Vitamin D

Vitamin D is genuinely essential, real deficiency is worth finding and correcting, and the people most likely to be low are identifiable in advance. The claim you have probably seen — that 95% of Americans are deficient — is a statistic about food, not about blood, and it traces back to an analysis written by employees of a supplement manufacturer.² Measured in blood, about 5% of Americans are at risk of deficiency.³

Read the grid above narrowly. Each grade answers one question: does taking a vitamin D supplement deliver on this specific promise? It does not ask whether vitamin D matters — you would have rickets without it. Most of the letters are D because the largest randomized trials tested the nutrient in people who were not short of it, and came back null. We sell nothing on this page, and every number below carries its source.

What vitamin D actually is, and why it is barely a vitamin

A vitamin is something you have to eat. Vitamin D mostly is not. Your skin makes it from ultraviolet light, the liver converts it to 25-hydroxyvitamin D — 25(OH)D, or calcidiol, the storage form a blood test measures — and the kidney converts that into the active hormone calcitriol.¹ Its central job is to promote calcium absorption in the gut and hold serum calcium and phosphate at levels that allow normal bone mineralisation; severe, sustained shortfall causes rickets in children and osteomalacia in adults, both diagnosable medical conditions and the reason this nutrient has a recommended intake at all.¹

Two facts about supply do most of the work here. Very few foods contain vitamin D naturally, so fortified foods — mainly milk, at about 3 mcg (120 IU) per cup — supply most of what Americans get from diet.¹⁶ And the amount your skin makes swings with season, latitude, cloud cover, smog, age, sunscreen and melanin: the darker your skin, the less you make from the same sunlight.¹⁶

So what you eat is a poor proxy for what you have. Serum 25(OH)D is the main indicator of vitamin D status, and it reflects vitamin D made in the body as well as what came from food and supplements.¹ Dietary intake covers only the second half of that.

The "95% of Americans are deficient" claim — what is actually true

This is the most repeated number in the category, and it is being used to mean something it does not mean.

Where it comes from. A 2020 analysis in Nutrients estimated usual micronutrient intakes in 26,282 US adults aged over 19 from the 2005–2016 NHANES surveys, and reported that 95% fell below the estimated average requirement for vitamin D.² That is real, correctly calculated and peer-reviewed. It is also, per the section heading it sits under, a figure for intake from food only.²

The number the same paper reports next. Include dietary supplements and inadequacy in that population drops from 95% to 65%.² The 95% is almost never quoted with the 65% beside it.

Who wrote it. All five authors are employed by Pharmavite LLC, a dietary supplement manufacturer, as the paper discloses in its conflict-of-interest statement; the research received no external funding, and its own conclusion is that dietary supplements can help address these gaps.² None of that makes the arithmetic wrong. It does mean the most-cited statistic in vitamin D marketing originated inside the industry doing the marketing, about a food supply that was never the main source in the first place.

What blood says. The CDC's National Center for Health Statistics measured serum 25(OH)D in 16,180 Americans aged 1 and over in NHANES 2011–2014. 5.0% (95% CI 4.1–6.2) were at risk of deficiency, below 30 nmol/L (12 ng/mL), and 18.3% (16.2–20.6) at risk of inadequacy, 30–49 nmol/L (12–19 ng/mL).³ Among adults the deficiency figure peaks at 7.6% in the 20–39 band and falls to 2.9% at 60 and over.³

Why the two differ. 95% is a statement about diets; 5% is a statement about people. Anyone quoting the first while implying the second is either confused or selling something.

How blood levels are graded

Serum 25(OH)D		Interpretation
Below 30 nmol/L	Below 12 ng/mL	Associated with vitamin D deficiency
30 to under 50 nmol/L	12 to under 20 ng/mL	Generally considered inadequate
50 nmol/L and above	20 ng/mL and above	Generally considered adequate
Above 125 nmol/L	Above 50 ng/mL	Linked to potential adverse effects

Thresholds from the NIH Office of Dietary Supplements.¹ Labs report both units, and mixing them up by a factor of 2.5 is a common error — 30 in one column is a serious shortfall, 30 in the other is fine.

Who is actually likely to be low

The gap is large: at-risk-of-deficiency runs 17.5% among non-Hispanic Black Americans, 7.6% among non-Hispanic Asian Americans, 5.9% among Hispanic Americans and 2.1% among non-Hispanic white Americans.³ That eight-fold spread is the most useful targeting fact on this topic, and it is missing from almost every consumer page about vitamin D.

