Vitamin D3 (Cholecalciferol)

A guideline grades other people's evidence rather than producing any, and most of its positions are worded as suggestions, which is how low certainty gets expressed. It sets aside anyone with an established medical reason to be tested or treated, so it speaks only about people without one. Two corrections to it were published in 2025 and its paediatric advice has drawn published rebuttals.

Each of these trials recruited a different population and measured a different outcome, so what they share is a pattern rather than one replicated result. Three of the four used large intermittent doses, a schedule some analyses argue is the least effective. In one of them, the group whose levels reached the recommended target fell more often than the group whose levels did not.

The outcome in both studies is a blood level; neither measured whether anyone felt or functioned better. The 87 percent figure comes from 33 people on a single-blind protocol at a weekly dose, which makes it a statement about that regimen rather than about a daily D3 tablet against a daily D2 one. The pooled comparison searched the literature only to July 2011.

Participants were picked for being unusually healthy for their age, with no major health event in five years and good mobility, so a frailer group might respond differently. Testing six outcomes at once, the authors demanded a p value below 0.01, a strict bar that would hide a modest real effect. Three years is short for both fractures and cognition.

Nobody was recruited for having low vitamin D, low bone mass or osteoporosis, which is exactly the group the supplement is normally prescribed to. Fractures were reported by participants once a year and then checked against medical records. Hip fractures were few enough that their estimate spans a 30 percent reduction to a 47 percent increase.

Both trials gave very large intermittent doses, so nothing here suggests a daily tablet carries the same risk. The falls trial enrolled 200 people and had no placebo group, comparing one dose against another rather than against nothing. The annual-dose trial's fracture interval starts at exactly no effect, and its explanation for the timing, a spike in blood levels just after each dose, is a hypothesis it raises rather than one it tests.

Neither trial screened people in for low vitamin D, so both test adding more to people who mostly had enough rather than correcting a shortfall. The Australian participants averaged 77 nanomoles per litre before they started, in a country with a great deal of sun. Five years may be short for cancers that take longer than that to surface.

The over-75 recommendation rests on a very small reduction in deaths, graded as high certainty but never translated into a number of lives. Most of the guideline's positions are worded as suggestions rather than recommendations, the wording used when certainty is low. The trials behind it used widely varying doses and often let participants keep taking their own vitamin D, so the right dose is left unresolved.

Only 155 people took part, so a small real effect could hide inside that. Entry required a level in the insufficient band rather than severe deficiency, which leaves open whether people much further down would respond. What makes the result carry more than its size suggests is that not one of nine outcomes moved in either direction.

These are standing pharmacological facts from a clinical reference chapter rather than findings from a trial. The chapter does not say how long supplementation takes to raise a level, or the dose at which toxicity begins. Nothing in it settles what level to aim for, which is a separate and open question.

The authors state plainly that the study lacked the statistical power to answer its own question. Depression was diagnosed against structured criteria at in-person assessments, which is a real strength, but too few cases arose for that rigour to help. A result this imprecise is an absence of evidence, not evidence of absence.

This is a snapshot of two existing groups at one research unit, so it shows which mathematical curve fits best rather than testing the mechanism directly. Ruling out fat trapping the vitamin rests on that better fit, which is suggestive rather than conclusive. The advice to scale a dose by body weight follows from the model and was not itself tested here.

That upper limit was set on the risk of high blood calcium with an allowance for uncertainty, not on any observed pattern of harm at 4,000 international units. Reaching a toxic blood level takes very large doses sustained over a long period. Safety at 2,000 international units daily says nothing about the very large intermittent doses that were followed by more falls in other trials.

A subgroup result is a comparison the trials were never designed to make, which is why it is the kind of finding that most often fails to repeat. Those 255 deficient people produced an odds ratio of 0.60, an interval running from a 72 percent reduction to a 30 percent increase. Only 2.4 percent of that trial's participants were deficient at all.

Whether vitamin D protects severely deficient people from respiratory infection was never properly tested. A 2017 estimate of 70 percent lower odds did not reappear in larger analyses, and trials enrol too few deficient people to tell.

The 2025 funnel plot leans left, the pattern seen when small trials that found nothing were never published. All three versions pool trials running from infancy to age 95 across many countries, so the pooled group is not one population. The 2017 version worked from individual participant records and the later two from summary numbers, which is a weaker footing for any subgroup.

Protection against respiratory infection shrank each time the same analysis was rerun with more trials. Whether the effect was never there, or the newer trials enrolled people with less to gain, these analyses cannot separate.

The p value lands exactly on 0.05 and the interval nearly touches no effect, which makes this a fragile positive rather than a settled one. The count pools rheumatoid arthritis, polymyalgia rheumatica, autoimmune thyroid disease and psoriasis together and tests none of them individually. Participants averaged 67 years old.

Almost nobody in the trial was depressed at the start, so this tests whether vitamin D3 stops depression arriving, not whether it treats depression already present. Participants were not selected for low vitamin D either. Several of the authors declare payments from supplement companies, which the paper states.

Large trials found no mood benefit; the pooled benefit rests on small, short trials, some counted twice. What observational studies still cannot separate is whether low vitamin D brings on depression or depression keeps people indoors.

The four-week trial scored fatigue on a self-reported questionnaire, a measure that moves easily on expectation, in people averaging 29 years old. The multiple sclerosis pooling covered five trials and 345 people, and its benefit runs from small down to almost nothing. The palliative trial lost 94 of its 244 participants, mostly to death from cancer.

This is a snapshot of blood levels taken in 2005 and 2006, so it shows who sits below a line and never what being below it does to them. Race here stands in for skin pigmentation, sun exposure, diet and income all at once, rather than for anything measured biologically. Black Americans measure lowest without correspondingly worse bone density or fracture rates, which is the part the number does not explain.

The 20 nanogram figure was set as a population planning number covering the needs of nearly everyone, not as a line for diagnosing one person, and the committee that set it has said so in print. The Institute of Medicine derived it from bone health alone, having found the evidence on cancer, heart disease, diabetes and infection too thin to set requirements on. Which number a person is measured against decides whether they are told they have a problem.

Where vitamin D deficiency begins is unsettled: expert lines sit at 30, 20 and 12 nanograms per millilitre. Evidence cannot choose between them, because no trial has tested whether measuring a level and acting on it does anyone good.