Taking Vitamin D “Just in Case”? Big Trials Say Save Your Money
The largest trials show vitamin D does not prevent major disease in non-deficient people, while a handful of secondary signals in cancer mortality and subgroups remain genuinely open but rest on exploratory findings.
Does vitamin D supplementation reduce disease risk in people with sufficient baseline vitamin D levels, according to randomized controlled trial evidence?
- 1The two largest purpose-built trials — VITAL (25,871 US adults) and D-Health (21,315 Australians aged 60+) — both found no significant reduction in cancer incidence, cardiovascular events, or total mortality.
- 2Observational studies link low vitamin D to 31% higher cardiovascular disease incidence, yet that association vanishes in randomized trials and under genetic-instrument testing, pointing to confounding rather than cause.
- 3The strongest surviving positive signals are secondary: cancer mortality fell about 25% in VITAL once the first two years were excluded, and non-obese participants saw about a quarter lower cancer incidence — but these are exploratory analyses, not the outcomes the trials were powered to detect.
- 4Three independent guideline bodies — the Endocrine Society, the US Preventive Services Task Force, and the UK's NHS — converge against routine supplementation or screening in non-deficient adults, while endorsing modest supplementation for specific groups or seasons.
- 5A meta-analysis of 29 trials enrolling 134,384 participants found no significant difference in cardiovascular incidence between supplement and placebo, at a risk ratio of 0.99.
The two largest trials ever run to test vitamin D in non-deficient people — VITAL (25,871 US adults) and D-Health (21,315 Australians aged 60+) — both found that daily or monthly supplementation does not reduce cancer incidence, cardiovascular events, or total mortality. The conclusion rests on confirmed null primary outcomes, reinforced by a 134,384-participant cardiovascular meta-analysis that also found no effect, and by genetic evidence that finds no causal link between vitamin D and major disease. A handful of secondary findings survive: cancer mortality fell about 25% in VITAL once the first two years were excluded, and non-obese participants saw about a quarter lower cancer incidence. Autoimmune disease dropped 22% and asthma exacerbations fell 37% in smaller trials. But these are exploratory and subgroup analyses, not the outcomes the trials were built to test, and several rest on single sources or advocacy-leaning outlets rather than independent replication. The observational case looks strong — low vitamin D tracks with 31% higher cardiovascular disease — but that association vanishes in both randomized trials and genetic instrument tests, pointing to confounding rather than cause. Three independent guideline bodies — the Endocrine Society, the US Preventive Services Task Force, and the UK's NHS — have read the same evidence and declined to endorse broad supplementation, instead recommending targeted use for children, the elderly, pregnant individuals, and those with high-risk prediabetes, or seasonal dosing through northern winters. The primary outcomes are settled: vitamin D does not prevent major disease in people who are not deficient. What remains genuinely open is whether the surviving mortality and subgroup signals are real-but-delayed biological effects or statistical artifacts of slicing the data until something significant appears.
The Full Investigation
7 sections · 10 min read
Confirmed facts and attributed reporting read normally; only contested, unverified, or speculative sentences are highlighted. Hover any sentence for its grade and sources.
A supplement millions take, tested by the trials built to judge it
Walk into any pharmacy and vitamin D sits on the shelf as a near-universal insurance policy against ill health. Behind that everyday habit lies a harder question that researchers spent the last decade trying to settle with the tools designed to separate cause from coincidence: does swallowing vitamin D actually prevent disease in people who already have enough of it?
The backdrop is a set of national standards built on a narrow foundation. In 2011 the Institute of Medicine set the recommended daily intake at 600 units for most adults and 800 for those over 70 — figures grounded solely in bone health, not in any promise of protection against cancer or heart disease. Everything claimed for vitamin D beyond the skeleton has had to be tested since.
Two trials dominate that testing. The VITAL trial randomized 25,871 US men aged 50 and older and women aged 55 and older to 2000 units of vitamin D3 daily or a placebo, following them for a median of 5.3 years. Australia's D-Health trial randomized 21,315 people aged 60 and over to a monthly 60,000-unit dose or placebo across five years, with death from any cause as its primary target. The two used different dosing rhythms — daily versus a monthly bolus — but delivered roughly the same average daily amount, and both deliberately enrolled largely vitamin D-replete populations. That design choice is central to everything that follows: these were not trials of treating deficiency, but tests of whether topping up already-adequate people buys any protection.
