Prospective association of 25(OH)D with metabolic syndrome

Authors

  • Sheena Kayaniyil,

    1. Department of Nutritional Sciences, University of Toronto, Toronto, ON, Canada
    Search for more papers by this author
  • Stewart B. Harris,

    1. Centre for Studies in Family Medicine, Western University, London, ON, Canada
    Search for more papers by this author
  • Ravi Retnakaran,

    1. Division of Endocrinology, Department of Medicine, University of Toronto, University of Toronto, Toronto, ON, Canada
    Search for more papers by this author
  • Reinhold Vieth,

    1. Department of Nutritional Sciences, University of Toronto, Toronto, ON, Canada
    Search for more papers by this author
  • Julia A. Knight,

    1. Dalla Lana School of Public Health, University of Toronto, Toronto, ON, Canada
    2. Samuel Lunenfeld Research Institute, Mount Sinai Hospital, Toronto, ON, Canada
    Search for more papers by this author
  • Hertzel C. Gerstein,

    1. Division of Endocrinology and Metabolism and the Population Health Research Institute, Department of Medicine, McMaster University and Hamilton Health Sciences, Hamilton, ON, Canada
    Search for more papers by this author
  • Bruce A. Perkins,

    1. Division of Endocrinology, Department of Medicine, University of Toronto, University of Toronto, Toronto, ON, Canada
    Search for more papers by this author
  • Bernard Zinman,

    1. Division of Endocrinology, Department of Medicine, University of Toronto, University of Toronto, Toronto, ON, Canada
    2. Samuel Lunenfeld Research Institute, Mount Sinai Hospital, Toronto, ON, Canada
    Search for more papers by this author
  • Anthony J. Hanley

    Corresponding author
    1. Department of Nutritional Sciences, University of Toronto, Toronto, ON, Canada
    2. Division of Endocrinology, Department of Medicine, University of Toronto, University of Toronto, Toronto, ON, Canada
    3. Dalla Lana School of Public Health, University of Toronto, Toronto, ON, Canada
    • Correspondence: Anthony Hanley, Department of Nutritional Sciences, University of Toronto, FitzGerald Building, 150 College Street, Room 341, Toronto, ON M5S 3E2, Canada. Tel.: 416 978 3616;Fax: 416 978 5882; E-mail: anthony.hanley@utoronto.ca

    Search for more papers by this author

Summary

Context

Vitamin D may play a role in the aetiology of the metabolic syndrome (MetS), yet the majority of previous studies have been cross-sectional, and the limited number of prospective studies has yielded inconsistent results.

Objective

To examine the prospective association of vitamin D [25-hydroxyvitamin D, 25(OH)D] with MetS in a multi-ethnic cohort of adults in Ontario, Canada.

Design

Nondiabetic individuals with pre-existing MetS risk factors were recruited for participation in the PROspective Metabolism and ISlet cell Evaluation (PROMISE) cohort study, a longitudinal study of the determinants of insulin resistance and MetS.

Methods

Of the 654 participants enrolled at baseline, 489 attended a 3-year follow-up visit. There were 301 participants eligible for the analysis of 25(OH)D with incident MetS (age 49·2 ± 9·3 years old, 75·4% female), after excluding 188 (38·5%) prevalent MetS cases at baseline. Longitudinal change in MetS components was assessed in the entire follow-up cohort.

Results

There were 76 (15·5%) participants who developed MetS over the 3-years of follow-up. Multivariate logistic regression analyses indicated a decreased risk of MetS at follow-up per standard deviation increase in baseline 25(OH)D after adjustment for sociodemographics, season, baseline and change in supplement use and physical activity and insulin resistance (OR = 0·63, 95% CI 0·44–0·90). Multivariate linear regression analyses revealed a significant inverse association of baseline 25(OH)D with fasting glucose at follow-up (β = −0·0005, P = 0·025).

Conclusions

There was a significant inverse association of baseline 25(OH)D with incident MetS, which may be partly driven by its association with glucose homoeostasis.

