I’ve written about vitamin D as a potential treatment for covid-19 a couple of times before. In September of 2020 I wrote about a Spanish randomized trial that showed a massive reduction in ICU admissions in hospitalized patients treated with 25-hydroxyvitamin D (a part-activated form of vitamin D). However, that study had some major weaknesses – it was completely unblinded and it was small. In other words, although the results were promising, they hardly constituted conclusive proof of a benefit.
Then, in January, I wrote about a much larger double-blind Brazilian trial that failed to show any benefit when hospitalized patients were treated with vitamin D. For many in the mainstream medical community, this study constituted conclusive proof that vitamin D is ineffective as a treatment for covid-19. However, participants in the trial weren’t given vitamin D until late in the disease course, and unlike in the previous Spanish study, they were given regular vitamin D, not the part-activated form used in the Spanish study. Since it takes several days for regular vitamin D to become activated and usable by the body, the study was more or less designed to fail from the start – whether intentionally or unintentionally. There was no realistic chance that it was ever going to show a benefit, even if one exists.
So, when I last wrote about vitamin D, in January, it was still unclear whether it had any role in the treatment of covid-19 or not. Well, has anything changed since then?
Let’s start by taking a look at the observational data. Although the randomized trials of vitamin D have been few and far between, there has been a massive amount of observational data produced. In recent months, two meta-analyses of cohort studies have been published that look at the relationship between vitamin D levels and death, one in the journal Nutrients, and another in the Nutrition Journal. For those who are unaware, a meta-analysis is a pooled study, where you take lots of different studies and pool their results together in to one, in order to get more statistically significant results. And a cohort study is a type of observational study in which you take two or more groups that vary in some specific way, such as in their vitamin D levels, and then follow them over time to see if they have different outcomes.
Interestingly, the two meta-analyses reach the exact opposite conclusions, with one claiming that if we can just get everyone up to 50 ng/ml of vitamin D in the blood stream, then we can literally prevent all covid deaths, while the other says that there is no correlation whatsoever between vitamin D levels and covid mortality.
How is this possible?
Well, the first thing to note is that the two meta-analyses vary in terms of which particular studies they include. The Nutrients study performed its data search in March, while the Nutrition Journal study performed its data search in June. This means that the Nutrition Journal meta-analysis had access to a few extra studies, which weren’t available when the first data search was conducted in March.
But that isn’t the whole explanation. The researchers have also used somewhat different search strategies, which means that each includes some studies that the other lacks. In fact, there is amazingly little overlap between the two meta-analyses in terms of which studies are included. Only two of the seven studies in the Nutrients meta-analysis are included in the Nutrition Journal meta-analysis, and only two of the eleven studies in the Nutrition Journal meta-analysis are included in the Nutrients meta-analysis. No wonder they’re able to reach such divergent results!
This really showcases how easy it is to manipulate meta-analyses so that they show what you want them to show, just by choosing the date on which you extract data and by choosing which specific search terms to use. It’s easy to keep modifying search terms slightly until you get a list that includes the studies you want in, and excludes the studies you want out. Which is why we should always be skeptical of meta-analyses, just as we are with other types of studies.
This explains why we saw a similar phenomenon earlier this year, when half the meta-analyses of ivermectin seemed to show massive benefit and the other half seemed to show no benefit whatsoever.
Meta-analyses are often considered to be the pinnacle of evidence based medicine, but considering how easy they are to game, I think that is wrong. I’d rather have a single large, well done study than a meta-analysis that consists of lots of little studies, even if that meta-analysis includes more participants overall. If all scientists were honest, then meta-analyses would be an excellent tool for determining the truth. But since we know that many scientists aren’t honest, their use is far more limited. At the end of the day we all have to do our own due diligence. The only meta-analyses that I trust fully are the ones I do myself, such as one I did earlier this year on ivermectin.
It is interesting to note though, that the Nutrition Journal study, which came to the conclusion that vitamin D had no effect, actually had results that did suggest benefit. All the included studies showed fewer deaths in the group with a higher vitamin D level, although the results were not able to reach statistical significance overall.
So we have one meta-analysis which shows a large benefit, and one which shows a trend towards benefit. Which is encouraging. Of course, these are observational studies, and so can’t really say anything about cause and effect. People with low vitamin D levels probably spend less time outdoors, which means they’re probably less physically active. And they probably consume a different diet from people with high vitamin D levels. Correlation is not causation, and even if a correlation is seen between vitamin D and death from covid in observational studies, that doesn’t mean it’s the vitamin D that’s preventing the deaths.
So, what we need is good data from randomized trials. As mentioned, the data from randomized trials that existed last time I wrote about vitamin D was limited and mixed, with the Spanish study that gave 25-hydroxyvitamin D showing massive benefit, while the Brazilian study that gave regular inactivated vitamin D showed no benefit whatsoever.
Since then, three new randomized trials have been published, one from India, one from Mexico, and one from Saudi Arabia. Unfortunately, all three were small, with the largest of the three only including 87 patients. Additionally, all three gave regular inactivated vitamin D, not the part-activated form that was found to have an effect in the earlier Spanish study. In other words, the new studies don’t add anything on top of the store of knowledge that we already had in January.
So, we’re actually more or less in the same situation that we were in regarding vitamin D back in January. The observational data suggests that there is a benefit to supplementing with regular vitamin D for prophylaxis, which is in line with a systematic review that was published in the British Medical Journal in the pre-covid era, which found that people with low vitamin D levels who supplement daily with vitamin D reduce the frequency of respiratory infections by half. And the limited randomized trial data that exists suggests that the part-activated 25-hydroxyvitamin D formulation can reduce the risk of bad outcomes if given on admission to hospital. But the evidence is still too weak to draw any firm conclusions.
It’s quite shocking that more data isn’t available to answer this question conclusively at this late stage in the pandemic. I do personally think though, based on the evidence that is available, that it makes sense to take a daily vitamin D supplement. 4,000 IU (100 mcg) is a reasonable dose. It’s safe, it’s cheap, it might well help, and it can’t hurt.