Associations Between Vitamin D Status, Adiposity, and Inflammatory Biomarkers in Young Women (18 – 30 Years)

University of Massachusetts Amherst ScholarWorks@UMass Amherst

Doctoral Dissertations Dissertations and Theses

Summer November 2014

Associations Between Vitamin D Status, Adiposity, And Inflammatory Biomarkers In Young Women (18 – 30 Years)

Adolphina Addo-Lartey University of Massachusetts Amherst

Follow this and additional works at: https://scholarworks.umass.edu/dissertations_2

Part of the Women's Health Commons

Recommended Citation Addo-Lartey, Adolphina, "Associations Between Vitamin D Status, Adiposity, And Inflammatory Biomarkers In Young Women (18 – 30 Years)" (2014). Doctoral Dissertations. 150. https://doi.org/10.7275/6054079.0 https://scholarworks.umass.edu/dissertations_2/150

This Open Access Dissertation is brought to you for free and open access by the Dissertations and Theses at ScholarWorks@UMass Amherst. It has been accepted for inclusion in Doctoral Dissertations by an authorized administrator of ScholarWorks@UMass Amherst. For more information, please contact [email protected].

ASSOCIATIONS BETWEEN VITAMIN D STATUS, ADIPOSITY, AND

INFLAMMATORY BIOMARKERS IN YOUNG WOMEN (18 – 30 YEARS)

A Dissertation Presented

ADOLPHINA A. ADDO-LARTEY

Submitted to the Graduate School of the

University of Massachusetts Amherst in partial fulfillment

of the requirements for the degree of

DOCTOR OF PHILOSOPHY

SEPTEMBER 2014

Public Health Nutrition

ASSOCIATIONS BETWEEN VITAMIN D STATUS, ADIPOSITY, AND

INFLAMMATORY BIOMARKERS IN YOUNG WOMEN (18 – 30 YEARS)

A Dissertation Presented

ADOLPHINA A. ADDO-LARTEY

Approved as to style and content by:

______Dr. Alayne Ronnenberg, Committee Chair

______Dr. Elizabeth Bertone-Johnson, Committee Member

______Dr. Richard Wood, Committee Member

______Dr. Carol Bigelow, Committee Member

______Dr. Nancy L. Cohen, Head Department of Nutrition

DEDICATION

To God be the Glory,

Great Things He hath done…

Greater Things He will do.

ACKNOWLEDGEMENTS

I am grateful to my advisor and committee chair, Dr. Alayne Ronnenberg, for her guidance and encouragement as I gathered research ideas for my dissertation. Her immense input during our progress/review meetings helped shaped my research questions and writing of manuscripts. I am thankful to Dr. Elizabeth Bertone Johnson and Dr. Alayne Ronnenberg, Principal Investigators of the UMass Amherst Vitamin D

Status Study, for making the study data available for my use. I would also like to thank all members of my committee; Dr. Elizabeth Bertone-Johnson, Dr. Richard

Wood, and Dr. Carol Bigelow, for sharing their time and expertise over the course of this project.

My sincere appreciation goes to all the women who volunteered for the UMass

Amherst Vitamin D Status Study; as well as Dr. Sofija Zagarins, Biki Takashima-

Uebelbör, and Joycelyn Faraj for their assistance with data collection. Last but not least, I would like to express my heartfelt gratitude to my amazing husband (Dr.

Michael Lartey), as well as all my family and friends for their constant encouragement and support. Without you to cheer me on, this journey would have seemed even more arduous. I am so very thankful for all of you, God richly bless you.

ABSTRACT

ASSOCIATIONS BETWEEN VITAMIN D STATUS, ADIPOSITY, AND

INFLAMMATORY BIOMARKERS IN YOUNG WOMEN (18 – 30 YEARS)

SEPTEMBER 2014

ADOLPHINA A. ADDO-LARTEY, B.Sc., UNIVERSITY OF GHANA, LEGON

Ph.D., UNIVERSITY OF MASSACHUSETTS AMHERST

Supervised by: Dr. Alayne Ronnenberg

Calcium and vitamin D intakes have been linked with reduced adiposity in older adults and in obese women (body mass index, BMI, ≥ 30 kg/m2). However, very little is known about the extent of these associations in apparently healthy, non-obese, young women. Emerging research also suggest that vitamin D insufficiency (serum

25-hydroxyvitamin D, 25-OHD, < 75 nmol/L) and increased adiposity may be associated with higher levels of inflammatory biomarkers in obese/older adults, and chronically ill patients. We conducted a cross-sectional analysis among 270 (18- to

30-year old) female participants in the UMass Amherst Vitamin D Status Study (n =

270) to assess the extent to which dietary intakes of calcium and vitamin D are associated with obesity markers. We also evaluated the association between serum 25-

OHD concentrations and both adiposity and inflammatory biomarkers. Study participants were mostly Caucasians (84.5%) with normal BMI, although about half of women had high adiposity (total body fat ‘TBF’≥ 32%). Women reporting adequate intakes of calcium (≥ 1000 mg/day) but low intakes of vitamin D (< 600

IU/day) were more than twice as likely to have a high percentage of TBF compared to women with adequate intakes of both calcium and vitamin D. In addition, women vi

with lower calcium intake from supplements were twice as likely to have a waist circumference ≥ 80 cm (OR = 2.04; 95% CI: 1.04 – 3.99) compared to women in the highest tertile of calcium intake. The magnitude of this association is important since among young women 18-30 years old, a waist circumference greater than 80 cm indicates central obesity and suggests increased visceral adiposity, which contributes to hyperlipidemia and other obesity-related chronic conditions. Among all women, total vitamin D, food vitamin D, and supplemental vitamin D intake were not associated with serum 25-OHD concentration (P > 0.05). However, among supplement users only, intake of vitamin D supplements was positively correlated with serum 25-OHD levels (ß = 0.03 ± 0.01 nmol/L, P = 0.05). These findings support the notion that serum levels of 25-OHD are influenced by other factors besides the vitamin D content of foods, including the use of vitamin D supplements. Serum 25-

OHD concentration tended to be correlated with hs-CRP levels (r = 0.14, P = 0.06), but was not significantly associated with adiposity and inflammatory biomarkers.

Among women with low 25-OHD (< 75 nmol/L), serum 25-OHD level was inversely associated with IL-2 and GM-CSF concentrations, and marginally associated with IL-

6 and IL-7 concentrations. Additional prospective studies in more heterogeneous populations will help to characterize the relationship between vitamin D status, inflammation and obesity.

vii

TABLE OF CONTENTS

Page

ACKNOWLEDGEMENTS ...... v ABSTRACT………...... vi LIST OF TABLES...... xii LIST OF FIGURES ...... xv

INTRODUCTION ...... 1 CHAPTER 1. LITERATURE REVIEW ...... 3 1.1 Background on Calcium and Vitamin D ...... 3

1.1.2 Absorption, Transport, Metabolism and Excretion...... 6

1.1.2.1 Calcium Absorption and Transport ...... 6

1.1.2.2 Calcium Metabolism and Excretion...... 9

1.1.2.3 Vitamin D Absorption and Transport ...... 10

1.1.2.4 Vitamin D Metabolism and Excretion ...... 11

1.1.3 Functions and Mechanisms of Action...... 12

1.1.4 Assessment of Nutritional Status ...... 13

1.1.5 Food Sources and Recommended Dietary Intakes ...... 17

1.2 Relationship between Dietary Calcium, Vitamin D and Adiposity ...... 23

1.2.1 Dietary Calcium and Adiposity ...... 23

1.2.2 Dietary Vitamin D and Adiposity ...... 33

1.3 Relationship between Adiposity and Serum 25-OHD Concentration ...... 34

1.4 Cytokines and Immune Function ...... 40

1.4.1 C-Reactive Protein and Inflammation ...... 45

1.5 Vitamin D and Inflammation ...... 49

1.6 Relationship between Serum 25-OHD, Adiposity and Inflammation ...... 50

viii

2. SPECIFIC AIMS AND HYPOTHESES ...... 56 2.1 Calcium, Vitamin D Intakes and Adiposity in Young Women ...... 56

2.1.1 Specific Aims ...... 56

2.1.2 Hypotheses ...... 56

2.2 Serum 25-OHD Concentration and Adiposity in Young Women ...... 57

2.2.1 Specific Aims ...... 57

2.2.2 Hypotheses ...... 58

2.3 Serum 25-OHD, Adiposity and Inflammation in Young Women ...... 58

2.3.1 Specific Aims ...... 58

2.3.2 Hypotheses ...... 59

2.4 Additional Covariates ...... 60

2.5 Human Subjects Proctection ...... 62

2.6 Permission to Access Data ...... 62

3. ASSOCIATIONS BETWEEN CALCIUM, VITAMIN D INTAKES AND OBESITY MARKERS IN YOUNG WOMEN (18 - 30 YEARS) ...... 63 3.1 Introduction ...... 64

3.2 Materials and Methods ...... 66

3.2.1 Study Design and Population ...... 66

3.2.2 Anthropometry and Lifestyle Factors ...... 67

3.2.3 Dietary Assessment ...... 69

3.2.4 Statistical Analysis ...... 69

3.3 Results ...... 72

3.4 Discussion ...... 74

3.5 Conclusions and Significance ...... 80

3.6 References ...... 82

4. ASSOCIATIONS BETWEEN SERUM 25-HYDROXYVITAMIN D CONCENTRATION AND ADIPOSITY MEASURES IN YOUNG WOMEN (18 - 30 YEARS) ...... 116 4.1 Introduction ...... 117

4.2 Materials and Methods ...... 119

4.2.1 Study Design and Population ...... 119

4.2.2 Assessment of Serum 25-OHD Status ...... 119

4.2.3 Dietary Assessment ...... 120

4.2.4 Assessment of Covariates ...... 120

4.2.5 Statistical Analysis ...... 122

4.3 Results ...... 124

4.4 Discussion ...... 125

4.5 Conclusions and Significance ...... 131

4.6 References ...... 132

5. ASSOCIATIONS BETWEEN SERUM 25-HYDROXYVITAMIN D CONCENTRATION, ADIPOSITY AND INFLAMMATORY BIOMARKERS IN YOUNG WOMEN (18 - 30 YEARS) ...... 158 5.1 Introduction ...... 159

5.2 Materials and Methods ...... 161

5.2.1 Study Design and Population ...... 161

5.2.2 Assessment of Serum 25-OHD Status ...... 161

5.2.3 Assessment of Inflammatory Biomarkers ...... 162

5.2.4 Assessment of Covariates ...... 162

5.2.5 Statistical Analysis ...... 164

5.3 Results ...... 166

5.4 Discussion ...... 168

5.4.1 Serum 25-OHD and Cytokine Levels ...... 169

5.4.2 Serum 25-OHD and C-Reactive Protein ...... 172

5.4.3 Adiposity and Inflammation ...... 173

5.5 Conclusions and Significance ...... 176

5.6 References ...... 178

6. SUMMARY, CONCLUSIONS AND FUTURE DIRECTIONS ...... 228 6.1 Summary and Conclusions ...... 228

6.2 Future Directions ...... 231

APPENDICES ...... 233 APPENDIX A. Permission to Access Data ...... 233

APPENDIX B. Normality and Distribution of Covariates ...... 235

APPENDIX C. Linearity and Spline Model Fit for Dietary Intakes ...... 242

APPENDIX D. Linearity and Spline Model Fit for Adiposity Measures...... 278

APPENDIX E. Linearity and Spline Model Fit for Serum 25-OHD Levels ...... 332

BIBLIOGRAPHY ...... 338

LIST OF TABLES

Table Page

1. IOM╧ Recommended Serum 25-Hydroxyvitamin D (25-OHD) Concentration ...... 16

2. Alternate Recommendation of Different Levels of Serum 25-OHD ...... 16

3. Selected Food Sources of Calcium ...... 18

4. Selected Food Sources of Vitamin D ...... 19

5. Recommended Dietary Allowances (RDAs) for Calcium ...... 20

6. Recommended Dietary Allowances (RDAs) for Vitamin D ...... 21

7. Characteristics of Participants (n = 270): UMass Amherst Vitamin D Status Study,

2006 - 2011 ...... 90

8. Unadjusted Dietary Intakes of Participants...... 92

9a. Variations in Characteristics of Participants by Adequacy of Total Calcium

Intake Ϫ, Defined by Recommended Dietary Allowance (RDA) ...... 93

9b. Variations in Characteristics of Participants by Adequacy of Total Vitamin D

Intake Ϫ, Defined by Recommended Dietary Allowance (RDA) ...... 95

10a. Variations in Characteristics of Participants by Total Body Fat (TBF) Percentage ..97

10b. Variations in Dietary Intakes of Participants by TBF Percentage ...... 99

11a. Variations in Characteristics of Participants by Body Mass Index (BMI) ...... 100

11b. Variations in Dietary Intakes of Participants by BMI ...... 102

12a. Variations in Characteristics of Participants by Waist Circumference (WC)

Measurement ...... 103

12b. Variations in Dietary Intakes of Participants by WC Measurement ...... 105

13. Correlations between Participant Characteristics and Adiposity Measures ...... 106

xii

14. Associations (ß ± SE) ϖ of Adiposity (Total Body Fat Percentage, Body Mass

Index and Waist Circumference) and Other Parameters ...... 107

15. Crude and Adjusted Indepenent Associtaions with Adiposity ß ± SE) ϖ in Young

Women ...... 108

16. Prevalence Odds Ratio for the Association of Calcium and Vitamin D Adequacy

with with Events of Adiposity ...... 111

17. Association of Calcium and Vitamin D Intakes with Event of High Adiposity ...... 112

18. Association of Women’s Combined Total Calcium and Vitamin D Intakes

with Events of High Adiposity ...... 115

19. Distribution of Serum 25-Hydroxyvitamin D (25-OHD) Levels ...... 141

20. Variations in Characteristics of Participants by Serum 25-OHD Levels ...... 142

21. Dietary Intakes of Participants by Serum 25-OHD Levels ...... 144

22. Mean Serum 25-OHD Levels by Oral Contraceptive Use, Alcohol Intake and

Smoking Status ...... 145

23. Mean Serum 25-OHD Levels by Adiposity...... 153

24. Unadjusted and Adjusted Mean Values of Adiposity by Serum 25-OHD Levels .....154

25. Linear Relationship between Serum 25-OHD Levels and Other Parameters ...... 156

26. Linear Relationship between Serum 25-OHD Levels and Adiposity ...... 157

27. Median and Inter-Quartile Range Distribution of Inflammatory Biomarkers ...... 190

28. Variations in Median and Inter-Quartile Range Distribution of Inflammatory

Biomarkers by Serum 25-OHD Status ...... 191

29. Variations in Median and Inter-Quartile Range Distribution of Inflammatory

Biomarkers by Waist Circumference (WC) Measurement ...... 192

xiii

30. Variations in Median and Inter-Quartile Range Distribution of Inflammatory

Biomarkers by Body Mass Index (BMI) ...... 193

31. Variations in Median and Inter-Quartile Range Distribution of Inflammatory

Biomarkers by Total Body Fat Percentage (TBF %) ...... 194

32. Median Inflammatory Biomarker Concentration by Categories of Oral

Contraceptive Use, Smoking Status and Alcohol Intake ...... 195

33. Pearson Product Correlations between Log Transformed Inflammatory Factors .....197

34. Correlations between Log Transformed Inflammatory Factors and Serum 25-OHD

Levels ...... 198

35. Mean Serum 25-OHD Concentration across Categories of High Sensitive C –

Reactive Protein (hs-CRP) Measures...... 199

36. Univariate Linear Relationship between logTransformed Inflammatory

Biomarkers and Other Parameters ...... 214

37. Association of Serum 25-OHD Concentrations with log Transformed

Inflammatory Biomarkers ...... 221

xiv

LIST OF FIGURES

Figure Page

1. Vitamin D Synthesis and Metabolism in Humans ...... 4

2. Role of Vitamin D in Calcium Metabolism ...... 7

3. Proposed Dietary Modulation of Adiposity via 1,25-dihydroxyvitamin D Mediation

of Intracellular Calcium Influx ...... 24

4. Prevalence at Risk for Vitamin D Deficiency by Age and Sex: United States, 2001

– 2006...... 35

5. Age and Season-Adjusted Prevalence at Risk of Deficiency and Inadequacy among

Persons aged 1 year and over: United States, 2001-2006 ...... 35

6. Mechanisms of Innate and Adaptive Immunity ...... 41

7. Activation of Inflammatory Response in Tissues ...... 46

8. CRP Ligand and Receptor Binding in Cells ...... 47

9. Possible Mechanism of CRP Modulating Inflammation ...... 48

10. Genes Involved in the Vitamin D Pathway...... 50

11. A DXA Scan in Progress ...... 68

12. A Dual-Energy X-ray Absorptiometry (DXA) Scan Report Showing Different

Body Composition Measures ...... 121

13. A DXA Scan showing Regional Adiposity Measures ...... 122

14. Covariation and Pearson Product Coefficient Correlation of Vitamin D Intake and

Serum 25-OHD Concentration ...... 146

15. Pearson Product Coefficient Correlation of Serum 25-OHD and Body Mass Index.147

16. Pearson Product Coefficient Correlation of Serum 25-OHD and Waist

Circumference ...... 148

17. Pearson Product Coefficient Correlation of Serum 25-OHD and Total Body Fat

Percentage ...... 149

18. Pearson Product Coefficient Correlation of Serum 25-OHD and Regional

Adiposity Measures (Arm and Leg Fat) ...... 150

19. Pearson Product Coefficient Correlation of Serum 25-OHD and Regional

Adiposity Measures (Trunk, Android and Gynoid Fat) ...... 151

20. Pearson Product Coefficient Correlation of Serum 25-OHD and Android to

Gynoid Fat ratio (AGF ratio) ...... 152

21. Lowess Regression Curve for the Association between High Sensitive C -

Reactive Protein (hs-CRP) and Serum 25-OHD Concentration ...... 200

22. Lowess Regression Curve for the Association between Interleukin-1β (IL-1β) and

Serum 25-OHD Concentration ...... 201

23. Lowess Regression Curve for the Association between Interleukin-2 (IL-2) and

Serum 25-OHD Concentration ...... 202

24. Lowess Regression Curve for the Association between Interleukin-4 (IL-4) and

Serum 25-OHD Concentration ...... 203

25. Lowess Regression Curve for the Association between Interleukin-5 (IL-5) and

Serum 25-OHD Concentration ...... 204

26. Lowess Regression Curve for the Association between Interleukin-6 (IL-6) and

Serum 25-OHD Concentration ...... 205

27. Lowess Regression Curve for the Association between Interleukin-7 (IL-7) and

Serum 25-OHD Concentration ...... 206

28. Lowess Regression Curve for the Association between Interleukin-8 (IL-8) and

Serum 25-OHD Concentration ...... 207

xvi

29. Lowess Regression Curve for the Association between Interleukin-10 (IL-10) and

Serum 25-OHD Concentration ...... 208

30. Lowess Regression Curve for the Association between Interleukin-12 (IL-12p70)

and Serum 25-OHD Concentration ...... 209

31. Lowess Regression Curve for the Association between Interleukin-13 (IL-13) and

Serum 25-OHD Concentration ...... 210

32. Lowess Regression Curve for the Association between Granulocyte Macrophage

Colony Stimulating Factor (GM-CSF) and Serum 25-OHD Concentration ...... 211

33. Lowess Regression Curve for the Association between Tumor Necrosis Factor

Alpha (TNF-α) and Serum 25-OHD Concentration ...... 212

34. Lowess Regression Curve for the Association between Interferon Gamma (IFN-γ)

and Serum 25-OHD Concentration ...... 213

xvii

INTRODUCTION

Recent estimates from the World Health Organization show that about 1.4 billion people worldwide are overweight and at least 500 million are obese (WHO, 2008).

Data from the Centers for Disease Control and Prevention indicate that in 2009 –

2010, more than one-third (35.7%) of U.S. adults were obese (body mass index, BMI,

> 30 kg/m2), (CDC/NCHS, 2012). The prevalence of obesity in young women, ages

20 to 39 years was also considerably high (31.9 %), (Flegal et al., 2012). This emerging obesity pandemic is worrisome, mostly because obesity is an independent risk factor for several obesity-related health problems, including cardiovascular disease (CVD), metabolic syndrome, type II diabetes mellitus (T2DM), and some kinds of cancers (Flegal et al., 2012; CDC/NCHS, 2012).

Although obesity is generally associated with sedentary lifestyle and increased caloric intake (energy imbalance), micronutrient status may also influence obesity risk

(Holick, 2006). The preponderance of available literature suggest that there is an inverse relationship between intakes of calcium/vitamin D and markers of adiposity, including BMI, waist circumference (WC), and total body fat percentage (% TBF), particularly in older and/or obese women (Zemel, 2000; Lin et al., 2000; Davies et al.,

2000; Buchowski et al., 2002; Zemel 2003; Zemel et al., 2004; Jacqmain et al., 2003;

Tidwell and Valliant, 2011). Most of the studies that have examined the role of dietary calcium and vitamin D intake in adiposity have relied on proxies such as BMI, waist-to-hip-ratio and skinfold measurement (Zemel, 2000; Lin et al., 2000; Davies et al., 2000; Buchowski et al., 2002) because they are easily obtained and non-invasive.

However, the use of these proxy measures may underestimate true body fat

percentage, especially among young women who are physically active (Roche et al.,

1981; Wellens et al., 1996; Flegal et al., 2009). In contrast, dual energy x-ray absorptiometry (DXA) is considered a more accurate method of estimating percentage body fat (Haarbo et al., 1991; Bolanowski and Nilsson, 2001) and its use among young women has been previously validated (Bolanowski and Nilsson, 2001).

Studies in both obese and non-obese adults, as well as post-menopausal women and ill-patients, show that serum 25-hydroxyvitamin D [25-OHD] concentration is inversely associated with inflammatory biomarkers such as C- reactive protein (CRP) and cytokines e.g., interleukins 4, 6, and 10 (Varghese et al.,

2012; Park et al., 2005; Bellia et al., 2011). Abdominal obesity (android to gynoid ratio) is also independently positively associated with CRP and IL-6 levels in obese adults (Kim et al., 2008). However, there is an important gap in our understanding about the extent of these associations in apparently healthy, non-obese, young women.

This lack of knowledge is problematic because an understanding of the modifiable dietary and lifestyle factors that contribute to increased adiposity and/or inflammation is important for the prevention and early intervention of chronic conditions such as metabolic syndrome, T2DM, and coronary heart disease. This dissertation used cross- sectional data from 270 healthy young women participating in the UMass Amherst

Vitamin D Status study to evaluate the extent to which vitamin D status and adiposity measures are associated with inflammatory biomarkers in young women, ages 18 to

30 years.

CHAPTER 1

LITERATURE REVIEW

1.1 Background on Calcium and Vitamin D

Calcium is the most abundant mineral in the body, representing about 1.5 - 2% of total body weight. Calcium is essential in living organisms, mainly in cell physiology, where movement of calcium ion (Ca2+) into and out of the cytoplasm functions as a signal for many cellular processes (Pietrobon et al., 1990; Chattopadhyay et al.,

1996). Calcium is required for nerve transmission and muscle stimulation (Karakia

1997; Utichel et al., 1992; Onimaru et al., 2003). It also plays a role in promoting cross-linking by binding to Gla- or hydroxyaspartate residues in pro-clotting enzymes

(Stipanuk, 2006). Almost all (99%) body’s calcium is stored in bones and teeth, where it works together with phosphorus to maintain skeletal strength and function. The other 1% is distributed in soft tissues, intra- and extracellular fluids and blood.

Calcium in foods and dietary supplements is mostly in the form of insoluble salts.

These salts can be solubilized at an acidic pH in the stomach; however, solubilization does not necessarily ensure better absorption because free calcium can bind to other dietary constituents, limiting its bioavailability.

Vitamin D is a fat-soluble vitamin that is naturally present in very few foods.

The term vitamin D refers collectively to two major nutritionally relevant chemical forms known as vitamin D2 (ergocalciferol), present in yeast and plants, and vitamin

D3 (cholecalciferol), which can be found in few foods, but is mainly synthesized from

7- dehydrocholesterol (7-DHC) in the skin. 7-DHC, also referred to as pro-vitamin D3, is an intermediate precursor of cholesterol that is synthesized in the sebaceous gland of the skin and is evenly distributed throughout the epidermis and dermis. During

exposure to sunlight, 7-DHC absorbs ultraviolet B radiation (UVB) photons with wavelengths between 290 and 315 nm. This process causes a transformation of 7-

DHC to pre-vitamin D3, which is then converted to vitamin D via thermal transformation (Holick, 1994). Conversion to vitamin D changes the structure of the molecule, facilitating its translocation from the skin into the blood stream. Once in the bloodstream, vitamin D binds to vitamin D binding protein (DBP), also called α-2- globulin (Fig 1).

Figure 1. Vitamin D Synthesis and Metabolism in Humans

Source: Holick MF. J Clin Invest 2006; 116: 2062-72.

Cutaneous synthesis of vitamin D is influenced by several factors, especially time of day, season of the year, and latitude. At latitudes above 42◦N (Boston) and

◦ below 42 S, very little pre-vitamin D3 is synthesized in skin from November through

February because the incident angle of sunlight at these latitudes during this time of year causes blockage of the UVB photons needed to convert 7-DHC to pre-vitamin

D3 (Webb et al., 1988). This period is extended at more extreme latitudes, such as in

Edmonton, Canada, (52◦N), to include October and March (Webb et al., 1988).

During spring and summer months, more UVB photons are able to penetrate the ozone layer because the sun is directly overhead. For the average fair skinned person, spending about 15 to 20 minutes in the midday sun (or when the sun is strongest) in minimal clothing (e.g., shorts and a tank top with no sunscreen) will give enough radiation to produce about 10,000 IU of vitamin D (Clemens et al., 1982; Holick,

2010). Excessive exposure to sunlight does not result in vitamin D toxicity because any pre-vitamin D3 that is not converted into vitamin D3 is degraded into biologically inert photoproducts such as luminsterol and tachysterol (Holick, 1994; Holick, 2006).

Excessive exposure to sunlight also causes some vitamin D3 to be converted to its non-active forms e.g., supersterol I, supersterol II, and 5, 6-trans-vitamin D3 (Holick,

1994; Holick, 2006).

Other factors that may influence vitamin D synthesis include age, skin tone, clothing, and use of sunscreen. Aging decreases the synthesis of 7-DHC in skin, thereby reducing the production of vitamin D3 by approximately 75% by age 70 years compared to younger adults (Holick et al., 1989). The skin pigment melanin is a natural sunblock that competes with 7-DHC for UVB photons (Clemens et al., 1982).

Hence, very dark- skinned people require almost ten times more sun exposure to make the same amount of vitamin D as light-skinned people (Clemens et al., 1982). Using sunscreen and wearing clothing that completely covers the body can also block

ultraviolet rays, reducing vitamin D synthesis (Wolpowitz and Gilchrest, 2006). In middle aged and older adults living in Boston for example, approximately 5-15 minutes of skin exposure (hands, arms, and legs) to sunlight two to three times a week is recommended for obtaining sufficient amounts of vitamin D

1.1.2 Absorption, Transport, Metabolism and Excretion

1.1.2.1 Calcium Absorption and Transport

Calcium absorption occurs in the small intestine. Most adults can absorb approximately 30% of the calcium in foods, but the efficiency of absorption tends to decrease as calcium intake increases (FNB/IOM, 2010). Absorption of calcium occurs in two ways: intracellular (active) absorption in the duodenum and proximal jejunum involving an epithelial calcium channel (TRPV6) and the calcium-binding transport protein Calbindin D9K (CaBP); and paracellular (passive) absorption that occurs throughout the small intestine, but mostly in the ileum and jejunum (Sheikh et al.,

1987).

Intracellular energy- dependent absorption is the primary mechanism for calcium uptake and transport in the body. Calcium is absorbed into the intestinal cell across the brush border via a channel (TRPV6) (Fig 2). When activated by

1,25(OH)2D, the nuclear VDR interacts with specific gene promoter regions in DNA and affects transcription of these vitamin D-specific genes, such as increasing expression of the enterocyte membrane calcium channel (TRPV6, also called CaT1

2+ (Wood et al., 2001) and Calbindin-D9K, intracellular Ca binding protein that facilitates the absorption of calcium (Bronner, 2009).

Figure 2. Role of Vitamin D in Calcium Metabolism

Source: http://bigtomato.org/endo/bcp/vitamind.php

Calbindin D9K transports calcium across the cytoplasm of the enterocyte to the basolateral (serosal) membrane. Calcium leaves the enterocyte with the assistance of a

Ca2+ - Mg2+ ATPase (an enzyme that hydrolyzes ATP and releases energy for pumping Ca2+ out of the cell as Mg2+ moves in). The Ca2+ - Mg2+ ATPase pump is also stimulated by vitamin D. Para-cellular calcium absorption, which occurs between cells, rather than through them, predominates when high calcium concentrations are present in the lumen. This forms a calcium concentration gradient between the lumen and the interstitial fluid, allowing for increased diffusion of calcium through tight junctions of the intestinal epithelial cells.

Calcium is absorbed in its ionized form; hence it must be released from the insoluble salts in which it is usually found in food and dietary supplements. Even though most calcium salts are dissolved in acidic pH of the stomach, some dietary calcium may still form insoluble complexes with other dietary components within the more alkaline pH of the small intestine, limiting its bioavailability. Fructose, oligosaccharides, inulin, and other non-digestible saccharides may enhance paracellular calcium absorption (Ziegler and Fomon, 1983), while vitamin D status, age and life stage, influence active calcium transport. In vitamin D deficient states, only

10 - 15% of dietary calcium is absorbed (Holick, 2006; Holick and Garabedian, 2006;

DeLuca, 2004). In infants and young children, net calcium absorption is as high as

75% because of the need to build bone (FNB/IOM, 1997). Calcium absorption usually decreases to 15% – 20% in older adults, mostly because of decreased renal production of 1,25(OH)2 D. Estrogen deficiency at menopause also decreases vitamin

D- mediated calcium absorption in the gut (Gallagher et al., 1980; Heaney et al.,

1989; Breslau, 1994). In a mixed diet, other components in food (e.g., phytic acid and

oxalic acid) may bind to calcium and inhibit its absorption (FNB/IOM, 2010). Foods such as spinach, collard greens, sweet potatoes, and rhubarb contain high levels of oxalic acid while whole-grain products, wheat bran, beans, seeds, nuts, and soy isolates contain high amounts of phytic acid (FNB/IOM, 2010), and the extent to which these compounds affect calcium absorption varies. For instance, eating spinach and drinking milk concurrently reduces the absorption of calcium in milk (Weaver and Heaney, 1991), but eating spinach and wheat products at the same time (with the exception of wheat bran) does not appear to lower calcium absorption (Weaver et al.,

1991). In general, the effects of these inhibitor-compounds are usually minimal and may not significantly impact nutritional status because people tend to eat a variety of foods during a meal. In contrast to the inhibitors described earlier, increased intake of dietary protein can facilitate calcium absorption (Kerstetter et al., 2005) due to greater trans-cellular calcium absorption (Gaffney-Stromberg et al., 2010).

1.1.2.2 Calcium Metabolism and Excretion

In order to maintain calcium homeostasis in the body, some of the calcium ingested is eliminated from the body in urine and feces, and an average of 60 mg may be lost daily through sweat (Charles et al., 1991). This process is necessary. Calcium is filtered and reabsorbed by the kidney such that urinary calcium losses range from about 100 - 240 mg/day with an average of about 170 mg/day (Charles et al., 1991;

Calvo et al., 1991). The amount of calcium excreted is also affected by many factors, including intakes of sodium, protein, phosphorus, caffeine and alcohol. Caffeine is a stimulant found in coffee and tea, and it can modestly reduce calcium absorption while increasing its excretion (Barrett-Connor et al., 1994). A comparison of data

from epidemiologic surveys and animal studies showed that for younger adult women consuming adequate calcium, moderate caffeine intakes (1 cup of coffee or 2 cups of tea per day) may have little or no harmful effects on bone metabolism (Massey and

Whiting, 1993). Alcohol consumption can also affect calcium status by reducing its absorption (Hirsch and Peng, 1996) and by inhibiting enzymes in the liver that help convert vitamin D into its pro-active form, 25-OHD (Laitinen et al., 1990; Turner,

2000). High sodium or protein intake increases urinary calcium excretion (Weaver et al., 1999; Heaney, 1996), although recent reports suggest that high protein intake also increases intestinal calcium absorption, thus effectively counteracting its effect on calcium excretion (Kerstetter et al., 2005).

A few observational studies have reported that consumption of carbonated soft drinks with high levels of phosphate is associated with reduced bone mass and increased fracture risk (bones remodeling may be hampered if calcium is inadequate, causing bones to become brittle or less dense) (Calvo, 1993; Heaney and Rafferty,

2001), although it is possible that this effect is due to replacement of milk with soda rather than the intake of phosphorus itself (Calvo, 1993; Heaney and Rafferty, 2001).

However, these studies did not control for dietary calcium intake in their analysis.

1.1.2.3 Vitamin D Absorption and Transport

Dietary vitamin D is absorbed from micelles together with fat and bile salts by way of passive diffusion in the intestinal cell; about 50% of dietary vitamin D3 is absorbed in this manner (Gropper et al., 2008). Although the rate of absorption is most rapid in the duodenum, the largest quantity of vitamin D is absorbed in the distal small intestinal. Within the intestinal cell, vitamin D is incorporated primarily into

chylomicrons, which then enter the lymphatic system with subsequent entry into the blood and transport to the liver. Chylomicrons transport about 40% of the cholecalciferol in blood, while the rest is bound to DBP for transport to extra hepatic tissues. Even though the vitamin D bound to DBP travels primarily to the liver, some may be picked up by other tissues, especially muscle and adipose tissue, primarily for storage, before hepatic uptake (Gropper et al., 2008).

1.1.2.4 Vitamin D Metabolism and Excretion

Vitamin D is stored in the body’s adipose tissue. Vitamin D obtained from sun exposure, food, and supplements is biologically inert and must undergo two hydroxylations in the body for activation (Fig 2). The first hydroxylation occurs in the liver, where the enzyme 25 hydoxylase (CYP27A) converts vitamin D to 25- hydroxyvitamin D [25-OHD], also known as calcidiol. The second hydroxylation is carried out by the enzyme 1-α-hydroxylase (CYP27B) in the kidney and forms the biologically active 1, 25-dihydroxyvitamin D (1, 25(OH)2 D), also known as calcitriol

(DeLuca, 2004; Holick, 2006; Holick and Garabedian, 2006). Once 1, 25(OH)2 D completes its actions in target tissues, it is principally excreted in bile. Excess amounts of 1, 25(OH)2 D also induce the expression of the enzyme 25- hydroxyvitamin D-24-hydroxylase (24-OHase), which enhances calcitriol catabolism

(Holick, 2007). Both 25-OHD and 1,25(OH)2 D undergo a 24-hydroxylation to form

24,25-dihydroxyvitamin D and 1,24,25-trihydroxyvitamin D, respectively. These hydroxylated metabolites undergo further hydroxylations in their side chain, resulting in the cleavage of the side chain between C-23 and C-24 and formation of the

biologically inert and water- soluble calcitroic acid, which is subsequently excreted in urine.

1.1.3 Functions and Mechanisms of Action

The major biological function of vitamin D is to promote calcium absorption in the gut and to maintain calcium homeostasis with the help of parathyroid hormone (PTH).

Low concentrations of serum calcium trigger PTH release from the parathyroid glands, which then stimulates 1-hydroxylase activity in kidney, converting 25-OHD to

1, 25(OH)2 D. The active form of vitamin D travels in the blood attached to vitamin D

Binding protein (DBP). In target tissues, 1,25(OH)2D binds to the vitamin D receptor

(VDR), which triggers a conformational change that increases receptor affinity for vitamin D response elements (VDRE) on specific genes. The 1,25(OH)2D together with the VDR and VDRE complex then influence transcription of genes to promote calcium absorption in the intestine and renal phosphorus-calcium reabsorption in the kidneys (Fig 2). Vitamin D is also required for bone growth and bone remodeling by osteoblasts and osteoclasts (Cranney et al., 2007).

Calcium is needed for bone growth, vascular contraction and vasodilation, muscle function, nerve transmission, intracellular signaling and hormonal secretion.

Maintenance of serum calcium and phosphorus concentrations within in their physiologic ranges is essential to support normal mineralization of bone and to prevent hypercalcemia and hypocalcemic tetany (a disease caused by abnormally low concentrations of calcium in the blood). Hypocalcemic tetany is characterized by hyper-excitability of the neuromuscular system and results in carpopedal spasms.

Adequate vitamin D intake also prevents rickets in children and osteomalacia in adults

(Wharton and Bishop, 2003; Jones, 2008). Together with calcium, vitamin D helps protect older adults from osteoporosis. Most tissues in the body (e.g., gut, kidney, gonads, heart, muscle, brain, and skin) have a VDR as well as the capacity to convert

25-OHD to 1,25(OH)2D (Holick, 2006; Bosse et al., 2009). Recent reports also suggest that vitamin D may have additional roles in the body, including modulation of cell growth (Holick, 2006; Krishnan and Feldman, 2011), neuromuscular and immune function (Marantes et al., 2011; Menant et al., 2012), and reduction of inflammation

(Norman and Henry, 2006).

1.1.4 Assessment of Nutritional Status

Although calcium concentration can be accurately measured in blood, measurement of serum total and ionic calcium status is inadequate for “true” assessment of calcium status because serum calcium concentration is tightly regulated and does not fluctuate with changes in dietary intakes. The body uses bone tissue as a reservoir and a source of calcium to maintain constant concentrations of calcium in blood, muscle, and intercellular fluid (Gropper et al., 2008). Normal serum calcium concentration usually ranges from 8.5 to 10.5 mg/d (Gropper et al., 2008). Low serum calcium levels may not also reflect calcium deficiency since serum calcium concentrations are tied to serum albumin status (albumin is a protein that binds calcium in plasma).

Abnormal levels of biochemical measures associated with calcium homeostasis are also not necessarily proof of dietary calcium insufficiency, because vitamin D deficiency, bone diseases, and other hormonal imbalances can produce similar symptoms (Stipanuk, 2006). Calcium status can also be measured using 24-hour urine sampling with adjustment for urinary creatinine (Weaver, 1990). Unfortunately, this

procedure is not very effective in assessing calcium status either, as some drugs may increase (e.g., antacids, anticonvulsants, carbonic anhydrase inhibitor diuretics, loop diuretics) or decrease (e.g., adrenocorticosteroids, birth control pills, and thiazide diuretics) calcium content in urine (Bringhurst et al., 2007; Wysolmerski et al., 2007).

Both 25-hydroxyvitamin D and 1,25-dihydroxyvitamin D levels can be specifically measured in serum or plasma (Holick, 1990; Chen et al., 1990). The concentration of 25-OHD is considered the best indicator of vitamin D status because it represents a summation of total cutaneous production of vitamin D and oral ingestion of either vitamin D2 or D3. Serum 25-OHD also has a fairly long circulating half-life of about 15 days (Jones, 2008) as compared to 1, 25(OH)2 D, which has a relatively short half-life of between 4 to 6 hours. Serum 1, 25(OH)2 D concentrations are also closely regulated by PTH, calcium, and phosphate (Jones, 2008); hence, levels of 1,25(OH)2D do not typically decrease until vitamin D deficiency is severe

(Cranney et al., 2007; Holick, 2007). Recent findings suggest that vitamin D3 may be more efficient at increasing serum 25-hydroxyvitamin D levels as compared to vitamin D2 (Malham et al., 2012). One of the difficulties associated with vitamin D status assessment lies in the actual measurement of serum 25-OHD concentration.

Presently, there is substantial variability among the various assays available (the two most common methods being antibody- based or liquid chromatography based) and among laboratories that conduct the analyses (Hollis, 2004; Carter, 2009). This means that compared with the actual concentration of 25-OHD in a blood sample, a falsely low or falsely high value may be obtained depending on the type of assay or the laboratory used (Binkley et al., 2004). A standard reference for serum 25-OHD assessment became available in July 2009. This standard material permits

standardization of values across laboratories and may improve method-related variability (NIST, 2009). Table 1 shows the Institute of Medicine (IOM)-proposed cut offs for assessing vitamin D sufficiency, deficiency, and toxicity using serum 25-

OHD concentrations.

While the 25-OHD cutoffs proposed by IOM serve as the “traditional” reference limits, many vitamin D researchers suggests that the optimal serum concentration of25-OHD is probably much higher than what is officially recommended for adults. Table 2 gives the proposed alternate 25-OHD reference limits considered “acceptable” by leading researchers in the field (Hollis and Wagner,

2004; Hollis and Wagner, 2004; Heaney, 2004; Vieth, 2004; Hollis, 2005; Hanley and

Davison, 2005; Bischoff-Ferrari et al., 2006; Holick, 2009). Serum 25-OHD level generally increases in response to increased vitamin D intake, but the relationship is not entirely linear (IOM/FNB, 2010). The increase varies, for example, by baseline serum levels and the duration of supplementation. For instance, increasing serum 25-

OHD from a baseline of ≥ 50 nmol/L requires more vitamin D than increasing levels from a baseline < 50 nmol/L. In persons with 25-OHD > 50 nmol/L, when the dose is

≥ 1,000 IU/day, the rise in serum 25-OHD is approximately 1 nmol/L for each 40 IU of vitamin D intake. In contrast, among subjects with 25-OHD < 50 nmol/L, a dose of

600 IU/day is associated with approximately 2.3 nmol/L increase in serum 25-OHD for each 40 IU of vitamin D consumed (IOM/FNB, 2010).

Table 1. IOM╧ Recommended Serum 25-Hydroxyvitamin D (25-OHD) Concentration 25-OHD Level 25-OHD Level Health Status (ng/mL)* (nmol/L)** < 12 < 30 Associated with vitamin D deficiency, leading to rickets in infants and children and osteomalacia in adults 12 - 20 30 - 50 Generally considered inadequate for bone and overall health in healthy individuals ≥ 20 ≥ 50 Generally considered adequate for bone and overall health in healthy individuals > 50 > 125 Emerging evidence links potential adverse effects to such high levels, particularly > 150 nmol/L ( > 60 ng/mL) ╧Source: Institute of Medicine, Food and Nutrition Board. Dietary Reference Intakes for Calcium and Vitamin D. Washington, DC: National Academy Press, 2010. * Serum concentrations of 25-OHD are reported in both nanomoles per liter (nmol/L) and nanograms per milliliter (ng/mL). ** 1 ng/mL = 2.496 nmol/L

Table 2. Alternate Recommendation of Different Levels of Serum 25-OHD 25-OHD Level 25-OHD Level Health Implications (ng/mL)* (nmol/L)** < 20 < 50 Deficiency

20 - 32 50 - 80 Insufficiency

32 - 100 80 - 250 Sufficiency

54 - 90 135 - 225 Normal in sunny countries

> 100 > 250 Excess

> 150 > 325 Intoxication

Source: Grant WB, Holick MF. Benefits and requirements of vitamin D for optimal health: A review. Altern Med Rev 2005; 10: 94-111. * Serum concentrations of 25-OHD are reported in both nanomoles per liter (nmol/L) and nanograms per milliliter (ng/mL). ** 1 ng/mL = 2.496 nmol/L

1.1.5 Food Sources and Recommended Dietary Intakes

In the U.S, the major contributor of dietary calcium for most individuals is dairy foods

(e.g., milk, yogurt, and cheese) (Briefel, 2004; Miller et al., 2007; FNB/IOM, 2010).

Plant foods such as Chinese cabbage, kale, broccoli, almonds and soybeans also provide substantial amounts of dietary calcium. Most cereals and grains provide very little calcium unless they are fortified; however, they can be considered as substantial contributors of overall dietary calcium intake because even though they contain small amounts of calcium, people tend to consume them frequently. Table 3 shows a list of selected food sources and their calcium content.

Naturally occurring vitamin D is rare in foods. The most significant dietary sources of vitamin D are fatty fish ( especially wild salmon, mackerel, and catfish), beef or veal liver, and fish oils, including cod- and tuna-liver oils (Cranney et al.,

2007; Ovesen et al., 2003; Chen et al., 2007). Cheese and egg yolk contain small amounts of vitamin D3 (Ovesen et al., 2003), and certain mushroom species also provide dietary vitamin D2 in variable amounts (Mattila et al., 1994; Calvo et al.,

2004; Philips et al., 2011). Table 4 shows the vitamin D content of some selected foods.

Table 3. Selected Food Sources of Calcium Food Milligrams (mg) per serving Yogurt, plain, low fat, 8 ounces 415 Orange juice, calcium-fortified, 6 ounces 375 Yogurt, fruit, low fat, 8 ounces 338 - 384 Mozzarella, part skim, 1.5 ounces 333 Sardines, canned in oil, with bones, 3 ounces 325 Cheddar cheese, 1.5 ounces 307 Milk, nonfat, 8 ounces** 299 Milk, reduced-fat (2% milk fat), 8 ounces 293 Milk, buttermilk, 8 ounces 282-350 Milk, whole (3.25% milk fat), 8 ounces 276 Tofu, firm, made with calcium sulfate, ½ cup*** 253 Salmon, pink, canned, solids with bone, 3 ounces 181 Cottage cheese, 1% milk fat, 1 cup 138 Ready-to-eat cereal, calcium-fortified, 1 cup 100 - 1,000 Turnip greens, fresh, boiled, ½ cup 99 Kale, raw, chopped, 1 cup 90 Soy beverage, calcium-fortified, 8 ounces 80 - 500 Chinese cabbage, bok choi, raw, shredded, 1 cup 74 Broccoli, raw, ½ cup 21

Source: U.S. Department of Agriculture, Agricultural Research Service. 2011. USDA National Nutrient Database for Standard Reference, Release 24. Nutrient Data Laboratory. Home Page, http://www.ars.usda.gov/ba/bhnrc/ndl. ** Calcium content varies slightly by fat content; the more fat, the less calcium the food contains. *** Calcium content is for tofu processed with a calcium salt. Tofu processed with other salts does not provide significant amounts of calcium. The U.S. Department of Agriculture recommends that persons aged 9 years and older eat 3 cups of foods from the milk group per day (ChooseMyPlate.gov, 2011). A cup is equal to 1 cup (8 ounces) of milk, 1 cup of yogurt, 1.5 ounces of natural cheese (such as Cheddar), or 2 ounces of processed cheese (e.g., American cheese).

Table 4. Selected Food Sources of Vitamin D Food IUs per serving* Cod liver oil, 1 tablespoon 1,360 Salmon (sockeye), cooked, 3 ounces 447 Tuna fish, canned in water, drained, 3 ounces 154 Orange juice fortified with vitamin D, 1 cup Approx. 137 Milk, nonfat, reduced fat, and whole, vitamin D-fortified, 1 cup 115 - 124 Yogurt, fortified with 20% of the DV for vitamin D, 6 ounces Approx. 80 Liver, beef, cooked, 3 ounces 42 Egg, 1 large (vitamin D is found in yolk) 41 Ready-to-eat fortified cereal, 0.75-1 cup Approx. 40 Cheese, Swiss, 1 ounce 6

Source: U.S. Department of Agriculture, Agricultural Research Service. 2011. USDA National Nutrient Database for Standard Reference, Release 24. Nutrient Data Laboratory. * IUs = International Units.

Recommended intakes for calcium and vitamin D are provided in the Dietary

Reference Intakes (DRIs) developed by the Food and Nutrition Board (FNB). DRI is a

general term for a set of reference values used for planning and assessing the nutrient intakes

of healthy people. These values, which vary by age and gender, include:

 Recommended Dietary Allowance (RDA): The average daily level of intake sufficient

to meet the nutrient requirements of nearly all (97% - 98%) healthy individuals.

 Adequate Intake (AI): established when evidence is insufficient to develop an RDA

and is set at a level assumed to ensure nutritional adequacy.

 Estimated Average Requirement (EAR): The average daily level of intake estimated

to meet the requirements of 50% of healthy individuals. It is usually used to assess the

adequacy of nutrient intakes in populations but not individuals.

 Tolerable Upper Intake Level (UL): maximum daily intake unlikely to cause adverse

health effects.

Table 5. Recommended Dietary Allowances (RDAs) for Calcium Age Male Female Pregnant Lactating 0 - 6 months* 200 mg 200 mg ------

7 - 12 months* 260 mg 260 mg ------

1 - 3 years 700 mg 700 mg ------

4 - 8 years 1,000 mg 1,000 mg ------

9 - 13 years 1,300 mg 1,300 mg ------

14 - 18 years 1,300 mg 1,300 mg 1,300 mg 1,300 mg

19 - 50 years 1,000 mg 1,000 mg 1,000 mg 1,000 mg

51 - 70 years 1,000 mg 1,200 mg ------

71 + years 1,200 mg 1,200 mg ------

Source: Committee to Review Dietary Reference intakes for Vitamin D and Calcium, Food and Nutrition Board, Institute of Medicine. Dietary Reference Intakes for Calcium and Vitamin D. Washington, DC: National Academy Press, 2010. * Adequate Intake (AI)

The RDA for calcium intake provided by the FNB was estimated based on the

amounts of calcium needed for bone health and to maintain adequate rates of calcium

retention in healthy people. The effects of inhibitor compounds and differences in

absorption of calcium from mixed diets are also factored into the overall estimation of calcium RDAs (Table 5).

Table 6. Recommended Dietary Allowances (RDAs) for Vitamin D Age Male Female Pregnancy Lactation 0 - 12 400 IU 400 IU ------months* (10 mcg) (10 mcg) 1 - 13 years 600 IU 600 IU ------(15 mcg) (15 mcg) 14 - 18 years 600 IU 600 IU 600 IU 600 IU (15 mcg) (15 mcg) (15 mcg) (15 mcg) 19 - 50 years 600 IU 600 IU 600 IU 600 IU (15 mcg) (15 mcg) (15 mcg) (15 mcg) 51 - 70 years 600 IU 600 IU ------(15 mcg) (15 mcg) > 70 years 800 IU 800 IU ------(20 mcg) (20 mcg)

Fortified foods can provide considerable amounts of vitamin D in a diet. In the

United States, foods such as milk, yogurt, cheese, margarine, as well as some brands of orange juice are often fortified with vitamin D. Milk and orange juice, for example, are fortified with about 400-500 IU (10 µg) of vitamin D per quart, providing 100 -

125 IU per eight ounce glass, while most cereals are fortified with about 40 - 140 IU

(1 - 3.5 µg) of vitamin D per serving (Calvo et al., 2004). Dairy products such as

yogurt may also be fortified with about 40 - 140 IU (1 - 3.5 µg) of vitamin D per serving, and recently, some kinds of bread are also being fortified with vitamin D. The

RDA for vitamin D represents a daily intake that is sufficient to maintain normal calcium homeostasis and bone health in healthy individuals with minimal sun exposure. RDAs for vitamin D are set listed in Table 6 in both International Units

(IUs) and micrograms (mcg); the biological activity of 40 IU is equal to 1 mcg.

Findings from the Continuing Survey of Food Intakes by Individuals (CSFII,

1994 -1996, 1998), NHANES 1999-2000, and NHANES 2005 – 2006 indicate that while vitamin D intake among the general population has steadily increased over the years, a high proportion of premenopausal women in the U.S. are still not meeting their vitamin D requirements (Bailey et al., 2010; Moore et al., 2005). Mean intake of vitamin D from food sources for female ages 19 to 50 years was 152 ± 4 IU/day in

CSFII, 168 ± 7 IU/day in NHANES 1999-2000, and 160 ± 12 IU/day in NHANES

2005-2006 (Bailey et al., 2010), but these amounts are still much lower than the newly established allowance of 600 IU/day (FNB/IOM, 2010). Results from

NHANES 2005-2006 also show that among female supplement users (19 - 30 years), the mean vitamin D intake from supplemental sources is insufficient (300 ± 28

IU/day) to meet current recommended dietary allowances (FNB/IOM, 2010). Despite the variety of both plant and animal dietary sources of calcium, data from the National

Health and Nutrition Examination Survey (NHANES 2003 – 2006), indicate that the mean daily calcium intake from food sources for young women (19 - 30 years) is inadequate (838 ± 25 mg/day) and below the current recommended intake level of

1000 mg/day for adults females (20 - 39 years) (FNB/IOM, 2010, Bailey et al., 2010).

Reports from NHAHES 2003 – 2006, also show that among female supplement users

(19 - 30 years), the mean daily contribution of calcium from supplemental sources averages approximately 283 ± 19 mg/day (Bailey et al., 2010).

1.2 Relationship between Dietary Calcium, Vitamin D and Adiposity

1.2.1 Dietary Calcium and Adiposity

In addition to its role in maximizing bone mineral mass in young adults and reducing the risk of osteoporosis during menopausal years (Nguyen et al., 2000; Cheng et al.,

2005; Rizzoli et al., 2010; Khadilkar et al., 2012), several epidemiologic studies have described an inverse association between dietary calcium intake and measures of adiposity (BMI, waist circumference, and abdominal visceral or subcutaneous fat), particularly in older or obese women. Zemel and associates pioneered research in the role of dietary calcium in energy/weight management when an “accidental” discovery was made during a clinical trial investigating the effects of dairy products on hypertension in obese African-Americans (Zemel et al., 1990; Zemel, 1994). They found that addition of calcium-rich dairy foods (yogurt) to the daily diet resulted in a sustained reduction in intracellular calcium concentrations (Zemel et al., 1990; Zemel,

1994), and significant reductions in body fat and circulating insulin (Zemel et al.,

2000). Twelve months of yogurt supplementation (sufficient to raise daily calcium intake from approximately 400 to 1,000 mg/day) led to significant decreases in body fat (from 32.3 ± 2.6 kg to 27.4 ± 3.1 kg, P < 0.01) measured using bioelectrical impedance (Zemel et al., 2000).

