Serum 25-Hydroxyvitamin D: A Predictor of Macrovascular and Microvascular Complications in Patients With Type 2 Diabetes

Diabetes is among the leading causes of death and affects 347 million people worldwide. The World Health Organization expects a 50% increase in deaths from diabetes over the next 10 years, and by 2030, diabetes is projected to be the seventh leading cause of death (1). Most major complications involve large vessel (macrovascular) or small vessel (microvascular) disease.

Recent reports have suggested associations between vascular disease (2–5), diabetes (6–8), and vitamin D deficiency. Higher rates of cardiovascular disease with lower vitamin D levels have also been reported, but most of these studies have been cross-sectional or based on unadjudicated end points. Rigorous large-scale prospective studies are needed to evaluate the relationship between circulating 25 hydroxyvitamin D (25OH-D) concentrations and vascular disease in diabetic patients, and, more particularly, no large-scale studies have addressed the issue of microvascular diseases.

The Fenofibrate Intervention and Event Lowering in Diabetes (FIELD) study provides a unique opportunity to examine the relationship of blood vitamin D concentration with macrovascular and microvascular events (9–11). In this study, we report the relationships between baseline blood 25OH-D concentrations and prospective macrovascular and microvascular events in the FIELD cohort.

The FIELD trial was a multinational, double-blind, placebo-controlled trial in 63 centers in Australia, New Zealand, and Finland. All patients gave written informed consent. The study protocol was approved by local and national ethics committees and undertaken in accordance with the Declaration of Helsinki and Good Clinical Practice guidelines.

The study population (Supplementary Fig. 1) consisted of 9,795 participants aged 50–75 years with type 2 diabetes diagnosed according to World Health Organization criteria who were not taking lipid-modifying therapy at study entry. A detailed description of the study design has been published (12). Patients recruited from hospital clinics and community sources were randomly allocated to once-daily micronized fenofibrate 200 mg (Laboratoires Fournier, Dijon, France) or matching placebo. Exclusion criteria included renal impairment (blood creatinine >130 μmol/L), known chronic liver disease or symptomatic gall bladder disease, and a cardiovascular event within the 3 months before recruitment. Blood 25OH-D results at baseline were available from 9,524 FIELD participants (Supplementary Fig. 1).

Fasting blood samples were collected into serum tubes. After clotting, serum was separated and frozen on site. Frozen samples were transported to one of two central laboratories, in Adelaide (Australia) and Helsinki (Finland), where they were aliquoted and stored at −80°C until analysis. For the measurement of 25OH-D, frozen samples were transported to the Department of Biochemistry at Royal Prince Alfred Hospital in Sydney (Australia). Only 271 (2.8%) samples were unavailable for analysis. Blood 25OH-D was assayed by a chemiluminescence immunoassay (DiaSorin Spa, Saluggia, Italy) on a Liaison automated analyzer (DiaSorin Spa). This assay detects 25OH-D₂ and 25OH-D₃ in an equimolar fashion. The lower limit of detection with this assay was 10 nmol/L; undetectable concentrations were assigned a value midway between this and zero (5 nmol/L). The assay used in this study was a premarket version that has been shown to compare closely with higher-order methods, including liquid chromatography-tandem mass spectrometry and radioimmunoassay (13). In addition, it has been shown to have performance identical to that of the current market version of the DiaSorin Liaison assay. Assay performance was monitored over time by internal quality-control procedures including the use of high- and low-concentration pooled sera. Levels of 25OH-D <65 nmol/L have been associated with rising levels of parathyroid hormone (14). While this may imply a broader agreement for the bone-related actions of 25OH-D, cutoff levels for many of the other actions of 25OH-D have not been established as yet. In agreement with previous recommendations from the Institute of Medicine, 25OH-D concentrations <50 nmol/L were deemed to be low for the purposes of this report.

