Vitamin D and Diabetes: The 15% Solution

A cheap pill, a scary diagnosis, and a number that travels well

If you have prediabetes — and roughly one in three American adults does — you have almost certainly been told to lose weight, move more, and cut the refined carbs. What you may also have been told, somewhere in a podcast or a wellness reel, is that there's a five-dollar bottle on the pharmacy shelf that cuts your odds of crossing into full diabetes. That bottle is vitamin D.

The pitch is seductive for a reason. Vitamin D is cheap, familiar, and carries a halo of "natural." Diabetes is frightening and expensive. And there is a real, peer-reviewed number floating around to anchor the claim: a 15% reduction in progression from prediabetes to type 2 diabetes.² That number is not made up. It comes from a serious 2023 analysis in Annals of Internal Medicine, pooling three rigorous randomized trials.

So why hasn't your endocrinologist handed you a prescription? Because the story underneath that 15% is far more interesting — and far more conditional — than the headline lets on. Every single one of those three trials, taken on its own, failed to show a statistically significant benefit. The win only appears when you stack them together. And the benefit, real as it is, is small in absolute terms, invisible in the general population, and strongest only at blood levels most supplement users never reach. Let's look at what the data actually says.

Why anyone thought this would work

The mechanistic case for vitamin D in diabetes is genuinely plausible, which is part of why this hypothesis has had such staying power. The hormone — and vitamin D is really a pro-hormone, not a vitamin in the classical sense — acts through the vitamin D receptor, which shows up in tissues you wouldn't expect, including the insulin-producing beta cells of the pancreas.¹

From there, the biology branches in two directions. Inside the beta cell, vitamin D appears to support insulin synthesis and secretion, partly by helping regulate intracellular calcium, the same calcium flux that triggers insulin release. Out in muscle and fat tissue, vitamin D has been linked to improved insulin sensitivity and to dampening the low-grade inflammation that blunts the body's response to insulin. Put those together and you have a tidy theory: top up a deficient system, and you might help the pancreas keep pace with demand while making the rest of the body listen to insulin a little better.

Tidy theories are exactly the kind The Corneum exists to pressure-test. Elegant mechanisms have launched a thousand supplements that did nothing in a human body. The vitamin D receptor being present in beta cells tells you the door exists. It does not tell you that walking through it changes whether a real person develops diabetes. For that, you need trials — and the trials are where this story gets honest.

Three big trials. Three misses.

Here is the fact that almost never survives the trip to social media: the major randomized trials of vitamin D for diabetes prevention all missed their primary endpoint. Not one of them, on its own, crossed the line of statistical significance. If you only ever read the individual papers, you would walk away thinking vitamin D does nothing.

The flagship was D2d, published in the New England Journal of Medicine in 2019. It is the trial everyone cites, and the trial almost nobody quotes accurately.

Then came the other two. The Norwegian Tromsø study gave 511 people with prediabetes 20,000 IU per week for five years and titled its paper, with admirable bluntness, "Vitamin D 20,000 IU per Week for Five Years Does Not Prevent Progression From Prediabetes to Diabetes."⁴ The hazard ratio was 0.90, not significant. And Japan's DPVD trial tested eldecalcitol, a prescription active vitamin D analog, in 1,256 adults with impaired glucose tolerance.³

And to close the door on the "everyone should take it" version of the claim, there is VITAL — the largest of them all, 25,871 general-population adults on 2,000 IU/day. Its diabetes analysis was flatly null: hazard ratio 0.91, not significant, with no benefit no matter how you sliced the subgroups.⁵ In people who are already vitamin D replete and not specifically at high risk, the answer is a clean zero.