Although the exact mechanism explaining the relationship between dietary calcium- vitamin D- intake and adiposity is yet to be fully elucidated, Zemel and colleagues have hypothesized that the observation could be mediated through plasma

2+ concentrations of 1, 25 - dihydroxyvitamin D3 . Since intracellular Ca concentration can be modulated by calcitropic hormones (1, 25(OH)2D and PTH), Shi et al., used primary cultures of human adipocytes to measure the effects of 1, 25(OH)2D concentration on Ca2+ and lipid metabolism (Shi et al., 2001). They observed that treating human adipocytes with 1, 25(OH)2D resulted in a coordinated activation of fatty acid synthase and marked inhibition of lipolysis, suggesting that dietary calcium could potentially reduce adipocyte mass. Zemel et al., proposed that suppression of 1,

25(OH)2D activation with high-calcium diets would reduce adipocyte intracellular

Ca2+ concentration, inhibit fatty acid synthase, and activate lipolysis, thus exerting an anti-obesity effect (Zemel et al., 2000; Zemel, 2003),(Figure 3).

Figure 3. Proposed Dietary Modulation of Adiposity via 1,25-dihydroxyvitamin D

Mediation of Intracellular Calcium Influx

Low +

Source: Zemel, MB. J Am Coll Nutr 2001; 20: 428-35S.

Zemel et al. also used transgenic mice expressing the agouti aP2 gene in adipocytes to explore how dietary calcium regulates adiposity in vivo (Zemel et al.,

2000). Agouti mice were selected because, in addition to its role in determining coat color (Jackson, 1991; Bultman et al., 1992), agouti protein is also known as the obesity gene (Jones et al., 1996). Agouti protein in mice is expressed in adipocytes

(Jones et al., 1996; Xue et al., 1998), where it stimulates the expression and activity of fatty acid synthase, through a Ca2+ dependent mechanism (Jones et al., 1996; Xue et al., 1998). Human adipocytes have a similar agouti signaling protein called ASIP

(Bonilla et al., 2005).

Zemel and associates confirmed their earlier hypothesis by placing the mice on a modified diet containing suboptimal calcium (0.4%), sucrose, and fat increased to

25% of energy with lard. All mice were then individually randomized (not pair fed) to one of four groups. Group 1 (basal group) continued this diet with no modifications;

Group 2 (high calcium group) received the basal diet supplemented with CaCO3 to increase dietary calcium by threefold to 1.2%; Group 3 (medium dairy group), received a diet in which 25% of the protein was replaced by non-fat dry milk and dietary calcium was increased to 1.2%; Group 4 (high dairy group) had 50% of the protein replaced by non-fat dry milk, increasing calcium to 2.4%. Food intake and spillage was measured daily, and animals were weighed weekly. At the end of six weeks, the agouti mice treated with the basal diet had a weight gain of 24%, while the high calcium, medium dairy, and high dairy diets resulted in a weight reduction of

26%, 29% and 39%, respectively (P < 0.04). The basal diet resulted in a 2.6-fold increase in fatty acid synthase activity while the high-calcium-dairy diets stimulated

lipolysis (3.4 - 5.2 fold, P < 0.01), with greater effects from the high-dairy diets than from the high calcium diet. Assessment of fat mass also showed that all three high calcium and dairy diets triggered a 36% reduction in mass of the epididymal, abdominal, perirenal, and subscapular adipose tissue compartments (P < 0.001). This led Zemel and colleagues to speculate that the augmented effect of calcium from dairy sources vs. non-dairy calcium is likely attributable to additional bioactive compounds in dairy that work together with calcium to attenuate body fat (Zemel et al., 2003).

They proposed that branched-chain amino acids, which are abundant in whey proteins

(Shah, 2000), may act independently or synergistically with calcium to attenuate lipogenesis, accelerate lipolysis, and/or affect nutrient partitioning between adipose tissue and skeletal muscle (Zemel, 2003; Ha and Zemel, 2003). Whey proteins also contain angiotensin-converting enzyme, a metalloproteinase that is central to the blood coagulation cascade by converting angiotensin I to angiotensin II, which is required for angiotensin receptor activation and vessel constriction (Mullally et al.,

1997; Pihlanto-Leppala et al., 2000; Zamal et al., 2002). Animal studies have shown that adipocytes contain an autocrine/paracrine renin-angiotensin system, suggesting that adipocyte lipogenesis may be regulated in part by angiotensin II (Morris et al.,

2001).

Zemel et al., conducted a cross-sectional analysis of NHANES III data (1988 -

1994) in 380 adult women to determine whether the results from previous animal studies are relevant in defining a role for dietary calcium in modulation of adiposity at the population level (Zemel et al., 2000). Mean age, body mass index (BMI), and percent body fat (assessed using bioelectrical impedance) in this population were 28.7

± 0.4 years, 25.7 ± 0.4 kg/m2, and 32.7 ± 0.6 %, respectively. Mean dietary calcium

intake was 720 ± 52 mg/day and calcium intakes by quartiles were 255 ± 20, 484 ±

13, 773 ± 28, and 1346 ± 113 mg/day, for 1st, 2nd, 3rd, and 4th quartiles respectively.

After adjusting for energy intake, activity level, age, race, and ethnicity, the odds of being in the highest quartile of body fat (42.49 ± 0.44 %) were significantly reduced as the quartile of calcium intake increased, with a relative risk of 1.00 for 1st quartile;

nd rd th and 0.75, 0.40, 0.16 for 2 , 3 , and 4 quartiles respectively (Ptrend < 0.001).

Several observational and clinical trials have described similar results with respect to dietary calcium and BMI or body weight (Flemming and Yanovski, 1994;

Davies et al., 2000; Buchowski et al., 2002; Jacqmain et al., 2003; Zemel et al.,

2004), total or abdominal fat (Flemming and Yanovski, 1994; Zemel et al., 2005;

Lovejoy et al., 2001; Jacqmain et al., 2003; Zemel et al., 2004), and prevalence of obesity (Fleming and Yanovski, 1994). A cross-sectional analyses of 235 women

(ages 20 - 65 years) enrolled in the Quebec Family Study showed that calcium intake

(assessed with 3-day diet records) was negatively correlated with body fat mass (r = -

0.17, P < 0.05) and abdominal adipose tissue (r = -0.17, P < 0.05), estimated using computed tomography scans (Jacqmain et al., 2003). These correlations were all adjusted for confounding variables such as age, energy intake, dietary fat and protein intake, and markers of socioeconomic status. Investigators also reported that percentage body fat was significantly lower in women consuming ≥ 600 mg of calcium per day (31.3 ± 0.9 vs. 37.3 ± 1.6 %, P < 0.05), as compared to women with lower intakes (Jacqmain et al., 2003). Similar results were obtained with respect to

BMI and waist circumference (WC) in this cohort (BMI 27.0 ± 0.7 vs. 31.8 ± 1.2 kg/m2, P < 0.05; and WC of 82.0 ± 1.6 vs. 93.6 ± 2.6 cm, P < 0.05). While Jacqmain et al. study provides important information of the relationship between dietary

calcium intake and body fat assessed with computed tomography scans, use of a three diet recall may not adequately reflect usual dietary intakes of individuals.

The HERITAGE Family Study also examined the association between calcium intake (derived from a Willett FFQ) and adiposity in 201 Black and 261 Caucasian women aged 17 to 65 years (Loos et al., 2004). Participants were healthy but sedentary and had a BMI < 40 kg/m2. Percent body fat (% BF), total abdominal fat

(TAF), abdominal visceral fat (AVF) and abdominal subcutaneous fat (ASF) were assessed using computerized tomography scans. Subjects were divided into tertiles of energy-adjusted calcium intake and adiposity measures across tertiles were compared by ANOVA. Adiposity measures were also regressed against the energy-adjusted calcium intake to test for a linear trend. Their results showed that mean BMI, waist circumference (WC), and % BF were significantly higher in Blacks than in White women (BMI of 28.4 ± 0.5 vs. 25.0 ± 0.3 kg/m2, P < 0.001; WC of 90.2 ± 1.1 vs. 86.2

± 0.9 cm, P= 0.01; and BF of 36.0 ± 0.7 vs. 30.1 ± 0.6 %, P < 0.001, respectively).

Self-reported total calcium intake was also significantly higher in White women compared with Black women (1060 ± 34 vs. 825 ± 42 mg/day, P < 0.001). Among

Caucasian women, energy-adjusted calcium intake was negatively correlated with

BMI (P = 0.02), % BF (P < 0.01), TAF (P = 0.01), AVF (P = 0.03), and ASF (P =

0.01). In contrast, Black women in the high calcium intake group tended to have a higher BMI (P = 0.05) and WC (P = 0.01) and had significantly more FFM (P =

0.02) compared with Black women in the low calcium intake group. Loos et al’s compared the role of dietary calcium in modulating adiposity in black and Caucasian women, but their results did not control for vitamin D intake. This is important because calcium and vitamin D intake tend to be highly correlated in most diet. It is

important that one controls for the effects of vitamin D in order to tease out the independent effects of dietary calcium on adiposity.

Studies in premenopausal African American and Caucasian women have also described inverse associations between calcium intakes and adiposity. Lovejoy et al., recruited and compared dietary intakes and body fat in 97 Caucasian women (mean age 46.7 ± 0.3 years) and 56 African American women (mean age 47.7 ± 0.2 years)

(Lovejoy et al., 2001). Dietary intakes were assessed with 4-day food records and dual-energy X-ray absorptiometry (DXA) was used to measure percent body fat (%

BF), fat mass (FM), and lean mass (LM). In this 4-year prospective study, dietary intakes of calcium were significantly lower in African American women than in

Caucasian women (518 ± 34 vs. 758 ± 25 mg/day, P < 0.05). Partial correlation

(adjusted for total energy intake) showed inverse correlations between calcium intake and percent body fat (r = -0.01, P > 0.05 in Black women, and r = -0.25, P < 0.05 in

White women) or BMI (r = -0.13, P > 0.05 in Black women, and r = -0.21, P < 0.05 in White women). Lovejoy et al’s study adjusted for total energy intake but not vitamin D intake. This is important because calcium and vitamin D intake tend to be highly correlated in most diets, hence it is important that one controls for the effects of vitamin D in order to tease out the independent effects of dietary calcium on adiposity. Lovejoy et al also used a four day food record to assess dietary intakes, and this may not adequately reflect usual dietary intakes of individuals.

A prospective analysis of the effect of calcium intake on changes in body composition during a two year exercise intervention among 18 - 31 year old

Caucasian women in the U.S. also showed similar results (Lin et al., 2000). Fifty four sedentary and normal weight women (mean age 24.6 ± 3.3 years) were randomized

into either three sessions of resistance exercise plus 60 minutes of jumping rope per week or a control group for 24 months. None of the women had participated in more than two hours a week of exercise in the year prior to enrollment. Dietary intake was assessed with 3-day diet records at baseline and every six months for two years, and

DXA was used to estimate body fat and lean mass. They observed that mean calcium intake was below the recommended allowance of 1000 mg/day for this age group (781

± 212 mg/day). Total calcium (adjusted for energy intake) was also negatively correlated with changes in body weight (r = -0.35, P < 0.05) and body fat (r = -0.34, P

< 0.05). This prospective study by Lin et al was important in understanding the causal relationship between dietary calcium intake and obesity. However, use of a 3-day food record to assess dietary intake is a limitation because it may not adequately reflect usual patterns of dietary intake.

In a randomized control trial of 32 obese adults (27 women and 5 men, mean age 49 ± 6 years), participants were kept for 24 weeks on balanced deficit diets (500 kcal/day deficit) and assigned to one of three diets; a standard diet (400 - 500 mg of dietary calcium/d supplemented with placebo), a high-calcium diet (standard diet supplemented with 800 mg of calcium/day), or high-dairy diet (1200 to 1300 mg of dietary calcium/day supplemented with placebo) (Zemel et al., 2004). All subjects had an initial BMI between 30.0 - 39.9 kg/m2 and a low-calcium diet (500 - 600 mg/d) at study entry, and no more than 3 kg weight change over the preceding 12 weeks.

Participants kept daily diet diaries throughout the study, and compliance with treatment was assessed by weekly interview, review of the diet diary, and pill counts.

Subjects were also asked to maintain their physical activity level and tobacco use at pre-study (baseline) levels throughout the study. DXA was used to assess total fat

mass, regional fat, and percent lean / fat mass at the beginning of the study and at weeks 12 and 24 weeks. The results showed that while all subjects lost some amount of weight due to daily energy restriction of 500 kcal/day, subjects on the high-dairy diet lost the largest amount (70%) of body weight, followed by the high-calcium diet and the low-calcium control diet (26%),(P < 0.01). Quantitatively these losses represented 10.9 ± 1.6%, 8.6 ± 1.1%, and 6.4 ± 2.5% of body weight respectively. Fat loss also followed a similar trend with subjects losing 14.1 ± 2.4% of their body fat on the high-dairy diet and 11.6 ± 2.2% on the high-calcium diet compared to 8.1 ± 2.3% on the low-calcium control diet. Finally, changes in abdominal fat, which is indicative of central obesity occurred in all subjects, but those on the low-calcium diet lost the least amount (5.3 ± 2.3%) of abdominal fat compared to 12.9 ± 2.2% for those on the high-calcium diet and 14.0 ± 2.3% for the high-dairy diet (P < 0.025). While this clinical trial shed some light on how different calcium diets influence changes in body weight, subjects were all older adults and obese (BMI of 30.0 - 39.9 kg/m2) at the start of the intervention. Participants also had low-calcium diets (500 - 600 mg/d) at the start of the intervention. Since the mechanisms underlying how calcium might modulate adiposity could vary by BMI, data on these associations in non-obese subjects who are not calcium deficient would be useful.

Another clinical trial examined the effects of dairy rich foods on body weight and fat in 34 obese African-American adults (Zemel et al., 2005). In phase one of the study, researchers investigated weight maintenance by randomizing subjects to a low calcium (500 mg/day)/low dairy (< 1 serving/day) or high dairy (1200 mg Ca/day diet including 3 servings of dairy) diet with no change in energy or macronutrient intake.

In the second phase examining the effect of dairy/calcium on weight loss, subjects

were similarly randomized to the low or high dairy diets and placed on a caloric restriction regimen (deficit of 500 kcal/day). All subjects were aged 26 to 55 years, had an initial BMI of 30 - 40 kg/m2, and were on a low calcium (< 600 mg/d) and low dairy (< 1 serving/d) diet as determined by food frequency and diet history at study entry. Total fat mass (TFM) and percent lean (% LM) and fat mass (% FM) were assessed using DXA. The results from phase 1 showed no significant changes in body weight over the 24-week period in both groups. Compared to the low-dairy-low- calcium group, the high dairy diet resulted in significant decreases in total body fat

(2.16 kg, P < 0.01) and trunk fat (1.03 kg, P < 0.01), with a corresponding increase in

% LM (1.08 kg, P < 0.04). Phase 2 of the study showed that although both diets produced significant changes in body weight and fat loss, weight and fat loss were almost 2-fold greater in subjects consuming the high-dairy-high calcium diet compared with those consuming the low-dairy-low calcium diet (P < 0.01). Loss of lean body mass was also noticeably reduced in subjects consuming the high-dairy- high-calcium diet than in those on the low-dairy-low calcium (P < 0.001). This study included various body fat mass measures which is important as the site of fat deposition may influence the relationship between dietary calcium and adiposity.

However, the investigators failed to account for the effects of dietary vitamin D. This is important because calcium and vitamin D intake tend to be highly correlated in most diets, hence it is essential to control for the effects of vitamin D in order to tease out the independent effects of dietary calcium on adiposity.

1.2.2 Dietary Vitamin D and Adiposity

Presently, epidemiologic evidence on the relationship between dietary vitamin D intake and adiposity in young women is limited. Tidwell and Valliant used a cross- sectional design to assess the correlation between dietary intakes (by 24-hr recalls) and total body fat (by DXA) in a cohort of 100 premenopausal women aged 18 to 40 years (Tidwell and Valliant, 2011). A large proportion (71%) of participants were overweight or obese (BMI ≥ 25 Kg/m2) and mean BMI and % BF were 29.8 ± 6.9

Kg/m2 and 36.9 ± 7.3 % respectively. Average dietary vitamin D and calcium intake were also below the recommended dietary allowance for adult women (mean intakes of 176 ± 40 IU/day for vitamin D and 720 ± 263 mg/day for calcium). Partial correlation between % BF and vitamin D intakes (adjusted for fat, carbohydrate and protein intakes) showed that women with lower intakes of dietary vitamin D (≤ 152 ±

36 IU/day) were more likely to have excessive body fat (% BF > 37.9%) compared to women with higher intakes of dietary vitamin D (> 200 ± 32 IU/day) (r = -0.46, P <

0.01). This study did not adjust for calcium intake which could potentially confound any observed relationship between dietary vitamin D and body fat. Controlling for calcium is important because most food sources of calcium also provide some amounts of dietary vitamin D; hence food calcium and food vitamin D are likely to be highly correlated in any diet. Adjustment for dietary calcium in this instance is necessary in order to assess the independent effects of food vitamin D.

Although there is wealth of information supporting the role of dietary calcium and dairy products in fat/weight loss, research on the association between dietary vitamin D intake and adiposity in non-obese young women remains scarce. Since health risks tend to increase with increasing adiposity, it is important to determine

modifiable dietary factors that may significantly impact body weight. Previous studies examining the role of dietary calcium and vitamin D intake in body weight have relied on BMI, waist circumference, and measures of abdominal fat (e.g., use of abdominal skin fold thickness) to assess adiposity; however, these proxy measures are not always precise in estimating body fat composition. Dual energy x-ray absorptiometry (DXA) is considered a more accurate method of estimating percent body fat, (Haarbo et al.,

1991; Bolanowski and Nilsson, 2001), and its use among young women has been previously validated (Bolanowski and Nilsson, 2001). The first study of this dissertation examined the associations between dietary intakes of calcium and vitamin

D, and adiposity in young women. We also determined whether the observed association differs for body fat measured with dual x-ray absorptiometry (DXA) compared with BMI or WC. Since food calcium and food vitamin D are likely to be highly correlated in any diet, we included these covariates together in one same adjustment to assess the independent effects of each nutrient on adiposity.

1.3 Relationship between Adiposity and Serum 25-OHD Concentration

Low serum 25-hydroxyvitamin D (25-OHD < 75 nmol/L) is associated with increased risk of metabolic syndrome (Ford et al., 2005; Beydoun et al., 2010), type-II-diabetes

(Scragg et al., 2004; Pittas et al.,2006), hypertension (Scragg et al., 2007), coronary heart disease (Giovannucci et al., 2008; Poole et al., 2006; Cigolini et al., 2006;

Martins et al., 2007), and cancer (Banerjee and Chatterjee, 2003). Reports from

NHANES 2001 - 2006 show that about one-fourth of the US population (aged ≥ 1 year) is at risk of vitamin D inadequacy [serum 25-OHD of 30 – 49 nmol/L], while

8% were at risk of vitamin D deficiency [serum 25-OHD < 30 nmol/L] (Looker et al.,

2008). The risk of vitamin D deficiency also differs by age and gender (Fig 4), and by race and ethnicity (Fig 5).

Figure 4. Prevalence at Risk for Vitamin D Deficiency by Age and Sex in the United

States, 2001 – 2006

Source: CDC NCHS. NHANES data for ages 1 - 70 years from 2001 - 2006. *P < 0.05 compared with preceding age group within sex. At risk of deficiency is defined as serum 25-OHD < 30 nmol/L. Serum 25-OHD values have been adjusted for season of blood draw.

Figure 5. Age and Season-Adjusted Prevalence at Risk of Deficiency and Inadequacy among Persons aged 1 year and over in the United States, 2001-2006

Source: CDC NCHS. National Health and Nutrition Examination Survey (NHANES) data for ages 1 - 70 years from 2001 - 2006. P < 0.05 compared with males. P < 0.05 compared with non-Hispanic white persons. P < 0.05 compared with pregnant or lactating women.

In the U.S, the prevalence of vitamin D deficiency is highest in non-Hispanic blacks (31%) followed by Mexican-Americans (12%) and non-Hispanic White

(3.2%). Since natural food sources of vitamin D are limited, addition of vitamin D fortified foods (e.g., dairy foods) to the diet is important in preventing vitamin D insufficiency. Unfortunately, most adults (65 %) in the U.S. do not consume adequate amounts of milk and milk products (Forrest and Stuhldreher, 2011). Concern about sun exposure and skin cancer risk also prevents people from staying out in the sun for long periods, which increases the risk of vitamin D insufficiency (Linos et al., 2012;

Holick, 2001).

Accumulating epidemiologic evidence suggests an inverse relationship between serum 25-OHD levels and body fat (Zemel et al., 2000; Davies et al., 2000;

Lovejoy et al., 2001; Vilarrasa et al., 2007; Kremer et al., 2009; Beydoun et al., 2010) or BMI (Bell et al., 1985; Liel et al., 1988; Worstman et al., 2000; Rodriguez-

Rodriguez et al., 2009). Among older adults (32 – 62 years), Konradsen et al., observed that subjects with BMI > 39.9 kg/m2 had 24% lower serum 25-OHD levels

2 and 18% lower 1,25(OH)2D levels than those with BMI < 25 kg/m (Konradsen et al.,

2008). Studies in elderly women (69 - 87 years) also showed that total body fat is a negative predictor of serum 25-OHD levels (β = -0.247, P = 0.016), even after adjusting for age, lifestyle, and serum intact parathyroid hormone (Jungert et al.,

2012). Among healthy Caucasians and African-American adults (25 – 48 years),

Parikh et al., observed that serum 25-OHD concentration was negatively correlated with BMI (r = -0.40; P < 0.001) and percentage body fat (r = -0.41; P < 0.001), measured by dual energy x-ray absorptiometry (Parikh et al., 2004). Arunabh et al., also reported that after adjusting for race, age, season, and dietary vitamin D intake in

20 - 80 year old women (mean BMI 23.9 ± 2.9 kg/m2) ,serum 25-OHD level was inversely correlated with percentage body fat (r = -0.13, P = 0.013), measured using dual energy x-ray absorptiometry (Arunabh et al., 2003).

Among Caucasian and African-American adults (25 – 48 years), serum 25-

OHD concentration was significantly lower (58.7 ± 30.5 vs. 77.4 ± 35.9 nmol/liter; P

< 0.0001) in obese subjects (mean BMI 37.3 ± 5.8 kg/m2) compared with non-obese subjects (mean BMI 25.6 ± 2.9 kg/m2),(Parikh et al., 2004). A recent study in 66

Spanish women aged 20 – 35 years also examined the relationship between vitamin D status and anthropometric indices in overweight/obese young women (Rodriguez –

Rodriguez et al., 2009). A questionnaire was used to collect demographic and health information, and a 3-day food record was used to assess dietary intake. Women were divided into two groups depending on their serum vitamin D concentrations: low-D

(25-OHD < 90 nmol/L) and high-D (25-OHD ≥ 90 nmol/L). The results showed that mean serum 25-OHD concentration in the High-D group was almost three times higher compared with the Low-D group (130.2 ± 52.3 vs.47.9 ± 18.4 nmol/L, P <

0.001). While intakes of vitamin D, calcium, and supplements did not differ significantly between groups, BMI and waist circumference were significantly lower than in the Low-D subjects (26.0 ± 1.3 kg/m2 and 79.4 ± 3.4 cm compared to

28.6 ± 3.2 kg/m2 and 86.2 ± 9.3 cm, respectively; P < 0.05) in the high-D subjects.

Most of the afore-mentioned studies included obese adult women and does not provide data on how low-25-OHD is associated with adiposity in non-obese young women.

Cross-sectional studies in menopausal-aged women have reported that elevated dietary vitamin D intake and serum 25-OHD values maybe be related to

lower visceral adiposity and omental adipocyte size. Caron-Jobin et al., measured adipocyte size in 43 menopausal women using adipocyte suspensions from collagenase-digested fat tissues (Caron-Jobin et al., 2011). Total and visceral adiposity were assesses by DXA and computed tomography, respectively, while serum 25-OHD concentration was measured using radio-immunoassay. They found that dietary vitamin D intake was inversely associated with visceral adipose tissue area (r = -0.34, P ≤ 0.05). Serum 25-OHD concentration was also negatively correlated with visceral adipose tissue area (r = -0.32), total adipose tissue area (r = -

0.44), subcutaneous adipose tissue area (r = -0.36), BMI (r = -0.43) and total body fat mass (r = -0.41, P ≤ 0.05). This study effectively assessed the role of both dietary vitamin D and serum 25-OHD on total and visceral adiposity, but the results are in menopausal women and cannot be generalized to pre-menopausal aged women.

The low levels of serum 25-OHD observed in obesity have been attributed to multiple factors including decreased exposure to sunlight because of limited mobility, negative feedback from elevated 1, 25-hydroxyvitamin D and PTH levels on hepatic synthesis of 25-OHD (Bell et al., 1985), and excessive storage of vitamin D in the adipose tissue (Liel et al., 1988; Shi et al., 2001). Lumb et al., were among the first researchers to hypothesize that after absorption in the small intestine, vitamin D is sequestered and stored in adipose tissue and muscle and then slowly released into circulation (Lumb et al., in 1971). The same group went on to demonstrate this hypothesis using animal models and human studies. Vitamin D in all body tissues was radioactively-labeled by supplementing weanling rats (previously made completely

14 vitamin D-deficient) with oral vitamin D3-4- C for two weeks. Measurement of radioactivity and vitamin D content in a several organs and tissues showed that the

largest amount of radioactivity was located in adipose tissue (Rosenstreich et al.,

1971). Similarly, when 60 individuals were injected with radioactively labeled vitamin D3, the highest concentration of biological activity and radioactivity was observed in fat tissue (Mawer et al., 1972).

Cross-sectional studies in morbidly obese subjects have suggested that the decline in serum 25-OHD concentration in obesity may be secondary to alteration in tissue distribution resulting from increase in adipose mass (Liel et al., 1988). More recently, it has been confirmed that obesity-related vitamin D insufficiency is most likely a result of decreased bioavailability of vitamin D3 from cutaneous and dietary sources due to its deposition in adipose tissue (Worstman et al., 2000; Shi et al.,

2001). Shi et al., have shown that elevated concentrations of 1, 25(OH)2D stimulate lipogenesis and inhibit lipolysis in cultured human adipocytes, leading to accumulation of fat in adipocytes (Shi et al., 2001). Additionally, 1,25-(OH)2D can inhibit the expression of adipocyte uncoupling protein-2 (UCP2), which causes a decrease in the metabolic efficiency of adipocytes (Shi et al., 2002).

The second study of this dissertation evaluated the extent to which serum 25-

OHD levels is associated with measures of adiposity, and the influence of lifestyle factors (e.g., oral contraceptive use, smoking status, and alcohol intake) on this association in young women. We also examined whether the relationship between vitamin D status and adiposity differs with the site of fat deposition (i.e. android, gynoid, arm and leg fat). Android fat is located primarily around organs within the abdominal cavity while gynoid fat is found around the hips, thighs, and buttocks.

Since vitamin D can be sequestered in fat stores (adipocytes), our analysis of the association between regional fat distribution and serum 25-OHD levels can provide

insight on the relationship between vitamin D status and different forms of obesity, and inform directions for future epidemiologic research in this field.

1.4 Cytokines and Immune Function

The human immune system comprises two major sides: the innate or non-specific immune system, and the adaptive or specific immune system (Fig 6). The innate immune system is the body’s first line of defense against microorganisms such as bacteria that penetrate the epithelial surface (Kaisho, 2007). Innate immunity consists of epithelial barriers, circulating phagocytes (primarily neutrophils and macrophages), and other cytotoxic cells (e.g., natural killer (NK) cells), as well as constitutive and inflammation-induced serum proteins (e.g., complement proteins and positive acute- phase reactant proteins) (Kaisho, 2007). Activated macrophages secrete small soluble hormone-like proteins (cytokines) that allow for communication between cells and the external environment (Janeway et al., 2005). T Activated macrophages also release proteins known as chemokoines that attract cells with specific chemokine receptors such as neutrophils and monocytes from the blood stream (Janeway et al., 2005).

Figure 6. Mechanisms of Innate and Adaptive Immunity

Source: Kaisho T. Elucidating the mechanism behind immunity using dendritic cells. Riken Research. 2007;2(8). Available at http://www.rikenresearch.riken.jp/eng/frontline/5028.

The term cytokines broadly encompasses a large family of proteins, including interleukins, colony-stimulating factors, interferons, tumor necrosis factor, and chemokines (Leonard, 1999). Cytokines are secreted by white blood cells as well as a variety of other cells (fibroblasts, endothelial cells, epithelial cells, etc.) and they work in tandem with specific soluble cytokine receptors and cytokine inhibitors to mediate and regulate the human immune response.

Cytokines can regulate cellular activity in a coordinated interactive way due to the following attributes (Leonard, 1999; Stipanuk, 2006):

i). Pleiotrophy – one cytokine has many different functions.

ii). Redundancy – several different cytokines can mediate the same or similar

functions.

iii). Synergism – occurs when the combined effect of two cytokines on cellular

activity is greater than the additive effects of individual cytokines.

iv). Antagonism- the effects of one cytokine inhibits the effects of another.

Although cytokines are produced by many cell types, the majority are produced by helper T-cells (Th) and macrophages (Leonard, 1999). The innate immune system is regulated by pro-inflammatory cytokines e.g., IL-1 and TNF-α, as well as anti-inflammatory cytokines e.g., IL-4 and IL-10, which down-regulate inflammation after pathogens have been eliminated (Packard et al., 2009; Damania and Blackbourn, 2012). The macrophages and neutrophils of the innate immune system cannot always eliminate infectious microorganisms, which is why the lymphocytes of the adaptive immune system have evolved to provide a more versatile means of defense. The adaptive immune system consists of antibody responses and cell-mediated responses, which are carried out by specific cytokines produced by different lymphocyte cells, B cells and T cells, respectively (Siffrin et al., 2007;

Fatourou and Koskinas, 2009; Andersson et al., 2010).

Pro-inflammatory cytokines function primarily to induce inflammation as a result of infection, trauma, ischemia, toxins etc. Interleukin-1 (IL-1) and tumor necrosis factor (TNF-α) typically initiate the cascade of inflammatory mediators by

targeting the endothelium. TNF-α is produced by macrophages and monocytes and is involved in phagocytic cell activation (in macrophages) and tumor cell cytotoxicity. It is the primary cytokine that mediates acute inflammation, activates platelets, and initiates cachexia (Bazzoni and Beutler, 1996; O'Dell, 1999). IL-1 and IL-8 are primarily synthesized by macrophages, but, IL-1 can also be made by B-cells and dendritic cells. The IL-1 family gene encodes three different peptides; IL-1α, IL-1β, and IL-1 receptor antagonist (IL-1ra) (Mao et al., 2000). In normal homeostasis, the actions of IL-1 are maintained in balance by IL-1ra and other IL-1 inhibitors. Like

TNF-α, IL-1 production is induced in response to inflammatory stimuli from target cells like NK-cells, T-cells, and B-cells, while IL-8 production is induced in response to inflammatory stimuli from neutrophils (Tayal and Kalra, 2008). IL-1 functions in cell proliferation and differentiation and production of fever (Tayal and Kalra, 2008), while IL-8 is a neutrophil chemoattractant that triggers neutrophil degranulation and release of cytotoxic (Dinarello, 2000). Both IL-1β and TNF-α stimulate expression of adhesion molecules, including vascular cell adhesion molecule-1, which causes endothelial dysfunction (Jialal et al., 2004; Elkind et al., 2002). IL-1 and TNF-α have also been strongly implicated in the pathogenesis of inflammatory conditions like rheumatoid arthritis (Badolato and Oppenheim, 1996; Feldman et al., 1996) and

Crohn’s disease (Van Dullemen et al., 1995)

Anti-inflammatory cytokines block the inflammatory cascade initiated by pro- inflammatory cytokines in the host. They work together with specific cytokine inhibitors and soluble cytokine receptors to regulate the human immune response.

Some of the most studied anti-inflammatory cytokines include interleukin (IL)-1 receptor antagonist, IL-4, IL-6, IL-10, IL-11, and IL-13 (Opal and DePalo, 2000). IL-

4 is a highly pleiotropic (it can act on many different types of cells) cytokine produced by mature Th-2 cells, mast cells and basophils. It drives Th-2 responses, mediates the recruitment and activation of mast cells, and stimulates the production of antibodies via the differentiation of B-cells into IgE-secreting cells (Wang et al., 1995; Brown and Hural, 1997). IL-4 also kills parasites in the body and suppresses macrophage- derived nitric oxide production (Vannier et al., 1992). IL-4 is able to affect a variety of structural cells. For instance, it can enhance proliferation of vascular endothelium and skin fibroblasts, or decrease the proliferation of vascular smooth muscle cells (Toi et al., 1991; Brown and Hural, 1997). It can also induce a potent cytotoxic response against tumor cell formation (Tepper et al., 1992; Toi et al., 1992).

IL-10 is synthesized by Th-2 lymphocyte cells, CD4+ cells, monocytes, and

B-cells. It can inhibit cytokine production by neutrophils and natural killer cells. IL-

13 is secreted by activated T lymphocytes and has profound effects on the expression of surface molecules on both monocytes and macrophages (de Waal Malefyt et al.,

1993). IL-4, IL-10 and IL-13 have marked inhibitory effects on the expression and release of monocyte-derived pro-inflammatory cytokines e.g., TNF-α, IL-1, IL-6 and

IL-8, (te Velde et al., 1990; Paul, 1991; Gerard et al., 1992; de Waal Malefyt et al.,

1993; Marchant et al., 1994; Zurawski and de Vries, 1994; Brandtzaeg et al., 1996;

Clarke et al., 1998). They down-regulate the pro-inflammatory cytokines by promoting the degradation of their messenger RNA (Opal et al., 1998). IL-6 is often used as a marker for systemic activation of pro-inflammatory cytokines (Barton,

1997). IL-6 production is induced by lipopolysaccharide stimulation, along with TNF-

α and IL-1 (Opal and DePalo, 2000). Although IL-6 is generally considered to be a pro-inflammatory cytokine, it also possesses anti-inflammatory properties. IL-6 is a

potent inducer of the acute-phase protein response (Barton et al., 1996). It also inhibits the production of pro-inflammatory cytokines such as GM-CSF and IFN-γ

(Barton, 1997).

1.4.1 C-Reactive Protein and Inflammation

When macrophages encounter microbes such as bacteria in tissues, they release cytokines that increase the permeability blood vessels, allowing fluids and proteins to pass into the tissues. The macrophages also release proteins called chemokines, which direct the migration of neutrophils to the site of infection. Cytokines and complement fragments have important effects on the adhesive properties of the endothelium, causing leukocytes to stick to the endothelial cells of the blood vessel wall. The dilation and increased permeability of the blood vessels leads to increased blood flow and leakage of fluid which accumulates at the site of infection (Janeway et al., 2005).

This process is generally described as inflammation and is defined by four Latin words, calor, dolor, rubor, and tumor: meaning heat, pain, redness, and swelling, all of which reflect the effects of cytokines and other inflammatory mediators. Figure 7 below depicts how infection triggers an inflammatory response in tissues.

Figure 7. Activation of Inflammatory Response in Tissues

Source: Janeway et al., 2005. Immunology biology, the immune system in health and disease.

C-reactive protein (CRP) is an acute-phase reactant protein synthesized by the liver in response to a variety of inflammatory cytokines produced by macrophages

(Pepys and Hirschfield, 2003) and adipocytes (Lau et al., 2005). CRP is a 224-residue protein located on chromosome 1q21-q23. It has a molecular mass of 25106Da and is a member of the pentaxin family of proteins (Du Clos, 2003). CRP binds to Fc- receptors and phosphocholine residues (Fig 8) on the surface of microbes, damaged tissue, nuclear antigens, in order to activate the humoural, adaptive immune system

(via phagocytosis by macrophages) (Marnel et al., 2005).

Figure 8. CRP Ligand and Receptor Binding in Cells

Source: Marnel et al. Clin Immunol 2005; 117: 104-11.

Interaction of CRP with Fc-receptors leads to the generation of pro- inflammatory cytokines (e.g., TNF-α, IL-1, and IL-6), that enhance the inflammatory response (Fig 9). Consequently, the level of CRP in plasma increase rapidly in response to infection, trauma, tissue injury or necrosis, infection, or other inflammatory stimuli (e.g., autoimmune disorders).

Figure 9. Possible Mechanism of CRP Modulating Inflammation

Source: Marnel et al. Clin Immunol 2005; 117: 104-11.

CRP is also important in innate immunity, where it acts as an early defense system against infections. During acute phase response, levels of CRP rapidly increase within the first two hours of infection, reaching a peak within forty-eight hours. Since its half-life is constant (18 hours), the level of CRP in blood is a direct reflection of the severity of inflammation or infection. Even though measurement of

CRP concentration is not specific enough to diagnose any particular disease, it does serve as a general marker for infection and inflammation, which is important in the early detection and treatment of chronic diseases such as diabetes mellitus (Pradhan et

al., 2001; Dehdhan et al., 2007), hypertension (Kong et al., 2012; Cheung et al.,

2012), and cardiovascular disease (Danesh et al., 2004; Swardfager et al., 2012).

1.5 Vitamin D and Inflammation

Few studies have examined the regulatory influence of vitamin D on cytokines in humans. Research in elderly populations suggests that changes in serum 25-OHD levels may not be associated with alterations in pro- or anti-inflammatory cytokines

(Barnes et al., 2011). However, during pathophysiological conditions (e.g., severe obesity and end-stage renal disease), some pro-inflammatory cytokines such as (TNF-

α) were elevated while anti-inflammatory cytokines like IL-10 were reduced in subjects with low-serum 25-OHD concentrations (Bellia et al., 2011; Kelly et al.,

2011).

Animal studies and population based studies have proposed that 1,25-

(OH)2D may help to suppress inflammation by meditating the release of specific cytokines and expression of cytokine receptors (Canning et al., 2001; Zhu et al., 2005;

Shab-Bidar et al., 2012). Figure 10, shows a proposed mechanism by which calcitriol could influence gene transcription of specific cytokines. After 1,25(OH)2D is produced in the kidney, it travels from the blood (bound to vitamin D binding protein) into the nucleus of cells, where it binds to the vitamin D receptor (VDR). The ligand/receptor complex then binds to vitamin D response element (VDRE) located in the promoter region of target genes. The DNA-bound complex subsequently interacts with nuclear co-regulators, such as SKIIP (MacDonald et al., 2004), and alters the rate of gene transcription of certain cytokines, including IL-10 and IL-8 (Bosse et al.,

2009).

Figure 10. Genes Involved in the Vitamin D Pathway

Source: Bosse et al. Respir Res 2009; 10: 98.

1.6 Relationship between Serum 25-OHD, Adiposity and Inflammation

Visceral obesity is associated with a chronic, low-grade inflammation of white adipose tissue, characterized by increased production of pro-inflammatory cytokines

(e.g., IL-1, IL-6, and tumor necrosis factor- α) and acute phase reactant proteins (e.g,.

C - reactive protein) (Hotamisligil et al., 1993; Vozarova et al., 2001; Xu et al., 2003;

Moriseset et al., 2008), some of which are actually secreted by adipose tissue itself

(Mohamed-Ali et al., 1997; Fried et al., 1998; Yudkin et al., 2000; Galic et al., 2010).

These cytokines may have local physiological effects on white adipose tissue physiology as well as systemic effects on other organs (Bastard et al., 2006).

The distribution of body fat in obesity is also of importance because an excess of abdominal fat is associated with high amounts of visceral fat which is thought to be metabolically active and can cause dysmetabolism of fatty acids, thus increasing the influx of free fatty acids into the splanchnic circulation (Bergman et al., 2007; Nielsen et al., 2004). High amounts of unbound free fatty acids in the blood can lead to hyperlipidemia and hypertriglyceridemia, both of which can promote the onset of cardiac complications. Varghese et al. conducted a case control study in patients with coronary artery disease (mean age cases: 43 ± 11.4 and controls 40.8 ± 9.3 years) to access the importance of fat distribution in heart disease (Varghese et al., 2012).

Controls were age-and sex matched to subjects and computer tomography scans were used to measure subcutaneous and visceral fat area. Result showed that visceral fat area was highly correlated with CRP (r = 0.45, P = 0.001) as well as a significant predictor of cardiovascular risk (P = 0.009). Among healthy older Korean adults

(mean age 41.3 ± 13 years), visceral fat was found to be the most important predictor of hs-CRP concentration, explaining 19.6% of the variance in high sensitive CRP (β =

0.009 mg/dL, P < 0.001) (Kim et al., 2008). Another cross-sectional study of 100

Korean adults (ages 20-60 years) showed that in both obese and non-obese adults, serum concentrations of CRP, TNF-α, and IL-6 were significantly positively correlated with body weight , BMI, waist circumference, and waist to hip-ratio (Park et al., 2005). Regression analysis further showed a significant association between

BMI and CRP (β = 0.158 mg/dL, P = 0.028), and a positive association between visceral adiposity and elevated cytokine levels (IL-6) in obese subjects (β = 0.012 pg/mL, P = 0.042). Even though the above studies reported of an inverse association

between measures of adiposity and inflammation (hs-CRP, TNF-α, and IL-6), all these studies were conducted in older adults or critically ill patients.

Some cross-sectional studies in obese and healthy U.S. adults have shown an inverse association between serum 25-OHD levels and plasma CRP concentrations

(Bellia et al., 2011; Beydoun et al., 2010). Elevated CRP concentration is a public health concern because it is an independent risk factor for several chronic diseases, including cardiovascular disease (Pradhan et al., 2001; Singh et al., 2005; Dehghan et al., 2007; Lopez-Garcia et al., 2005; Pickup and Mattock, 2003; Ridker, 2001; Ridker et al., 2000; Rost et al., 2001), and type II diabetes (Schmidt et al., 1999; Pradhan et al., 2001; Freeman et al., 2002). Studies in post-menopausal women (Ridker et al.,

1998) have also shown that women with high baseline CRP levels (> 7.3 mg/L) were almost 5-times more likely to experience any vascular event (Relative Risk 4.8; 95%

CI, 2.3 - 10.1) compared to women with low baseline CRP concentrations (< 1.5 mg/L). This association remained significant even after adjusting for body mass index, diabetes, hypertension, hypercholesterolemia, physical activity, and family history of cardiovascular disease (Ridker et al., 1998).

Nationally representative data from the National Health and Nutrition

Examination Surveys (NHANES) 2001 to 2006 showed that after controlling for the effects of age, gender, and race, there was an inverse association between 25-OHD and CRP concentrations (ß = -0.30 mg/dL, P < 0.001) in asymptomatic adults with low vitamin D levels (below the population median, i.e. 25-OHD < 52.5 nmol/L),

(Amer and Quayyum, 2012). However, this association became non-significant once the serum 25-OHD increased above the population median (≥ 52.5 nmol/L), (ß = -

0.05 mg/dl, P = 0.08). Garcia-Bailo et al. recently reported of a cross-sectional study

(n = 1403) among adult men and women (20-29 years) that showed positive associations between plasma concentrations of 25-OHD and hs-CRP levels (ß = 0.011

± 0.002 mg/L, P < 0.0001) after controlling for factors such as age, gender, waist circumference, physical activity, ethnicity and season in the overall population

(García-Bailo et al., 2013). They also observed that among individuals with 25-OHD

> 51.9 nmol/L, the positive association between 25-OHD and hs-CRP was no longer significant (ß = 0.003 ± 0.004 mg/L, P = 0.29), after adjustment for hormonal contraceptive (HC) use. HC use was therefore a confounder of the association between 25-OHD and hs-CRP in this population.

Supplementation with vitamin D has been effective in down-regulating the expression of pro-inflammatory cytokines in some populations (Zittermann et al.,

2009; Schleithoff et al., 2006; Shab-Bidar et al., 2010), but not others (Jorde et al.,

2010; Beilfuss et al., 2012). A clinical trial by Zittermann et al. supplemented 200 healthy-but-overweight subjects participating in a weight-reduction program with either 83 mcg (3320 IU) of vitamin D daily or placebo for 12 months. All subjects all had low 25-OHD concentrations (< 30 nmol/L) at the time of treatment randomization. They found that even though weight loss was not significantly affected by vitamin D supplementation, mean 25-OHD and 1,25(OH)2 D concentrations increased by 55.5 nmol/L and 40.0 pmol/L, respectively, in the treatment group compared to 11.8 nmol/L and 9.3 nmol/L, respectively, in the placebo group (P < 0.001). Subjects receiving vitamin D supplements also had significantly lower concentrations of TNF-α compared with the placebo supplemented group (7.04

± 2.25 pg/mL vs. 7.90 ± 2.80, P = 0.049).

Another study in older men with congestive heart failure also showed that a daily regimen of 50 μg vitamin D3 plus 500 mg calcium for 9 months was beneficial in suppressing TNF-α while increasing the levels of anti-inflammatory cytokines like

IL-10 (Schleithoff et al., 2006). In a randomized controlled trial among 100 patients with type II diabetes mellitus, subjects were assigned either a plain yogurt drink (PD; containing 170 mg calcium and no detectable vitamin D/250 mL) or a vitamin D3- fortified yogurt drink (FD; containing 170 mg calcium and 500 IU/250 mL) twice a day (Shab-Bidar et al., 2010). Following treatment, patients who received the vitamin

D fortified drink had a significant increase in serum 25-OHD concentration (before

38.5 ± 20.2 vs. after 72.0 ± 23.5 nmol/L; P < 0.001) compared with patients who got the plain drink (before 38.0 ± 22.8 vs. after 33.4 ± 22.8 nmol/L, P = 0.28). Intake of

FD was also associated with a significant decrease in hs-CRP concentration (before

2.04 ± 2.69 vs. after 1.60 ± 1.04 mg/L; P < 0.001).

Beilfuss et al. investigated the relationship between vitamin D status (and such pro-inflammatory cytokines as tumor necrosis factor-alpha (TNF-α), interleukin 6 (IL-

6) and high sensitive C-reactive protein (hs-CRP) in overweigh/obesity patients. A total of 332 subjects (21-70 years), with BMI between 28.0 and 47.0 kg/m2 were included in this 12 month randomized control trial. Hs-CRP and cytokines were measured using high sensitivity ELISAand serum levels of 25-OHD were measured by radioimmunoassay. Participants received either 40,000 IU vitamin D

(cholecalciferol) per week (DD), or 20,000 IU vitamin D per week (DP), or placebo.

No associations between IL-6, TNF-α and 25-OHD were found at baseline. After the one year intervention, IL-6 levels were significantly reduced in DD and DP subjects

(delta -0.14; 95% CI (4.98 - 27.85), compared to placebo group (delta 0.02; 95% CI (-

8.08 - 7.72). Hs-CRP levels were however significantly increased in DD and DP subjects (delta 0.11; 95% CI (-9.91 - 59.28), compared to placebo group (delta -0.08;

95% CI (-18.89 - 37.04).

Information on the relationship between erum 25-OHD status and inflammation remains largely conflicted, with the majorities of studies being carried out in post-menopausal, obese, or chronically ill populations. Studies examining relationships between vitamin D status and inflammation in young women remain scare. Few studies have controlled for the potential confounding effects of factors such as oral contraceptive use, alcohol intake, or smoking status, especially in young women. Our third study examined the association between vitamin D status (serum

25-OHD levels), adiposity, and inflammatory factors in apparently healthy premenopausal aged women.

CHAPTER 2

SPECIFIC AIMS AND HYPOTHESES

2.1 Calcium, Vitamin D Intakes and Adiposity in Young Women

2.1.1 Specific Aims

1. To evaluate the extent to which dietary calcium and vitamin D intakes are associated with adiposity in young women enrolled in the UMass Amherst Vitamin D

Status Study.

2. To determine whether the above association varies with different types of adiposity, e.g., total body fat percentage (TBF %) measured with dual xray absorptiometry

(DXA), compared to body mass index (BMI) or waist circumference (WC)

 Primary Exposure - Dietary Intakes: Vitamin D, Calcium, and Dairy.

 Outcome – Obesity Markers: Total body fat percentage (% TBF), Waist

Circumference (WC), and body mass index (BMI).

2.1.2 Hypotheses

1. Dietary calcium, vitamin D, and dairy intakes will be inversely associated with measures of adiposity (% TBF, BMI, WC) in this population of apparently healthy young .women (i.e. women with higher food calcium, dairy, or vitamin D intakes will be more likely to have lower adiposity measures, evidenced by low BMI, WC, or %

TBF (TBF ≤ 32%, BMI < 25 Kg/m2, or waist circumference ≤ 80 cm] compared to women with lower intakes of calcium, dairy, or vitamin D).

2. The hypothesized inverse association between calcium-vitamin D- intakes and body composition will be stronger with total body fat percentage (%TBF) measured by

DXA as compared to BMI.

2.2 Serum 25-OHD Concentration and Adiposity in Young Women

2.2.1 Specific Aims

1. To determine the extent to which the relationship between vitamin D status and adiposity is influenced by lifestyle factors (e.g., oral contraceptive use, smoking status, alcohol intake and physical activity level) in young women.

2. To determine whether the relationship between vitamin D status and adiposity varies with the site of fat deposition (i.e. android, gynoid, arms, legs, and trunk fat).

3. To estimate the prevalence of vitamin D insufficiency in a population of premenopausal aged women using two different sets of cut-off values for serum 25-

OHD concentration (i.e. Traditional (FNB/ IOM, 2010) cut-off values: serum 25-

OHD < 25.0 nmol/L = Deficient, 25.0 – 49.9 nmol/L = Insufficient , and ≥ 50.0 nmol/L = Adequate. Current evidenced based (Bischoff-Ferrari et al., 2006; Holick et al., 2011) proposed cut-offs, serum 25-OHD < 50.0 nmol/L = Deficient, 50.0 – 74.9 nmol/L = Insufficient, and ≥ 75.0 nmol/L = Adequate).

4. To evaluate the correlation between dietary vitamin D intake and serum 25-OHD concentration in this population.

 Primary Exposure – Adiposity: Total body fat percentage (% TBF),

Android Fat (ANF), Gynoid Fat (GYF), Android to Gynoid Fat Ratio

(AGFR), Arm Fat (AF), Leg Fat (LF), Trunk Fat (TF), Body Mass Index

(BMI), Waist Circumference (WC).

 Outcome - Vitamin D Status: Serum 25-OHD Concentration in nmol/L.

2.2.2 Hypotheses

1. Women with higher adiposity (TBF > 32%, BMI ≥ 25 Kg/m2, or waist circumference > 80 cm) will be more likely to have lower serum 25-OHD concentrations, compared to women with lower adiposity measures (TBF ≤ 32%, BMI

< 25 Kg/m2, or waist circumference < 80 cm). Oral contraceptive use, smoking status and alcohol intake may confound the association between adiposity and 25-OHD in young women.

2. The hypothesized inverse association between adiposity and serum 25-OHD levels will be stronger with AN, GY, and %TBF than AF, LF, or TF.

3. The prevalence of vitamin D deficiency (25-OHD < 75 nmol/L) will be low (< 15

%) in this population of young women as a result of seasonal variation in 25-OHD levels and vitamin D supplement intake.

4. Dietary vitamin D intake will be correlated with serum 25-OHD concentration, but may not be necessarily be a significant predictor of serum 25-OHD concentrations.

2.3 Serum 25-OHD, Adiposity and Inflammation in Young Women

2.3.1 Specific Aims

1. To determine the extent to which serum 25-OHD level is associated with inflammatory biomarkers (including highly sensitive C-reactive protein (hs-CRP), interleukin (IL)-1β, 2, 4, 5, 6, 7, 8, 10, 12p70, 13, granulocyte-macrophage-colony- stimulating factor (GM-CSF), interferon-gamma (IFN-γ) and tumor necrosis factor- alpha (TNF-α) in apparently healthy young women.

2. To determine whether adiposity is associated with inflammation in young women.

3. To determine whether adiposity modifies the relationship between 25-OHD status and inflammation in young women.

 Primary Exposure: Vitamin D Status- Serum 25-OHD concentration

 Outcome – Inflammation: High sensitive C-reactive protein (hs-CRP),

interleukin (IL)-1β, 2, 4, 5, 6, 7, 8, 10, 12p70, 13, granulocyte-

macrophage-colony-stimulating factor (GM-CSF), interferon-gamma

(IFN-γ), and tumor necrosis factor-alpha (TNF-α).

 Other Covariates: Adiposity - Total body fat percentage (% TBF), Android

Fat (AN), Gynoid Fat (GY), Arm Fat (AF), Leg Fat (LF), Trunk Fat (TF),

Body Mass Index (BMI), Waist Circumference (WC), Waist-to-Height-

Ratio (WHtR).

2.3.2 Hypotheses

1. Serum 25-OHD levels will be inversely associated with the inflammatory biomarker (hs-CRP) in young women.

2. High adiposity (TBF > 32%, BMI ≥ 25 Kg/m2, or waist circumference > 80 cm) will confound/modify the relationship between serum 25-OHD and inflammation in this population.

3. Regional fat mass (AN, GY, AF, LF, TF) will be positively correlated with pro- inflammatory biomarkers (e.g., TNF-α, IL-1β, IL-6, IL-12p70) in young women.

2.4 Additional Covariates

Previous studies have suggested that cigarette smoking and high levels of alcohol use is linked to impaired intestinal calcium absorption (Berg, 2008; Sampson, 2002;

Saggese et al., 1995). We used the Harvard FFQ described earlier to collect information on alcohol use (i.e. beer, red and white wine, and liquor) in our population. We also collected information on self-reported history of smoking with our study questionnaire.

