The current study found that baseline vitamin D status was an independent predictor of better β-cell function and AUCglucose after 3 years of follow-up in the PROMISE study cohort. This is the first study to examine the prospective association of serum 25(OH)D with β-cell function.
Previous studies assessing the association between 25(OH)D and β-cell function have used cross-sectional designs and have reported inconsistent findings (14
). These inconsistencies may be attributed to the use of less-detailed, fasting-based surrogate measures of β-cell function (e.g., HOMA-β or C-peptide) in the majority of previous studies (16
). Only two studies have used gold-standard methods, including the hyperglycemic clamp (14
) and the IVGTT (21
), but no significant association between 25(OH)D and β-cell function was found after adjusting for potential confounders. In contrast, the findings in the current study support our cross-sectional results in the PROMISE study cohort (19
), in which OGTT-based measures of β-cell function, specifically IGI/IR and ISSI-2, were significantly associated with 25(OH)D. Other cross-sectional studies also have reported a similar significant association of 25(OH)D with β-cell function (17
). In addition to the observational literature, a limited number of intervention studies have examined the effect of vitamin D supplementation on measures of β-cell function (17
); these studies have similarly yielded inconsistent results. However, most studies included small sample sizes, a short duration of intervention, variation in vitamin D doses, and surrogate measures of β-cell function, including C-peptide and HOMA-β, both of which on their own do not account for background IR. Three previous intervention studies (35
) used gold-standard IVGTT-based measures of insulin secretion, with one study reporting that supplementation of 1,332 IU vitamin D3
per day resulted in increased first-phase insulin secretion (35
). However, given that this study was not a randomized controlled trial, that it included only 10 study participants, and that the IVGTTs were not performed according to standard procedures, their finding should be interpreted with caution. The remaining two studies using IVGTT (37
) found no effect of supplementation with a synthetic analog of the active vitamin D metabolite, calcitriol [i.e., 1,25(OH)D], on insulin secretion in subjects with IGT. It is clear that current evidence is limited and inconsistent regarding the association of 25(OH)D with β-cell function.
In contrast to the findings regarding β-cell function, the current study did not find a significant association of baseline 25(OH)D with follow-up measures of IR. Most cross-sectional studies have found significant inverse associations between 25(OH)D and IR (12
), including our recent study in the PROMISE study cohort (19
). However, some studies have reported no association (20
). In addition, only two prospective studies have been conducted to date (6
). Forouhi et al. (23
) reported a significant inverse association of baseline serum 25(OH)D with HOMA-IR after 10 years of follow-up in white subjects from the U.K. More recently, Gagnon et al. (6
) found a significant positive association of baseline 25(OH)D with insulin sensitivity (HOMA-S) at 5 years in adults participating in the Australian Diabetes, Obesity, and Lifestyle Study. Likewise, we also report an initial significant inverse association of baseline 25(OH)D with follow-up HOMA-IR in the current study, which was attenuated to nonsignificance after adjustment for obesity. Given that our study population is more obese than the populations in these previous studies, it is possible that obesity was a stronger determinant of IR than 25(OH)D in this population.
The current study also found a significant inverse association of baseline serum 25(OH)D with AUCglucose
at follow-up, indicating that those with higher baseline 25(OH)D had significantly better glucose homeostasis during the follow-up OGTT, even after adjusting for baseline AUCglucose
. Most previous studies have reported significant inverse associations of 25(OH)D with various continuous measures of glycemia, including fasting or 2-h glucose during the OGTT (14
). Forouhi et al. (23
) also reported a significant inverse association of baseline 25(OH)D with 2-h OGTT glucose, but not fasting glucose, after 10 years of follow-up, with multivariate adjustment.
In addition to the continuous outcome measures assessed, this study also examined the association between baseline serum 25(OH)D and the risk of progression to dysglycemia at follow-up. The initial multivariate logistic regression analyses indicated a significant reduced risk of progression to dysglycemia with greater baseline 25(OH)D, but this association was attenuated to nonsignificance with additional adjustment for BMI in model 3. Although there is limited evidence, most previous studies have found an inverse association between vitamin D and diabetes risk (7
), but negative findings also have been reported (10
). As was documented in the current study, some investigators also have reported attenuation of an initial significant association after BMI adjustment (10
), but most previous studies have reported a significant association between 25(OH)D and diabetes even after accounting for body composition (6
). Vitamin D is a fat-soluble vitamin, and the consistently observed inverse association between 25(OH)D and adiposity is thought to be largely a result of the sequestering of 25(OH)D in adipose tissue, where it is no longer bioavailable (49
). PROMISE study participants are primarily overweight or obese, with 72.8% having a BMI ≥27 kg/m2
, and thus the sequestering effect of adipose tissue on vitamin D bioavailability is one potential explanation for the nonsignificant association with dysglycemia, after BMI adjustment in this cohort. However, given that the current study did find a significant prospective association of baseline 25(OH)D with β-cell function and continuously measured glycemia, increased power provided through a longer follow-up duration may be needed to detect a significant association of 25(OH)D with risk of progression to dysglycemia.
The current study has a number of potential limitations. First, only baseline 25(OH)D was collected, and an additional serum 25(OH)D measurement at follow-up to examine the effect of longitudinal changes in 25(OH)D on the outcome measures would have strengthened the study. Second, no information on diet was collected, but we did have information on participants’ vitamin D supplement use, which is an important contributor to 25(OH)D levels. In addition, gold-standard measures of IR and β-cell function were not used because these procedures are costly and invasive and therefore not feasible for large epidemiological studies. However, the current study used extensively validated proxy measures to determine IR and β-cell dysfunction based on glucose and insulin values from multiple time points in the OGTT. In addition, we did not have glucose data for 60 and 90 min during the OGTT, which would have allowed for increased accuracy in the calculation of AUCglucose
. It also is important to note possible bias in our results given that those who returned for the follow-up clinic visit were more likely to be older, female, and white than those who did not return. However, we did adjust for these variables in our multivariate analyses. Last, because this was an observational study, residual confounding is possible because unmeasured confounders may impact the association of serum 25(OH)D with the outcomes. Strengths of this study include its prospective design, which allows for the temporality of the associations to be observed. In addition, the current study examined a multiethnic cohort, whereas most previous studies have focused solely on white populations. Examining non-white populations is valuable, considering these individuals are at high risk for type 2 diabetes and are known to have low 25(OH)D concentrations. The current study also included the direct measurement of serum 25(OH)D, versus reliance on diet and sun-exposure data. In addition, multivariate analyses were adjusted for numerous potential confounders, including vitamin D supplement use, which has been excluded in most previous studies (14
In conclusion, the current study found that higher baseline 25(OH)D independently predicted better β-cell function and lower AUCglucose after 3 years of follow-up, even after adjustment for baseline β-cell function and AUCglucose, respectively. Higher 25(OH)D levels also were associated with a reduced risk of progressing to dysglycemia, although this association was not statistically significant after adjustment for obesity (adjusted odds ratio 0.78 [95% CI 0.59–1.02]). Longer follow-up of this cohort may reveal a significant inverse association of 25(OH)D with risk of type 2 diabetes. These results support a potential role for vitamin D in the etiology of type 2 diabetes.