Insulin secretion and sensitivity in healthy adults with low vitamin D are not affected by high-dose ergocalciferol administration: a randomized controlled trial^,

Abstract

Background: Epidemiologic data suggest that low serum 25-hydroxyvitamin D [25(OH)D] increases insulin resistance and the risk of type 2 diabetes. Few interventional trials have assessed the effect of vitamin D on insulin metabolism, and published results are discordant.

Objective: The goal of this study was to perform a detailed assessment of the effect of ergocalciferol administration on glucose and insulin metabolism in healthy people with low total 25(OH)D_total.

Design: This was a 12-wk, double-blinded, randomized controlled trial. We enrolled 90 healthy volunteers aged 18–45 y with serum 25(OH)D ≤20 ng/mL (by immunoassay) and administered 50,000 IU ergocalciferol/wk or placebo for 12 wk. Primary endpoints were change in first-phase insulin response and insulin sensitivity as measured by intravenous glucose tolerance test. Secondary endpoints included change in homeostasis model assessment of insulin resistance; fasting glucose, insulin, and lipids; body mass index (BMI); and blood pressure.

Results: On-study 25(OH)D_total was assessed by liquid chromatography–tandem mass spectrometry. In the treated group, 25(OH)D_total rose from 18 ± 7 to 43 ± 12 ng/mL (P < 0.001) with no change in the placebo group. Despite this increase, at 12 wk, there were no between-group differences in either insulin response or insulin sensitivity; nor were there differences in any measured secondary endpoints. There was no evidence of effect modification by sex, race, glucose tolerance status, baseline 25(OH)D_total, or BMI.

Conclusion: In healthy persons with low 25(OH)D_total, ergocalciferol administration for 12 wk normalizes 25(OH)D_total but does not improve insulin secretion, insulin sensitivity, or other markers of metabolic health. This trial was registered at clinicaltrials.gov as NCT00491322.

INTRODUCTION

The pathogenesis of type 2 diabetes is multifactorial, with genetics and environmental factors playing key roles. The prevalence of type 2 diabetes is increasing rapidly and is associated with substantial morbidity and mortality as well as an increase in health care costs (1). Public health interventions to decrease the risk of type 2 diabetes are urgently needed.

Type 2 diabetes arises in the setting of increased insulin resistance coupled with an inability of the pancreatic β cell to compensate by increasing insulin secretion (2). Vitamin D has been proposed to modify both insulin resistance and secretion. Substantial expression of the vitamin D receptor has been found in β cells, suggesting that vitamin D may play an important role in the regulation of β-cell function (3). In animal models of vitamin D deficiency, stimulated insulin secretion is impaired independent of circulating serum calcium and can be restored by repleting vitamin D (4, 5). Insulin resistance is likely also affected by vitamin D. Notably, polymorphisms in both the gene encoding the vitamin D receptor and the gene encoding the vitamin D binding protein are associated with insulin resistance (6, 7). In addition, vitamin D may play a role in modulating the inflammatory response to obesity, which contributes to insulin resistance (8).

Several longitudinal observational human studies have demonstrated an association between low dietary vitamin D intake or low serum 25-hydroxyvitamin D [25(OH)D]⁷ and incident type 2 diabetes (9–14). Recent meta-analyses have reported that the risk of developing type 2 diabetes is 40% lower in subjects with higher serum 25(OH)D than in those with lower concentrations (15, 16). Randomized controlled trials of vitamin D administration and glycemic outcomes have, however, had conflicting results (17–23). Most published trials have evaluated glycemic outcomes by using either fasting measures of glucose and insulin or indices derived from oral-glucose-tolerance tests (OGTTs) (15, 24). Intravenous-glucose-tolerance tests (IVGTTs), however, may assess altered β-cell function more accurately and reproducibly than data derived from OGTTs (25, 26). Both the Institute of Medicine and the Endocrine Society have called on the scientific community to rigorously study the potential beneficial nonskeletal effects of vitamin D (27, 28). In this randomized controlled trial, we administered high-dose ergocalciferol (vitamin D₂) or placebo to healthy adults with low total serum 25(OH)D [25(OH)D_total] (≤20 ng/mL) and assessed the effects on fasting glucose homeostasis and IVGTT-derived measures of insulin sensitivity and secretion. Ergocalciferol was chosen because the high-dose form was readily available by prescription in the United States at the time the study was performed and thus has potential to be a public health intervention.