Physical activity, specifically vigorous activities and high-intensity exercise have been linked with lower amounts of subcutaneous skinfold thicknesses and fat distribution (Obarzanek et al., 1994). Higher amounts of leisure time physical activity has also been positively associated with the low/ normal BMI and lower body fat range in women (Ball et al., 2001). In our study, women were asked to report their average weekly physical activity using a modified assessment based on that used by the Nurses’ Health Study II. Activities listed on the questionnaire included walking, jogging, running, bicycling, aerobics/dance/rowing machine, tennis/racket sports, swimming, yoga/pilates, and resistance exercises. All responses ranged from zero to eleven or more hours per week. We assigned metabolic equivalent (MET) scores using procedures described in literature (Ainsworth et al., 1993). The validity of these physical activity questions have been previously assessed in a representative sample of the Nurses’ Health Study II Cohort (n = 147), with a correlation of 0.62 reported for the questions listed above collected over four weeks using activity diaries and a correlation of 0.79 reported for the questions of physical activity recalled in the past week (Wolf, 1994).

Current smoking status and increasing BMI have been independently associated with a higher risk of serum 25-OHD deficiency (Odds Ratio 1.59; 95% CI

1.35 - 1.87 for smoking status and Odds Ratio 1.04; 95% CI 1.02 - 1.06) (Melamed et al., 2008). Cutaneous synthesis of vitamin D also decreases with increasing age

(Holick et al., 1989); and people with darker skin color require almost ten times more sunlight exposure in order to make the same amount of vitamin D as light skinned people because of melanin in the skin which blocks the absorption of UVB radiation

(Clemens et al., 1982; Holick and Chen, 2008). Vitamin D insufficiency is also more common at the end of winter than at the end of summer (Millen and Bodnar, 2008).

Among free-living healthy young adults in Boston, Massachusetts, the prevalence of vitamin D insufficiency (25-OHD ≤ 20 ng/mL) was more common at the end of winter than at the end of summer (30 % vs. 11% respectively) (Tangricha et al.,

2002). There was also a significant seasonal variation in the 25-OHD concentrations between the winter and summer (35 ± 10 ng/mL at the end of summer, compared to

30 ± 10 ng/mL at the end of winter; P <0.01). Seasonal variation in 25-OHD levels was highest in subjects aged 18 to 29 years (28 ± 10 ng/mL at the end of winter, vs.

36 ± 10 ng/mL at the end of summer) and people who took multivitamin / vitamin D supplement were also more likely to have high serum 25-OHD concentrations than those who did not. Subjects who took a daily multivitamin supplement increased their serum 25-OHD levels by 30% compared to those who did not (37 ng/mL vs. 29 ng/mL, P < 0.01) and this association was true in both summer and winter seasons

(Tangricha et al., 2002). We included all the above covariates in our analyses to assess for potential interaction/confounding effects.

2.5 Human Subjects Protection

Ethical clearance for this dissertation will was not done separately because this dissertation was conducted as part of the UMass Amherst Vitamin D Status Study, which had already obtained ethical clearance from the Human Subjects Committee of the University of Massachusetts at Amherst. Potential participants were provided with information describing the study and were made aware that participation was entirely voluntary. Each participant read and signed an informed consent document prior to her participation. One potential risk of participation was breach of privacy by study personnel. To minimize this risk, subjects were assigned a unique code to be used on forms instead of their names. Only project investigators had access to records linking subjects’ name and codes and all identifying information was kept separate from analysis databases. All study staff were educated on the need for confidentiality and all records were kept in a locked filing cabinet. Sensitive digital information was kept in password-protected computer files.

2.6 Permission to Access Data

Prior to accessing the data to be used for this dissertation analyses, written permission was obtained from the principal investigators of the UMass Amherst Vitamin D Status

Study; Dr. Alayne Ronnenberg and Dr. Elizabeth Bertone-Johnson (Appendix A).

CHAPTER 3

ASSOCIATIONS BETWEEN CALCIUM, VITAMIN D INTAKES AND

OBESITY MARKERS IN YOUNG WOMEN (18 – 30 YEARS)

A. A. Addo-Lartey §, E. R. Bertone-Johnson ‡, C. Bigelow ‡, R. J. Wood §, S. E.

Zagarins‡, S. H. Gehlbach ‡, A. G. Ronnenberg §

§ Department of Nutrition, School of Public Health and Health Sciences, University of

Massachusetts Amherst, MA, 01003, USA

‡ Division of Epidemiology and Biostatistics, Department of Public Health, School of

Public Health and Health Sciences, University of Massachusetts Amherst, MA,

01003, USA

Abstract: Calcium and vitamin D intakes have been linked with reduced body mass

index (BMI ≥ 30 kg/m2) and body fat percentage in older adults and in obese women.

However, very little is known about the extent of these associations in healthy non-

obese young women. We evaluated these relationships among 270 women aged 18-30

years (Mean BMI: 22.9 ± 3.3 kg/m2) in a cross-sectional analysis within the UMass

Amherst Vitamin D Status Study (2006-2011). Dietary information and total body fat

percentage (%TBF) were obtained using a validated food frequency questionnaire and

dual energy x-ray absorptiometry, respectively. After adjustment for total energy

intake, physical activity level and smoking status, women reporting adequate intakes

of calcium (≥ 1000 mg/day) but low intakes of vitamin D (< 600 IU/day) were over

two-fold more likely to have high TBF (≥ 32%); [Odds Ratio, OR, 95% Confidence

Interval, CI: 2.27; 1.02 - 5.04], compared to those with adequate intakes of both

calcium (≥ 1000 mg/day) and vitamin D (≥ 600 IU/day). Low intake of calcium from

dietary supplements (225 to 500 mg/day) was associated with risk of increased WC (≥

80 cm) [OR, 95% CI: 2.27, 1.17 - 4.39]. Furthermore, the risk of being overweight/obese (BMI ≥ 25 kg/m2) increased in a linear manner across tertiles of decreasing supplemental calcium intake (Ptrend = 0.02). Our findings suggest that addition of calcium supplements to inadequate diets could reduce adiposity risk in young women. Adequate dietary intakes of both calcium and vitamin D may also be necessary to reduce body fat percentage, even among non-obese young women.

Further studies are needed to examine these associations and also investigate whether dietary intakes of calcium and/or vitamin D modify adiposity levels in young women.

Keywords: Calcium, Vitamin D, Adiposity, Percentage body fat, Body mass index,

Waist circumference, Young women

3.1 Introduction

In addition to their well-established roles in bone metabolism [1-3], intakes of calcium and vitamin D have been inversely associated with adiposity in some populations [4-

6]. In the U.S., about 50% of dietary calcium is derived from dairy products [7, 8], although other foods, including broccoli, kale, almonds, and soybeans, can provide substantial amounts of calcium, particularly for vegetarians [9]. While most circulating vitamin D is derived from endogenous cutaneous synthesis following exposure to ultraviolet rays [10-12], fortified foods can provide considerable amounts of vitamin D. Milk, for instance, is usually fortified with about 400 IU (10 µg) of vitamin D per quart, providing 100 IU per eight ounce glass, while most cereals are fortified with about 40 - 140 IU (1 to 3.5 µg) of vitamin D per serving [13]. Fatty fish

(especially wild salmon, mackerel, and catfish), fish oils, oysters, and certain

mushroom species also provide substantial amounts of dietary vitamin D [11-14].

Data from the National Health and Nutrition Examination Survey (NHAHES) 2003 –

2006 [15] show that despite the variety of both plant and animal dietary sources of calcium, mean daily calcium intake from food sources for young women (19 to 30 years) remains inadequate (838 ± 25 mg/day) and below the current recommended dietary allowance (RDA) of 1000 mg/day for adults females (20 - 39 years) [16].

Similarly, among premenopausal aged women (19 - 50 years), mean daily intake of vitamin D from food sources was around 160 ± 12 IU/day, [15] which is much lower than the newly established RDA of 600 IU/day for young women [16].

While many epidemiologic studies have described an inverse relationship between calcium-vitamin D intake and adiposity, particularly in older or obese women [17-25], the exact mechanism explaining this association is yet to be fully elucidated. Zemel and colleagues have hypothesized that the association could be mediated through plasma concentrations of 1, 25 dihydroxyvitamin D3 (calcitriol), the metabolically active form of vitamin D [17-20]. This hypothesis suggests that low dietary calcium intake increases production of calcitriol, which, in turn, stimulates intracellular calcium (Ca2+) influx. The increased intracellular Ca2+ concentration then enhances expression of lipogenic genes while inhibiting lipolysis, leading to increased fat cell mass (adiposity). Additional studies by the same group indicate that the inverse association between dietary calcium and adiposity appears stronger for calcium consumed from dairy products than from other calcium sources [17-20, 26], and in the United States, dairy products are typically the primary source of both calcium and vitamin D [7, 8].

Most of the studies assessing the relationship between dietary intakes of calcium, vitamin D and adiposity have relied on proxy measures of adiposity such as body mass index (BMI), waist circumference, and skinfold measures of abdominal fat. However, the use of these proxy measures may underestimate true body fat percentage [27-29]. In contrast, dual energy x-ray absorptiometry (DXA) is considered a more accurate method of estimating percentage body fat [30, 31] and its use among young women has been previously validated [31]. Given that health risks increase dramatically with increasing adiposity, identifying modifiable dietary factors in early adulthood may significantly impact risk of obesity and related conditions later in life. For the present analyses, we examined the association between dietary intakes of calcium, vitamin D or dairy products and three measures of adiposity (DXA estimates of percentage body fat, BMI and waist circumference) among young women enrolled in the UMass Amherst Vitamin D Status Study.

3.2 Materials and Methods

3.2.1 Study Design and Population

The University of Massachusetts Amherst Vitamin D Status Study is a cross-sectional study intended to assess vitamin D status in young women and to identify its dietary, environmental, and lifestyle determinants, as well as its association with other health measurements. Between March 2006 and September 2011, 288 healthy women aged

18 to 30 years from the Amherst, Massachusetts, area were enrolled in the study.

Women were recruited into the study using advertisement and fliers posted around college. Women were eligible to participate if they were not currently pregnant; not experiencing untreated depression; did not have a history of high blood pressure or

elevated cholesterol, kidney or liver disease, bone disease (e.g., osteomalacia), digestive disorders, multiple sclerosis, rheumatoid arthritis, thyroid disease, hyperparathyroidism, cancer, Type 1 or Type 2 diabetes mellitus, or polycystic ovaries; and were not taking corticosteroids, anabolic steroids, anticonvulsants, propranolol, or cimetidine. The present analyses include 270 women who met eligibility criteria and had dual energy x-ray absorptiometry (DXA) measurements. This study was approved by the Institutional Review Board at the

University of Massachusetts, Amherst. Written informed consent was obtained from all participants.

3.2.2 Anthropometry and Lifestyle Factors

All study measurements were completed in a single clinic visit scheduled for the late luteal phase of each participant's menstrual cycle. Information on lifestyle and demographic factors was collected by self-reported questionnaire. During the study visit, women’s weight was measured using a calibrated scale with subjects wearing light clothing and no shoes. Height and waist circumference were measured to the nearest 0.1 cm using a wall mounted stadiometer and a flexible tape measure, respectively. Body mass index (BMI) was calculated as weight in kilograms divided by the square of height in meters. Percentage of total body fat (%TBF) was measured by dual energy x-ray absorptiometry (Fig 11) using the total body scan mode on a narrow angle fan GE Lunar Prodigy scanner (GE Lunar Corp., Madison, WI). The

DXA scan was performed with the participant lying in a supine position with her arms and legs placed according to the manufacturer’s specifications. Measurements were completed by the same technician, and the machine was calibrated daily using the

standard calibration method provided by the manufacturer. The coefficient of variation for repeated scans of the manufacturer’s phantom is less than 1%.

Figure 11. A DXA Scan in Progress

Source: http://www.bodycompositioncenter.com/

To estimate level of physical activity, we asked participants to report the time they spent each week engaged in specific activities including walking, jogging, running, bicycling, aerobics/dancing, tennis/racket sports, swimming, yoga/pilates and weight training. These questions were based on those used in the Nurses’ Health

Study II and have been previously validated in that population [32]. We then calculated total metabolic equivalents (MET), which were recorded in hours of activity per week. The MET per week score measures the intensity of a physical activity, with 1 MET being defined as the amount of energy expended when a person is at rest [33]. We also used self-reported questionnaires to collect information on race, anthropometric and lifestyle factors, including smoking status, alcohol intake and use of oral contraceptives.

3.2.3 Dietary Assessment

Dietary intake of food and supplements in the previous two months was assessed using a modified version of the Harvard Food Frequency Questionnaire (FFQ), which has been extensively validated for use in U.S. women [34, 35]. Calcium and vitamin

D-rich food items on the modified questionnaire included regular and fortified dairy foods and fortified orange juice and breakfast cereals, dark meat fish, shellfish, leafy green vegetables and multivitamins or other supplements containing calcium or vitamin D. FFQs were analyzed at Harvard University, where nutrient intakes were calculated by multiplying the frequency of intake of a specified portion size of each food by its nutrient content and then summing across all food items. Adequacy of dietary intakes of calcium and vitamin D were determined by comparing reported intakes with the Recommended Dietary Allowance (RDA) for women in this age range [16]. Intake of dairy foods was assessed using self-reported intake of selected dairy foods (skim, low fat or whole milk, sherbet or ice milk, ice cream, yogurt, cottage cheese and other cheeses). Average number of servings per day was estimated by dividing the frequency of intake of each food by the number of days. For instance, if a participant reported eating yogurt once a week, the number of servings per day was estimated as 1/7 days = 0.14 servings per day.

3.2.4 Statistical Analysis

All analyses were completed using STATA version 12.0 (Stata Corporation., College

Station, TX.). The distribution of participant characteristics were assessed using descriptive statistics (means, variances, percentiles) and graphs (histograms and scatter plots (Appendix B). We compared participant characteristics by adequacy of

calcium and vitamin D intakes [16] using student t-tests and chi-squared tests. Only one subject reported an implausible total energy intake (> 5000 kcal/day); however, this participant was not excluded from our analyses as her physical activity level was high (>100 metabolic equivalents [METs]/week).

Total body fat percentage was categorized as “low” (TBF < 32%) or “high”

(TBF ≥ 32%) based on gender- and age-specific cut-offs [36] and also on the current

American Council on Exercise recommendation for maximum body fat percentage in women who are non-athletes [37]. BMI was classified as “low” (BMI < 25 kg/m2) or

“high” (BMI ≥ 25 kg/m2) based on current recommendations by the World Health

Organization [38] and National Heart, Lung and Blood Institute and the North

American Association for the Study of Obesity [39]. The waist circumference (WC) cut-off point for high central obesity was defined as WC ≥ 80 cm, based on guidelines proposed by the International Diabetes Federation [40].

We assessed the nature and strength of associations among covariates, dietary intakes, and adiposity variables using normal throery multiple linear regression.

Linear and restricted cubic spline model fit were also obtained to assess departure from linearity (Appendix C). Model adequacy was assessed using residual versus predicted values, and Cook’s distance plots were used to identify influential data points. Crude and adjusted associations between dietary calcium, vitamin D intakes and adiposity were estimated from unadjusted and multivariable adjusted linear regression. The dependent variable in these analyses was adiposity defined as total body fat percentage (TBF), body mass index (BMI) and waist circumference (WC).

Total energy and physical activity level were considered a priori to be likely confounders of adiposity-calcium-vitamin D relationships. Selected covariates were

considered for inclusion in our final multivariable models based on evidence in the literature suggesting their potential role as confounders or effect modifiers. We conducted model assessment with and without each of these covariates, and included in the final regression model any variable with P < 0.20 or led to at least a 10% change in the beta coefficient for each adiposity measure. Our final model was further evaluated for multicollinearity between covariates. This was necessary to ensure that the eﬀect of one variable in our model does not depend on the level of another variable in that same model.

Logistic regression was used to estimate odds ratios and 95% confidence intervals for the association between intakes of dietary calcium, vitamin D, dairy products, and event of high adiposity, defined as TBF ≥ 32%, BMI ≥ 30 kg/m2, and

WC ≥ 80 cm. Calcium and vitamin D were each evaluated as follows: total intake, intake from food sources only, and intake from supplements only. Covariates were included in the final regression model based on likelihood ratio tests and the changes they effected in the betas for the primary predictors. Selected important variables, including total energy intake and physical activity level, were included a priori. The

Hosmer-Lemeshow goodness of fit test for our main effects model had a P-value of

0.25, indicating adequate goodness of fit. To evaluate linear trends, we used a Mantel extension test, modeling the medians of each tertile of dietary intake as a continuous variable. All P-values were two-sided and considered statistically significant if P <

0.05.

3.3 Results

The characteristics of the study participants are presented in Table 7. The majority of participants were Caucasian (86.2%). About one-half (51.5%) of women had high adiposity (%TBF ≥ 32) while one-third (33.3%) had a WC measurement ≥ 80 cm.

About 21.1% were overweight or obese (BMI ≥ 25 kg/m2). Mean calcium and vitamin

D intakes were 1220 ± 616 mg/day and 375 ± 298 IU/day, respectively (Table 8). A large proportion of women (83.3%) had total vitamin D intakes less than the current

RDA of 600 IU/day (FNB/IOM, 2010), and about 40% of women had inadequate total calcium intakes (< RDA of 1000 mg/day).

Overall, 37.4% of women had low intakes of both vitamin D and calcium.

Women reporting adequate calcium or vitamin D intakes reported higher intakes of total energy, protein and fat (P < 0.01). From Table 9a, women with adequate calcium intakes were more likely to be non-smokers (P = 0.01) and exercised more than those with low total calcium intakes (P < 0.01). Compared to women with adequate vitamin

D intake, women with low vitamin D intakes were more likely to have high total body fat percentage (P = 0.04) or use oral contraceptive (P = 0.01), (Table 9b). Participants with low TBF (< 32 %) reported higher intakes of food vitamin D compared to women with TBF ≥ 32 % (257 ± 19 vs. 205 ± 12, P = 0.02), (Table 10b). Some of the major dietary contributors to vitamin D intake in our cohort were milk (28.0%), skim milk (12.1%), cream (16.1%), yogurt (10.9%), eggs (4.3%) and cold breakfast cereals

(1.5%). Dairy products thus accounted for over two-thirds of dietary vitamin D intake

(data not shown).

Study participants who were smokers had low BMI (< 25kg/m2), (Table 11a).

Correlation analyses (Table 13) indicated that adiposity measures were highly

correlated with each other; %TBF and BMI (r = 0.76; P < 0.01), %TBF and WC (r =

0.72; P < 0.01) and BMI and WC (r = 0.84; P < 0.01). Physical activity level was inversely correlated with %TBF (r = -0.14; P = 0.02), but not BMI (r = 0.01, P =

0.92) or WC (r = 0.01, P = 0.86). Physical activity level was also an independent predictor of total body fat (ß: -0.03 ± 0.1%, P = 0.02), (Table 14).

Energy adjusted and multivariable adjusted linear regression showed no significant associations between adiposity (TBF and BMI) and intakes of calcium or vitamin D (Table 15). Among all women, supplemental calcium intake (adjusted for total energy intake) was inversely associated with waist circumference (ß: -0.004 ±

0.002 cm, P = 0.045). However, the strength of this association became marginal after adjustment for physical activity level, smoking status and vitamin D intake (ß: -0.005

± 0.003 cm, P = 0.07). In the subset of calcium supplement users, indicated that for every 100 mg increase in calcium supplement intake, waist circumference was reduced by 0.8 ± 0.4 cm (P = 0.04).

Logistic regression analyses of the event of high adiposity (TBF ≥ 32 % , BMI

≥ 30kg/m2 , and WC ≥ 80 cm) (Tables 16 & 17), indicated that among all participants, there was a slight inverse (beneficial) trend in relative odds of high adiposity with increasing calcium or vitamin D intake; however, these associations were not statistically significant (P > 0.05). Intakes of calcium and vitamin D were significantly correlated (Table 13): total calcium and total vitamin D (r = 0.51, P <

0.01); food calcium and food vitamin D (r = 0.75, P < 0.01); calcium supplement and vitamin D supplement (r = 0.46, P < 0.01). We therefore adjusted vitamin D and calcium for each other as appropriate to assess the independent effects of each nutrient. Compared to women in the highest tertile of calcium supplement intake

(Table 17), women with lower calcium intake from supplements were more likely to have waist circumference ≥ 80 cm (OR = 2.04; 95% CI: 1.04 – 3.99). We also found evidence of a linear dose-relationship across tertiles of decreasing supplemental calcium intake and risk of high BMI (Ptrend = 0.02). From Table 18, both energy adjusted and multivariable logistic regression models showed that the risk of having a high body fat percentage (%TBF ≥ 32) was greater among women reporting a combination of high calcium intake (≥ 1000 mg/day) but low intake of vitamin D (<

600 IU/day), compared to women with high intakes of both calcium (≥ 1000 mg/day) and vitamin D (≥ 600 IU/day) [OR = 2.37; 95% CI: 1.09 - 5.12 and OR = 2.27; 95%

CI:1.02 – 5.04, respectively)

3.4 Discussion

In this cross-sectional study among college-aged women, linear regression analyses showed that low intake of calcium supplements was significantly associated with events of waist circumference ≥ 80 cm. Among young women 18-30 years old, a waist circumference greater than 80 cm indicates central obesity (i.e., an excess of abdominal fat) and suggests increased visceral adiposity. Visceral fat is more metabolically active than subcutaneous fat [41, 42], and large deposits can disrupt fatty acid metabolism, increasing the influx of free fatty acids entering splanchnic circulation [43, 44]. Excessive amounts of unbound fatty acids in the blood contribute to hyperlipidemia and hypertriglyceridemia, both of which can promote or exacerbate obesity-related chronic conditions [45-47]. Our findings suggest that increased intake of supplemental calcium may help reduce the risk of central obesity (high waist circumference) among pre-menopausal Caucasian women. However, the cross-

sectional design of our study precludes causal inferences (i.e., it is impossible to determine whether high waist circumference influenced calcium supplement intake or whether the reverse was true). Additional prospective and intervention studies utilizing larger cohorts are needed to further evaluate these associations and to examine the mechanism(s) by which calcium intake, especially calcium from dietary supplements, may modulate adiposity in normal- and overweight young women.

In our samples, women in the second tertile of total calcium intake (median intake of 1143 mg/day) had significantly higher risk of increased waist circumference measurement (OR = 2.27; 95% CI: 1.17 - 4.39), compared to women in the highest tertile of total calcium intake (median intake of 1658 mg/day). Similarly, total calcium intake was marginally associated with increased risk of high total body fat percentage

(OR = 1.82; 95% CI: 0.99 - 3.37) among women in the second tertile of total calcium intake, but this association was not present with respect to BMI. This observation possibly lends support to the notion that BMI is a poor proxy for estimating adiposity, especially in a cohort of relatively active young adult women, where high BMI (≥ 25 kg/m2) may lead to misclassification of heavily muscled women as overweight or obese [27-29]. In our cohort, waist circumference was strongly correlated with total body fat percentage (r = 0.72; P < 0.01). Since WC is less expensive and invasive as compared to dual energy x-ray absorptiometry (DXA), its use as a proxy for fat mass in youg women warantly further investigation.

Our logistic regression analyses showed that calcium intake (total calcium, calcium from food sources only and calcium supplements) was inversely correlated with %TBF, BMI, and WC, although these associations were not statistically significant. This finding may have been due to the fact that the ranges of BMI, WC,

and %TBF were relatively narrow in our study compared to those in other studies of older individuals. Mean intake of total calcium in our study (1220 ± 616 mg/day) was also above the RDA (1000 mg/day) for women 19 - 50 years, which contrasts with several reports from similar cohorts [15, 17, 18, 23, and 24]. In one example, Lin et al. conducted a prospective analysis of the effect of calcium intake on changes in body composition during a two-year exercise intervention among 54 sedentary but normal weight Caucasian US women (18 - 31 years) [22]. Dietary intake was assessed with 3-day diet records at baseline and every 6 months, and DXA was used to estimate body fat percentage and lean mass. They reported that total calcium intake, adjusted for total energy intake, was negatively correlated with changes in body weight (r = -0.35, P < 0.05) and body fat percentage (r = -0.34, P < 0.05). Mean calcium intake was also estimated as 781 ± 212 mg/day, which is lower than the current RDA for adult women.

In the Quebec Family Study (n = 235 women, ages 20 - 65 years), Jacqmain et al. observedthat percentage body fat, BMI and WC were lower in women consuming

600 mg or more of calcium per day (BF: 31.3 ± 0.9 % vs. 37.3 ± 1.6 %, P < 0.05;

BMI: 27.0 ± 0.7 vs. 31.8 ± 1.2 kg/m2, P < 0.05; and WC: 82.0 ± 1.6 vs. 93.6 ± 2.6 cm,

P < 0.05) as compared to women with lower intakes [24]. In our sample, conversely, when we used the current RDA of 1000 mg/day to assess adequacy of calcium intake in the present study, we found no significant differences in adiposity measures

(%TBF, BMI or WC) for women consuming ≥ 1000 mg/day of calcium compared to those consuming less calcium (%TBF: 32.0 ± 7.8 vs. 32.2 ± 8.7 %, P = 0.81; BMI:

22.7 ± 2.9 vs. 23.1 ± 3.7 kg/m2, P = 0.38; and WC: 76.9 ± 8.3 vs. 78.2 ± 9.9 cm, P =

0.26). After adjusting for age, energy intake, percentage dietary fat, dietary protein

intake and markers of socioeconomic status, Jacqmain et al. reported that dietary calcium intake (assessed with 3-day food records) was negatively correlated with computed tomography estimates of percentage body fat (r = -0.17, P < 0.05) [24]. We did not observe an association in our sample (r = -0.04, P = 0.47).

In our cohort, women reporting adequate intakes of calcium (≥ 1000 mg/day) but low intakes of vitamin D (< 600 IU/day) were twice as likely to have high body fat (%TBF ≥ 32), compared to those with adequate intakes of both calcium (≥ 1000 mg/day) and vitamin D (≥ 600 IU/day). In independent studies of agouti-mice as well as overweight/obese adult women, Zemel et al. have hypothesized that suppression of

1,25-dihydroxyvitamin D activation with high-calcium diets would reduce adipocyte intracellular Ca2+ concentration, inhibit fatty acid synthase activity and activate lipolysis, thus exerting an anti-obesity effect [17-20]. In our sample, we were unable to assess the relationship between vitamin D intake from food sources and serum concentrations of 25-hydroxyvitamin D3. However, previous findings from a sub- sample of our cohort (n = 186) suggested a statistically significant modest correlation between dietary vitamin D intake and serum 25-dihydroxy vitamin D3 levels (r =

0.18, P = 0.01)[49].

Women in our study were similar to each other with respect to some characteristics known to be associated with adiposity, including smoking, alcohol use, physical activity levels and intakes of total calcium and energy [50-54], but we still adjusted for these covariates as they could potentially confound any observed associations. However, it is still possible that there may be residual confounding by other unknown factors.

Few studies have examined the association between dietary vitamin D intake and adiposity. Our findings did not corroborate the observations of Tidwell and

Valliant, who recently reported an inverse correlation between dietary vitamin D intake and percentage body fat (measured using DXA) in African American women

[25]. Using cross-sectional data from 100 pre-menopausal women (18 - 40 years), they estimated partial correlations between vitamin D intake and body fat, controlling for fat, carbohydrate and protein intakes. In their study, women with lower intakes of dietary vitamin D (152 ± 36 IU/day) were statistically significantly more likely to have excessive body fat (%BF > 37.9 %) than were women with higher intakes of dietary vitamin D (200 ± 32 IU/day) (r = -0.46, P < 0.01). Our results may have differed from that of Tidwell & Valliant because of differences in sample characteristics, which could have precluded the discovery of associations between nutrient intake and adiposity. We also controlled for dietary calcium intake (which was highly correlated with vitamin D from food sources) in order to assess the independent effects of dietary vitamin D intake on adiposity. Tidwell and Valliant’s cohort included primarily overweight and obese African-American women (mean

BMI: 29.8 ± 6.9 kg/m2 and % BF: 36.9 ± 7.3 %). In contrast, women in our study were predominately Caucasian with relatively low BMI (22.9 ± 3.2 kg/m2) and TBF

(31.9 ± 7.8 %). This dissimilarity in women’s body composition may be central in defining how dietary calcium and vitamin D intake is related to adiposity in young women. Use of a comprehensive FFQ compared to a one-time 24-hr dietary recall

[25] also enabled us to assess usual dietary intake over a longer period, which may be more predictive of adiposity than recent intake.

In contrast to what has been reported in other studies [6, 17-20], we did not find any significant associations between intake of dairy foods and measures of adiposity (percentage body fat, BMI and waist circumference) in our cohort.

Some of the strengths of our study include a comprehensive assessment of nutrient intake using a validated food frequency questionnaire. Overall, we expected some degree of misclassification of dietary intake because it is challenging to accurately estimate dietary intake over time without error. Although it is especially difficult to estimate and recall past portion size accurately, research has shown that it is much easier for people to recall the frequency of intake, which is well correlated with portion size [55]. It is also unlikely that any misclassification of dietary intake in our study would be related to adiposity (total body fat percentage) in a systematic way, as participants were unaware of their body fat statistics prior to having the DXA scan.

We assessed adiposity using DXA scans, which have been previously validated and found to accurately differentiate between women with acceptable/normal body fat and those with high or excess adiposity [30, 31]. DXA scans estimate body composition by measuring the amount of energy absorbed from two x-ray beams through the body. These two beams have different energy levels, such that the differential absorption of the beans allows for distinction of bone mineral mass from soft tissue mass [56]. The accuracy of body composition measurement is therefore dependent in part on the accuracy of soft tissue measurements.

DXA assessment of body composition is largely accurate in non-obese persons; however, some studies have indicated that DXA measurements may be less accurate in obese individuals [57-58]. Although DXA is widely accepted as a

benchmark method for the measurement of body composition [59], there is some potential for non-differential misclassification of body composition. This is because all methods that assess body composition in vivo only provide indirect estimates of body composition. Only chemical analysis and cadaver dissection can offer accurate direct measures of body composition [60-61]. In our analyses, we included all participants with DXA scan information because body fat percentage was one of our outcomes of interest hence we cannot justifiably exclude any subject based on their body fat percentage. This study design therefore suggests that if DXA- estimated percentage body fat values are less accurately detected in obese persons (BMI > 30 kg/m2) then our findings related to percentage of body fat in this population of young women may be underestimated.

Some limitations of our study include the homogeneity of our population. Our population was comprised of mostly non-obese and few obese individuals and all participants were between 18 to 30 years. Subjects were also predominately were

Caucasian. Our findings are therefore limited to mostly non-obese young Caucasian women and may not be generalizable to women in other age groups, BMI categories, or ethnicities. In addition, because all exposures and outcome variables were assessed at one time-point, causal relationships cannot be inferred from our study. Prospective studies are needed to better undertand the role of dietary calcium and vitamin D intake in obesity development.

3.5 Conclusions and Significance

Calcium and vitamin D intakes have been linked with reduced adiposity in older adults and in obese women (body mass index, BMI, ≥ 30 kg/m2), but few studies

have examined these associations in healthy young women. Obesity is a precursor of a host of chronic conditions, including cardiovascular disease, type II diabetes mellitus, and several cancers. Understanding and identifying modifiable dietary factors associated with obesity (increased adiposity) is therefore important in formulating nutrition interventions targeted at obesity reduction/prevention. The results of this study suggest that even though obesity is generally associated with sedentary lifestyle and energy imbalance, micronutrient status may also influence obesity risk. Our findings indicate that a dietary high intake of both calcium and vitamin D may protect against high body fat percentage, even among non-obese young women. Women with lower intakes of total and supplemental calcium were also showed increased risk of high waist circumference, an indicator of central obesity, compared to women with higher calcium intakes. Globally, obesity is a growing public health concern; and the possibility that adiposity risk in young women may be influenced by a high intake of calcium, especially calcium obtained from supplements, warrants further study.

3.6 References

1. Khadilkar A, Kadam N, Chiplonkar S, Fischer PR, Khadilkar V. School-based

Calcium vitamin D with micronutrient supplementation enhances bone mass in

underprivileged Indian premenarchal girls. Bone 2012; 51: 1-7.

2. Cheng S, Lyytikainen A, Kroger H, Lamberg-Allardt C, Alen M, Koistinen A, Wang

QJ, Suuriniemi M, Suominen H, Mahonen A, Nicholson P, Ivaska KK, Korpela R,

Ohlsson C, Vaananen KH, Tylavsky F. Effects of calcium, dairy product, and vitamin

D supplementation on bone mass accrual and body composition in 10-12-y-old girls: a

2-y randomized trial. Am J Clin Nutr 2005; 82: 1115-26.

3. Lanou AJ, Berkow SE, Barnard ND. Calcium, dairy Products, and Bone Health in

Children and Young Adults: A reevaluation of the evidence. Pediatrics 2005; 115:

736-43.

4. Loos RJF, Rankinen T, Leon AS, Skinner JS, Wilmore JH, Rao DC, Bouchard C.

Calcium Intake Is Associated with Adiposity in Black and White Men and White

Women of the HERITAGE Family Study. J Nutr 2004; 134: 1772-8.

5. Lovejoy JC, Champagne CM, Smith SR, de Jonge L, Xie H. Ethnic differences in

dietary intakes, physical activity, and energy expenditure in middle-aged,

premenopausal women, the Healthy Transitions Study. Am J Clin Nutr 2001; 74: 90-

6. Zemel MB, Richards J, Milstead A, Campbell P. Effects of calcium and dairy on body

composition and weight loss in African American adults. Obes Res 2005; 13: 1218-

25.

7. Briefel RR. Secular trends in dietary intake in the United States. Annu Rev Nutr 2004;

24: 401-31.

8. Miller GD, Jarvis JK, McBean LD. Handbook of dairy foods and nutrition. Third

Edition (2007). McBean CRC Press. Boca Raton, FL.

9. Weaver CM, Plawecki KL. Dietary calcium adequacy of a vegetarian diet. Am J Clin

Nutr 1994; 59: 1238-41S.

10. Grant WB, Holick MF. Benefits and requirements of vitamin D for optimal health, a

review. Altern Med Rev 2005; 10: 94-111.

11. Chen TC, Chimeh F, Lu Z, Mathieu J, Person KS, Zhang A, Kohn N, Martinello S,

Berkowitz R, Holick MF. Factors that influence the cutaneous synthesis and dietary

sources of vitamin D. Arch Biochem Biophys 2007; 460: 213-7.

12. Holick MF, Chen TC. Vitamin D deficiency, a worldwide problem with health

consequences. Am J Clin Nutr 2008; 87: 1080-6S.

13. Calvo MS, Whiting SJ, Barton CN. Vitamin D fortification in the United States and

Canada, current status and data needs. Am J Clin Nutr 2004; 80: 1710-6S.

14. Phillips KM, Ruggio DM, Horst RL, Minor B, Simon RR, Feeney MJ, Byrdwell WC,

Haytowitz DB. Vitamin D and sterol composition of 10 types of mushrooms from

retail suppliers in the United States. J Agric Food Chem 2011; 59: 7841-53.

15. Bailey RL, Dodd KW, Goldman JA, Gahche JJ, Dwyer JT, Moshfegh AJ, Sempos

CT, Picciano MF. Estimation of total usual calcium and vitamin D intakes in the

United States. J Nutr 2010; 140: 817-22.

16. Food and Nutrition Board / Institute of Medicine. Dietary Reference Intakes for

Calcium and Vitamin D (2010). 2101 Constitution Avenue, N.W. Washington,

DC 20418. National Academy Press.

17. Zemel MB. Mechanisms of dairy modulation of Adiposity. J Nutr 2003; 133: 252-6S.

18. Zemel MB. Role of dietary calcium and dairy products in modulating adiposity.

Lipids 2003; 38: 139-46.

19. Zemel MB, Shi H, Greer B, Dirienzo D, Zemel PC. Regulation of adiposity by dietary

calcium. FASEB J 2000; 14: 1132-8.

20. Zemel MB, Warren T, Milstead A, Morris K, Campbell P. Calcium and dairy

acceleration of weight and fat loss during energy restriction in obese adults. Obes Res

2004; 12: 582-90.

21. Davies KM, Heaney RP, Recker RR, Lappe JM, Barger-Lux MJ, Rafferty K, Hinders

S. Calcium Intake and Body Weight. J Clin Endocrinol Metab 2000; 85: 4635-8.

22. Lin YC, Lyle RM, McCabe LD, McCabe GP, Weaver CM, Teegarden D. Dairy

calcium is related to changes in body composition during a two-year exercise

intervention in young women. J Am Coll Nutr 2000; 19: 754-60.

23. Buchowski MS, Semenya J, Johnson AO. Dietary calcium intake in lactose

maldigesting intolerant and tolerant African-American Women. J Am Coll Nutr 2002;

21: 47-54.

24. Jacqmain M, Doucet E, Despres JP, Bouchard C, Tremblay A. Calcium intake, body

composition, and lipoprotein-lipid concentrations in adults. Am J Clin Nutr 2003; 77:

1448-52.

25. Tidwell DK, Valliant MW. Higher amounts of body fat are associated with inadequate

intakes of calcium and vitamin D in African American women. Nutr Res 2011; 31:

527-36.

26. Ha E, Zemel MB. Functional properties of whey, whey components, and essential

amino acids, mechanisms underlying health benefits for active people. J Nutr

Biochem 2003; 14: 251-58.

27. Roche AF, Sievogel RM, Chumlea WC, Webb P. Grading body fatness from limited

anthropometric data. Am J Clin Nutr 1981; 34: 2831-8.

28. Wellens RI, Roche AF, Khamis HJ, Jackson AS, Pollock ML, Siervogel RM.

Relationships between the body mass index and body composition. Obes Res 1996; 4:

35-44.

29. Flegal KM, Shepherd JA, Looker AC, Graubard BI, Borrud LG, Ogden CL, Harris

TB, Everhart JE, Schenker N. Comparisons of percentage body fat, body mass index,

waist circumference, and waist-stature ratio in adults. Am J Clin Nutr 2009; 89: 500-8.

30. Haarbo J, Gotfredsen A, Hassager C, Christiansen C. Validation of body composition

by dual energy X-ray absorptiometry (DEXA). Clin Physiol 1991; 11: 331-41.

31. Bolanowski M, Nilsson BE. Assessment of human body composition using dual-

energy x-ray absorptiometry and bioelectrical impedance analysis. Med Sci Monit

2001; 7: 1029-33.

32. Wolf AM, Hunter DJ, Colditz GA, Manson JE, Stamoer MJ, Corsano KA, Rosner B,

Kriska A, Willet WC. Reproducibility and validity of a self-administered physical

activity questionnaire. Int J Epidemiol 1994; 23: 991-9.

33. Ainsworth BE, Haskell WL, Leon AS, Jacobs DRjr, Montoye HJ, Sallis JF,

Paffenbarger RS jr. Compendium of physical activities, classification of energy costs

of human physical activities. Med Sci Sports Exerc 1993; 25: 1575-81.

34. Willet WC, Sampson L, Stampler MJ, Rosner B, Bain C, Witschi J, Hennekens CH,

Speizer FE. Reproducibility and validity of a semi-quantitative food frequency

questionnaire. Am J Epidemiol 1985; 122: 51-65.

35. Feskanich D, Rimm EB, Giovannucci EL, Colditz GA, Stampfer MJ. Reproducibility

and validity of food intake measurements from a semi-quantitative food frequency

questionnaire. J Am Diet Assoc 1993; 93: 790-6.

36. Jackson AS, Pollack ML. Practical assessment of body composition. Physician and

Sportsmed 1985; 13: 76-90.

37. American Council on Exercise. What are the guidelines for percentage of body fat

loss, 2009? Available online: http://www.acefitness.org/blog/112/what-are-the

guidelines-for-percentage-of-body-fat (accessed 30th March, 2013).

38. World Health Organization. Preventing and Managing the Global Epidemic, 2000.

WHO, Geneva. WHO Obesity Technical Report Series, Obesity, Number 894.

39. National Heart, Lung, and Blood Institute and the North American Association for the

Study of Obesity. The Practical Guide: Identification, Evaluation, and Treatment of

Overweight and Obesity in Adults (2000). NIH Publication Number 00-4084.

Available online: http://www.nhlbi.nih.gov/guidelines/obesity/prctgd_c.pdf (accessed

30th March, 2013).

40. International Diabetes Federation. The IDF Consensus Worldwide Definition of the

Metabolic Syndrome (2006), Avenue Emile De Mot 19, B-1000 Brussels, Belgium.

Available online: http//www.idf.org/webdata/docs/IDF_Meta_def_final.pdf (accessed

30th March, 2013).

41. Mauriege P, Galitzky J, Berlan M, Lafontan M. Heterogeneous distribution of beta

and alpha-2 adrenoceptor binding sites in human fat cells from various fat deposits:

functional consequences. Eur J Clin Invest 1987; 17: 156-65.

42. Rebuffe-Scrive M, Andersson B, Olbe L, Bjorntorp P. Metabolism of adipose tissue

in intra-abdominal depots of non-obese men and women. Metabolism 1989; 38: 453-

43. Bergman BC, Cornier MA, Horton TJ, Bessesen DH. Effects of fasting on insulin

action and glucose kinetics in lean and obese men and women. Am J Physiol

Endocrinol Metab 2007; 293: 1103-11E.

44. Nielsen S, Guo Z, Johnson CM, Hensrud DD, Jensen MD. Splanchnic lipolysis in

human obesity. J Clin Invest 2004; 113: 1582-8.

45. Zhu S, Wang Z, Heshka S, Heo M, Faith MS, Heymsfield SB. Waist circumference

and obesity-associated risk factors among whites in the third National Health and

Nutrition Examination Survey: clinical action thresholds. Am J Clin Nutr 2002; 76:

743-9.

46. Clark VR, Greenberg B, Harris TS, Carson BL. Body mass index and waist

circumference predictors of cardiovascular risk in African Americans. Ethn Dis

2012; 22: 162-7.

47. Shields M, Tremblay MS, Connor Gorber S, Janssen I. Abdominal obesity

and cardiovascular disease risk factors within body mass index categories. Health Rep

2012; 23: 7-15.

48. Lin YC, Lyle RM, McCabe LD, McCabe GP, Weaver CM, Teegarden D. Dairy

calcium is related to changes in body composition during a two-year exercise

intervention in young women. J Am Coll Nutr 2000; 19: 754-60.

49. Bertone-Johnson ER, Chocano-Bedoya PO, Zagarins SE, Micka AE, Ronnenberg

AG. Dietary vitamin D intake, 25-hydroxyvitamin D3 levels and premenstural

syndrome in a college-aged population. J Steroid Biochem Mol Biol 2010; 121: 434-7.

50. Taylor RW, Keil D, Gold EJ, Williams SM, Goulding A. Body mass index, waist

girth, and waist-to-hip ratio as indexes of total and regional adiposity in women,

evaluation using receiver operating characteristic curve. Am J Clin Nutr 1988; 67: 44-

51. Kimm SYS, Glynn NW, Aston CE, Poehlman ET, Daniels SR. Effects of Race,

Cigarette Smoking, and Use of Contraceptive Medications on Resting Energy

Expenditure in Young Women. Am J Epidemiol 2001; 154: 718-24.

52. Rolls BJ, Morris EL, Roe LS. Portion size of food affects energy intake in normal-

weight and overweight men and women. Am J Clin Nutr 2002; 76: 1207-13.

53. Parikh SJ, Yanovski JA. Calcium intake and adiposity. Am J Clin Nutr 2003; 77: 281-

54. Dorn JM, Hovey K, Muti P, Freudenheim JL, Russell M, Nochajski TH, Trevisan M.

Alcohol Drinking Patterns Differentially Affect Central Adiposity as Measured by

Abdominal Height in Women and Men. J Nutr 2003; 133: 2655-62.

55. Willett W. Nutritional epidemiology. 2nd Edition. New York, New York: Oxford

University Press, 1998. pp 74-156.

56. Berger A. Bone mineral density scans. BMJ 2002; 325: 484.

57. Svendsen OL, Haarbo J, Hassager C, Christiansen C. Accuracy of measurements of

body composition by dual-energy x-ray absorptiometry in vivo. Am J Clin Nutr 1993;

57: 605-8.

58. Svendsen OL, Hassager C, Skodt V, Christiansen C. Impact of soft tissue on in vivo

accuracy of bone mineral measurements in the spine, hip, and forearm: A human

cadaver study. J Bone Miner Res 1995; 10: 868-73.

59. Ellis KJ. Human body composition: In vivo methods. Physiol Rev. 2000; 80: 649-80.

60. Clary JP, Martin AD, Drinkwater DT, Marfell-Jones MJ. The skinfold: Myth and

reality. J Sports Sci 1987; 5:3-33.

61. Martin AD, Drinkwater DT. Variability in the measures of body fat, assumptions or

technique? Sports Med 1991; 11:277-88.

Table 7. Characteristics of Participants: UMass Amherst Vitamin D Status Study, 2006 – 2011

Covariate Mean ± Standard Deviation Min - Max Age, years 21.4 ± 2.8 18.0 - 30.9 BMI, kg/m2 22.9 ± 3.3 16.7 - 37.7 Waist Circumference, cm 77.4 ± 9.0 58.4 - 114.3 Waist : Height, ratio 0.5 ± 0.1 0.4 - 0.7 Total Body Fat, % 32.1 ± 8.2 14.8 - 56.6 Arm Fat, % 30.6 ± 9.3 8.2 - 52.5 Leg Fat, % 35.4 ± 7.5 8.7 - 54.2 Trunk Fat, % 31.9 ± 9.6 11.7 - 60.7 Android Fat, % 34.5 ± 10.7 12.5 - 62.9 Gynoid Fat, % 40.5 ± 6.7 16.1 - 59.9 Android : Gynoid Fat, ratio 0.8 ± 0.2 0.4 - 1.8 Physical Activity, MET-hrs/week╔ 49.6 ± 43.4 0 - 206.9 Systolic Blood Pressure, mm Hg 108.1 ± 10.8 77 - 160 Diastolic Blood Pressure, mm Hg 69.9 ± 7.7 50 - 105 Fasting Blood Glucose, mg/dLϖ 81.8 ± 10.0 42 - 134 Serum 25-OHD, nmol/L‡ 79.2 ± 32.3 21.7 - 174.3 ╔ MET, metabolic equivalent hours of activity per week, n = 261 for physical activity. 50 MET-hrs / week is approximately equivalent to about 25-30 minutes of moderate intensity exercise per day, e.g., walking outdoors or on a treadmill. ϖ n = 257 for fasting blood glucose. ‡ n = 174 for serum 25-OHD.

Table 7. Continued

Covariate n (%) Race┴ Caucasian 232 (86.2) Other 37 (13.8) Education Some College 224 (83.0) College Graduate 46 (17.0) Currently Drink Alcohol No 70 (25.9) Yes 200 (74.1) Currently Smoke┴ No 254 (94.4) Yes 15 (5.6) Currently Use Oral Contraceptive No 155 (57.4) Yes 115 (42.6) BMI, kg/m2 < 25 213 (78.9) ≥ 25 57 (21.1) Waist Circumference, cm < 80 180 (66.7) ≥ 80 90 (33.3) Total Body Fat, % < 32 131 (48.5) ≥ 32 139 (51.5) Android: Gynoid Fat, ratio < 1 202 (74.8) ≥ 1 68 (25.2) Serum 25-OHD, nmol/L‡ < 50 28 (16.1) 50 - 75 65 (37.4) ≥ 75 81 (46.5) ‡ n = 174 for serum 25-OHD. ┴ n = 269 for race and smoking status.

Table 8. Unadjusted Dietary Intakes of Participants

Mean ± Min - Max Dietary Intake Standard Error Energy, kcal/day 2192 ± 54 474 - 6765 Total Protein, g/day 98 ± 3 22 - 315 Total Fat, g/day 72 ± 2 15 - 236 Total Calcium Intake, mg/day 1220 ± 37 224 - 4601 Food calcium 1091 ± 35 180 - 4601 Supplemental calcium (n = 93 users) 377 ± 30 11 - 1334 Vitamin A, IU/day 16120 ± 651 791 - 79213 Thiamin, mg/day 1.9 ± 0.1 0.1 - 6.1 Riboflavin, mg/day 2.6 ± 0.1 0.4 - 9.8 Niacin, mg/day 27.5 ± 0.9 1.3 - 101.7 Pantothenic acid, mg/day 6.7 ± 0.2 0.5 - 34.9 Vitamin B6, mg/day 2.5 ± 0.1 0.1 - 10.4 Folate, µg/day 661 ± 23 28 - 2338 Vitamin B12, µg/day 6.3 ± 0.3 0.1 - 34.3 Vitamin C, mg/day 153 ± 6 1 - 1054 Total Vitamin D Intake, IU/day 375 ± 18 3 - 1806 Food vitamin D 230 ± 11 3 - 1006 Supplemental vitamin D (n = 110 users) 357 ± 26 29 - 1400 Vitamin E, mg/day 152 ± 143 1 - 38535 Iron, mg/day 17.9 ± 0.6 1.3 - 73.0 Zinc, mg/day 13.6 ± 0.4 2.9 - 54.2 Linoleic acid, g/day 24.2 ± 0.8 4.6 - 102.2 Linolenic acid, g/day 12.7 ± 0.4 1.7 - 70.8 Dairy Intake, total servings/day₸ 1.9 ± 0.1 0 - 7.2 § IU, International Units: 1 IU = 0.025 µg. ₸ n = 169 for dairy intake responders. 1 serving dairy = 8 fluid oz. (1 cup) of milk or yogurt; 1.5 oz of natural cheese; 2 oz of processed cheese, 2 cups of cottage cheese.

Table 9a. Variations in Characteristics of Participants by Adequacy of Total Calcium Intake Ϫ, Defined by Recommended Dietary Allowance (RDA)

Total Calcium Intake (mg/day) Low (< 1000) Adequate (≥ 1000) P ⱶ Covariate (n = 108) (n = 162) Mean ± Standard Deviation Age, y 21.4 ± 2.8 21.4 ± 2.8 0.97 Waist Circumference, cm 78.2 ± 9.9 76.9 ± 8.3 0.26 < 80 (n = 180) 72.4 ± 4.1 72.3 ± 5.0 0.95 ≥ 80 (n = 90) 88.8 ± 8.4 86.5 ± 5.1 0.12 BMI, kg/m2 23.1 ± 3.7 22.7 ± 2.9 0.38 < 25 (n = 213) 21.5 ± 2.1 21.6 ± 1.8 0.59 ≥ 25 (n = 57) 28.4 ± 2.9 27.2 ± 2.4 0.08 Total Body Fat, % 32.2 ± 8.7 32.0 ± 7.8 0.81 < 32 (n = 131) 25.1 ± 4.3 25.4 ± 4.2 0.66 ≥ 32 (n = 139) 38.9 ± 6.0 38.2 ± 4.9 0.49 Physical Activity, MET-hrs/week╧ 39.5 ± 35.0 56.6 ± 47.3 < 0.01 Energy, kcal/day 1714 ± 621 2511 ± 885 < 0.01 Total Protein, g/day 77 ± 42 111 ± 43 < 0.01 Animal Protein 46 ± 38 63 ± 36 < 0.01 Plant Protein 31 ± 19 48 ± 28 < 0.01 Total Fat, g/day 58 ± 25 82 ± 35 < 0.01 Animal Fat 26 ± 16 36 ± 21 < 0.01 Non-animal Fat 32 ± 16 46 ± 28 < 0.01 Ϫ RDA for women ages 19 – 50 years is 1000 mg/day. This is the average daily level of intake sufficient to meet the nutrient requirements of nearly all (97% - 98%) healthy individuals. ⱶ P-values for any difference across categories estimated using student t-test. ╧ 50 MET-hrs / week is equivalent to about 25-30 minutes of moderate intensity exercise per day, e.g., walking outdoors or on a treadmill.

Table 9a. Continued

Total Calcium Intake (mg/day) Low (< 1000) Adequate (≥ 1000) P ⱶ Covariate (n = 108) (n = 162) n (%) Race┴ 0.68 Caucasian 92 (85.2) 140 (87.0) Other 16 (14.8) 21 (13.0) Education 0.64 Some College 91 (82.1) 186 (82.7) College Graduate 17 (17.9) 39 (17.3) Currently Drink Alcohol 0.26 No 24 (22.2) 46 (28.4) Yes 84 (77.8) 116 (71.6) Currently Smoke ┴ 0.01 No 96 (89.7) 158 (97.5) Yes 11 (10.3) 4 (2.5) Currently Use Oral Contraceptive 0.45 No 65 (60.2) 90 (55.6) Yes 43 (39.8) 72 (44.4) BMI, kg/m2 0.50 < 25 83 (76.9) 130 (80.2) ≥ 25 25 (23.1) 32 (19.8) Waist Circumference, cm 0.60 < 80 70 (64.8) 110 (67.9) ≥ 80 38 (35.2) 52 (32.1) Total Body Fat, % 0.92 < 32 52 (48.1) 79 (48.8) ≥ 32 56 (51.9) 83 (51.2) Android: Gynoid Ratio 0.95 < 1 81 (75.0) 121 (74.7) ≥ 1 27 (25.0) 41 (25.3) Serum 25-OHD, nmol/L‡,§ 0.89 < 75 40 (54.0) 47 (51.7) ≥ 75 34 (46.0) 44 (48.4) ⱶ P-values for any difference across categories estimated using chi-square test for proportions. ┴ N = 269 for smoking status.

