Why vitamin D probably won’t reduce your risk and the Promise of Immortality

Vitamin D is pretty contentious. If you’ve ever written about the glorious sun vitamin before, you’ll know that people are strangely passionate about it to an extent that you’ll rarely see it in any arena other than sports.

It’s been proposed as a cure for everything from heart disease to cancer, and we’ve run more studies on D than almost any other supplement you’d find in the second most expensive aisle of a supermarket.

People have been claiming that vitamin D is an effective treatment for Covid throughout the pandemic, based on a succession of poor-quality, biased, and sometimes even fraudulent research papers.

Unfortunately, the best-quality evidence has generally not found large benefits for supplementing vitamin D, although there aren’t many good trials so it’s hard to say either way if it works.

Recently, however, the idea that vitamin D could reduce your risk of Covid-19 death to 0% has been going viral online, which is quite odd to see. Rigorous randomized trials have thus far largely failed to find a mortality benefit for the vitamin, so how could it reduce your risk to absolutely nothing?

The answer is, it turns out, that this is almost certainly not true. Take vitamin D if you want, but it’s not going to eliminate the risk of Covid-19 entirely.

Bad Science

The study in question is a review paper looking at observational research of vitamin D and mortality and has recently been published in the journal Nutrients.

The authors collated the results from seven papers looking at hospitals and one ecological paper that looked at population-level vitamin D and used these to examine the correlation between the vitamin and Covid-19 death.

In particular, they plotted all of the points onto a linear regression which showed a weak, statistically significant correlation (p=0.01, r² = 0.19) between vitamin D levels and death.

The authors then used this linear regression to genuinely argue that “The regression for the combined data intersects the D3 axis at 50.7 ng/mL, suggesting a theoretical point of zero mortality”.

Fascinating stuff

There are so many issues here that it’s hard to know where to start. Any review is only as good as the included studies, but for some reason, the authors only looked at some of the lowest-quality research.

For example, the ecological trial, which they reference and use in the regression, itself uses data on vitamin D status from yet another paper from 2018, and that paper references the original estimates of vitamin D levels in various countries that are sometimes decades old.

If we look at the estimate of vitamin D levels in Switzerland used in the ecological trial, 46ng/mL, which is taken from the 2018 paper, the original reference is this 1992 study.

This is obviously not even close to being useful as an estimate of the vitamin D levels of Swiss people in 2020/21 but is merely one example of how problematic some of this data is.

Most of the included observational studies are tiny, for example, this 29-person case series. While interesting, papers like this obviously cannot tell us much about even the correlation between vitamin D levels and mortality, if for no other reason than we know that getting sick can impact your vitamin D levels so if we don’t have people’s values from before they caught Covid-19 it’s hard to know which direction the causality runs.

There’s also no real correction in most of these studies for confounding (as you’d expect given the tiny samples), which means that we don’t know if the mortality is correlated with vitamin D, or if perhaps there’s some third factor causing both low vitamin levels and an increased risk of death from Covid-19.

There are also clear issues with the study’s search methodology. The authors describe this as a systematic review, but even a cursory glance shows that there are papers that are not there that meet the inclusion criteria.

It’s hard to know how this happened because the entire review methodology is one short paragraph and the paper does not meet even the most basic guidelines for systematic reviews.

On top of all of this, there’s the completely incorrect linear regression analysis to consider. Firstly, most of these studies didn’t provide exact values for vitamin D levels, they only gave a mean/median, so even the data points are not really correct.

But even more fascinating is the claim that, because the regression line hits zero around 50ng/mL, there is a “theoretical point of zero mortality” if your vitamin D levels are high enough.

This is just a very basic statistical error. You cannot extrapolate from linear regressions in this way, as was perhaps most famously demonstrated in discussions about athletes’ running times in that excellent post by the Calling Bullshit authors Carl Bergstrom and Jevin West.

Even a basic think about this regression line shows how absurd it is — what happens at levels of 60ng/mL? I have vitamin D levels about that high, does that mean that I am at a negative risk of Covid-19 death?

Moreover, looking at the graph closely shows how bad this analysis is. The mean value for vitamin D levels in most places varies, but even in European countries, it’s often above 50ng/mL.

However, this graph has every datapoint below 40ng/mL. The obvious reason here is that the authors only looked at studies of people who had been admitted to hospital, and were therefore far sicker than the general population — this makes any extrapolation to people who have not been hospitalized largely meaningless.

Bottom Line

I could go on, but at a certain point, you have to wonder what the point is. The study itself is filled with holes, and should never have been used as evidence for anything.

At best, it’s a collation of some rather inconclusive and/or terrible papers published that look at very vague correlations between people’s vitamin D levels when they enter the hospital and whether they die of Covid-19.

What we’re left with is, somehow, the same situation that we were in when I last wrote about vitamin D and Covid-19 early in 2021: there is a plausible benefit, and the supplement is low-risk, but the evidence to date is woeful.

According to a recent systematic review, there is still only a single large, reasonably well-conducted randomized trial of supplementation with vitamin D, and it did not find a benefit.

Even this study, however, was far too small to assess the impact of the vitamin on the Covid-19 death well, which means we still don’t have a good answer to the question.

As I wrote months ago and probably will have to write again the next time the vitamin D hype comes to town, you can take a D supplement if you want.

It’s not hugely harmful, although everything has risks, it also probably isn’t going to eliminate your risk from Covid-19. Vitamin D is great, but supplementing it is unlikely to be a panacea for anything, and the high-quality evidence to date — what little there is — is not promising in its use for Covid-19.