Vitamin D against colds: the effect has just vanished

Does vitamin D protect against colds? In 2017 I would have said yes, with a large study to back me up. Today the answer comes out differently, and this article is about how that happens and why it is a good sign.

How it began

In 2017 the British Medical Journal published a meta-analysis that caused a stir. Rightly so. It analysed not only the published summaries of 25 randomised trials, but the data of the individual participants. 10,933 people, from infants to 95-year-olds. That is the most demanding form of meta-analysis and the most informative.

The result: those given vitamin D had a 12 percent lower risk of suffering at least one acute respiratory infection.

Martineau AR et al. BMJ. 2017;356:i6583. PMID 28202713.

Two details mattered more than the headline. First: the protection showed up with daily or weekly dosing. With large single doses, so-called bolus doses, it disappeared. Second: in people with severe deficiency the effect was large, in the well-supplied it was small.

Both details were lost in public perception. What remained was: vitamin D protects against colds.

How it continued

In 2021 came the update. More studies, more people: 46 randomised trials, 75,541 participants. The protection was still there — and had grown smaller. From 12 percent risk reduction, it became 8.

The authors wrote themselves that the reduction was small. And they added a sentence rarely stated this plainly: the relevance of these findings to COVID-19 is not known.

Jolliffe DA et al. Lancet Diabetes Endocrinol. 2021;9(5):276–292. PMID 33798465.

That question has since been answered. In a randomised trial with 6,200 adults, a test-and-treat programme led neither to fewer respiratory infections nor to fewer COVID-19 illnesses.

Jolliffe DA et al. BMJ. 2022;378:e071230. PMID 36215226.

How it looks today

In 2025 the next update from the same group appeared. 40 studies, 61,589 participants, including one very large new trial.

The result: a risk reduction of 6 percent. And a confidence interval that includes zero. p = 0.057. The effect is therefore no longer statistically significant.

Jolliffe DA et al. Lancet Diabetes Endocrinol. 2025;13(4):307–320. PMID 39993397.

The authors additionally report that the funnel plot is asymmetrical. That is a sign that small studies with a positive result were more likely to be published than small studies without one. A familiar pattern. It means: the true effect is likely somewhat smaller still than the calculated one.

Other analyses fit the picture. In people over 65: no effect. In healthy children: no effect.

Jia H et al. PLoS One. 2024;19(5):e0303495. PMID 38787821. Fang Q et al. Front Nutr. 2023;10:1188958. PMID 37408991.

What just happened here

An effect has, over eight years and three analyses, shrunk from clear to small to no longer statistically detectable.

That is not a scandal. That is the normal case.

Early studies are smaller. Small studies with a spectacular result are more likely to be published. When large, carefully planned studies then join, the estimate moves towards zero. Anyone following science only by its headlines experiences this as a betrayal. Anyone measuring it by its method experiences it as self-correction.

The question is not: did they get it wrong in 2017? They correctly summarised what was available in 2017.

The question is: what holds now?

My placement

I write about this finding as plausible, not proven. What those three labels mean I explained in the opening article.

There is a traceable mechanism (vitamin D acts on the innate immune defence). There are many human studies. And the pooled effect on what matters is, at the current state, no longer statistically secured.

Where a residue remains, it is where it was largest from the start: in people with real, pronounced deficiency. Anyone who sees barely any sun for months in a central European winter, eats little fish and spends a lot of time indoors can have a low level. For them, vitamin D is not a question of cold protection but of basic supply, for bones and muscles. And there the evidence is different.

What the data do not support: that a well-supplied, healthy adult lowers their cold frequency by taking a capsule. And certainly not that a high single dose helps. The bolus doses failed to work back in 2017.

What I take from this

If someone sells you a supplement with the line "studies show it protects against infections," ask them for the year of the study. And for the study after it.

Knowledge has an expiry date. That is not a weakness of science. It is its only advantage over opinion.

---

This article is not medical advice. Whether you have a deficiency is shown by a blood test; whether you should do anything about it you decide with a doctor. Very high doses over a long time can cause harm.