In Edinburgh the sun stops making vitamin D in your skin somewhere around the end of this month and does not start again until spring. That much is not in dispute. Almost everything else about vitamin D is — and the argument is being had badly by both sides.
Every September I have a version of the same conversation. Someone tells me they’ve started taking 5,000 IU of vitamin D a day because they saw something online about immunity, or mood, or cancer. And someone else tells me they’ve stopped taking any at all because they saw something online about a huge trial showing it does nothing.
Both of them read something true. Both of them drew a conclusion the evidence does not support. And neither of them has ever had the one number that would tell them which of the two situations they are actually in.
So this month’s supplement is vitamin D, and the point of the piece is not to tell you to take it or not take it. It is to show you what the trials found, including the parts that are inconvenient for the side I’m more sympathetic to, and then to explain why the dose argument and the outcome argument are two different arguments that keep getting fused into one.
Your skin makes vitamin D when UVB radiation at roughly 290–315 nm hits 7-dehydrocholesterol in the epidermis. The amount of UVB reaching the ground depends on the angle of the sun. Above about 50° of latitude, for several months either side of the winter solstice, the sun never gets high enough in the sky for meaningful amounts of that wavelength to get through the atmosphere.
Edinburgh sits at 55.95°N. Glasgow, 55.86°N. Aberdeen, 57.15°N. For practical purposes, from roughly October to March, you can stand outside in Scotland all day and make close to none.
This is why the UK has a national recommendation at all. The Scientific Advisory Committee on Nutrition reported on this in 2016 and set a Reference Nutrient Intake of 10 µg (400 IU) per day for everyone in the UK aged 4 and over, on the basis that blood 25(OH)D concentrations below 25 nmol/L (10 ng/mL) carry increased risk of poor musculoskeletal health. The NHS translates that into the advice most people have heard: consider supplementing through the autumn and winter, and all year round if you are housebound, live in a care home, usually cover your skin outdoors, or have darker skin.
RNI: 10 µg / 400 IU per day — SACN, Vitamin D and Health, 2016.
Risk threshold: below 25 nmol/L (10 ng/mL) — same source, musculoskeletal endpoints.
Upper limit for adults: 100 µg / 4,000 IU per day — NHS, current guidance.
Hold on to that last one. It matters later.
So far, so uncontroversial. Now the fight starts.
Between 2018 and 2022 three very large randomised placebo-controlled trials reported on vitamin D and hard outcomes. They are the reason you keep seeing headlines saying vitamin D does nothing.
VITAL randomised 25,871 US adults — men over 50, women over 55 — to 2,000 IU/day of cholecalciferol or placebo, for a median of 5.3 years. Invasive cancer of any type: 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). Null, null, null.
VITAL-DEP, the depression arm, followed 18,353 of those participants. Risk of depression or clinically relevant depressive symptoms: hazard ratio 0.97 (0.87–1.09). Change in mood score on the PHQ-8: a mean difference of 0.01 points, against a minimal clinically important difference of 0.5. That is not a small effect. That is no effect, measured precisely enough to say so.
D-Health randomised 21,315 Australians aged 60 and over and followed them for a median of 5.7 years. All-cause mortality: hazard ratio 1.04 (0.93–1.18). Cardiovascular death: 0.96 (0.72–1.28).
And a fourth, because bone is the claim people assume is safe: the VITAL fracture ancillary confirmed 1,991 fractures in 1,551 participants and found nothing. Total fractures 0.98 (0.89–1.08). Hip fractures 1.01 (0.70–1.47). In adults who were not selected for deficiency, low bone mass or osteoporosis, supplementation did not prevent fractures.
If you only read that far, “vitamin D does nothing” looks like a fair summary. It isn’t, and the reasons are not the ones the supplement industry usually gives.
Three findings from these same trials rarely make the summaries.
One. VITAL reported a hazard ratio of 0.83 (0.67–1.02) for death from cancer. That confidence interval crosses 1, so the trial cannot claim the effect is real. But a 17% point estimate on cancer mortality is not nothing, and it is in the same trial that gets cited as proof of nothing. I state it because leaving it out is the mirror image of the thing I’m criticising.
