Vitamin D: Supplement or Sunlight?

The observational literature on vitamin D is enormous and remarkably consistent. People with low 25-hydroxyvitamin D have more cancer, more heart disease, more infections, more autoimmune disease, more fractures and shorter lives. Run the correlation in almost any cohort and it appears.

The problem is that the exposure is not independent of everything else. People who are chronically ill, housebound, sedentary, poorly nourished or carrying excess weight all have lower vitamin D — because of those things, not before them. Mendelian randomisation makes the direction explicit for body weight: across 21 cohorts and up to 42,024 adults, a genetically instrumented 10% higher BMI was associated with 4.2% lower 25-hydroxyvitamin D, while genetically instrumented vitamin D had no effect on BMI at all.^[8] Low vitamin D is, in large part, a marker of being unwell rather than a cause of it.

That is the hypothesis the big trials were built to test.

VITAL randomised 25,871 American adults — men 50 and over, women 55 and over — to 2,000 IU of vitamin D3 daily or placebo for a median 5.3 years. Invasive cancer: hazard ratio 0.96 (95% CI 0.88–1.06). Major cardiovascular events: 0.97 (0.85–1.12). Death from any cause: 0.99 (0.87–1.12).^[1] The pre-planned fracture analysis, published separately, confirmed 1,991 fractures and found nothing: total fractures 0.98 (0.89–1.08), hip fractures 1.01 (0.70–1.47), with no effect modification by baseline 25-hydroxyvitamin D level.^[2]

D-Health gave 21,315 older Australians 60,000 IU monthly for five years. All-cause mortality: hazard ratio 1.04 (0.93–1.18). In an exploratory analysis excluding the first two years, cancer mortality ran numerically higher in the vitamin D arm, 1.24 (1.01–1.54) — not a result to over-read, but not the direction anyone expected.^[3]

D2d randomised 2,423 adults with prediabetes to 4,000 IU daily. New-onset diabetes: hazard ratio 0.88 (0.75–1.04), p = 0.12.^[4]

One outcome did move. The VITAL autoimmune ancillary found 123 confirmed autoimmune diseases on vitamin D versus 155 on placebo — hazard ratio 0.78 (0.61–0.99), p = 0.05.^[5] That is a single significant result at the margin, out of a very large number of outcomes examined, and it deserves confirmation rather than a headline.

The respiratory-infection literature sits in between. A meta-analysis of 46 trials and 75,541 participants found a small protective effect — odds ratio 0.92 (0.86–0.99) for having at least one acute respiratory infection — concentrated in daily dosing of 400–1,000 IU and in children aged 1 to 16.^[6] Real, but small, and not what a megadose capsule is sold on.

Read those trials carefully and the pattern is not "vitamin D does nothing." It is "giving more vitamin D to people who already have enough does nothing."

VITAL did not recruit on the basis of deficiency — participants were explicitly not selected for low vitamin D, low bone mass or osteoporosis.^[2] In D-Health, mean 25-hydroxyvitamin D in the placebo group was 77 nmol/L, comfortably sufficient.^[3] In D2d, mean baseline was 27.7 ng/mL, about 69 nmol/L.^[4] These trials asked whether sufficiency plus more is better than sufficiency. The answer is no.

The genetic evidence gives the shape of the curve. In 307,601 UK Biobank participants with 18,700 deaths over 14 years, the association between genetically predicted 25-hydroxyvitamin D and all-cause mortality was L-shaped: risk fell steeply with rising concentration up to about 50 nmol/L and then levelled off. Someone at 25 nmol/L had an estimated 25% higher odds of death than someone at 50 nmol/L — odds ratio 1.25 (1.16–1.35).^[7]

And when trials did enrol genuinely deficient people, they worked. The Decalyos trial gave 3,270 institutionalised French women averaging 84 years old 800 IU of vitamin D3 plus calcium daily; hip fractures fell 43% and all non-vertebral fractures 32% over 18 months, with parathyroid hormone down 44%.^[10] A pooled participant-level analysis of 11 trials and 31,022 people aged 65 and over found nothing overall, but a 30% reduction in hip fracture — hazard ratio 0.70 (0.58–0.86) — in the quartile with the highest actual intake, a median of 800 IU daily.^[9]

Same molecule. Different starting point. Entirely different answer.