The NIH lists the groups most likely to fall short: breastfed infants, older adults, people with limited sun exposure, people with dark skin, people with conditions that limit fat absorption such as coeliac disease, Crohn's disease, ulcerative colitis and cystic fibrosis, and people with obesity or a history of gastric bypass surgery.¹

Notice what that list is made of — skin tone, daylight, age, surgical history, absorption. It is a list of circumstances, not symptoms. There is no "tired and achy? you might be deficient" checklist here, because that is diagnosis dressed as nutrition. Low vitamin D is confirmed with a blood test or not at all.

What the evidence actually supports, grade by grade

Correcting a shortfall and supporting bone mineralisation

AStrong evidence

Two As, and they carry everything else. Supplementing raises serum 25(OH)D — every trial below confirms it as a side effect of testing something else — and the physiology is not in dispute: vitamin D promotes calcium absorption in the gut and maintains the serum calcium and phosphate concentrations normal bone mineralisation requires.¹ Grade A is not a claim that vitamin D does everything; it is the plain statement that a vitamin D supplement fixes a vitamin D shortfall. The rest of this page asks whether fixing it changes anything else.

Bone mineral density

DInsufficient evidence

A systematic review pooled 23 randomized trials of vitamin D and bone mineral density in 4,082 participants. Density was measured at up to five skeletal sites per study, producing 70 tests of statistical significance: six showed benefit, two showed harm, and the rest were null.⁶ Pooled, there was a small benefit at the femoral neck — a weighted mean difference of 0.8% (95% CI 0.2–1.4), with substantial heterogeneity — and no effect at any other site, including the total hip, with a bias toward positive results recorded at exactly those two sites.⁶

Six hits in 70 tests, at one site, with publication bias detected, is what noise looks like. The authors concluded that continued widespread use for osteoporosis prevention in community-dwelling adults without specific risk factors for deficiency "seems to be inappropriate."⁶ We grade this D rather than the C most summaries imply, because the pooled estimate does not survive contact with how it was produced.

Fractures

DInsufficient evidence

Density only matters if it moves the outcome anyone cares about. A fracture ancillary to VITAL followed 25,871 adults on 2,000 IU (50 mcg) of vitamin D3 a day or placebo for a median 5.3 years and confirmed 1,991 fractures. Total fractures HR 0.98 (95% CI 0.89–1.08, P=0.70), nonvertebral 0.97 (0.87–1.07, P=0.50), hip 1.01 (0.70–1.47, P=0.96).⁵ Flat.

Two caveats, in opposite directions. In fairness to vitamin D, participants were not recruited on the basis of deficiency, low bone mass or osteoporosis, so this establishes "no benefit in a broadly replete population," not "no benefit ever." Cutting the other way, and far less often quoted: there was no modification of the treatment effect by baseline serum 25(OH)D, or by age, sex, race or BMI.⁵ The lower-vitamin-D participants did not do better. The USPSTF likewise found the evidence insufficient to evaluate the benefits and harms of supplementation to prevent fractures.¹

Falls

DInsufficient evidence

The USPSTF reviewed seven trials in community-dwelling adults aged 65 and over without osteoporosis or vitamin D deficiency, and concluded with moderate certainty that supplementation does not reduce the number of falls or the injuries resulting from them.¹ "Moderate certainty of no effect" is stronger than "we don't know," and it is the answer.

Cancer — incidence and mortality

DInsufficient evidence

Two questions with two different answers, and conflating them is how this claim gets oversold.

Do you get cancer less often? No. In VITAL, invasive cancer of any type occurred at HR 0.96 (95% CI 0.88–1.06, P=0.47).⁴ A meta-analysis of ten trials covering 6,537 cases agreed, at a summary relative risk of 0.98 (0.93–1.03, P=0.42) — still null where attained blood levels exceeded 100 nmol/L (40 ng/mL).⁷ Grade D.

Do people die of cancer less often? Here there is a signal. The same meta-analysis pooled five trials covering 1,591 cancer deaths and found a summary relative risk of 0.87 (0.79–0.96, P=0.005) — a 13% reduction, largely attributable to daily rather than bolus dosing.⁷ Note the arithmetic: incidence rests on ten trials, mortality on five. VITAL alone pointed the same way without reaching significance, at HR 0.83 (0.67–1.02).⁴

Both can be true — a supplement that does not change whether a tumour appears could still change how a diagnosed cancer behaves. Hence C for cancer mortality in the grid: a real, replicated, modest signal from a narrower evidence base than the headline suggests. What nobody should write, and we will not, is that vitamin D stops people getting cancer. Ten trials say it does not.