Introduction

Emerging evidence suggests that vitamin D [25-hydroxyvitamin D; 25(OH)D] may play a role in the aetiology of the metabolic syndrome (MetS). Although most studies to date have reported an inverse association of 25(OH)D and the MetS,[1-11] some studies have reported no association.[12-15] Notably, the majority of previous studies have been cross-sectional, with only four prospective studies published.[9-12] Three of these studies reported a significant inverse association of baseline 25(OH)D[10, 11] or vitamin D intake[9] with MetS at follow-up. However, Forouhi et al.[12] found no significant association of baseline 25(OH)D with MetS risk at the 10-year follow-up after multivariate adjustment. Therefore, given the limited prospective data available and inconsistent findings reported to date, the objective of the current study was to examine the association of baseline 25(OH)D with incident MetS after 3 years of follow-up in a well-characterized cohort at risk of the MetS.

Materials and methods

A detailed methodology of the PROspective Metabolism and ISlet cell Evaluation (PROMISE) cohort study has been published previously.[5, 16] Briefly, participants, aged 30 years and older, were recruited from Toronto and London, Ontario, Canada, between May 2004 and December 2006. Participants were recruited based on the presence of one or more risk factors for diabetes and the MetS including obesity, hypertension, family history of diabetes, history of gestational diabetes or birth of a macrosomic infant.[5] At baseline, 654 individuals without diabetes participated. Participants were contacted annually after the baseline visit to update contact information and collect data on major health events. Of the 654 participants at baseline, contact was maintained with n = 549 (84%), and 496 (76%) attended the 3-year follow-up clinic visits. Of the 496 participants who returned for follow-up, 489 (99%) had a baseline 25(OH)D measurement. Those who attended the follow-up examination were more likely to be older, female and Caucasian (P < 0·02) than those who did not attend, but there were no significant differences in body mass index (BMI), measures of insulin resistance or β-cell function or prevalent MetS (P ≥ 0·07).

Measures

As part of the baseline and 3-year clinic assessments, fasting blood samples were collected and 75-gram oral glucose tolerance tests were conducted. Blood samples were immediately processed for the determination of serum glucose and remaining samples were processed and frozen at minus 70 °C. Glucose was measured by standard laboratory procedures. Specific insulin was measured using the Elecsys 1010 immunoassay analyzer (Roche Diagnostics, Basel, Switzerland) and electrochemiluminescence immunoassay. This assay shows 0·05% cross-reactivity with intact human proinsulin and the Des 31, 32 circulating split form. Analyses for high-density lipoprotein cholesterol (HDL-C) and triglyercides were performed using Roche Modular's enzymatic colorimetric tests (Mississauga, ON, Canada).

Baseline 25(OH)D was measured in serum using the DiaSorin 25-OH Vitamin D TOTAL competitive chemiluminescence immunoassay on the automated LIAISON® analyzer (Stillwater, MN, USA). This assay measures both 25(OH)D2 and 25(OH)D3 equally, with a detection limit of 10 nm. Intra- and inter-assay coefficients of variation were 6·7%, 11·6%, respectively. In addition, the laboratory in which this assay was conducted participates in the International External Quality Assessment Scheme for Vitamin D Metabolites (DEQAS, Northwest Thames, UK), and it has been reported that the 25(OH)D results from this laboratory were consistently within one SD of the group mean in the international DEQAS proficiency surveys.[17]