Table 9b. Variations in Characteristics of Participants by Adequacy of Total Vitamin D Intake Ϫ, Defined by Recommended Dietary Allowance (RDA)

Total Vitamin D Intake (IU/day) § Low (< 600) Adequate (≥ 600) P ⱶ Covariate (n = 225) (n = 45) Mean ± Standard Deviation Age, y 21.3 ± 2.7 21.7 ± 3.5 0.41 Waist Circumference, cm 77.7 ± 9.0 75.8 ± 8.8 0.19 < 80 (n = 180) 72.5 ± 4.4 71.8 ± 5.5 0.42 ≥ 80 (n = 90) 87.4 ± 7.1 88.2 ± 3.9 0.70 BMI, kg/m2 22.9 ± 3.3 22.6 ± 3.2 0.54 < 25 (n = 213) 21.6 ± 1.9 21.5 ± 1.9 0.71 ≥ 25 (n = 57) 27.7 ± 2.7 27.9 ± 2.3 0.89 Total Body Fat, % 32.3 ± 8.1 30.8 ± 8.6 0.24 < 32 (n = 131) 25.3 ± 4.4 25.2 ± 3.8 0.91 ≥ 32 (n = 139) 38.3 ± 5.2 39.9 ± 5.9 0.24 Physical Activity, MET-hrs/week₸ 47.7 ± 41.2 59.4 ± 52.8 0.11 Energy, kcal/day 2094 ± 769 2684 ± 1197 < 0.01 Total Protein, g/day 93 ± 40 123 ± 63 < 0.01 Animal Protein 53 ± 34 73 ± 50 < 0.01 Plant Protein 40 ± 26 50 ± 24 0.01 Total Fat, g/day 70 ± 32 84 ± 40 0.02 Animal Fat 31 ± 18 38 ± 26 0.02 Non-animal Fat 39 ± 25 45 ± 24 0.13 Ϫ RDA for women ages 19 – 50 years is 600 IU/day. This is the average daily level of intake sufficient to meet the nutrient requirements of nearly all (97% - 98%) healthy individuals. ⱶ P-values for any difference across categories estimated using student t-test. § IU (International Units): 1 IU = 0.025 µg. ₸ 50 MET-hrs / week is approximately equivalent to about 25-30 minutes of moderate intensity exercise per day, e.g., walking outdoors or on a treadmill.

Table 9b.Continued

Total Vitamin D Intake (IU/day) § Low (< 600) Adequate (≥ 600) P ⱶ Covariate (n = 225) (n = 45) n (%) Race┴ 0.70 Caucasian 194 (86.6) 38 (84.4) Other 30 (13.4) 7 (15.6) Education 0.77 Some College 186 (82.7) 38 (84.4) College Graduate 39 (17.3) 7 (15.6) Currently Drink Alcohol 0.56 No 60 (26.7) 10 (22.2) Yes 165 (73.3) 35 (77.8) Currently Smoke ┴ 0.29 No 213 (95.1) 41 (91.1) Yes 11 (4.9) 4 (8.9) Currently Use Oral Contraceptive 0.01 No 137 (60.9) 18 (40.0) Yes 88 (39.1) 27 (60.0) BMI, kg/m2 0.55 < 25 176 (78.2) 37 (82.2) ≥ 25 49 (21.8) 8 (17.8) Waist Circumference, cm 0.17 < 80 146 (64.9) 34 (75.6) ≥ 80 79 (35.1) 11 (24.4) Total Body Fat, % 0.04 < 32 103 (45.8) 28 (62.2) ≥ 32 122 (54.2) 17 (37.8) Android: Gynoid Ratio 0.80 < 1 169 (75.1) 33 (73.3) ≥ 1 56 (24.9) 12 (26.7) Serum 25-OHD, nmol/L‡,§ 0.84 < 75 77 (53.1) 16 (51.2) ≥ 75 68 (44.9) 13 (44.8) ⱶ P-values for any difference across categories estimated using chi-square test for proportions. § IU (International Units): 1 IU = 0.025 µg. ┴ N = 269 for smoking status.

Table 10a. Variations in Characteristics of Participants by Total Body Fat (TBF) Percentage

Low TBF < 32 % High TBF ≥ 32 % P ⱶ Covariate (n = 131) (n = 139) Mean ± Standard Deviation Age, y 21.5 ± 3.0 21.3 ± 2.7 0.60 BMI, kg/m2 21.0 ± 2.0 24.6 ± 3.2 < 0.01 Waist Circumference, cm 72.2 ± 5.8 82.3 ± 8.7 < 0.01 Waist to Height, ratio 0.44 ± 0.03 0.50 ± 0.05 < 0.01 Total Body Fat, % 25.3 ± 4.2 38.5 ± 6.8 < 0.01 Arm Fat, % 23.4 ± 5.9 37.3 ± 6.3 < 0.01 Leg Fat, % 29.5 ± 4.8 41.0 ± 4.8 < 0.01 Trunk Fat, % 24.1 ± 5.0 39.2 ± 6.9 < 0.01 Android Fat, % 26.1 ± 6.1 42.5 ± 7.4 < 0.01 Gynoid Fat, % 35.4 ± 4.6 45.4 ± 4.4 < 0.01 Android : Gynoid Fat, ratio 0.74 ± 0.16 0.94 ± 0.12 < 0.01 Physical Activity, MET-hrs/week╔ 57.4 ± 46.9 42.8 ± 39.1 < 0.01 Systolic Blood Pressure, mm Hg 105.9 ± 10.0 110.1 ± 11.2 < 0.01 Diastolic Blood Pressure, mm Hg 68.3 ± 7.9 71.4 ± 7.1 < 0.01 Fasting Blood Glucose, mg/dL ϖ 80.1 ± 10.1 83.4 ± 9.7 0.01 Serum 25-OHD, nmol/L‡,§ 78.0 ± 31.9 80.4 ± 32.8 0.63 ⱶ P-values for test of Null: Equal means, estimated with student t-test. ╔ MET, metabolic equivalent of activity hours per week, n = 261 for physical activity. ϖ n = 257 for fasting blood glucose. ‡ n = 174 for serum 25-OHD, § 2.496 nmol/L = 1 ng/mL.

Table 10a. Continued

Low TBF < 32 % High TBF ≥ 32 % P ⱶ Covariate (n = 131) (n = 139) n (%) Race┴ 0.69 Caucasian 111 (85.4) 121 (87.0) Other 19 (14.6) 18 (13.0) Education 0.83 Some College 108 (82.4) 116 (83.5) College Graduate 23 (17.6) 23 (16.5) Currently Drink Alcohol 0.99 No 34 (26.0) 36 (25.9) Yes 97 (74.0) 103 (74.1) Currently Smoke┴ 0.14 No 120 (92.3) 134 (96.4) Yes 10 (7.7) 5 (3.6) Currently Use Oral Contraceptive 0.35 No 79 (60.3) 76 (54.7) Yes 52 (39.7) 63 (45.3) BMI, kg/m2 < 0.01 < 25 125 (95.4) 88 (63.3) ≥ 25 6 (4.6) 51 (36.7) Waist Circumference, cm < 0.01 < 80 119 (90.8) 61 (43.9) ≥ 80 12 (9.2) 78 (56.1) Android: Gynoid Ratio < 0.01 < 1 126 (96.2) 76 (54.7) ≥ 1 5 (3.8) 63 (45.3) Serum 25-OHD, nmol/L‡,§ 0.62 < 75 46 (55.4) 47 (51.7) ≥ 75 37 (44.6) 44 (48.4) ⱶ P-values for chi-squared tests. ┴ n = 269 for race and smoking status. ‡ n = 174 for serum 25-OHD, § 2.496 nmol/L = 1 ng/mL.

Table 10b. Variations in Dietary Intakes of Participants by TBF Percentage

Low TBF High TBF P ⱶ < 32 % ≥ 32 % Covariate (n = 131) (n = 139) Mean ± Standard Error Energy, kcal/day 2152 ± 84 2136 ± 68 0.28 Total Protein, g/day 102 ± 4 93 ± 4 0.13 Total Fat, g/day 75 ± 3 70 ± 3 0.30 Total Calcium Intake, mg/day 1281 ± 63 1163 ± 42 0.12 Food calcium 1157 ± 59 1028 ± 40 0.07 Supplemental calcium (n = 93 users) 370 ± 46 383 ± 40 0.82 Vitamin A, IU/day 17758 ± 1072 14731 ± 744 0.02 Thiamin, mg/day 2.0 ± 0.1 1.7 ± 0.1 0.01 Riboflavin, mg/day 2.8 ± 0.1 2.4 ± 0.1 0.01 Niacin, mg/day 29 ± 1 26 ± 1 0.04 Pantothenic acid, mg/day 7.3 ± 0.4 6.1 ± 0.2 0.01 Vitamin B6, mg/day 2.7 ± 0.1 2.4 ± 0.1 0.06 Folate, µg/day 724 ± 38 601 ± 25 0.01 Vitamin B12, µg/day 6.9 ± 0.5 5.8 ± 0.3 0.04 Vitamin C, mg/day 163 ± 11 143 ± 7 0.12 Total Vitamin D Intake, IU/day§ 409 ± 29 344 ± 22 0.07 Food vitamin D 257 ± 19 205 ± 12 0.02 Supplemental vitamin D (n = 110 users) 363 ± 39 350 ± 35 0.81 Iron, mg/day 19 ± 1 17 ± 1 0.01 Zinc, mg/day 15 ± 0.7 13 ± 0.4 0.02 Linoleic acid, g/day 25 ± 1 24 ± 1 0.37 Linolenic acid, g/day 14 ± 1 12 ± 1 0.09 Dairy Intake, total servings/day₸ 1.9 ± 0.2 1.9 ± 0.1 0.87 ⱶ P-values for test of Null: Equal means, estimated with student t-tests. § IU, International Units: 1 IU = 0.025 µg. ₸ n = 169 for dairy intake responders. 1 serving dairy = 8 fluid oz. (1 cup) of milk or yogurt; 1.5 oz of natural cheese; 2 oz of processed cheese, 2 cups of cottage cheese.

Table 11a. Variations in Characteristics of Participants by Body Mass Index (BMI)

Low BMI < 25 kg/m2 High BMI ≥ 25 kg/m2 P ⱶ Covariate (n = 213) (n = 57) Mean ± Standard Deviation Age, y 21.4 ± 2.8 21.5 ± 2.9 0.85 BMI, kg/m2 21.6 ± 1.9 27.7 ± 2.7 < 0.01 Waist Circumference, cm 74.3 ± 6.2 89.1 ± 8.0 < 0.01 Waist to Height, ratio 0.45 ± 0.04 0.54 ± 0.05 < 0.01 Total Body Fat, % 29.7 ± 6.7 40.9 ± 7.2 < 0.01 Arm Fat, % 28.4 ± 8.4 38.6 ± 8.0 < 0.01 Leg Fat, % 33.7 ± 6.4 41.9 ± 7.6 < 0.01 Trunk Fat, % 28.9 ± 7.7 42.9 ± 8.0 < 0.01 Android Fat, % 31.4 ± 8.9 46.4 ± 8.1 < 0.01 Gynoid Fat, % 39.1 ± 5.8 46.1 ± 7.0 < 0.01 Android : Gynoid Fat, ratio 0.8 ± 0.1 1.0 ± 0.1 < 0.01 Physical Activity, MET-hrs/week╔ 48.9 ± 41.5 52.2 ± 50.3 0.61 Systolic Blood Pressure, mm Hg 107.1 ± 10.7 111.8 ± 10.7 < 0.01 Diastolic Blood Pressure, mm Hg 69.2 ± 7.6 72.6 ± 7.3 < 0.01 Fasting Blood Glucose, mg/dL ϖ 80.6 ± 10.3 86.2 ± 7.4 < 0.01 Serum 25-OHD, nmol/L‡,§ 77.6 ± 31.1 84.5 ± 35.8 0.23 ⱶ P-values for test of Null: Equal means, estimated with student t-tests. ╔ MET, metabolic equivalent of activity hours per week, n = 261 for physical activity. ϖ n = 257 for fasting blood glucose. ‡ n = 174 for serum 25-OHD, § 2.496 nmol/L = 1 ng/mL.

100

Table 11a. Continued

Covariate Low BMI < 25 High BMI ≥ 25 P ⱶ kg/m2 kg/m2 (n = 213) (n = 57) n (%) Race┴ 0.62 Caucasian 184 (86.8) 48 (84.2) Other 28 (13.2) 9 (15.8) Education 0.91 Some College 177 (83.1) 47 (82.5) College Graduate 36 (16.9) 10 (17.5) Currently Drink Alcohol 0.79 No 56 (26.3) 14 (24.6) Yes 157 (73.7) 43 (75.4) Currently Smoke ┴ 0.04 No 197 (92.9) 57 (100.0) Yes 15 (7.1) 0 (0.0) Currently Use Oral Contraceptive 0.49 No 120 (56.3) 35 (61.4) Yes 93 (43.7) 22 (38.6) Waist Circumference, cm < 0.01 < 80 176 (82.6) 4 (7.0) ≥ 80 37 (17.4) 53 (93.0) Total Body Fat, % < 0.01 < 32 125 (58.7) 6 (10.5) ≥ 32 88 (41.3) 51 (89.5) Android: Gynoid Ratio < 0.01 < 1 183 (85.9) 19 (33.3) ≥ 1 30 (14.1) 38 (66.7) Serum 25-OHD, nmol/L‡,§ 0.61 < 75 72 (54.6) 21 (50.0) ≥ 75 60 (45.4) 21 (50.0) ⱶ P-values estimated with chi-squared tests. ╔ MET, metabolic equivalent of activity hours per week, n = 261 for physical activity. ϖ n = 257 for fasting blood glucose. ┴ n = 269 for race and smoking status. ‡ n = 174 for serum 25-OHD, § 2.496 nmol/L = 1 ng/mL.

101

Table 11b. Variations in Dietary Intakes of Participants by BMI

Low BMI High BMI P ⱶ < 25 kg/m2 ≥ 25 kg/m2 Dietary Intake (n = 213) (n = 57) Mean ± Standard Error Energy, kcal/day 2215 ± 61 2106 ± 113 0.41 Total Protein, g/day 98 ± 3 97 ± 6 0.89 Total Fat, g/day 73 ± 2 69 ± 4 0.43 Total Calcium Intake, mg/day 1243 ± 43 1135 ± 75 0.24 Food calcium 1102 ± 40 1048 ± 75 0.54 Supplemental calcium (n = 93 users) 407 ± 35 261 ± 43 0.05* Vitamin A, IU/day 16663 ± 762 14469 ± 1169 0.17 Thiamin, mg/day 1.9 ± 0.1 1.8 ± 0.1 0.29 Riboflavin, mg/day 2.6 ± 0.1 2.5 ± 0.2 0.43 Niacin, mg/day 27.7 ± 1.0 26.6 ± 2.1 0.60 Pantothenic acid, mg/day 6.8 ± 0.3 6.2 ± 0.4 0.29 Vitamin B6, mg/day 2.5 ± 0.1 2.5 ± 0.2 0.91 Folate, µg/day 672 ± 26 619 ± 45 0.34 Vitamin B12, µg/day 6.3 ± 0.3 6.4 ± 0.6 0.83 Vitamin C, mg/day 155 ± 7 144 ± 13 0.53 Total Vitamin D Intake, IU/day§ 383 ± 21 350 ± 34 0.48 Food vitamin D 231 ± 12 228 ± 25 0.92 Supplemental vitamin D (n = 110 users) 362 ± 31 333 ± 45 0.66 Iron, mg/day 18.3 ± 0.6 16.9 ± 1.4 0.32 Zinc, mg/day 13.7 ± 0.5 13.1± 0.8 0.51 Linoleic acid, g/day 24.5 ± 0.9 23.3 ± 1.5 0.52 Linolenic acid, g/day 13.0 ± 0.5 11.6 ± 0.9 0.21 Dairy Intake, total servings/day₸ 1.8 ± 0.1 2.2 ± 0.3 0.09 ⱶ P-values for test of Null: Equal means, estimated with student t-tests. * Estimated P-value is borderline significant (P = 0.048). § IU, International Units: 1 IU = 0.025 µg. ₸ n = 169 for dairy intake responders. 1 serving dairy = 8 fluid oz. (1 cup) of milk or yogurt; 1.5 oz of natural cheese; 2 oz of processed cheese, 2 cups of cottage cheese.

102

Table 12a. Variations in Characteristics of Participants by Waist Circumference (WC) Measurement

Low WC < 80 cm High WC ≥ 80 cm P ⱶ Covariate (n = 180) (n = 90) Mean ± Standard Deviation Age, y 21.4 ± 2.8 21.5 ± 3.0 0.75 BMI, kg/m2 21.3 ± 1.9 26.0 ± 3.2 < 0.01 Waist Circumference, cm 72.4 ± 4.6 87.5 ± 6.8 < 0.01 Waist to Height, ratio 0.44 ± 0.03 0.52 ± 0.05 < 0.01 Total Body Fat, % 28.6 ± 6.4 39.0 ± 6.8 < 0.01 Arm Fat, % 27.3 ± 8.3 37.0 ± 7.7 < 0.01 Leg Fat, % 32.7 ± 6.2 40.8 ± 6.9 < 0.01 Trunk Fat, % 27.5 ± 7.2 40.5 ± 8.0 < 0.01 Android Fat, % 38.3 ± 5.8 45.0 ± 6.2 < 0.01 Gynoid Fat, % 40.5 ± 7.3 41.1 ± 6.4 < 0.01 Android : Gynoid Fat, ratio 0.77 ± 0.14 0.98 ± 0.13 < 0.01 Physical Activity, METs/week╔ 49.3 ± 41.3 50.4 ± 47.4 0.84 Systolic Blood Pressure, mm Hg 107.7 ± 11.0 108.9 ± 10.6 0.38 Diastolic Blood Pressure, mm Hg 69.3 ± 7.8 71.2 ± 7.6 0.06 Fasting Blood Glucose, mg/dL ϖ 80.4 ± 10.4 84.6 ± 8.4 < 0.01 Serum 25-OHD, nmol/L‡,§ 78.3 ± 31.6 80.5 ± 33.3 0.67 ⱶ P-values for test of Null: Equal means, estimated with student t-test. ╔ MET, metabolic equivalent of activity hours per week, n = 261 for physical activity. ϖ n = 257 for fasting blood glucose. ‡ n = 174 for serum 25-OHD, § 2.496 nmol/L = 1 ng/mL.

103

Table 12a. Continued

Low WC < 80 cm High WC ≥ 80 cm P ⱶ Covariate (n = 180) (n = 90) n (%) Race┴ 0.33 Caucasian 157 (87.7) 75 (83.3) Other 22 (12.3) 15 (16.7) Education 0.82 Some College 150 (83.3) 74 (82.2) College Graduate 30 (16.7) 16 (17.8) Currently Drink Alcohol 0.17 No 42 (23.3) 28 (31.1) Yes 138 (76.7) 62 (68.9) Currently Smoke┴ 0.09 No 166 (92.7) 88 (97.8) Yes 13 (7.3) 2 (2.2) Currently Use Oral Contraceptive 0.26 No 99 (55.0) 56 (62.2) Yes 81 (45.0) 34 (37.8) BMI, kg/m2 < 0.01 < 25 176 (97.8) 37 (41.1) ≥ 25 4 (2.2) 53 (58.9) Total Body Fat, % < 0.01 < 32 119 (66.1) 12 (13.3) ≥ 32 61 (33.9) 78 (86.7) Android: Gynoid Ratio < 0.01 < 1 166 (92.2) 36 (40.0) ≥ 1 14 (7.8) 54 (60.0) Serum 25-OHD, nmol/L‡,§ 0.74 < 75 54 (54.6) 39 (52.0) ≥ 75 45 (45.4) 36 (48.0) ⱶ P-values estimated with chi-squared tests. ┴ n = 269 for race and smoking status. ‡ n = 174 for serum 25-OHD, § 2.496 nmol/L = 1 ng/mL.

104

Table 12b. Variations in Dietary Intakes of Participants by WC Measurement

Low WC High WC P ⱶ < 80 cm ≥ 80 cm Covariate (n = 180) (n = 90) Mean ± Standard Error Energy, kcal/day 2162 ± 65 2253 ± 96 0.42 Total Protein, g/day 93 ± 3 106 ± 6 0.03 Total Fat, g/day 72 ± 2 74 ± 4 0.61 Total Calcium Intake, mg/day 1249 ± 48 1162 ± 57 0.27 Food calcium 1093 ± 44 1085 ± 59 0.92 Supplemental calcium (n = 93 users) 420 ± 38 266 ± 35 0.02 Vitamin A, IU/day 16677 ± 851 15245 ± 956 0.30 Thiamin, mg/day 1.9 ± 0.1 1.9 ± 0.1 0.99 Riboflavin, mg/day 2.6 ± 0.1 2.7 ± 0.2 0.64 Niacin, mg/day 27 ± 1 29 ± 2 0.31 Pantothenic acid, mg/day 6.5 ± 0.2 7.0 ± 0.5 0.31 Vitamin B6, mg/day 2.5 ± 0.1 2.6 ± 0.2 0.32 Folate, µg/day 659 ± 30 664 ± 41 0.91 Vitamin B12, µg/day 6.2 ± 0.3 6.4 ± 0.5 0.72 Vitamin C, mg/day 154 ± 8 150 ± 10 0.76 Total Vitamin D Intake, IU/day§ 394 ± 24 338 ± 25 0.14 Food vitamin D 229 ± 13 233 ± 20 0.86 Supplemental vitamin D (n = 110 users) 372 ± 33 315 ± 39 0.33 Iron, mg/day 18 ± 1 18 ± 1 0.57 Zinc, mg/day 13 ± 1 14 ± 1 0.38 Linoleic acid, g/day 24 ± 1 24 ± 1 0.92 Linolenic acid, g/day 13 ± 1 12 ± 1 0.59 Dairy Intake, total servings/day₸ 1.7 ± 0.1 2.1 ± 0.2 0.08 ⱶ P-values for test of Null: Equal means, estimated with student t-tests. § IU, International Units: 1 IU = 0.025 µg. ₸ n = 169 for dairy intake responders. 1 serving dairy = 8 fluid oz. (1 cup) of milk or yogurt; 1.5 oz of natural cheese; 2 oz of processed cheese, 2 cups of cottage cheese.

105

Table 13. Correlations between Participant Characteristics and Adiposity Measures

Correlation Coefficients (r) ₸ Total Body Fat (%) BMI (kg/m2) Waist Circumference (cm) Covariate r P r P r P Physical activity, METs/week - 0.14 0.02 0.01 0.92 0.01 0.86 Total body fat, % ------0.76 < 0.01 0.72 < 0.01 Waist circumference, cm 0.72 < 0.01 0.84 < 0.01 ------BMI, kg/m2 0.76 < 0.01 ------0.84 < 0.01 Energy, kcal - 0.05 0.43 - 0.06 0.37 0.02 0.78 Total Calcium, mg/day - 0.05 0.41 - 0.08 0.18 - 0.06 0.31 Food - 0.04 0.47 - 0.04 0.47 - 0.01 0.82 Supplement - 0.02 0.72 - 0.10 0.10 - 0.12 0.04 Total Vitamin D, IU/day - 0.09 0.13 - 0.06 0.34 - 0.08 0.18 Food - 0.09 0.14 - 0.01 0.91 - 0.004 0.95 Supplement - 0.05 0.46 - 0.07 0.29 - 0.10 0.11 Dairy Products, total servings/day 0.04 0.64 0.09 0.24 0.12 0.11 ₸ r (Correlation Coefficient) and P-value for significance estimated using Pearson’s Product. Additional Pearson Product Correlation Estimates: Total calcium and total vitamin D (r = 0.51, P < 0.01) Food calcium and food vitamin D (r = 0.75, P < 0.001) Calcium supplement and vitamin D supplement (r = 0.46, P < 0.01).

106

Table 14. Associations (ß ± SE) ϖ of Adiposity (Total Body Fat Percentage, Body Mass Index and Waist Circumference) and Other Parametersϖ

Total Body Fat %

Unadjusted Independent Predictor ß ± SE P-value Age, years -0.23 ± 0.18 0.19 College graduate - 2.1 ± 1.32 0.11 Currently smoke -3.01 ± 2.17 0.17

Currently drink alcohol 1.39 ± 1.13 0.22 Currently use oral contraceptives 0.03 ± 1.01 0.98 Physical activity level, MET-hrs/week -0.03 ± 0.01 0.02 Body Mass Index, kg/m2

Age, years 0.04 ± 0.07 0.57 College graduate - 0.34 ± 0.53 0.52 Currently smoke -1.16 ± 0.87 0.18 Currently drink alcohol 0.60 ± 0.45 0.19

Currently use oral contraceptives -0.26 ± 0.40 0.52 Physical activity level, MET-hrs/week 0.001 ± 0.01 0.92 Waist Circumference, cm Age, years 0.04 ± 0.07 0.57

College graduate - 0.34 ± 0.53 0.52 Currently smoke -1.16 ± 0.87 0.18 Currently drink alcohol 0.60 ± 0.45 0.19 Currently use oral contraceptives -0.26 ± 0.40 0.52

Physical activity level, MET-hrs/week 0.001 ± 0.01 0.92 ϖ Regression coefficients (ß) ± standard errors (SE). All analyses were estimated using simple linear regressions.

107

Table 15. Crude and Adjusted Indepenent Associtaions with Adiposity (ß ± SE) ϖ in Young Women

Total Body Fat (%) Dietary Intake β Std. Err P § Energy Adjusted Total Calcium, mg/day -0.0004 0.001 0.68

Food Calcium, mg/day -0.0003 0.001 0.83 Supplemental Calcium, mg/day -0.001 0.002 0.66 *Supplemental Calcium, mg/day -0.003 0.003 0.44 Total Vitamin D, IU/day -0.002 0.002 0.18 Food Vitamin D, IU/day -0.004 0.003 0.21 Supplemental Vitamin D, IU/day -0.002 0.002 0.44 *Supplemental Vitamin D, IU/day -0.002 0.003 0.45 Multivariable Adjusted Ж Total Calcium, mg/day 0.001 0.001 0.66

Food Calcium, mg/day 0.002 0.002 0.27 Supplemental Calcium, mg/day -0.002 0.002 0.76 * Supplemental Calcium, mg/day -0.004 0.003 0.20 Total Vitamin D, IU/day -0.003 0.002 0.22 Food Vitamin D, IU/day -0.006 0.004 0.14 Supplemental Vitamin D, IU/day -0.001 0.002 0.64 *Supplemental Vitamin D, IU/day 0.003 0.004 0.44 ϖ Regression coefficients (ß) ± standard errors (SE). All analyses were estimated using simple or multiple linear regressions. § Energy adjusted by including total energy intake in the regression model. Ж Multivariable adjusted for total energy intake, physical activity level, smoking status and calcium or vitamin D intake as appropriate. * Among supplement users only.

108

Table 15. Continued

Body Mass Index (kg/m2) Dietary Intake β Std. Err P § Energy Adjusted Total Calcium, mg/day -0.0004 0.0004 0.32 Food Calcium, mg/day -0.0001 0.001 0.93 Supplemental Calcium, mg/day -0.001 0.001 0.08 *Supplemental Calcium, mg/day -0.002 0.001 0.17 Total Vitamin D, IU/day -0.001 0.001 0.46 Food Vitamin D, IU/day 0.0004 0.001 0.72 Supplemental Vitamin D, IU/day -0.001 0.001 0.27 *Supplemental Vitamin D, IU/day -0.001 0.001 0.60 Multivariable Adjusted Ж Total Calcium, mg/day -0.0004 0.001 0.39 Food Calcium, mg/day -0.0003 0.001 0.72 Supplemental Calcium, mg/day -0.002 0.001 0.11 *Supplemental Calcium, mg/day -0.002 0.001 0.12 Total Vitamin D, IU/day -0.0001 0.001 0.90 Food Vitamin D, IU/day 0.001 0.002 0.64 Supplemental Vitamin D, IU/day -0.0001 0.001 0.94 *Supplemental Vitamin D, IU/day 0.002 0.002 0.35 ϖ Regression coefficients (ß) ± standard errors (SE). All analyses were estimated using simple or multiple linear regressions. § Energy adjusted by including total energy intake in the regression model. Ж Multivariable adjusted for total energy intake, physical activity level, smoking status and calcium or vitamin D intake as appropriate. * Among supplement users only.

109

Table 15. Continued

Waist Circumference (cm) Dietary Intake β Std. Err P Energy Adjusted§

Total Calcium, mg/day -0.002 .001 0.11 Food Calcium, mg/day -0.001 .001 0.53 Supplemental Calcium, mg/day -0.004 .002 0.05* ¥ Supplemental Calcium, mg/day -0.006 .003 0.05 Total Vitamin D, IU/day -0.003 .002 0.14 Food Vitamin D, IU/day -0.001 .003 0.82 Supplemental Vitamin D, IU/day -0.003 .002 0.11 Supplemental¥ Vitamin D, IU/day -0.003 .003 0.31 Multivariable Adjusted Ж

Total Calcium, mg/day -0.002 .001 0.24 Food Calcium, mg/day -0.001 .002 0.61

Supplemental Calcium, mg/day -0.005 .003 0.07 Supplemental¥ Calcium, mg/day -0.008 .004 0.04 Total Vitamin D, IU/day -0.001 .002 0.66 Food Vitamin D, IU/day 0.001 .010 0.80 Supplemental Vitamin D, IU/day -0.001 .003 0.79 Supplemental¥ Vitamin D, IU/day 0.003 .004 0.75 ϖ Regression coefficients (ß) ± standard errors (SE). All analyses were estimated using simple or multiple linear regressions. * Esitmated P-value is borderline significant (P = 0.045). § Energy adjusted by including total energy intake in regression model. Ж Multivariable adjusted for total energy intake, physical activity level, smoking status and calcium or vitamin D intake as appropriate.

¥ Among supplement users only.

110

Table 16. Prevalence Odds Ratio for the Association of Calcium and Vitamin D Adequacy with Events of Adiposity

Odds Ratio

(95% Confidence Interval)

Energy Multivariable Adequacy of Control / Cases Adjusted § Adjusted ¤ Dietary Intake Total Body Fat (≥ 32%) Total Calcium, mg/day < 1000 52 / 56 1.13 (0.65 - 1.94) 1.23 (0.69 - 2.20) ≥ 1000 79/ 83 1.0 1.0 Total Vitamin D, IU/day < 600 103 / 122 0.54 (0.27 - 1.06) 0.55 (0.27 - 1.10) ≥ 600 28/ 17 1.0 1.0 Body Mass Index (≥ 25 kg/m2) Total Calcium, mg/day < 1000 83 / 25 0.90 (0.46 - 1.74) 0.80 (0.41 - 1.59) ≥ 1000 130 / 32 1.0 1.0 Total Vitamin D, IU/day < 600 176 / 49 0.83 (0.36 - 1.95) 0.90 (0.37 - 2.16) ≥ 600 37 / 8 1.0 1.0 Waist circumference (≥ 80 cm) Total Calcium, mg/day < 1000 70 / 38 0.75 (0.42 - 1.33) 0.75 (0.41 - 1.36) ≥ 1000 110 / 52 1.0 1.0 Total Vitamin D, IU/day < 600 146 / 79 0.53 (0.25 - 1.14) 0.57 (0.26 - 1.24) ≥ 600 34 / 11 1.0 1.0 All models were constructed using logistic regression. Outcome variable coding was as follows: percentage total body fat, (Control = 0 is < 32%, Case = 1 is ≥ 32%); body mass index, (Control = 0 is < 25 kg/m2, Case = 1 is ≥ 25 kg/m2); waist circumference, (Control = 0 is < 80 cm, Case = 1 is ≥ 80 cm). § Energy adjusted by including total energy intake in the regression model. ¤ Multivariable adjusted for total energy intake, physical activity level, smoking status and calcium or vitamin D intake as appropriate.

111

Table 17. Association of Calcium and Vitamin D Intakes with Event of High Adiposity

Total Body Fat (≥ 32%) Odds Ratio (95% Confidence Interval) Exposure Controls / Cases Energy Adjusted § Multivariable Adjusted ¤ Total Calcium, (tertile medians in mg/day) T1 (692) 46 / 44 1.02 (0.52 - 2.00) 0.80 (0.37 - 1.74) T2 (1143) 35 / 55 1.82 (0.99 - 3.37) 1.56 (0.80 - 3.02) T3 (1658) 50 / 40 1.0 1.0

P trend 0.86 0.64 Food Calcium, (tertile medians in mg/day) T1 (597) 45 / 45 1.08 (0.52 - 2.27) 0.72 (0.30 - 1.71) T2 (980) 36 / 54 1.72 (0.90 - 3.30) 1.28 (0.61 - 2.67) T3 (1498) 50 / 40 1.0 1.0 P trend 0.72 0.48 Supplemental Calcium, (tertile medians in mg/day) T1 (93) 46 / 44 1.02 (0.52 - 2.00) 0.96 (0.46 - 1.99) T2 (450) 35 / 55 1.82 (0.99 - 3.37) 1.74 (0.92 - 3.31) T3 (623) 50 / 40 1.0 1.0 * P trend 0.50 0.81 Total Vitamin D, (tertile medians in IU/day) T1 (109) 40 / 50 1.29 (0.71 - 2.35) 1.16 (0.57 - 2.38) T2 (299) 44 / 46 1.10 (0.61 - 1.99) 1.05 (0.55 - 2.00) T3 (600) 47/ 43 1.0 1.0

P trend 0.42 0.70 Food Vitamin D, (tertile medians in IU/day) T1 (71) 39 / 51 1.62 (0.87 - 3.04) 1.50 (0.71 - 3.17) T2 (188) 41 / 49 1.51 (0.83 - 2.77) 1.34 (0.68 - 2.71) T3 (369) 51 / 39 1.0 1.0 P trend 0.11 0.28 Supplemental Vitamin D, (tertile medians in IU/day) T1 (85) 40 / 50 1.29 (0.71 - 2.35) 1.40 (0.70 - 2.79) T2 (400) 44 / 46 1.10 (0.61 - 1.99) 1.18 (0.61 - 2.28) T3 (800) 47 / 43 1.0 1.0 * P trend 0.86 0.45 All models were constructed using logistic regression. Outcome variable coding (Percentage total body fat, Control = 0 is < 32%, Case = 1 is ≥ 32%). § Energy adjusted by including total energy intake in the regression model. ¤ Multivariable adjusted for total energy intake, physical activity level, smoking status and calcium or vitamin D intake as appropriate. * Among supplement users only.

112

Table 17. Continued

Body Mass Index (≥ 25 kg/m2) Odds Ratio (95% Confidence Interval) Exposure Controls / Cases Energy Adjusted § Multivariable Adjusted ¤ Total Calcium, (tertile medians in mg/day) T1 (692) 71 / 19 1.31 (0.55 - 3.12) 1.48 (0.56 - 3.87) T2 (1143) 66 / 24 1.88 (0.88 - 4.03) 1.84 (0.83 - 4.10) T3 (1658) 76 / 14 1.0 1.0

P trend 0.52 0.41 Food Calcium, (tertile medians in mg/day) T1 (597) 71 / 19 1.19 (0.47 - 3.05) 1.53 (0.53 - 4.37) T2 (980) 67 / 23 1.60 (0.72 - 3.57) 1.86 (0.77 - 4.54) T3 (1498) 75 / 15 1.0 1.0 P trend 0.67 0.41 Supplemental Calcium, (tertile medians in mg/day) T1 (93) 71 / 19 1.31 (0.55 - 3.12) 1.38 (0.56 - 3.42) T2 (450) 66 / 24 1.88 (0.88 - 4.03) 1.80 (0.84 - 3.89) T3 (623) 76 / 14 1.0 1.0 * P trend 0.046 0.02 Total Vitamin D, (tertile medians in IU/day) T1 (109) 68 / 22 1.06 (0.52 - 2.15) 0.81 (0.35 - 1.88) T2 (299) 75 / 15 0.67 (0.32 - 1.42) 0.59 (0.27 - 1.30) T3 (600) 70 / 20 1.0 1.0

P trend 0.99 0.52 Food Vitamin D, (tertile medians in IU/day) T1 (71) 67 / 23 1.08 (0.52 - 2.26) 1.01 (0.43 - 2.38) T2 (188) 76 / 14 0.60 (0.28 - 1.31) 0.53 (0.23 - 1.25) T3 (369) 70 / 20 1.0 1.0 P trend 0.95 0.99 Supplemental Vitamin D, (tertile medians in IU/day) T1 (85) 68 / 22 1.06 (0.52 - 2.15) 0.70 (0.31 - 1.57) T2 (400) 75 / 15 0.67 (0.32 - 1.42) 0.49 (0.22 - 1.11) T3 (800) 70 / 20 1.0 1.0 * P trend 0.82 0.27 All models were constructed using logistic regression. Outcome variable coding (body mass index, Control = 0 is < 25 kg/m2, Case = 1 is ≥ 25 kg/m2). § Energy adjusted by including total energy intake in the regression model. ¤ Multivariable adjusted for total energy intake, physical activity level, smoking status and calcium or vitamin D intake as appropriate. * Among supplement users only.

113

Table 17. Continued

Waist Circumference (≥ 80 cm) Odds Ratio (95% Confidence Interval) Exposure Controls / Cases Energy Adjusted § Multivaria ble Adjusted ¤ Total Calcium, (tertile medians in mg/day) T1 (692) 44 / 31 1.58 (0.75 - 3.34) 1.36 (0.59 - 3.10) T2 (1143) 40 / 32 2.27 (1.17 - 4.39) 1.92 (0.96 - 3.84) T3 (1658) 51 /22 1.0 1.0

P trend 0.20 0.44 Food Calcium, (tertile medians in mg/day) T1 (597) 47 / 26 1.22 (0.55 - 2.70) 1.26 (0.52 - 3.05) T2 (980) 38 / 36 1.64 (0.83 - 3.25) 1.61 (0.76 - 3.40) T3 (1498) 50 / 23 1.0 1.0 P trend 0.55 0.58 Supplemental Calcium, (tertile medians in mg/day) T1 (93) 50 / 26 1.58 (0.75 - 3.34) 1.39 (0.63 - 3.05) T2 (450) 38 / 33 2.27 (1.17 - 4.39) 2.04 (1.04 - 3.99) T3 (623) 47 / 26 1.0 1.0 * P trend 0.09 0.01 Total Vitamin D, (tertile medians in IU/day) T1 (109) 45 / 32 1.31 (0.69 - 2.47) 0.84 (0.39 - 1.77) T2 (299) 46 / 26 1.16 (0.62 - 2.19) 0.93 (0.47 - 1.83) T3 (600) 44 / 27 1.0 1.0

P trend 0.41 0.65 Food Vitamin D, (tertile medians in IU/day) T1 (71) 47 / 29 1.16 (0.66 - 2.24) 0.92 (0.43 - 1.96) T2 (188) 44 / 31 0.87 (0.67 - 1.66) 0.68 (0.33 - 1.39) T3 (369) 44 / 25 1.0 1.0 P trend 0.72 0.80 Supplemental Vitamin D, (tertile medians in IU/day) T1 (85) 47 / 29 1.31 (0.69 - 2.47) 0.82 (0.40 - 1.68) T2 (400) 39 / 34 1.16 (0.62 - 2.19) 0.83 (0.42 - 1.66) T3 (800) 49 / 22 1.0 1.0 * P trend 0.55 0.47 All models were constructed using logistic regression. Outcome variable coding (waist circumference, Control = 0 is < 80 cm, Case = 1 is ≥ 80 cm). § Energy adjusted by including total energy intake in the regression model. ¤ Multivariable adjusted for total energy intake, physical activity level, smoking status and calcium or vitamin D intake as appropriate. * Among supplement users only.

114

Table 18. Association of Women’s Combined Total Calcium and Vitamin D Intakes with Events of High Adiposity

Odds Ratio

(95% Confidence Interval) Energy Multivariable Combined Total Intakes Controls / Cases Adjusted § Adjusted ¤ Total Body Fat (≥ 32%) Low Calcium₸ & Low 49 / 52 1.80 (0.76 - 4.21) 1.69 (0.70 - 4.09) Vitamin D┬ Low Calcium₸ & High 3 / 4 2.22 (0.41 - 12.03) 2.13 (0.37 - 12.16) Vitamin D┴ High Calcium╧ & Low 54 / 70 2.37 (1.09 - 5.12) 2.27 (1.02 - 5.04) Vitamin D┬ High Calcium╧ & High 25 / 13 1.0 1.0 Vitamin D┴ Body Mass Index (≥ 25 kg/m2) Low Calcium₸ & Low 77 / 24 1.21 (0.43 - 3.40) 1.33 (0.46 - 3.81) Vitamin D┬ Low Calcium₸ & High 6 / 1 0.64 (0.06 - 6.49) 0.89 (0.08 - 9.84) Vitamin D┴ High Calcium╧ & Low 1.06 (0.41 - 2.73) 1.05 (0.40 - 2.73) Vitamin D┬ 99 / 25 High Calcium╧ & High 31 / 7 1.0 1.0 Vitamin D┴ Waist circumference (≥ 80 cm) Low Calcium₸ & Low 65 / 34 2.42 (0.93 - 6.31) 2.51 (0.95 - 6.66) Vitamin D┬ Low Calcium₸ & High 5 / 2 1.80 (0.28 - 11.62) 2.17 (0.32 - 14.88) Vitamin D┴ High Calcium╧ & Low 1.96 (0.83 - 4.65) 1.94 (0.81 - 4.67) Vitamin D┬ 81 / 43 High Calcium╧ & High 29 / 9 1.0 1.0 Vitamin D┴ All models were constructed using logistic regression. Outcome variable coding was as follows: percentage total body fat, (Control = 0 is < 32%, Case = 1 is ≥ 32%); body mass index, (Control = 0 is < 25 kg/m2, Case = 1 is ≥ 25 kg/m2); waist circumference, (Control = 0 is < 80 cm, Case = 1 is ≥ 80 cm). § Energy adjusted by including total energy intake in the regression model. ¤ Multivariable adjusted for total energy intake, physical activity level and smoking status. ╧ High Calcium = Total Calcium intake ≥ 1000 mg/day. ₸ Low Calcium = Total Calcium intake < 1000 mg/day. ┴ High Vitamin D = Total Vitamin D intake ≥ 600 IU/day. ┬ Low Vitamin D = Total Vitamin D intake < 600 IU/day.

115

CHAPTER 4

ASSOCIATIONS BETWEEN SERUM 25-HYDROXYVITAMIN D

CONCENTRATION AND ADIPOSITY MEASURES IN YOUNG WOMEN

(18 – 30 YEARS)

A. A. Addo-Lartey §, E. R. Bertone-Johnson ‡, C. Bigelow ‡, R. J. Wood §, S. E. Zagarins‡,

S. H. Gehlbach ‡, A. G. Ronnenberg §

§ Department of Nutrition, School of Public Health and Health Sciences, University of

Massachusetts Amherst, MA, 01003, USA

‡ Division of Epidemiology and Biostatistics, Department of Public Health, School of

Public Health and Health Sciences, University of Massachusetts Amherst, MA, 01003,

USA

Abstract: Some cross-sectional studies have suggested that serum 25-hydroxyvitamin D levels [25-OHD] are consistently lower in overweight/obese women (body mass index,

BMI, ≥ 25 kg/m2) compared to women with low BMI (< 25 kg/m2). Our objective was to determine the extent of association between serum 25-OHD concentrations and adiposity measures in apparently healthy young women. This was a cross-sectional analysis among

174 pre-menopausal aged women (18-31 years) enrolled the UMass Amherst Vitamin D

Status Study (2006 - 2011). Serum 25-hydroxyvitamin D levels [25-OHD] were assessed using radioimmunoassay. Total and regional body fat percentages were measured by dual energy x-ray absorptiometry. Our results showed that vitamin D insufficiency (25-OHD <

75 nmol/L) was common (53.4%). Overal, total, food, and supplemental vitamin D intakes were not associated with serum 25-OHD concentrations (P > 0.05). However, among supplement users only (44.3% of women), intake of vitamin D supplements was positively

116

associated with 25-OHD levels (ß = 0.03 ± 0.01, P = 0.05). We found no significant correlations of 25-OHD levels with adiposity measures (P > 0.05). After adjusting for season, physical activity level and smoking status, adiposity was not associated with risk of low 25-OHD (< 75 nmol/L). Prospective studies are needed to determine if adiposity plays a causal role in modifying 25-OHD levels in young women.

Keywords: Vitamin D, Serum 25-hydroxyvitamin D, Adiposity, Body fat percentage,

Young women

4.1 Introduction

Vitamin D is naturally synthesized in the skin, but can also be found in some foods, especially wild salmon, mackerel, beef or veal liver, and fish oils [1-3]. Vitamin D obtained from sun exposure, food, and supplements is biologically inert and must undergo undergo two hydroxylations for activation. The first hydroxylation occurs in the liver, where the enzyme 25-hydoxylase (CYP27A) converts the vit D to 25-OHD (calcidiol). The second hydroxylation is carried out by the enzyme 1-α-hydroxylase (CYP27B) in the kidney and forms the biologically active 1, 25D (calcitriol).

Sub-optimal concentrations of serum/plasma 25-OHD (< 75 nmol/L) has been associated with increased risk of metabolic syndrome [4-5], and chronic diseases such as type-II-diabetes [6-7], hypertension [8], coronary heart disease [9-12], and cancer [13].

Reports from NHANES (2001 – 2006) show that about one-fourth of the US population

(aged ≥ 1 year) is at risk of vitamin D inadequacy (25-OHD of 30 – 49 nmol/L), while 8% of the population are at risk of vitamin D deficiency [25-OHD < 30 nmol/L] [14]. Among

117

young adult women (19 - 30 years), the percentage at risk for vitamin D deficiency (25-

OHD < 30 nmol/L) is estimated to be about 11% [14].

Accumulating epidemiologic evidence is conflicting on the relationship between serum 25-OHD levels and adiposity. Some cross-sectional studies in adult men and women indicate that 25-OHD concentrations are lower in obese compared with non-obese subjects

[15-19], and negatively associated with BMI [15-22] and body fat percentage [15-17, 23-25] measured by dual energy x-ray absorptiometry (DXA). Other studies in similar populations have however failed to confirm these observations [26, 27]. The low levels of 25-OHD observed in obesity have been attributed to multiple factors including, i) decreased exposure to sunlight because of limited mobility [28], ii) negative feedback from elevated 1,25- dihydroxyvitamin D (1,25-(OH)2D) and parathyroid hormone levels on hepatic synthesis of

25-OHD [28, 29], and iii) decreased bioavailability of vitamin D3 from cutaneous and dietary sources due to its deposition in adipose tissue [30, 31]. It is unclear whether the use of oral contraceptives and lifestyle choices such as cigarette smoking and alcohol use can influence the association between 25-OHD levels and adiposity in reproductive aged women. The aim of this cross-sectional study was to evaluate the association between serum

25-OHD levels and regional adiposity measures and the influence of oral contraceptive use or lifestyle factors on this association in young women enrolled in the UMass Amherst

Vitamin D Status Study.

118

4.2 Materials and Methods

4.2.1 Study Design and Population

Between March 2006 and September 20011, 288 healthy women aged 18 to 30 years from the Amherst, Massachusetts, area were enrolled in the University of Massachusetts Amherst

Vitamin D Status Study. Recruitment criteria have been described in detail previously

(section 3.2.1). Assays for serum 25-hydroxyvitamin D (25-OHD) concentrations were available for 174 women providing blood samples. This study was approved by the

Institutional Review Board at the University of Massachusetts, Amherst. Written informed consent was obtained from all participants.

4.2.2 Assessment of Serum 25-OHD Status

All study measurements were completed in a single clinic visit scheduled for the mid-luteal phase of participant's menstrual cycle. Women provided a fasting venous blood sample which was immediately processed and stored at −80 °C, usually within two hours of draw.

Assays for serum 25-hydroxyvitamin D concentrations (25-OHD) were available for 175 women providing blood samples. Serum 25-OHD levels were determined using a commercially available radioimmunoassay kit (25-hydroxyvitamin D 125I RIA Kit; DiaSorin

Inc., Stillwater, MN, USA), [32]. Gamma irradiation of I125 (in counts per minute; CPM) was quantified using a Beckman Gamma 4000 counter (Beckman Coulter, California,

USA), and CPM were converted into concentration units using GraphPad Prism version 4.0 for Windows (GraphPad Software, San Diego, CA). In light of the on-going debate regarding use of a higher 25-OHD cut-off, vitamin D status was classified as follows; serum

119

25-OHD < 50 nmol/L [< 20 ng/mL] = deficient, 25-OHD of 50 - 75 nmol/L [20 - 50 ng/mL]

= insufficient, and 25-OHD ≥ 75 nmol/L [50 ng/mL] = sufficient [33, 34].

4.2.3 Dietary Assessment

Dietary intake of food and supplements in the previous two months was assessed using a modified version of the Harvard Food Frequency Questionnaire (FFQ). This FFQ has been previously validated for use in U.S. women [35, 36]. Vitamin D-rich food items on the modified questionnaire included regular and fortified dairy foods and fortified orange juice and breakfast cereals, dark meat fish, shellfish, leafy green vegetables and multivitamins or other supplements containing vitamin D. Alcohol intake was measured using FFQs which were analyzed at Harvard University. Nutrient intakes were calculated by multiplying the frequency of intake of a specified portion size of each food by its nutrient content and then summing across all food items. Adequacy of dietary vitamin D intake was determined by comparing reported intakes with the Estimated Average Requirement (EAR) or

Recommended Dietary Allowance (RDA) for women in this age range [37].

4.2.4 Assessment of Covariates

Information on lifestyle and demographic factors was collected by self-reported questionnaire. Women’s weight, height, and waist circumference were directly measured using a calibrated scale, wall mounted stadiometer, and a flexible tape measure, respectively.

Body mass index (BMI) was calculated as weight in kilograms divided by the square of height in meters. Body composition was measured by dual energy x-ray absorptiometry

120

(DXA) using the total body scan mode on a narrow angle fan GE Lunar Prodigy scanner

(GE Lunar Corp., Madison, WI). DXA scans were performed with the participant lying in a supine position with arms and legs placed flat by the side. Measurements were completed by the same technician, and the machine was calibrated daily using the standard calibration method provided by the manufacturer. The coefficient of variation for repeated scans of the manufacturer’s phantom was less than 1%. DXA scans also provided direct estimates of regional adiposity measurements (fat mass and percentage fat mass), including the arms, legs, and trunk, android and gynoid regions (Fig 12). Android fat is the fat stored in the midsection of the body, predominantly in the abdomen, but also the in the chest and upper arms. Gynoid fat is stored primarily around the hips, thighs, and buttocks. Individuals with disproportionate amounts of android or gynoid fat are usually referred to as apple-or pear- shape, respectively (Fig 13). Total body fat percentage (TBF %) was obtained directly from

DXA scans (estimated using the ratio of whole-body fat mass to whole-body total mass).

Figure 12. A Dual-Energy X-ray Absorptiometry (DXA) Scan Report Showing Different

Body Composition Measures

Source: http://www.rugkliniek.nu/het-onderzoek/wervelkolom-idxa-scan/

121

Figure 13. A DXA Scan showing Regional Adiposity Measures

Apple Shape vs. Pear shape

Dorgan et al. Breast Cancer Research 2012; 14:.R107.

To measure physical activity levels, participants were asked to report the time they spent each week engaged in specific activities including walking, jogging, running, bicycling, aerobics/dancing, tennis/racket sports, swimming, yoga/pilates and weight training. These questions were based on those used in the Nurses’ Health Study II and have been previously validated in that population [38]. We then calculated total metabolic equivalents (MET), which were recorded in MET-hours of activity per week. The MET-hours per week score measures the intensity of a physical activity, with 1 MET being defined as the amount of energy expended when a person is at rest [39]. We also collected information on race, use of oral contraceptives and smoking status.

4.2.5 Statistical Analysis

All analyses were completed using STATA version 12.0 (Stata Corporation., College

Station, TX.). We compared demographic, dietary and life style characteristics of women using student t-tests for normally distributed continuous data, and chi-squared tests for

122

categorical data. BMI was classified as “low” (< 25 kg/m2) or “high” (≥ 25 kg/m2) based on current recommendations by the World Health Organization [40] and National Heart, Lung and Blood Institute and the North American Association for the Study of Obesity [41].

Waist circumference (WC) cut-off point for high central obesity was defined as WC ≥ 80 cm, using guidelines proposed by the International Diabetes Federation [42]. Total body fat percentage was categorized as “low” (< 32%) or “high” (≥ 32%) based on gender- and age- specific cut-offs [43] and also on the current American Council on Exercise recommendation for maximum body fat percentage in women who are non-athletes [44].

One way analysis of variance (ANOVA) was used to test between-group differences for anthropometric and adiposity measures with Scheffe post-hoc tests for differences in adiposity based on categories of 25-OHD concentration.

Correlation between adiposity measures and serum 25-OHD was assessed using

Pearson’s Product Correlation test. We assessed the nature and strength of associations among covariates, 25-OHD and adiposity variables using normal throery multiple linear regression. Linear and restricted cubic spline model fit were also obtained to assess departure from linearity (Appendix D). Model adequacy was assessed using residual versus predicted values, and Cook’s distance plots were used to identify influential data points.

Crude and adjusted associations between 25-OHD and adiposity were estimated from unadjusted and multivariable adjusted linear regression. The dependent variable in these analyses was vitamin D staus defined as serum 25-OHD < 50 nmol/L, 50-75 nmol/L and ≥

75 nmol/L.

Selected covariates were considered for inclusion in our multivariable linear regression models based on evidence in the literature suggesting their potential role as

123

confounders or effect modifiers. Season of blood draw and physical activity level were considered a priori to be likely confounders of the association between serum 25-OHD and adiposity. Additional covariates including age, smoking status, alcohol intake, and use of oral contraceptives were considered for inclusion in our final multivariable models. We conducted model assessment with and without each of these covariates, and included in the final regression model any variable with P < 0.20 or led to at least a 10% change in the beta coefficient for serum 25-OHD concentration. Our final model was further evaluated for multicollinearity between covariates. To evaluate linear trends, we used a Mantel extension test, modeling the medians of each tertile of adiposity measure as a continuous variable. All

P-values were two-sided and considered statistically significant if P < 0.05.