Cardiovascular events

DInsufficient evidence

VITAL's other primary endpoint, a composite of myocardial infarction, stroke or cardiovascular death: HR 0.97 (95% CI 0.85–1.12, P=0.69) across 25,871 participants over 5.3 years.⁴ Secondaries ran the same way — myocardial infarction 0.96, stroke 0.95, cardiovascular death 1.11, none significant — and all-cause mortality was 0.99 (0.87–1.12).⁴

Mood and depression

DInsufficient evidence

VITAL-DEP randomized 18,353 adults aged 50 and over to 2,000 IU (50 mcg) a day or placebo for a median 5.3 years. Depression or clinically relevant depressive symptoms: HR 0.97 (95% CI 0.87–1.09, P=0.62). Change in PHQ-8 mood score versus placebo: 0.01 points (95% CI −0.04 to 0.05), against a minimal clinically important difference of 0.5.⁹ That is not a small effect; it is no effect, measured precisely enough that a real one would have shown.

The objection is that this trial enrolled people without clinically relevant depressive symptoms at baseline. Fair — but a meta-analysis of nine trials in 4,923 adults who had been diagnosed with depression or depressive symptoms also found no significant reduction in symptoms.¹ Low vitamin D is genuinely associated with depression in observational data; giving people vitamin D has not moved it.

Testosterone

DInsufficient evidence

The most recent meta-analysis of randomized trials pooled 11 comparisons and found no clear effect on total testosterone (mean difference 0.47 nmol/L, 95% CI −0.50 to 1.44), nor on sex hormone-binding globulin or free androgen index, with GRADE certainty rated low.¹³

Two trials from the same group in Graz closed the loophole people reach for. The first gave 20,000 IU of vitamin D3 weekly for 12 weeks to 98 men with normal testosterone whose 25(OH)D averaged 53.3 nmol/L (21.3 ng/mL): no effect on total testosterone, and a significant decrease in the QUICKI insulin-sensitivity index.¹¹ The second ran the identical protocol in 94 men with low testosterone whose 25(OH)D averaged 56.3 nmol/L (22.5 ng/mL): no effect on total testosterone or any secondary outcome.¹² Note what those baselines are: both cohorts averaged above the 50 nmol/L adequacy line in the table above. The trials screened at under 75 nmol/L, not for deficiency. The closest anyone has come is the second trial's subgroup of 39 men who genuinely were below 50 — and that came back null as well, at p=0.725.¹² Frankly deficient men, below 30 nmol/L, have never been enrolled.

The full argument is in our guide to vitamin D and testosterone; ashwagandha for men covers the other end of that shelf.

Two things we have not graded, and why

Respiratory infections. A single letter would misrepresent this literature. A meta-analysis of 46 trials in 75,541 people found fewer participants on vitamin D had at least one acute respiratory infection than on placebo — 61.3% versus 62.3%, OR 0.92 (95% CI 0.86–0.99).¹⁰ Significant; an absolute difference of one percentage point. The subgroups look far better — OR 0.78 for daily dosing, 0.70 at daily equivalents of 400–1,000 IU (10–25 mcg) — but the paper reports no significant interaction between treatment allocation and dose, dose frequency, study duration or age, and no effect in any subgroup defined by baseline 25(OH)D.¹⁰ The flattering subgroup numbers are not statistically distinguishable from the unflattering overall one, and the authors call the risk reduction small.¹⁰

Immune function generally, autoimmune disease, diabetes and COVID-19. All are claimed for vitamin D. We have not assessed them at primary sources, so we are not grading them. An ungraded topic is a gap we will fill, not a benefit we are hinting at.

Why the observational and randomized evidence disagree

Observational studies consistently link low vitamin D to worse outcomes. Randomized trials keep coming back null. Both bodies of evidence are large, so the disagreement is structural, not a case of one side being sloppy.

The most parsimonious reading: low vitamin D is partly a marker of poor health rather than a cause of it. People who are ill, sedentary, indoors or carrying more body fat have lower 25(OH)D for reasons a capsule does not fix — the NIH notes plainly that people with a BMI of 30 or more have lower serum 25(OH)D.¹ A cohort study cannot separate "low vitamin D made them sick" from "being sick made their vitamin D low." A trial can, which is why trials are the harder test and why they keep failing it.