Two. The VITAL autoimmune arm was positive. Same trial, same 2,000 IU/day, same 5.3 years, autoimmune diseases confirmed by medical record review rather than self-report: 123 cases on vitamin D against 155 on placebo, hazard ratio 0.78 (0.61–0.99). A 22% reduction, and it reached significance. Published in the BMJ in 2022. This is the single best randomised evidence that vitamin D does something at a clinical endpoint in a general adult population, and it comes from the trial people cite to say it doesn’t.
Three. D-Health’s cancer signal ran the wrong way. Cancer mortality hazard ratio 1.15 (0.96–1.39). Excluding the first two years of follow-up, 1.24 (1.01–1.54, p=0.05). The per-protocol analysis gave an odds ratio of 1.18 (1.00–1.40, p=0.06). The authors concluded that “the precautionary principle would suggest that this dosing regimen might not be appropriate in people who are vitamin D-replete.”
That is not a comfortable finding for anyone. It does not sit well with VITAL’s 0.83 for cancer death, and both are exploratory. But symmetry cuts both ways: if I’m going to quote VITAL’s non-significant benefit, I have to quote D-Health’s non-significant harm.
Two design differences separate them, and both are large enough to matter.
VITAL gave 2,000 IU daily. D-Health gave 60,000 IU once a month. Those deliver a similar average intake and a completely different physiology: a monthly bolus produces a spike in 25(OH)D followed by weeks of decline, and there is evidence that intermittent high doses behave differently from daily ones.
The respiratory-infection literature shows it most clearly. In a 2021 meta-analysis of 46 randomised trials and 75,541 participants, vitamin D reduced the risk of having at least one acute respiratory infection — odds ratio 0.92 (0.86–0.99), a genuine but small effect. When the trials were split by regimen, protection appeared in the daily trials (0.78, 0.65–0.94) and at 400–1,000 IU/day (0.70, 0.55–0.89). It did not appear in the bolus trials. The Endocrine Society’s own 2024 evidence review reached a related conclusion: high-dose intermittent dosing may increase falls compared with lower-dose daily dosing.
In blood samples taken during D-Health follow-up, the placebo group’s mean 25(OH)D was 77 nmol/L (31 ng/mL). The vitamin D group reached 115 nmol/L (46 ng/mL).
A trial of correcting a deficiency, run in a population that does not have the deficiency, is a trial of something else. Participants were recruited off the Australian electoral roll and were explicitly not screened for vitamin D status. What D-Health tested, and answered well, is whether topping up already-adequate older Australians helps. The answer was no, with a hint of harm. It is not evidence about a depleted Scot in February, and neither its supporters nor its critics usually say so.
Both of those caveats are what a careful practitioner says and a debunker drops. They also do not rescue the enthusiast position, which is the next section.
The version doing the rounds this month claims that 5,000 IU a day of vitamin D will lift depression, cut cancer risk and protect your heart.
Take those one at a time against the evidence above.
So the slide is wrong about the outcomes. But notice what it is not wrong about, at least not demonstrably: no trial here tested 5,000 IU daily. VITAL used 2,000. D-Health used a monthly bolus. The respiratory meta-analysis found its clearest signal at 400–1,000. Nobody has run a five-year hard-outcome trial at 5,000 IU a day, so nobody can tell you what it does.
That is not a defence of the slide. It is the opposite. The slide is wrong about the outcomes and unsupported about the dose — and those are two separate failures, which is exactly why picking a number off the internet is not a plan.
5,000 IU is 125 µg. The NHS upper limit for adults is 100 µg — 4,000 IU a day. A slide circulating in the UK is recommending a daily intake above the stated national upper limit, without mentioning it, without a blood test, and without knowing anything about the person reading it. Whatever you think of the outcome evidence, that is the part I would want flagged if it were my mother following it.
Having spent four paragraphs on somebody else’s slide, here is the version of it I was responsible for.
Until August this year, AIdan — the clinical assistant on this site — was telling people that the functional optimal range for 25(OH)D is 125–175 nmol/L (50–70 ng/mL). It stated it confidently, in the same register as trial data, and it stated it consistently: in an internal evaluation the figure appeared identically in every single panel interpretation it produced.
That range had no authoritative source. Not a weak one — none. It was in the knowledge base because it had come in from the functional-medicine literature at some point and had never been asked for its evidence.
The NIH Office of Dietary Supplements states that serum concentrations “>125 nmol/L (>50 ng/mL)” are “linked to potential adverse effects, particularly at >150 nmol/L (>60 ng/mL).” The bottom of the range being recommended as optimal is where mainstream guidance starts flagging risk, and the top half of it is above the level that guidance flags particularly.