Anthropometric measurements including height and weight were conducted for the calculation of BMI, with participants in light clothing and shoes removed. Waist circumference was measured at the natural waist, defined as the narrowest part of the torso, as viewed from behind, or the minimal circumference between the umbilicus and xiphoid process as viewed from the front. Blood pressure (BP) was measured in the right arm with the subject seated after five minutes resting using an automated sphygmomanometer. Each measure was determined twice using standardized procedures, with the average used in the analysis. Ethnicity and smoking were assessed using structured questionnaires. Physical activity was determined using a version of the Modifiable Activity Questionnaire (MAQ),[18] which collects information on both leisure and occupational activity over the past year (including measures of frequency and duration). The MAQ has been shown to have good reliability and validity.[18] Each reported activity from the MAQ is weighted by its relative intensity, referred to as a metabolic equivalent of task (MET), thereby deriving MET-hours per week (MET/h/week) as the final unit of expression. Season was defined using the date participants completed their baseline assessment and was categorized as follows: May-October (summer/early fall); November-April (winter/early spring). Supplement use, including multivitamins, vitamin D, or vitamin D + calcium, was obtained through an open-ended question on current medication use.

Metabolic syndrome

The harmonized definition of the MetS was used.[19] Metabolic syndrome was defined as present if the subject had at least three of the following criteria: elevated waist circumference (≥102 cm for men and ≥88 cm for women if of European origin; ≥90 cm for men and ≥80 cm for women if of non-European origin), elevated triglycerides (≥1·7 mm or drug treatment), reduced HDL-C (<1·0 mm for men and <1·3 mm in women, or drug treatment), elevated BP [systolic (SBP) ≥130 mmHg and/or diastolic (DBP) ≥85 mmHg, or drug treatment] and elevated fasting glucose (≥5·6 mm or drug treatment).

Statistical analysis

SAS Version 9.2 (Cary, NC, USA) was used for all analyses and a P-value <0·05 denoted statistical significance. Continuous variables were reported as mean ± SD or median with interquartile range in the case of skewed distributions, while categorical variables were reported as n (%). Natural logarithmic transformations were applied for all non-normally distributed variables. Student's t-tests and chi-square tests were used to examine differences at baseline for continuous and categorical variables respectively between incident MetS cases and those without MetS onset at follow-up. Multivariate logistic regression analyses were conducted to assess the association of baseline serum 25(OH)D with incident MetS at the 3-year follow-up, after excluding those with prevalent MetS at baseline. Odds ratios (OR) are presented to indicate the risk of incident MetS per SD increase in 25(OH)D. Multivariate linear regression analyses were then conducted in the entire follow-up cohort to investigate the association of baseline serum 25(OH)D with the individual MetS components at follow-up, analysed as continuous variables. Separate models were used for each outcome variable, which included adjustment for the baseline value of the outcome measure being assessed. Model 1 adjusted for sex, age, ethnicity and season of the 25(OH)D measurement; model 2 additionally adjusted for baseline and change in both vitamin D supplement use and physical activity. Model 3 for the logistic regression analyses included additional adjustment for HOMA-IR. For the linear regression analyses, model 3 included additional adjustment for baseline MetS status, and model 4 further adjusted for baseline and change in BMI. Sensitivity analyses were conducted to examine potential effect modifiers including prevalent MetS at baseline, ethnicity, season and BMI.

Results

Baseline participant characteristics for the entire follow-up PROMISE cohort are presented in Table 1. Participants in the PROMISE cohort were on average 50 years of age, with a high proportion of females and Caucasians. In addition, the mean serum 25(OH)D concentration at baseline was 58·0 ± 23·3 nm. As previously reported,[16] based on the most recent Institute of Medicine (IOM) recommendations for vitamin D, approximately 11%, 22% and 63% of the PROMISE cohort had deficient (<30 nm), insufficient (31–39 nm) and sufficient (>50 nm) 25(OH)D levels, respectively. Based on The Endocrine Society recommendations,[20] 37%, 32% and 21% had deficient (<50 nm), insufficient (52·5–72·5 nm) and sufficient (≥75 nm) 25(OH)D levels. After excluding those with prevalent MetS at baseline (n = 188, 38·5%), there were 301 participants remaining at risk of MetS, and therefore eligible for the incidence analysis, of which 76 (15·5%) developed MetS at follow-up. Table 1 also presents baseline characteristics stratified by incident MetS status at follow-up. Overall, those with incident MetS were significantly older, had higher BMI and waist circumference, higher BP and triglycerides and lower HDL-C compared with those without MetS onset. In addition, those with incident MetS at follow-up had significantly lower serum 25(OH)D concentrations compared with those without MetS onset.

