Thread by @MonicaGandhi9

Our group including the very smart @MattSpinelliMD was asked to comment on the DANMASK study found here: acpjournals.org/doi/10.7326/M2…. Others have already smartly commented that implementation science is the way to study mask wearing but here goes:

First the study team (many cardiologists in Denmark) should be commended for performing a randomized controlled trial of a non-pharmaceutical intervention (mask-wearing) in the early days of a pandemic; there are study design flaws that will impair our ability to draw conclusions

The study was a non-blinded study of providing 50 surgical face masks tor community participants spending >3; hours outside of the home in April-May (intervention group) versus the standard of care of Denmark not recommending widespread use of face masks in the community at time

Flaw#1: Underpowering: This study conducted during a time of declining incidence in Denmark (…id19-country-overviews.ecdc.europa.eu) so study underpowered (classic Type II error) to look at the outcome. Indeed -few outcomes in this trial, which demonstrates this limit in statistical power,

regardless of the sample size calculations. Flaw #2: Randomization at individual rather than community level: Other studies have shown that the most common source of infection with SARS-CoV-2 is from household contacts. By designing the study as an individual-level rather than a

Flaw 3: Flawed outcome metrics: Use of a home point-of-care antibody test with limited sensitivity (sensitivity is likely even less than 83% given that samples used in validation studies typically come from symptomatic participants with positive PCR tests) decreased study power.

and limited number of outcomes further. Hypothesis that mask-wearing could increase the proportion of asymptomatic infections if infection occurs mentioned, but study is not designed to show this difference - we have talked about T cells (not antibodies) to detect past infection!

Participants had to prick own finger for antibody home testing, and then mail their results back in (subject to underreporting, forgetting) and (fewer times) collect their own OP/NP swabs

Flaw #4: Self-reporting of results introduces bias: Use of self-reported results on this test introduces further measurement error and potential bias, in an unknown direction, which is impossible to sort out without confirmatory testing

Flaw #5: Given low community-level adherence to masking during this time period, masks were at a disadvantage, given that they are more effective at preventing infection from an infectious source.

Flaw #6: Community norms and global recommendations could influence mask wearing in each group, leading to misreporting: Participants in the intervention early on said they experienced social harms from wearing the masks, possibly lead to low adherence and over-reporting

Later in the study, the control group was likely to be influenced by increasing promotion of masking as a non-pharmaceutical intervention worldwide, and likely under-reported their mask wearing. 20% of participants did not complete the study.

And finally, to be fair, the small reduction in incidence seen in this study was acknowledged, as were most of the limitations so the authors should be commended on this important RCT! But don't think it changes the recommendation/importance of masking.

Here is an editorial published simultaneously with the paper covering many of the same points and stressing this RCT does not mean that community-based mask wearing is not an important pillar of control:
acpjournals.org/doi/10.7326/M2…

And one more important editorial published simultaneously by Dr. Thomas Frieden that ends with "Masks have been shown to protect others and, despite the reported results of this study, probably protect the wearer. Maximum benefit of masking is likely to result from combination

of source control and wearer protection. If everyone wears a mask when near others, everyone is safer.
acpjournals.org/doi/10.7326/M2…

I am really bad at twitter right? that was so long!

Consulted with the amazing @dvgbiostat and a few more flaws to the study.
Flaw#7: Assessment of risk: There is no assessment of the nature of SARS-CoV exposure in the two groups. They are spending >3 hours out of the house? Doing what? Occupational? country partial lockdown

Flaw #8: Definitions of adherence: Adherence to mask wearing in the intervention group is quite low (47%) and we know from many other fields and studies that self-reported adherence usually overestimates actual adherence. Also, no assessment of adherence in relation to exposure.

Flaw #9: Since they mentioned our viral inoculum hypothesis: In terms of exploring whether mask wearers could have milder infections, there are no details on the clinical course of the detected infections- so not investigated

In sum, testing this ? via an RCT (rather an observational or implementation design) is odd because of contamination issues; it is easy for those randomized to not wear masks to “drop in” and adherence in those randomized to wearing masks is both hard - and critical - to assess