4.3 Results

Demographic characteristics of participants has been reported previously (section 3.3). Body mass index (BMI), total body fat percentage (TBF %) and waist circumference (WC) ranged from 16.7 – 37.7 kg/m2, 14.8 – 56.6 %, and 58.4 – 114.3 cm, respectively (Table 7). Serum vitamin D levels (25-OHD) ranged from 21.7 - 174.3 nmol/L (Table 19) with a mean of

79.2 ± 32.3 nmol/L. Overall, characteristics of participants, including demographic, lifestyle factors, and dietary intakes did not vary by serum 25-OHD status (Table 20 - 21). Mean serum 25-OHD levels also did not differ by oral contraceptive use, alcohol intake or smoking status (Table 22).

Figure 14 shows the covariation and Pearson product correlation between vitamin D intake and serum 25-OHD levels. Overal, total, food, and supplemental vitamin D intake were not associated with serum 25-OHD concentrations (P > 0.05). However, further

124

analysis among women using supplements containing vitamin D (44.3% of women), showed that use of vitamin D supplements was positively associated with serum 25-OHD levels (ß = 0.03 ± 0.01, P = 0.05). Thus for every 100 IU/day increase in vitamin D supplement intake, serum 25-OHD concentrations subsequently improved by 3 nmol/L.

Figures 15-20 show no significant correlation between serum 25-OHD levels and adiposity measures. Mean serum levels of 25-OHD did not differ by total body fat percentage, BMI, or waist circumference (Table 23). Crude and adjusted mean measures of adiposity did not differ vary by serum 25-OHD status. There was no evidence of a linear dose-relationship between adiposity and serum 25-OHD levels (Table 24). Simple and multivariable adjusted regression analyses (Table 26), adjusted for season, physical activity level and smoking status also showed no significant associations between serum 25-OHD levels and adiposity measures.

4.4 Discussion

To our knowledge, this is the first study to assess the association between serum 25-OHD and adiposity in a general population of pre-dominantly non-obese physically active young women. After adjusting for physical activity level, smoking status, and season of blood draw, we did not find serum 25-OHD concentration to be associated with adiposity measures in apparently healthy young women.

Our results showed that 16.1% of women in our population were at risk for vitamin

D deficiency (serum 25-OHD < 50 nmol/L). Our estimate is slightly higher than reports from NHANES data which indicates that about 11% of young adult women (19 – 30 years) in the U.S. are at risk of vitamin D deficiency (25-OHD < 30 nmol/L) [14]. It is not entirely

125

surprising that our estimate of vitamin D deficiency prevalence is larger than reports from

NHANES. This is because NHANES data is a nationally representative assessment while our cohort is from the Northeastern Region of the U.S. Cutaneous synthesis of vitamin D is influenced by several factors, including time of day, season of year, and latitude. At latitudes above 42◦N (In Massachusetts, New York, Connecticut, and Pennsylvania), very little pre-vitamin D3 is synthesized in skin (using 7-dehydrocholesterol) from November through February (winter months) because the incident angle of sunlight at these latitudes during this time of year causes blockage of the UVB photons needed to convert 7- dehydrocholesterol to pre-vitamin D3 [45]. During spring and summer months, more UVB photons are able to penetrate the ozone layer because the sun is directly overhead [45]. We did not find serum 25-OHD levels to be associated with season of blood draw. This could be because only very few women (1.7 %) provided blood samples for serum 25-OHD assays in summer months.

We also assessed the relationship between vitamin D intake and serum concentrations of 25-OHD and the results showed a relatively poor correlation between vitamin D intake from foods and serum 25-OHD levels (r = -0.02, P = 0.75), demonstrating that food vitamin D intake (natural and fortified) might only be modestly related to vitamin

D status. Vitamin D from supplement intake also accounted for only 4% of variation in serum 25-OHD concentrations. Our observations suggest that other factors apart from vitamin D intake might influence serum 25-OHD status in young women. Additional studies are needed to investigate these relationships.

In the present study, we observed no significant correlations between serum 25-OHD levels and BMI, TBF%, WC, or regional adiposity measures, including percentage of fat in the arms legs, trunk, android, and gynoid regions. Unadjusted and adjusted generalized 126

linear regression analyses also showed no significant associations between serum 25-OHD concentrations and BMI, waist circumference (WC), or total body fat percentage (TBF %), which does not support the hypothesis that adipose tissue exerts an important effect on serum 25-OHD levels [46-49]. Shi et al., have demonstrated that elevated concentrations of

1,25-(OH)2D stimulates lipogenesis and inhibit lipolysis in cultured human adipocytes, leading to accumulation of fat in adipocytes [31]. Also, 1, 25(OH)2D can inhibit the expression of adipocyte uncoupling protein-2 (UCP2), which causes a decrease in the metabolic efficiency of adipocytes [48].

Among healthy Caucasians and African-American adults (25 – 48 years), Parikh et al., observed that serum 25-OHD concentration was negatively correlated with body mass index (BMI: r = -0.40; P < 0.001) and percentage body fat (BF%: r = -0.41; P < 0.001), measured by dual energy x-ray absorptiometry [15]. A cross-sectional study by Rodriguez –

Rodriguez et al. in pre-menopausal overweight/obese women (20 – 35 years ) also found that while intakes of vitamin D, calcium, and supplements did not differ significantly between groups, BMI and WC measures were significantly lower in women with low 25-

OHD (< 90 nmol/L) compared to women with high 25-OHD concentrations (≥ 90 nmol/L), i.e. BMI: 26.0 ± 1.3 kg/m2 vs. 28.6 ± 3.2 kg/m2 and WC: 79.4 ± 3.4 cm vs. 86.2 ± 9.3 cm, respectively; P < 0.05) [18]. One potential explanation for this inconsistency in our findings compared to previous studies [15, 18] is that previous studies have focused on 25-OHD levels in predominately overweight or obese women; while our population comprised largely of non-obese women (75.9% of women had BMI < 25 kg/m2). Our range of adiposity was therefore very narrow; therefore it is likely that we may have had little power to evaluate serum 25-OHD levels across a wide range of adiposity values. Rodriguez-

127

Rodriguez et al. also used a relatively higher cut-off (≥ 90 nmol/L) to assess low 25-OHD, compared to our evidence-based cut-off of ≥ 75 nmol/L [33, 34]. This variation in categorization of vitamin D insufficiency may have influenced the percentage of women identified as cases in our study, and this in turn may have limited our power to detect any associations of serum 25-OHD and adiposity.

A recent cross-sectional study among ninety post-menopausal women (69 - 87 years) showed that total body fat is a negative predictor of serum 25-OHD levels (β = -0.247, P =

0.016), even after adjusting for age, lifestyle, and serum intact parathyroid hormone [50].

Moschonis et al. [51] have also reported of a cross-sectional study in post-menopausal women (mean age, 60.3 ± 5.0 years; mean BMI, 29.5 ± 4.8 kg/m) showing that after controlling for age, serum intact parathyroid hormone, insulin-like growth factor I levels, ultraviolet B radiation exposure, and physical activity levels, there was an independent inverse association between serum 25-OHD and DXA measurements of total body fat (β = -

0.28, P = 0.012), trunk fat (β = -0.20, P = 0.061), and leg fat mass (β = -0.30, P = 0.005),

[50].

We observed no associations between 25-OHD levels and adiposity measures, which contrast previous findings [50, 51]. However, this dissimilarity in findings is totally not unexpected, especially since young healthy women are more likely to be physically active, regardless of their BMI; whereas the association of physical activity and BMI is higher in older women [52, 53]. The median BMI in Jungert et al’s study was in the overweight category (26.3 kg/m2) compared to the low/normal median BMI of 22.6 kg/m2 observed in our study. We controlled for potential confounding factors including season of blood draw, smoking status, and physical activity level which have been independently associated with

128

serum 25-OHD concentrations in previous studies [54-58], although not in our study.

Physical activity, specifically vigorous activities and high-intensity exercise, has also been linked with lower amounts of subcutaneous skinfold thicknesses and fat distribution [59], and previous findings from our cohort indicated significant associations between physical activity level and total body fat percentage (r = -0.14, P = 0.02). Eventhough women with high serum 25-OHD (≥ 75 nmol/L) also reported higher intakes of supplemental calcium

(420 ± 55 vs. 275 ± 39 g/day, respectively, P = 0.04) as compared to women with low 25-

OHD (< 75 nmol/L), supplemental calcium intake was not a significant predictor of vitamin

D status.

We also considered the effects of oral contraceptive (OC) use in the association between adiposity and serum 25-OHD, but OC use did not influence the extent of association between serum 25-OHD concentration and adiposity and was excluded from or final analyses. Oral contraceptives usually contain progestin and/or synthetic estrogen [60], which has been associated with elevated 25-OHD in pre-menopausal women, ages 20-40 years [61]. While the effects of progesterone based OC on vitamin D metabolism are not well understood, experimental in vivo studies suggest that estrogen may up-regulate the expression of 1- α -hydroxylase, the enzyme responsible for converting 25-OHD to

1,25(OH)2D in the kidney, as well as the vitamin D receptor and vitamin D binding protein

[62-64]. Estrogen may also down-regulate 24-hydroxylase, the main enzyme that degrades

25-OHD in cells [62]. Together, these effects of estrogen may result in elevated circulating

25-OHD, although the clinical significance of such an effect remains unknown. Additional studies with larger cohorts are needed to further investigate whether oral contraceptive use modifies 25-OHD levels in young women.

129

Some of the strengths of our study include assessment of vitamin D status with a commonly used biomarker, serum levels of 25-OHD [64]. While both 25-OHD and 1, 25- dihydroxyvitamin D concentrations can be specifically measured in serum or plasma [65,

66], the concentration of 25-OHD is considered the best indicator of vitamin D status because it represents a summation of total cutaneous production of vitamin D and intake of vitamin D from foods [65]. Serum 25-OHD also has a fairly long circulating half-life (~ 15 days) (Jones, 2008) as compared to 1, 25(OH)2 D, which has a relatively short half-life of between 4 to 6 hours [67]. Serum 1, 25(OH)2 D concentrations are also closely regulated by parathyroid hormone (PTH), calcium, and phosphate [68]; hence, levels of 1,25(OH)2D do not typically decrease until vitamin D deficiency is severe [1, 69].

Our population comprised of non-obese and obese individuals; and we controlled for several factors (including season of blood draw, smoking status and physical activity level), that could potentially influence a relation between 25-OHD levels and adiposity.

Some important limitations of our study include the cross-sectional design, and the fact that serum 25-OHD level and adiposity measures were assessesed only once. Causal relationships between 25-OHD levels and adiposity can therefore not be inferred from our results. A one-time measurement of adiposity and 25-OHD may not also reflect long term changes in obesity markers and 25-OHD levels. Future studies that evaluate predictors of serum 25-OHD status should employ a prospective design and enroll larger cohorts as this will help to better understand if serum 25-OHD status differs among women with relatively little adiposity.

Our population was also fairly homogenous; i.e., most women had low BMI (< 25 kg/m2), and all participants were between 18 to 30 years. Subjects were also predominately

130

were Caucasian. Our findings are therefore limited to mostly non-obese young Caucasian women and may not be generalizable to women in other age groups, BMI categories, or ethnicities.

4.5 Conclusions and Significance

Presently, the evidence is conflicting on the relationship between serum 25-OHD levels and adiposity in young women. While some cross-sectional studies suggest that 25-OHD concentrations are lower in obese compared with non-obese subjects, other studies in similar populations have failed to confirm these observations. Previous studies have also not explored whether use of oral contraceptives, alcohol intake, or cigarette smoking can influence the association between 25-OHD levels and adiposity risk in reproductive aged women. Obesity is an independent risk factor for many chronic conditions, including cardiovascular disease, type II diabetes mellitus, and several cancers. Understanding and identifying any modifiable lifestyle factors that may be associated with obesity (increased adiposity) could help in formulating nutrition interventions targeted at obesity reduction/prevention. In the present study, serum 25-OHD levels were not associated with adiposity measures in young women ages 18-30 years. Further studies are needed to corroborate our fingings. In the meantime, we suggest that young women (18 to 30 years) should still seek to maintain optimal levels of serum 25-OHD (via food and supplement intake), since vitamin D is important for attaining peak bone mass; which in turn decreases the risk of osteoporosis and osteoporotic fractures in later life.

131

4.6 References

1. Cranney A, Horsley T, O'Donnell S, Weiler H, Puil L, Ooi D, Atkinson S, Ward L, Moher

D, Hanley D, Fang M, Yazdi F, Garritty C, Sampson M, Barrowman N,Tsertsvadze

A, Mamaladze V. Effectiveness and safety of vitamin D in relation to bone health. Evid Rep

Technol Assess 2007; 158: 1-235.

2. Ovesen L, Brot C, Jakobsen J. Food contents and biological activity of 25-hydroxyvitamin

D: a vitamin D metabolite to be reckoned with? Ann Nutr Metab 2003; 47: 107-13.

3. Chen TC, Chimeh F, Lu Z, Mathieu J, Person KS, Zhang A, Kohn N, Martinello S,

Berkowitz R, Holick MF. Factors that influence the cutaneous synthesis and dietary sources

of vitamin D. Arch Biochem Biophys 2007; 460, 213-217.

4. Ford ES, Ajani UA, McGuire LC, Liu S. Concentrations of serum vitamin D and the

metabolic syndrome among U.S. adults. Diabetes Care 2005; 28: 1228-30.

5. Beydoun MA, Boueiz A, Shroff MR, Beydoun HA, Wang Y, Zonderman AB.Associations

among 25-Hydroxyvitamin D, Diet Quality, and Metabolic Disturbance Differ by Adiposity

in Adults in the United States. J Clin Endocrinol Metab 2010; 95: 3814-27.

6. Scragg R, Sowers MF, Bell C. Serum 25-hydroxyvitamin D, diabetes, and ethnicity in the

Third National Health and Nutrition Examination Survey. Diabetes Care 2004; 27: 2813-8.

7. Pittas AG, Dawson-Hughes B, Li T, Van Dam RM, Willett WC, Manson JE, Hu FB.

Vitamin D and calcium intake in relation to type 2 diabetes in women. Diabetes Care 2006;

29: 650-6.

8. Scragg R, Sowers MF, Bell C. Serum 25-hydroxyvitamin D, ethnicity, and blood pressure in

the Third National Health and Nutrition Examination Survey. Am J Hypertens 2007; 20:

713-9.

132

9. Giovannucci E, Liu Y, Hollis BW, Rimm EB. 25-Hydroxyvitamin D and Risk of

Myocardial Infarction in Men; A Prospective Study. Arch Intern Med 2008; 168: 1174-80.

10. Poole KES, Loveridge N, Barker PJ, Halsall DJ, Rose C, Reeve J, Warburton EA. Reduced

vitamin D in acute stroke. Stroke 2006; 37:243-245.

11. Cigolini M, Iagulli MP, Miconi V, Galiotto M, Lombardi S, Targher G. Serum 25

hydroxyvitamin D3 concentrations and prevalence of cardiovascular disease among type 2

diabetic patients. Diabetes Care 2006; 29: 722-4.

12. Martins D, Wolf M, Pan D, Zadshir A, Tareen N, Thadhani R, Felsenfeld A, Levine

B, Mehrotra R, Norris K. Prevalence of cardiovascular risk factors and the serum levels of

25-hydroxyvitamin D in the United States: data from the Third National Health and

Nutrition Examination Survey. Arch Intern Med 2007; 167: 1159-65.

13. Banerjee P, Chatterjee M. Anti-proliferative role of vitamin D and its analog-a brief

overview. Mol Cell Biochem 2003; 253: 247-54.

14. Looker AC, Pfeiffer CM, Lacher DA, Schleicher RL, Picciano MF, Yetley EA. Serum 25-

hydroxyvitamin D status of the U.S. population: 1988–1994 compared with 2000–2004. Am

J Clin Nutr 2008; 88: 1519-27.

15. Zemel MB, Shi H, Greer B, Dirienzo D, Zemel PC. Regulation of adiposity by dietary

calcium. FASEB J 2000; 14, 1132-38.

16. Konradsen S, Ag H, Lindberg F, Hexeberg S, Jorde R. Serum 1, 25-dihydroxy vitamin D is

inversely associated with body mass index. Eur J Nutr 2008; 47: 87-91.

17. Davies KM, Heaney RP, Recker RR, Lappe JM, Barger-Lux MJ, Rafferty K, Hinders S.

Calcium Intake and Body Weight. J Clin Endocrinol Metab 2000; 85: 4635-8.

133

18. Lovejoy JC, Champagne CM, Smith SR, de Jonge L, Xie H. Ethnic differences in dietary

intakes, physical activity, and energy expenditure in middle-aged, premenopausal women,

the Healthy Transitions Study. Am J Clin Nutr 2001; 74, 90-5.

19. Vilarrasa N, Maravall J, Estepa A, Sanchez R, Masdevall C, Navarro MA, Alia P, Soler

J, Gomez JM. Low 25-hydroxyvitamin D concentrations in obese women: their clinical

significance and relationship with anthropometric and body composition variables. J

Endocrinol Invest 2007; 30: 653-8.

20. Kremer R, Campbell PP, Reinhardt T, Gilsanz V. Vitamin D Status and Its Relationship to

Body Fat, Final Height, and Peak Bone Mass in Young Women. J Clin Endocrinol Metab

2009; 94: 67-73.

21. Bell NH, Epstein S, Greene A, Shary J, Oexmann MJ, Shaw S. Evidence for alteration of

the vitamin D-endocrine system in obese subjects. J Clin Invest 1985; 76: 370-3.

22. Liel Y, Ulmer E, Shary J, Hollis BW, Bell NH. Low circulating vitamin D in obesity. Calcif

Tissue Int 1988; 43: 199-201.

23. Wortsman J, Matsuoka LY, Chen TC, Lu Z, Holick MF. Decreased bioavailability of

vitamin D in obesity. Am J Clin Nutr 2000; 72: 690-3.

24. Rodriguez-Rodriguez E, Navia B, Lopez-Sobaler AM, Ortega RM. Vitamin D in

overweight / obese women and its relationship with dietetic and anthropometric

variables. Obesity 2009; 17: 778-82.

25. Parikh SJ, Edelman M, Uwaifo GI, Freedman RJ, Semega-Janneh M, Reynolds J, Yanovski

JA. The Relationship between Obesity and Serum 1, 25-Dihydroxy Vitamin D

Concentrations in Healthy Adults. J Clin Endocrinol Metab 2004; 89: 1196-9.

134

26. Maetani M, Maskarinec G, Franke AA, Cooney RV. Association of leptin, 25-

hydroxyvitamin D, and parathyroid hormone in women. Nutr Cancer 2009; 61: 225-31.

27. Van der Wielen RPJ, Lowik MRH, van den Berg H, de Groot LC, Haller J, Moreiras O, van

Staveren WA. Serum vitamin D concentrations among elderly people in Europe. Lancet

1995; 346: 207-10.

28. Arunabh S, Pollack S, Yeh J, Aloia JF. Body Fat Content and 25-Hydroxyvitamin D Levels

in Healthy Women. J Clin Endocrinol Metab 2003; 88: 157-61.

29. Caron-Jobin M, Morisset AS, Tremblay A, Huot C, Legare D, Tchernof A. Elevated serum

25-OHD concentrations, vitamin D, and calcium intakes are associated with reduced

adipocyte size in women. Obesity 2011; 19: 1335-41.

30. Wortsman J, Matsuoka LY, Chen TC, Lu Z, Holick MF. Decreased bioavailability of

vitamin D in obesity. Am J Clin Nutr 2000; 72: 690-3.

31. Shi H, Norman AW, Okamura WH, Sen A, Zemel MB. 1, 25-Dihydroxyvitamin D3

Modulates Human Adipocyte Metabolism via Nongenomic Action. FASEB J 2001; 15:

2751-3.

32. Hollis, B. Comparison of commercially available (125) I-based RIA method for the

determination of circulating 25-hydroxyvitamin D. Clin Chem 2000; 46: 1657-61.

33. Holick MF, Binkley NC, Bischoff-Ferrari HA, Gordon CM, Hanley DA, Heaney RP, Murad

MH, Weaver CM: Evaluation, treatment, and prevention of vitamin D deficiency: an

Endocrine Society clinical practice guideline. J Clin Endocrinol Metab 2011; 96:1911-30.

34. Norman AW, Bouillon R, Whiting SJ, Vieth R, Lips P. 13th Workshop consensus for the

vitamin D nutritional guidelines. J Steroid Biochem Mol Bio 2007; 103:204-5.

135

35. Willet WC, Sampson L, Stampler MJ, Rosner B, Bain C, Witschi J, Hennekens CH, Speizer

FE. Reproducibility and validity of a semi-quantitative food frequency questionnaire. Am J

Epidemiol 1985; 122, 51-65.

36. Feskanich D, Rimm EB, Giovannucci EL, Colditz GA, Stampfer MJ. Reproducibility and

validity of food intake measurements from a semi-quantitative food frequency

questionnaire. J Am Diet Assoc 1993; 93, 790-6.

37. Food and Nutrition Board (FNB) / Institute of Medicine (IOM). Dietary Reference Intakes

for Calcium and Vitamin D: 2010. 2101 Constitution Avenue, N.W. Washington,

DC 20418. National Academy Press.

38. Wolf AM, Hunter DJ, Colditz GA, Manson JE, Stamoer MJ, Corsano KA., Rosner B,

Kriska A, Willet WC. Reproducibility and validity of a self-administered physical activity

questionnaire. Int J Epidemiol 1994; 23, 991-9.

39. Ainsworth BE, Haskell WL, Leon AS, Jacobs DR jr, Montoye HJ, Sallis JF, Paffenbarger

RS, jr. Compendium of physical activities: classification of energy costs of human physical

activities. Med Sci Sports Exerc. 1993; 25: 1575-81.

40. World Health Organization. Preventing and Managing the Global Epidemic. WHO, Geneva.

WHO Obesity Technical Report Series, Obesity 2000; Number 894.

41. National Heart, Lung, and Blood Institute (NHLBI) and the North American Association for

the Study of Obesity (NAASO). The Practical Guide: Identification, Evaluation, and

Treatment of Overweight and Obesity in Adults, 2000. NIH Publication Number 00-4084.

Available online: http://www.nhlbi.nih.gov/guidelines/obesity/prctgd_c.pdf.

136

42. International Diabetes Federation.The IDF Consensus Worldwide Definition of the

Metabolic Syndrome, Avenue Emile De Mot 19, B-1000 Brussels, Belgium: 2006.

Available online: http//www.idf.org/webdata/docs/IDF_Meta_def_final.pdf.

43. Jackson AS, Pollack ML. Practical assessment of body composition. Physician and

Sportsmed 1985; 13: 76-90.

44. American Council on Exercise. What are the guidelines for percentage of body fat loss,

2009? Available online: http://www.acefitness.org/blog/112/what-are-the guidelines-for-

percentage-of-body-fat.

45. Webb AR, DeCosta BR, Holick MF. Sunlight regulates the cutaneous production of vitamin

D3 by causing its photodegradation. J Clin Endocrinol Metab 1989; 68: 882-7.

46. Ding C, Parameswaran V, Blizzard L, Burgess J, Jones G. Not a simple fat soluble vitamin:

changes in serum 25-(OH)D levels are predicted by adiposity and adipocytokines in older

adults. J Intern Med 2010; 268: 501-10.

47. Cheng S, Massaro JM, Fox CS, Larson MG, Keyes MJ, McCabe EL, Robins SJ, O’Donnell

CJ, Hoffmann U, Jacques PF, Booth SL, Vasan RS, Wolf M, Wang TJ. Adiposity,

cardiometabolic risk, and vitamin D status: the Framingham Heart Study. Diabetes 2010;

59: 242-8.

48. Chacko SA, Song Y, Manson JE, Van Horn L, Eaton C, Martin LW, McTiernan A, Curb

JD, Wylie-Rosett J, Phillips LS, Plodkowski RA, Liu S. Serum 25-hydroxyvitamin D

concentrations in relation to cardiometabolic risk factors and metabolic syndrome in

postmenopausal women. Am J Clin Nutr 2011; 94: 209-17.

137

49. Shi H, Norman AW, Okamura WH, Anintida S, Zemel MB. 1α, 25-dihydroxyvitamin

D3 inhibits uncoupling protein 2 expression in human adipocytes. FASEB J 2002; 16: 1808-

10.

50. Jungert A, Roth HJ, Neuhauser-Berthold M. Serum 25-hydroxyvitamin D3 and body

composition in an elderly cohort from Germany: a cross-sectional study. Nutr Metab 2012;

9: 42.

51. Moschonis G, Tanagra S, Koutsikas K, Nikolaidou A, Androutsos O, Manios Y.

Association between serum 25-hydroxyvitamin D levels and body composition in

postmenopausal women: the Postmenopausal Health Study. Menopause 2009; 16:701-7.

52. Du H, Bennett D, Li L, Whitlock G, Guo Y, Collins R, Chen J, Bian Z, Hong LS, Feng

S, Chen X, Chen L, Zhou R, Mao E, Peto R, Chen Z; China Kadoorie Biobank

Collaborative Group. Physical activity and sedentary leisure time and their associations with

BMI, waist circumference, and percentage body fat in 0.5 million adults: the China

Kadoorie Biobank study. Am J Clin Nutr 2013; 97: 487-96.

53. Dunton GF, Berrigan D, Ballard-Barbash R, Graubard B, Atienza AA. Joint associations of

physical activity and sedentary behaviors with body mass index: results from a time use

survey of US adults. Int J Obes (Lond) 2009; 33: 1427-36.

54. Need AG, Morris HA, Horowitz M, Nordin C. Effects of skin thickness, age, body fat, and

sunlight on serum 25-hydroxyvitamin D. Am J Clin Nutr 1993; 58: 882-5.

55. Millen AE, Bodnar LM. Vitamin D assessment in population-based studies: a review of the

issues. Am J Clin Nutr 2008; 87: 1102S-5S.

56. Tangpricha V, Pearce EN, Chen TC, Holick MF. Vitamin D Insufficiency among Free-

Living Healthy Young Adults. Am J Med 2002; 112: 659-62.

138

57. Brot C, Jorgensen NR, Sørensen OH. The influence of smoking on vitamin D status and

calcium metabolism. Eur J Clin Nutr 1999; 53: 920-6.

58. Ilich JZ, Brownbill RA, Tamborini L, Crncevic-Orlic Z: To drink or not to drink: how are

alcohol, caffeine and past smoking related to bone mineral density in elderly women? J Am

Coll Nutr 2002; 21: 536-44.

59. Obarzanek E, Schreiber GB, Crawford PB, Goldman SR, Barrier PM, Frederick MM,

Lakatos E. Energy intake and physical activity in relation to indexes of body fat: the

National Heart, Lung, and Blood Institute Growth and Health Study. Am J Clin Nutr 1994;

60: 15-22.

60. Williams JK. Rationale for new oral contraceptive dosing. Int J Fertil Womens Med 2004;

49: 30-5.

61. Harris SS, Dawson-Hughes B. The association of oral contraceptive use with plasma 25-

hydroxyvitamin D levels. J Am Coll Nutr 1998; 17: 282-4.

62. Abbas S, Linseisen J, Slanger T, Kropp S, Mutschelknauss EJ, Flesch-Janys D, Chang-

Claude J. Serum 25-hydroxyvitamin D and risk of postmenopausal breast cancer, results of

a large case-control study. Carcinogenesis 2008; 29: 93-9.

63. Schwartz B, Smirnoff P, Shany S, Liel Y. Estrogen controls expression and

bioresponse of 1,25dihydroxyvitamin D receptors in the rat colon. Mol Cell Biochem 2000;

203: 87-93.

64. Buchanan JR, Santen R, Cauffman S, Cavaliere A, Greer RB, Demers LM. The effect of

endogenous estrogen fluctuation on metabolism of 25-hydroxyvitamin D. Calcif Tissue

Int 1986; 39: 139-44.

139

65. Brannon PM, Yetley EA, Bailey RL, Picciano MF. Overview of the conference “Vitamin D

and health in the 21st century: an update”. Am J Clin Nutr 2008; 88:483-90S.

66. Holick MF. The use and interpretation of assays for vitamin D and its metabolites. J

Nutr 1990; 120: 1464-9.

67. Chen TC, Turner AK, Holick MF. Methods for the determination of the circulating

concentration of 25-hydroxyvitamin D. J Nutr Biochem 1990; 1: 315-9.

68. Jones G. Pharmacokinetics of vitamin D toxicity. Am J Clin Nutr 2008; 88: 582S-6S.

69. Holick MF. Vitamin D status: measurement, interpretation and clinical application. Ann

Epidemiol 2009; 19: 73-8.

140

Table 19. Distribution of Serum 25-Hydroxyvitamin D (25-OHD) Levels

Mean ± Standard Deviation Range Serum 25-OHD, nmol/L 79.2 ± 32.3 21.7 - 174.3

n (%) Serum 25-OHD, nmol/L‡ (< 50 nmol/L) 28 (16.1) (50 - 75 nmol/L) 65 (37.4) (≥ 75 nmol/L) 81 (46.5)

‡ 2.496 nmol/L = 1 ng/mL.

141

Table 20. Variations in Characteristics of Participants by Serum 25-OHD Levels

Covariate Low 25-OHD High 25-OHD P ⱶ < 75 nmol/L§ ≥ 75 nmol/L§ (n = 93) (n = 81) Mean ± Standard Deviation Age, y 21.7 ± 3.3 21.5 ± 3.1 0.54 BMI, kg/m2 22.8 ± 3.3 23.1 ± 3.3 0.65 Waist Circumference, cm 78.2 ± 8.6 79.1 ± 9.2 0.51 Waist to Height, ratio 0.47 ± 0.05 0.48 ± 0.07 0.53 Total Body Fat, % 31.6 ± 8.4 32.8 ± 8.2 0.34 Arm Fat, % 29.3 ± 9.5 30.5 ± 9.0 0.40 Leg Fat, % 35.7 ± 8.2 36.2 ± 7.2 0.66 Trunk Fat, % 31.0 ± 9.7 32.8 ± 9.8 0.24 Android Fat, % 33.8 ± 10.6 35.8 ± 10.6 0.23 Gynoid Fat, % 40.5 ± 7.3 41.1 ± 6.4 0.60 Android : Gynoid Ratio 0.82 ± 0.18 0.86 ± 0.16 0.25 Physical Activity, METs/week╔ 43.7 ± 36.3 56.5 ± 49.6 0.05 Systolic Blood Pressure, mm Hg 103.8 ± 7.6 105.9 ± 10.2 0.12 Diastolic Blood Pressure, mm Hg 68.5 ± 6.8 69.3 ± 8.1 0.44 Fasting Blood Glucose, mg/dL ϖ 79.5 ± 9.0 79.9 ± 10.8 0.82 Serum 25-OHD, nmol/L 55.7 ± 13.1 106.3 ± 25.8 < 0.01 ⱶ P-values for test of Null: Equal means, estimated with student t-test. ╔ METs, metabolic equivalents, n = 173 for physical activity. ϖ n = 163 for fasting blood glucose. ‡ 2.496 nmol/L = 1 ng/mL.

142

Table 20. Continued

Low 25-OHD High 25-OHD P ⱶ < 75 nmol/L§ ≥ 75 nmol/L§ Covariate (n = 93) (n = 81) n (%) BMI, kg/m2 0.61 < 25 72 (77.4) 60 (74.1) ≥ 25 21 (22.6) 21 (25.9) Waist Circumference, cm 0.74 < 80 54 (58.1) 45 (55.6) ≥ 80 39 (41.9) 36 (44.4) Total Body Fat, % 0.62 < 32 46 (49.5) 37 (45.7) ≥ 32 47 (50.5) 44 (54.3) Android: Gynoid Ratio 0.29 < 1 72 (77.4) 57 (70.4) ≥ 1 21 (22.6) 24 (29.6) Race 0.86 Caucasian 79 (85.0) 68 (84.0) Other 14 (15.0) 13 (16.0) Education 0.91 Some College 74 (79.6) 65 (84.2) College Graduate 19 (20.4) 16 (19.8) Currently Drink Alcohol 0.43 No 34 (36.6) 25 (30.9) Yes 59 (63.4) 56 (69.1) Currently Smoke 0.13 No 86 (92.5) 79 (97.5) Yes 7 (7.5) 2 (2.5) Currently Use Oral Contraceptive 0.07 No 63 (67.7) 44 (54.3) Yes 30 (32.3) 37 (45.7) ⱶ P-values for chi-squared tests. ‡ 2.496 nmol/L = 1 ng/mL.

143

Table 21. Dietary Intakes of Participants by Serum 25-OHD Levels

Low 25-OHD High 25-OHD P ⱶ < 75 nmol/L§ ≥ 75 nmol/L§ Dietary Intake (n = 93) (n = 81) Mean ± Standard Error Energy, kcal/day 2205 ± 84 2223 ± 87 0.88 Total Protein, g/day 104 ± 6 102 ± 4 0.75 Total Fat, g/day 72 ± 3 73 ± 3 0.69 Total Calcium Intake, mg/day 1130 ± 54 1195 ± 61 0.43 Food calcium 1042 ± 51 1023 ± 58 0.81 Supplemental calcium (n = 63 users) 275 ± 39 420 ± 55 0.04 Vitamin A, IU/day Ϫ 16701 ± 1069 15532 ± 1121 0.45 Thiamin, mg/day 1.9 ± 0.1 1.8 ± 0.1 0.52 Riboflavin, mg/day 2.7 ± 0.1 2.5 ± 0.1 0.48 Niacin, mg/day 29 ± 2 28 ± 1 0.56 Pantothenic acid, mg/day 7.1 ± 0.4 6.4 ± 0.3 0.20 Vitamin B6, mg/day 2.6 ± 0.1 2.5 ± 0.1 0.52 Folate, µg/day 695 ± 40 647 ± 40 0.39 Vitamin B12, µg/day 6.1 ± 0.4 6.2 ± 0.5 0.88 Vitamin C, mg/day 150 ± 10 144 ± 9 0.61 Total Vitamin D Intake, IU/day Ϫ 355 ± 27 403 ± 34 0.26 Food vitamin D 223 ± 19 222 ± 20 0.99 Supplemental vitamin D (n = 77 users) 307 ± 37 396 ± 40 0.10 Iron, mg/day 19 ± 1 18 ± 1 0.85 Zinc, mg/day 14 ± 1 14 ± 1 0.78 Linoleic acid, g/day 23 ± 1 24 ± 1 0.44 Linolenic acid, g/day 12 ± 1 13 ± 1 0.65 Dairy Intake, total servings/day₸ 2.0 ± 0.2 1.8 ± 0.2 0.56 § 2.496 nmol/L = 1 ng/mL. ⱶ P-values for test of Null: Equal means, estimated with student t-tests. Ϫ IU, International Units: 1 IU = 0.025 µg. ₸ n = 166 for dairy intake responders. 1 serving dairy = 8 fluid oz. (1 cup) of milk or yogurt; 1.5 oz of natural cheese; 2 oz of processed cheese, 2 cups of cottage cheese.

144

Table 22. Mean Serum 25-OHD Levels by Oral Contraceptive Use, Alcohol Intake and Smoking Status

Oral Contraceptive Use Yes (n = 67 ) No (n = 107) Mean ± Standard Deviation P ⱶ Serum 25-OHD, nmol/L§ 83.0 ± 34.0 76.9 ± 31.1 0.23

Smoking status Yes (n = 9 ) No (n = 165) Mean ± Standard Deviation P ⱶ Serum 25-OHD, nmol/L§ 65.1 ± 21.9 80.0 ± 32.6 0.18

Drink Alcohol Yes (n = 115 ) No (n = 59) Mean ± Standard Deviation P ⱶ Serum 25-OHD, nmol/L§ 83.2 ± 32.3 77.5 ± 32.5 0.61

ⱶ P-values for test of Null: Equal means, estimated with student t-tests. § 2.496 nmol/L = 1 ng/mL.

145

Figure 14. Covariation and Pearson Product Correlation of Vitamin D Intake and Serum 25-OHD Concentration

200 200 ß = 0.002, P = 0.83 ß = -0.004, P = 0.75 200 ß = 0.005, P = 0.61

150 150 r = 0.02 r = -0.02 150 r = 0.04

100 100

100

Serum 25-OHD, nmol/L

Serum 25-OHD, nmol/L 25-OHD, Serum

Serum 25-OHD,nmol/L

0 0

0 500 1000 1500 0 200 400 600 800 1000 0 500 1000 1500 Total Vitamin D Intake, IU/day Vitamin D from Food Sources, IU/day Vitamin D from Dietary Supplement Intake, IU/day 95% CI Fitted values* 95% CI Fitted values* 95% CI Fitted values Serum 25-OHD Serum 25-OHD Serum 25-OHD

* Fitted values are simple Linear Regression.

146

Figure 15. Covariation and Pearson Product Correlation of Serum 25-OHD and Body Mass Index

All women Low BMI < 25 kg/m2 High BMI >= 25 kg/m2 (n = 42)

(n = 174) (n = 132)

200 200

200

r = 0.07, P = 0.37 r = 0.04, P = 0.67 r = -0.11, P = 0.51

150

100

Serum 25-OHD, nmol/L

0 0

15 20 25 30 35 16 18 20 22 24 26 25 30 35 Body Mass Index, kg/m2 Body Mass Index, kg/m2 Body Mass Index, kg/m2

95% CI Fitted values* 95% CI Fitted values* 95% CI Fitted values* Serum 25-OHD concentration Serum 25-OHD concentration Serum 25-OHD concentration

* Fitted values are simple Linear Regression.

147

Figure 16. Covariation and Pearson Product Correlation of Serum 25-OHD and Waist Circumference

All women Low WC < 80 cm High WC >= 80 cm

(n = 174) (n = 99) (n = 75)

200 200

r = 0.04, P = 0.57 r = -0.003, P = 0.97 200 r = 0.06, P = 0.61

150

100

Serum 25-OHD, nmol/L

0 0

60 80 100 120 60 65 70 75 80 80 90 100 110 120 Waist Circumference, cm Waist Circumference, cm Waist Circumference, cm

95% CI Fitted values* 95% CI Fitted values* 95% CI Fitted values* Serum 25-OHD concentration Serum 25-OHD concentration Serum 25-OHD concentration

* Fitted values are simple Linear Regression.

148

Figure 17. Covariation and Pearson Product Correlation of Serum 25-OHD Total Body Fat percentage

All women Low TBF < 32 % High TBF >= 32 %

(n = 174) (n = 83) (n = 91)

200 200

r = 0.05, P = 0.53 r = -0.07, P = 0.55 200 r = 0.11, P = 0.32

150 150

150

100 100

100

50 50

Serum 25-OHD, nmol/L Serum 25-OHD, nmol/L

Serum 25-OHD, nmol/L

0 0 0

10 20 30 40 50 15 20 25 30 35 30 35 40 45 50 55 Total Body Fat, % Total Body Fat, % Total Body Fat, % 95% CI Fitted values* 95% CI Fitted values* 95% CI Fitted values* Serum 25-OHD concentration Serum 25-OHD concentration Serum 25-OHD concentration

* Fitted values are simple Linear Regression.

149

Figure 18. Covariation and Pearson Product Correlation of Serum 25-OHD and Regional Adiposity Measures (Arm and Leg Fat)

(n = 174) (n = 174)

200 200

r = 0.02, P = 0.85 r = 0.02, P = 0.83

150

100

Serum 25-OHD, nmol/L

0 0

10 20 30 40 50 10 20 30 40 50 Arm Fat, % Leg Fat, %

95% CI Fitted values* 95% CI Fitted values* Serum 25-OHD concentration Serum 25-OHD concentration

* Fitted values are simple Linear Regression.

150

Figure 19. Covariation and Pearson Product Correlation of Serum 25-OHD and Regional Adiposity Measures (Trunk, Android and Gynoid Fat)

(n = 174) (n = 174) (n = 174)

200 200

r = 0.07, P = 0.37 r = 0.06, P = 0.40 200 r = 0.01, P = 0.86

150

100

Serum 25-OHD, nmol/L

0 0

10 20 30 40 50 60 10 20 30 40 50 60 20 30 40 50 60 Trunk Fat, % Android Fat, % Gynoid Fat, %

95% CI Fitted values* 95% CI Fitted values* 95% CI Fitted values* Serum 25-OHD concentration Serum 25-OHD concentration Serum 25-OHD concentration

* Fitted values are simple Linear Regression.

151

Figure 20. Covariation and Pearson Product Correlation of Serum 25-OHD and Android to Gynoid Fat ratio (AGF ratio)

All women Low AGF ratio < 1 High AGF ratio >= 1

(n = 174) (n = 129) (n = 45)

200 200

r = 0.07, P = 0.38 r = 0.01, P = 0.89 200 r = 0.07, P = 0.64

150

100

Serum 25-OHD, nmol/L

0 0

.5 1 1.5 2 .4 .6 .8 1 1 1.2 1.4 1.6 1.8 Android:Gynoid Fat, ratio Android:Gynoid Fat, ratio Android:Gynoid Fat, ratio 95% CI Fitted values* 95% CI Fitted values* 95% CI Fitted values* Serum 25-OHD concentration Serum 25-OHD concentration Serum 25-OHD concentration

* Fitted values are simple Linear Regression.

152

Table 23. Mean Serum 25-OHD Levels by Adiposity

Low BMI < 25 kg/m2 High BMI ≥ 25 kg/m2 P-value ⱶ (n = 132) (n = 42) Mean ± Standard Deviation § Serum 25 -OHD, nmol/L 77.6 ± 31.1 84.5 ± 35.8 0.23

Low WC < 80 cm High WC ≥ 80 cm P-value ⱶ (n = 99) (n = 75) Mean ± Standard Deviation Serum 25-OHD, nmol/L§ 78.3 ± 31.6 80.5 ± 33.3 0.67

Low TBF < 32 % High TBF ≥ 32% P-value ⱶ (n = 83) (n = 91) Mean ± Standard Deviation Serum 25-OHD, nmol/L§ 78.0 ± 31.9 80.4 ± 32.8 0.63

§ 2.496 nmol/L = 1 ng/mL. ⱶ P-values for test of Null: Equal means, estimated with student t-test.

153

Table 24. Unadjusted and Adjusted Mean Values of Adiposity by Serum 25-OHD Levels

(X = Serum 25-OHD Concentration) < 50 nmol/L 50-75 nmol/L ≥ 75 nmol/L P diff† P trend‡ (Y = Adiposity) (n = 28 ) (n = 65 ) (n = 81 ) Unadjusted Means ± Standard Error BMI, kg/m2 22.3 ± 0.6 23.1 ± 0.4 23.1 ± 0.4 0.57 0.52 Waist Circumference, cm 76.8 ± 1.7 78.8 ± 1.1 79.1 ± 1.0 0.50 0.78 Total Body Fat, % 30.9 ± 1.6 31.9 ± 1.0 32.8 ± 0.9 0.58 0.21 Arm Fat, % 28.5 ± 1.8 29.6 ± 1.2 30.5 ± 1.0 0.61 0.49 Leg Fat, % 35.5 ± 1.5 35.7 ± 1.0 36.2 ± 0.9 0.90 0.74 Trunk Fat, % 29.9 ± 1.8 31.5 ± 1.2 32.8 ± 1.1 0.39 0.19 Android Fat, % 32.5 ± 2.0 34.4 ± 1.3 35.8 ± 1.2 0.36 0.44 Gynoid Fat, % 40.5 ± 1.3 40.5 ± 0.9 41.1 ± 0.8 0.87 0.72 Android : Gynoid Fat, ratio 0.79 ± 0.03 0.84 ± 0.02 0.86 ± 0.02 0.23 0.60 † Analysis of variance (ANOVA) was used to compare group differences. ‡ P-value for trend estimated with simple linear regression using median of tertiles of adiposity.

154

Table 24. Continued

(X = Serum 25-OHD Concentration) < 50 nmol/L 50-75 nmol/L ≥ 75 nmol/L P diff† P trend ѣ (Y = Adiposity) (n = 28 ) (n = 65 ) (n = 81 ) Adjusted Means ± Standard Error ¤ BMI, kg/m2 22.2 ± 0.8 23.0 ± 0.7 22.7 ± 0.7 0.57 0.28 Waist Circumference, cm 75.6 ± 2.2 77.9 ± 1.8 77.5 ± 1.8 0.52 0.28 Total Body Fat, % 29.4 ± 2.0 30.5 ± 1.7 31.3 ± 1.7 0.39 0.14 Arm Fat, % 27.7 ± 2.3 28.8 ± 1.9 29.6 ± 1.9 0.27 0.15 Leg Fat, % 34.6 ± 1.9 34.7 ± 1.6 35.3 ± 1.5 0.41 0.05 Trunk Fat, % 27.8 ± 2.4 29.5 ± 2.0 30.6 ± 1.9 0.37 0.19 Android Fat, % 30.1 ± 2.6 32.1 ± 2.1 33.2 ± 2.1 0.43 0.29 Gynoid Fat, % 39.1 ±1.7 39.0 ± 1.4 39.6 ± 1.4 0.26 0.10 Android : Gynoid Fat, ratio 0.76 ± 0.04 0.82 ± 0.03 0.82 ± 0.03 0.39 0.58 † Analysis of variance (ANOVA) was used to compare group differences. ¤ Mean values estimated with ANOVA and adjusted for season, smoking status and physical activity level. ѣ P-value for trend estimated with multiple linear regression using median of tertiles of adiposity and adjusted for season, smoking status and physical activity level.

155

Table 25. Linear Relationship between Serum 25-OHD Levels and Other Parameters

Serum 25-OHD

Independent Predictor ß ± SE P-value

Age, y -0.13 ± 0.77 0.86 Education a - 0.13 ± 6.12 0.98 Season of blood draw b -7.27 ± 6.04 0.23 Winter (n = 36) Reference Reference Spring (n = 98) -9.70 ± 6.29 0.13 Summer (n = 3) -3.78 ± 19.38 0.85 Fall (n = 37) -1.11 ± 7.55 0.88 Smoking status c -14.92 ± 11.03 0.18 Alcohol intake c 2.68 ± 5.18 0.61

Oral contraceptive use c 6.06 ± 5.02 0.23 Physical activity level, MET-hrs/week 0.10 ± 0.06 0.08 Total calcium intake, g/day -0.002 ± 0.01 0.67 Food calcium intake, g/day -0.004 ± 0.01 0.43 Supplemental calcium intake, g/day 0.01 ± 0.01 0.47 Total vitamin D intake, IU/day 0.002 ± 0.01 0.83 Food vitamin D intake, IU/day -0.004 ± 0.01 0.75 Supplemental vitamin D intake, IU/day 0.01 ± 0.01 0.61 ϖ Regression coefficients (ß) ± standard errors (SE). All analyses were estimated using simple linear regressions. a Dummy variable (0 = some college / 1 = college graduate). b Dummy variable (0 = winter / 1 = non-winter, i.e. Spring, Summer, Fall). c Dummy variable (0 = no / 1 = yes).

156

Table 26. Linear Relationship between Serum 25-OHD Levels and Adiposity

Serum 25-OHD Concentration

§ Unadjusted Multivariable Exposure ß ± SE P-value Adjusted P-value ß ± SE BMI, kg/m2 0.67 ± 0.75 0.37 0.39 ± 0.75 0.60 Waist Circumference, cm 0.16 ± 0.28 0.57 0.06 ± 0.28 0.83

Total Body Fat, % 0.19 ± 0.30 0.53 0.17 ± 0.30 0.56 Arm Fat, % 0.05 ± 0.27 0.85 0.04 ± 0.27 0.89 Leg Fat, % 0.07 ± 0.32 0.83 0.09 ± 0.32 0.77

Trunk Fat, % 0.22 ± 0.25 0.37 0.20 ± 0.25 0.43 Android Fat, % 0.19 ± 0.23 0.40 0.15 ± 0.23 0.51 Gynoid Fat, % 0.06 ± 0.36 0.86 0.07 ± 0.36 0.86 Android : Gynoid Fat, ratio 12.6 ± 14.2 0.38 7.84 ± 14.29 0.58

All analyses were estimated using simple or multiple linear regressions. §Multivariable adjusted for season, smoking status and physical activity level.

157

CHAPTER 5

ASSOCIATIONS BETWEEN SERUM 25-HYDROXYVITAMIN D

CONCENTRATION, ADIPOSITY AND INFLAMMATORY BIOMARKERS IN

YOUNG WOMEN (18 – 30 YEARS)

A. A. Addo-Lartey §, E. R. Bertone-Johnson ‡, C. Bigelow ‡, R. J. Wood §, B. Whitcomb‡,

S. H. Gehlbach ‡, A. G. Ronnenberg §

§ Department of Nutrition, School of Public Health and Health Sciences, University of

Massachusetts Amherst, MA, 01003, USA

‡ Division of Epidemiology and Biostatistics, Department of Public Health, School of

Public Health and Health Sciences, University of Massachusetts Amherst, MA, 01003,

USA

Abstract: Vitamin D deficiency (serum 25-hydroxyvitamin D, 25-OHD, < 50 nmol/L) and increased adiposity have been linked with higher levels of inflammatory biomarkers in obese/older adults and chronically ill patients. We used cross-sectional data from 270 healthy young women (18-30 years) participating in the UMass Amherst Vitamin D

Status Study to evaluate the extent to which serum 25-OHD levels are associated with inflammatory biomarkers. 25-OHD levels and inflammatory factors including highly sensitive C-reactive protein (hs-CRP); interleukins (IL)-1β, 2, 4 , 5, 6, 7, 8, 10, 12p70,

13; granulocyte macrophage colony stimulating factor (GM-CSF); interferon-gamma

(IFN-γ) and tumor necrosis factor-alpha (TNF-α) were assayed in serum using enzyme immunosorbent assays or a multiplex microsphere-bead array. Most inflammatory factors were strongly correlated with each other (P < 0.01); however, hs-CRP was not correlated

158

with other inflammatory markers (P > 0.05). Levels of IL-1β, IL-7, GM-CSF, and TNF-α were higher in women with waist circumference < 80 cm (P < 0.01). Mean 25-OHD levels were also greater among women with high hs-CRP (≥ 5 mg/L), compared to women with low hs-CRP (< 5 mg/L), (91.6 ± 35.8 vs. 76.4 ± 31.1 mg/L, P = 0.02). After adjustment for season, physical activity level, adiposity, alcohol and multivitamin intake,

25-OHD (< 75 nmol/L) was negatively associated with IL-2 and GM-CSF concentrations

(ß = -0.02 ± 0.01 pg/mL, P = 0.04), and marginally associated with IL-6 and IL-7 concentrations (ß = -0.02 ± 0.01 pg/mL, P = 0.05). These findings suggest that IL-2, GM-

CSF, IL-6, and IL-7 levels may be reduced in young women with 25-OHD < 75 nmol/L.

Further studies are needed to investigate causal relations between 25-OHD status and inflammation in young women.

Keywords: Vitamin D, Serum 25-hydroxyvitamin D, Adiposity, Inflammation, Young women

Definition: CRP and hs-CRP assays measure the same molecule in blood; but, hs-CRP is a more sensitive test for the detection of very small amounts of CRP (as low as 0.025 mg/L).

5.1 Introduction

Viseral obesity is commonly associated with a chronic, low-grade inflammation of white adipose tissue, characterized by increased production of pro-inflammatory cytokines

(e.g., interleukin-1, interleukin-6, and tumor necrosis factor- α) and acute phase reactant proteins (e.g., C - reactive protein) [1-4]., some of which are actually secreted by adipose tissue itself [5-8]. These cytokines may have local physiological effects on white adipose tissue physiology as well as systemic effects on other organs [9]. Among patients

159

with coronary artery disease, Verghese et al. found that visceral fat area is highly correlated with CRP levls [10]. Cross-sectional studies in Korean adults showed a positive association between visceral adiposity and elevated CRP and/or cytokine levels

(TNF-α, and IL-6) in obese subjects [11]. Abdominal obesity, assessed by android to gynoid fat ratio, was an independent predictor of CRP and IL-6 levels in obese adults

[11]. Additional studies in similar cohorts have suggested that visceral fat mass may be the most important predictor of high sensitive-CRP (hs-CRP) concentration [12].

Elevated CRP levels is undesirable because it is an independent risk factor for several chronic diseases, including cardiovascular disease [15, 22], and type II diabetes [15, 23],

Vitamin D modulates immune response; and cross-sectional studies in obese but healthy U.S. adults have shown an inverse association between serum 25-OHD concentrations and plasma CRP concentrations [13-14]. Some have proposed that improving vitamin D status (1-alpha, 25-dihydroxyvitamin D3 concentration, (1,

25(OH)2D3), may help to suppress inflammation by meditating the release of specific cytokines and expression of cytokine receptors [24-26]. Although accumulating research strongly implicates vitamin D in the modulation of systemic inflammation, most of these findings have been reported in post-menopausal, obese, or chronically ill populations. We used cross-sectional data from 170 healthy young women participating in the UMass

Amherst Vitamin D Status Study to evaluate the extent to which serum 25-OHD concentrations are associated with inflammatory biomarkers in apparently healthy young women.