METHODS

Subjects

Healthy subjects aged 18–45 y were recruited via advertisements and mass mailings. Results of the effect of ergocalciferol on hormones of mineral metabolism from this cohort have been reported previously (29). Screening 25(OH)D_total was measured by chemiluminescent immunoassay (CLIA) (Diasorin), and subjects were eligible if 25(OH)D_total was ≤20 ng/mL. The CLIA assay is 100% cross-reactive to 25-hydroxyvitamin D₂ [25(OH)D₂] and 25-hydroxyvitamin D₃ [25(OH)D₃] as per the manufacturer. Subjects with a history of clinically significant cardiac, hepatic, gastrointestinal, oncologic, or thyroid disease or any disorder known to affect vitamin D metabolism were excluded. In addition, male subjects were required to have a serum testosterone in the reference range, and females were required to have regular menses. Use of oral contraceptives was allowed. Subjects who met these criteria then underwent a second screening visit at which a 2-h OGTT with 75 g dextrose was administered; subjects were classified as having either normal glucose metabolism (fasting glucose <100 mg/dL and 2-h glucose <140 mg/dL) or impaired glucose metabolism (fasting glucose 100–125 mg/dL and/or 2-h glucose 140–199 mg/dL). No subject had OGTT results consistent with diabetes (fasting glucose ≥126 mg/dL and/or 2-h glucose ≥200 mg/dL). Subjects self-identified race and ethnicity. This study was approved by the Partners Human Research Committee, and informed consent was obtained from all subjects. This trial was registered as NCT00491322 at clinicaltrials.gov.

Randomization

Subjects were randomly allocated in a 1:1 ratio to ergocalciferol treatment or to matching placebo by computer-generated assignment, in randomly varying blocks of 2, 4, or 6. The study statistician communicated the allocation sequence to a research pharmacist who dispensed study drug containers with identically appearing capsules. Before randomization, subjects were stratified by sex, by screening 25(OH)D_total [25(OH)D_total ≤10 ng/mL compared with 25(OH)D_total >10 ng/mL], and by glucose metabolism status (normal compared with impaired). Both subjects and investigators were blinded to treatment allocation.

Study design

In this 12-wk, double-blinded, randomized controlled trial, subjects were given either ergocalciferol 50,000 IU/wk or matching placebo to be taken orally for 12 wk. Study drug was dispensed by a research pharmacist in sequentially numbered containers and was given to research subjects by study staff. Both subjects and study staff were blinded to the assignment. Subjects in the placebo arm were given 50,000 IU/d for 7 d on completion of the study protocol. Calcium intake was maintained at 1000–1500 mg/d in both groups with use of calcium supplements, as necessary, based on dietary intake questionnaires. This study was conducted at a clinical research center of a tertiary care hospital. Subjects were recruited from June 2006 through November 2007, and study visits occurred from June 2006 through February 2008. As noted in Table 1, subjects were recruited year-round (baseline visits in spring, n = 15; summer, n = 20; fall, n = 33; and winter, n = 22).

TABLE 1.