The fair counter-argument is that those trials largely enrolled people who were not deficient.⁵ True — but where anyone has looked directly, the deficient subgroups have not rescued the result: no effect modification by baseline 25(OH)D in VITAL's fractures⁵, none in any baseline-status subgroup of the respiratory meta-analysis¹⁰, and a null in the one testosterone-trial subgroup that was actually below the 50 nmol/L line.¹² That is not proof that correcting a real deficiency is pointless; it is a reason to be careful about calling it powerful.

Vitamin D2 vs D3 — which should you take?

D3 (cholecalciferol) is the form your skin makes, found in animal foods and most supplements. D2 (ergocalciferol) comes from plants and fungi, and is used in some high-dose prescription products.

Both absorb well. They differ in what happens next: most evidence indicates D3 raises serum 25(OH)D to a greater extent than D2 and maintains the higher level for longer.¹

The verdict: choose D3. The exception is if you want a fully plant-based product — and even then, D3 sourced from lichen exists, so vegan and D3 are no longer mutually exclusive. There is no good reason for someone with a choice to take D2.

How much vitamin D do you need — and how much is too much

The recommended dietary allowance is 600 IU (15 mcg) a day for adults 19 to 70, rising to 800 IU (20 mcg) from 71 — and the same 600 IU (15 mcg) in pregnancy and while breastfeeding. The tolerable upper intake level for adults is 4,000 IU (100 mcg) a day from all sources.¹

Which brings us to the dose people actually search for. Is 5,000 IU too much? Three things are true at once, and most pages tell you only one:

1. It is well above the recommended intake — more than eight times the 600 IU (15 mcg) RDA.¹
2. It is above the upper limit of 4,000 IU (100 mcg) a day, the level set as safe for indefinite daily use in the general population.¹
3. It is well below where toxicity appears. The NIH's framing is that signs of toxicity are unlikely at daily intakes below 250 mcg (10,000 IU), while cautioning that lower intakes could still have adverse effects over the long term.¹

So 5,000 IU (125 mcg) is not a poisoning risk, and not a casual choice either. It sits between "more than anyone has established you need" and "enough to make you ill" — the territory where you want a blood level and a clinician rather than a shelf label.

Can you take too much vitamin D?

Yes — and unlike most vitamins, this is not theoretical.

Because vitamin D increases calcium absorption, excess causes hypercalcemia: blood calcium above 11.1 mg/dL against a normal range of 8.4–10.2 mg/dL, typically alongside serum 25(OH)D above 375 nmol/L (150 ng/mL).¹ The effects the NIH describes run from nausea, vomiting, muscle weakness, excessive thirst and kidney stones through, in extreme cases, renal failure, soft-tissue calcification including coronary vessels and heart valves, arrhythmias and death.¹ That is a description, not a self-management guide: hypercalcemia is diagnosed and treated by a clinician, and if you recognise yourself in that list the response is a phone call, not a change to your supplement routine.

Two reassurances worth stating precisely. Sun does not do this — experts do not believe excessive sun exposure causes vitamin D toxicity, because heat converts previtamin D3 in the skin into forms that limit how much vitamin D3 can be made.¹ Food does not either: toxicity is almost always the result of excessive intake from supplements.¹ It is a supplement problem, exclusively.

The interaction callout near the top of this page is the other part worth reading: orlistat, statins, corticosteroids and thiazide diuretics all interact with vitamin D.¹ If any are in your medicine cabinet, this is a pharmacist conversation.

When and how to take it

Vitamin D is fat-soluble, so take it with a meal that contains some fat. That is the whole of the mechanical advice, and the part that changes how much you absorb.

Time of day matters far less than consistency. Daily and weekly regimens both maintain status — the Graz trials above used 20,000 IU (500 mcg) once a week and raised levels fine¹¹ — but where dosing frequency has mattered to an outcome, it favoured daily over bolus dosing, in both the cancer-mortality⁷ and respiratory¹⁰ analyses. Absent a reason to prefer weekly, take it daily with a meal.

Vitamin D with K2, magnesium and calcium

Magnesium has the clearest mechanism: the enzymes that metabolise vitamin D in the liver and kidney appear to require magnesium as a cofactor, which is a real reason not to be short of both at once.¹⁵ Whether adding magnesium to a vitamin D supplement improves any outcome has not been shown, so treat it as a reason to meet your magnesium requirement rather than to buy a combination product — our magnesium hub grades that mineral on the same A-to-D scale.