Table 1. Baseline characteristics of the participants for the entire follow-up cohort and those with no MetS at baseline
CharacteristicFollow-up PROMISE cohort (n = 489)No MetS at baseline (n = 301)P-value
Incident MetS at follow-up (n = 76)No MetS onset at follow-up (n = 225)
  1. Data are n (%) for categorical variables, mean ± SD for continuous variables, or median (25% and 75% interquartiles) for non-normally distributed variables. P-values are tests for proportions for categorical variables or tests for equality for continuous variables to assess whether baseline characteristics differ between those who did and did not develop MetS.

Age50·19 ± 9·6751·74 ± 8·5448·31 ± 9·410·0053
Sex (% females)357 (73·01)57 (75·00)170 (75·56)0·92
Ethnicity
Caucasian347 (70·96)53 (69·74)156 (69·33)0·67
Hispanic58 (11·86)7 (9·21)30 (13·33)
South Asian33 (6·75)7 (9·21)14 (6·22)
Other51 (10·43)9 (11·84)25 (11·11)
Anthropometry
BMI (kg/m2)30·33 (26·72, 34·57)31·42 ± 5·6928·48 ± 5·31<0·0001
Waist circumference (cm)98·43 ± 15·4399·00 ± 13·4591·08 ± 13·15<0·0001
Physical activity (MET-h/week)19·59 (7·39, 53·52)20·41 (11·45, 69·26)24·23 (10·01, 57·35)0·88
Smoking (% current)30 (6·29)4 (5·33)15 (6·91)0·85
Vitamin D Supplement Use (% yes)212 (43·35)30 (39·47)110 (48·89)0·15
25(OH)D (nm)58·0 ± 23·355·0 ± 20·361·7 ± 22·50·022
Blood pressure
SBP (mm Hg)125·92 ± 16·03129·1 ± 18·08120·7 ± 15·23<0·0001
DBP (mm Hg)80·16 ± 10·3281·49 ± 10·9477·72 ± 10·370·0073
Triglycerides (mm)1·32 (0·94, 1·86)1·36 (1·08, 1·58)1·07 (0·78, 1·30)<0·0001
HDL-C (mm)1·30 (1·1, 1·6)1·42 (1·20, 1·60)1·53 (1·30, 1·70)0·0060
Fasting glucose (mm)4·9 (4·6, 5·2)5·00 (4·65, 5·30)4·70 (4·50, 5·00)<0·0001
HOMA-IR1·88 (1·19, 3·09)1·99 (1·32, 2·87)1·33 (0·85, 1·98)<0·0001

Multivariate logistic regression analyses of incident MetS per SD increase in baseline serum 25(OH)D are presented in Table 2. Overall, there was a significant inverse association of baseline 25(OH)D with MetS at follow-up after adjustment for age, sex, season, ethnicity, and baseline and change in both supplement use and physical activity (Model 2: OR 0·62 per SD, 95% CI 0·44–0·88). Additional adjustment for HOMA-IR resulted in essentially similar findings (Model 3). Findings were also similar in a sensitivity analysis investigating the association of 25(OH)D with incident MetS, defined excluding the glucose component (data not shown). Additional sensitivity analyses, which included adjustment for PTH, or baseline and change in beta-cell function, resulted in essentially similar findings (data not shown).

Table 2. Multivariate logistic regression analysis of associations of baseline 25(OH)D with incident MetS at follow-up (n = 301)
25(OH)DOR (95% CI)P-value
  1. Model 1: Adjusted for age, sex, ethnicity and season.

  2. Model 2: Adjusted as in Model 1 plus baseline supplement use and change in supplement use, baseline physical activity, change in physical activity.