Baseline demographic and clinical characteristics¹

	Placebo (n = 50)	Ergocalciferol (n = 40)	P value
Age, y	29 ± 92	28 ± 7	0.46
Female sex, n (%)	31 (62)	24 (60)	0.85
Race, n (%)
Asian	6 (12)	11 (28)	0.15
Black	18 (36)	8 (20)
White	20 (40)	18 (45)
Other	6 (12)	3 (7)
Ethnicity, n (%)
Hispanic	6 (12)	2 (5)	0.29
Non-Hispanic	44 (88)	38 (95)
Family history of type 2 diabetes, n (%)	27 (54)	24 (60)	0.57
Baseline visit season, n (%)
Winter (January–March)	11 (22)	11 (28)	0.91
Spring (April–June)	9 (18)	6 (15)
Summer (July–September)	12 (24)	8 (20)
Fall (October–December)	18 (36)	15 (37)
BMI, kg/m2	26.1 ± 6.7	25.2 ± 4.5	0.46
BMI ≥25, n (%)	25 (50)	16 (40)	0.34
Waist-to-hip ratio	0.86 ± 0.08	0.84 ± 0.09	0.41
Systolic blood pressure, mm Hg	113 ± 13	114 ± 13	0.71
Diastolic blood pressure, mm Hg	71 ± 9	72 ± 10	0.77
Screening 25(OH)D
CLIA, ng/mL	15 ± 4	14 ± 3	0.54
≤10 ng/mL, n (%)	8 (16)	6 (15)	0.90
Week 0 (LC-MS/MS), ng/mL
25(OH)Dtotal	18 ± 7	18 ± 7	0.83
25(OH)D2	1 ± 2	1 ± 1	0.31
25(OH)D3	17 ± 7	17 ± 7	0.98
Fasting glucose, mg/dL	85 ± 8	87 ± 7	0.11
Fasting insulin, μU/mL	10.0 ± 6.6	9.3 ± 4.5	0.75
HOMA-IR	2.2 ± 1.7	2.0 ± 1.1	0.65
Impaired glucose metabolism, n (%)	8 (16)	4 (10)	0.54
Total cholesterol, mg/dL	161 ± 39	156 ± 32	0.59
HDL cholesterol, mg/dL	58 ± 15	56 ± 15	0.50
LDL cholesterol, mg/dL	86 ± 36	84 ± 29	0.77
Triglycerides, mg/dL	83 ± 47	84 ± 51	0.95

¹P values were calculated by t test for continuous variables and by χ² or Fisher exact test as appropriate for categorical variables. CLIA, chemiluminescence immunoassay; LC-MS/MS, liquid chromatography–tandem mass spectrometry; 25(OH)D_total, total 25-hydroxyvitamin D; 25(OH)D₂, 25-hydroxyvitamin D₂; 25(OH)D₃, 25-hydroxyvitamin D₃.

²Mean ± SD (all such values).

After the screening visits, subjects were seen at 0, 4, 8, and 12 wk. At each visit, fasting serum samples were obtained between 0700 and 0900. Blood pressure, height, weight, and waist measurements were obtained. At the 0- and 12-wk visits, subjects underwent a modified IVGTT as previously described (30). Briefly, an intravenous dextrose bolus of 0.3 g/kg was administered over 30 s. Blood samples for insulin and glucose measurements were obtained at 10 and 1 min before infusion and then at 2, 4, 6, 8, 10, 12, 14, 19, 25, 30, and 40 min. First-phase insulin response, a measure of insulin secretion, was determined by using values from 0 to 10 min after intravenous dextrose administration, and insulin sensitivity was determined by using values from 10 to 40 min after administration. These calculations correlate well with insulin sensitivity as determined by euglycemic-hyperinsulinemic clamp and by the minimal model in standard IVGTT (31). HOMA-IR was calculated at each visit as previously described (32).

Laboratory testing

Screening serum 25(OH)D_total was measured by CLIA (Diasorin), which was the available 25(OH)D assay at our institution at that time. On-study 25(OH)D_total [including 25(OH)D₂ and 25(OH)D₃] was measured by liquid chromatography–tandem mass spectrometry with a lower limit of detection of 6 ng/mL and an interassay CV of 6–9%. We have previously reported on the very good agreement between these assays in our study population (29). Serum insulin was measured by radioimmunoassay with an interassay CV of 3–6% (Linco). Parathyroid hormone was measured by using a 2-site immunoradiometric assay (Nichols Institute Diagnostics) with a sensitivity of 1 ng/L and intra- and interassay CVs of 2–3% and 6%, respectively. Serum samples for insulin and 25(OH)D via liquid chromatography–tandem mass spectrometry were frozen at −80°C, and all samples from each subject were measured in a single batch. Glucose concentrations were measured on plasma samples in real time via enzymatic assay (Abbot). All batched laboratory analyses were performed in 2008 and 2009.

Statistical analyses

On the basis of published means and variances of insulin resistance as measured by IVGTT (33), we designed the study to have 80% power to detect 30% improvement in insulin sensitivity with vitamin D repletion with an α level of 0.05 (thus 40 subjects were needed per group); given an anticipated 20% loss to follow-up rate, we planned to enroll 100 subjects.

