Vitamin D: What the Big Trials Actually Found

Vitamin D is the clearest example in modern nutrition of a nutrient that was mistaken for a cause when it was mostly a marker. Over about fifteen years it went from an obscure bone vitamin to a supposed shield against cancer, heart disease, diabetes, depression and infection — and then, one large trial at a time, most of that came apart.

The story is worth telling accurately, because the marketing has not updated and the guidelines have.

Are most people deficient? No

This is the founding claim of the vitamin D category and it rests on a definitional trick.

Using the sufficiency threshold the National Institutes of Health actually applies — 20 ng/mL — American survey data from 2011 to 2014 found roughly 5 per cent of people at risk of deficiency and about 18 per cent at risk of inadequacy. Not “most.”

The “most Americans are deficient” figure comes from applying a 30 ng/mL target instead, which reclassifies a large slice of a healthy population as insufficient overnight. That target came from the Endocrine Society’s 2011 guideline — and in 2024 the Endocrine Society withdrew it. More on that below.

There is a real disparity underneath the averages: 17.5 per cent of non-Hispanic Black Americans fell below 12 ng/mL compared with 2.1 per cent of non-Hispanic White Americans.

Who genuinely is at higher risk

Breastfed infants not receiving supplemental vitamin D
Older adults, whose skin synthesises less
People with minimal sun exposure — housebound, institutionalised, or fully covered by clothing
People with darker skin pigmentation
People with fat-malabsorption conditions
People with a BMI of 30 or above, and people after bariatric surgery

If none of those describe you and you spend some time outdoors, the base rate of a genuine problem is low.

What the big trials found

Three large randomised trials tested the broad hypothesis properly. They are worth knowing by name because they are the reason the guidance changed.

VITAL (Manson and colleagues, 2019) randomised 25,871 US adults over 50 to 2,000 IU of D3 daily for a median of 5.3 years. Invasive cancer: no effect. Major cardiovascular events: no effect. All-cause mortality: no effect. Crucially, no benefit appeared even in participants who started below 20 ng/mL. A 2022 follow-up on fractures found nothing either — total fractures, non-vertebral fractures and hip fractures were all unchanged, with no modification by age, sex, race, BMI, baseline level, or prior fracture.

D-Health (Neale and colleagues, 2022) gave 21,315 Australians over 60 a large monthly dose for five years. Blood levels separated substantially between groups, so this was not a failure of dosing. All-cause mortality was unchanged. A post-hoc analysis excluding the first two years suggested a possible increase in cancer mortality — exploratory, hypothesis-generating, not a demonstration of harm, but a data point vitamin D marketing never mentions.

DO-HEALTH (Bischoff-Ferrari and colleagues, 2020) tested 2,000 IU daily in 2,157 European adults over 70 across three years. All six primary endpoints were null for vitamin D: blood pressure, leg function, cognition, non-vertebral fractures and infection rate. Subsequent analyses from the same trial found no benefit for falls, no benefit for cardiovascular outcomes, and no improvement in bone density or fracture.

The colds claim, which quietly expired

For years the best surviving consumer claim was that vitamin D modestly reduces respiratory infections. A 2021 meta-analysis of 37 trials found a small significant effect. The 2024 update, expanded to 46 trials and 64,086 participants, found the same point estimate but a confidence interval that now crosses one — no longer statistically significant, with no effect modification by age, baseline status, or dosing regimen.

That is a good teaching example of what it looks like when a real-but-tiny effect fails to survive more data. The estimate barely moved. The certainty around it collapsed.

The 2024 guideline change most people have not heard about

In 2024 the Endocrine Society — the body that set the 30 ng/mL target in the first place — published a new guideline, co-sponsored by four other professional bodies, that did two remarkable things.

It stated that the Society no longer endorses the 30 ng/mL target. And it stated that it no longer endorses specific blood levels to define sufficiency, insufficiency and deficiency at all.

