How much vitamin D
is officially recommended?

Almost every question about vitamin D ends up at the same one : how much should I take? The answers circulating range from a few hundred to several thousand units a day, and the gap between them is not a rounding difference. This article sets out what the official bodies actually recommend, where they differ, and why we deliberately do not add a figure of our own.

A note up front : this article reports official recommendations and is educational, not medical advice. StoaVita gives no dosing recommendations of its own. For infants, toddlers, during pregnancy and breastfeeding, and for the conditions named further down, this question belongs in medical hands.

The official figures side by side

Two things need separating before you read the table. The German estimated value applies to the case where the body produces none of its own, so it is a nutritional reference value, not a treatment dose. The tolerable upper intake level in the last column is the opposite of a target : it marks the point up to which long-term intake is considered unproblematic.

Correcting a shortfall is not the same as optimising

Most of the confusion in this field comes from mixing up two questions that only look similar. The first : someone is genuinely short, what brings them back into range? The second : someone is adequately supplied, does more bring an additional benefit? For the first question the evidence is solid and the answer is yes. For the second, the randomised trials have repeatedly come back empty, and that is the part of the debate this article is really about.

Daily rather than in one big dose

If you take only one practical point from this page, take this one : the pattern of intake appears to matter more than the exact target level. Trials that reported clinical harm used large single doses given monthly or yearly. Daily regimes have not produced that signal. This point is better supported than any of the disputed threshold values, and it gets far less attention because it is less interesting to argue about.

Magnesium and K2 as co-players

On magnesium you will often read that the converting enzymes depend on it. In that strong form the sentence does not hold : the reviews it traces back to write that these enzymes appear to require magnesium. Better documented is a connection one level up, through the parathyroid hormone axis. Vitamin K, by contrast, activates proteins that bind calcium, among them osteocalcin in bone and matrix Gla protein in vessel walls ; that part is established biochemistry.

What does not follow automatically is that adding them improves outcomes in people who already get enough of both. Mechanism explains why something could work. It does not show that it does. We name the mechanism and leave the recommendation to those who may give one.

One claim in particular does not hold, and it is the one most often used to sell the combination : that vitamin D without K2 is dangerous, because the calcium would otherwise end up in the arteries. In humans this is not established. The idea traces back essentially to a single animal experiment that combined warfarin with a calcium-raising dose of vitamin D, and no trial with a hard endpoint has shown that harm in people.

Too much : where the actual risk sits

Vitamin D toxicity is rare, and it is worth being precise about it rather than dramatic. In a review of 13 documented case reports, the lowest level at which intoxication occurred was around 150 ng/ml. That is the lowest value in that series, not a proven floor for everyone, and levels in the range this article discusses are far below it. The German Federal Institute for Risk Assessment nevertheless warns specifically about high-dose single products sold as food supplements. In its opinion of September 2025 it proposes a maximum of 20 µg, that is 800 IU, per daily dose in a supplement, and it notes that the tolerable upper level was never meant to be applied to large single doses at all.

There is a second finding that sits closer to everyday use. A meta-analysis of 22 randomised trials with almost 13,000 participants examined the range of roughly 3,200 to 4,000 IU daily and found more hypercalcaemia and more falls. That is not the far end of the scale : it is the range in which many freely available high-dose products sit.

When our interpretation does not apply at all

In a number of conditions calcium regulation no longer works through the feedback loop described in the companion article. In those cases hypercalcaemia can occur at far lower levels, and the reassuring sentence about 150 ng/ml does not apply. This concerns sarcoidosis and other granulomatous diseases, certain lymphomas, a CYP24A1 mutation, Williams-Beuren syndrome, primary hyperparathyroidism and a history of kidney stones.

We name these cases rather than glossing over them, for a simple reason : knowing where our own interpretation stops is part of doing it properly. If one of them applies to you, this is a medical question from the outset.

