Vitamin D receptors are ubiquitous in the body( Reference Holick 1 ), indicating widespread effects of vitamin D in health and disease. Adverse effects of vitamin D deficiency, defined by low circulating concentrations of 25-hydroxyvitamin D (25(OH)D), on bone health have long been known( Reference Cranney, Weiler and O’Donnell 2 ). More recently, potential associations of vitamin D with other chronic diseases, including cancer, CVD, autoimmune disease and type 2 diabetes (T2D), have been demonstrated( Reference Holick 3 ). For instance, prospective cohort studies have found associations between baseline low circulating 25(OH)D concentrations and incident myocardial infarction, stroke and all-cause mortality( Reference Giovannucci, Liu and Hollis 4 – Reference Wang, Pencina and Booth 6 ). However, findings from randomized clinical trials of vitamin D supplementation to reduce hypertension and cardiovascular mortality were inconsistent( Reference Judd and Tangpricha 7 ), suggesting more complex relationships between vitamin D and cardiovascular risk development.
A clear understanding of relationships between vitamin D and known cardiometabolic risk factors (CMR), such as fasting blood glucose and related parameters, blood pressure (BP) and lipid levels, as well as the role of potential confounders or mediators, such as obesity, is lacking, particularly among young adults free from overt disease. Further, sex-specific differences have not been fully explored, although it is recognized that men tend to have higher 25(OH)D levels than women( Reference Fiscella and Franks 8 ) while having an earlier age of onset of CMR development( Reference van Lennep, Westerveld and Erkelens 9 ).
We examined cross-sectional relationships between circulating 25(OH)D and CMR among young adult participants of a longitudinal birth cohort study, the Jerusalem Perinatal Study (JPS). We also evaluated whether these relationships differ by sex or are confounded by current BMI.
Methods
Overall setting and study population
The JPS was conducted among women who gave birth as residents of Jerusalem between 1974 and 1976 and who were interviewed within 2 d postpartum (n 17 003). The JPS Family Follow-Up study (JPS-1), conducted between 2007 and 2009, was a study designed to determine maternal and offspring genetic risk factors that account for associations between maternal obesity and adulthood CMR among the offspring. The JPS-1 study population consisted of mother–offspring dyads identified and recruited using a stratified sampling of term, singleton live births from the original JPS cohort. Participants were oversampled for low (≤2500 g) and high (≥4000 g) birth weight and for maternal obesity (pre-pregnancy BMI ≥27 kg/m2). Mother–offspring dyads with offspring who had congenital disorders at birth were excluded.
The current study was conducted among offspring JPS-1 participants, 32 (sd 1·3) years old on average. Individuals who self-reported taking medication to lower BP (n 13), lipids (n 14) or to treat diabetes (n 5) were excluded from the analyses. Individuals with a fasting time before blood draw of less than 8 h (n 97), fasting glucose outside the range 60–126 mg/dl (n 135) or TAG higher than 600 mg/dl (n 58) were excluded. The final analytic study population consisted of 1204 offspring (609 males and 595 females). Study procedures were approved by the Institutional Review Boards of the University of Washington, in Seattle, and the Hebrew University Hadassah Medical Center in Jerusalem. All participants provided informed consent.
Data collection
Data collection procedures have been published previously( Reference Hochner, Friedlander and Calderon-Margalit 10 ). Briefly, data on anthropomorphic measurements, lifestyle and sociodemographic characteristics were collected during offspring interviews conducted between 2007 and 2009. BMI was calculated as weight (kg) divided by squared height (m2). BP was measured as the average of three consecutive measurements performed after sitting for 5 min (Omron M7 automated sphygmomanometer). Fasting blood samples were taken using standard procedures. Plasma glucose, HDL-cholesterol (HDL-C) and TAG were measured on the VITROS 5,1FS Chemistry System (Ortho Clinical Diagnostics, Raritan, NJ, USA). Plasma insulin levels were determined using RIA with the Human Insulin-Specific RIA Kit (Millipore, Billerica, MA, USA). Plasma vitamin D (25(OH)D) was measured using LC–tandem MS at the Nutrition and Obesity Research Center, University of Washington (Seattle, WA, USA). The sum of 25(OH)D2 and 25(OH)D3 was used in analyses to represent total plasma vitamin D levels.
