physician-scientist, author, editor
The underpowered Danish randomized mask trial, w/ only 46% of those in the intervention group adherence, and only focused on the mask wearer (no insight about transmission) is published @AnnalsofIM acpjournals.org/doi/10.7326/M2…
✓ commentary by @DrTomFrieden and editors (why publish?) pic.twitter.com/bnrviqT2pt
Major issue I think: detection of 85% of cases by antibody tests. Many false positives likely, even with baseline negatives, excluding some w/cross-reactive antibodies. In our editorial, we reanalyzed from the available data without antibody tests: 66% mask effectiveness, P=0.04. twitter.com/EricTopol/stat…
Thread. Importance of reading studies vs. headlines & abstracts: COVID-19 mask edition. Seeing people w/ no science background inaccurately barking abt this study as validation. They're wrong. For those who do this for a living, article fairly reports many study limitations. 1/ twitter.com/EricTopol/stat…
2/n .@EricTopol's tweet above gets at some of the fundamental issues related to study design and implications for drawing conclusions from the findings, but authors were clear about study limitations (see below). Unlike Topol, I don't have issues that it was published. pic.twitter.com/J9r64jeS2R
3/3 Design issues aside, incl. user-centered endpts & low stat. power (ability to detect stat. sig. difference; they aimed for rather high 50% at outset), study shows how setting matters for making sense of findings. @DrTomFrieden commentary hilites this: acpjournals.org/doi/10.7326/M2…
Disappointing result on Danish mask RCT: 0.3% difference in COVID infections (P=0.38).
Several limitations w/ design.
I'm also concerned this was turned down at Big3 journals while many incremental RCTs of big pharma/device products are accepted.
Will stir debate.
"a recommendation to wear a surgical mask when outside the home among others did not reduce...incident #COVID19" acpjournals.org/doi/10.7326/M2…
"should not be used to conclude that a recommendation for everyone to wear masks in the community would not be effective" pic.twitter.com/el5iRGMZmv
Glad this study was published (go science!!) but there are SO MANY CAVEATS that it literally changes NOTHING about our knowledge about recs to #MaskUp .
You'll be hearing a lot about this, so let me break it down.🧵 twitter.com/CaulfieldTim/s…
1) Most impt: Very wide confidence intervals (-43% to +23%) suggest that they were underpowered, likely due to low community prevalence (not enough people exposed to possible infection).
This makes it impossible to draw conclusions.
2) The intervention was sub-par.
They didn't follow best practice. It was SINGLE masking. Despite extensive evidence that BOTH PEOPLE wearing masks = much better:
as they say "the trial did not test the role of masks in source control of SARS-CoV-2 infection"
3) Additionally, there were a lot of confounders:
* only 46% of ppl wore the masks as intended!! 🤦🏽♀️
* the trial started DURING LOCKDOWN when everyone was distanced
* face mask group took fewer social precautions
4) I do like that follow-up included both PCR + antibody tests. A nice touch that increases confidence in results.
5) But weirdly, the % with household contacts who were +, who got infected, was ridiculously low c/w other studies?
The Danish Mask Study is out & mask-skeptics have a field day. 2 things:
i) masks protect others!
ii) 1-month infection rate in trial >10x rate in 🇩🇰 in same period → false positives in both arms, 0.8%? Subtract these ⇒ masks ~30%↓ in infections🤔 acpjournals.org/doi/10.7326/M2…pic.twitter.com/ijGuwqwZe9
Landmark study reports that being instructed to wear a face mask might reduce the wearer's infection by 46% over 30 days, or equally increase it by 23%, with the best guess being that it reduces infection by 21%. acpjournals.org/doi/10.7326/M2…
It's easy for headlines to muddle 'doesn't have large effect' with 'has no effect'. Below study was designed to test for 50%+ reduction in infection risk, so smaller reduction (e.g. 10-20%) wouldn't show up as significant (even though v useful effect to have at population level). twitter.com/theosanderson/…pic.twitter.com/3rieKNuSvR