1 INTRODUCTION

Vitamin D, a prohormone, is essential for bone mineral metabolism but several non-skeletal functions are being recognized, as vitamin D receptors have been identified in various tissues such as muscle and pancreatic tissue.1-3 Low 25-hydroxyvitamin D concentrations have consequently been associated with a variety of diseases such as obesity, insulin resistance (IR), sarcopenia, immune deficiency, cancer, cardiovascular disease, and increased fall risk.4-8 Vitamin D—in contrast to other vitamins—can be produced in the skin and is found in certain foods such as oily fish and eggs. Nevertheless, vitamin D insufficiency is common, due to a low vitamin D content in most foods and a lack of UVB-induced vitamin D synthesis at high latitudes.9-11 In older adults in particular, dietary deficiency of vitamin D is particularly common, especially in the case of malnutrition or alcohol consumption.12,13 Besides the inadequate dietary intake, diminished exposure to sunlight, reduced skin thickness and impaired intestinal absorption in the old lead to an increased risk for vitamin D insufficiency.12,14,15 In addition, impaired hydroxylation of 25-hydroxyvitamin D in the liver and kidney leads to altered metabolism of vitamin D in the old.13 The metabolic syndrome (MetS) is a cluster of at least three out of five important cardiovascular risk factors: these risk factors include hypertension, dyslipidemia, elevated waist circumference, and hyperglycemia. The prevalence of MetS increases with age16 and has been linked to a variety of diseases such as diabetes, osteoporosis, lung function decline, or cardiovascular disease and outcomes.16-21

Notably, several studies suggested that vitamin D insufficiency increases the risk for MetS in children, adolescents, and older people.22-29 Although, recent studies suggested that this association may be indirect and mainly driven by the frequently increased BMI and/or IR among individuals who are affected by the MetS. For example, in a large Korean study among 5559 adult participants, the association between vitamin D status and MetS vanished when BMI was included in the multivariate model.30 Similarly, Huang et al. reported that after adjustment for IR and BMI, the inverse relationship between vitamin D status and prevalence of the MetS disappeared among young subjects without diabetes.25 Thus, to date, the association between vitamin D and MetS remains unclear. Moreover, the influence of age on the association between vitamin D status and MetS has not been studied conclusively.31

To point out age-specific differences, within the current study of Berlin Aging study II (BASE-II) data, we assessed the odds of the MetS in vitamin D insufficiency among young (n = 416; age 22–37 years) and older adults (n = 1357; age 60–84 years, with and without T2D) including important covariate information on BMI, IR, and dietary as well as supplement-based vitamin D intake.

2 METHODS 2.1 Study population

For the current cross-sectional analysis, we included 1773 participants (a young cohort [n = 416; 52.4% female] and an old cohort [n = 1357; 51.6% female]) of BASE-II, a prospective epidemiological study described previously in detail elsewhere.32, 33 In brief, BASE-II is a population-based sample of community-dwelling participants, living in the greater metropolitan area of Berlin, Germany (the latitude of Berlin is 52°31′27″ N). All subjects gave written informed consent to participate in the study. The study was conducted according to the Declaration of Helsinki. The study was approved by the ethics committee of Charité–University Medicine Berlin (project number: EA2/029/09).

2.2 Exclusion criteria

Only subjects with complete data on MetS (waist circumference, triglycerides, high-density lipoprotein [HDL] cholesterol, blood pressure, glucose, HbA1c, and T2D) were included in the data analysis. This resulted in a sample of 1357 participants with and without T2D, aged between 60 and 84 years, and 416 young subjects without diabetes (20–37 years old). In order to prevent bias due to disease state, we analysed anamnestic information on history of Cushing disease, hyperparathyroidism, or severe liver disease, however no participant in the collected sample reported such a disease. With respect to the regression models calculated, we excluded subjects taking vitamin D supplements and subjects with potentially severe hypo-/hyperthyroidism (fT3 <1,80 or fT3 >5,70 pg/ml).