Statistical analyses
We examined participant characteristics in the entire study cohort and stratified by sex. Means and standard deviations were calculated for continuous variables; numbers and percentages for categorical variables. We used weighted multiple linear regression models to examine associations between offspring 25(OH)D and CMR (BMI, waist circumference, waist:hip ratio, fasting glucose, fasting insulin, homeostatic model assessment of insulin resistance (HOMA-IR), lipids and BP). Ninety-five per cent confidence intervals and P values were calculated using robust standard errors. BMI, TAG, LDL-cholesterol (LDL-C) and HDL-C were natural log-transformed to improve model fit. We first examined the relationships among the entire study population with sex as a covariate and an adjustment for multiple testing (critical P value α=0·0045). Then, we used an interaction term between sex and 25(OH)D to evaluate statistical interaction (at the P=0·05 level) between vitamin D and sex on CMR.
We fit two models to evaluate the associations. Model 1 included season of blood draw (December–February, March–May, June–August, September–November) and additional covariates for smoking status (current, former and never), alcohol abstinence (yes/no), intense physical activity participation (yes/no), employment status (yes/no), religiosity (ultra-orthodox, religious and traditionalist/secular/other), ethnic origin (Israel, Morocco, Other Africa, Iraq, Iran, Kurdistan, Yemen, Other Asia and Ashkenazi) and parental history of relevant disease (see below) as confounders and/or to increase precision. The models for fasting blood glucose, insulin and HOMA-IR were adjusted for parental history of diabetes. The models for LDL-C, HDL-C and TAG were adjusted for parental history of heart disease, stroke or elevated cholesterol. The models for systolic and diastolic BP were adjusted for parental history of high BP and stroke. In Model 2, we adjusted for Model 1 variables and BMI, except for the model evaluating BMI as an outcome. The aim of Model 2 was to assess the extent to which BMI mediates or confounds observed associations.
Analyses were performed using the statistical software package Stata v11·0.
Results
Selected characteristics of the study participants are shown in Table 1. Mean total 25(OH)D level was 21·7, 22·9 and 20·5 mmol/l among all participants, males and females, respectively. Overall, 44 % of participants, 36 % of males and 52 % of females, had vitamin D deficiency based on the cut-off of 20 ng/ml( Reference Holick, Binkley and Bischoff-Ferrari 11 ). Males had higher BMI, fasting blood glucose, TAG, LDL-C, systolic BP, diastolic BP and HOMA-IR compared with females. They also had lower HDL-C levels.
Table 1 Characteristics of the study participants by sex and quartile of plasma vitamin D; offspring (n 1204) in the Jerusalem Perinatal Study (JPS; 1974–1976), interviewed and examined at age 32 years in the JPS Family Follow-Up study (JPS-1; 2007–2009)
25(OH)D, 25-hydroxyvitamin D; LDL-C, LDL-cholesterol; HDL-C, HDL-cholesterol; BP, blood pressure; HOMA-IR, homeostatic model assessment of insulin resistance.
† May not sum to total due to missing data.
‡ Defined as exercise that brings about laboured breathing, increased pulse rate and sweating, lasting for at least 20 min.
Among all study participants (Table 2), the association between 25(OH)D with TAG was nearly statistically significant in Model 1 after adjustment for other confounders (P>0·0045 after multiple testing correction). Each 1 mmol/l increase in 25(OH)D was associated with a 1 % decrease in TAG levels (P=0·005). However, the association was attenuated and became non-significant after adjustment for current BMI (P=0·041). Among females (Table 3), 25(OH)D was not significantly associated with any outcome of interest. Among males (Table 4), 25(OH)D was inversely associated with insulin and HOMA-IR (P=0·003 and 0·002, respectively). Associations of 25(OH)D with HOMA-IR persisted even after adjustment for current BMI (P-value =0·004). Significant statistical interactions were observed between 25(OH)D and sex on insulin and HOMA-IR (P=0·015 and 0·012, respectively). We did not observe associations between 25(OH)D and fasting glucose level, LDL-C or BP overall, or among either sex.