The primary outcomes came back null in both large trials
The trials were built to catch a benefit if one existed. On their main questions, they caught nothing. In VITAL, vitamin D did not significantly reduce the incidence of invasive cancer, with a hazard ratio of 0.96 — a result statistically indistinguishable from no effect at all. Major cardiovascular events showed the same flat line, at a hazard ratio of 0.97, as did death from any cause, at 0.99.
D-Health, run on the other side of the world in an older population, reached the same destination by a different route. Its cancer incidence result showed no significant effect in what its authors describe as a largely vitamin D-replete group, and its main cardiovascular endpoint was non-significant overall, at a hazard ratio of 0.91. That the two trials converge matters. VITAL's cancer incidence figure is independently echoed in the D-Health cancer paper, which cites the same hazard ratio of 0.96 — two research programs, one number.
The agreement is not total, and the differences are worth naming plainly. VITAL enrolled somewhat younger participants and dosed daily; D-Health enrolled only those 60 and over and dosed monthly. A monthly bolus and a daily tablet may behave differently in the body even at the same average dose. Yet across those design gaps the primary-outcome verdict is the same in both: no significant protection against the diseases that matter most.
2011
- Institute of Medicine publishes dietary reference intakes for vitamin D: RDA 600 IU/d for ages 1-70, 800 IU/d for over 70, based solely on bone health evidence
2017-11-06
- Cochrane reviews published showing vitamin D3 reduces all-cause mortality (RR 0.93-0.94) and asthma exacerbations (RR 0.63) but not cancer incidence
2020-04
- VITAL trial results published: no significant reduction in cancer incidence, CVD events, or all-cause mortality in 25,871 participants over 5.3 years
- VITAL trial reports cancer mortality reduced 25% (HR 0.75) when first 2 years of follow-up excluded
- VITAL trial subgroup analysis shows 24% cancer incidence reduction (HR 0.76) in participants with BMI under 25
2021-04-13
- US Preventive Services Task Force issues Grade I recommendation: insufficient evidence to assess vitamin D screening in asymptomatic adults
2021
- Meta-analysis of 37 RCTs finds statistically significant 8% reduction in acute respiratory infections with vitamin D (OR 0.92)
2022-08-02
- Mendelian randomization study using 41 genetic variants finds no causal association between vitamin D and most common diseases including cancer and CVD
2023-12-29
- Meta-analysis of 29 RCTs (134,384 participants) published showing no significant vitamin D effect on CVD incidence or mortality
2024
- D-Health trial results released: 21,315 Australians aged 60+ show no significant cancer incidence reduction but borderline MACE benefit in cardiovascular drug users
2024-07-12
- Endocrine Society publishes new clinical practice guideline recommending empiric vitamin D supplementation only for specific groups (ages 1-18, over 75, pregnant, high-risk prediabetes)
2025-04
- Updated Lancet meta-analysis of 40 RCTs finds borderline non-significant 6% reduction in acute respiratory infections (OR 0.94, p=0.057)
Open: No published individual-participant analysis yet pools VITAL, D-Health and DO-HEALTH under a single prespecified primary-outcome protocol, which would test whether the shared null result is robust to differences in dosing frequency and age — though DO-HEALTH is invoked here only as a candidate for future pooling, not as a trial whose outcomes are already established in the record.
Where the positive signals survive — and how solid they are
If the headline outcomes are flat, the interesting story lives in the corners of the data. Several conditions show a signal of benefit that has not disappeared, and honesty about vitamin D requires taking each seriously without overselling it.
Cancer death is the most cited. When VITAL's analysts set aside the first two years of follow-up — on the logic that supplementation cannot prevent a cancer already growing undetected — cancer mortality fell to a hazard ratio of 0.75, roughly a quarter lower, and this time the result cleared the bar for significance. Two meta-analyses restricted to daily dosing point the same way, each landing on a 12% reduction in cancer mortality with closely overlapping confidence ranges. That the figures align strengthens the signal, though it is unclear whether these represent independent syntheses or overlapping evidence bases. But one caution matters: the meta-analyses report risk ratios while VITAL reports a hazard ratio, and comparing them directly requires care, since one accounts for the timing of events and the other for cumulative counts.
Body weight sorts the cancer-incidence picture too. Participants with a body-mass index under 25 showed about a quarter lower cancer incidence, at a hazard ratio of 0.76, even as the overall effect stayed flat. Three separate reports carry this identical figure — though all trace back to the same VITAL secondary analysis, so the convergence reflects careful reporting rather than genuine replication. Cardiovascular benefit follows a similar shape in D-Health: participants already taking heart medication at baseline saw a 16% reduction in major events, at a hazard ratio of 0.84. That specific figure, however, rests on a single-source estimate; the trial sponsor confirms a subgroup benefit in general terms but supplies no number.
