Vitamin D and Type 2 Diabetes: The VDR Genotype That May Decide Whether You Respond

The vitamin D evidence base has spent two decades in a familiar pattern: observational studies show low 25(OH)D levels track with worse outcomes across nearly every disease, large randomized trials test supplementation in unselected adults, and the trials come back null or modestly positive. The disconnect has been one of the cleaner case studies in why "low X is associated with bad outcome" rarely translates cleanly into "supplementing X improves outcome."

A new analysis published in JAMA Network Open on April 23, 2026 offers a possible explanation for the disconnect — at least for the diabetes-prevention question — that has direct implications for anyone running a personal vitamin D experiment. The short version: the receptor matters. About 30 percent of people may not respond to vitamin D supplementation in this context, regardless of how much they take or how high they get their 25(OH)D level.

This post walks through the new finding, what it does and doesn't establish, and how to think about a personal vitamin D experiment in light of it.

The Trial in Question

The Vitamin D and Type 2 Diabetes (D2d) study was the largest randomized trial ever designed to test whether vitamin D supplementation prevents progression from prediabetes to type 2 diabetes. The protocol: 4,000 IU per day of vitamin D3 versus placebo in 2,423 adults with prediabetes, median follow-up about 2.5 years, primary endpoint new-onset type 2 diabetes. Published in the New England Journal of Medicine in 2019.

The headline result was a 12% relative reduction in diabetes progression with vitamin D versus placebo — directionally what the observational literature had predicted, but with a confidence interval that crossed 1.0. Statistically null. The result has shaped how clinicians and guideline bodies discuss vitamin D and metabolic health ever since: in adults who are not deficient, supplementation does not reliably prevent diabetes on average.

That last word is doing a lot of work, and the new paper is what teases it apart.

What the New Genotype Analysis Found

Lead author Bess Dawson-Hughes and senior author Anastassios Pittas — the same Tufts team that designed and ran D2d — went back to the trial cohort, genotyped participants at four well-studied VDR polymorphisms (ApaI rs7975232, BsmI rs1544410, FokI rs2228570, and TaqI rs731236), and re-ran the trial's primary analysis stratified by genotype.

The result that drove the headlines was about ApaI rs7975232:

• Adults carrying at least one C allele at ApaI (genotypes AC or CC, ~70% of the cohort) showed a 19% reduction in progression to type 2 diabetes with vitamin D versus placebo. This subgroup result was statistically significant.
• Adults homozygous for the A allele (genotype AA, ~30% of the cohort) showed no measurable benefit. Their progression rates on vitamin D and placebo were statistically indistinguishable.

An accompanying JAMA Network Open editorial called the polymorphism a "D-terminant" — the genetic variable that may decide whether vitamin D supplementation matters in this context. The authors of the editorial were careful, and the finding deserves the same care:

1. It is a secondary analysis of a negative trial. The original D2d primary endpoint did not reach significance. The genotype-stratified result is hypothesis-generating, not confirmatory. It says something interesting happened inside the trial when the cohort is sliced by genotype — not that vitamin D prevents diabetes in carriers as a settled matter.
2. It needs prospective replication. The most rigorous test would randomize prediabetic adults to vitamin D or placebo after identifying their VDR genotype, with diabetes progression as the primary endpoint in the AC/CC subgroup. That trial doesn't yet exist.
3. The mechanism is plausible, not proven. ApaI is in a non-coding region of the VDR gene; the functional effect is thought to be on receptor expression or mRNA stability rather than on the receptor protein structure itself. Biology consistent with the result, but mechanistic chain not fully traced.

Why This Matters for a Personal Experiment

The Prova audience is the demographic most likely to already have raw 23andMe data, an active relationship with Quest or LabCorp's direct-to-consumer testing, and a habit of running vitamin D experiments on themselves. If you are in that group, the new paper is directly actionable as a framing tool — not as a treatment protocol.

The reframing: instead of asking "is my vitamin D supplementation working for me?" — which is a vague question — you can ask a more specific one. "Given my ApaI genotype and my baseline labs, what should I expect to see, and what is worth tracking to find out?"

Your raw 23andMe data file (or the equivalent from AncestryDNA) almost certainly includes rs7975232. Several free or low-cost services (Promethease, Genetic Lifehacks, SelfDecode, Nutrigenomix) will surface it from a raw data export. None of these services are validated as diagnostic tools for the diabetes-prevention question — but they are sufficient for the educational question of "what variant do I carry?"