160

5.2 Materials and Methods

5.2.1 Study Design and Population

The UMass Amherst Vitamin D Status Study was a cross-sectional study designed to assess vitamin D status in young women and to identify factors (dietary, environmental, or lifestyle) with which it is associated. The study protocol has been described in detail previously (section 3.2.1). The majority of participants completed all study questionnaires and tests during a single clinic visit, scheduled on a morning during the late luteal phase of their menstrual cycle. Information on lifestyle and demographic factors, including smoking and drinking habits and history of oral contraceptive use, was collected using self-reported questionnaire. The present analyses include 265 women with inflammation data of which 170 women had serum 25-OHD levels available. The study was approved by the Institutional Review Board at the University of

Massachusetts, Amherst. Written informed consent was obtained from all participants.

5.2.2 Assessment of Serum 25-OHD Status

All study measurements were completed in a single clinic visit scheduled for the mid-late luteal phase of each participant's menstrual cycle. Each participant provided a fasting venous blood sample which was immediately processed and stored at −80 °C, usually within two hours of draw. Serum 25-OHD concentrations were determined using a commercially available radioimmunoassay kit (25-hydroxyvitamin D 125I RIA Kit;

DiaSorin Inc., Stillwater, MN, USA), which has been previously validated [30]. Gamma irradiation of I125 (in counts per minute; CPM) was quantified using a Beckman Gamma

4000 counter (Beckman Coulter, California, USA), and CPM were converted into

161

concentration units using GraphPad Prism version 4.0 for Windows (GraphPad Software,

San Diego, CA). Serum vitamin D status was classified as follows: 25-OHD < 50 nmol/L

(< 20 ng/mL) = deficient, 25-OHD of 50 - 75 nmol/L (20 - 50 ng/mL) = suboptimal, and

25-OHD ≥ 75 nmol/L (50 ng/mL) = optimal [31, 32].

5.2.3 Assessment of Inflammatory Biomarkers

Highly sensitive C - reactive protein (hs-CRP) was measured in serum using enzyme immunosorbent assay (EIA) kits (BioCheck, Inc., Foster City, CA, USA). Cytokine levels in serum were determined with a standard human inflammation panel using a multiplex microsphere-bead array (Assay Gate, Ijamsville, MD, USA), and quantified using a Bio-Plx 200 Bead Reader System (Bio-Rad, Hercules, CA). Inflammatory factors assayed included interleukins (IL) -1β, 2, 4, 5, 6, 7, 8, 10, 12p70, 13, tumor necrosis factor-alpha (TNF-α), granulocute macrophage colony stimulating factor (GM-CSF), and interferon-gamma (IFN-γ). Laboratory methods for the multiplexed ELISA-based approach have been described elsewhere [33, 34]. Samples were run in duplicate with internal standards. Coefficients of variation for all cytokines were < 10%.

5.2.4 Assessment of Covariates

All study measurements were completed in a single clinic visit scheduled for the late luteal phase of each participant's menstrual cycle. Dietary intake and multivitamin use was assessed using a validated food frequency questionnaire (previously described in section 3.2.3). Information on demographic factors, oral contraceptive use and lifestyle factors (e.g., smoking status and alcohol intake) was collected by self-reported

162

questionnaire. Body mass index (BMI) was calculated as weight in kilograms divided by the square of height in meters. Percentage of total body fat (%TBF) was measured by dual energy x-ray absorptiometry (DXA) using the total body scan mode on a narrow angle fan GE Lunar Prodigy scanner (GE Lunar Corp., Madison, WI). DXA scans were performed with the participant lying in a supine position with arms and legs placed flat by the side. Measurements were completed by the same technician, and the machine was calibrated daily using the standard calibration method provided by the manufacturer. The coefficient of variation for repeated scans of the manufacturer’s phantom is less than 1%.

The DXA scan report also provided estimates of sub-regional adiposity measurements

(fat mass and percentage fat mass), including the arms, legs, trunk, android and gynoid regions. Total body fat percentage was obtained directly from DXA scans (estimated using the ratio of whole-body fat mass to whole-body total mass).

To estimate physical activity levels, we asked participants to report the time they spent each week engaged in specific physical activities, including walking, jogging, running, bicycling, aerobics/dancing, tennis/racket sports, swimming, yoga/pilates and weight training. These questions were based on those used in the Nurses’ Health Study II and have been previously validated in that population [35]. We then estimated total metabolic equivalents (MET), which were recorded in hours of activity per week. The

MET per week score measures the intensity of a physical activity, with 1 MET being defined as the amount of energy expended when a person is at rest [36] .

163

5.2.5 Statistical Analysis

STATA version 12.0 (Stata Corporation., College Station, TX.) was used for all statistical analyses. Continuous variables that were not normally distributed (e.g., inflammatory factors) were log transformed to reduce skewness and improve normality.

Inflammation factors were expressed as median (inter-quartile range), and Mann-Whitney

U tests were used to compare group differences by adiposity, oral contraceptive use, smoking status and alcohol intake.

Total body fat percentage was categorized as “low” (TBF < 32%) or “high” (TBF

≥ 32%) based on gender-and-age specific cut-offs [37] and also on the current American

Council on Exercise recommendation for maximum body fat percentage in women who are non-athletes [38]. BMI was classified as “low” (BMI < 25 kg/m2) or “high” (BMI ≥

25 kg/m2) based on current recommendations by the World Health Organization [39] and

National Heart, Lung and Blood Institute and the North American Association for the

Study of Obesity [40]. The waist circumference (WC) cut-off point for high central obesity was defined as WC ≥ 80 cm, based on guidelines proposed by the International

Diabetes Federation [41]. While current anthropometric guidelines do not state a specific cut-off for waist to height ratio (WHtR), previous research indicated that among individuals less than 40 years, WHtR less 0.5 is generally considered healthy [42].

Pearson’s Product Correlation between serum 25-OHD levels and inflammatory factors were determined. The nature and strength of associations among covariates, dietary intakes, and adiposity variables were assessed using normal throery multiple linear regression. Linear and restricted cubic spline model fit were also obtained to assess departure from linearity (Appendix E). Model adequacy was assessed using

164

residual versus predicted values, and Cook’s distance plots were used to identify influential data points.

We started model building with an unadjusted simple linear regression (Model 1) using serum 25-OHD as the predictor variable and inflammatory factors as the outcome variables. Selected covariates were considered for inclusion in our final multivariable models. Season of blood draw was considered a priori to be likely confounder of a serum

25-OHD-inflammation relation. We conducted model assessment with and without each selected covariate, and included in the final regression model any variable with P < 0.20 or whose inclusion in the model led to at least a 10% change in the beta coefficient for each adiposity measure. This was followed with two different models of multivariable linear regressions. Model 2 adjusted for season of blood draw and adiposity (BMI, WC,

TBF, or WHtR), while Model 3 adjusted for all variables in Model 2 plus physical activity level, multivitamin intake, and either smoking status, alcohol intake, or oral contraceptive use. To assess effect modification (whether the relationship between hs-

CRP and 25-OHD varies with 25-OHD levels), stratified analyses were conducted to compare the above associations among individuals with low 25-OHD (< 75, nmol/L) or high 25-OHD (≥ 75 nmol/L). Our final model was evaluated for multicollinearity between covariates. All P-values were two-sided and considered statistically significant if

P < 0.05.

165

5.3 Results

Characteristics of participants are reported in Table 27. We did not find significant variations in median and interquartile distribution of inflammatory biomarkers by serum

25-OHD status (Table 28). However, women with low waist circumference (WC < 80 cm) had significantly higher levels of cytokines such as interleukin-1β (IL-1β), interleukin-7 (IL-7), granulocyte-macrophage-colony-stimulating factor (GM-CSF) and tumor necrosis factor-alpha (TNF-α), compared to women with WC > 80 cm; P < 0.01

(Table 29). Compared to women with lower adiposity (Table 30), those with high BMI (≥

25 kg/m2) had a 30.6% higher IL-6 levels (P = 0.02) while women with high total body fat, TBF, percentage (≥ 32%) exhibited a 19.2% increase in IL-6 levels (P = 0.03), Table

31.

Median levels of IL-2 and IL-4 were marginally higher among women consuming alcohol compared to non-drinkers (P = 0.07, and P = 0.05, respectively), Table 32.

Inflammatory factors were strongly correlated with each other (P < 0.01), except for correlations of hs-CRP with other inflammatory markers (Table 33), which were not significantly correlated.

Table 34 shows the Pearson product correlation between serum 25-OHD concentration and inflammatory factors. Overall, correlations of 25-OHD levels with inflammation were non-significant for all inflammatory factors (P > 0.05). Mean 25-

OHD levels were also significantly greater among women with high inflammation (hs-

CRP ≥ 5 mg/L), compared to women with low inflammation (hs-CRP < 5 mg/L), (91.6 ±

35.8 vs. 76.4 ± 31.1, P = 0.02), Table 35.

166

Figures 21-34 show the lowess regression curve and linear fit for the association between inflammatory factors and 25-OHD concentration. The gray line shows the linear prediction and the shaded band gives the confidence interval for the linear fit. The red line shows the lowess regression curve for this same association if a linear fit is not assumed. While the lowess regression curve closely approximated the least squares linear regression line over most of the observed data, there was a slight suggestion of a mild departure from the linear model. Further analysis showed that fitting a spline model did not significantly explain perceived associations any better than a linear model fit

(Appendix E).

We observed significant relationships between adiposity and several cytokines

(Table 36). For instance, unadjusted linear regressions indicated that IL-6 levels was positively associated with BMI, TBF, and WC (ß = 0.08 ± 0.02, P < 0.01; ß = 0.03 ±

0.01, P < 0.01; and ß = 0.02 ± 0.01 pg/L, P = 0.02, respectively). IL-1ß and GM-CSF were negatively associated with WC (ß = -0.02 ± 0.01, P = 0.03; and ß = -0.02 ± 0.01 pg/L, P = 0.02, respectively) while IL-10 was positively associated with BMI (ß = 0.05 ±

0.02 pg/L, P = 0.02). IL-13 was positively associated with both BMI (ß = 0.06 ± 0.03 pg/L, P = 0.03) and TBF (ß = 0.03 ± 0.01 pg/L, P = 0.02). Season of blood draw was also an important predictor of hs-CRP levels in all women. Non winter seasons (i.e. spring, summer and fall) were associated with lower hs-CRP concentraion (ß = -0.67 ±

0.30 mg/L, P = 0.03).

Crude and adjusted linear associations between serum 25-OHD and inflammatory factors are shown in Table 37. Among all women, results adjusted for season, physical activity level, adiposity, multivitamin intake, and lifestyle (alcohol, smoking, or oral

167

contraceptive use), showed no significant associations between with 25-OHD levels and log-transformed inflammatory factors. However, among women with low 25-OHD (< 75 nmol/L) (Table 31), 25-OHD levels were negatively associated with IL-2 concentrations after controlling for season, physical activity level, total body fat percentage, alcohol and multivitamin intake (ß = -0.02 ± 0.01pg/mL, P = 0.04). Thus, among women with low

25-OHD, an increase of 1 nmol/L in serum 25-OHD concentration is associated with a

2.1% lower change in IL-2 level [(i.e. e-0.021 – 1) x 100]. Similarly, among women with low 25-OHD (< 75 nmol/L), serum 25-OHD levels were negatively associated with GM-

CSF concentrations (Table 33), after adjustment for season, physical activity level, waist to height ratio, alcohol intake and multivitamin intake (ß = -0.022 ± 0.011 pg/mL, P =

0.04). Finally, multivariable adjusted regressions also showed that among women with

25-OHD < 75 nmol/L, serum 25-OHD levels were associated with both IL-6 and IL-7 (ß

= -0.020 pg/mL, P = 0.05), (Table 35).

5.4 Discussion

Vitamin D modulates inflammatory cytokines in isolate immune cells, but it is unclear whether serum 25-OHD status impacts circulating cytokine levels in young women. In this population of young healthy women (18-30 years), we found that among women with evidence of vitamin D insufficiency (25-OHD < 75 nmol/L), serum 25-hydroxyvitamin D level was negatively associated with pro-inflammatory cytokines such as interleukin-(IL-

2) and granulocute macrophage colony stimulating factor (GM-CSF) after adjustment for factors such as season, physical activity level, adiposity, multivitamin intake, and alcohol intake (P = 0.04). Low 25-OHD was also marginally associated with decreases in serum

168

concentrations of the inflammatory cytokines IL-6 and IL-7 (P = 0.05). No significant associations were observed for 25-OHD and concentrations of IL-1ß, IL-4, IL-5, IL-8,

IL-10, IL-12, IL-13, TNF-α and IFN-γ (P > 0.05).

5.4.1 Serum 25-OHD and Cytokine Levels

The role of vitamin D in modulating cytokines levels is largely supported by research indicating that supplemental vitamin D intake increases the levels of anti-inflammatory cytokines (especially IL-10), and simultaneously lowers or attenuates the increase of pro- inflammatory cytokines, notably IL-1β and TNF-α [43-46]. Pro-inflammatory cytokines function primarily to induce inflammation as a result of infection or trauma. They typically initiate the cascade of inflammatory mediators by targeting endothelial cells

[47]. Anti-inflammatory cytokines on the other hand block the inflammatory cascade initiated by pro-inflammatory cytokines [47, 48]. They work together with specific cytokine inhibitors and soluble cytokine receptors to regulate the human immune response [47, 48].

Although the exact mechanism(s) explaining the role of vitamin D in immune modulation of humans is not fully known, in- vivo studies indicate that 1,25 dihydroxyvitamin D (1,25(OH)2D), which is the biologically active form of vitamin D in blood, could influence gene transcription of specific cytokines [24-26]. 1,25(OH)2D is thought to exert its anti-inflammatory effects by binding to the vitamin D receptor

(VDR), which is present in most immune cells [46, 49]. This ligand/receptor complex then binds to vitamin D response element (VDRE) in the promoter region of target tissues

169

to and alters the rate of gene transcription of certain cytokines (e.g., IL-10 and IL-8) and cytokine receptors (e.g., IL1-RL1) that are involved in innate immunity [46, 49, 50, 51].

Studies examining the association between vitamin D and inflammation have produced conflicting reports in overweight and obese adults, post-menopausal women, and in young apparently healthy young men and women [52-57]. One reason for this inconsistency in findings may be attributed to age-related differences. Differences in ages of subjects are important because the expression of the VDR decreases with age which can in turn reduce the capacity of vitamin D to mediate its biological activities [58].

Barker et al. conducted a prospective study over two successive winter seasons among 28 non-smoking, physically active men and women (25 - 42 years) to evaluate whether circulating cytokines varies with alterations in serum 25-hydroxyvitamin D (25-

OHD) concentrations [59]. Participants provided fasting blood samples and a multiplex microsphere-bead array was used to measure circulating inflammatory cytokines. Vitamin D insufficiency was set at 25-OHD ≤ 80 nmol/L and sufficiency at

25-OHD > 80 nmol/L. They found that levels of pro-inflammatory cytokines (interleukin

(IL)-2, IL-1ß, TNF-α and IFN- γ) were significantly greater in subjects with 25-OHD ≤

80 nmol/L. Conversely, concentrations of the anti-inflammatory cytokine (i.e., IL-10) did not significantly differ by 25-OHD status. Our results are similar but also contrast the reports of Barker et al. We observed that among women with 25-OHD < 75 nmol/L, 25-

OHD was inversely associated with and macrophage colony stimulating factor (GM-

CSF) concentrations, but concentrations of IL-1ß, TNF-α IFN- γ, and Il-10 remained unchanged with vitamin D status after adjustment for season, physical activity level, adiposity, alcohol and multivitamin intake .

170

In the present study, serum 25-OHD levels were negatively associated with GM-

CSF concentrations among women with 25-OHD < 75 nmol/L, after adjustment for season, physical activity level, waist to height ratio, alcohol intake and multivitamin intake (ß = -0.022 ± 0.011 pg/mL, P = 0.04). Granulocyte-macrophage colony- stimulating factor (GM-CSF) is an important immune modulator with profound effects on the functional activities of various circulating cytokines. Inflammatory cytokines, such as

IL-1 and TNF-α serve as potent inducers of GM-CSF, but GM-CSF expression can be inhibited by IL-4, IL-10, and IFN- γ [60-63]. Multivariable adjusted regressions also showed that serum 25-OHD levels were marginally associated with the inflammatory markers IL-6 and IL-7 (ß = -0.020 ± 0.010 pg/mL, P = 0.05) in women with vitamin D insufficiency (25-OHD < 75 nmol/L). IL-6 is often used as a marker for systemic activation of pro-inflammatory cytokines [64] and its production is induced by lipopolysaccharide stimulation, along with presence of TNF-α and IL-1 [48]. Although

IL-6 is generally considered to be a pro-inflammatory cytokine, it also possesses anti- inflammatory properties (pleiotrophic). IL-6 is a potent inducer of acute-phase protein response [64]. It also inhibits the production of pro-inflammatory cytokines such as GM-

CSF and IFN-γ [65].

In our study, inflammatory factors were significantly correlated with one-another

(P < 0.01), except for correlations of hs-CRP with other inflammatory markers. Levels of

IL-1 and TNF-α were positively correlated with GM-CSF (r = 0.93 and r = 0.71, P < 0.01 respectively) as were correlations between GM-CSF and IFN- γ, IL-4, and IL-10 (r =

0.71, r = 0.53, r = 0.20; P < 0.01 respectively). These strong correlations between cytokines make it impossible for us to tease out the effects of indiviual cytokines.

171

Further more, because our analyses made several comparisons between 25-OHD status and inflammatory factors, we are cautious in making any definite conclusions on associations between 25-OHD levels and inflammatory factors. Simply stated, our findings merely suggest that sub-optimal vitamin D levels may impact inflammatory response in young women. Prospective studies with larger cohorts are needed to determine if there is a causal relationship between low 25-OHD and inflammatory factors in young women.

5.4.2 Serum 25-OHD and C-Reactive Protein

C-reactive protein (CRP) is an acute-phase reactant protein synthesized by the liver in response to a variety of inflammatory cytokines [66]. Our results contrast with previous research that found positive associations between 25-OHD and hs-CRP concentraions in young women [67, 68]. Cross-sectional analyses of nationally representative data

(National Health and Nutrition Examination Surveys, NHANES, 2001 to 2006) showed that after controlling for the effects of age, gender, and race, there was an inverse association between 25-OHD and CRP concentrations (ß = -0.30, P < 0.001) in asymptomatic adults with low vitamin D levels (below the population median, i.e. 25-

OHD < 52.5 nmol/L), [67]. However, this association became non-significant once the serum 25-OHD increased above the population median (≥ 52.5 nmol/L), (ß = -0.05, P =

0.08). Our results contrast that of Amer and Qayyum, in that hs-CRP levels were not associated with 25-OHD concentration, even among women with 25-OHD < 75 nmol/L.

Further adjustment for season BMI, physical activity level, alcohol intake, and multivitamin use did not reveal any significant changes in linear relations. We advise that

172

any comparison of our findings to other studies be done with caution, especially since our sample size was relatively small and statistical power may have limited.

Garcia-Bailo et al. carried out a cross-sectional study among adult men (n = 428) and women (n = 975) aged 20-29 years that showed positive associations between plasma concentrations of 25-OHD and hs-CRP levels in women (ß = 0.011 ± 0.002, P < 0.0001) after controlling for factors such as age, gender, waist circumference, physical activity, ethnicity and season in the overall population [68]. Garcia-Bailo et al. also reported that women using hormonal contraceptives (HC) had significantly higher circulating 25-OHD

(78.1 ± 33.9 vs. 51.1 ± 24.2 nmol/L, P < 0.0001) and hs-CRP (2.7 ± 3.3 vs. 0.8 ± 1.9 mg/L, P < 0.0001) than non-HC users. We found no associations between 25-OHD and hs-CRP levels (ß = 0.011 ± 0.002 mg/L, P < 0.0001), and 25-OHD status and hs-CRP concentrations did not also differ between oral contraceptive (OC) users and non-users

(83.0 ± 34.0 vs. 76.9 ± 31.1 nmol/L, P = 0.23; and 0.9 ± 2.8 vs. 0.8 ± 3.4 mg/L, P = 0.91, respectively). While the populations of both studies remain fairly similar with regard to mean age and BMI, our results contrast that of Garcia-Bailo et al., possibly because of differences in ethnicity. About 49.6% of women in Garcia-Bailo et al’s study were

Caucasian compared to 84.5% in our study. Future studies should consider examining associations between ethnicity and lifestyle factors (e.g., OC use) in young women as this may be important confounders of the association between 25-OHD and hs-CRP levels.

5.4.3 Adiposity and Inflammation

In this population of young women, uadjusted linear regressions showed significant relationships between adiposity and several cytokines (P < 0.05). To our knowledge, this

173

is the first study to examine associations of obesity markers (body mass index, BMI, total body fat percentage, TBF, waist cirfumference, WC, and waist to heigh ratio, WHtR) with inflammatory factors in young women. Marques-Vidal et al., evaluated associations between obesity markers (BMI, WC and TBF) and inflammatory markers IL-1β, IL-6,

TNF-α and hs-CRP in Caucasian adult ages 35-75 years [69]. After adjusting for age, physical activity, and smoking, they observed that obese women (BMI ≥ 30 kg/m2) had higher hs-CRP levels than overweight or normal women (mean hs-CRP: 2.24 ± 1.05 vs.

1.47 ± 1.03 vs. 0.93 ± 1.03 mg/L, respectively; P < 0.001). In addition, BMI was negatively associated with IL-1β levels (ß = -0.005 ± 0.007 pg/mL, P < 0.05), and women with high TBF (≥ 32) or WC (≥ 80 cm) had higher levels of IL-6 (TBF: ß = 0.010

± 0.003 pg/mL, P < 0.01; WC: ß = 0.011 ± 0.002 pg/mL, P < 0.001). Finally, positive associations also existed between TNF-α levels and BMI, TBF, and WC (BMI: ß = 0.019

± 0.003 pg/mL, P < 0.001; TBF: ß = 0.006 ± 0.002 pg/mL, P < 0.01; WC: ß = 0.007 ±

0.001 pg/mL, P < 0.001).

BMI and TBF are common indicators of obesity, but WC and WHtR are often discounted as useful indicators of obesity. Among young adults, 18-30 years old, a waist circumference greater than 80 cm indicates central obesity (i.e., an excess of abdominal fat) and suggests increased visceral adiposity. Visceral fat is more metabolically active than subcutaneous fat [70, 71], and large deposits can disrupt fatty acid metabolism, increasing the influx of free fatty acids entering splanchnic circulation [72, 73]. Excessive amounts of unbound fatty acids in the blood contribute to hyperlipidemia and hypertriglyceridemia, both of which can promote or exacerbate obesity-related chronic conditions [74-76].

174

WHtR is a measure of the distribution of body fat [77, 78] and is especially useful in the assessment of health status for individuals with a higher percentage of muscle and a lower percentage of body fat (e.g., athletes). Among individuals less than 40 years,

WHtR ≤ 0.5 is generally considered healthy [42]. In addition to BMI, future studies on obesity and inflammation should consider including WC and WHtR as proxies for obesity since both measures are easy to obtain and less intrusive/expensive than measuring dual energy x-ray absorptiometry.

Some of the strengths of our study include assessment of several inflammatory biomarkers e.g., highly sensitive C-reactive protein (hs-CRP), interleukin (IL)-1β, 2, 4, 5,

6, 7, 8, 10, 12p70, 13, granulocyte-macrophage-colony-stimulating factor (GM-CSF), interferon-gamma (IFN-γ), and tumor necrosis factor-alpha (TNF-α) using multiplex microsphere-based assays. Radioimmuno assays were used assess serum levels of 25-

OHD, which is considered the best indicator of vitamin D status because it represents a summation of total cutaneous production of vitamin D and intake of vitamin D from foods [79-81]. Our population was comprised of mostly non-obese and few obese individuals; and we determined adiposity levels using DXA scans, which have been previously validated and found to accurately differentiate between women with acceptable/normal body fat and those with high or excess adiposity [82, 83]. An important limitation of our study is its cross-sectional design, which prevents us from speculating on cause-effect relationships between 25-OHD status, obesity markers, and inflammatory factors. Inflammation factors and 25-OHD status were assessed using a single serum sample which may not perfectly capture long-term changes of biomarkers, although previous research indicates that inflammation levels remain stable over time

175

[91]. We controlled for several factors including season of blood draw, physical activity level, adiposity, multivitamin use, and life style factors ( i.e. alcohol intake, smoking status and oral contraceptive use) as these parameters could confound associations of 25-

OHD with inflammation [84-90]. However, it is still possible that there may be residual confounding by other unknown factors. Any comparison or interpretation of results should also be done with caution. Prospective studies are needed to better understand the role of 25-OHD and adiposity in inflammation. Our population was also fairly homogenous; with mostly non-obese and few obese individuals. All participants were also between 18 to 30 years, and the majorities were Caucasian. Our findings are therefore limited to young, non-obese, Caucasian women and may not be generalizable to women in other BMI categories, age groups or ethnicities.

5.5 Conclusions and Significance

Vitamin D insufficiency and increased adiposity have been linked with higher levels of inflammatory biomarkers in obese/older adults, and chronically ill patients. However, few studies have examined these associations in healthy young women. In the present study, we observed that IL-2, IL-6, IL-7, and GM-CSF concentrations were inversely associated with serum 25-OHD levels among women with vitamin D insufficiency (25-

OHD < 75 nmol/L). Significant positive associations were also observed between obesity markers and inflammatory factors in young women.

Increased adiposity, high inflammation, and central obesity all tend to increase an individual’s risk for cardiovascular disease (CVD), whether or not the individual exhibits traditional CVD risk factors such as elevated lipid profile (triglyceride, cholesterol, and

176

low-density lipoprotein levels), smoking history, and presence of hypertension or diabetes. Prospective studies with larger cohorts are needed to establish causality, and to establish whether findings from our third study could be useful in informing routine screening in clinic settings for the prevention, early detection, and management of inflammation, an important cardiovascular disease (CVD) risk factors in young women.

177

5.6 References

1. Hotamisligil GS, Shargill NS, Spiegelman BM. Adipose expression of tumor necrosis

factor-alpha: direct role in obesity-linked insulin resistance. Science 1993; 259: 87-91.

2. Vozarova B, Weyer C, Hanson K, Tataranni PA, Bogardus C, Pratley RE. Circulating

interleukin-6 in relation to adiposity, insulin action, and insulin secretion. Obes Res 2001;

9 :414-7.

3. Xu H, Barnes TG, Yang Q, Tan G, Yang D, Chou CJ, Sole J, Nichols A, Ross JS,

Tartaglia LA, Chen H. Chronic inflammation in fat plays a crucial role in the

development of obesity-related insulin resistance. J Clin Invest 2003; 112: 1821-30.

4. Moriseset AS, Huot C, Legare D. Tcernof A. Circulating IL-6 concentrations and

abdominal adipocyte is isoproterenol-stimulated lipolysis in women. Obesity 2008; 16:

1487-92.

5. Mohamed-Ali V, Goodrick S, Rawesh A, Katz DR, Miles JM, Yudkin JS, Klein

S, Coppack SW. Subcutaneous adipose tissue releases interleukin-6, but not tumor

necrosis factor-alpha, in vivo. J Clin Endocrinol Metab 1997; 82: 4196–200.

6. Fried SK, Bunkin DA, Greenberg AS. Omental and subcutaneous adipose tissues of

obese subjects release interleukin-6: depot difference and regulation by glucocorticoid. J

Clin Endocrinol Metab 1998; 83: 847-50.

7. Yudkin JS, Kumari M, Humphries SE, Mohamed-Ali V. Inflammation, obesity, stress

and coronary heart disease: is interleukin-6 the link? Atherosclerosis 2000; 148: 209-14.

8. Galic S, Oakhill JS, Steinberg GR. Adipose tissue as an endocrine organ. Mol Cell

Endocrinol 2010; 316:129-39.

178

9. Bastard JP, Maachi M, Lagathu C, Kim MJ, Caron M, Vidal H, Capeau J, Feve B.

Recent advances in the relationship between obesity, inflammation, and insulin

resistance. Eur Cytokine Netw 2006: 17: 4-12.

10. Varghese B, Swamy S, Srilakshmi MA, Santhosh MJ, Shetty GG, Varghese K, Patil CB,

Iyengar SS. Visceral adiposity in young patients with coronary artery disease-a case

control study. Indian Heart J 2012; 64: 284-9.

11. Park HS, Park JY, Yu R. Relationship of obesity and visceral adiposity with serum

concentrations of CRP, TNF-α and IL-6. Diabetes Res Clin Pract 2005; 69: 29-35.

12. Kim K, Valentine RJ, Shin Y, Gong K. Associations of visceral adiposity and exercise

participation with C-reactive protein, insulin resistance, and endothelia dysfunction in

Korean healthy adults. Metabolism 2008; 57: 1181-9.

13. Bellia A, Garcovich C, D'Adamo M, Lombardo M, Tesauro M, Donadel G, Gentileschi

P, Lauro D, Federici M, Lauro R, Sbraccia P. Serum 25-

hydroxyvitaminD levels are inversely associated with systemic inflammation in severe

obese subjects. Intern Emerg Med Published online: 25 March 2011.

14. Beydoun MA, Boueiz A, Shroff MR, Beydoun HA, Wang Y, Zonderman AB.

Associations among 25-Hydroxyvitamin D, Diet Quality, and Metabolic Disturbance

Differ by Adiposity in Adults in the United States. J Clin Endocrinol Metab 2010; 95:

3814-27.

15. Pradhan AD, Manson JE, Rifai N, Buring JE, Ridker PM. C-reactive protein, interleukin

6, and risk of developing type 2 diabetes mellitus. JAMA 2001; 286: 327-34.

16. Singh U, Devaraj S, Jialal I. Vitamin E, oxidative stress, and inflammation. Annu Rev

Nutr 2005; 25: 151-74.

179

17. Dehghan A, Kardys I, de Maat MP, Uitterlinden AG, Sijbrands EJ, Bootsma AH, Stijnen

T, Hofman A, Schram MT, Witteman JC. Genetic variation, C-reactive protein levels,

and incidence of diabetes. Diabetes 2007; 56: 872-8.

18. Lopez-Garcia E, Schulze MB, Meigs JB, Manson JE, Rifai N, Stampfer MJ, Willett WC,

Hu FB. Consumption of trans fatty acids is related to plasma biomarkers of inflammation

and endothelial dysfunction. J Nutr 2005; 135: 562-6.

19. Pickup J, Mattock M. Activation of the innate immune system as a predictor of

cardiovascular mortality in Type 2 diabetes mellitus. Diabet Med 2003; 20: 723-6.

20. Ridker PM, Rifai N, Stampfer MJ, Hennekens CH. Plasma concentration of interleukin-6

and the risk of future myocardial infarction among apparently healthy men. Circulation

2000; 101: 1767-72.

21. Ridker PM. High-sensitivity C-reactive protein: potential adjunct for global risk

assessment in the primary prevention of cardiovascular disease. Circulation 2001; 103:

1813-8.

22. Rost NS, Wolf PA, Kase CS, Kelly-Hayes M, Silbershatz H, Massaro JM, D'Agostino

RB, Franzblau C, Wilson PW. Plasma concentration of C-reactive protein and risk of

ischemic stroke and transient ischemic attack: the Framingham study. Stroke 2001; 32:

2575-9.

23. Freeman DJ, Norrie J, Caslake MJ, Gaw A, Ford I, Lowe GD O'Reilly DS, Packard CJ,

Sattar N. C-reactive protein is an independent predictor of risk for the development of

diabetes in the West of Scotland Coronary Prevention Study. Diabetes 2002; 51: 159-

600.

180

24. Canning MO, Grotenhuis K, de Wit H, Ruwhof C, Drexhage HA. 1-alpha, 25-

Dihydroxyvitamin D3 (1,25(OH)2D3) hampers the maturation of fully active immature

dendritic cells from monocytes. Eur J Endocrinol 2001; 145: 351-7.

25. Zhu Y, Mahon BD, Froicu M, Cantorna MT. Calcium and 1alpha, 25-dihydroxyvitamin

D3 target the TNF-α pathway to suppress experimental inflammatory bowel disease. Eur

J Immunol 2005; 35: 217-24.

26. Shab-Bidar S, Neyestani TR, Djazayery A, Eshraghian MR, Houshiarrad A, Kalayi A,

Shariatzadeh N, Khalaji N, Gharavi A. Improvement of vitamin D status resulted in

amelioration of biomarkers of systemic inflammation in the subjects with type 2 diabetes.

Diabetes Metab Res Rev 2012; 28: 424-30.

27. Willet WC, Sampson L, Stampler MJ, Rosner B, Bain C, Witschi J, Hennekens CH,

Speizer FE. Reproducibility and validity of a semi-quantitative food frequency

questionnaire. Am J Epidemiol 1985; 122, 51-65.

28. Feskanich D, Rimm EB, Giovannucci EL, Colditz GA, Stampfer MJ. Reproducibility and

validity of food intake measurements from a semi-quantitative food frequency

questionnaire. J Am Diet Assoc 1993; 93, 790-6.

29. Food and Nutrition Board (FNB) / Institute of Medicine (IOM). Dietary Reference

Intakes for Calcium and Vitamin D (2010). 2101 Constitution Avenue,

N.W. Washington, DC 20418. National Academy Press.

30. Hollis, B. Comparison of commercially available (125) I-based RIA method for the

determination of circulating 25-hydroxyvitamin D. Clin Chem 2000; 46: 1657-61.

31. Holick MF, Binkley NC, Bischoff-Ferrari HA, Gordon CM, Hanley DA, Heaney RP,

Murad MH, Weaver CM: Evaluation, treatment, and prevention of vitamin D deficiency:

181

an Endocrine Society clinical practice guideline. J Clin Endocrinol Metab 2011; 96:1911-

30.

32. Norman AW, Bouillon R, Whiting SJ, Vieth R, Lips P. 13th Workshop consensus for the

vitamin D nutritional guidelines. J Steroid Biochem Mol Bio 2007; 103:204-5.

33. Carson RT, Vignali DA. Simultaneous quantitation of 15 cytokines using a multiplexed

flow cytometric assay. J Immunol Methods 1999; 227: 41-52.

34. Paolo JA, Lee LS, Wu B, Banks PA, Steen H, Conwell DL. Cytokine profiling of

pancreatic fluid using the ePFT collection method in tandem with a multiplexed

microarray assay. J Immunol Methods 2011; 369: 98-107.

35. Wolf AM, Hunter DJ, Colditz GA, Manson JE, Stamoer MJ, Corsano KA., Rosner B,

Kriska A, Willet WC. Reproducibility and validity of a self-administered physical

activity questionnaire. Int J Epidemiol 1994; 23, 991-9.

36. Ainsworth BE, Haskell WL, Leon AS, Jacobs DR jr, Montoye HJ, Sallis JF, Paffenbarger

RS, jr. Compendium of physical activities: classification of energy costs of human

physical activities. Med Sci Sports Exerc 1993; 25: 1575-81.

37. Jackson AS, Pollack ML. Practical assessment of body composition. Physician and

Sportsmed 1985; 13, 76-90.

38. American Council on Exercise. What are the guidelines for percentage of body fat loss,

2009? Available online: http://www.acefitness.org/blog/112/what-are-the guidelines-for-

percentage-of-body-fat.

39. World Health Organization. Preventing and Managing the Global Epidemic. WHO,

Geneva, 2000. WHO Obesity Technical Report Series, Obesity; Number 894.

182

40. National Heart, Lung, and Blood Institute (NHLBI) and the North American Association

for the Study of Obesity (NAASO). The Practical Guide: Identification, Evaluation, and

Treatment of Overweight and Obesity in Adults (2000). NIH Publication Number 00-

4084. Available online: http://www.nhlbi.nih.gov/guidelines/obesity/prctgd_c.pdf.

41. International Diabetes Federation.The IDF Consensus Worldwide Definition of the

Metabolic Syndrome (2006), Avenue Emile De Mot 19, B-1000 Brussels, Belgium.

Available online: http//www.idf.org/webdata/docs/IDF_Meta_def_final.pdf.

42. Browning LM, Hsieh SD, Ashwell M. A systematic review of waist-to-height ratio as a

screening tool for the prediction of cardiovascular disease and diabetes: 0·5 could be a

suitable global boundary value. Nutr Res Rev 2010; 23: 247-69.

43. Schleithoff SS, Zittermann A, Tenderich G, Berthold HK, Stehle P, Koerfer R. Vitamin

D supplementation improves cytokine profiles in patients with congestive heart failure: a

double-blind, randomized, placebo-controlled trial. Am J Clin Nutr 2006; 83: 754-9.

44. Zittermann A, Frisch S, Berthold HK, Gotting C, Kuhn J, Kleesiek K, Stehle P, Koertke

H, Koerfer R. Vitamin D supplementation enhances the beneficial effects of weight loss

on cardiovascular disease risk markers. Am J Clin Nutr 2009; 89: 1321-7.

45. Stubbs JR, Idiculla A, Slusser J, Menard R, Quarles LD. Cholecalciferol supplementation

alters calcitriol-responsive monocyte proteins and decreases inflammatory cytokines in

ESRD. J Am Soc Nephrol 2010; 21: 353-61.

46. Wang Y, DeLuca HF. Is the vitamin d receptor found in muscle? Endocrinology 2011;

152:354-63.

47. O'Dell JR. Anticytokine therapy-a new era in the treatment of rheumatoid arthritis? N

Engl J Med 1999; 340: 310-12.

183

48. Opal SM, DePalo VA. Anti-inflammatory cytokines. Chest 2000; 117: 1162-72.

49. Opal, SM, Wherry, JC, Grint, P Interleukin-10: potential benefits and possible risks in

clinical infectious diseases. Clin Infect Dis 1998; 27: 1497-507.

50. Bosse Y, Lemire M, Poon AH, Daley D, He JQ, Sandford A, White JH, James AL, Musk

AW, Palmer LJ, Raby BA, Weiss ST, Kozyrskyj AL, Becker A, Hudson TJ, Laprise C.

Asthma and genes encoding components of the vitamin D pathway. Respir Res 2009; 10:

98.

51. Ramagopalan SV, Heger A, Berlanga AJ, Maugeri NJ, Lincoln MR, Burrell

A, Handunnetthi L, Handel AE, Disanto G, Orton SM, Watson CT, Morahan

JM,Giovannoni G, Ponting CP, Ebers GC, Knight JC. A ChIP-seq defined genome-wide

map of vitamin D receptor binding: associations with disease and evolution. Genome

Res 2010; 20:1352-60.

52. Carrillo AE, Flynn MG, Pinkston C, Markofski MM, Jiang Y, Donkin SS, Teegarden D.

Vitamin D supplementation during exercise training does not alter inflammatory

biomarkers in overweight and obese subjects. Eur J Appl Physiol 2012; 112: 3045-52.

53. Jorde R, Sneve M, Torjesen PA, Figenschau Y, Goransson LG, Omdal R. No effect of

supplementation with cholecalciferol on cytokines and markers of inflammation in

overweight and obese subjects. Cytokine 2010; 50:175-80.

54. Gannage-Yared MH, Azoury M, Mansour I, Baddoura R, Halaby G, Naaman R. Effects

of a short-term calcium and vitamin D treatment on serum cytokines, bone markers,

insulin and lipid concentrations in healthy post-menopausal women. J Endocrinol Invest

2003; 26:748-53.

184

55. Khoo AL, Chai LY, Koenen HJ, Sweep FC, Joosten I, Netea MG, van der Ven AJ.

Regulation of cytokine responses by seasonality of vitamin D status in healthy

individuals. Clin Exp Immunol 2011; 164:72-9.

56. Barnes MS, Horigan G, Cashman KD, Hill TR, Forsythe LK, Lucey AJ, McSorley EM,

Kiely M, Bonham MP, Magee PJ, Strain JJ, Wallace JM. Maintenance of wintertime

vitamin D status with cholecalciferol supplementation is not associated with alterations in

serum cytokine concentrations among apparently healthy younger or older adults. J Nutr

2011; 141: 476-81.

57. Barker T, Martins TB, Hill HR, Kjeldsberg CR, Henriksen VT, Dixon BM, Schneider

ED, Dern A, Weaver LK. Different doses of supplemental vitamin D maintain

interleukin-5 without altering skeletal muscle strength: a randomized, double-blind,

placebo-controlled study in vitamin D sufficient adults. Nutr Metab (Lond) 2012; 9:16.

58. Bischoff-Ferrari HA, Borchers M, Gudat F, Durmuller U, Stahelin HB, Dick W. Vitamin

D receptor expression in human muscle tissue decreases with age. J Bone Miner Res

2004; 19:265-9.

59. Barker T, Martins TB, Hill HR, Kjeldsberg CR, Dixon BM, Schneider ED, Henriksen

VT, Weaver LK. Circulating pro-inflammatory cytokines are elevated and peak power

output correlates with 25-hydroxyvitamin D in vitamin D insufficient adults. Eur J Appl

Physiol 2013; 113: 1523-34.

60. Sagawa K, Mochizuki M, Sugita S, Nagai K, Sudo T, Itoh K. Suppression by IL-10 and

IL-4 of cytokine production induced by two-way autologous mixed lymphocyte

reaction. Cytokine 1996; 8:501–6.

185

61. Ozawa H, Aiba S, Nakagawa, Tagami H. Interferon-gamma and interleukin-10 inhibit

antigen presentation by Langerhans cells for T helper type 1 cells by suppressing their

CD80 (B7-1) expression. Eur J Immunol 1996; 26:648-52.

62. Jansen JH, Wientjens GJ, Fibbe WE, Willemze R, Kluin-Nelemans HC. Inhibition of

human macrophage colony formation by interleukin 4. J Exp Med 1989; 170:577-82.

63. Akashi K, Shibuya T, Harada M, Takamatsu Y, Uike N, Eto T, Niho Y. Interleukin 4

suppresses the spontaneous growth of chronic myelomonocytic leukemia cells. J Clin

Invest 1991; 88:223-30.

64. Barton, BE, Shortall, J, Jackson, JV Interleukins 6 and 11 protect mice from mortality in

a staphylococcal enterotoxin-induced toxic shock model. Infect Immun 1996; 64: 714-

718.

65. Barton, BE IL-6: insights into novel biological activities. Clin Immunol Immunopathol

1997; 85: 16-20.

66. Pepys MB, Hirschfield GM. C-reactive protein: a critical update. J Clin Invest 2003; 111:

1805-12.

67. Amer M, Qayyum R. Relation between serum 25-hydroxyvitamin D and C-reactive

protein in asymptomatic adults (from the continuous National Health and Nutrition

Examination Survey 2001 to 2006). Am J Cardiol 2012; 109: 226-30.

68. García-Bailo B, Josse AR, Jamnik J, Badawi A, El-Sohemy A. Positive Association

Between 25-Hydroxyvitamin D and C-Reactive Protein is Confounded by Hormonal

Contraceptive Use. J Womens Health 2013; 22:417-25.

69. Marques-Vidal P, Bochud M, Bastardot F, Lüscher T, Ferrero F, Gaspoz JM, Paccaud

F, Urwyler A, von Känel R, Hock C, Waeber G, Preisig M, Vollenweider P. Association

186

between inflammatory and obesity markers in a Swiss population-based sample (CoLaus

Study). Obes Facts 2012; 5: 734-44.

70. Mauriege P, Galitzky J, Berlan M, Lafontan M. Heterogeneous distribution of beta and

alpha-2 adrenoceptor binding sites in human fat cells from various fat deposits: functional

consequences. Eur J Clin Invest 1987; 17: 156-65.

71. Rebuffe-Scrive M, Andersson B, Olbe L, Bjorntorp P. Metabolism of adipose tissue in

intra-abdominal depots of non-obese men and women. Metabolism 1989; 38: 453-8.

72. Bergman BC, Cornier MA, Horton TJ, Bessesen DH. Effects of fasting on insulin action

and glucose kinetics in lean and obese men and women. Am J Physiol Endocrinol Metab

2007; 293: 1103-11E.

73. Nielsen S, Guo Z, Johnson CM, Hensrud DD, Jensen MD. Splanchnic lipolysis in human

obesity. J Clin Invest 2004; 113: 1582-8.

74. Zhu S, Wang Z, Heshka S, Heo M, Faith MS, Heymsfield SB. Waist circumference and

obesity-associated risk factors among whites in the third National Health and Nutrition

Examination Survey: clinical action thresholds. Am J Clin Nutr 2002; 76: 743-9.

75. Clark VR, Greenberg B, Harris TS, Carson BL. Body mass index and waist

circumference predictors of cardiovascular risk in African Americans. Ethn Dis 2012; 22:

162-7.

76. Shields M, Tremblay MS, Connor Gorber S, Janssen I. Abdominal obesity

and cardiovascular disease risk factors within body mass index categories. Health Rep

2012; 23: 7-15.

187

77. Hsieh SD, Muto T. The superiority of waist-to-height ratio as an anthropometric index to

evaluate clustering of coronary risk factors among non-obese men and women. Prev Med

2005; 40: 216-20.

78. Hsieh SD, Yoshinaga H, Muto T. Waist-to-height ratio, a simple and practical index for

assessing central fat distribution and metabolic risk in Japanese men and women. Int J

Obes Relat Metab Disord 2003; 27: 610-6.

79. Brannon PM, Yetley EA, Bailey RL, Picciano MF. Overview of the conference “Vitamin

D and health in the 21st century: an update”. Am J Clin Nutr 2008; 88:483-90S.

80. Holick MF. The use and interpretation of assays for vitamin D and its metabolites. J

Nutr 1990; 120: 1464-9.

81. Chen TC, Turner AK, Holick MF. Methods for the determination of the circulating

concentration of 25-hydroxyvitamin D. J Nutr Biochem 1990; 1: 315-9.

82. Haarbo J, Gotfredsen A, Hassager C, Christiansen C. Validation of body composition by

dual energy X-ray absorptiometry (DEXA). Clin Physiol 1991; 11: 331-41.

83. Bolanowski M, Nilsson BE. Assessment of human body composition using dual-energy

x-ray absorptiometry and bioelectrical impedance analysis. Med Sci Monit 2001; 7: 1029-

84. Need AG, Morris HA, Horowitz M, Nordin C. Effects of skin thickness, age, body fat,

and sunlight on serum 25-hydroxyvitamin D. Am J Clin Nutr 1993; 58: 882-5.

85. Millen AE, Bodnar LM. Vitamin D assessment in population-based studies: a review of

the issues. Am J Clin Nutr 2008; 87: 1102S-5S.

86. Tangpricha V, Pearce EN, Chen TC, Holick MF. Vitamin D Insufficiency among Free-

Living Healthy Young Adults. Am J Med 2002; 112: 659-62.

188

87. Ilich JZ, Brownbill RA, Tamborini L, Crncevic-Orlic Z: To drink or not to drink: how are

alcohol, caffeine and past smoking related to bone mineral density in elderly women? J

Am Coll Nutr 2002; 21: 536-44.

88. Taylor RW, Keil D, Gold EJ, Williams SM, Goulding A. Body mass index, waist girth,

and waist-to-hip ratio as indexes of total and regional adiposity in women, evaluation

using receiver operating characteristic curve. Am J Clin Nutr 1988; 67: 44-9.

89. Kimm SYS, Glynn NW, Aston CE, Poehlman ET, Daniels SR. Effects of Race, Cigarette

Smoking, and Use of Contraceptive Medications on Resting Energy Expenditure in

Young Women. Am J Epidemiol 2001; 154: 718-24.

90. Dorn JM, Hovey K, Muti P, Freudenheim JL, Russell M, Nochajski TH, Trevisan M.

Alcohol Drinking Patterns Differentially Affect Central Adiposity as Measured by

Abdominal Height in Women and Men. J Nutr 2003; 133: 2655-62.

91. Pischon T, Hankinson SE, Hotamisligil GS, Rifai N, Willett WC, Rimm EB. Habitual

dietary intake of n-3 and n-6 fatty acids in relation to inflammatory markers among US

men and women. Circulation 2003; 108: 155-60.

189

Table 27. Median and Inter-Quartile Range Distribution of Inflammatory Biomarkers

Inflammatory Factor (pg/mL) Median (Inter-Quartile Range) Min - Max IL-1ß 1.48 (2.14) 0.13 - 948.78 IL-2 5.51 ( 6.33) 0.09 - 2581.44 IL-4 0.18 (0.24) 0.13 - 41.29 IL-5 2.26 (3.03) 0.16 - 324.25 IL-6 2.00 (3.69) 0.14 - 50.92 IL-7 1.72 (1.04) 0.14 - 91.73 IL-8 10.52 (14.71) 0.50 - 1735.52 IL-10 3.57 (7.47) 0.15 - 7093.24 IL-12p70 2.28 (3.18) 0.15 - 227.51 IL-13 7.08 (22.19) 0.12 - 238.37 GM-CSF 4.66 (5.65) 0.51 - 64.26 TNF-α 2.36 (3.84) 0.12 - 620.00 IFN-γ 0.80 (3.30) 0.10 - 42.80 hs-CRP (mg/L) ѣ 0.71 (2.30) 0.13 - 48.18 ѣ n = 171 for hs-CRP. Definition: Interleukin (IL), granulocyte-macrophage-colony-stimulating factor (GM-CSF), tumor necrosis factor-alpha (TNF-α), interferon-gamma (IFN-γ), and C-reactive protein (hs-CRP).

190

Table 28. Variations in Median and Inter-Quartile Range Distribution of Inflammatory Biomarkers by Serum 25-OHD Status

Inflammatory Factor Low 25-OHD < 75 nmol/L§ High 25-OHD ≥ 75 nmol/L§ P Ϫ (pg/mL) (n = 92) (n = 78) Median (Inter-Quartile Range) IL-1ß 0.24 (0.80) 0.20 (0.67) 0.68 IL-2 1.06 (1.58) 1.09 (1.71) 0.93 IL-4 4.45 ( 5.37) 4.15 ( 5.55) 0.32 IL-5 0.16 (0.21) 0.20 (0.22) 0.48 IL-6 2.01 (2.90) 1.79 (3.64) 0.36 IL-7 1.14 (2.27) 1.22 (2.73) 0.89 IL-8 1.56 (0.79) 1.57 (0.94) 0.58 IL-10 10.28 (16.32) 12.03 (21.46) 0.30 IL-12p70 3.01 (8.87) 2.91 (9.58) 0.88 IL-13 1.88 (3.52) 1.73 (3.84) 0.77 GM-CSF 2.93 (8.87) 2.33 (8.10) 0.66 TNF-α 3.09 (2.90) 3.05 (3.20) 0.96 IFN-γ 1.57 (3.03) 1.37 (3.10) 0.77 hs-CRP (mg/L) ѣ 0.80 (3.20) 0.80 (5.00) 0.76 § 2.496 nmol/L = 1 ng/mL. Ϫ P-values compare groups using Mann-Whitney tests. Ѣ n = 168 for hs-CRP. Definition: Interleukin (IL), granulocyte-macrophage-colony-stimulating factor (GM-CSF), tumor necrosis factor-alpha (TNF-α), interferon-gamma (IFN-γ), and C-reactive protein (hs-CRP).

191

Table 29. Variations in Median and Inter-Quartile Range Distribution of Inflammatory Biomarkers by Waist Circumference (WC) Measurement

Inflammatory Factor Low WC < 80 cm High WC ≥ 80 cm P Ϫ (pg/mL) (n = 176) (n = 89) Median (Inter-Quartile Range) IL-1ß 1.17 (2.40) 0.24 (0.83) < 0.01 IL-2 1.52 (2.25) 1.20 (2.09) 0.37 IL-4 5.38 ( 5.60) 5.98 ( 7.36) 0.58 IL-5 0.17 (0.22) 0.21 (0.26) 0.18 IL-6 2.15 (2.60) 2.58 (3.83) 0.08 IL-7 2.49 (3.66) 1.01 (3.03) < 0.01 IL-8 1.76 (1.28) 1.69 (1.09) 0.67 IL-10 9.92 (11.56) 13.48 (16.26) 0.08 IL-12p70 3.29 (6.05) 3.98 (12.61) 0.29 IL-13 2.29 (2.66) 2.28 (3.84) 0.62 GM-CSF 11.63 (24.47) 2.95 (8.22) < 0.01 TNF-α 5.18 (6.02) 3.22 (3.24) < 0.01 IFN-γ 2.70 (3.61) 1.70 (4.09) 0.11 hs-CRP (mg/L) ѣ 0.70 (3.55) 0.90 (3.30) 0.95 Ϫ P-values compare groups using Mann-Whitney tests. Ѣ n = 171 for hs-CRP. Definition: Interleukin (IL), granulocyte-macrophage-colony-stimulating factor (GM-CSF), tumor necrosis factor-alpha (TNF-α), interferon-gamma (IFN-γ), and C-reactive protein (hs-CRP).

192

Table 30. Variations in Median and Inter-Quartile Range Distribution of Inflammatory Biomarkers by Body Mass Index (BMI)

Inflammatory Factor Low BMI < 25 kg/m2 High BMI ≥ 25 kg/m2 P Ϫ (pg/mL) (n = 208) (n = 57) Median (Inter-Quartile Range) IL-1ß 0.80 (2.32) 0.25 (2.06) 0.08 IL-2 1.50 (2.24) 1.20 (1.85) 0.36 IL-4 5.49 ( 6.10) 5.63 ( 6.37) 0.85 IL-5 0.18 (0.24) 0.20 (0.22) 0.95 IL-6 2.06 (3.02) 2.97 (4.57) 0.02 IL-7 2.13 (3.83) 1.57 (3.43) 0.24 IL-8 1.72 (1.12) 1.75 (0.75) 0.55 IL-10 10.02 (16.34) 12.99 (12.16) 0.16 IL-12p70 3.33 (6.05) 4.60 (13.21) 0.13 IL-13 2.29 (3.05) 2.28 (3.49) 0.52 GM-CSF 7.86 (23.36) 3.37 (16.04) 0.05* TNF-α 4.86 (5.79) 3.87 (4.80) 0.55 IFN-γ 2.46 (3.80) 2.16 (3.86) 0.61 hs-CRP (mg/L) ѣ 0.80 (3.90) 0.60 (1.30) 0.10 Ϫ P-values compare groups using Mann-Whitney tests. Ѣ n = 171 for hs-CRP. * Estimated P-value is borderline significant (P = 0.049). Definition: Interleukin (IL), granulocyte-macrophage-colony-stimulating factor (GM-CSF), tumor necrosis factor-alpha (TNF-α), interferon-gamma (IFN-γ), and C-reactive protein (hs-CRP).