2.3 Diabetes mellitus (T2D)

T2D was defined according to the guidelines of the European Society of Cardiology (ESC),34 MetS was defined as suggested by the International Diabetes Federation/American Heart Association/National Heart, Lung and Blood Institute (IDF/AHA/NHLBI 2009).16 Vitamin D insufficiency was defined as 25(OH)D below 50 nmol/L according to the Institute of Medicine (IOM) guidelines.35 Blood pressure was measured with an electronic sphygmomanometer (boso-medicus memory, Jung Willingen, Germany), waist circumference was assessed using a non-elastic tape measure at the level of the umbilicus and elevated waist circumference was classified as ≥94 cm in men and ≥80 cm in women.16

2.4 Laboratory measurements

After a fasting period of at least 8 h, blood was collected from the subjects, subsequently stored at 4°C–8°C and prepared for transport and measurement on the same day. The laboratory parameters were analyzed by a certified laboratory (Labor 28 GmbH, Berlin and Labor Berlin). Serum triglycerides and HDL cholesterol were measured with enzymatic colour tests (Roche/Hitachi Modular; device: ACN 435 und ACN 781). Insulin concentrations (NaF-tube) were analyzed by chemiluminescence measurements; IR was calculated using fasting glucose and insulin concentrations in the homeostasis model of insulin resistance (HOMA-IR) as (fasting glucose (mg/dl) × fasting insulin (mu/ml))/405.36 25-hydroxyvitamin D serum concentrations were assessed with two different methods (Serum-tubes): Automated chemiluminescence immunoassays (IDS-iSYS 25-Hydroxy Vitamin D Immunoassay [IDS] and LIAISON 25 OH Vitamin D TOTAL Assay [DiaSorin]). A within-assay CV of 8%–21% and a between-assay CV of 8%–34% for DiaSorin and an intraassay CV <7.3% and an interassay CV <8.9% for IDS have been described previously.37 Moreover, variations between these two methods have been described.38, 39 The measurement of 25-hydroxyvitamin D serum concentrations was conducted by Labor Berlin, which performed a method comparison on 74 samples. An adjustment factor based on regression models was calculated and we performed an adjustment as: 25-hydroxyvitamin D serum concentrations (IDS) = 1.617 × 25-hydroxyvitamin D serum concentrations (Diasorin). Serum concentrations of C—reactive protein were analyzed using immunological turbidity test.

2.5 Co-variables

Regular alcohol intake (yes/no), current smoking status (yes/no), and physical activity (yes/no) were assessed by standardized questions. As part of the medical examination, diagnoses were obtained through participant reports, with select diagnosis (e.g., diabetes mellitus) being verified by additional laboratory tests. Diagnoses were used to compute a morbidity index largely based on the categories of the Charlson index, which is a weighted sum of moderate to severe, mostly chronic physical illnesses, including cardiovascular (e.g., congestive heart failure), cancer (e.g., lymphoma), and metabolic diseases (e.g., diabetes mellitus).40, 41 The medical examination included an anamnestic assessment of medications including vitamin D supplementation. To estimate dietary vitamin D intake, participants completed a validated, self-administered 146-item food frequency questionnaire (European Prospective Investigation into Cancer and Nutrition).42, 43

2.6 Statistical analysis

The statistical analysis was carried out using the software package SPSS 21 for Windows (IBM Corp. Released 2012. IBM SPSS Statistics for Windows, Version 21.0). We used the Kolmogorov-Smirnov test to examine normal distribution of all included variables. Normally distributed variables were analyzed using the parametric Student's t-test, variables with a skewed distribution were analyzed with the non-parametric Mann-Whitney-U-Test. Groups were compared by Chi2-test or Fisher's exact test. Regarding Tables 1 and 2, data are either shown as mean ± SD or number (%). p-values in Table 1 refer to the comparison of subjects with 25-hydroxyvitamin D serum concentrations <50 nmol/L and subjects with 25-hydroxyvitamin D serum concentrations ≥50 nmol/L, independently calculated for the young cohort and the old cohort. Subjects with vitamin D supplementation and potentially difficult hypo-/hyperthyroidism (fT3 <1,80 or fT3 >5,70 pg/ml) were excluded from the regression models. p-values in Table 2 refer to the comparison of subjects with MetS and without MetS independently calculated for the young cohort (subjects without diabetes) and the two old cohorts (subjects with diabetes and without diabetes). Finally, binary logistic regression models with MetS as a dependent variable were calculated to assess the association of vitamin D insufficiency with MetS (Table 3; p-values from binary logistic regression models excluding subjects with vitamin D supplementation). Regression models were independently calculated for the young cohort (subjects without diabetes) and the two old cohorts (subjects with diabetes and without diabetes). In a first step, we adjusted for an increasing number of potential co-variables. Model 1 is adjusted for age, and sex; Model 2 is additionally adjusted for physical activity, smoking status, alcohol intake, total energy intake/day, dietary vitamin D intake/day, morbidities, and CRP. Models for subjects with diabetes were additionally adjusted for antidiabetic medication (none, insulin, metformin, other oral antidiabetic medication). Moreover, we recalculated these models including HbA1c as a proxy for metabolic control of patients with T2D (Model 5). Finally, we included HOMA-IR and BMI in these binary logistic regression models (Model 3, Model 4, and Model 5) to work out the influence of obesity and IR on the association between MetS with vitamin D insufficiency. p-values for the association between BMI with MetS and for the association between HOMA-IR with MetS are from binary logistic regression models (adjusted for age, sex, physical activity, smoking status, alcohol intake, total energy intake/day, dietary vitamin D intake/day, morbidities, and CRP).

