is no exception.
But let’s poke away! Aside from receiving a lot of attention when it first appeared online, it gives us a chance to review some interesting clinical research principles.
Here’s the study question — does using a medical mask while caring for a person with confirmed or suspected COVID-19 provide healthcare workers “noninferior” protection to an N95 mask?
That “noninferior” term is often confusing, so best to translate it into plain English, which is what Harvard biostatistician Michael Hughes kindly did for me many years ago. Noninferior simply means “not too much worse than.” Noninferiority studies are great when the thing you’re testing has other advantages to the standard of care — it might be simpler, or cheaper, or both. As a result, a noninferiority design is quite appropriate for comparing surgical masks (cheaper, easier) to N95s.
Now one thing the statisticians ask us clinicians is to define the noninferiority margin — in other words, if not-too-much-worse is a key determinant, how much worse would we tolerate? In this study, if the upper bound of the 95% confidence interval of the hazard ratio of surgical to N95 masks was less than 2, then they’d be declared noninferior.
That sounds like a lot — would we really tolerate something that gives us only half the protection of an N95 mask? One thing to remember about noninferiority margins is that the smaller the margin, the bigger the required sample size. I suspect that anything smaller would have made the study impractically large.
The study was conducted in 29 healthcare facilities in Canada, Israel, Pakistan, and Egypt from May 2020 to March 2022, with 1009 participants. It’s important to scrutinize the dates of all COVID-19 studies because the vaccines, prior COVID-19 (with residual immunity), and variants have greatly changed the nature of SARS-CoV-2’s transmissibility, severity, and our response to it. As I’ve noted previously, the post-Omicron era includes vastly more people who had COVID, and vastly more people who stopped preventive measures while out and about in society.
So finally, let’s get to the primary results. RT-PCR–confirmed COVID-19 occurred in 52 of 497 (10.46%) participants in the medical mask group versus 47 of 507 (9.27%) in the N95 respirator group. You don’t have to be a statistician to conclude that these numbers are pretty darn close.
This yields a hazard ratio of 1.14, with a 95% confidence interval of 0.77 to 1.69. That means the surgical masks could be as much as 69% worse (but less than twofold worse, the non-inferiority margin), or even 23% better, at protecting healthcare workers.
In short (drumroll), the strategy of wearing surgical masks was noninferior to N95s among people caring for people with confirmed or suspected COVID-19.
Perspectives on this study:
The critical view: The study was sloppy and, frankly, unethical. It’s already been proven in several models that filtration of respiratory viruses is more effective with a well-fitted N95 than regular masks — and COVID-19 is clearly transmitted by an airborne respiratory virus.
A twofold noninferiority margin is way too high. Even if they’re only 69% worse, why should healthcare workers take the chance?
The study didn’t even test the efficacy of the masks, since undoubtedly many of the participants got COVID while not even caring for COVID-19 patients and not wearing the N95s — either elsewhere in the hospital or (even more likely) in the community. This is especially true in Egypt, which accounted for many of the cases in the study during the post-Omicron period. How can we say the masks didn’t work when infections were occurring outside the patient room?
Let’s look at another country, how about Canada? There, surgical masks were more than twofold worse than N95s. Shouldn’t that be our model?
Finally, the study took a long time to enroll and had several modifications before completion. Doesn’t that alone make the results unreliable?
The supportive view: This study proves that a policy of recommending uncomfortable, expensive N95 over surgical masks is pointless. The highest form of evidence — the randomized clinical trial — shows they’re noninferior to cheap surgical ones.
Lots of clinicians hate N95s. When the study was first posted online, one of the smartest doctors I know asked me flat out — “So can I finally ditch these things? By the end of the day, I feel like my face has been in a vice.”
Yes, there are differences between countries, but importantly this was a post-hoc analysis. We should ignore these analyses because if you measure something frequently enough with smaller and smaller sample sizes, you’re bound to find something that’s statistically significant that supports your hypothesis.
And if you’re going to focus on a country, isn’t the current landscape of COVID much more like Egypt (during Omicron and high community transmission) than Canada (early in the study, very low event rates)?
As for those filtration studies? Remember, clinical trials enroll human beings — not mannequins or robots wearing masks.
My take: Both sides have excellent points. I learned a lot from reading insightful commentaries taking both sides of this debate — both praising the study (here and here) and criticizing it. Plus, there’s an excellent accompanying editorial.
As for what I think?
I believe that if the study were large enough, if the N95 masks were properly worn and correctly fit tested, if in-hospital COVID-19 transmission in break rooms during snacks and lunch could be excluded, and (an even bigger task) if transmission at restaurants and weddings and concerts and gyms (meaning in the community) could be excluded, then this study would have shown that surgical masks are not as good as N95 masks in protecting healthcare workers.
In other words, they’d be too much worse to make up for the lower cost and greater comfort. More than twofold worse, as defined by the study.
But that’s a lot of ifs, and is not the real world. The real world is messy, and such stipulations would be impossible. In the real world, the surgical masks in this randomized clinical trial were noninferior for the primary endpoint of PCR-diagnosed COVID-19 in the healthcare workers.
Using N95 masks in the post-Omicron era is like giving someone an excellent umbrella during a rainstorm, but only during the brief downpours when dashing from the car to the front door — the times of highest direct exposure. The rest of the day, with steady and frequent rain, they use either a broken umbrella or none at all. Plenty of chances to get wet.
No wonder the study showed surgical masks to be noninferior. COVID-19 is now everywhere, and patient-to-healthcare provider transmission of the virus is a small fraction of the exposures happening globally.
Do I still wear an N95 when caring for patients with confirmed or suspected COVID-19? Yes. After all, I still use an umbrella when dashing from the car to the door during a downpour.
Do I also believe that getting COVID-19 is much more likely at a restaurant or party or medical meeting than in the patient’s room?
Also yes. That’s just the world we live in now.