Vitamin K2 deserves more scepticism than it gets. The rationale is coherent: vitamin K activates proteins that direct calcium into bone, so pairing it with a nutrient that increases calcium absorption sounds sensible. The evidence is thinner than the pairing's popularity. The trials most often cited for K2 and fractures used MK-4 at 15 or 45 mg a day, a pharmacological dose used in Japan as an osteoporosis treatment, against average dietary vitamin K intakes of 122 mcg for women and 138 mcg for men — roughly 300-fold apart.¹⁴ Later trials found no effect of vitamin K supplementation on bone mineral density in older men or women, and the NIH calls vitamin K's role in arterial calcification and coronary heart disease risk unclear.¹⁴ A typical D3+K2 capsule delivers around 100 mcg of K2, not 45 mg. We are not aware of evidence that adding it improves any outcome, and we will not imply otherwise.

Calcium is what vitamin D exists to help you absorb, so meeting your calcium intake is part of the same picture — a food question before a supplement question.¹

Do you need a vitamin D blood test?

The test is serum 25-hydroxyvitamin D, a nutrient-status measurement rather than a diagnosis, reflecting both what your skin made and what you ate.¹ It was also, by one count, the fifth most commonly ordered laboratory test among Medicare Part B payments.³

That volume is the problem. The US Preventive Services Task Force's 2021 assessment of screening asymptomatic, community-dwelling, nonpregnant adults is an "I statement" — the current evidence is insufficient to assess the balance of benefits and harms.⁸ That is not a green light with caveats. It means the Task Force could not establish that finding and correcting low vitamin D in people without symptoms makes them better off.

So the honest recommendation is targeted, not universal:

• Worth discussing with a clinician if you are in one of the risk groups above — limited sun exposure, darker skin, older age, obesity or a history of gastric bypass, or a condition affecting fat absorption.¹
• Not supported as routine screening if you are a healthy adult with none of those factors. Testing everyone is a habit, not a recommendation.

"Get tested" is the most commercially convenient sentence available on this topic. It is also the one the evidence supports least, and this page has no test to sell.

Frequently asked questions

What does vitamin D do?

It lets your body absorb calcium and phosphorus and keeps bone mineralisation normal.¹ Your skin makes most of it from sunlight; food and supplements make up the rest.¹

Are 95% of Americans really vitamin D deficient?

No. That figure describes how many fall short of the recommended dietary intake from food alone, and most vitamin D comes from skin rather than food.² Measured in blood, about 5% of Americans are at risk of deficiency and another 18% at risk of inadequacy.³

How much vitamin D should I take?

The recommended intake is 600 IU (15 mcg) a day for adults up to 70 and 800 IU (20 mcg) from 71, with an upper limit of 4,000 IU (100 mcg) a day from all sources.¹

Is 5,000 IU of vitamin D too much?

It is above both the recommended intake and the 4,000 IU (100 mcg) upper limit, though still well below the roughly 10,000 IU (250 mcg) a day at which signs of toxicity appear.¹ It is a dose worth discussing with a clinician rather than picking off a shelf.

What is the difference between vitamin D and D3?

D3 (cholecalciferol) is the form your skin makes; D2 (ergocalciferol) comes from plants and fungi. Most evidence indicates D3 raises blood levels further and sustains them longer, so D3 is the default choice.¹

When should you take vitamin D?

With a meal containing some fat, since it is fat-soluble.¹ The time of day matters much less than taking it consistently.

Can you take too much vitamin D?

Yes. Excess raises blood calcium, and it almost always comes from supplements rather than sun or food.¹ That is a medical situation for a clinician, not something to manage yourself.

Do I need a vitamin D blood test?

It is worth discussing if you are in a higher-risk group — limited sun exposure, darker skin, older age, obesity, or a condition affecting fat absorption.¹ For healthy adults with no risk factors, routine screening is not currently recommended; the USPSTF rates the evidence insufficient.⁸

The bottom line

Vitamin D is essential, and correcting a genuine shortfall is cheap and sensible. What it is not is the population-wide emergency the category describes. The 95% figure is about diets, comes from a supplement manufacturer's own analysis, and sits next to a 65% figure in the same paper that nobody quotes.² Measured properly, about 5% of Americans are at risk of deficiency — concentrated heavily among people with darker skin, less sun and more body fat.³

If that is you, a modest daily D3 with a meal is reasonable, and worth a blood test conversation. If it is not, the largest randomized trials ever run on this nutrient say a supplement will not lower your risk of fractures, falls, cancer, cardiovascular events or depression.⁴⁵⁹

The honest uncertainty, which we would rather leave visible: those trials mostly enrolled people who were not deficient, so that question is not fully closed. Where anyone has looked directly, though, the deficient subgroups have not come out ahead. We will update this page when that changes.