This was a per-protocol analysis; data from the 7 subjects who did not complete the study were not included. Pearson correlations were used to examine univariate associations of baseline 25(OH)D_total with other variables. The primary endpoints were 12-wk change in first-phase insulin response and change in insulin sensitivity as measured by modified IVGTT. Change was analyzed by ANCOVA after control for baseline values of each outcome respectively. The effect of ergocalciferol compared with placebo on HOMA-IR was assessed by comparison of the AUC as defined by values at each of the 4 time points. In prespecified analyses, we examined the interactions of race and glucose homeostasis (normal compared with impaired) with first-phase insulin secretion and insulin sensitivity by ANCOVA. Post hoc analyses were also performed to examine the effect of sex and baseline BMI (in kg/m²) on these parameters. The effect of treatment on fasting insulin and fasting glucose was examined by using a mixed-model ANOVA with random slopes. Unpaired t tests were used to examine the effect of ergocalciferol on the 12-wk change in BMI, waist-to-hip ratio, serum lipids (total cholesterol, HDL cholesterol, LDL cholesterol, and triglycerides), and blood pressure. Analyses were conducted by using SAS version 9.2 (SAS Institute).

RESULTS

Enrollment and protocol adherence

The study flow diagram is shown in Figure 1. Of the 97 subjects who were randomly allocated, 90 completed all 4 study visits. Compliance with the intervention was monitored by medication diaries and by pill counts of returned bottles. Of subjects randomly allocated to ergocalciferol who completed the study, 38 (95%) reported taking every pill, and 2 (5%) reported missing one dose each. Of subjects randomly allocated to placebo who completed the study, 47 (94%) reported taking every pill, 2 (4%) missed 1 dose, and 1 (2%) missed 2 doses.

FIGURE 1.

Flow diagram of study subjects. OGTT, oral-glucose-tolerance test; 25(OH)D_total, total 25-hydroxyvitamin D.

Subject characteristics

Baseline characteristics are described in Table 1. Subjects were well matched for demographic, clinical, and laboratory parameters. As previously reported, baseline 25(OH)D_total was negatively correlated with parathyroid hormone (r = −0.29, P = 0.006) (29). Baseline 25(OH)D_total was not significantly associated with fasting glucose, fasting insulin, HOMA-IR, lipids, BMI, or blood pressure with univariate testing (data not shown). Twelve subjects (13%) had impaired glucose metabolism: 2 had elevated fasting glucose, 9 had impaired glucose tolerance, and 1 had both elevated fasting glucose and impaired glucose tolerance.

Ergocalciferol administration and measures of glucose homeostasis

As we have previously reported, administration of high-dose ergocalciferol increased the mean 25(OH)D_total concentration significantly from 18 ± 7 to 43 ± 12 ng/mL (these and all subsequent data are presented as means ± SDs), whereas 25(OH)D_total did not change significantly in the placebo group (18 ± 7 to 20 ± 10 ng/mL; P < 0.001 for comparison between groups) (Figure 2A) (29). In the ergocalciferol group, 25(OH)D₂ increased from 1 ± 1 to 35 ± 12 ng/mL, 25(OH)D₃ decreased from 17 ± 7 to 8 ± 4 ng/mL (P < 0.001 for both), and neither 25(OH)D₂ nor 25(OH)D₃ changed significantly in the placebo group (1 ± 2 to 1 ± 1 ng/mL, P = 0.16, and 17 ± 7 to 19 ± 9 ng/mL, P = 0.13, respectively). There was no significant difference in the increment in 25(OH)D_total among subjects in the treatment arm by season (P = 0.34 by ANOVA). The change in HOMA-IR did not differ between the treatment and placebo groups as assessed by AUC comparison (P = 0.60) (Figure 2B). HOMA-IR did not change significantly within the treated group (within-group change of −0.2; 95% CI: −0.5, 0.2; P = 0.62) or the placebo group (within-group change of −0.1; 95% CI: −0.5, 0.2; P = 0.78). Additional analyses revealed no effect of ergocalciferol on fasting insulin or fasting glucose (P = 0.85 and P = 0.69 for between-group differences, respectively).