  3. Model 3: Adjusted as in Model 2 plus HOMA-IR.

Outcome per SD increase in 25(OH)D
Model 10·62 (0·45, 0·85)0·0031
Model 20·62 (0·44, 0·88)0·0065
Model 30·63 (0·44, 0·90)0·011

Multivariate linear regression analyses were then conducted to assess the association of baseline 25(OH)D with MetS components at follow-up in the entire cohort (n = 489; Table 3) which included adjustment for the baseline value of each outcome variable. Overall, baseline 25(OH)D was not significantly associated with waist circumference, triglycerides, HDL, SBP or DBP at follow-up. However, there was a significant inverse association of baseline 25(OH)D with fasting glucose at follow-up (β = −0·0005, P = 0·025) after adjustment for sociodemographics, season of 25(OH)D measurement, baseline fasting glucose, baseline and change in physical activity, supplement use as well as BMI and adjustment for baseline MetS status. In addition, tests for potential effect modifiers including prevalent MetS at baseline, ethnicity, season and BMI were not significant.

Table 3. Multiple linear regression analysis of associations of baseline 25(OH)D (nm) with components of the MetS at the 3-year follow-up, n = 489
Outcome per unit increase in baseline 25(OH)DModel 1Model 2Model 3Model 4
β (95% CI)P-valueβ (95% CI)P-valueβ (95% CI)P-valueβ (95% CI)P-value
  1. Model 1: adjusted for age, sex, ethnicity, season of 25D measurement and baseline outcome variable.

  2. Model 2: adjusted as in Model 1 plus baseline physical activity, change in physical activity, baseline supplement use, change in supplement use.

  3. Model 3: adjusted as in Model 2 plus baseline MetS status.

  4. Model 4: adjusted as in Model 3 plus baseline body mass index (BMI) and change in BMI.

  5. a

    Log transformations.

Waist circumference−0·001 (−0·03, 0·03)0·940·001 (−0·03, 0·03)0·940·001 (−0·03, 0·03)0·94  
Triglycerides (mm)a−0·0004 (−0·002, 0·001)0·56−0·0004 (−0·002, 0·001)0·61−0·0004 (−0·002, 0·001)0·57−0·001 (−0·002, 0·001)0·20
HDL (mm)a0·0001(−0·0006, 0·0007)0·880·00002 (−0·0007, 0·0007)0·96−0·0001 (−0·0007, 0·0006)0·87−0·0001 (−0·0008, 0·0005)0·70
SBP (mmHg)a0·005 (−0·05, 0·05)0·850·010 (−0·04, 0·06)0·700·010 (−0·04, 0·06)0·700·012 (−0·04, 0·07)0·67
DBP (mmHg)a−0·010 (−0·04, 0·02)0·57−0·008 (−0·04, 0·03)0·68−0·006 (−0·04, 0·03)0·76−0·006 (−0·04, 0·03)0·74
Fasting glucose (mm)a−0·0005 (−0·001, −0·0001)0·021−0·0006 (−0·001, −0·0001)0·0086−0·0005 (−0·001, −0·0001)0·012−0·0005 (−0·0009, −0·0001)0·025

Discussion

This study found a 37% reduced risk of incident MetS per SD increase in baseline 25(OH)D after 3-years of follow-up. There was also a significant inverse association of baseline 25(OH)D with fasting glucose levels at follow-up.