TABLE 1. Baseline characteristics of study participants according to vitamin D insufficiency and age Young subjects (n = 415) Old subjects (n = 1357) Vitamin D <50 nmol/L (n = 211) Vitamin D ≥50 nmol/L (n = 205) p Value Vitamin D <50 nmol/L (n = 609) Vitamin D ≥50 nmol/L (n = 748) p Value Age [years] 28 ± 3 28 ± 3 0.872 68 ± 4 68 ± 3 0.248 BMI [kg/m2] 23.6 ± 4.6 23.0 ± 3.9 0.305 27.3 ± 4.5 26.4 ± 3.8 0.001 Waist circumference [cm] 83.9 ± 12.6 80.2 ± 10.5 0.003 97.7 ± 12.3 94.7 ± 11.3 0.001 Vitamin D [nmol/L] 73.8 ± 21.9 29.7 ± 10.0 0.001 71.5 ± 20.6 31.4 ± 10.0 0.001 SBP [mmHg] 124 ± 15 120 ± 14 0.009 145 ± 19 141 ± 20 0.001 DBP [mmHg] 79 ± 10 77 ± 10 0.006 84 ± 11 82 ± 12 0.040 HOMA - IR 1.5 ± 1.8 1.7 ± 1.7 0.098 2.9 ± 4.9 2.2 ± 2.9 0.001 TAG [mg/dl] 96 ± 54 89 ± 47 0.250 118 ± 63 106 ± 61 0.001 HDL [mg/dl] 59 ± 17 62 ± 16 0.028 61 ± 17 64 ± 17 0.001 MetS [n; %] 22 (10.7) 10 (4.7) 0.017 251 (41.2) 232 (31.0) 0.001 T2D [n; %] - - - 84 (13.8) 88 (11.8) 0.150 Smoking status [n; %] 57 (27.8) 71 (33.6) 0.257 69 (11.3) 57 (7.6) 0.064 Regular alcohol intake [n; %] 188 (91.7) 191 (90.5) 0.410 533 (87.5) 687 (91.8) 0.006 Self-reported physical inactivity [n; %] 19 (9.5) 18 (8.6) 0.438 73 (12.1) 52 (7.1) 0.001 Energy intake [kcal/d] 2420 ± 806 2297 ± 798 0.151 2288 ± 750 2222 ± 671 0.224 Dietary vitamin D intake [µg/day] 4.1 ± 2.7 4.1 ± 2.5 0.976 4.8 ± 2.7 5.2 ± 3.0 0.001 CRP [mg/L] 1.4 ± 2.2 1.9 ± 4.3 0.683 2.1 ± 2.7 1.8 ± 2.7 0.002 Note: Data are either mean ± SD or number (%). Abbreviations: BMI, Body Mass Index; CRP, C-reactive protein; DBP, diastolic blood pressure; HDL, high-density lipoprotein; HOMA-IR, homeostasis model assessment of insulin resistance; MetS, metabolic syndrome; SBP, systolic blood pressure; T2D, diabetes mellitus; TAG, triglycerides; vitamin D insufficiency, vitamin D <50 nmol/L. TABLE 2. Baseline characteristics of study participants according to metabolic syndrome, diabetes, and age Non-diabetic young (n = 415) p Non-diabetic old (n = 1185) p Diabetic old (n = 172) p No MetS (n = 383) MetS (n = 32) No MetS (n = 852) MetS (n = 333) No MetS (n = 22) MetS (n = 150) Age [years] 28 ± 3 29 ± 3 0.593 68 ± 4 68 ± 4 0.944 68 ± 3 68 ± 4 0.587 BMI [kg/m2] 22.7 ± 3.6 29.8 ± 5.24 0.001 25.5 ± 3.6 28.6 ± 4.1 0.001 25.1 ± 4.2 30.1 ± 4.4 0.001 Waist circumference [cm] 80.6 ± 10.6 99.5 ± 10.7 0.001 92.2 ± 10.8 102.1 ± 10.1 0.001 89.5 ± 12.2 105.5 ± 9.9 0.001 Vitamin D [nmol/L] 53.1 ± 28.1 39.5 ± 22.0 0.003 55.2 ± 25.9 51.4 ± 26.7 0.072 59.3 ± 27.8 48.0 ± 24 0.003 Vitamin D insufficiency [n; %] 182 (47.5) 22 (68.8) 0.021 352 (41.3) 173 (52.0) 0.001 6 (27.3) 78 (52.0) 0.030 Vitamin D supplementation [n; %] - - 59 (6.9) 19 (5.7) 0.447 - 7 (4.7) - Current smoker [n; %] 119 (31.1) 9 (28.1) 0.416 79 (9.3) 27 (8.1) 0.369 4 (18.2) 16 (10.7) 0.590 Regular alcohol intake [n; %] 350 (91.4) 28 (87.5) 0.459 759 (89.1) 309 (92.8) 0.054 19 (86.4) 133 (88.7) 0.753 Self-reported physical inactivity [n; %] 30 (8.0) 7 (23.3) 0.005 56 (6.7) 42 (12.9) 0.001 4 (20.0) 23 (15.4) 0.601 Energy intake [kcal/d] 2361 ± 814 2306 ± 689 0.865 2206 ± 676 2360 ± 755 0.001 2360 ± 604 2252 ± 766 0.257 Dietary vitamin D intake [µg/day] 4.1 ± 2.6 4.4 ± 2.7 0.481 5.0 ± 2.9 5.1 ± 2.9 0.644 3.2 ± 4.9 4.9 ± 2.7 0.134 HOMA - IR 1.5 ± 1.6 3.2 ± 2.3 0.001 1.6 ± 0.8 3.1 ± 2.0 0.001 3.7 ± 5.2 6.1 ± 10.3 0.001 CRP [mg/L] 1.5 ± 3.45 3.4 ± 2.9 0.001 1.7 ± 2.6 2.3 ± 2.9 0.001 2.0 ± 2.4 2.5 ± 2.5 0.278 Note: Data are either mean ± SD or number (%). Abbreviations: BMI, Body Mass Index; CRP, C-reactive protein; HOMA-IR, homeostasis model assessment of insulin resistance; MetS, metabolic syndrome; T2D, diabetes mellitus type II; vitamin D insufficiency, vitamin D <50 nmol/L. TABLE 3. Association of vitamin D insufficiency and metabolic syndrome according to diabetes and age Young participants without diabetes (n = 415) Old participants without diabetes (n = 1126) Old participants with diabetes (n = 165) OR (95% CI) p OR (95% CI) p OR (95% CI) p Model 1 2.3 (1.1–5.0) 0.036 1.6 (1.2–2.1) 0.002 3.7 (1.3–11.1) 0.018 Model 2 2.8 (1.2–7.0) 0.026 1.5 (1.1–2.1) 0.006 7.4 (1.4–39.4) 0.019 Model 3 2.9 (1.0–8.2) 0.043 1.5 (1.1–2.1) 0.014 4.4 (0.9–27.0) 0.072 Model 4 3.2 (1.2–8.4) 0.019 1.3 (0.09–1.8) 0.216 9.3 (1.5–56.0) 0.015 Model 5 3.2 (1.0–8.7) 0.042 1.3 (0.9–1.8) 0.203 6.1 (1.0–36.8) 0.047 Note: Data show OR and 95% CIs. p for trend is from binary logistic regression models. Models were calculated excluding subjects with vitamin D supplementation and subjects with fT3 <1,80 or fT3 >5,70 pg/ml. Model 1: adjusted for age, sex, antidiabetic medication (in subjects with diabetes); Model 2: Model 1 + physical activity, smoking status, alcohol intake, energy intake/day, dietary vitamin D intake, co-morbidities (modified Charlson morbidity index), CRP; Model 3: Model 2 + BMI; Model 4: Model 3 + HOMA-IR; Model 5: Model 2 + BMI + HOMA-IR. 3 RESULTS