Other biomarkers used for adjustment in analysis were measured using standardized routine methods as described previously (10). Measurement of hs-CRP used an automated immune-turbidometric assay from Roche Diagnostics run on a Modular E170 analyzer (Roche Diagnostics, Mannheim, Germany). Baseline physical activity levels were recorded in response to a self-reported questionnaire including the categories self-care/home (e.g., house work, lawn mowing), occupational (type of work, e.g., sitting, home building), recreation (sports), and physical conditioning (aspects of general fitness). All categories were recorded on a scale from 1 (very light) to 5 (very heavy) (15).

A detailed description of all events and how they were verified has been published elsewhere (10,12). Macrovascular events included myocardial infarction, stroke, cardiovascular death, and coronary or carotid revascularization. Incident events were verified by an outcome assessment committee masked to study treatment allocation. Incident microvascular events were classified as follows: peripheral neuropathy, abnormal monofilament test; nephropathy, urinary albumin-to-creatinine ratio ≥2.5 mg/mmol for men and ≥3.5 mg/mmol for women; retinopathy, on-study laser treatment for diabetic retinopathy (including macular edema); and amputation or minor amputation without known peripheral vascular disease in the same limb. Measurements of urinary albumin excretion and the monofilament test were performed at baseline, at 2 and 5 years, and at study close.

Normally distributed variables were summarized as means and SDs and nonnormally distributed variables as medians and interquartile ranges. χ² tests were used to compare categorical variables. For continuous variables, groups were compared using t tests or Wilcoxon rank-sum tests if the variables were nonnormally distributed. Spearman correlations were used for bivariate analysis of continuous variables.

Prospective associations with incident events during follow-up were assessed using unadjusted and adjusted Cox proportional hazards analysis for macrovascular events and logistic regression for microvascular events (as some of the latter were assessed only at specific time points). Analyses were stratified by study treatment allocation and adjusted for baseline event status. Adjusted models also included sex, age, diabetes duration, HbA_1c, systolic blood pressure, BMI, lipids (triglycerides, HDL cholesterol, and LDL cholesterol), smoking, baseline use of oral hypoglycemic agents, and baseline use of insulin.

In exploratory analyses, we also adjusted for physical activity score, hs-CRP, and season. For each patient, date of sample collection and country of residence (northern or southern hemisphere) were used to identify the season at the time of sample collection (winter, spring, autumn, and summer). Season was then included in the multivariable model as a factor variable.

As well as vitamin D in the models as a continuous variable (with results reported for increments of 50 nmol/L), quartiles of vitamin D were also examined. Cumulative risk curves of macrovascular and microvascular events were derived from the Kaplan-Meier method.

Two-sided P values <0.05 were considered statistically significant. All analyses used SAS (version 9.2) and Analysis of Censored and Correlated Data software (version 1.6.3).

The study treatment groups were well matched for baseline characteristics, such as age, BMI, smoking habits, blood pressure, and duration of diabetes (10). Both groups had similar cardiovascular histories. The use of glucose-lowering and cardiovascular medication was as expected for a cohort of people with diabetes and did not differ between the two groups. The median 25OH-D concentration in the entire cohort was 49 nmol/L (range 5–196 nmol/L [Fig. 1]), and there was no significant difference between the two groups. Eight percent, 52%, and 89% had blood 25OH-D concentrations <25 nmol/L, <50 nmol/L, and <75 nmol/L, respectively. Vitamin D supplements were used by 0.1%, but only one patient used a dose >700 IU daily.

In patients who had low vitamin D at baseline (blood 25OH-D <50 nmol/L), cardiovascular disease, hypertension, retinopathy, and nephropathy were more prevalent (Table 1). In addition, these patients had a higher baseline prevalence of dyslipidemia and microalbuminuria. On average, patients with low 25OH-D also had 0.3% (3 mmol/mol) higher HbA_1c, 0.3 mg/L higher hs-CRP, and 1.2 kg/m² higher BMI. There were significantly fewer former and current smokers among the participants with low 25OH-D.

