This study in PLOS One is being reported (and used by a lot of people on twitter) as proving definitively that Vitamin D supplementation can treat COVID-19

So, thought I'd do a peer-review on twitter 1/n

2/n The study is here if you want to read it - a very simple retrospective epidemiological paper: https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0239799

3/n Basically, the authors took a sample of people who'd been hospitalized with COVID-19, and split them up into 2 groups - normal vs low vitamin D (or, more technically, "sufficient" vs "insufficient")

4/n They found that those with low vit D had a higher risk of some adverse outcomes, in particular that people >40 years with low vit D died more often than people >40 years with normal vit D

5/n Sounds good right?

Well, some issues

Firstly, sample size: it was VERY SMALL

6/n The study used a dataset with 611 patients, but only 235 of them had information on their vitamin D status, so the final sample for the study was only 235

7/n For that main, headline finding - that vit D reduces the risk of death - the authors were looking at a subset of this number (about 150 people)

That's not a lot!

8/n It's also very odd when you actually look at the results, because the findings seem...wildly unimpressive

This graph, for example, comparing vit D levels with risk of death. It's about as null a finding as you can get at first glance

9/n If we look just at the patients who the authors found a significant relationship for - those over 40 - you can kind of see a relationship but it's VERY slim

10/n And when you look at all of the other outcomes the authors analyzed, a similar pattern emerges

Low vit D increased the risk of hypoxia, but not shortness of breath. It DECREASED (not significantly) the risk of chest pain

11/n You could write this study up as demonstrating that there was no correlation between vit D and most COVID-19 outcomes, and therefore we still don't know if it does anything for the disease

12/n Instead, the authors use the pretty tenuous observational link between low vit D in people over 40 and death to posit that supplementation should be used for all patients which isn't really supported by their results

13/n While I could go into the results a bit more - the statistical analysis is a little bit meaningless, the confounding factors not really even considered - but honestly there's not really much point I think

14/n The results show a vague correlation between vit D and some outcomes, with a smattering of statistical significance if you run the analyses in certain ways

Not much that you can take away from that, I think

15/n If nothing else, the small sample size makes it hard to conclude anything from these results other than "we need BETTER research"