When measuring is worth it at all

This is where we have to quote something that runs against our own commercial interest. The current Endocrine Society guideline advises against routine measurement in healthy adults, and not only in the 19 to 74 age group but also for those over 75, in pregnancy, with obesity and with darker skin, because no threshold could be tied to a specific benefit. So a measurement makes sense when there is a reason for it : signs of a shortfall, little exposure to daylight, conditions affecting absorption or the kidneys, an existing high-dose intake that nobody is checking, or the wish to know where you stand as part of a full workup rather than in isolation.

Why we do not land at 40 to 60 ng/ml and 5,000 IU

If you have read about vitamin D anywhere else, you will have met a higher target range than ours, usually 40 to 60 ng/ml, often together with a daily intake of 5,000 IU. That position is not a fringe invention, and it deserves to be named with its actual authors rather than through whoever popularised it. It comes from the Scientists' Call to Action published by GrassrootsHealth, signed by 48 researchers working in the field.

Their strongest argument is genuinely strong, and it is worth stating properly before answering it. The large trials that found no benefit largely enrolled people who were already adequately supplied. In VITAL, the participants whose blood was measured averaged 30.8 ng/ml at baseline and only about 12.7 percent were below 20 ng/ml. Anyone already supplied can hardly benefit from more, so, the argument goes, those trials never tested the question that matters. That objection is fair.

The answer to it is not a counter-opinion but a different kind of study : trials that randomise two achieved levels against each other, exactly as the criticism demands. The clearest of them is VIDAMS. It compared 600 against 5,000 IU daily over 96 weeks in people with relapsing-remitting multiple sclerosis, explicitly aiming at 60 ng/ml, and it separated the groups : the achieved levels were 30.3 against 54.0 ng/ml. This is the direct test of the higher target, and it came back empty : a hazard ratio of 1.17 with a confidence interval from 0.67 to 2.05 and a p value of 0.57. Read correctly, that does not say the high dose was worse. It says no difference was found. Two caveats belong with it : this was a trial in a specific disease, so it cannot simply be transferred to healthy people, and both arms started at around 29 ng/ml, so it tested adding more on top of an adequate supply rather than correcting a shortfall. On its own it does not settle the question. As one of five trials all pointing the same way, it carries weight.

VIDAMS does not stand alone. Burt 2019 compared 400, 4,000 and 10,000 IU daily over three years and found a dose-dependent loss of bone density at the radius, though bone strength itself did not differ significantly between the groups. Lappe 2017 found no significant effect on cancer incidence, D2d none on diabetes incidence in prediabetes, FIND none in older adults. Five trials built the way the criticism asks for, all of them null or pointing against the higher level.

Two things belong here that make the picture less tidy, and leaving them out would be the easier but worse choice. First, conflicts of interest exist on both sides. GrassrootsHealth sells testing offers and receives a share of the sales made through its partner brands, so it is not economically independent of the field its recommendation concerns. And the consensus paper whose range we ourselves adopt is not clean either : every one of its authors declares fees from a single pharmaceutical company, which also funded and organised the work, and the authors write themselves that a funding bias cannot be entirely ruled out. We take their figure, not their independence.

Second, a high dose is demonstrably right in a specific situation. Hollis 2015 showed that 6,400 IU daily in a breastfeeding mother covers the infant's requirement through breast milk, as a randomised alternative to dosing the child directly. A high dose can be well founded in a particular case without being suitable as a default. Two limitations belong with it : the middle arm at 2,400 IU was stopped in 2009 because the supply to the infants was insufficient, and only around 28 percent of mothers breastfed exclusively through month seven. This too therefore belongs with a paediatrician rather than in a coaching article.

And the point that supports neither side : the most recent guideline, from the Endocrine Society in 2024, withdrew the 30 ng/ml mark set in 2011 and did not replace it with a new one, because no threshold could be tied to a demonstrated benefit. Our range is therefore not the correct answer to a question the guideline settled. It is a reasoned working range, not a proven optimum, and we would rather say that plainly than dress a number up as more than it is.