It also suggested against routine vitamin D testing in all the populations it considered, including healthy adults, people with obesity, people with darker skin, and pregnant people. For adults under 50 and adults aged 50 to 74, it suggested against supplementation beyond the standard dietary intake. It suggested empiric supplementation for children and adolescents, adults 75 and over, pregnancy, and high-risk prediabetes.

Every one of those recommendations is graded conditional, several on low or very low certainty, and the guideline is contested — a 2025 rebuttal in Frontiers in Endocrinology pushed back on the paediatric recommendations in particular. But note what the argument is now about: children, pregnancy, and whether “no testing” is practical. Almost nobody is defending the 30 ng/mL target.

So what number should you aim for?

The most defensible reading of the evidence:

Above 20 ng/mL is where the evidence for benefit is solid, and that benefit is about bone.
20 to 30 ng/mL is a genuine grey zone where experts disagree.
Above 30 ng/mL the evidence for additional benefit is thin to absent.
Above 50 ng/mL there are signals of potential harm.

For most people, the recommended intake of 600 to 800 IU plus honest attention to the risk factors above beats testing and chasing a number. If you are in one of the risk groups, that is a conversation with your doctor, and one where testing may well be appropriate.

Is more better? No — and there is a trial that shows it

Burt and colleagues (2019, JAMA) randomised 311 adults aged 55 to 70 to 400, 4,000 or 10,000 IU daily for three years and measured bone density directly.

Radius bone density fell 1.2, 2.4 and 3.5 per cent across the three groups — dose-dependently worse. Tibia followed the same pattern. Bone strength showed no benefit at any dose. Hypercalcaemia occurred in 0, 4 and 9 per cent of participants; hypercalciuria in 17, 22 and 33 per cent.

For the one outcome vitamin D is most confidently supposed to help, three years of megadosing made the measured result slightly worse and the side effects substantially more common.

The tolerable upper intake level is 4,000 IU a day for anyone aged nine and over. Toxicity generally requires blood levels above 150 ng/mL and is essentially always supplement-driven — sunlight cannot cause it, because skin synthesis is self-limiting. Case reports of serious harm from protocols in the tens of thousands of IU are accumulating in the medical literature, and the fat-soluble nature of the vitamin means symptoms can persist for months after stopping.

D3 or D2, and the cofactor claims

D3 versus D2: this one holds up, with a caveat. D3 raises blood levels more effectively, clearly so for intermittent or bolus dosing, and modestly to negligibly at equivalent daily doses. D3 is the sensible default; lichen-derived D3 exists for vegans. Worth remembering that raising the number is not the same as improving an outcome — see every trial above.

Magnesium: magnesium is genuinely a cofactor for the enzymes that metabolise vitamin D. That is textbook biochemistry. But the clinical evidence that adding magnesium improves anything is essentially one nested trial in 250 people, which found magnesium moved vitamin D levels in opposite directions depending on where they started. Magnesium is worth adequate intake on its own merits — we have written about which forms are worth buying — but it is not a required activator without which your vitamin D sits inert.

Vitamin K2: the claim is that vitamin D without K2 drives calcium into your arteries. No trial has demonstrated that vitamin D supplementation causes arterial calcification in humans that K2 then prevents. The premise itself is unestablished. K2 may turn out to have a role in vascular calcification — a 2026 trial found a modest slowing of coronary calcium scoring, funded by K2 manufacturers and measuring no clinical events — but that is a separate research question. The specific syllogism functions mainly to sell a second capsule.

What survives

Vitamin D prevents rickets and osteomalacia. Genuine deficiency is a real clinical problem in identifiable groups, and supplementation reliably corrects it. Those are not small things.

What did not survive is vitamin D as general-purpose disease insurance for people who already have enough. Low blood levels travel with obesity, illness, immobility, indoor living and age — the very conditions that predict poor outcomes. Correcting the number does not correct the conditions.

That distinction — marker versus cause — is the single most useful thing to carry out of this article, and it applies well beyond vitamin D. It is the same reason we are cautious about reading too much into deficiency symptoms and about what a multivitamin can reasonably be expected to do. Our own multivitamin contains vitamin D at a sensible dose, and we would rather you understood why that dose is sensible than believed it was doing more than it is.