FIGURE 2.

Mean ± SEM changes in 25(OH)D_total (A) and HOMA-IR (B) with ergocalciferol administration. Open squares: placebo group (n = 50); filled squares: ergocalciferol group (n = 40). P values are for between-group comparisons by repeated-measures ANOVA (A) or by ANOVA of AUC (B). 25(OH)D_total, total 25-hydroxyvitamin D.

Among the group in the active arm, 25(OH)D_total concentrations after 12 wk of ergocalciferol treatment were negatively associated with HOMA-IR (r = −0.35, P = 0.03), as were 25(OH)D₂ concentrations (r = −0.34, P = 0.03). We found no association of 25(OH)D₃ concentration with HOMA-IR at week 12 (r = 0.02, P = 0.90). There was no association of 25(OH)D_total, 25(OH)D₂, or 25(OH)D₃ with HOMA-IR at 12 wk in the placebo group. These associations were not significant after Bonferroni correction for multiple comparisons. In post hoc within-group analyses, there were no associations of the AUC for HOMA-IR with 25(OH)D_total, 25(OH)D₂, or 25(OH)D₃ concentrations in the active arm by linear regression.

Effect of ergocalciferol on IVGTT measures

At baseline and at week 12, IVGTTs were performed and first-phase insulin secretory response and insulin sensitivity were calculated (30). As shown in Figure 3A, there was no significant difference in the 12-wk change in first-phase insulin secretion (decrease of 7.8 ± 31.8 mU · L⁻¹ · min/mmol · L⁻¹ in the treatment group and decrease of 3.7 ± 22.1 mU · L⁻¹ · min/mmol · L⁻¹ in the placebo group, P = 0.48). As shown in Figure 3B, there was no significant difference in the 12-wk change in insulin sensitivity (increase of 0.3 ± 3.0 min⁻¹/mU⁻¹ · min in the treatment group and change of 0.0 ± 1.9 min⁻¹/mU⁻¹ · min in the placebo group, P = 0.52, with higher values indicating increased insulin sensitivity). In a prespecified analysis, we found that the effect of ergocalciferol on insulin secretion and on insulin sensitivity was not modified by baseline measures of glucose homeostasis; in particular, we observed no difference between glucose-tolerant and glucose-intolerant subjects. In a second prespecified analysis, we found no evidence that race modified these same parameters. In post hoc analyses, we also found no significant interaction of sex, severity of baseline vitamin D deficiency (25(OH)D_total ≤10 ng/mL compared with 25(OH)D_total >10 ng/mL), or BMI (<25 compared with ≥25) with treatment for either of these outcomes. In within-group analyses, we found no association of 25(OH)D_total, 25(OH)D₂, or 25(OH)D₃ concentrations with change in insulin secretion or change in insulin sensitivity in the active arm.

FIGURE 3.

Mean ± SEM changes in first-phase insulin secretory response (mU ^· L^{−1 ·} min/mmol ^· L⁻¹) (A) and insulin sensitivity (min⁻¹/mU⁻¹ · min) (B) as measured by IVGTT with ergocalciferol administration (30). Open squares: placebo group (n = 50); filled squares: ergocalciferol group (n = 40). P values are for between-group difference in change in insulin secretion (A) or insulin sensitivity (B) by ANCOVA with control for baseline value. IVGTT, intravenous-glucose-tolerance test.

Ergocalciferol administration and additional metabolic outcomes

We found no significant between-group differences in the 12-wk change in BMI, waist-to-hip ratio, systolic or diastolic blood pressure, or concentrations of total cholesterol, LDL cholesterol, HDL cholesterol, or triglycerides (Table 2).

TABLE 2.