193

Table 31. Variations in Median and Inter-Quartile Range Distribution of Inflammatory Biomarkers by Total Body Fat Percentage (TBF %):

Inflammatory Factor Low TBF < 32 % High TBF ≥ 32% P Ϫ (pg/mL) (n = 131) (n = 139) Median (Inter-Quartile Range) IL-1ß 0.85 (2.34) 0.56 (2.16) 0.22 IL-2 1.45 (2.01) 1.51 (2.30) 0.55 IL-4 5.36 ( 5.38) 5.89 ( 6.83) 0.34 IL-5 0.18 (0.24) 0.19 (0.24) 0.51 IL-6 2.06 (2.46) 2.55 (3.69) 0.03 IL-7 2.35 (4.11) 1.77 (3.44) 0.46 IL-8 1.71 (1.08) 1.75 (1.01) 0.42 IL-10 9.91 (11.64) 11.72 (17.92) 0.14 IL-12p70 3.25 (6.03) 4.01 (11.83) 0.34 IL-13 2.13 (2.26) 2.41 (3.70) 0.15 GM-CSF 8.47 (23.23) 5.47 (20.03) 0.34 TNF-α 4.57 (5.98) 4.74 (4.93) 0.92 IFN-γ 2.63 (3.87) 2.20 (3.88) 0.42 hs-CRP (mg/L) ѣ 0.80 (4.40) 0.80 (2.90) 0.56 Ϫ P-values compare groups using Mann-Whitney tests. Ѣ n = 171 for hs-CRP. Definition: Interleukin (IL), granulocyte-macrophage-colony-stimulating factor (GM-CSF), tumor necrosis factor-alpha (TNF-α), interferon-gamma (IFN-γ), and C-reactive protein (hs-CRP).

194

Table 32. Median Inflammatory Biomarker Concentration by Categories of Oral Contraceptive Use, Smoking Status and Alcohol Intake

Oral Contraceptive Use Smoking Status Alcohol Intake Inflammatory Factor (pg/mL) No (152) Yes (113) No (249) Yes (15) No ( 112) Yes(59) Median P Ϫ Median P Ϫ Median P Ϫ (Inter-Quartile Range) (Inter-Quartile Range) (Inter-Quartile Range) IL-1ß 0.64 (2.41) 0.74 (2.21) 0.81 0.72 (2.32) 0.43 (0.72) 0.57 0.77 (2.24) 0.70 (2.34) 0.62 IL-2 1.38 (2.00) 1.52 (2.21) 0.33 1.47 (2.16) 1.48 (1.55) 0.84 1.02 (2.17) 1.53 (2.18) 0.07 IL-4 5.61 (6.40) 5.45 (5.61) 0.67 5.45 (5.97) 6.57 (7.63) 0.36 4.15 (6.57) 5.97 (6.15) 0.05 IL-5 0.17 (0.24) 0.19 (0.23) 0.52 0.18 (0.23) 0.17 (0.36) 0.64 0.18 (0.26) 0.19 (0.22) 0.96 IL-6 2.04 (2.86) 2.64 (3.29) 0.08 2.21 (2.96) 2.47 (5.19) 0.59 2.01 (3.18) 2.37 (2.99) 0.40 IL-7 1.85 (3.72) 2.17 (3.62) 0.24 2.11 (3.75) 1.70 (3.02) 0.82 2.17 (3.51) 1.97 (4.15) 0.78 IL-8 1.72 (1.08) 1.75 (1.04) 0.86 1.72 (1.02) 1.54 (2.24) 0.88 1.68 (0.98) 1.74 (1.13) 0.73 Ϫ P-values compare groups using Mann-Whitney tests. Definition: Interleukin (IL), granulocyte-macrophage-colony-stimulating factor (GM-CSF), tumor necrosis factor-alpha (TNF-α), interferon-gamma (IFN-γ), and C-reactive protein (hs-CRP).

195

Table 32. Continued

Oral Contraceptive Use Smoking Status Alcohol Intake Inflammatory Factor (pg/mL) No ( 152) Yes (113) No (249) Yes (15 ) No ( 112) Yes( 59) Median P Ϫ Median P Ϫ Median P Ϫ (Inter-Quartile Range) (Inter-Quartile Range) (Inter-Quartile Range) IL-10 11.07 (13.43) 9.91 (16.69) 0.59 10.52 (14.26) 9.92 (18.03) 0.89 10.04 (20.24) 10.92 (13.41) 0.81 IL-12p70 3.25 (5.80) 4.19 (10.23) 0.24 3.37 (7.44) 4.65 (11.39) 0.62 3.57 (8.54) 3.55 (7.27) 0.77 IL-13 2.23 (3.21) 2.40 (3.13) 0.50 2.28 (3.16) 2.49 (3.86) 0.79 1.77 (2.10) 2.43 (3.25) 0.06 GM-CSF 6.35 (23.22) 7.59 (21.35) 0.59 7.15 (23.15) 5.31 (11.77) 0.91 7.29 (19.87) 6.96 (23.54) 0.35 TNF-α 4.24 (5.57) 5.09 (5.36) 0.09 4.82 (5.75) 3.67 (3.51) 0.44 4.35 (5.30) 4.83 (5.85) 0.93 IFN-γ 2.18 (3.96) 2.96 (3.49) 0.05 2.36 (3.88) 2.29 (3.15) 0.95 2.16 (4.35) 2.43 (3.84) 0.69 hs-CRP (mg/L) ѣ 0.75 (3.80) 0.90 (2.80) 0.91 0.80 (3.30) 0.60 (1.50) 0.67 1.00 (4.50) 0.60 (2.45) 0.19 Ϫ P-values compare groups using Mann-Whitney tests. Ѣ n = 171 for hs-CRP. Definition: Interleukin (IL), granulocyte-macrophage-colony-stimulating factor (GM-CSF), tumor necrosis factor-alpha (TNF-α), interferon-gamma (IFN-γ), and C-reactive protein (hs-CRP).

196

Table 33. Pearson Product Correlations between Log Transformed Inflammatory Factors

log log log log log log log log log log log log log log il_1ß il_2 il_4 il_5 il_6 il_7 il_8 il_10 il_12 il_13 gm_csf tnf_α ifn_γ hs-crp log il_1ß --- log il_2 0.65* --- log il_4 0.54* 0.63* --- log il_5 0.37* 0.43* 0.37* --- log il_6 0.27* 0.44* 0.36* 0.29* --- log il_7 0.78* 0.63* 0.53* 0.43* 0.55* --- log il_8 0.41* 0.45* 0.49* 0.32* 0.41* 0.48* --- log il_10 0.25* 0.43* 0.33* 0.46* 0.65* 0.51* 0.40* --- log il_12 0.31* 0.57* 0.40* 0.36* 0.80* 0.57* 0.36* 0.69 --- log il_13 0.47* 0.60* 0.53* 0.45* 0.61* 0.71* 0.43* 0.57* 0.66* --- log gm_csf 0.93* 0.66* 0.53* 0.34* 0.26* 0.77* 0.38* 0.20* 0.29* 0.46* --- log tnf_α 0.85* 0.53* 0.43* 0.34* 0.27* 0.72* 0.41* 0.19* 0.23* 0.39* 0.82* --- log ifn_γ 0.71* 0.76* 0.60* 0.48* 0.45* 0.74* 0.45* 0.41* 0.50* 0.66* 0.71* 0.62* --- log hs-crp 0.02 0.01 0.10 0.04 0.004 -0.04 0.13 0.01 0.02 -0.01 0.03 -0.004 0.02 --- (0.76) (0.85) (0.19) (0.64) (0.96) (0.63) (0.11) (0.89) (0.77) (0.92) (0.66) (0.96) (0.80)

All values in parentheses are P-values * Indicates P-values are significant at P < 0.001.

197

Table 34. Correlations between Log Transformed Inflammatory Factors and Serum 25-OHD Levels

Serum 25-OHD Levels Inflammatory Factor (pg/mL) Correlation Coefficient (r) P-value log IL-1ß -0.03 0.72 log IL-2 -0.07 0.34 log IL-4 -0.05 0.56 log IL-5 0.04 0.58 log IL-6 -0.07 0.33 log IL-7 -0.01 0.85 log IL-8 0.02 0.80 log IL-10 0.08 0.29 log IL-12p70 -0.03 0.66 log IL-13 -0.04 0.67 log GM-CSF -0.09 0.26 log TNF-α -0.003 0.97 log IFN-γ -0.04 0.63 log hs-CRP (mg/L) ѣ 0.10 0.21 Ϫ ₸ r (Correlation Coefficient) and P-value for significance estimated using Pearson’s Product. Ѣ n = 171 for hs-CRP.

198

Table 35. Mean Serum 25-OHD Concentration across Categories of High Sensitive C - Reactive Protein (hs-CRP) Measures

hs-CRP Concentration (mg/L)

< 1 mg/L (n = 94) ≥ 1 mg/L (n = 74) Mean ± Standard Deviation P-value ⱶ Serum 25-OHD, nmol/L 76.7 ± 32.1 82.8 ± 33.0 0.23

< 3 mg/L (n = 119) ≥ 3 mg/L (n = 49) Mean ± Standard Deviation P-value ⱶ Serum 25-OHD, nmol/L 77.1 ± 32.2 85.1 ± 33.0 0.15

< 5 mg/L (n = 135) ≥ 5 mg/L (n = 33) Mean ± Standard Deviation P-value ⱶ Serum 25-OHD, nmol/L 76.4 ± 31.1 91.6 ± 35.8 0.02 ⱶ P-values compare groups using student t-tests

199

Figure 21. Lowess Regression Curve for the Association between High Sensitive C - Reactive Protein (hs-CRP) and Serum 25-OHD Concentration

Lowess smoother Lowess smoother

40 40

30 30

20 20

10 10

hs-C ReactiveProtein(mg/L)

hs-CReactive Protein(mg/L)

0 0

0 50 100 150 200 0 50 100 150 200 Serum 25-OHD (nmol/L) Serum 25-OHD (nmol/L) hs-crp_mgl hs-crp_mgl lowess: hs-crp_mgl lowess: hs-crp_mgl Fitted values 95% CI Fitted values

bandwidth = .3 bandwidth = .3

Lowess smoother Lowess smoother

4 4

2 2

0 0

loghs-C ReactiveProtein(mg/L)

-2 -2

0 50 100 150 200 0 50 100 150 200 Serum 25-OHD (nmol/L) Serum 25-OHD (nmol/L)

ths-crp_mgl lowess: ths-crp_mgl ths-crp_mgl lowess: ths-crp_mgl Fitted values 95% CI Fitted values

bandwidth = .3 bandwidth = .3

200

Figure 22. Lowess Regression Curve for the Association between Interleukin-1β (IL-1β) and Serum 25-OHD Concentration

Lowess smoother Lowess smoother

50 50

Interleukin-1b(pg/ml) Interleukin-1b(pg/ml)

0 0

0 50 100 150 200 0 50 100 150 200 Serum 25-OHD (nmol/L) Serum 25-OHD (nmol/L)

IL_1b lowess: IL_1b IL_1b lowess: IL_1b Fitted values 95% CI Fitted values

bandwidth = .3 bandwidth = .3

Lowess smoother Lowess smoother

4 4

2 2

0 0

logInterleukin-1b(pg/ml) logInterleukin-1b(pg/ml)

-2 -2

0 50 100 150 200 0 50 100 150 200 Serum 25-OHD (nmol/L) Serum 25-OHD (nmol/L)

til_1b lowess: til_1b til_1b lowess: til_1b Fitted values 95% CI Fitted values

bandwidth = .3 bandwidth = .3

201

Figure 23. Lowess Regression Curve for the Association between Interleukin-2 (IL-2) and Serum 25-OHD Concentration

Lowess smoother Lowess smoother

1000 1000

800

600

400

Interleukin-2(pg/ml) Interleukin-2(pg/ml)

200

0 0

0 50 100 150 200 0 50 100 150 200 Serum 25-OHD (nmol/L) Serum 25-OHD (nmol/L)

IL_2 lowess: IL_2 IL_2 lowess: IL_2 Fitted values 95% CI Fitted values

bandwidth = .3 bandwidth = .3

Lowess smoother Lowess smoother

6 6

4 4

2 2

0 0

logInterleukin-2(pg/ml) logInterleukin-2(pg/ml)

-2 -2 0 50 100 150 200 0 50 100 150 200 Serum 25-OHD (nmol/L) Serum 25-OHD (nmol/L)

til_2 lowess: til_2 til_2 lowess: til_2 Fitted values 95% CI Fitted values

bandwidth = .3 bandwidth = .3

202

Figure 24. Lowess Regression Curve for the Association between Interleukin-4 (IL-4) and Serum 25-OHD Concentration

Lowess smoother Lowess smoother

800 800

600

400

Interleukin-4(pg/ml) Interleukin-4(pg/ml)

200

0 0

0 50 100 150 200 0 50 100 150 200 Serum 25-OHD (nmol/L) Serum 25-OHD (nmol/L)

IL_4 lowess: IL_4 IL_4 lowess: IL_4 Fitted values 95% CI Fitted values

bandwidth = .3 bandwidth = .3

Lowess smoother Lowess smoother

6 6

4 4

2 2

0 0

logInterleukin-4(pg/ml) logInterleukin-4(pg/ml)

-2 -2

0 50 100 150 200 0 50 100 150 200 Serum 25-OHD (nmol/L) Serum 25-OHD (nmol/L)

til_4 lowess: til_4 til_4 lowess: til_4 Fitted values 95% CI Fitted values

bandwidth = .3 bandwidth = .3

203

Figure 25. Lowess Regression Curve for the Association between Interleukin-5 (IL-5) and Serum 25-OHD Concentration

Lowess smoother Lowess smoother

Interleukin-5(pg/ml) Interleukin-5(pg/ml)

0 0

0 50 100 150 200 0 50 100 150 200 Serum 25-OHD (nmol/L) Serum 25-OHD (nmol/L)

IL_5 lowess: IL_5 IL_5 lowess: IL_5 Fitted values 95% CI Fitted values

bandwidth = .3 bandwidth = .3

Lowess smoother Lowess smoother

4 4

2 2

0 0

logInterleukin-5(pg/ml) logInterleukin-5(pg/ml)

-2 -2

0 50 100 150 200 0 50 100 150 200 Serum 25-OHD (nmol/L) Serum 25-OHD (nmol/L)

til_5 lowess: til_5 til_5 lowess: til_5 Fitted values 95% CI Fitted values

bandwidth = .3 bandwidth = .3

204

Figure 26. Lowess Regression Curve for the Association between Interleukin-6 (IL-6) and Serum 25-OHD Concentration

Lowess smoother Lowess smoother

100 100

Interleukin-6(pg/ml) Interleukin-6(pg/ml)

0 0

0 50 100 150 200 0 50 100 150 200 Serum 25-OHD (nmol/L) Serum 25-OHD (nmol/L)

IL_6 lowess: IL_6 IL_6 lowess: IL_6 Fitted values 95% CI Fitted values

bandwidth = .3 bandwidth = .3

Lowess smoother Lowess smoother

4 4

2 2

0 0

logInterleukin-6(pg/ml) logInterleukin-6(pg/ml)

-2 -2 0 50 100 150 200 0 50 100 150 200 Serum 25-OHD (nmol/L) Serum 25-OHD (nmol/L)

til_6 lowess: til_6 til_6 lowess: til_6 Fitted values 95% CI Fitted values

bandwidth = .3 bandwidth = .3

205

Figure 27. Lowess Regression Curve for the Association between Interleukin-7 (IL-7) and Serum 25-OHD Concentration

Lowess smoother Lowess smoother

40 40

30 30

20 20

Interleukin-7(pg/ml) Interleukin-7(pg/ml)

10 10

0 0

0 50 100 150 200 0 50 100 150 200 Serum 25-OHD (nmol/L) Serum 25-OHD (nmol/L)

IL_7 lowess: IL_7 IL_7 lowess: IL_7 Fitted values 95% CI Fitted values

bandwidth = .3 bandwidth = .3

Lowess smoother Lowess smoother

4 4

2 2

0 0

logInterleukin-7(pg/ml) logInterleukin-7(pg/ml)

-2 -2 0 50 100 150 200 0 50 100 150 200 Serum 25-OHD (nmol/L) Serum 25-OHD (nmol/L)

til_7 lowess: til_7 til_7 lowess: til_7 Fitted values 95% CI Fitted values

bandwidth = .3 bandwidth = .3

206

Figure 28. Lowess Regression Curve for the Association between Interleukin-8 (IL-8) and Serum 25-OHD Concentration

Lowess smoother Lowess smoother

100 100

Interleukin-8(pg/ml) Interleukin-8(pg/ml)

0 0

0 50 100 150 200 0 50 100 150 200 Serum 25-OHD (nmol/L) Serum 25-OHD (nmol/L)

IL_8 lowess: IL_8 IL_8 lowess: IL_8 Fitted values 95% CI Fitted values

bandwidth = .3 bandwidth = .3

Lowess smoother Lowess smoother

6 6

4 4

2 2

0 0

logInterleukin-8(pg/ml) logInterleukin-8(pg/ml)

-2 -2 0 50 100 150 200 0 50 100 150 200 Serum 25-OHD (nmol/L) Serum 25-OHD (nmol/L)

til_8 lowess: til_8 til_8 lowess: til_8 Fitted values 95% CI Fitted values

bandwidth = .3 bandwidth = .3

207

Figure 29. Lowess Regression Curve for the Association between Interleukin-10 (IL-10) and Serum 25-OHD Concentration

Lowess smoother Lowess smoother

1500

1000

500

Interleukin-10(pg/ml) Interleukin-10(pg/ml)

0 0

0 50 100 150 200 0 50 100 150 200 Serum 25-OHD (nmol/L) Serum 25-OHD (nmol/L)

IL_10 lowess: IL_10 IL_10 lowess: IL_10 Fitted values 95% CI Fitted values

bandwidth = .3 bandwidth = .3

Lowess smoother Lowess smoother

8 8

6 6

4 4

2 2

logInterleukin-10(pg/ml) logInterleukin-10(pg/ml)

0 0

0 50 100 150 200 0 50 100 150 200 Serum 25-OHD (nmol/L) Serum 25-OHD (nmol/L)

til_10 lowess: til_10 til_10 lowess: til_10 Fitted values 95% CI Fitted values

bandwidth = .3 bandwidth = .3

208

Figure 30. Lowess Regression Curve for the Association between Interleukin-12 (IL-12p70) and Serum 25-OHD Concentration

Lowess smoother Lowess smoother

3000 3000

2000

1000

Interleukin-12p70(pg/ml) Interleukin-12p70(pg/ml)

0 0

0 50 100 150 200 0 50 100 150 200 Serum 25-OHD (nmol/L) Serum 25-OHD (nmol/L)

IL_12p70 lowess: IL_12p70 IL_12p70 lowess: IL_12p70 Fitted values 95% CI Fitted values

bandwidth = .3 bandwidth = .3

Lowess smoother Lowess smoother

8 8

6 6

4 4

2 2

0 0

logInterleukin-12p70(pg/ml) logInterleukin-12p70(pg/ml)

-2 -2 0 50 100 150 200 0 50 100 150 200 Serum 25-OHD (nmol/L) Serum 25-OHD (nmol/L)

til_12p70 lowess: til_12p70 til_12p70 lowess: til_12p70 Fitted values 95% CI Fitted values

bandwidth = .3 bandwidth = .3

209

Figure 31. Lowess Regression Curve for the Association between Interleukin-13 (IL-13) and Serum 25-OHD Concentration

Lowess smoother Lowess smoother

250 250

200

150

100

Interleukin-13(pg/ml) Interleukin-13(pg/ml)

0 0

0 50 100 150 200 0 50 100 150 200 Serum 25-OHD (nmol/L) Serum 25-OHD (nmol/L)

IL_13 lowess: IL_13 IL_13 lowess: IL_13 Fitted values 95% CI Fitted values

bandwidth = .3 bandwidth = .3

Lowess smoother Lowess smoother

6 6

4 4

2 2

0 0

logInterleukin-13(pg/ml) logInterleukin-13(pg/ml)

-2 -2 0 50 100 150 200 0 50 100 150 200 Serum 25-OHD (nmol/L) Serum 25-OHD (nmol/L)

til_13 lowess: til_13 til_13 lowess: til_13 Fitted values 95% CI Fitted values

bandwidth = .3 bandwidth = .3

210

Figure 32. Lowess Regression Curve for the Association between Granulocyte Macrophage Colony Stimulating Factor (GM-CSF) and Serum 25-OHD Concentration

Lowess smoother Lowess smoother

250 250

200 200

150 150

100 100

50 50

0 0

0 50 100 150 200 0 50 100 150 200 Serum 25-OHD (nmol/L) Serum 25-OHD (nmol/L)

GM_CSF lowess: GM_CSF GM_CSF lowess: GM_CSF

Fitted values 95% CI Fitted values Granulocyte-macrophagecolony-stimulating factor(pg/ml) Granulocyte-macrophagecolony-stimulating factor(pg/ml) bandwidth = .3 bandwidth = .3

Lowess smoother Lowess smoother

6 6

4 4

2 2

0 0

-2 -2

0 50 100 150 200 0 50 100 150 200 Serum 25-OHD (nmol/L) Serum 25-OHD (nmol/L)

tgm_csf lowess: tgm_csf tgm_csf lowess: tgm_csf Fitted values 95% CI Fitted values

bandwidth = .3 bandwidth = .3 logGranulocyte-macrophage colony-stimulating factor(pg/ml) logGranulocyte-macrophage colony-stimulating factor(pg/ml) 211

Figure 33. Lowess Regression Curve for the Association between Tumor Necrosis Factor Alpha (TNF-α) and Serum 25-OHD Concentration

Lowess smoother Lowess smoother

60 60

40 40

20 20

Tumornecrosis alpha(pg/ml)factor Tumornecrosis alpha(pg/ml)factor

0 0

0 50 100 150 200 0 50 100 150 200 Serum 25-OHD (nmol/L) Serum 25-OHD (nmol/L)

TNF_a lowess: TNF_a TNF_a lowess: TNF_a Fitted values 95% CI Fitted values

bandwidth = .3 bandwidth = .3

Lowess smoother Lowess smoother

4 4

3 3

2 2

1 1

0 0

-1 -1

logTumornecrosis alpha(pg/ml)factor logTumornecrosis alpha(pg/ml)factor 0 50 100 150 200 0 50 100 150 200 Serum 25-OHD (nmol/L) Serum 25-OHD (nmol/L)

ttnf_a lowess: ttnf_a ttnf_a lowess: ttnf_a Fitted values 95% CI Fitted values

bandwidth = .3 bandwidth = .3

212

Figure 34. Lowess Regression Curve for the Association between Interferon Gamma (IFN-γ) and Serum 25-OHD Concentration

Lowess smoother Lowess smoother

600 600

400

200

Interferon gamma(pg/ml) Interferon gamma(pg/ml)

0 0

0 50 100 150 200 0 50 100 150 200 Serum 25-OHD (nmol/L) Serum 25-OHD (nmol/L)

IFN_g lowess: IFN_g IFN_g lowess: IFN_g Fitted values 95% CI Fitted values

bandwidth = .3 bandwidth = .3

Lowess smoother Lowess smoother

6 6

4 4

2 2

0 0

logInterferon gamma(pg/ml) logInterferon gamma(pg/ml)

-2 -2

0 50 100 150 200 0 50 100 150 200 Serum 25-OHD (nmol/L) Serum 25-OHD (nmol/L)

tifn_g lowess: tifn_g tifn_g lowess: tifn_g Fitted values 95% CI Fitted values

bandwidth = .3 bandwidth = .3

213

Table 36. Univariate Linear Relationship between log Transformed Inflammatory Biomarkers and Other Parameters

log hs-CRP log IL-1β Unadjusted Unadjusted Predictor ß ± SE P-value ß ± SE P-value Age, y -0.04 ± 0.04 0.32 -0.02 ± 0.03 0.57 Body Mass Index, kg/m2 -0.04 ± 0.04 0.26 -0.03 ± 0.03 0.28 Total Body Fat, % -0.01 ± 0.01 0.68 -0.01 ± 0.01 0.34 Waist Circumference, cm -0.01 ± 0.01 0.59 -0.02 ± 0.01 0.03 Waist to Height Ratio -1.05 ± 2.35 0.66 -3.66 ± 1.63 0.03 Education a - 0.06 ± 0.30 0.83 - 0.21 ± 0.23 0.37 Season of blood draw b -0.67 ± 0.30 0.03 -0.08 ± 0.24 0.73 Smoking status c 0.27 ± 0.53 0.61 0.21 ± 0.38 0.58 Alcohol intake c -0.35 ± 0.25 0.16 -0.30 ± 0.20 0.88 Oral Contraceptive use c -0.04 ± 0.25 0.88 -0.05 ± 0.18 0.77 Physical activity level, MET-hrs/week 0.002 ± 0.003 0.54 -0.02 ± 0.002 0.41 Multivitamin intake, g/day c -0.20 ± 0.24 0.40 -0.34 ± 0.18 0.06 ϖ Regression coefficients (ß) ± standard errors (SE). All analyses were estimated using simple linear regressions. Beta coefficient represents change in inflammatory factor level per unit change in predictor. a Dummy variable (0 = Some College, 1= College Graduate) b Dummy variable (0 = winter/ 1 = non-winter, i.e. Spring, Summer, Fall). c Dummy variable (0 = no / 1 = yes).

214

Table 36. Continued

log IL-2 log IL-4 Unadjusted Unadjusted Predictor ß ± SE P-value ß ± SE P-value

Age, y -0.02 ± 0.03 0.37 -0.01 ± 0.03 0.77 Body Mass Index, kg/m2 0.01 ± 0.02 0.56 0.04 ± 0.02 0.09 Total Body Fat, % 0.01 ± 0.01 0.20 0.02 ± 0.01 0.06 Waist Circumference, cm -0.001 ± 0.01 0.87 0.01 ± 0.01 0.16

Waist to Height Ratio -4.62 ± 1.48 0.76 -1.55 ± 1.34 0.25 Education a - 0.24 ± 0.21 0.25 - 0.12 ± 0.19 0.54 Season of blood draw b -0.33 ± 0.22 0.13 -0.05 ± 0.20 0.80 Smoking status c 0.28 ± 0.34 0.40 0.24 ± 0.31 0.44 c Alcohol intake 0.17 ± 0.18 0.34 0.16 ± 0.16 0.34 Oral Contraceptive use c 0.13 ± 0.16 0.40 -0.01 ± 0.14 0.97 Physical activity level, MET-hrs/week -0.002 ± 0.002 0.19 -0.001 ± 0.002 0.38 Multivitamin intake, g/day c -0.20 ± 0.16 0.22 -0.24 ± 0.15 0.10

ϖ Regression coefficients (ß) ± standard errors (SE). All analyses were estimated using simple linear regressions. Beta coefficient represents change in inflammatory factor level per unit change in predictor. a Dummy variable (0 = Some College, 1= College Graduate) b Dummy variable (0 = winter/ 1 = non-winter, i.e. Spring, Summer, Fall). c Dummy variable (0 = no / 1 = yes).

215

Table 36. Continued

log IL-5 log IL-6

Unadjusted Unadjusted Predictor ß ± SE P-value ß ± SE P-value Age, y -0.03 ± 0.02 0.13 0.03 ± 0.03 0.30 Body Mass Index, kg/m2 0.01 ± 0.02 0.67 0.08 ± 0.02 < 0.01 Total Body Fat, % 0.01 ± 0.01 0.26 0.03 ± 0.01 < 0.01

Waist Circumference, cm 0.01 ± 0.01 0.32 0.02 ± 0.01 0.03 Waist to Height Ratio -0.82 ± 0.99 0.41 3.10 ± 1.53 0.04 Education a -0.20 ± 0.14 0.16 0.12 ± 0.22 0.57 Season of blood draw b 0.01 ± 0.15 0.94 0.21 ± 0.23 0.36 c Smoking status -0.03 ± 0.23 0.89 0.17 ± 0.35 0.62 Alcohol intake c 0.004 ± 0.12 0.98 0.01 ± 0.19 0.96 Oral Contraceptive use c 0.07 ± 0.11 0.54 0.18 ± 0.17 0.28 Physical activity level, MET-hrs/week 0.001 ± 0.001 0.43 -0.002 ± 0.002 0.26

Multivitamin intake, g/day c -0.20 ± 0.11 0.07 -0.09 ± 0.17 0.58 ϖ Regression coefficients (ß) ± standard errors (SE). All analyses were estimated using simple linear regressions. Beta coefficient represents change in inflammatory factor level per unit change in predictor. a Dummy variable (0 = Some College, 1= College Graduate) b Dummy variable (0 = winter/ 1 = non-winter, i.e. Spring, Summer, Fall). c Dummy variable (0 = no / 1 = yes).

216

Table 36. Continued

log IL-7 log IL-8 Unadjusted Unadjusted Predictor ß ± SE P-value ß ± SE P-value

Age, y 0.001 ± 0.03 0.97 -0.01 ± 0.02 0.44 2 Body Mass Index, kg/m 0.003 ± 0.03 0.90 0.02 ± 0.01 0.09 Total Body Fat, % 0.002 ± 0.01 0.83 0.01 ± 0.01 0.37 Waist Circumference, cm -0.01 ± 0.01 0.11 0.01 ± 0.01 0.07 Waist to Height Ratio -2.16 ± 1.55 0.16 1.30 ± 0.83 0.12 Education a - 0.16 ± 0.22 0.46 - 0.03 ± 0.11 0.83 Season of blood draw b 0.17 ± 0.23 0.47 -0.05 ± 0.12 0.68 Smoking status c 0.24 ± 0.35 0.50 0.48 ± 0.19 0.01

Alcohol intake c 0.35 ± 0.19 0.86 -0.10 ± 0.10 0.34 Oral Contraceptive use c 0.24 ± 0.17 0.15 0.03 ± 0.09 0.72 Physical activity level, MET-hrs/week -0.003 ± 0.002 0.15 0.001 ± 0.001 0.58 Multivitamin intake, g/day c -0.18 ± 0.17 0.29 -0.03 ± 0.09 0.70 ϖ Regression coefficients (ß) ± standard errors (SE). All analyses were estimated using simple linear regressions. Beta coefficient represents change in inflammatory factor level per unit change in predictor. a Dummy variable (0 = Some College, 1= College Graduate) b Dummy variable (0 = winter/ 1 = non-winter, i.e. Spring, Summer, Fall). c Dummy variable (0 = no / 1 = yes).

217

Table 36. Continued

log IL-10 log IL-12p70 Unadjusted Unadjusted Predictor ß ± SE P-value ß ± SE P-value Age, y 0.03 ± 0.02 0.24 0.01 ± 0.04 0.81 2 Body Mass Index, kg/m 0.05 ± 0.02 0.02 0.07 ± 0.04 0.05 Total Body Fat, % 0.02 ± 0.01 0.06 0.02 ± 0.01 0.09 Waist Circumference, cm 0.01 ± 0.01 0.10 0.01 ± 0.01 0.35 Waist to Height Ratio 1.90 ± 1.27 0.14 1.53 ± 2.17 0.48 Education a 0.11 ± 0.18 0.54 - 0.02 ± 0.31 0.94 Season of blood draw b -0.08 ± 0.19 0.69 0.10 ± 0.32 0.76 Smoking status c -0.05 ± 0.28 0.86 0.22 ± 0.49 0.65 c Alcohol intake -0.02 ± 0.16 0.91 -0.01 ± 0.27 0.97 c Oral Contraceptive use -0.05 ± 0.14 0.69 0.19 ± 0.23 0.41 Physical activity level, MET-hrs/week 0.0003 ± 0.002 0.83 -0.0003 ± 0.003 0.93 Multivitamin intake, g/day c -0.02 ± 0.14 0.88 -0.34 ± 0.23 0.15 ϖ Regression coefficients (ß) ± standard errors (SE). All analyses were estimated using simple linear regressions. Beta coefficient represents change in inflammatory factor level per unit change in predictor. a Dummy variable (0 = Some College, 1= College Graduate) b Dummy variable (0 = winter/ 1 = non-winter, i.e. Spring, Summer, Fall). c Dummy variable (0 = no / 1 = yes).

218

Table 36. Continued

log IL-13 log GM-CSF Unadjusted Unadjusted Predictor ß ± SE P-value ß ± SE P-value Age, y -0.003 ± 0.03 0.93 -0.03 ± 0.03 0.38 Body Mass Index, kg/m2 0.06 ± 0.03 0.03 -0.03 ± 0.03 0.27 Total Body Fat, % 0.03 ± 0.01 0.02 -0.01 ± 0.01 0.66 Waist Circumference, cm 0.01 ± 0.01 0.40 -0.02 ± 0.01 0.02 Waist to Height Ratio 1.63 ± 1.61 0.31 -4.28 ± 1.78 0.02 Education a -0.20 ± 0.23 0.38 - 0.28 ± 0.25 0.27 Season of blood draw b 0.31 ± 0.24 0.19 -0.05 ± 0.27 0.86 Smoking status c 0.16 ± 0.37 0.67 0.03 ± 0.41 0.93

Alcohol intake c 0.19 ± 0.20 0.32 0.26 ± 0.22 0.23 Oral Contraceptive use c 0.16 ± 0.17 0.36 0.16 ± 0.19 0.41 Physical activity level, MET-hrs/week -0.003 ± 0.002 0.13 -0.003 ± 0.002 0.26 Multivitamin intake, g/day c -0.24 ± 0.17 0.17 -0.23 ± 0.20 0.23 ϖ Regression coefficients (ß) ± standard errors (SE). All analyses were estimated using simple linear regressions. Beta coefficient represents change in inflammatory factor level per unit change in predictor. a Dummy variable (0 = Some College, 1= College Graduate) b Dummy variable (0 = winter/ 1 = non-winter, i.e. Spring, Summer, Fall). c Dummy variable (0 = no / 1 = yes).

219

Table 36. Continued

log TNF-α log IFN-γ Unadjusted Unadjusted Predictor ß ± SE P-value ß ± SE P-value Age, y -0.01 ± 0.02 0.50 -0.01 ± 0.03 0.82 Body Mass Index, kg/m2 0.001 ± 0.01 0.93 -0.004 ± 0.03 0.87 Total Body Fat, % 0.002 ± 0.01 0.68 0.0003 ± 0.01 0.97

Waist Circumference, cm -0.01 ± 0.01 0.33 -0.01 ± 0.01 0.19 Waist to Height Ratio -0.82 ± 0.86 0.34 -2.05 ± 1.65 0.21 Education a - 0.08 ± 0.12 0.52 - 0.14 ± 0.23 0.55 Season of blood draw b 0.04 ± 0.13 0.76 -0.15 ± 0.25 0.53 c Smoking status -0.14 ± 0.20 0.47 0.40 ± 0.38 0.29 Alcohol intake c 0.01 ± 0.11 0.90 -0.02 ± 0.20 0.93 Oral Contraceptive use c 0.16 ± 0.09 0.08 0.36 ± 0.18 0.05 Physical activity level, MET-hrs/week -0.0001 ± 0.001 0.91 -0.0003 ± 0.002 0.90

Multivitamin intake, g/day c -0.08 ± 0.09 0.42 -0.10 ± 0.18 0.59 ϖ Regression coefficients (ß) ± standard errors (SE). All analyses were estimated using simple linear regressions. Beta coefficient represents change in inflammatory factor level per unit change in predictor. a Dummy variable (0 = Some College, 1= College Graduate) b Dummy variable (0 = winter/ 1 = non-winter, i.e. Spring, Summer, Fall). c Dummy variable (0 = no / 1 = yes).

220

Table 37. Association of Serum 25-OHD Concentrations with Log Transformed Inflammatory Biomarkers (Beta ± SE)§ in Young Women: UMass Amherst Vitamin D Status Study, 2006 – 2011

Y = log (high sensitive C-Reactive Protein Concentration, mg/L) 25-OHD, nmol/L 25-OHD < 75 nmol/L 25-OHD ≥ 75 nmol/L All women (n = 168) (n = 90) (n = 78) ß ± SEϖ P Ϫ ß ± SEϖ P Ϫ ß ± SEϖ P Ϫ Model 1Ж 0.005 ± 0.004 0.21 0.014 ± 0.012 0.23 0.008 ± 0.007 0.31 Model 2 ╧ 0.004 ± 0.004 0.27 0.013 ± 0.011 0.27 0.008 ± 0.007 0.29 Model 3 ₸ 0.004 ± 0.004 0.33 0.014 ± 0.011 0.22 0.009 ± 0.008 0.26

Y = log (Interleukin-1ß Concentration, pg/mL) 25-OHD, nmol/L 25-OHD < 75 nmol/L 25-OHD ≥ 75 nmol/L All women (n = 170) (n = 92) (n = 78) ß ± SEϖ P Ϫ ß ± SEϖ P Ϫ ß ± SEϖ P Ϫ Model 1Ж -0.001 ± 0.003 0.72 -0.013 ± 0.010 0.19 -0.002 ± 0.006 0.76 Model 2 ┴ -0.002 ± 0.003 0.54 -0.013 ± 0.009 0.17 -0.002 ± 0.006 0.80 Model 3 ┬ -0.001 ± 0.003 0.67 -0.013 ± 0.009 0.15 -0.003 ± 0.006 0.64 ϖ Regression coefficients: Beta (ß) ± standard errors (SE). Beta coefficient represents change in inflammatory factor level per 1-nmol/L change in serum 25-OHD concentration. Ϫ P-values estimated using simple or multiple linear regressions. Ж Model 1: Unadjusted. ╧ Model 2: Adjusted for season and body mass index. ₸ Model 3: Adjusted for season and body mass index, multivitamin intake, alcohol intake and physical activity level. ┴ Model 2: Adjusted for season and waist circumference. ┬ Model 3: Adjusted for season, waist circumference, multivitamin intake, smoking status and physical activity level.

221

Table 37. Continued

Y = log (Interleukin-2 Concentration, pg/mL) 25-OHD, nmol/L 25-OHD < 75 nmol/L 25-OHD ≥ 75 nmol/L All women (n = 170) (n = 92) (n = 78) ß ± SEϖ P Ϫ ß ± SEϖ P Ϫ ß ± SEϖ P Ϫ Model 1Ж -0.003 ± 0.003 0.34 -0.021 ± 0.010 0.05* -0.006 ± 0.006 0.33 Model 2 ╧ -0.004 ± 0.003 0.20 -0.020 ± 0.010 0.05 -0.006 ± 0.006 0.35 Model 3 ₸ -0.003 ± 0.003 0.32 -0.021± 0.010 0.04 -0.006 ± 0.006 0.36

Y = log (Interleukin-4 Concentration, pg/mL) Model 1Ж -0.002 ± 0.003 0.54 -0.005 ± 0.010 0.60 -0.003 ± 0.005 0.61 Model 2 ╧ -0.002 ± 0.003 0.38 -0.005 ± 0.010 0.65 -0.002 ± 0.005 0.64 Model 3 ₸ -0.002 ± 0.003 0.49 -0.004 ± 0.011 0.69 -0.003 ± 0.005 0.49 ϖ Regression coefficients: Beta (ß) ± standard errors (SE). Beta coefficient represents change in inflammatory factor level per 1-nmol/L change in serum 25-OHD concentration. Ϫ P-values estimated using simple or multiple linear regressions. * Esitmated P-value is borderline significant (P = 0.049). Ж Model 1: Unadjusted. ╧ Model 2: Adjusted for season and total body fat percentage. ₸ Model 3: Adjusted for season and total body fat percentage, multivitamin intake, alcohol intake and physical activity level.

222

Table 37. Continued

Y = log (Interleukin-5 Concentration, pg/mL) 25-OHD, nmol/L 25-OHD < 75 nmol/L 25-OHD ≥ 75 nmol/L All women (n = 170) (n = 92) (n = 78) ß ± SEϖ P Ϫ ß ± SEϖ P Ϫ ß ± SEϖ P Ϫ Model 1Ж 0.001 ± 0.002 0.58 0.006 ± 0.006 0.32 -0.003 ± 0.004 0.45 Model 2 ╧ 0.001 ± 0.002 0.64 0.006 ± 0.006 0.32 -0.003 ± 0.004 0.46 Model 3 ₸ 0.001 ± 0.002 0.80 0.006 ± 0.006 0.33 -0.004 ± 0.004 0.34

Y = log (Interleukin-6 Concentration, pg/mL) Model 1Ж -0.003 ± 0.003 0.33 -0.020 ± 0.010 0.05 -0.013 ± 0.007 0.85 Model 2 ┴ -0.004 ± 0.003 0.23 -0.020 ± 0.010 0.05* -0.003 ± 0.007 0.69 Model 3 ┬ -0.003 ± 0.003 0.41 -0.020 ± 0.010 0.05 -0.003 ± 0.007 0.68 ϖ Regression coefficients: Beta (ß) ± standard errors (SE). Beta coefficient represents change in inflammatory factor level per 1-nmol/L change in serum 25-OHD concentration. Ϫ P-values estimated using simple or multiple linear regressions. * Esitmated P-value is borderline significant (P = 0.048). Ж Model 1: Unadjusted. ╧ Model 2: Adjusted for season and waist circumference. ₸ Model 3: Adjusted for season, waist circumference, multivitamin intake, smoking status and physical activity level. ┴ Model 2: Adjusted for season and waist circumference. body mass index ┬ Model 3: Adjusted for season, waist circumference, multivitamin intake, oral contraceptive use and physical activity level.

223

Table 37. Continued

Y = log (Interleukin-7 Concentration, pg/mL) 25-OHD, nmol/L 25-OHD < 75 nmol/L 25-OHD ≥ 75 nmol/L All women (n = 170) (n = 92) (n = 78) ß ± SEϖ P Ϫ ß ± SEϖ P Ϫ ß ± SEϖ P Ϫ Model 1Ж -0.001 ± 0.003 0.85 -0.019 ± 0.010 0.06 0.002 ± 0.006 0.81 Model 2 ╧ -0.001 ± 0.003 0.76 -0.019 ± 0.010 0.06 0.002 ± 0.007 0.80 Model 3 ₸ -0.001 ± 0.003 0.87 -0.020 ± 0.010 0.05 0.002 ± 0.007 0.77

Y = log (Interleukin-8 Concentration, pg/mL) Model 1Ж 0.005 ± 0.002 0.80 -0.005 ± 0.007 0.45 0.005 ± 0.003 0.15 Model 2 ┴ 0.0001 ± 0.002 0.94 -0.005 ± 0.007 0.45 0.004 ± 0.003 0.16 Model 3 ┬ 0.001 ± 0.002 0.60 -0.004 ± 0.006 0.49 0.004 ± 0.003 0.22 ϖ Regression coefficients: Beta (ß) ± standard errors (SE). Beta coefficient represents change in inflammatory factor level per 1- nmol/L change in serum 25-OHD concentration. Ϫ P-values estimated using simple or multiple linear regressions. Ж Model 1: Unadjusted. ╧ Model 2: Adjusted for season and waist circumference. ₸ Model 3: Adjusted for season and waist circumference, multivitamin intake, oral contraceptive use and physical activity level. ┴ Model 2: Adjusted for season and waist circumference. ┬ Model 3: Adjusted for season, waist circumference, multivitamin intake, smoking status and physical activity level.

224

Table 37. Continued

Y = log (Interleukin-10 Concentration, pg/mL) 25-OHD, nmol/L 25-OHD < 75 nmol/L 25-OHD ≥ 75 nmol/L All women (n = 170) (n = 92) (n = 78) ß ± SEϖ P Ϫ ß ± SEϖ P Ϫ ß ± SEϖ P Ϫ Model 1Ж 0.003 ± 0.003 0.29 -0.002 ± 0.008 0.80 0.002 ± 0.005 0.72 Model 2 ╧ 0.002 ± 0.003 0.36 -0.002 ± 0.008 0.81 0.002 ± 0.005 0.78 Model 3 ₸ 0.003 ± 0.003 0.31 -0.002 ± 0.008 0.85 0.002 ± 0.006 0.75

Y = log (Interleukin-12 Concentration, pg/mL) Model 1Ж -0.002 ± 0.005 0.67 -0.019 ± 0.015 0.21 -0.009 ± 0.010 0.36 Model 2 ╧ -0.003 ± 0.005 0.55 -0.018 ± 0.015 0.21 -0.011 ± 0.010 0.29 Model 3 ₸ -0.003 ± 0.005 0.58 -0.019 ± 0.015 0.21 -0.012 ± 0.010 0.25 ϖ Regression coefficients: Beta (ß) ± standard errors (SE). Beta coefficient represents change in inflammatory factor level per 1-nmol/L change in serum 25-OHD concentration. Ϫ P-values estimated using simple or multiple linear regressions. Ж Model 1: Unadjusted. ╧ Model 2: Adjusted for season and body mass index. ₸ Model 3: Adjusted for season and body mass index, multivitamin intake, oral contraceptive use and physical activity level.

225

Table 37. Continued

Y = log (Interleukin-13 Concentration, pg/mL) 25-OHD, nmol/L 25-OHD < 75 nmol/L 25-OHD ≥ 75 nmol/L All women (n = 170) (n = 92) (n = 78) ß ± SEϖ P Ϫ ß ± SEϖ P Ϫ ß ± SEϖ P Ϫ Model 1Ж -0.002 ± 0.004 0.65 -0.007 ± 0.011 0.53 -0.002 ± 0.007 0.83 Model 2 ╧ -0.002 ± 0.004 0.58 -0.005 ± 0.011 0.55 -0.002 ± 0.007 0.82 Model 3 ₸ -0.001 ± 0.004 0.84 -0.006 ± 0.012 0.63 -0.002 ± 0.008 0.77

Y = log (GM-CSF Concentration, pg/mL) Model 1Ж -0.004 ± 0.004 0.26 -0.021 ± 0.010 0.06 -0.003 ± 0.007 0.66 Model 2 ┴ -0.005 ± 0.004 0.17 -0.022 ± 0.011 0.04 -0.003 ± 0.007 0.70 Model 3 ┬ -0.004 ± 0.004 0.27 -0.022 ± 0.011 0.04 -0.003 ± 0.007 0.65 ϖ Regression coefficients: Beta (ß) ± standard errors (SE). Beta coefficient represents change in inflammatory factor level per 1-nmol/L change in serum 25-OHD concentration. Ϫ P-values estimated using simple or multiple linear regressions. Ж Model 1: Unadjusted. ╧ Model 2: Adjusted for season and total body fat percentage. ₸ Model 3: Adjusted for season, total body fat percentage, multivitamin intake, alcohol intake and physical activity level. ┴ Model 2: Adjusted for season and waist to height ratio. ┬ Model 3: Adjusted for season, waist to height ratio, multivitamin intake, alcohol intake and physical activity level.

226

Table 36. Continued

Y = log (TNF-α Concentration, pg/mL) 25-OHD, nmol/L 25-OHD < 75 nmol/L 25-OHD ≥ 75 nmol/L All women (n = 170) (n = 92) (n = 78)

ß ± SEϖ P Ϫ ß ± SEϖ P Ϫ ß ± SEϖ P Ϫ Ж Model 1 -0.0004 ± 0.002 0.97 -0.009 ± 0.005 0.09 -0.001 ± 0.003 0.77 Model 2 ╧ -0.0004 ± 0.002 0.79 -0.009 ± 0.005 0.10 -0.001 ± 0.003 0.82

Model 3 ₸ -0.0005 ± 0.002 0.77 -0.008 ± 0.005 0.10 -0.001 ± 0.003 0.81

Y = log (IFN-γ Concentration, pg/mL) Model 1Ж -0.002 ± 0.004 0.63 -0.018 ± 0.012 0.13 -0.002 ± 0.007 0.75 Model 2 ╧ -0.002 ± 0.004 0.58 -0.019 ± 0.012 0.14 -0.002 ± 0.007 0.79 Model 3 ₸ -0.003 ± 0.004 0.47 -0.018 ± 0.012 0.15 -0.002 ± 0.007 0.79 ϖ Regression coefficients: Beta (ß) ± standard errors (SE). Beta coefficient represents change in inflammatory factor level per 1-nmol/L change in serum 25-OHD concentration. Ϫ P-values estimated using simple or multiple linear regressions. Ж Model 1: Unadjusted. ╧ Model 2: Adjusted for season and waist circumference. ₸ Model 3: Adjusted for season, waist circumference, multivitamin intake, oral contraceptive use and physical activity level.

227

CHAPTER 6

SUMMARY, CONCLUSIONS AND FUTURE DIRECTIONS

6.1 Summary and Conclusions

We conducted a cross-sectional analysis among two hundred and seventy 18- to 30-year old female participants in the UMass Amherst Vitamin D Status Study (n = 270) to assess the extent to which dietary intakes of calcium and vitamin D are associated with obesity markers. We also evaluated the association between serum 25-OHD concentrations and both adiposity and inflammatory biomarkers. To our knowledge, ours is the first study to examine these associations in healthy young women using multiple measures of adiposity.

Study participants were mostly Caucasian women (84.5%) with a normal BMI, although about half of women had high adiposity (total body fat ‘TBF’ ≥ 32%). Most women (83.3%) had total vitamin D intakes less than the current RDA of 600 IU/day

(FNB/IOM, 2010), and about 40% of women had inadequate total calcium intakes (<

RDA of 1000 mg/day). These findings confirm that large proportions of young women likely have inadequate intakes of both calcium and vitamin D.

In multivariable analyses, women reporting adequate intakes of calcium (≥ 1000 mg/day) but low intakes of vitamin D (< 600 IU/day) were more than twice as likely to have a high percentage of TBF compared to women with adequate intakes of both calcium and vitamin D. The odds of being overweight/obese (BMI ≥ 25 kg/m2) also increased in a linear manner across tertiles of decreasing supplemental calcium intake

(Ptrend = 0.02). In addition, multivariable logistic regression analyses showed that women with lower calcium intake from supplements were twice as likely to have a waist circumference ≥ 80 cm (OR = 2.04; 95% CI: 1.04 – 3.99) compared to women in the

228

highest tertile of calcium intake. The magnitude of this association is important since among young women 18-30 years old, a waist circumference greater than 80 cm indicates central obesity and suggests increased visceral adiposity, which contributes to hyperlipidemia and other obesity-related chronic conditions (Zhu et al., 2002; Shields et al., 2012; Clark et al., 2012). Additional studies are needed to better characterize the association between supplemental calcium and central adiposity and to determine whether calcium supplementation can influence subsequent disease risk.

Among all women, total vitamin D, food vitamin D, and supplemental vitamin D intake were not associated with serum 25-OHD concentration (P > 0.05). However, among supplement users only, intake of vitamin D supplements was positively correlated with serum 25-OHD levels (ß = 0.03 ± 0.01 nmol/L, P = 0.05). Vitamin D insufficiency

(25-OHD < 75 nmol/L) was common (53.4%), but serum 25-OHD was not significantly correlated with adiposity measures. These findings support the notion that serum levels of

25-OHD are influenced by other factors besides the vitamin D content of foods, including the use of vitamin D supplements. They also suggest that, contrary to reports in obese populations, 25-OHD levels are not associated with adiposity among women with relatively normal body composition.

Serum 25-OHD concentration tended to be correlated with hs-CRP levels (r =

0.14, P = 0.06), but was not significantly associated with other inflammatory biomarkers.

Mean 25-OHD levels were greater among women with high inflammation (hs-CRP ≥ 5 mg/L) as compared to women with low inflammation. We also found significant associations between markers of obesity and inflammatory factors. Serum levels of IL-

1β, IL-7, GM-CSF, and TNF-α were higher in women with high waist circumference

229

(WC < 80 cm, P < 0.01). In multivariable analyses, we found that among women with low 25-OHD (< 75 nmol/L), serum 25-OHD level was inversely associated with IL-2 and

GM-CSF concentrations, and marginally associated with IL-6 and IL-7 concentrations.

Among these women, an increase of 1 nmol/L in serum 25-OHD concentration would be associated with a 2% reduction in IL-2, IL-6, IL-7 and GM-CSF concentrations.

Additional prospective studies in more heterogeneous populations will help to characterize the relationship between vitamin D status, inflammation and obesity.

A major strength of our study is the use of dual energy x-ray absorptiometry

(DXA) to objectively quantify adiposity. We also assessed dietary intake using a comprehensive and validated food frequency questionnaire that captured the frequency of dietary intake in the previous two months. In addition to dietary vitamin D intake, we assessed serum vitamin D status with a biomarker (25-OHD). Assays were also available for several inflammatory biomarkers. We controlled for several factors known to be associated with adiposity, serum 25-OHD levels, and inflammation; however, it is still possible that there may be residual confounding by other unknown factors.

Some limitations of our study include the homogeneity of our population and the cross-sectional design of our study. Our population comprised mostly of Caucasian women who were non-obese and between 18 to 30 years. Our findings are therefore limited to such populations and may not be generalizable to women in other age groups,

BMI categories, or ethnicities. In addition, because all exposures and outcome variables were assessed at one time-point, temporal and/or causal associations cannot be inferred from our findings.

230

To the best of our knowledge, this research is among the first to assess associations between vitamin D status, adiposity and inflammation in a general population of pre-dominantly non-obese physically active young women. Since health risks tend to rise with increases in adiposity, identifying modifiable dietary and lifestyle factors could significantly impact adiposity in young women. Information pertaining to associations between vitamin D status and inflammatory factors in young women could also be vital in identifying their susceptibility to future life-altering chronic conditions.