The pool that changed the story

Here is where it gets genuinely interesting. The three prediabetes trials — D2d, Tromsø, and DPVD — were designed from the start so their data could be combined at the level of individual participants, not just summary numbers. That is the gold standard for this kind of synthesis. In 2023, that pooled analysis landed in Annals of Internal Medicine.²

So the signal is real. Three independent trials, one rigorous pooling method, a consistent direction, and a biologically coherent companion finding — more people drifting back to normal glucose — that you would expect if the drug were doing something. This is why the rating here is Promising and not "Insufficient Data." A 15% relative reduction that replicates across cohorts is not noise.

But "real" and "meaningful for you" are different questions, and the gap between them is the whole game.

Relative percentages are flattering. Absolute ones are honest.

A 15% reduction sounds like a lot. It is doing a great deal of rhetorical work in every post that cites it, because relative reductions always sound bigger than they feel. The number that actually describes your odds is the absolute risk reduction, and over three years it was about 3.3 percentage points — diabetes in 22.7% of the vitamin D group versus 25.0% of placebo.²

Translate that into people. You would need to treat roughly 30 adults with prediabetes for three years to prevent one case of diabetes. (That number-needed-to-treat is my arithmetic off the published 3.3% figure, not a value the paper states outright — but it follows directly.) For a five-dollar supplement with a clean safety record in monitored trials, an NNT of 30 is not nothing. For a "diabetes-preventing miracle," it's a reality check.

And there's a final twist that the supplement aisle loves and that deserves a heavy asterisk. Within the pooled cholecalciferol data, the people who benefited most were those who pushed their blood levels of 25-hydroxyvitamin D very high — to around 125 nmol/L, or roughly 50 ng/mL. In that group the risk reduction reached 76%.² That sounds like a green light to megadose. It isn't. This was a non-randomized, on-treatment comparison: people who reliably hit high levels also tend to be the most adherent, best-absorbing, healthiest participants. Adherence is a notorious confounder. The 76% is a hypothesis, not a dosing instruction.

Cheap and well-tolerated — until it isn't

Give credit where due: in the trials, with proper monitoring, 4,000 IU/day of vitamin D3 was safe. The pooled analysis found no statistically significant excess of kidney stones, hypercalcemia, or hypercalciuria.² The dedicated D2d safety paper reached the same conclusion.¹¹ This is not a dangerous molecule at studied doses in monitored people.

The phrase carrying the weight there is "in monitored people." Vitamin D toxicity is real, follows a U-shaped curve, and shows up at the high end as hypercalcemia — too much calcium in the blood — which can mean kidney stones, nausea, and in extreme cases kidney damage. The broader literature pegs the hypercalcemia risk from supplementation at roughly double baseline.¹² And remember that D2d deliberately excluded people with a history of stones or high calcium. The reassuring safety numbers come from a screened population taking a defined dose under observation — not from someone chasing a 50 ng/mL blood level on their own with megadose capsules.

When the experts disagree, that's the story

If the evidence were truly decisive in either direction, the major guideline bodies would agree. They don't — and that disagreement is itself the most honest summary of where this stands.

The American Diabetes Association, in its 2024 Standards of Care, declines to recommend vitamin D for diabetes prevention in the general population. Its language is careful: D2d "showed no significant benefit," while "post hoc analyses and meta-analyses suggest a potential benefit in specific populations," and "further research is needed."⁹ Translation: not yet.

The Endocrine Society, in its 2024 vitamin D guideline, goes a step further. For adults with high-risk prediabetes, on top of lifestyle change, it suggests empiric vitamin D to reduce progression — while explicitly rating this a conditional recommendation built on low-to-moderate-certainty evidence, and noting that healthy adults under 75 are unlikely to benefit from doses above the recommended dietary allowance.¹⁰ Translation: reasonable, in the right person, with eyes open.

Both are reading the same trials. One emphasizes that every primary endpoint was missed; the other emphasizes that the pooled signal is real and the intervention is cheap and safe. Neither is wrong. That is what "Promising" looks like in practice — not a verdict everyone has reached, but a signal serious enough that thoughtful people land on different sides of the same fence.

Dr. Cole's read