[epmaybe]

I would like to point out the limitations of the linked paper (you can follow along in the discussion section of the paper). Hopefully this can generate enough discussion on how to better design studies like this in the future:

1. Statistically significant difference in patients with documented hypertension in the no-vitamin-d group (p value 0.002, 15 vs 11 patients)

2. 2:1 randomization, reducing statistical power. They do not document the sample size needed from a power analysis looking for power > 0.80. They basically assumed that twice as many people without vitamin D would end up in ICU, when in reality 13x as many people ended up ICU. In general, I wish we could have a larger sample size for studies like this, but I know it isn't practical.

3. Biggest for me is no baseline serum vitamin D level recorded, just an assumption that patients selected were overall vitamin D deficient based on population. Seems like a simple enough test that could have been run as a send out test (no need to have the level during hospitalization, just for data gathering purposes).

4. Not placebo controlled, would have helped as placebo effect is real. For all we know, giving the research medication could have meant that the patient received more care and supervision, potentially aiding in their hospital course.

[YZF]

Sure. It's not a perfect experiment. But the result of 2% vs. 50% ICU admission is astounding. And sure, the sample size is small so it could have arisen by chance (not very likely though). Personally until I saw this result I listened very patiently and mostly agreed with the people who argued correlation vs. causation, lack of data, or random other non-data backed arguments. I think now the balance of the scale has shifted towards until proven otherwise I'm gonna be supplementing vitamin D and if I do get sick with Covid19 I'm going to be asking for this as part of my treatment. And if placebo is so strong here, then I'll get my placebo effect anyways ;) Until there's conflicting evidence or more studies I think that's a reasonable position/risk balance.

[epmaybe]

I think that's a pretty fair assessment, there's generally minimal risk at normal dosages you find in the store (1000IU or whatever). If you're really concerned, a quick visit to the doctor to get baseline labs is a good way to make sure you're not gonna verge on toxicity.

[andruby]

Correct. The sample population was also limited. IIRC the main conclusion of the study is that larger scale studies should be conducted.

Every study has limitations. Do you feel that the limitations you stated mostly invalidate the study?

Nevertheless, like the professor of medicine at Harvard Medical School JoAnn Manson stated in May 2020 [0]: "The evidence is becoming quite compelling [that good Vitamin D status will protect against severe complications]".

[0] https://www.medscape.com/viewarticle/930152

[ekianjo]

Effect size matters as well, not just sample size. The effect size in this paper is huge.