Here the Canadian geography matters more than anything in the supplement aisle.

In the classic latitude experiment, human skin and 7-dehydrocholesterol exposed to cloudless sunlight in Edmonton at 52°N produced no previtamin D3 at all from October through March — six months of "vitamin D winter." In Boston at 42.2°N the dead zone ran November through February. At 34°N and 18°N, midwinter sunlight still worked.^[15] No amount of time outdoors in a Canadian January produces vitamin D, because the UVB simply is not reaching the ground at a useful angle.

The population data track that exactly. In the Canadian Health Measures Survey, 8% of Canadians overall were below 30 nmol/L — but 13% from November through March versus 5% from April through October. Among adults taking a vitamin D supplement, 3% were below 30 nmol/L; among those not taking one, 11%.^[17]

That said, sun is not merely a less convenient delivery system for a molecule. Irradiating human skin with UVA — which makes no vitamin D whatsoever — lowered blood pressure in 24 healthy volunteers, with falling circulating nitrate and rising nitrite, and the forearm blood-flow response persisted when nitric oxide synthase was blocked. The effect came from pre-formed nitric oxide stores in the upper epidermis, released by light, independently of vitamin D.^[16] Daylight also anchors circadian timing in a way no capsule does.

So sunlight plausibly does things a supplement cannot. It also causes skin cancer. The Canadian Dermatology Association's position is unambiguous: oral intake is "the optimal and safest way to maintain adequate levels," and "it is unsafe and unnecessary to increase sun exposure in order to maintain vitamin D levels."^[19] Tanning beds for vitamin D are indefensible.

The resolution is not to pick a side. Get daylight for its own reasons — morning light, movement, mood, blood pressure — without burning and without deliberately extending exposure. Do not rely on it for vitamin D between October and March in Canada.

The dose-response for vitamin D is not a straight line, and the upper end is not benign.

A three-year randomised trial in Calgary gave 311 healthy adults aged 55 to 70 either 400, 4,000 or 10,000 IU daily. Radial volumetric bone density fell in every group, but fell more with higher doses: −1.2% at 400 IU, −2.4% at 4,000 IU, −3.5% at 10,000 IU. Bone strength did not differ.^[12] That is the opposite of what megadosing is supposed to achieve.

Intermittent megadoses look worse. A single annual 500,000 IU dose given to 2,256 older Australian women increased falls — incidence rate ratio 1.15 (1.02–1.30) — and fractures, 1.26 (1.00–1.59), with the excess concentrated in the three months after dosing.^[11] D-Health's monthly 60,000 IU did not reduce falls either, and in participants with a BMI under 25 the odds of falling were higher on vitamin D, 1.25 (1.09–1.43).^[13]

Health Canada's reference values are modest by comparison: 400 IU daily for infants, 600 IU for ages 1 to 70, 800 IU over 70, with a tolerable upper intake level of 4,000 IU for anyone 9 and older — and the table explicitly assumes minimal sun exposure.^[18]

The 2024 Endocrine Society guideline on vitamin D for disease prevention is the most useful recent document on this question, and its conclusions are narrower than most people expect. It suggests empiric supplementation for children and adolescents aged 1 to 18, for adults over 75, in pregnancy, and in high-risk prediabetes. It suggests against empiric supplementation above the dietary reference intake in healthy adults under 75. And it found no clinical trial evidence supporting routine 25-hydroxyvitamin D screening in the general population — including in people with obesity or darker skin — so it suggests against routine testing in all the populations it considered.^[14]

That is not an argument against ever measuring it. Testing is reasonable where deficiency is likely or consequential: malabsorption, bariatric surgery, osteoporosis or recurrent fracture, chronic kidney or liver disease, certain medications, housebound or institutionalised patients, and people with very limited sun exposure living at Canadian latitudes. What the evidence does not support is screening everyone, or chasing a high "optimal" number once you are comfortably above the deficiency threshold.

One practical note on form: a 2025 systematic review found that people given vitamin D2 had measurably lower 25-hydroxyvitamin D3 than those who were not — roughly 9 to 18 nmol/L lower depending on the analysis — even though D2 raises total vitamin D status.^[20] If you are supplementing, D3 is the sensible default.