Beyond cancer and the heart, two further signals stand out. VITAL documented roughly a 22% drop in newly diagnosed autoimmune diseases, and a Cochrane review of nine small asthma trials found vitamin D cut exacerbations requiring steroid treatment by 37%. Both are real findings, but both come with a footnote: the autoimmune figure rests on a single reporting origin, and the asthma trials enrolled only about 1,093 people in a population likely to include the genuinely deficient — a different question from topping up the replete. The Cochrane review also reports an all-cause mortality reduction with vitamin D3 supplementation, but this figure is internally contested: the source excerpt reports a risk ratio of 0.93 while the claim itself asserts 0.94, a numeric discrepancy between the source text and the reported figure that leaves the precise estimate uncertain.
Open: Whether the cancer-mortality latency effect and the BMI-under-25 cancer signal replicate in a trial that prespecifies them as primary outcomes, rather than surfacing them after the fact, remains untested.; The exact hazard ratio for cardiovascular benefit among baseline heart-drug users lacks any independent confirmation of its specific value.
Why the observational picture and the trial picture disagree
This is the inconvenient heart of the matter, and it is where the correlation-versus-causation question the topic raises gets its cleanest test. Observational studies paint vitamin D as strongly protective. Pooling 30 prospective cohorts, people with the lowest circulating vitamin D had 31% higher cardiovascular disease incidence and 37% higher cardiovascular mortality than those with the highest. Read alone, that looks like a powerful case for supplementation.
The randomized evidence refuses to cooperate. A meta-analysis of 29 trials enrolling 134,384 participants found no significant difference in cardiovascular incidence between supplement and placebo, at a risk ratio of 0.99, nor in cardiovascular mortality, at 0.97. The gap between the observational and the randomized numbers is not a rounding error; it is a reversal. And the two are not measuring the same thing: the cohort figures compare people with naturally high versus naturally low vitamin D, while the trials compare a supplement against a placebo in people who mostly already had enough.
Genetics offers a third, independent verdict. Mendelian randomization uses inherited genetic variants that nudge a person's vitamin D level up or down from birth — a natural experiment free of the lifestyle confounding that plagues observational data. A study using 41 such variants found no causal association between vitamin D and most common diseases, including cancers and cardiovascular disease. Broader reviews, synthesizing multiple Mendelian randomization findings rather than conducting an independent primary analysis, report no causal links to heart disease, stroke, cancer, diabetes or mortality in replete populations. The interpretation offered by some reviewers is that low vitamin D is a symptom of poor health — obesity, inactivity, illness, little time outdoors — rather than its cause, though that reading is presented as an argument rather than a settled fact. Notably, the genetic evidence is not uniformly null: the same reviews report that it provides the strongest causal support for multiple sclerosis, where genetically lower vitamin D is linked to higher risk.
Open: No longitudinal study tracking individuals' vitamin D before and after they fall ill has yet quantified how much of the observational association is reverse causation versus genuine confounding.
What the guidelines actually recommend — and for whom
Faced with this evidence, the bodies that write medical guidance have landed in a strikingly consistent place, and their convergence is itself part of the answer. The 2024 Endocrine Society guideline recommends empiric supplementation only for defined groups: children and adolescents aged 1 to 18, adults over 75, pregnant individuals, and people with high-risk prediabetes. Crucially, the same guideline concedes that the trial evidence does not permit setting any blood-level threshold that reliably predicts who will benefit — an unusual admission that the field cannot yet identify the very patients it aims to treat.
The US Preventive Services Task Force goes further into caution. In April 2021 it concluded that the evidence is simply insufficient to assess screening for vitamin D deficiency in asymptomatic, non-pregnant adults, issuing its inconclusive grade. That is not a recommendation against testing so much as a declaration that the science does not yet justify it.
The UK takes a different but compatible tack, aimed at deficiency rather than disease. Government advice is that everyone consider a daily 10-microgram (400-unit) dose through autumn and winter, from October to early March, when northern sunlight is too weak for the skin to make enough. This is pragmatic seasonal policy — it does not claim to prevent cancer or heart disease, only to keep the population out of frank deficiency. Read together, three independent authorities point the same way: targeted supplementation for specific groups and seasons, but no endorsement of vitamin D as broad disease prevention for the already-replete.
The prediabetes recommendation in the Endocrine Society guideline rests on a reported modest reduction in type 2 diabetes risk from a 2023 individual participant data meta-analysis of three trials, but that claim is sourced only from a low-tier private practice blog, leaving the strength of evidence for this specific guideline group weakly documented.