Table 2 Overall associations between plasma 25(OH)D (mmol/l) and cardiometabolic risk factors in young adults; offspring (n 1204) in the Jerusalem Perinatal Study (JPS; 1974–1976), interviewed and examined at age 32 years in the JPS Family Follow-Up study (JPS-1; 2007–2009)
25(OH)D, 25-hydroxyvitamin D; LDL-C, LDL-cholesterol; HDL-C, HDL-cholesterol; BP, blood pressure; HOMA-IR, homeostatic model assessment of insulin resistance.
Model 1 adjusted for season of blood draw, sex, smoking status, alcohol abstinence, religious orthodoxy, physical activity, employment status, ethnic origin and parental history of disease.
Model 2 adjusted for Model 2 covariates, plus BMI.
*P value for 25(OH)D–sex interaction less than 0·05.
Table 3 Associations between plasma 25(OH)D (mmol/l) and cardiometabolic risk factors among adult females; offspring (n 595) in the Jerusalem Perinatal Study (JPS; 1974–1976), interviewed and examined at age 32 years in the JPS Family Follow-Up study (JPS-1; 2007–2009)
25(OH)D, 25-hydroxyvitamin D; LDL-C, LDL-cholesterol; HDL-C, HDL-cholesterol; BP, blood pressure; HOMA-IR, homeostatic model assessment of insulin resistance.
Model 1 adjusted for season of blood draw, smoking status, alcohol abstinance, religious orthodoxy, physical activity, employment status, ethnic origin and parental history of disease.
Model 2 adjusted for Model 1 covariates, plus BMI.
Table 4 Associations between plasma 25(OH)D (mmol/l) and cardiometabolic risk factors among adult males; offspring (n 609) in the Jerusalem Perinatal Study (JPS; 1974–1976), interviewed and examined at age 32 years in the JPS Family Follow-Up study (JPS-1; 2007–2009)
25(OH)D, 25-hydroxyvitamin D; LDL-C, LDL-cholesterol; HDL-C, HDL-cholesterol; BP, blood pressure; HOMA-IR, homeostatic model assessment of insulin resistance.
Model 1 adjusted for season of blood draw, smoking status, alcohol abstinance, religious orthodoxy, physical activity, employment status, ethnic origin and parental history of disease.
Model 2 adjusted for Model 1 covariates, plus BMI.
*P value less than 0·0045 (multiple-testing corrected).
Discussion
In the current study, we found that associations of total plasma 25(OH)D with fasting insulin concentration and HOMA-IR were present among males but not females. Interactions between sex and vitamin D on insulin and HOMA-IR were statistically significant. Among males, the association between fasting insulin concentration and 25(OH)D was attenuated after adjustment for BMI, such that it no longer met our stringent criteria for statistical significance.
The inverse association between 25(OH)D and TAG has been widely reported, although many studies report significant associations after adjustment for measures of adiposity, such as BMI( Reference Jorde and Grimnes 12 ). In one such study by Kayaniyil et al., the relationship was attenuated with the addition of BMI to the model but remained statistically significant in their slightly older, more obese and sicker population( Reference Kayaniyil, Vieth and Harris 13 ). It is therefore possible that the relationship between 25(OH)D and TAG is stronger and less subject to attenuation by adiposity among individuals with diagnosed metabolic syndrome, potentially explaining the differences between our study and that by Kayaniyil et al.
Evidence for direct effects of vitamin D on TAG levels is still conflicting( Reference Jorde and Grimnes 12 ). Only one randomized controlled trial has reported a significant decrease in TAG levels after vitamin D supplementation( Reference Zittermann, Frisch and Berthold 14 ), while four other randomized controlled trials have failed to duplicate these findings( Reference Jorde, Sneve and Torjesen 15 – Reference Nagpal, Pande and Bhartia 18 ), including one with both a larger sample size and a larger dose of vitamin D( Reference Jorde, Sneve and Torjesen 15 ). While indirect effects via vitamin D’s influence on intestinal Ca absorption are not thought to cause substantial changes in serum TAG( Reference Reid, Mason and Horne 19 ), it is plausible that vitamin D reduces serum TAG by suppressing pituitary hormone levels( Reference Zemel, Shi and Greer 20 ) or through its posited effects on glucose metabolism( Reference Kamycheva, Jorde and Figenschau 21 ), but these hypotheses require further study.
