Thread by @GidMK

A new study has hit the headlines claiming that eyeglasses can reduce your risk of catching COVID-19 by "2-3 times:

Unfortunately the science is...not good

Some peer-review on twitter 1/n

2/n The study is preprinted on medrxiv here. It is a single-author study on a survey done in India during COVID-19 medrxiv.org/content/10.110…

This will be a short thread, because...wow. Issues

3/n Some background here - traditionally, in epidemiology, to work out whether something is protective against disease, you need to know two basic things:

1. Likelihood of disease if exposed
2. Likelihood of disease if not exposed

4/n Essentially, to know whether glasses make you less likely to catch COVID-19, we'd need to know whether people with glasses caught the disease less than people without glasses

Simple, right?

5/n Ok, back to the study

What did the author do? Well, the asked 304 people with COVID-19 whether they wore glasses most/all of the time. 58 (19%) said yes

6/n Then, the author took the proportion of Indian adults estimated to wear glasses from a paper in 2019, to compare this sample to

7/n So far so problematic. You can't just compare to an out-of-study population like that, it makes no sense at all. If nothing else, the comparison group is for the whole of India, while this survey was done on a tiny subsample in one hospital 😬

8/n But then, we get to these calculations, which are described as "the risk of [catching] COVID-19" in glasses wearers vs non-wearers

But...that's just incorrect

9/n What the author has done here is compare the rate of glasses wearing in COVID-19 patients to the rate of glasses wearing in the general population

10/n What this calculation actually gives you is the likelihood of wearing glasses in COVID-19 vs non-COVID-19 people

In other words, what we've got here is the relative risk of glasses-wearing behaviour in COVID-19 patients compared to everyone else

11/n So not only is the study a tiny cross-sectional survey with no appropriate comparator, it's also not calculating what the headlines (and conclusions) say it is at all

12/n What the headlines should say is that COVID-19 patients are 2-3 times less likely to wear glasses than the general population, based on a small, biased survey

But I doubt that will get as much attention, because it's a bit meaningless

13/n Here is the conclusion of the study. This is not correct based on the methodology as described:

14/n Oh and in case anyone was wondering, according to Altmetric the study has been in 51 news articles so far