During the 5-year follow-up period, 1,250 patients (13.1% of the entire cohort) had at least one macrovascular event, and 2,395 (25.1% of the entire cohort) had at least one microvascular event. Most of the affected patients had one or two events during the 5-year follow-up. New or worsening nephropathy (1,632 or 17.1% of the entire cohort) and new neuropathy (651 or 6.8% of the entire cohort) were the most common microvascular complications. Retinopathy requiring laser surgery and amputations occurred in <5% of the entire cohort.

Blood 25OH-D was an independent predictor of incident macrovascular events over 5 years of follow-up. This relationship remained unchanged in multivariable analyses (Table 2). For each 50 nmol/L difference in baseline blood 25OH-D concentration, the unadjusted risk of incident cardiovascular events during follow-up was altered by 20% (95% CI 5–39%; P = 0.01). Multivariable adjustment confirmed a significant prospective association between baseline blood 25OH-D and incident cardiovascular events with a 23% (95% CI 6–42%; P = 0.007) higher risk per 50 nmol/L lower baseline concentration. Exploratory further adjustment for physical activity score, hs-CRP, or seasonality level did not materially change these relationships (Table 3). The association between quartiles of blood 25OH-D and incident macrovascular events is shown in Table 2. In the adjusted model, patients in the lowest quartile of blood 25OH-D (<36 nmol/L) showed 21% higher risk of macrovascular disease than those in the highest quartile (>63 nmol/L).

Patients with a low blood 25OH-D concentration (less than the median [50 nmol/L]) had a higher cumulative incidence of macrovascular events than those whose concentrations fell above the median (Fig. 2A). The cumulative incidence of microvascular disease was not significantly higher at the prespecified cut point (Fig. 2B), but a higher incidence was evident at levels of 25OH-D <36 nmol/L. Analysis using this cut point, which was not prespecified, revealed that the incidence was significantly higher in patients with 25OH-D levels <36 nmol/L (Supplementary Fig. 2).

Blood 25OH-D was significantly associated with new microvascular complications over 5 years of follow-up (Table 2). Compared with those whose blood 25OH-D concentrations fell above the median (50 nmol/L), the lower group had an absolute excess of events during follow-up (Fig. 2). Prospective analysis of incident events during follow-up revealed 18% (95% CI 5–32%; P = 0.006) higher unadjusted risk of a microvascular event for each 50 nmol/L difference in 25OH-D blood concentration and a 15% (95% CI 1–30%; P = 0.03) higher risk after multivariable adjustment. Exploratory further adjustment for HbA_1c, physical activity score, or seasonality abolished its significance and reduced the incremental risk to 11 to 12% (not significant, Table 3).

This large-scale prospective study demonstrates an inverse association between the serum concentration of 25OH-D and future macrovascular and microvascular disease that was independent of treatment and the duration of diabetes. Higher baseline blood 25OH-D was associated with reductions in the risk of macrovascular disease by 20% and microvascular disease by 18% per 50 nmol/L, which, after adjustment for classical risk predictors, including HbA_1c, became 23 and 11% (not significant), respectively. 25OH-D levels <36 nmol/L were significantly associated with future microvascular events, but analysis of this cut point was based on visualization of the distribution of events and was not prespecified. While it raised the possibility that more severe 25OH-D depletion may contribute directly to the risk of microvascular events, analysis adjusted for HbA_1c still fell short of significance (P = 0.07).