Twelve-week change (Δ) in selected metabolic parameters¹

	Placebo (n = 50)	Ergocalciferol (n = 40)	P value
Δ BMI, kg/m2	0.0 ± 0.8	0.3 ± 0.6	0.08
Δ Waist-to-hip ratio	−0.01 ± 0.03	0.01 ± 0.05	0.26
Δ Systolic blood pressure, mm Hg	−1 ± 11	−1 ± 7	0.87
Δ Diastolic blood pressure, mm Hg	0 ± 8	−1 ± 8	0.38
Δ Total cholesterol, mg/dL	2 ± 25	6 ± 21	0.38
Δ HDL cholesterol, mg/dL	−3 ± 10	−1 ± 6	0.21
Δ LDL cholesterol, mg/dL	5 ± 20	8 ± 17	0.50
Δ Triglycerides, mg/dL	1 ± 35	−3 ± 37	0.64

¹Values are means ± SDs of baseline values subtracted from week 12 values. P values were calculated by t test.

DISCUSSION

In this longitudinal trial of healthy adults with low 25(OH)D_total, we found no effect of ergocalciferol administration and consequent increase in 25(OH)D_total on measures of insulin secretion or resistance as assessed by IVGTT. In addition, we found no effect of ergocalciferol administration on other indices of glucose metabolism, including fasting glucose concentration, fasting insulin concentration, and HOMA-IR. Other markers of metabolic health, including blood pressure, serum lipids, and BMI, were similarly unaffected. HOMA-IR increased in both the placebo and treatment groups over the first 8 wk of the study and then declined to baseline at week 12. This variability was seen regardless of season of baseline visit. Although the reason for this fluctuation is not clear, the similarity between the treatment and placebo groups implies that this was not a result of ergocalciferol administration. Overall, these results imply that supplementation of unselected populations with ergocalciferol would not lead to decreased risk of the development of type 2 diabetes.

Published clinical trials of vitamin D administration on glycemic outcomes have had varied results. In the 2 largest clinical trials, supplementation with low daily doses (400–800 IU) of cholecalciferol (vitamin D₃) did not affect laboratory measures of glucose homeostasis or rates of incident type 2 diabetes (18, 19). The mean achieved 25(OH)D concentrations in these studies were estimated to be 26 and 25 ng/mL, respectively. In contrast, a smaller study found that administration of 700 IU vitamin D₃/d for 3 y attenuated the rise in fasting glucose and HOMA-IR in subjects with impaired fasting glucose at baseline (17). In this latter study, baseline 25(OH)D_total concentrations were higher (28–32 ng/mL), leading to a mean achieved 25(OH)D_total concentration of ∼40 ng/mL among subjects in the treatment arm.

A meta-analysis of studies investigating the effect of vitamin D administration on glycemic outcomes reported that the beneficial effect was limited to those with impaired glucose tolerance or diabetes (24). However, subsequent studies in this population have reported discordant results. In one recent study, supplementation with 2000 IU vitamin D₃ for 16 wk led to a beneficial increase in the glucose disposition index, an integrated measure of insulin secretion and sensitivity that reflects risk of type 2 diabetes (23, 34). This increase appeared to be driven by an increase in insulin secretion as measured by the first-phase insulin response. Conversely, in a study of African American subjects with prediabetes or mild diabetes, administration of 4000 IU vitamin D₃/d for 12 wk increased insulin secretion but decreased insulin sensitivity in response to OGTT, with no overall change in the glucose disposition index (35). Similarly, high-dose weekly vitamin D₃ supplementation for 12 mo among primarily Latino and black subjects with prediabetes showed no effect on measures derived from OGTT or on incident diabetes, with a significant but small (0.2%) decrease in hemoglobin A_1c (36). Mean achieved 25(OH)D_total in this study was ∼70 ng/mL. Although the number of subjects with glucose intolerance in our study was low, we saw no effect of ergocalciferol administration on glycemic outcomes in this subset of subjects. Interestingly, 2 recent studies reported that vitamin D administration, as either vitamin D₂ or vitamin D₃, improved glycemic outcomes in obese adolescents, suggesting that intervention at an earlier developmental stage may be more beneficial (37, 38).

One potential explanation for these discrepancies is the reliance on total 25(OH)D concentrations rather than free or bioavailable concentrations as measures of both deficiency and adequacy of therapy (39). Although renal reuptake of vitamin D occurs complexed to vitamin D–binding protein (40, 41), recent data suggest that bioavailable 25(OH)D may be a better marker of vitamin D sufficiency, at least for some outcomes (42–44). Similarly, genotypic variants of DBP, which encodes the vitamin D–binding protein, may predict the response of 25(OH)D concentrations to vitamin D supplementation (45). Our finding that the achieved 25(OH)D_total concentration at week 12 in the ergocalciferol arm correlated with the week 12 HOMA-IR may reflect a genotype-dependent effect.