Our analysis of 25(OH)D and concurrently measured fasting insulin in the JPS offspring population found an inverse association between vitamin D and fasting insulin in males, but not females. To confirm these findings in less constrained models, we also examined models using splines and categories of vitamin D insufficiency and insulin resistance (results not shown), but these models did not add anything to our report. Two other studies have examined these sex-specific associations( Reference Yin, Sun and Zhang 22 , Reference Delvin, Lambert and Levy 23 ) and reported associations in females as well as in males. The study by Delvin et al. examined French-Canadian adolescents who were 9, 13 and 16 years old during the study( Reference Delvin, Lambert and Levy 23 ). Due to the age difference, results from that study of adolescents and our study on JPS adults, approximately 32 years old, may not be directly comparable. In their study, Yin et al. excluded individuals who used vitamin D supplements, smoked or engaged in strenuous exercise, besides individuals excluded for other reasons. Furthermore, only 25 % of their study participants were female( Reference Yin, Sun and Zhang 22 ). While we adjusted for most of these factors in our study, study population differences, such as diet or genetics, may account for differences in findings from these studies compared with the current study. Other large, population-based studies have also found inverse associations between measures of vitamin D and fasting insulin( Reference Husemoen, Thuesen and Fenger 24 – Reference Liu, Meigs and Pittas 27 ) in both sexes; however, these studies did not evaluate sex-specific relationships.
We examined a vitamin D–sex interaction due to the literature supporting sex differences in vitamin D levels( Reference Fiscella and Franks 8 ) and cardiometabolic outcomes( Reference van Lennep, Westerveld and Erkelens 9 ). Despite differences between the results of the current analysis and previous studies( Reference Yin, Sun and Zhang 22 , Reference Delvin, Lambert and Levy 23 , Reference Del Gobbo, Song and Dannenbaum 28 ), some potential reasons supporting observed associations of 25(OH)D levels with fasting insulin and HOMA-IR among males, but not females, exist. Among non-diabetics, insulin resistance is higher among males than females( Reference Geer and Shen 29 ). Middle-aged males have been shown to have a higher risk of incident T2D compared with middle-aged females( Reference Choi, Kim and Kim 25 , Reference Lipscombe and Hux 30 , 31 ). Therefore, if indeed low vitamin D levels contribute to the pathogenesis of T2D, our analysis of 32-year-olds might be expected to show a stronger relationship in males than females, who are not expected to develop disease until later.
A prospective cohort study among non-diabetic participants found that lower concentration of baseline 25(OH)D is associated with higher fasting insulin and HOMA-IR, along with other indicators of impaired glucose metabolism( Reference Forouhi, Luan and Cooper 32 ). However, they were unable to study incident T2D due to a small number of cases. Conversely, a study by Mattila et al. reported that the association of 25(OH)D concentration with incident T2D was attenuated by the addition of concurrent BMI into the model, although the trend of increasing T2D diagnosis with decreasing quartile of serum 25(OH)D remained( Reference Mattila, Knekt and Mannisto 33 ). Unfortunately, they did not examine markers of insulin resistance and so these two studies are not directly comparable with ours. A randomized controlled trial of vitamin D supplementation found that supplemented, obese, non-diabetic African-American adolescents had improved fasting insulin and HOMA-IR compared with placebo after 6 months( Reference Belenchia, Tosh and Hillman 34 ), indicating that statistically significant improvements may take some time to be observable even in high-risk populations.