It was retired and replaced — Endocrine Society ≥75 nmol/L, IOM ≥50 nmol/L, and the NIH caution stated alongside them. This paragraph is the first time that correction has been made where a reader can see it, which is a fair criticism of me and not of the software.
And note what the trials can and cannot say about it. D-Health’s treatment arm reached a mean of 115 nmol/L — just below the bottom of that retired band — and found no mortality benefit, with a cancer signal running the wrong way. Nobody has run a five-year hard-outcome trial on people held at 125–175. The honest statement is not “that range is disproven.” It is that the range was never supported, it begins where NIH begins warning, and the closest any large trial has come to it found nothing good.
I am including this because a piece that takes apart somebody else’s number and stays quiet about its own is not applying one standard. It is doing PR.
My whole approach is test, don’t guess. So the honest thing to do here is tell you that the most recent major guideline disagrees with me about testing.
The Endocrine Society’s 2024 clinical practice guideline reviewed the trial evidence and concluded that there is no randomised evidence supporting routine 25(OH)D screening in the general population, including in people with obesity or darker skin, and no clear evidence defining an optimal target concentration for disease prevention. On that basis the panel suggested against routine 25(OH)D testing in all the populations it considered, and against empiric supplementation above the dietary reference intake in healthy adults under 75.
That is a direct challenge to what I do, from a body with no supplement to sell, and I am not going to bury it in a footnote.
Here is my answer, and you can judge whether it holds. The guideline is about screening — testing populations of well people to decide who should take a cheap, safe, 400 IU supplement. On that question they are plainly right: the supplement costs less than the blood test, and if the answer is “take 10 µg a day” either way, the test bought you nothing.
The question I face is different. It is somebody with fatigue, diffuse musculoskeletal pain, recurrent infection or an autoimmune diagnosis, who is already taking something, often at a dose they chose themselves, and who wants to know whether it is doing anything. That is not screening a well population. That is measuring before and after an intervention already under way — and the guideline explicitly excludes people with established indications from its scope.
The same guideline, incidentally, does suggest empiric supplementation for children and adolescents aged 1–18, adults over 75, pregnancy, and high-risk prediabetes — and, for people over 50 for whom vitamin D is indicated, suggests daily rather than intermittent high-dose administration. That last recommendation is the bolus finding showing up again.
If you do get tested, the number comes back in nmol/L in the UK and in ng/mL in most American writing. The conversion is ng/mL × 2.496 = nmol/L. Mixing them up is how a perfectly normal result becomes a crisis, and it happens more often than you would think.
25 nmol/L (10 ng/mL) — SACN 2016. Below this, increased risk of poor musculoskeletal health in the UK population. This is a population protective threshold, not a personal optimum.
50 nmol/L (20 ng/mL) — US Institute of Medicine, 2011. Sufficient to meet the bone-health requirements of 97.5% of the population.
75 nmol/L (30 ng/mL) — the 2011 Endocrine Society guideline’s sufficiency cut-off, widely quoted in functional medicine. Note that the same society in 2024 stated it found no clear evidence defining an optimal target for disease prevention.
What this means in practice: a result of 60 nmol/L is “fine” by two of those standards and “insufficient” by the third. Anyone who tells you there is one correct number is telling you which committee they prefer, not what the evidence shows.
You will be told that vitamin D must be taken with magnesium, with vitamin K2, or both. Applying the same standard to supplements as to drugs:
One caution that matters more than the rest. If you have sarcoidosis, another granulomatous disease, primary hyperparathyroidism, or a history of kidney stones or high blood calcium, vitamin D is not a casual supplement for you and the standard advice does not apply. Speak to your GP before starting.
Two people take vitamin D this winter. One is at 22 nmol/L, has been indoors since August, and is genuinely deficient by every threshold anyone has proposed. The other is at 130 nmol/L, already supplementing, and is about to add 5,000 IU on top because of a slide.
The trials cannot tell those two apart, because the trials did not try to — VITAL and D-Health both deliberately enrolled people regardless of status. That is exactly why their null results are real and also why they cannot answer the question in front of you.
One blood test can tell those two apart. That is the whole argument for measuring rather than picking a number.