A Self-Experiment Worth Running

If you are weighing vitamin D supplementation for prediabetes-related reasons — or already supplementing and curious whether it's accomplishing anything — the experiment Prova is built for here is fairly clean. The structure:

Baseline (week 0). Capture three things:

• VDR ApaI genotype. Pull rs7975232 from your raw 23andMe or AncestryDNA data. The reference SNP convention can flip orientation depending on the platform, so when in doubt, check the position on Chr12 and the minor allele frequency report. Note your genotype as AA, AC, or CC.
• Vitamin D status. A 25(OH)D blood test through a clinician, Quest, LabCorp, or a direct-to-consumer service. Record the value in ng/mL.
• Glycemic markers. Fasting glucose, HbA1c, and — if available — fasting insulin (so you can compute HOMA-IR externally). Your provider can order all three from a single blood draw.

Intervention (12 to 16 weeks). If you decide, with your clinician, to supplement, vitamin D3 in the 2,000 to 5,000 IU/day range with food and a fat source is the typical protocol — and 4,000 IU is what D2d used. Take consistently. Don't change other variables (diet, training, other supplements) more than usual; the value of an N-of-1 experiment shrinks as confounders pile up. (See confounding variables in supplements for the methodology.)

Re-test (week 12 to 16). Repeat the 25(OH)D test, fasting glucose, HbA1c, and fasting insulin. Compare to baseline.

Interpret. What you're looking for, in priority order:

• Did your 25(OH)D actually move? If your baseline was 28 ng/mL and you're now at 35 ng/mL, the dose moved your levels modestly. If you're still under 30, dose was too low or absorption is poor. If 25(OH)D didn't change, none of the downstream conclusions matter — you didn't run a vitamin D experiment, you ran a placebo experiment. (See our does vitamin D actually work deep-dive on absorption, magnesium cofactor, and dose-response.)
• Did your fasting glucose and HbA1c trend? A1c reflects roughly the prior 90 days of glycemic state, so a 12-week experiment is the minimum window for a meaningful read. Direction and magnitude matter more than statistical confidence in a single N-of-1.
• Does the direction match what your genotype predicts? This is the integrative read. AC/CC genotype + meaningful 25(OH)D rise + improvement in HbA1c is a coherent positive signal. AA genotype + meaningful 25(OH)D rise + no change in HbA1c is consistent with the new paper's null subgroup. AC/CC genotype + meaningful 25(OH)D rise + no change in HbA1c is also possible — the trial result was a population average within the subgroup, not a guarantee for any individual.

What This Doesn't Tell You

The new paper addresses one question — vitamin D and progression from prediabetes to type 2 diabetes — and stops there. A few things it specifically does not say:

• It does not say vitamin D prevents type 2 diabetes in people who are not prediabetic. The cohort was specifically prediabetic adults.
• It does not say vitamin D treats existing type 2 diabetes. The endpoint was progression, not glycemic control after onset.
• It does not say AA-homozygous adults shouldn't supplement vitamin D. The trial measured one outcome (diabetes progression). Vitamin D has other roles — bone health, immune function, mood — where the genotype-effect relationship may be entirely different. (See vitamin D and testosterone, vitamin D for lifters, and the D3/K2 stack guide for those other contexts.)
• It does not say the AC/CC benefit is large in absolute terms. A 19% relative risk reduction in a population with ~5% annual progression rate translates to about a 1 percentage-point absolute reduction per year. Real, but not dramatic.
• It does not say the result will replicate. Large secondary analyses of negative trials sometimes hold up, sometimes don't. The next two to three years of vitamin D research will tell that part.

Where This Slots Into Prova's Vitamin D Cluster

Existing posts:

• Does Vitamin D Actually Work? Beyond Deficiency — what the large trials show in unselected adults. The new VDR analysis is the personalized-medicine asterisk on that picture.
• Vitamin D for Lifters — performance, recovery, and strength angles. Genotype context matters less here; the bone/muscle vitamin D effects appear to be more population-general.
• Vitamin D and Testosterone — the T-axis literature. Different mechanism, different evidence base.
• Vitamin D3 + K2 Guide — the cofactor stack.

This new piece is the prediabetes-and-VDR angle that none of the others address. If you carry an AA genotype at ApaI and you're supplementing vitamin D primarily for muscle, immune, or hormonal reasons, the new paper has nothing to say about that decision. The diabetes-prevention question is what changed.

The Honest Summary

The new JAMA Network Open analysis is the most interesting vitamin D result of the year so far. It offers a credible explanation for why a long line of large, well-powered vitamin D trials in unselected adults have been disappointing: a substantial fraction of any randomized cohort may simply be unable to translate higher 25(OH)D levels into the clinical benefit being measured. If the personalized-medicine framing replicates, vitamin D research over the next decade will look very different from the previous two.

But the result is one secondary analysis of one negative trial. The right reaction is to update your priors slightly toward "personal genotype matters" — and to keep the rest of the conversation about vitamin D where it has been: between you, your labs, and your clinician.

If you're already running an N-of-1 vitamin D experiment, this is the kind of finding that gives the experiment more structure: a hypothesis to test, a genotype to know, a set of biomarkers to track, and a window of time to read the result. That's what the data lets you do today. Anything more confident than that is overreach.