At the population level, most cross-sectional studies of 25(OH)D and insulin sensitivity have found associations between the two( Reference Pittas, Lau and Hu 35 ). However, other cross-sectional studies have found that insulin sensitivity is independent of 25(OH)D in specific populations, such as African Americans( Reference Scragg, Sowers and Bell 36 , Reference Alemzadeh, Kichler and Babar 37 ) and the morbidly obese( Reference Manco, Calvani and Nanni 38 ). Beyond the associations between vitamin D and adiposity, evidence from basic science research points to mechanisms for potential associations between vitamin D and glucose homeostasis at the cellular level. Vitamin D receptors( Reference Johnson, Grande and Roche 39 ) and 1α-hydroxylase( Reference Bland, Markovic and Hills 40 ) are expressed on insulin-producing pancreatic β cells. The promoter of the human insulin gene has a vitamin D response element( Reference Maestro, Dávila and Carranza 41 ), while 1,25-dihydroxyvitamin D activates transcription of that gene( Reference Maestro, Molero and Bajo 42 ). Evidence has also been found in animal studies. A study in mice found that lack of a functional vitamin D response impaired insulin response( Reference Zeitz, Weber and Soegiarto 43 ). In an in vivo rat study, vitamin D deficiency impaired glucose-mediated insulin secretion, which was restored by supplementation( Reference Cade and Norman 44 ).
Intervention studies of vitamin D supplementation have not found an association between supplementation and hepatic insulin sensitivity nor with direct measures of insulin sensitivity( Reference Pittas, Lau and Hu 35 ). Some of these inconsistencies may be related to sex-specific or population-level differences in the relationships.
The major strength of the present study is the data on lifestyle characteristics, demographics and CMR in young adult offspring. That our analyses specifically examine sex-specific differences as well as the role of obesity in these relationships are additional strengths. By examining these relationships in a young, healthy population, we have fewer confounding factors and co-morbidities to consider. However, limitations of our study deserve mention. Levels of 25(OH)D and CMR were measured cross-sectionally and so the temporal relationship of observed associations remains unclear. We tested associations with eleven different CMR outcomes, leading to a substantial correction for multiple testing. Finally, specific characteristics of the study population (e.g. lifestyle and dietary characteristics) may limit generalizability of our findings. Similarities of our findings, particularly overall findings, with previous reports mitigate this concern.
Conclusion
In conclusion, we report that an inverse association of 25(OH)D with TAG is attenuated after adjustment by current BMI in young adults. We also report that 25(OH)D is associated with insulin sensitivity markers in young adult males, but not females. Inconsistencies that are observed, particularly in relation to sex-specific differences and the absence of similar associations in intervention studies, indicate that the relationship between vitamin D and insulin sensitivity may be more complex than anticipated. It would be interesting to conduct follow-up studies in this and similar populations as the participants age. Better understanding of these relationships may help direct future research for public health or clinical applications. Finally, well-designed randomized controlled trials of vitamin D supplementation and CMR are needed in well-defined at-risk populations to determine the utility of vitamin D supplementation in preventing cardiometabolic disease.
Acknowledgements
Financial support: This work was supported by grants R01HL088884 (Principal Investigator: D.S.S.) and K01HL103174 (Principal Investigator: D.A.E.) from the National Heart, Lung, and Blood Institute of the National Institutes of Health (NIH). The work was partially supported by the Nutrition and Obesity Research Center at the University of Washington (grant number P30 DK035816). A.M. received support from the National Institutes of Child Health and Development of the NIH (grant number T32 HD052462). The funding bodies had no role in the design, analysis or writing of this article. Conflicts of interest: None. Authorship: A.M. designed the research, analysed the data, wrote the paper and had primary responsibility for the final content. H.H. conducted the research and contributed to the paper. C.M.S. analysed the data and contributed to the paper. M.A.W. contributed to the paper. A.N.H. contributed to the paper. I.H.d.B. contributed to the paper. B.K. contributed to the paper. D.S.S. designed the research and contributed to the paper. Y.F. designed the research and contributed to the paper. D.A.E. designed the research and contributed to the paper. All authors read and approved the final manuscript. Ethics of human subject participation: Study procedures were approved by the Institutional Review Boards of the University of Washington, in Seattle, and the Hebrew University Hadassah Medical Center in Jerusalem. All participants provided informed consent.