And measuring is only half of it. A number you cannot source is a guess with a decimal point. That applies to 5,000 IU on a slide, to 125–175 nmol/L in my own knowledge base, and to every threshold either of us quotes at you. Ask where it came from. If the answer is “it is generally considered to be”, you have your answer.
Three pieces on this site go deeper in three different directions. “Everyone takes vitamin D for immunity” — here’s why that’s not enough takes the mechanism apart: the conversion steps, VDR variation, and what a capsule cannot replicate about sunlight. Vitamin D and multiple sclerosis is the cleanest example anywhere of the pattern this article is about — genetic evidence supporting a causal role in whether disease develops, and randomised trials failing to change it once established. Both findings real, and the distinction routinely lost. And the kidneys puts vitamin D where it actually sits: in a four-way loop with calcium, magnesium and PTH, which is why a lone number means less than people want it to.
The first Supplement of the Month was magnesium, which is not a coincidence — it is the co-factor this month’s supplement depends on, and the more useful test of the two if you can only have one.
If you’re taking vitamin D and have never seen a 25(OH)D result, you don’t know whether you’re correcting a deficiency or topping up someone who was already fine. Those are different situations with different answers. The free call takes twenty minutes.
Book a free 20-minute call More articles
Manson JE et al. Vitamin D supplements and prevention of cancer and cardiovascular disease (VITAL). N Engl J Med 2019;380(1):33–44. doi:10.1056/NEJMoa1809944 · PMID 30415629
Okereke OI et al. Effect of long-term vitamin D3 supplementation vs placebo on risk of depression or clinically relevant depressive symptoms and on change in mood scores (VITAL-DEP). JAMA 2020;324(5):471–480. doi:10.1001/jama.2020.10224 · PMID 32749491
Neale RE et al. The D-Health Trial: a randomised controlled trial of the effect of vitamin D on mortality. Lancet Diabetes Endocrinol 2022;10(2):120–128. doi:10.1016/S2213-8587(21)00345-4 · PMID 35026158
LeBoff MS et al. Supplemental vitamin D and incident fractures in midlife and older adults. N Engl J Med 2022;387(4):299–309. doi:10.1056/NEJMoa2202106 · PMID 35939577
Hahn J et al. Vitamin D and marine omega 3 fatty acid supplementation and incident autoimmune disease: VITAL randomized controlled trial. BMJ 2022;376:e066452. doi:10.1136/bmj-2021-066452 · PMID 35082139
Jolliffe DA et al. Vitamin D supplementation to prevent acute respiratory infections: a systematic review and meta-analysis of aggregate data from randomised controlled trials. Lancet Diabetes Endocrinol 2021;9(5):276–292. doi:10.1016/S2213-8587(21)00051-6 · PMID 33798465
Demay MB et al. Vitamin D for the prevention of disease: an Endocrine Society clinical practice guideline. J Clin Endocrinol Metab 2024;109(8):1907–1947. doi:10.1210/clinem/dgae290 · PMID 38828931
Shah VP et al. A systematic review supporting the Endocrine Society clinical practice guidelines on vitamin D. J Clin Endocrinol Metab 2024;109(8):1961–1974. doi:10.1210/clinem/dgae312 · PMID 38828942
Scientific Advisory Committee on Nutrition. Vitamin D and Health. London: TSO, 2016. RNI 10 µg/day for ages 4+; population protective concentration 25 nmol/L.
NHS. Vitamins and minerals — Vitamin D. Daily requirement 10 µg; upper limit for adults 100 µg (4,000 IU) per day.
Institute of Medicine. Dietary Reference Intakes for Calcium and Vitamin D. Washington DC: National Academies Press, 2011. Sufficiency 50 nmol/L (20 ng/mL).
Holick MF et al. Evaluation, treatment, and prevention of vitamin D deficiency: an Endocrine Society clinical practice guideline. J Clin Endocrinol Metab 2011;96(7):1911–1930. doi:10.1210/jc.2011-0385 — the origin of the 75 nmol/L figure, superseded in 2024.
Educational content, not medical advice. Vitamin D is not a neutral supplement for everyone: if you have sarcoidosis or another granulomatous disease, primary hyperparathyroidism, hypercalcaemia, or a history of kidney stones, speak to your GP before taking it. New, persistent or worsening symptoms warrant medical assessment rather than a supplement.