https://jamanetwork.com/journals/jamanetworkopen/fullarticle/2848109
A really interesting paper.
chatGPT:
Summary
This paper is a genetic re-analysis of the D2d randomized trial, asking whether variation in the vitamin D receptor gene helps explain why vitamin D supplementation appears to reduce diabetes risk in some adults with prediabetes but not others. The original D2d trial tested 4000 IU/day vitamin D3 vs placebo for a median of 2.5 years, but the overall intention-to-treat result was not statistically significant: HR 0.88, 95% CI 0.75β1.04. Prior D2d analyses suggested that people who achieved higher serum 25-hydroxyvitamin D [25(OH)D] levels, especially β₯40 ng/mL, had lower diabetes risk.
The authors genotyped three common VDR polymorphisms: ApaI rs7975232, BsmI rs1544410, and FokI rs2228570. They used a two-stage approach. First, a βdiscoveryβ analysis in 1903 participants examined whether genotype modified the association between achieved intratrial 25(OH)D and diabetes risk. Second, a βtestβ analysis in 2098 participants examined whether genotype modified the effect of randomized vitamin D3 supplementation itself.
The central finding is that the ApaI genotype split the population into apparent responders and nonresponders. Participants with ApaI AA did not benefit from vitamin D3: HR 1.02, 95% CI 0.72β1.44. Participants with ApaI AC or CC had a statistically borderline but significant reduction in diabetes risk: HR 0.81, 95% CI 0.66β0.99, equivalent to a 19% relative risk reduction.
The dose-response pattern with achieved 25(OH)D was striking. In the discovery analysis, those with ApaI CC had HR 0.29 at 40β49 ng/mL and HR 0.17 at β₯50 ng/mL compared with the 20β29 ng/mL reference group. Those with ApaI AC had HR 0.51 at 40β49 ng/mL and HR 0.26 at β₯50 ng/mL. By contrast, ApaI AA showed no meaningful reduction: HR 0.94 at 40β49 ng/mL and HR 0.86 at β₯50 ng/mL.
BsmI showed a similar pattern, but it overlapped almost completely with ApaI: 260 of 261 participants with the nonresponsive BsmI TT genotype also had ApaI AA. The authors therefore argue that ApaI genotyping alone may be sufficient to identify likely responders and nonresponders. FokI gave a less consistent pattern and was not taken forward.
What is novel?
The main novelty is not that vitamin D might reduce diabetes risk in prediabetes; that had already been suggested by D2d secondary analyses and meta-analysis. The novelty is the proposed pharmacogenetic stratification: vitamin D3 may reduce diabetes risk mainly in people with particular VDR genotypes, especially ApaI AC/CC, while giving little or no benefit to ApaI AA carriers.
This is clinically interesting because it turns a weak or borderline overall trial result into a more targeted claim: vitamin D may not be a general diabetes-prevention agent for everyone with prediabetes, but may be useful in a genetically defined subgroup. The authors suggest that a single inexpensive genotype marker could potentially identify the subgroup most likely to benefit.
A second novelty is that this work examines response to high-dose supplementation and achieved higher 25(OH)D levels, rather than merely baseline vitamin D status. The paper contrasts this with UK Biobank evidence, where lower baseline 25(OH)D and VDR genotype were studied observationally, but there were fewer people with high 25(OH)D levels comparable to those achieved in D2d.
A third novelty is that the paper suggests that ApaI, not only BsmI, may matter in this context. The prior UK Biobank signal was stronger for BsmI TT, whereas this trial-based analysis emphasizes ApaI AA as the apparent nonresponder genotype.
Critique
The result is interesting but should be treated as hypothesis-generating rather than practice-changing.
The biggest limitation is that this is a post hoc genetic association analysis of an existing trial. The D2d trial was randomized for vitamin D vs placebo, but it was not originally designed or powered primarily to test ApaI-defined treatment response. Once a trial is subdivided by genotype and by achieved vitamin D level, the risk of chance findings rises.
The statistical signal is also modest in the test phase. The responder group had HR 0.81 with a 95% CI of 0.66β0.99. That just crosses conventional statistical significance. The nonresponder group had HR 1.02, but with a wide CI of 0.72β1.44, meaning the study cannot precisely exclude moderate benefit or harm in that subgroup.
The discovery analysis uses achieved intratrial 25(OH)D, which is biologically sensible but less clean than pure randomization. Achieved 25(OH)D is influenced by adherence, body size, baseline status, supplement use, absorption, metabolism, sun exposure, and other behaviours. The authors adjusted for several covariates, including site, race/ethnicity, sex, age, BMI, physical activity, statin use, and weight change, but residual confounding is still possible.
There is also no direct mechanism shown. ApaI is not obviously demonstrated here to alter VDR function in pancreatic Ξ² cells or insulin-sensitive tissues. It may be a marker in linkage disequilibrium with another functional variant, or it may be tagging ancestry or other genomic structure. The paper acknowledges that it does not resolve the mechanism.
The racial and ethnic subgroup question is underpowered. This matters because vitamin D physiology, baseline 25(OH)D, VDR allele frequencies, diabetes risk, skin pigmentation, supplementation response, and social/environmental confounding may all vary by ancestry and context. The paper says it was too small to examine ApaI treatment response within individual racial or ethnic groups.
Another issue is clinical translation. The authors frame 4000 IU/day as well tolerated and potentially useful, but this is still a relatively high daily dose. The paper excluded people with hypercalcemia, nephrolithiasis, hyperparathyroidism, bariatric surgery, and related risks, so extrapolating to unscreened populations would be unwise. For an individual, serum calcium, kidney stone history, kidney function, baseline 25(OH)D, and other supplementation would matter.
Bottom line
This paper provides a plausible and clinically interesting explanation for why vitamin D supplementation has had mixed results in diabetes prevention: the benefit may be concentrated in adults with prediabetes who carry ApaI AC or CC VDR genotypes, while ApaI AA carriers may not benefit.
However, the finding needs independent confirmation in a trial designed prospectively around genotype-stratified vitamin D supplementation. Until then, it is best seen as a strong personalised-medicine hypothesis, not yet as definitive evidence that people with prediabetes should be genotyped and treated accordingly.