You are unlikely to get the right answer if you ask the wrong question. This mask study is not only underpowered, it has design issues.

Here are my thoughts on why this data is not useful. https://twitter.com/erictopol/status/1329082347480784897

Masks are a public health strategy. The authors are trying to measure its effect from an individual perspective. This is intrinsically wrong.
1. Masks very likely protect more the others than the user. Somebody else’s mask protects me.

2. The study fails to consider that masks have compounding effects in the community. While a single person wearing it is likely to have no impact, if everyone wear the effect multiplies, particularly as transmission has an (semi) exponential progression.

3. Using a randomization at the individual level it is impossible to account for the spill over effect to the other group (controls wearing masks) as well as the potential effect of others around to start (or stop) using masks influenced by the study subjects.

1, 2 and 3 could be fully addressed with a different design that approaches the question from a collective, public health perspective. The study should have randomized cities or communities in a cluster design. Each cluster goes to one group.

4. Loss to follow up is too high. Since those lost to follow up are unlikely to be missed at random or completely at random (MAR or MCAR), this likely leads to a biased result regardless of power.

5. The study is underpowered, though the number of events reached what was planned. The study was powered for a 50% event reduction. I think even a 20% reduction would be clinically meaningful. Thus, and OR of 0.82 with this power is just inconclusive.

6. Even if the effect was lower (e.g. 15%) but compounding if the other person also wears a mask this would lead to a much higher reduction in transmission, so real power of masks could easily be twice the 0.82 OR.

My take is the this could have been a great study with an appropriate design. As is it adds nothing but noise to the discussion.

Any public health community based intervention should be studied on a community level with cluster design.