The results are in keeping with cross-sectional (5) and prospective (4) studies showing an inverse relationship between 25OH-D and cardiovascular disease. A cross-sectional analysis of data from the third National Health and Nutrition Examination Survey showed a higher prevalence of vitamin D deficiency in patients with cardiovascular disease than in patients without (29.3 vs. 21.4%) (5). In addition, subjects with low 25OH-D had an odds ratio for pre-existing cardiovascular disease of 1.20 (95% CI 1.01–1.36; P = 0.03). However, the study was cross-sectional and lacked a rigorous verification of cardiovascular disease. In a nested case-control study including 454 patients with cardiovascular events and 900 matched control subjects, individuals with vitamin D deficiency at baseline had a 10-year relative risk of nonfatal myocardial infarction or fatal coronary heart disease of 2.09 (95% CI 1.24–3.54; P = 0.02 for trend) (4). Case-control studies typically overestimate risk, are affected by selection bias, and lack randomization. In contrast, FIELD was a prospective, randomized study with 1,250 cardiovascular events rigorously verified by an independent event review committee.

The results of our study expand existing knowledge by showing prospectively in a large cohort that low blood 25OH-D is also a risk factor for development of peripheral neuropathy, retinopathy, nephropathy, and amputation. Previous cross-sectional studies reported lower circulating 25OH-D concentrations in diabetic patients with microvascular disease events (16–19). However, all of these studies were limited by their size and design. In the National Health and Nutrition Examination Survey, for example, the diagnosis of peripheral neuropathy is based on self-reported symptoms of neuropathy, such as pain, tingling, numbness, or loss of feeling in hands or feet (19). In contrast, Kaur et al. (20) reported a higher prevalence of vitamin D deficiency in patients with diabetic retinopathy but found no association between vitamin D deficiency and neuropathy or nephropathy.

Potential mechanisms that explain the relationship between vitamin D deficiency and vascular disease (microvascular and macrovascular disease) include pancreatic β-cell dysfunction, peripheral insulin resistance, chronic inflammation, and endothelial dysfunction. In animal models, vitamin D deficiency impairs insulin synthesis (21,22), possibly via a reduced intracellular calcium concentration (23). The final statistical models presented in our study were adjusted for indices of glycemic control, such as glycated hemoglobin, fasting glucose, and insulin, so β-cell dysfunction and abnormal glucose homoeostasis may partly explain the increased risk of microvascular, but not macrovascular, disease, with decreasing blood concentrations of 25OH-D.

Despite the well-established link between immune function and vitamin D (24), reduction in blood 25OH-D concentrations may be a nonspecific marker of chronic illness, and the associations between blood 25OH-D and clinical outcomes may represent reverse causality. In our study, blood 25OH-D remained a significant predictor of macrovascular and microvascular events after adjustment for hs-CRP, a biomarker of chronic inflammation. Recent metanalyses failed to find benefits for survival, cardiovascular disease, or multiple outcomes (other than birth weight) following vitamin D supplementation (25,26). In contrast, metanalyses that differentiated between D₂ and D₃ supplementation demonstrated a selective benefit following the use of D₃ (27,28). In the current study, blood 25OH-D measurements reflected the concentration of the combination of D₂ and D₃.

Sedentary lifestyle characterized by insufficient physical activity is another established risk factor for cardiovascular disease and type 2 diabetes. Physical activity may increase blood 25OH-D concentrations as a consequence of an associated increase in sunlight exposure. Sarcopoenia and reduced muscle strength are associated with low blood 25OH-D concentrations (29), and a number of recent publications support the presence of the vitamin D receptor (VDR) in numerous tissues including skeletal muscle (30,31). The study by Autier et al. (25) detected a beneficial effect of lower doses of vitamin D among elderly women in whom mobility was more likely to be reduced. Adjustment for activity level affected the inverse association between vitamin D and microvascular, but not macrovascular, outcomes.

Direct effects of vitamin D on vessel function are another explanation for the relationship between vitamin D and vascular disease (32–34). However, Wang and DeLuca (35) recently reported that the VDR is not expressed in cardiac myocytes and vascular smooth muscle cells, which challenges the hypothesis of direct vascular effects of vitamin D. In fact, it appears that most of the antibodies used to identify VDR in these previous studies detect proteins other than VDR and may have led to false-positive results. In summary, existing data are insufficient to establish the mechanism responsible for the increased risk of vascular events in people with low vitamin D. However, the present results suggest that the relationship between serum 25OH-D and microvascular complications is partially due to the surrogacy for the effects of diabetes, season, and activity.