The majority of previous studies examining the association of 25(OH)D with the MetS have been cross-sectional, with most reporting a significant inverse association of 25(OH)D with MetS prevalence,[1-4, 6-8] including our study in the PROMISE cohort.[5] However, some studies have also reported no association.[13-15] Notably, there have been only four studies conducted to date examining the association of vitamin D on incident MetS.[9-12] Dietary and supplemental vitamin D intake was inversely related to incident MetS after 20 years of follow-up in a recent study,[9] but this study did not measure serum 25(OH)D concentration. Forouhi et al.[12] found that higher baseline 25(OH)D levels were significantly associated with lower MetS risk z-scores at the 10-year follow-up after adjustment for age, sex, smoking, season, BMI and baseline MetS z-score (P = 0·048).[12] But this association was attenuated to nonsignificance with further multivariate adjustment. It is important to recognize that Forouhi et al.[12] used a continuous MetS risk z-score, whereas our study used a dichotomous variable for MetS based on published harmonized criteria,[19] which may explain differences between the study findings. In contrast, Skaaby et al.[11] reported a significant inverse association of 25(OH)D and incident MetS (OR = 0·95, P < 0·05) after multivariate adjustment. In addition, in a population-based cohort study, Gagnon et al.[10] reported that for each 25 nm decrease in 25(OH)D, the risk of the MetS at the 5-year follow-up increased by 23% after multivariate adjustment. However, the association between 25(OH)D and MetS risk was attenuated to nonsignificance after further adjustment for HOMA-IR,[10] which suggests that insulin resistance may be mediating the association of 25(OH)D with MetS risk. The results of the current study extend these observations to individuals with pre-existing metabolic risk factors. Moreover, and in contrast to the results of Gagnon et al.,[10] the findings in this study indicated a significant inverse association of 25(OH)D with incident MetS, even after adjustment for HOMA-IR. It is conceivable that differences in insulin resistance or its determinants in the current study population may explain these contrasting observations. Further, a sensitivity analysis of the association of 25(OH)D with incident MetS, defined using nonglucose components, yielded results that were largely consistent with the main analysis, suggesting that the association is not entirely explained by the relationship of 25(OH)D with dysglycaemia. Nevertheless, low baseline 25(OH)D concentrations were significantly associated with greater fasting glucose levels at follow-up, which suggests that changes in glucose levels may be a major (although not the sole) determinant of the link between 25(OH)D and incident MetS. Previous studies have also found a significant inverse association of 25(OH)D with glucose homoeostasis and diabetes risk,[12, 21, 22] including our study in the PROMISE cohort at baseline,[23] which supports the findings in the current study. In addition to the demonstrated association with fasting glucose levels, unmeasured factors including inflammation and lifestyle variables may also modulate the relationship between baseline 25(OH)D and incident MetS.

We failed to find a significant association of baseline 25(OH)D with the other MetS components at follow-up. Although previous cross-sectional and prospective studies have reported inverse associations of 25(OH)D with several of the MetS components, results have not been consistent.[5, 6, 9-11, 24, 25] Further, a number of randomized controlled trials have investigated the effects of vitamin D supplementation on various metabolic syndrome components,[26-30] although results have been inconsistent, possibly due to inadequate statistical power, dose and/or the fact that these trials were not designed for specific MetS components as outcomes. Additional research in this area is therefore warranted.

Strengths of this study include its relatively low attrition rate, as well as its prospective design which allows for the examination of a temporal association of 25(OH)D and MetS risk, although the duration of follow-up was short. While we did adjust for several potential confounders, the observational design of this study does not precludes residual confounding. In addition, certain factors including the high proportion of women in this cohort and the high latitude study setting may limit the generalizability of the study findings. Further, the potential for bias cannot be ruled out given that 76% attended the follow-up clinic visit and these individuals were more likely to be older, female and Caucasian compared with those who did not return. However, there were no significant differences in BMI or prevalent MetS between those who returned vs those who did not return for the follow-up clinic visit.

In conclusion, this study supports a potential role for low 25(OH)D in the aetiology of the MetS, which may be partly driven by the association of 25(OH)D with glucose homoeostasis. Additional studies are needed to confirm these findings.

Acknowledgements

We would like to thank our study subjects for their participation. We also wish to thank Jan Neuman, Paula Van Nostrand, Stella Kink, Sheila Porter, Mauricio Marin and Annette Barnie for their dedication and expert technical assistance.

Competing interests/financial disclosure

Nothing to declare.

Ancillary