Mean 25(OH) concentrations were 52.1 ± 27.9 nmol/L in the young cohort and 53.5 ± 26.0 nmol/L in the old cohort. MetS was prevalent in 7.7% of the young and 35.6% of the old cohort and T2D occurred in 12.7% of the old cohort. We found that the distribution of MetS parameters differed between young and old. Whereas elevated blood pressure (93.8%), elevated waist circumference (87.5%), and low HDL (71.9%) were the most frequently identified parameters in young subjects with MetS, elevated waist circumference (99.1% of old subjects without diabetes, 99.3% of old subjects with diabetes), IR (60.6% of old subjects without diabetes, 100% of old subjects with diabetes) and elevated blood pressure (97.3% of old subjects without diabetes, 98% of old subjects with diabetes) were predominant in the old with MetS. Vitamin D insufficiency was present in 49.3% of the young and in 44.9% of the old (p = 0.064). Dietary intake of vitamin D was significantly lower in the old cohort with vitamin D insufficiency. Basic and clinical characteristics of both age groups according to vitamin D insufficiency are summarized in Table 1. Subjects with T2D had significantly higher BMI and HOMA-IR concentrations compared to old and young subjects without diabetes (p < 0.001). In addition, subjects with T2D had higher concentrations of triglycerides, exhibited more parameters of MetS, had higher CRP concentrations and lower concentrations of HDL (p < 0.001, data not shown).