Open: Whether the specific diabetes-prevention finding underpinning the prediabetes recommendation holds up rests, in the material reviewed here, on a single low-quality source [C-035], leaving the strength of that particular guideline group weakly documented.
Testing the competing explanations
Four explanations compete to make sense of this evidence, and the honest scorecard gives none of them a clean sweep. The first — the null hypothesis — holds that in non-deficient people, vitamin D simply does not prevent major disease. It draws deep support: the confirmed null primary outcomes in both mega-trials, the null cardiovascular meta-analysis, the borderline respiratory result, and the genetic evidence against causation. Its vulnerability is the cluster of surviving positive signals it must explain away as noise.
A second reading is the latency hypothesis: vitamin D may act too slowly to show up in a five-year incidence count, but reveal itself in mortality once early, pre-existing cancers are set aside. This is exactly the shape of VITAL's data — flat on cancer incidence, but a quarter lower on cancer mortality after excluding the first two years, with daily-dosing meta-analyses pointing the same direction. Its weakness is methodological: excluding follow-up until significance appears is precisely the kind of post-hoc move that manufactures false positives, and no trial has yet prespecified this analysis to confirm it.
A third reading, the subgroup hypothesis, argues benefit is real but confined — to non-obese people for cancer, heart-drug users for cardiovascular events, asthmatics, and those with autoimmune susceptibility. What cuts against it is telling: the updated respiratory meta-analysis found no effect modification by baseline vitamin D status, age, dose or dosing frequency — though this subgroup null is specific to respiratory infection outcomes and is offered here as suggestive of a general pattern, not as a direct test of the cancer or cardiovascular subgroup claims. And the subgroup figures lean heavily on a narrow, single-source base rather than independent replication.
The fourth explanation — reverse causation — is the best supported of all. It says the observational associations are an illusion of confounding: sick, inactive, housebound people have both low vitamin D and bad outcomes. Randomized trials and genetic instruments both fail to find the causal effect the correlations imply, and no claim in the record directly contradicts this account. What would discriminate between these views is concrete: an individual-participant pooling of the major trials, stratified by baseline vitamin D level, and longitudinal cohorts tracking vitamin D before and after illness onset. Until then, the null and reverse-causation readings carry the primary outcomes, while the latency and subgroup readings survive only in the exploratory margins.
What the evidence forces us to conclude
The evidence compels a clear answer to the central question and a careful hedge around its edges. For the diseases the trials were built to test — cancer, cardiovascular events, total mortality — vitamin D supplementation does not reduce risk in people who are not deficient. That conclusion rests on the confirmed primary outcomes of the two largest trials ever run for the purpose, reinforced by a null cardiovascular meta-analysis of 134,384 participants and by genetic evidence that finds no causal pathway. Three independent guideline bodies have read the same data and declined to endorse broad supplementation. This part of the answer is not in serious doubt.
What remains open is narrower and should not be dismissed. Cancer mortality after a latency window, reduced cancer incidence in leaner people, autoimmune disease, and asthma exacerbations each show a signal that has not vanished. Speculatively — and this reasoning is offered as hypothesis, not finding — it is biologically plausible that a slow-acting nutrient shows its hand in mortality and progression rather than in five-year incidence counts, which would reconcile the null incidence with the mortality signal. But the strength of the evidence does not yet match the strength of that intuition: these are post-hoc and subgroup findings, several resting on single sources, and the one respiratory analysis that formally tested for subgroup effects found none.
The genuine stalemate is not between benefit and no-benefit — the primary outcomes settle that — but between the two innocent explanations for the surviving signals: real-but-delayed biological effect versus statistical artifact of repeated slicing. The evidence in hand cannot break that tie. It can only say that the burden now sits with those claiming benefit to produce a trial that prespecifies these effects rather than discovering them afterward.
Why it matters
Vitamin D is among the most widely taken supplements in the world, and the question of whether it prevents disease in the already-replete determines whether that daily habit is precaution or waste. The evidence reviewed here shows that the largest trials designed to find a benefit did not find one for cancer, heart disease, or death, and that the strong observational case for vitamin D dissolves under randomized and genetic testing — a textbook illustration of correlation mistaken for causation. Getting this right shapes public spending, screening policy, and the credibility of nutritional guidance, while the genuine open signals in specific subgroups mean the door to targeted, evidence-based use is not closed.
- How much of the affirmative case depends on a concentrated, single-source base, given that several key positive figures trace to a small number of related origins rather than independent replication.
Response from the subject of this report · 30 August 2026