Our study had several important strengths. Subjects were selected to have low serum 25(OH)D_total, which would tend to highlight an effect of increasing 25(OH)D_total concentration. Our inclusion serum 25(OH)D_total threshold was consistent with the Institute of Medicine definition of low vitamin D (27), and mean 25(OH)D_total concentrations in the ergocalciferol-treated subjects increased to >30 ng/mL, the concentration recommended in some recent practice guidelines, in 90% of the treated subjects (46). We used indexes derived from IVGTT that are more reproducible than fasting or OGTT measures, particularly in the setting of decreased β-cell reserve, and can separate effects on insulin secretion from effects on insulin resistance (25, 47). Our cohort was racially and ethnically diverse, distinct from prior studies of the effects of vitamin D administration in subjects with normal glucose tolerance (48). Finally, we intentionally recruited a healthy cohort and used a method, modified IVGTT, that would rigorously assess the effects of ergocalciferol repletion on insulin handling because of the public health implications of this research. Our negative findings are important, nonetheless, because they support the Institute of Medicine’s recommendation that appropriate caution be used when attributing beneficial effects to vitamin D and that randomized clinical trials to separate associations from causality as well as to define patient subtypes that may benefit from an intervention are essential.

Our study also had limitations. We had few subjects with impaired glucose tolerance and were thus not powered to detect effects in this subgroup. However, we did not observe even a trend toward improved insulin secretion or sensitivity with ergocalciferol administration among these subjects, suggesting that a larger sample size would not have led to statistically significant differences. In addition, although IVGTT measures are more reproducible, they are arguably less physiologic than an OGTT or mixed meal challenge in which carbohydrates are absorbed enterally (47). Although we enrolled subjects with screening 25(OH)D_total concentrations ≤20 ng/mL, because of the use of a more sensitive assay (liquid chromatography–tandem mass spectrometry compared with CLIA) for the on-protocol visits, some subjects had 25(OH)D_total >20 ng/mL at the first on-protocol visit. Some studies that have found a positive effect of vitamin D treatment on glycemic outcomes have enrolled subjects with lower baseline 25(OH)D_total concentrations (20, 21). However, given the lack of effect modification by baseline 25(OH)D_total concentration on insulin secretion or sensitivity, it is unlikely that this contributed to our negative result. Finally, it is also possible that an alternative replacement strategy, such as use of daily instead of weekly supplements or administration of cholecalciferol instead of ergocalciferol, may have altered our findings, particularly because many studies have suggested that cholecalciferol is more effective than ergocalciferol at raising 25(OH)D_total concentrations, in part because ergocalciferol supplementation may induce a decrease in 25(OH)D₃ concentrations (49–51). However, the sustained elevations in 25(OH)D_total that we observed suggest that, as other authors have found, our treatment strategy was adequate (52, 53).

In summary, our data add to a growing body of literature that suggests that an increase in the concentration of 25(OH)D_total does not improve metabolic parameters predictive of the development of type 2 diabetes, at least in unselected populations. Broadly, these data have 3 potential interpretations. Vitamin D may play no physiologic role in the regulation of glucose homeostasis; the previously reported associations of higher 25(OH)D_total concentrations and decreased insulin resistance or incident type 2 diabetes may reflect confounding. Alternatively, appropriate glucose regulation may be vitamin D dependent but require relatively low concentrations; thus, impaired insulin secretion or sensitivity would be observed only at extremely low concentrations. Finally, vitamin D may influence glucose homeostasis over a longer time scale, the demonstration of which would require an extended duration of vitamin D administration. Thus, future studies are warranted to examine the role of vitamin D in particular high-risk populations such as those with pre-existing abnormalities in glucose homeostasis and those with extremely low 25(OH)D concentrations. In addition, an improved understanding of the genetic contributions to total and bioavailable 25(OH)D concentrations, as well as of the interaction of 25(OH)D concentrations with other genetic variants contributing to glucose and insulin metabolism, may identify a target population likely to respond to this intervention.