Irrespective of age group, participants with vitamin D insufficiency were more frequently diagnosed with MetS, as waist circumference and blood pressure were significantly higher in these subjects, and HDL-concentrations were lower. In addition, old participants with vitamin D insufficiency had higher concentrations of HOMA-IR, higher BMI, higher triglyceride concentrations, higher CRP-concentrations and reported more frequently to be physically inactive and to consume alcohol on a regular basis (Table 1).

In both age groups and independent from T2D, participants with MetS suffered more frequently from vitamin D insufficiency (Table 2). BMI, HOMA-IR and waist circumference were higher in subjects with MetS independent from age group or T2D. Young subjects without diabetes and old subjects with MetS moreover reported more frequently to be physically inactive and had higher concentrations of CRP, whereas these differences were not observed between the old subjects with diabetes and with or without MetS (Table 2).

To assess the association between vitamin D insufficiency and MetS, we calculated different regression models adjusted for an increasing number of covariates. As summarized in Table 3, in adjusted models, vitamin D insufficiency increased the risk for MetS in all cohorts (Models 1 and 2). In the young subjects without diabetes, vitamin D insufficiency was associated with higher odds of having MetS even in the fully adjusted model, that is, independent from HOMA-IR and BMI (Model 5). However, these results were not observed in the old cohort with diabetes, when BMI was taken into account (Model 3) and neither seen in the old cohort without diabetes after adjustment for HOMA-IR (Models 4 and 5). HOMA-IR showed a significant association with MetS (Model 5) in young subjects without diabetes (p = 0.011), in old subjects without diabetes (p < 0.001), but not in old subjects with diabetes (p = 0.122), whereas BMI had a significant association with MetS (Model 5) in young subjects without diabetes (p < 0.001), in old subjects without diabetes (p = 0.003) and in old subjects with diabetes (p < 0.006).

4 DISCUSSION

Our results from the cross-sectional analysis within BASE-II support earlier findings on the relationship between vitamin D insufficiency with MetS. Vitamin D insufficiency was more frequent in subjects with MetS, irrespective of age group or T2D. Moreover, metabolic load (represented by lower HDL concentrations, higher HOMA-IR concentrations, higher systolic/diastolic blood pressure, and elevated waist circumference) was higher in subjects with low vitamin D concentrations. These results remained stable in the young cohort even after adjustment for multiple co-variates. However, in the old cohorts, we obtained that when BMI was taken into account, the relationship between vitamin D insufficiency and MetS was no longer evident in the old subjects with diabetes. Similarly, in the old subjects without diabetes, the relationship between vitamin D and MetS vanished when the analysis was adjusted for HOMA-IR as marker for IR. The differing results between the young and the old group imply that age-related factors may play a modulating role in the association between vitamin D insufficiency and MetS. However, differing effects of the covariates on the relationship between vitamin D insufficiency and MetS in our three study cohorts are not easily understood.

The majority of cross-sectional studies have reported an inverse relationship between vitamin D with MetS, diabetes and beta-cell function7-12, 23 and longitudinal studies found a subsequent higher risk for MetS in subjects with low vitamin D concentrations.44-46 Also, studies analyzing the effect of preventive health programs promoting vitamin D supplementation found reduced risk of developing MetS with improved vitamin D concentrations.26, 31, 47 Most of the authors conclude that disturbed glucose homeostasis and obesity play a decisive role and recent studies even suggested that the association between MetS and vitamin D may mainly be driven by the frequently increased BMI and/or IR among individuals with MetS.25, 31, 46 Following this approach, Huang et al. found in their study on 355 young subjects without diabetes, that the association between MetS with vitamin D vanished, when BMI and HOMA-IR were taken into account.25 This in contrast to our findings. However, in old participants, Kim et al. reported similar results. Low vitamin D concentrations were not associated with higher risk of MetS when regression models were adjusted for BMI.30

Vitamin D-dependent processes modulate effects of insulin,2, 4, 48 that is, through upregulation of insulin dependent-receptors on muscle tissue and stimulation of receptors sensitivity to insulin.2, 8, 49-