Comment by timr | Hacker News Reader

Comment by timr | Hacker News Reader

Parent Full thread

It's notable that the only randomized controlled trial ever conducted for masks and Covid [1] was not listed under the personal protection section of this page, and was instead was inaccurately summarized:

> A community-based randomized control trial in Denmark during 2020 assessed whether the use of surgical masks reduced the SARS-CoV-2 infection rate among wearers (personal protection) by more than 50%. Findings were inconclusive,54 most likely because the actual reduction in infections was lower.

This is inaccurate, and should make anyone doubt the judgment of whomever wrote this summary. The paper in question was quite clear that there was no statistical significance between groups:

> The between-group difference was −0.3 percentage point (95% CI, −1.2 to 0.4 percentage point; P = 0.38) (odds ratio, 0.82 [CI, 0.54 to 1.23]; P = 0.33). Multiple imputation accounting for loss to follow-up yielded similar results. Although the difference observed was not statistically significant, the 95% CIs are compatible with a 46% reduction to a 23% increase in infection.

To characterize this as "inconclusive" is misleading at best, dishonest at worst. In general, I find this page to be a highly selective reading of the available literature, which was (and remains) predominantly inconclusive regarding the effectiveness of masks. Even the WHO meta-analysis of mask literature found only a weak positive effect after mixing together the results of studies ranging from cloth masks to respirators, in medical (most of the data) and non-medical settings [2]. This CDC page has instead chosen to lean on a few poorly controlled observational and/or correlative studies, and ignore the higher quality evidence that came before.

[1] https://www.acpjournals.org/doi/10.7326/M20-6817

[2] https://www.thelancet.com/journals/lancet/article/PIIS0140-6...

Ok so you say that characterizing the results as incoonclusive results is dishonest, but in the article the limitations sections literally lists "Inconclusive results" as their first item.

On top of that they state: "Although the difference observed was not statistically significant, the 95% CIs are compatible with a 46% reduction to a 23% increase in infection."

Which sounds like their saying they can say with 95% confidence that there is between a 46% reduction and a 21% increases, which sounds pretty damn inconclusive to me.

The between-group results are not "inconclusive" -- they found no statistically significant difference between the masked and unmasked cohorts. There's no reason to reject the null hypothesis ("masks are not protective") based on this data.

The authors say this:

> The most important limitation is that the findings are inconclusive, with CIs compatible with a 46% decrease to a 23% increase in infection.

This is not saying that that study is inconclusive. It's saying that the protective effect is inconclusive -- masks could be anything from slightly protective, to slightly harmful. The fact that the writer characterizes the entire paper as "inconclusive" is incriminating: it's an editorial bias, a complete misunderstanding of statistics, or a combination of both.

> Which sounds like their saying they can say with 95% confidence that there is between a 46% reduction and a 21% increases, which sounds pretty damn inconclusive to me.

No. They're stating the confidence interval on the point estimate. All point estimates have confidence intervals, and the existence of a confidence interval does not mean that the result is uncertain. To restate basic statistics: the top-line conclusion of the study is that the 95% confidence intervals of the two groups overlap to such a degree that you can't reject the conclusion that they're the same.

I didn't mean to say the fact that the interval exists makes it uncertain, I meant to say the fact that it is a 67% spread makes it uncertain. They are saying that they have narrowed the effect down to within that spread with a 95% chance. That combined with sources of uncertainty not covered by that intervale (Listed in their limitations: missing data, variable adherence, patient-reported findings on home tests, no blinding, and of course inconclusive results) definitely make it sound pretty inconclusive.

There is also the limitation that there was "no assessment of whether masks could decrease disease transmission from mask wearers to others."

The article explicitly states "The findings, however, should not be used to conclude that a recommendation for everyone to wear masks in the community would not be effective in reducing SARS-CoV-2 infections, because the trial did not test the role of masks in source control of SARS-CoV-2 infection."

> That combined with sources of uncertainty not covered by that intervale (Listed in their limitations: missing data, variable adherence, patient-reported findings on home tests, no blinding, and of course inconclusive results) definitely make it sound pretty inconclusive.

The study is not inconclusive. The study failed to reject the null hypothesis. That much is definitive.

Whether or not there might be some smaller difference that the study wasn't powered to detect...we don't know. It's still a definitive rebuttal of any claim that "masks reduce personal risk of infection by 50%", and the fact that it's not in the "personal protection" section of the CDC webpage is simply editorial bias. At the very least, this paper is the best study ever performed on masks and SARS-CoV2, and it severely limits any real-world claim of protectiveness.

> There is also the limitation that there was "no assessment of whether masks could decrease disease transmission from mask wearers to others."

The study wasn't designed to do that. It should still be in the "personal protection" section of the CDC webpage, and is not.

Reply on news.ycombinator.com