The median blood 25OH-D concentration in the FIELD cohort was 49 nmol/L, indicating that >50% of the participants patients had low 25OH-D. Furthermore, 25% had blood 25OH-D concentrations of ≤36 nmol/L. This supports previous reports showing a high prevalence of vitamin D deficiency in the general population and in elderly individuals worldwide (36–38). Considering the relationship of vitamin D with common chronic diseases, the high prevalence of vitamin D deficiency has socioeconomic implications, affecting individual patients and health care systems.

The main strengths of the study are its prospective and randomized design, the large number of participants, and the high number of verified microvascular and macrovascular events. Blood 25OH-D was measured with a well-characterized latest generation chemiluminescence immunoassay that compares well with higher-order methods, such as liquid chromatography-tandem mass spectrometry (13). This is important, as the performance of vitamin D assays varies substantially (13,39).

Our study also has some limitations. Fenofibrate treatment is known to reduce cardiovascular and microvascular events and may have biased our results. However, the effect of blood 25OH-D on microvascular and macrovascular risk was similar in the placebo and the fenofibrate groups, and no heterogeneity was detected. Individuals with low 25OH-D had a higher prevalence of microvascular and macrovascular disease at baseline, which predisposes them to a higher event rate during the study period, but there was no difference in the association seen among people with and without prior macrovascular and microvascular disease.

These relationships raise the question whether vitamin D supplementation can lower the risk of vascular events (25,26). Few prospective intervention studies with sufficient statistical power have addressed this question. Both the Women’s Health Initiative study, which followed >36,000 women, and the RECORD trial, which studied >5,000 patients, failed to show cardiovascular benefit (40,41), but metanalysis suggests the formulation of vitamin D may be important (27,28).

The findings also raise the question whether vitamin D supplementation can lower the risk of diabetic microvascular complications, possibly via an effect on HbA_1c. Short-term administration of a supraphysiologic dose of vitamin D does not affect insulin sensitivity (42). In contrast, supplementation with 2,000 IU of cholecalciferol for 16 weeks has been reported to improve β-cell function (43). A recent meta-analysis suggested that vitamin D administration did not improve HbA_1c or overall mortality rate (25). Together with our results, the negative outcomes of vitamin D supplementation studies highlight the importance of endogenous vitamin D synthesis and the need for further research into the role of vitamin D deficiency.

In conclusion, this study provides prospective evidence for serum vitamin D as a risk marker for future vascular events. The association of vitamin D with macrovascular disease was not explained by associated effects on β-cell function, glucose homeostasis, physical activity, or seasonality, while the novel relationship between serum vitamin D and microvascular disease was attributable to these covariates. The mechanisms underlying these relationships require further research and the results of large-scale supplementation trials.

Acknowledgments. The authors thank Dijana Bosnjak, Scott Chan, and Veronica Dy of the Department of Biochemistry at Royal Prince Alfred Hospital for vitamin D assays and Diasorin for access to assay equipment and technical support.

Funding and Duality of Interest. The FIELD study was supported by a grant from Laboratoires Fournier SA, Dijon, France (now part of Abbott Pharmaceuticals) and by the National Health and Medical Research Council. This substudy had no specific funding. The sponsors of the study had no role in data collection or analysis. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. M.H. and D.R.S. conceived the study and wrote the paper. A.-S.V. contributed to the design of the study and analyzed the data. T.M. and I.R.S. interpreted the data, revised the paper, and established and performed the laboratory analyses. R.S., M.L., D.T., A.J.J., and S.B. interpreted the data and revised the paper. A.C.K. conceived the study. A.C.K. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.