6.2 Future Directions

Obesity is a growing public health concern, and the possibility that central obesity risk in young women may be influenced by a high intake of calcium, especially calcium obtained from dietary supplements, warrants additional study. Our study design was cross-sectional, which means that we cannot make any causal inferences on the nature of association between calcium/vitamin D intakes and adiposity. Prospective studies with dietary intakes and adiposity assessed at several time points are needed to establish causal and/or temporal relations. For instance, we will be able to tell whether high waist circumference somehow influences calcium supplement intake or whether the reverse is true.

In our study, serum 25-OHD levels were not associated with adiposity measures in young women (ages 18-30 years). We recommend that additional studies be done to corroborate our findings. We also advise that future studies consider using a prospective study design as this will help to determine whether low serum 25-OHD concentration is a consequence of high adiposity or if it is in some way involved in its development. Larger

231

sample sizes may also guarantee a wider range of adiposity, and this will help to determine whether serum 25-OHD status differs among women with relatively little adiposity.

We found that IL-2, GM-CSF, IL-6, and IL-7 levels may be reduced in young women with low levels of 25-OHD (< 75 nmol/L). However, information on exposure and outcome were assessed simultaneously, making it impossible for us to determine whether 25-OHD status causes changes in inflammation, or whether inflammation influences 25-OHD levels. Prospective studies are necessary to investigate whether low

25-OHD status plays a causal role in the development of inflammation in young women.

Future research should also consider enrolling larger cohorts as this will help improve the precision and power of analyses.

232

APPENDIX A

PERMISSION TO ACCESS DATA

233

234

APPENDIX B NORMALITY AND DISTRIBUTION OF COVARIATES

Distribution of X=Total Calcium Intake, mg/day 60

58 60

40 36

Frequency 22

20 16 12

4 3

1 2 1 1 1 0 0 1000 2000 3000 4000 5000 calc

calc

Percentiles Smallest 1% 311.22 224.27 5% 467.69 299.37 10% 587.77 311.22 Obs 270 25% 776.76 316.41 Sum of Wgt. 270

50% 1142.665 Mean 1220.322 Largest Std. Dev. 615.7101 75% 1503.56 3440.94 90% 1982.785 3552.6 Variance 379098.9 95% 2210.32 4129.2 Skewness 1.731049 99% 3552.6 4600.61 Kurtosis 8.428664

235

Distribution of X=Food Calcium Intake, mg/day

77 80

60 56

52 40

Frequency 31

27 20

10 7 3 3

1 1 1 1 0 0 1000 2000 3000 4000 5000 calc_wo

calc_wo

Percentiles Smallest 1% 224.27 179.7 5% 444.57 214.02 10% 517.985 224.27 Obs 270 25% 714.97 299.37 Sum of Wgt. 270

50% 979.645 Mean 1090.554 Largest Std. Dev. 579.4937 75% 1333.82 3440.94 90% 1673.015 3552.6 Variance 335812.9 95% 2087.95 4111.35 Skewness 2.209983 99% 3552.6 4600.61 Kurtosis 11.55174

236

Distribution of X=Supplemental Calcium Intake, mg/day

190

200

150

100

Frequency 50

17 14 14 15 17 5

2 4 3 1 2 1 2 0 0 500 1000 1500 calc_suppl

calc_suppl

Percentiles Smallest 1% -21.70238 -54.32323 5% 0 -33.36703 10% 0 -21.70238 Obs 287 25% 0 0 Sum of Wgt. 287

50% 0 Mean 132.9097 Largest Std. Dev. 245.2884 75% 200 1064.28 90% 499.95 1200 Variance 60166.39 95% 600.08 1257.15 Skewness 2.211212 99% 1200 1334 Kurtosis 8.223289

237

Distribution of X=Total Vitamin D Intake, IU/day 52 52

50 47 40

29 30

Frequency 20

10 7 4 4 2

1 1 1 1 0 0 500 1000 1500 2000 vitd

Percentiles Smallest 1% 13.6 3.37 5% 43.45 4.16 10% 67.11 13.6 Obs 270 25% 163.15 17 Sum of Wgt. 270

50% 299.45 Mean 375.4516 Largest Std. Dev. 298.4194 75% 533.64 1292.84 90% 828.695 1445.52 Variance 89054.15 95% 950.44 1604.25 Skewness 1.465864 99% 1445.52 1806.07 Kurtosis 5.881516

238

Distribution of X=Food Vitamin D Intake, IU/day

60 56

44 41

39 40

27 Frequency

20 20 14 10 6 4

2 2 1 2 2 0 0 200 400 600 800 1000 vitd_wo

vitd_wo

Percentiles Smallest 1% 4.16 3.37 5% 25.61 4.01 10% 45.51 4.16 Obs 270 25% 110.94 7.47 Sum of Wgt. 270

50% 188.025 Mean 230.188 Largest Std. Dev. 180.639 75% 306.55 859.46 90% 443.72 872.65 Variance 32630.43 95% 555.79 963.67 Skewness 1.625135 99% 872.65 1006.2 Kurtosis 6.341862

239

Distribution of X=Supplemental Vitamin D Intake, IU/day 200

168

150

100 Frequency

50 37 30 25

9 5 5

3 2 1 2 0 0 500 1000 1500 vitd_suppl

vitd_suppl

Percentiles Smallest 1% 0 -35.77971 5% 0 -1.338881 10% 0 0 Obs 287 25% 0 0 Sum of Wgt. 287

50% 0 Mean 147.7056 Largest Std. Dev. 243.6279 75% 228.57 807.14 90% 400 1300 Variance 59354.58 95% 660 1400 Skewness 2.20605 99% 1300 1400 Kurtosis 9.049925

240

Distribution of Y=Serum 25-OHD, nmol/L

40 38

30 30 26

20 17

Frequency 15 13 11

10 10

5 4 3

2 0 0 50 100 150 200 serum_25d

serum_25d

Percentiles Smallest 1% 22 21.7 5% 37.3 22 10% 45.7 26.3 Obs 174 25% 55.4 26.9 Sum of Wgt. 174

50% 72.7 Mean 79.24368 Largest Std. Dev. 32.29236 75% 96.9 161.2 90% 126.4 162.5 Variance 1042.797 95% 150.7 164.6 Skewness .788026 99% 164.6 174.3 Kurtosis 3.189464

241

APPENDIX C LINEARITY AND SPLINE MODEL FIT FOR DIETARY INTAKES *Y =TOTAL BODY FAT; X = Total Calcium Intakes * Lowess Regression Curve and Linear Prediction Fit for the Association between Total Body Fat Percentage and Total Calcium Intake

Lowess smoother

TotalBody (%) Fat

20 10 0 1000 2000 3000 4000 5000 Total Calcium Intake (mg/day)

total_bf_percent lowess: total_bf_percent Fitted values

bandwidth = .3

Lowess smoother

TotalBody (%) Fat

20 10 0 1000 2000 3000 4000 5000 Total Calcium Intake (mg/day)

total_bf_percent lowess: total_bf_percent 95% CI Fitted values

bandwidth = .3

* Lacking specific knots – Define knots using evenly spaced percentiles Binom. Interp. Variable Obs Percentile Centile [95% Conf. Interval]

calc 270 0 224.27 224.27 224.27* 25 776.58 719.6616 851.1368 50 1142.665 1051.204 1221.064 75 1504.128 1393.055 1601.13 100 4600.61 4600.61 4600.61*

* Lower (upper) confidence limit held at minimum (maximum) of sample

242

* Note: Above suggests knots at p25=224.27, p50=776.58 and p75=1504.13

* LINEAR MODEL – Fit

Source SS df MS Number of obs = 270 F( 1, 268) = 0.68 Model 45.7150077 1 45.7150077 Prob > F = 0.4088 Residual 17901.1601 268 66.7953733 R-squared = 0.0025 Adj R-squared = -0.0012 Total 17946.8751 269 66.7170077 Root MSE = 8.1728

total_bf_p~t Coef. Std. Err. t P>|t| [95% Conf. Interval]

calc -.0006695 .0008093 -0.83 0.409 -.002263 .0009239 _cons 32.90039 1.105806 29.75 0.000 30.72322 35.07756

* LINEAR MODEL – Diagnostics

Shapiro-Wilk W test for normal data

Variable Obs W V z Prob>z

ehat_linea~c 270 0.99155 1.639 1.154 0.12417

* Normality of residuals check is not significant (null not rejected)*

* Linear Model Fit – Plot of ("Normality of Residuals")

Linear Model

Normality of Residuals

1.00

0.75

0.50

0.25

Normal F[(ehat_linear_calc-m)/s] Normal 0.00 0.00 0.25 0.50 0.75 1.00 Empirical P[i] = i/(N+1)

243

* Linear Model Fit – Plot of ("Jacknife Residuals versus Predicted")

Linear Model

Jacknife Residuals v Predicted

Studentized residuals Studentized

-1 -2

30 31 32 33 Linear prediction

* Linear Model Fit – Plot of ("Cook Distances")

Linear Model

Cook Distances

Cook's D Cook's

.2 0

0 100 200 300 subj_id

* Note: Linear prediction looks good. Biggest cook’s distance is less than 1

244

*RESTRICTED CUBIC SPLINES – Fit

Source SS df MS Number of obs = 270 F( 5, 264) = 1.44 Model 477.563292 5 95.5126585 Prob > F = 0.2090 Residual 17469.3118 264 66.1716355 R-squared = 0.0266 Adj R-squared = 0.0082 Total 17946.8751 269 66.7170077 Root MSE = 8.1346

total_bf_p~t Coef. Std. Err. t P>|t| [95% Conf. Interval]

calc 0 (omitted) calc2 -.0006498 .0002607 -2.49 0.013 -.0011632 -.0001365 calc3 1.29e-06 5.14e-07 2.50 0.013 2.74e-07 2.30e-06 knot_calcp25 -1.43e-06 5.70e-07 -2.51 0.013 -2.55e-06 -3.07e-07 knot_calcp50 1.55e-07 6.17e-08 2.51 0.013 3.36e-08 2.76e-07 knot_calcp75 -1.27e-08 7.42e-09 -1.72 0.087 -2.74e-08 1.87e-09 _cons 62.38243 12.60157 4.95 0.000 37.57006 87.19479

* RESTRICTED CUBIC SPLINES – Diagnostics

Shapiro-Wilk W test for normal data

Variable Obs W V z Prob>z

ehat_splin~c 270 0.99360 1.242 0.507 0.30614

* Normality of residuals check is not significant (null not rejected)

* Restricted Cubic Splines Fit - Plot of ("Normality of Residuals")

Restricted Cubic Splines Model

Normality of Residuals

1.00

0.75

0.50

0.25

Normal F[(ehat_spline_calc-m)/s] Normal 0.00 0.00 0.25 0.50 0.75 1.00 Empirical P[i] = i/(N+1)

245

* Restricted Cubic Splines Fit - Plot of ("Jacknife Residuals versus Predicted")

Restricted Cubic Splines Model

Jacknife Residuals v Predicted

Studentized residuals Studentized

-1 -2

25 30 35 40 45 Linear prediction

* Restricted Cubic Splines Fit - Plot of ("Cook Distances")

Restricted Cubic Splines Model

Cook Distances

Cook's D Cook's

.2 0

0 100 200 300 subj_id

* Note: Spline prediction looks good. Biggest cook’s distance is still less than 1

246

* Compare Linear (1) and Splines (2) side-by-side

(1) (2) total_bf_p~t total_bf_p~t calc -0.000670 (0.000809) o.calc 0 (.) calc2 -0.000650* (0.000261) calc3 0.00000129* (0.000000514) knot_calcp25 -0.00000143* (0.000000570) knot_calcp50 0.000000155* (6.17e-08) knot_calcp75 -1.27e-08 (7.42e-09)

Constant 32.90*** 62.38*** (1.106) (12.60)

Observations 270 270 R-squared 0.003 0.027 Adjusted R-squared -0.001 0.008 rmse 8.173 8.135

Standard errors in parentheses * p<0.05, ** p<0.01, *** p<0.001

* Partial F-test of Departure from Linearity*

( 1) calc2 = 0 ( 2) calc3 = 0 ( 3) knot_calcp25 = 0 ( 4) knot_calcp50 = 0 ( 5) knot_calcp75 = 0 Constraint 2 dropped Constraint 4 dropped

F( 3, 264) = 2.13 Prob > F = 0.0967 * Note: Departure from linearity is marginally significant. Proceed with the linear model

247

*Y =TOTAL BODY FAT; X = Food Calcium Intake

/* Lowess Regression Curve and Linear Prediction Fit for the Association between Total Body Fat Percentage and Food Calcium Intake

Lowess smoother

TotalBody (%) Fat

20 10 0 1000 2000 3000 4000 5000 Food Calcium Intake (mg/day)

total_bf_percent lowess: total_bf_percent Fitted values

bandwidth = .3

Lowess smoother

TotalBody (%) Fat

20 10 0 1000 2000 3000 4000 5000 Food Calcium Intake (mg/day)

total_bf_percent lowess: total_bf_percent 95% CI Fitted values

bandwidth = .3

* Lacking specific knots – Define knots using evenly spaced percentiles Binom. Interp. Variable Obs Percentile Centile [95% Conf. Interval]

calc_wo 270 0 179.7 179.7 179.7* 25 714.8375 665.6529 768.2664 50 979.645 891.4725 1063.545 75 1335.74 1282.706 1410.184 100 4600.61 4600.61 4600.61*

* Lower (upper) confidence limit held at minimum (maximum) of sample

248

* Note: Above suggests knots at p25=714.84, p50=979.65 and p75=1335.74

* LINEAR MODEL – Fit Source SS df MS Number of obs = 270 F( 1, 268) = 0.52 Model 34.9601664 1 34.9601664 Prob > F = 0.4702 Residual 17911.9149 268 66.8355033 R-squared = 0.0019 Adj R-squared = -0.0018 Total 17946.8751 269 66.7170077 Root MSE = 8.1753

total_bf_p~t Coef. Std. Err. t P>|t| [95% Conf. Interval]

calc_wo -.0006221 .0008602 -0.72 0.470 -.0023156 .0010714 _cons 32.76177 1.061827 30.85 0.000 30.67119 34.85235

* LINEAR MODEL – Diagnostics

Shapiro-Wilk W test for normal data

Variable Obs W V z Prob>z

ehat_linea~o 270 0.99178 1.596 1.091 0.13755 * Normality of residuals check is NOT significant (null not rejected)*

* Linear Model Fit – Plot of ("Normality of Residuals")

Linear Model

Normality of Residuals

1.00

0.75

0.50

0.25

Normal F[(ehat_linear_calc_wo-m)/s] Normal 0.00 0.00 0.25 0.50 0.75 1.00 Empirical P[i] = i/(N+1)

249

* Linear Model Fit – Plot of ("Jacknife Residuals versus Predicted")

Linear Model

Jacknife Residuals v Predicted

Studentized residuals Studentized

-1 -2

30 30.5 31 31.5 32 32.5 Linear prediction

* Linear Model Fit – Plot of ("Cook Distances")

Linear Model

Cook Distances

Cook's D Cook's

.2 0

0 100 200 300 subj_id

* Note: Linear prediction looks good. Biggest cook’s distance is less than 1

250

*RESTRICTED CUBIC SPLINES – Fit

Source SS df MS Number of obs = 270 F( 6, 263) = 1.29 Model 512.824331 6 85.4707219 Prob > F = 0.2623 Residual 17434.0507 263 66.2891663 R-squared = 0.0286 Adj R-squared = 0.0064 Total 17946.8751 269 66.7170077 Root MSE = 8.1418

total_bf_perc~t Coef. Std. Err. t P>|t| [95% Conf. Interval]

calc_wo -.2010608 .0996081 -2.02 0.045 -.3971916 -.00493 calc_wo2 .0003268 .000178 1.84 0.068 -.0000237 .0006773 calc_wo3 -1.68e-07 1.00e-07 -1.68 0.095 -3.65e-07 2.94e-08 knot_calc_wop25 1.97e-07 1.70e-07 1.16 0.248 -1.38e-07 5.31e-07 knot_calc_wop50 -1.48e-08 9.90e-08 -0.15 0.881 -2.10e-07 1.80e-07 knot_calc_wop75 -1.56e-08 2.54e-08 -0.61 0.540 -6.55e-08 3.44e-08 _cons 70.31099 17.43537 4.03 0.000 35.98032 104.6417

* RESTRICTED CUBIC SPLINES – Diagnostics

Shapiro-Wilk W test for normal data

Variable Obs W V z Prob>z

ehat_splin~o 270 0.99172 1.606 1.107 0.13415

* Normality of residuals check is NOT significant (null not rejected)

* Restricted Cubic Splines Fit - Plot of ("Normality of Residuals")

Restricted Cubic Splines Model

Normality of Residuals

1.00

0.75

0.50

0.25

Normal F[(ehat_spline_calc_wo-m)/s] Normal 0.00 0.00 0.25 0.50 0.75 1.00 Empirical P[i] = i/(N+1)

251

* Restricted Cubic Splines Fit - Plot of ("Jacknife Residuals versus Predicted")

Restricted Cubic Splines Model

Jacknife Residuals v Predicted

Studentized residuals Studentized

-1 -2

25 30 35 40 45 Linear prediction

* Restricted Cubic Splines Fit - Plot of ("Cook Distances")

Restricted Cubic Splines Model

Cook Distances

Cook's D Cook's

.2 0

0 100 200 300 subj_id

* Note: Spline prediction looks good. Biggest cook’s distance is still less than 1

252

* Compare Linear (1) and Splines (2) side-by-side

(1) (2) total_bf_p~t total_bf_p~t

calc_wo -0.000622 -0.201* (0.000860) (0.0996)

calc_wo2 0.000327 (0.000178)

calc_wo3 -0.000000168 (0.000000100)

knot_calc_wop25 0.000000197 (0.000000170)

knot_calc_wop50 -1.48e-08 (9.90e-08)

knot_calc_wop75 -1.56e-08 (2.54e-08)

Constant 32.76*** 70.31*** (1.062) (17.44)

Observations 270 270 R-squared 0.002 0.029 Adjusted R-squared -0.002 0.006 rmse 8.175 8.142

Standard errors in parentheses * p<0.05, ** p<0.01, *** p<0.001

* Partial F-test of Departure from Linearity

( 1) calc_wo = 0 ( 2) calc_wo2 = 0 ( 3) calc_wo3 = 0 ( 4) knot_calc_wop25 = 0 ( 5) knot_calc_wop50 = 0 ( 6) knot_calc_wop75 = 0 Constraint 3 dropped Constraint 4 dropped

F( 4, 263) = 1.84 Prob > F = 0.1223 * Note: Departure from linearity is NOT significant. Proceed with the linear model

253

*Y =TOTAL BODY FAT; X = Supplemental Calcium Intake * Lowess Regression Curve and Linear Prediction Fit for the Association between Total Body Fat Percentage and Supplemental Calcium Intake

Lowess smoother

TotalBody (%) Fat

20 10 0 500 1000 1500 Supplement Calcium Intake (mg/day)

total_bf_percent lowess: total_bf_percent Fitted values

bandwidth = .3

Lowess smoother

TotalBody (%) Fat

20 10 0 500 1000 1500 Supplement Calcium Intake (mg/day)

total_bf_percent lowess: total_bf_percent 95% CI Fitted values

bandwidth = .3

* Lacking specific knots – Define knots using evenly spaced percentiles Binom. Interp. Variable Obs Percentile Centile [95% Conf. Interval]

calc_supplr 93 0 11.43 11.43 11.43* 25 162 90.69199 200 50 362 257.14 450 75 500 499.9483 600.0162 100 1334 1334 1334*

* Lower (upper) confidence limit held at minimum (maximum) of sample

* Note: Above suggests knots at p25=162, p50=362 and p75=500

254

* LINEAR MODEL – Fit

Source SS df MS Number of obs = 93 F( 1, 91) = 0.42 Model 33.0682854 1 33.0682854 Prob > F = 0.5200 Residual 7213.4048 91 79.2681846 R-squared = 0.0046 Adj R-squared = -0.0064 Total 7246.47309 92 78.7660118 Root MSE = 8.9033

total_bf_p~t Coef. Std. Err. t P>|t| [95% Conf. Interval]

calc_supplr -.002084 .0032266 -0.65 0.520 -.0084932 .0043252 _cons 32.97654 1.526441 21.60 0.000 29.94445 36.00862

* LINEAR MODEL – Diagnostics Shapiro-Wilk W test for normal data

Variable Obs W V z Prob>z

ehat_linea~r 93 0.97752 1.747 1.233 0.10883

* Normality of residuals check is NOT significant (null not rejected)

* Linear Model Fit – Plot of ("Normality of Residuals")

Linear Model

Normality of Residuals

1.00

0.75

0.50

0.25

Normal F[(ehat_linear_calc_supplr-m)/s] Normal 0.00 0.00 0.25 0.50 0.75 1.00 Empirical P[i] = i/(N+1)

255

* Linear Model Fit – Plot of ("Jacknife Residuals versus Predicted")

Linear Model

Jacknife Residuals v Predicted

Studentized residuals Studentized

-1 -2

30 31 32 33 Linear prediction

* Linear Model Fit – Plot of ("Cook Distances")

Linear Model

Cook Distances

Cook's D Cook's

.2 0

0 100 200 300 subj_id

* Note: Linear prediction looks good. Biggest cook’s distance is less than 1

256

*RESTRICTED CUBIC SPLINES – Fit

Source SS df MS Number of obs = 93 F( 6, 86) = 0.32 Model 156.942066 6 26.157011 Prob > F = 0.9263 Residual 7089.53102 86 82.4364072 R-squared = 0.0217 Adj R-squared = -0.0466 Total 7246.47309 92 78.7660118 Root MSE = 9.0794

total_bf_percent Coef. Std. Err. t P>|t| [95% Conf. Interval]

calc_supplr .1243309 .20635 0.60 0.548 -.2858794 .5345411 calc_supplr2 -.0009688 .0017825 -0.54 0.588 -.0045122 .0025746 calc_supplr3 2.03e-06 4.38e-06 0.46 0.644 -6.68e-06 .0000107 knot_calc_supplrp25 -1.82e-06 5.29e-06 -0.34 0.732 -.0000123 8.70e-06 knot_calc_supplrp50 -7.67e-07 1.91e-06 -0.40 0.688 -4.55e-06 3.02e-06 knot_calc_supplrp75 6.27e-07 9.50e-07 0.66 0.511 -1.26e-06 2.51e-06 _cons 30.14036 5.712494 5.28 0.000 18.7843 41.49642

* RESTRICTED CUBIC SPLINES – Diagnostics Shapiro-Wilk W test for normal data

Variable Obs W V z Prob>z

ehat_splin~r 93 0.97590 1.873 1.386 0.08280

* Normality of residuals check is marginally significant (null not rejected)

* Restricted Cubic Splines Fit - Plot of ("Normality of Residuals")

Restricted Cubic Splines Model

Normality of Residuals

1.00

0.75

0.50

0.25

Normal F[(ehat_spline_calc_supplr-m)/s] Normal 0.00 0.00 0.25 0.50 0.75 1.00 Empirical P[i] = i/(N+1)

257

* Restricted Cubic Splines Fit - Plot of ("Jacknife Residuals versus Predicted")

Restricted Cubic Splines Model

Jacknife Residuals v Predicted

Studentized residuals Studentized

-1 -2

28 30 32 34 36 Linear prediction

* Restricted Cubic Splines Fit - Plot of ("Cook Distances")

Restricted Cubic Splines Model

Cook Distances

Cook's D Cook's

.2 0

0 100 200 300 subj_id

* Note: Generally looks good. One observation has cook’s distance is >1

258

* Compare Linear (1) and Splines (2) side-by-side

(1) (2) total_bf_p~t total_bf_p~t

calc_supplr -0.00208 0.124 (0.00323) (0.206)

calc_supplr2 -0.000969 (0.00178)

calc_supplr3 0.00000203 (0.00000438)

knot_calc_supplrp25 -0.00000182 (0.00000529)

knot_calc_supplrp50 -0.000000767 (0.00000191)

knot_calc_supplrp75 0.000000627 (0.000000950)

Constant 32.98*** 30.14*** (1.526) (5.712)

Observations 93 93 R-squared 0.005 0.022 Adjusted R-squared -0.006 -0.047 rmse 8.903 9.079

Standard errors in parentheses * p<0.05, ** p<0.01, *** p<0.001

/* Partial F-test of Departure from Linearity

( 1) calc_supplr = 0 ( 2) calc_supplr2 = 0 ( 3) calc_supplr3 = 0 ( 4) knot_calc_supplrp25 = 0 ( 5) knot_calc_supplrp50 = 0 ( 6) knot_calc_supplrp75 = 0 Constraint 3 dropped

F( 5, 86) = 0.35 Prob > F = 0.8789 * Note: Departure from linearity is NOT significant. Proceed with the linear model

259

*Y =TOTAL BODY FAT; X = Total Vitamin D Intake * Lowess Regression Curve and Linear Prediction Fit for the Association between Total Body Fat Percentage and Total Vitamin D Intake

Lowess smoother

TotalBody (%) Fat

20 10 0 500 1000 1500 2000 Total Vitamin D Intake (IU/day)

total_bf_percent lowess: total_bf_percent Fitted values

bandwidth = .3

Lowess smoother

TotalBody (%) Fat

20 10 0 500 1000 1500 2000 Total Vitamin D Intake (IU/day)

total_bf_percent lowess: total_bf_percent 95% CI Fitted values

bandwidth = .3

* Lacking specific knots – Define knots using evenly spaced percentiles Binom. Interp. Variable Obs Percentile Centile [95% Conf. Interval]

vitd 270 0 3.37 3.37 3.37* 25 162.59 133.8724 188.4364 50 299.45 257.0065 340.6129 75 533.7875 488.6483 571.3427 100 1806.07 1806.07 1806.07*

* Lower (upper) confidence limit held at minimum (maximum) of sample

* Note: Above suggests knots at p25=162.59, p50=299.45 and p75=533.79

260

* LINEAR MODEL – Fit

Source SS df MS Number of obs = 270 F( 1, 268) = 2.29 Model 151.875213 1 151.875213 Prob > F = 0.1316 Residual 17794.9998 268 66.3992532 R-squared = 0.0085 Adj R-squared = 0.0048 Total 17946.8751 269 66.7170077 Root MSE = 8.1486

total_bf_p~t Coef. Std. Err. t P>|t| [95% Conf. Interval]

vitd -.0025179 .0016649 -1.51 0.132 -.0057958 .00076 _cons 33.02869 .7978986 41.39 0.000 31.45774 34.59963

/* LINEAR MODEL – Diagnostics

Shapiro-Wilk W test for normal data

Variable Obs W V z Prob>z

ehat_linea~d 270 0.99119 1.710 1.253 0.10516 * Normality of residuals check is NOT significant (null not rejected)

* Linear Model Fit – Plot of ("Normality of Residuals")

Linear Model

Normality of Residuals

1.00

0.75

0.50

0.25

Normal F[(ehat_linear_vitd-m)/s] Normal 0.00 0.00 0.25 0.50 0.75 1.00 Empirical P[i] = i/(N+1)

261

* Linear Model Fit – Plot of ("Jacknife Residuals versus Predicted")

Linear Model

Jacknife Residuals v Predicted

Studentized residuals Studentized

-1 -2

28 29 30 31 32 33 Linear prediction

* Linear Model Fit – Plot of ("Cook Distances")

Linear Model Cook Distances

Cook's D Cook's .4

0 100 200 300 subj_id

* Note: Linear prediction looks good. Biggest cook’s distance is less than 1

262

*RESTRICTED CUBIC SPLINES – Fit

Source SS df MS Number of obs = 270 F( 6, 263) = 1.13 Model 452.867931 6 75.4779886 Prob > F = 0.3424 Residual 17494.0071 263 66.5171374 R-squared = 0.0252 Adj R-squared = 0.0030 Total 17946.8751 269 66.7170077 Root MSE = 8.1558

total_bf_p~t Coef. Std. Err. t P>|t| [95% Conf. Interval]

vitd .2434721 .1247986 1.95 0.052 -.0022595 .4892037 vitd2 -.0019495 .0010471 -1.86 0.064 -.0040113 .0001123 vitd3 4.52e-06 2.57e-06 1.76 0.080 -5.38e-07 9.58e-06 knot_vitdp25 -5.17e-06 3.23e-06 -1.60 0.111 -.0000115 1.20e-06 knot_vitdp50 6.43e-07 8.92e-07 0.72 0.472 -1.11e-06 2.40e-06 knot_vitdp75 2.20e-09 1.67e-07 0.01 0.989 -3.26e-07 3.30e-07 _cons 25.42006 4.050045 6.28 0.000 17.44542 33.3947

* RESTRICTED CUBIC SPLINES – Diagnostics Shapiro-Wilk W test for normal data

Variable Obs W V z Prob>z

ehat_splin~d 270 0.99068 1.809 1.384 0.08321

* Normality of residuals check is marginally significant (null not rejected)

* Restricted Cubic Splines Fit - Plot of ("Normality of Residuals")

Restricted Cubic Splines Model

Normality of Residuals

1.00

0.75

0.50

0.25

Normal F[(ehat_spline_vitd-m)/s] Normal 0.00 0.00 0.25 0.50 0.75 1.00 Empirical P[i] = i/(N+1)

263

* Restricted Cubic Splines Fit - Plot of ("Jacknife Residuals versus Predicted")

Restricted Cubic Splines Model

Jacknife Residuals v Predicted

Studentized residuals Studentized

-1 -2

26 28 30 32 34 Linear prediction

* Restricted Cubic Splines Fit - Plot of ("Cook Distances")

Restricted Cubic Splines Model

Cook Distances

Cook's D Cook's

.2 0

0 100 200 300 subj_id

* Note: Spline prediction looks good. Biggest cook’s distance is exactly 1 for one observation

264

* Compare Linear (1) and Splines (2) side-by-side

(1) (2) total_bf_p~t total_bf_p~t vitd -0.00252 0.243 (0.00166) (0.125) vitd2 -0.00195 (0.00105) vitd3 0.00000452 (0.00000257) knot_vitdp25 -0.00000517 (0.00000323) knot_vitdp50 0.000000643 (0.000000892) knot_vitdp75 2.20e-09 (0.000000167)

Constant 33.03*** 25.42*** (0.798) (4.050)

Observations 270 270 R-squared 0.008 0.025 Adjusted R-squared 0.005 0.003 rmse 8.149 8.156

Standard errors in parentheses * p<0.05, ** p<0.01, *** p<0.001

* Partial F-test of Departure from Linearity

( 1) vitd = 0 ( 2) vitd2 = 0 ( 3) vitd3 = 0 ( 4) knot_vitdp25 = 0 ( 5) knot_vitdp50 = 0 ( 6) knot_vitdp75 = 0 Constraint 3 dropped

F( 5, 263) = 1.00 Prob > F = 0.4208 * Note: Departure from linearity is NOT significant. Proceed with the linear model

265

*Y =TOTAL BODY FAT; X = Food Vitamin D Intake * Lowess Regression Curve and Linear Prediction Fit for the Association between Total Body Fat Percentage and Food Vitamin D Intake

Lowess smoother

TotalBody (%) Fat

20 10 0 200 400 600 800 1000 Food Vitamin D Intake (IU/day)

total_bf_percent lowess: total_bf_percent Fitted values

bandwidth = .3

Lowess smoother

TotalBody (%) Fat

20 10 0 200 400 600 800 1000 Food Vitamin D Intake (IU/day)

total_bf_percent lowess: total_bf_percent 95% CI Fitted values

bandwidth = .3

* Define knots - Lacking specific knots – Define knots using evenly spaced percentiles

Binom. Interp. Variable Obs Percentile Centile [95% Conf. Interval]

vitd_wo 270 0 3.37 3.37 3.37* 25 110.4275 86.71537 128.0679 50 188.025 169.6337 204.1505 75 306.695 279.1118 341.1421 100 1006.2 1006.2 1006.2*

* Lower (upper) confidence limit held at minimum (maximum) of sample

* Note: Above suggests knots at p25=110.43, p50=188.03 and p75=306.70

266

* LINEAR MODEL – Fit

Source SS df MS Number of obs = 270 F( 1, 268) = 2.18 Model 144.910253 1 144.910253 Prob > F = 0.1408 Residual 17801.9648 268 66.4252418 R-squared = 0.0081 Adj R-squared = 0.0044 Total 17946.8751 269 66.7170077 Root MSE = 8.1502

total_bf_p~t Coef. Std. Err. t P>|t| [95% Conf. Interval]

vitd_wo -.0040631 .0027509 -1.48 0.141 -.0094793 .001353 _cons 33.01862 .8043631 41.05 0.000 31.43495 34.60229

* LINEAR MODEL – Diagnostics Shapiro-Wilk W test for normal data

Variable Obs W V z Prob>z

ehat_linea~o 270 0.99165 1.620 1.127 0.12987

* Normality of residuals check is NOT significant (null not rejected)

* Linear Model Fit – Plot of ("Normality of Residuals")

Linear Model

Normality of Residuals

1.00

0.75

0.50

0.25

Normal F[(ehat_linear_vitd_wo-m)/s] Normal 0.00 0.00 0.25 0.50 0.75 1.00 Empirical P[i] = i/(N+1)

267

* Linear Model Fit – Plot of ("Jacknife Residuals versus Predicted")

Linear Model

Jacknife Residuals v Predicted

Studentized residuals Studentized

-1 -2

29 30 31 32 33 Linear prediction

* Linear Model Fit – Plot of ("Jacknife Residuals versus Predicted")

Linear Model

Cook Distances

Cook's D Cook's

.2 0

0 100 200 300 subj_id

* Note: Linear prediction looks good. Biggest cook’s distance is less than 1

268

*RESTRICTED CUBIC SPLINES – Fit

Source SS df MS Number of obs = 270 F( 6, 263) = 0.44 Model 176.811614 6 29.4686023 Prob > F = 0.8544 Residual 17770.0634 263 67.5667812 R-squared = 0.0099 Adj R-squared = -0.0127 Total 17946.8751 269 66.7170077 Root MSE = 8.2199

total_bf_perc~t Coef. Std. Err. t P>|t| [95% Conf. Interval]

vitd_wo .0834572 .1829118 0.46 0.649 -.2767008 .4436151 vitd_wo2 -.0009759 .0023706 -0.41 0.681 -.0056438 .0036919 vitd_wo3 3.39e-06 8.92e-06 0.38 0.704 -.0000142 .0000209 knot_vitd_wop25 -4.28e-06 .0000124 -0.35 0.730 -.0000287 .0000201 knot_vitd_wop50 1.09e-06 4.77e-06 0.23 0.819 -8.29e-06 .0000105 knot_vitd_wop75 -2.19e-07 1.05e-06 -0.21 0.835 -2.28e-06 1.84e-06 _cons 30.73728 3.827989 8.03 0.000 23.19987 38.27468

* RESTRICTED CUBIC SPLINES – Diagnostics Shapiro-Wilk W test for normal data

Variable Obs W V z Prob>z

ehat_splin~o 270 0.99197 1.558 1.036 0.15006

* Normality of residuals check is NOT significant (null not rejected)

* Restricted Cubic Splines Fit - Plot of ("Normality of Residuals")

Restricted Cubic Splines Model

Normality of Residuals

1.00

0.75

0.50

0.25

Normal F[(ehat_spline_vitd_wo-m)/s] Normal 0.00 0.00 0.25 0.50 0.75 1.00 Empirical P[i] = i/(N+1)

269

* Restricted Cubic Splines Fit - Plot of ("Jacknife Residuals versus Predicted")

Restricted Cubic Splines Model

Jacknife Residuals v Predicted

Studentized residuals Studentized

-1 -2

28 29 30 31 32 33 Linear prediction

* Restricted Cubic Splines Fit - Plot of ("Cook Distances")

Restricted Cubic Splines Model

Cook Distances

Cook's D Cook's

.2 0

0 100 200 300 subj_id

* Note: Spline prediction looks good. Biggest cook’s distance is still less than 1

270

* Compare Linear (1) and Splines (2) side-by-side

(1) (2) total_bf_p~t total_bf_p~t

vitd_wo -0.00406 0.0835 (0.00275) (0.183)

vitd_wo2 -0.000976 (0.00237)

vitd_wo3 0.00000339 (0.00000892)

knot_vitd_wop25 -0.00000428 (0.0000124)

knot_vitd_wop50 0.00000109 (0.00000477)

knot_vitd_wop75 -0.000000219 (0.00000105)

Constant 33.02*** 30.74*** (0.804) (3.828)

Observations 270 270 R-squared 0.008 0.010 Adjusted R-squared 0.004 -0.013 rmse 8.150 8.220

Standard errors in parentheses * p<0.05, ** p<0.01, *** p<0.001

* Partial F-test of Departure from Linearity

( 1) vitd_wo = 0 ( 2) vitd_wo2 = 0 ( 3) vitd_wo3 = 0 ( 4) knot_vitd_wop25 = 0 ( 5) knot_vitd_wop50 = 0 ( 6) knot_vitd_wop75 = 0

F( 6, 263) = 0.44 Prob > F = 0.8544

* Note: Departure from linearity is NOT significant. Proceed with the linear model

271

*Y =TOTAL BODY FAT; X = Supplemental Vitamin D Intake * Lowess Regression Curve and Linear Prediction Fit for the Association between Total Body Fat Percentage and Supplemental Vitamin D Intake

Lowess smoother

TotalBody (%) Fat

20 10 0 500 1000 1500 Supplement Vitamin D Intake (IU/day)

total_bf_percent lowess: total_bf_percent Fitted values

bandwidth = .3

Lowess smoother

TotalBody (%) Fat

20 10 0 500 1000 1500 Supplement Vitamin D Intake (IU/day)

total_bf_percent lowess: total_bf_percent 95% CI Fitted values

bandwidth = .3

* Define knots - Lacking specific knots – Define knots using evenly spaced percentiles Binom. Interp. Variable Obs Percentile Centile [95% Conf. Interval]

vitd_supplr 110 0 28.57 28.57 28.57* 25 228.57 57.15 228.57 50 400 228.57 400 75 400 400 555.0903 100 1400 1400 1400*

* Lower (upper) confidence limit held at minimum (maximum) of sample

* Note: Above suggests knots at p25=228.57, p50=400 and p75=400

272

* LINEAR MODEL – Fit

Source SS df MS Number of obs = 110 F( 1, 108) = 0.61 Model 42.8432637 1 42.8432637 Prob > F = 0.4348 Residual 7529.22232 108 69.7150215 R-squared = 0.0057 Adj R-squared = -0.0035 Total 7572.06558 109 69.4684916 Root MSE = 8.3496

total_bf_p~t Coef. Std. Err. t P>|t| [95% Conf. Interval]

vitd_supplr -.0022811 .0029098 -0.78 0.435 -.0080488 .0034866 _cons 32.74697 1.307738 25.04 0.000 30.1548 35.33913

* LINEAR MODEL – Diagnostics Shapiro-Wilk W test for normal data

Variable Obs W V z Prob>z

ehat_linea~r 110 0.98746 1.121 0.255 0.39918

* Normality of residuals check is NOT significant (null not rejected)

* Linear Model Fit – Plot of ("Normality of Residuals")

Linear Model

Normality of Residuals

1.00

0.75

0.50

0.25

Normal F[(ehat_linear_vitd_supplr-m)/s] Normal 0.00 0.00 0.25 0.50 0.75 1.00 Empirical P[i] = i/(N+1)

273

* Linear Model Fit – Plot of ("Jacknife Residuals versus Predicted")

Linear Model

Jacknife Residuals v Predicted

Studentized residuals Studentized

-1 -2

29 30 31 32 33 Linear prediction

* Linear Model Fit – Plot of ("Cook Distances")

Linear Model

Cook Distances

Cook's D Cook's

.2 0

0 100 200 300 subj_id

* Note: Looks good. Biggest cook’s distance is less than 1

274

*RESTRICTED CUBIC SPLINES – Fit

Source SS df MS Number of obs = 110 F( 5, 104) = 0.52 Model 183.550973 5 36.7101947 Prob > F = 0.7631 Residual 7388.51461 104 71.0434097 R-squared = 0.0242 Adj R-squared = -0.0227 Total 7572.06558 109 69.4684916 Root MSE = 8.4287

total_bf_percent Coef. Std. Err. t P>|t| [95% Conf. Interval]

vitd_supplr -.0784632 .280821 -0.28 0.780 -.6353419 .4784154 vitd_supplr2 .0003081 .0016802 0.18 0.855 -.0030238 .00364 vitd_supplr3 -3.47e-07 2.91e-06 -0.12 0.905 -6.11e-06 5.42e-06 knot_vitd_supplrp25 8.75e-08 3.61e-06 0.02 0.981 -7.08e-06 7.25e-06 knot_vitd_supplrp50 3.10e-07 7.87e-07 0.39 0.694 -1.25e-06 1.87e-06 knot_vitd_supplrp75 0 (omitted) _cons 36.63863 11.2204 3.27 0.001 14.38815 58.88912

* RESTRICTED CUBIC SPLINES – Diagnostics

Shapiro-Wilk W test for normal data

Variable Obs W V z Prob>z

ehat_splin~r 110 0.98904 0.980 -0.044 0.51759

* Normality of residuals check is NOT significant (null not rejected)

* Restricted Cubic Splines Fit - Plot of ("Normality of Residuals")

Restricted Cubic Splines Model

Normality of Residuals

1.00

0.75

0.50

0.25

Normal F[(ehat_spline_vitd_supplr-m)/s] Normal 0.00 0.00 0.25 0.50 0.75 1.00 Empirical P[i] = i/(N+1)

275

* Restricted Cubic Splines Fit - Plot of ("Jacknife Residuals versus Predicted")

Restricted Cubic Splines Model

Jacknife Residuals v Predicted

Studentized residuals Studentized

-1 -2

28 30 32 34 36 Linear prediction

* Restricted Cubic Splines Fit - Plot of ("Cook Distances")

Restricted Cubic Splines Model

Cook Distances

Cook's D Cook's

.2 0

0 100 200 300 subj_id

* Note: Looks good. Biggest cook’s distance is still less than 1

276

* Compare Linear (1) and Splines (2) side-by-side

(1) (2) total_bf_p~t total_bf_p~t

vitd_supplr -0.00228 -0.0785 (0.00291) (0.281)

vitd_supplr2 0.000308 (0.00168)

vitd_supplr3 -0.000000347 (0.00000291)

knot_vitd_supplrp25 8.75e-08 (0.00000361)

knot_vitd_supplrp50 0.000000310 (0.000000787)

o.knot_vitd_suppl~75 0 (.)

Constant 32.75*** 36.64** (1.308) (11.22)

Observations 110 110 R-squared 0.006 0.024 Adjusted R-squared -0.004 -0.023 rmse 8.350 8.429

Standard errors in parentheses * p<0.05, ** p<0.01, *** p<0.001

* Partial F-test of Departure from Linearity

( 1) vitd_supplr = 0 ( 2) vitd_supplr2 = 0 ( 3) vitd_supplr3 = 0 ( 4) knot_vitd_supplrp25 = 0 ( 5) knot_vitd_supplrp50 = 0 Constraint 3 dropped

F( 4, 104) = 0.53 Prob > F = 0.7150

* Note: Departure from linearity is NOT significant. Proceed with the linear model

277

APPENDIX D LINEARITY AND SPLINE MODEL FIT FOR ADIPOSITY MEASURES *Y =Serum 25-OHD; X = Total Body Fat Percentage * Lowess Regression Curve and Linear Prediction Fit for the Association between Serum 25-OHD Levels and Total Body Fat Percentage

Lowess smoother

200

150

100

Serum 25-OHD (nmol/L) 0

10 20 30 40 50 Total Body Fat (%)

serum_25d lowess: serum_25d Fitted values

bandwidth = .3

Lowess smoother

200

150

100

Serum 25-OHD (nmol/L) 0

10 20 30 40 50 Total Body Fat (%)

serum_25d lowess: serum_25d 95% CI Fitted values

bandwidth = .3

* Define knots - Lacking specific knots – Define knots using evenly spaced percentiles Binom. Interp. Variable Obs Percentile Centile [95% Conf. Interval]

total_bf_p~t 270 0 14.8 14.8 14.8* 25 26.1 25.2 27.63685 50 31.6 30.63828 32.8 75 37.675 36.28157 39.54393 100 56.6 56.6 56.6*

* Lower (upper) confidence limit held at minimum (maximum) of sample

278

* Note: Above suggests knots at p25=26.1, p50=31.6 and p75=37.68

* LINEAR MODEL – Fit

Source SS df MS Number of obs = 174 F( 1, 172) = 0.40 Model 413.511617 1 413.511617 Prob > F = 0.5304 Residual 179990.301 172 1046.45524 R-squared = 0.0023 Adj R-squared = -0.0035 Total 180403.813 173 1042.7966 Root MSE = 32.349

serum_25d Coef. Std. Err. t P>|t| [95% Conf. Interval]

total_bf_percent .1862036 .2962134 0.63 0.530 -.3984777 .770885 _cons 73.25745 9.833621 7.45 0.000 53.84733 92.66756

* LINEAR MODEL – Diagnostics Shapiro-Wilk W test for normal data

Variable Obs W V z Prob>z

ehat_linea~f 174 0.94879 6.770 4.369 0.00001

* Normality of residuals check is significant (null rejected)

* Linear Model Fit – Plot of ("Normality of Residuals")

Linear Model

Normality of Residuals

1.00

0.75

0.50

0.25

Normal F[(ehat_linear_tbf-m)/s] Normal 0.00 0.00 0.25 0.50 0.75 1.00 Empirical P[i] = i/(N+1)

279

* Linear Model Fit – Plot of ("Jacknife Residuals versus Predicted")

Linear Model

Jacknife Residuals v Predicted

Studentized residuals Studentized

-1 -2 76 78 80 82 84 Linear prediction

* Linear Model Fit – Plot of ("Cook Distances")

Linear Model

Cook Distances

Cook's D Cook's

.2 0

0 100 200 300 subj_id

* Note: Linear prediction looks good. Biggest cook’s distance is less than 1

280

*RESTRICTED CUBIC SPLINES – Fit

Source SS df MS Number of obs = 174 F( 6, 167) = 1.75 Model 10667.1193 6 1777.85322 Prob > F = 0.1126 Residual 169736.693 167 1016.38738 R-squared = 0.0591 Adj R-squared = 0.0253 Total 180403.813 173 1042.7966 Root MSE = 31.881

serum_25d Coef. Std. Err. t P>|t| [95% Conf. Interval]

total_bf_percent 125.535 87.95225 1.43 0.155 -48.10659 299.1766 total_bf_percent2 -5.456412 3.891785 -1.40 0.163 -13.13985 2.227026 total_bf_percent3 .0766644 .0562542 1.36 0.175 -.0343967 .1877255 knot_total_bf_percentp25 -.0925775 .1116539 -0.83 0.408 -.3130124 .1278575 knot_total_bf_percentp50 -.0162627 .1019666 -0.16 0.873 -.2175723 .185047 knot_total_bf_percentp75 .0360661 .069492 0.52 0.604 -.10113 .1732622 _cons -850.0428 647.3483 -1.31 0.191 -2128.084 427.9981

* RESTRICTED CUBIC SPLINES – Diagnostics Shapiro-Wilk W test for normal data

Variable Obs W V z Prob>z

ehat_splin~f 174 0.95906 5.413 3.858 0.00006

* Normality of residuals check is significant (null rejected)

* Restricted Cubic Splines Fit - Plot of ("Normality of Residuals")

Restricted Cubic Splines Model

Normality of Residuals

1.00

0.75

0.50

0.25

Normal F[(ehat_spline_tbf-m)/s] Normal 0.00 0.00 0.25 0.50 0.75 1.00 Empirical P[i] = i/(N+1)

* Restricted Cubic Splines Fit - Plot of ("Jacknife Residuals versus Predicted")

281

Restricted Cubic Splines Model

Jacknife Residuals v Predicted

Studentized residuals Studentized

-1 -2 40 50 60 70 80 90 Linear prediction

* Restricted Cubic Splines Fit - Plot of ("Cook Distances")

Restricted Cubic Splines Model

Cook Distances

Cook's D Cook's

.2 0

0 100 200 300 subj_id

* Note: Spline prediction looks good. Biggest cook’s distance is still less than 1*

282

* Compare Linear (1) and Splines (2) side-by-side

(1) (2) serum_25d serum_25d

total_bf_percent 0.186 125.5 (0.296) (87.95)

total_bf_percent2 -5.456 (3.892)

total_bf_percent3 0.0767 (0.0563)

knot_total_bf_per~25 -0.0926 (0.112)

knot_total_bf_per~50 -0.0163 (0.102)

knot_total_bf_per~75 0.0361 (0.0695)

Constant 73.26*** -850.0 (9.834) (647.3)

Observations 174 174 R-squared 0.002 0.059 Adjusted R-squared -0.004 0.025 rmse 32.35 31.88

Standard errors in parentheses * p<0.05, ** p<0.01, *** p<0.001

* Partial F-test of Departure from Linearity

( 1) total_bf_percent = 0 ( 2) total_bf_percent2 = 0 ( 3) total_bf_percent3 = 0 ( 4) knot_total_bf_percentp25 = 0 ( 5) knot_total_bf_percentp50 = 0 ( 6) knot_total_bf_percentp75 = 0

F( 6, 167) = 1.75 Prob > F = 0.1126

* Note: Departure from linearity is NOT significant. Proceed with the linear model

283

*Y =Serum 25-OHD; X = Body Mass Index * Lowess Regression Curve and Linear Prediction Fit for the Association between Serum 25-OHD Levels and Body Mass Index

Lowess smoother

200

150

100

Serum 25-OHD (nmol/L) 0

15 20 25 30 35 Body Mass Index (kg/m2)

serum_25d lowess: serum_25d Fitted values

bandwidth = .3

Lowess smoother

200

150

100

Serum 25-OHD (nmol/L) 0

15 20 25 30 35 Body Mass Index (kg/m2)

serum_25d lowess: serum_25d 95% CI Fitted values

bandwidth = .3

* Lacking specific knots – Define knots using evenly spaced percentiles Binom. Interp. Variable Obs Percentile Centile [95% Conf. Interval]

bmi 270 0 16.67 16.67 16.67* 25 20.5125 20.20841 20.93369 50 22.36 21.95531 22.92469 75 24.4925 24.00263 25.10439 100 37.72 37.72 37.72*

* Lower (upper) confidence limit held at minimum (maximum) of sample

* Note: Above suggests knots at p25=20.51, p50=22.36 and p75=24.49

284

* LINEAR MODEL – Fit Source SS df MS Number of obs = 174 F( 1, 172) = 0.81 Model 842.415473 1 842.415473 Prob > F = 0.3703 Residual 179561.397 172 1043.96161 R-squared = 0.0047 Adj R-squared = -0.0011 Total 180403.813 173 1042.7966 Root MSE = 32.31

serum_25d Coef. Std. Err. t P>|t| [95% Conf. Interval]

bmi .6725131 .7486518 0.90 0.370 -.8052148 2.150241 _cons 63.80506 17.36018 3.68 0.000 29.53863 98.07148

* LINEAR MODEL – Diagnostics

Shapiro-Wilk W test for normal data

Variable Obs W V z Prob>z

ehat_linea~i 174 0.95022 6.581 4.305 0.00001

* Normality of residuals check is significant (null rejected)

* Linear Model Fit – Plot of ("Normality of Residuals")

Linear Model

Normality of Residuals

1.00

0.75

0.50

0.25

Normal F[(ehat_linear_bmi-m)/s] Normal 0.00 0.00 0.25 0.50 0.75 1.00 Empirical P[i] = i/(N+1)

285

* Linear Model Fit – Plot of ("Jacknife Residuals versus Predicted")

Linear Model

Jacknife Residuals v Predicted

Studentized residuals Studentized

-1 -2 75 80 85 90 Linear prediction

* Linear Model Fit – Plot of ("Jacknife Residuals versus Predicted")

Linear Model

Cook Distances

Cook's D Cook's

.2 0

0 100 200 300 subj_id

* Note: Linear prediction looks good. Biggest cook’s distance is less than 1

286

*RESTRICTED CUBIC SPLINES – Fit

Source SS df MS Number of obs = 174 F( 6, 167) = 0.31 Model 1986.43746 6 331.07291 Prob > F = 0.9312 Residual 178417.375 167 1068.36752 R-squared = 0.0110 Adj R-squared = -0.0245 Total 180403.813 173 1042.7966 Root MSE = 32.686

serum_25d Coef. Std. Err. t P>|t| [95% Conf. Interval]

bmi 142.1807 1645.947 0.09 0.931 -3107.364 3391.725 bmi2 -6.854677 84.4944 -0.08 0.935 -173.6695 159.9602 bmi3 .1104872 1.441605 0.08 0.939 -2.735632 2.956606 knot_bmip25 .0448409 2.703609 0.02 0.987 -5.292817 5.382499 knot_bmip50 -.4534833 2.048367 -0.22 0.825 -4.497515 3.590548 knot_bmip75 .4108644 .841773 0.49 0.626 -1.251024 2.072752 _cons -908.745 10655.77 -0.09 0.932 -21946.12 20128.63

* RESTRICTED CUBIC SPLINES – Diagnostics* Shapiro-Wilk W test for normal data

Variable Obs W V z Prob>z

ehat_splin~i 174 0.95316 6.193 4.166 0.00002

* Normality of residuals check is significant (null rejected)

* Restricted Cubic Splines Fit - Plot of ("Normality of Residuals")

Restricted Cubic Splines Model

Normality of Residuals

1.00

0.75

0.50

0.25

Normal F[(ehat_spline_bmi-m)/s] Normal 0.00 0.00 0.25 0.50 0.75 1.00 Empirical P[i] = i/(N+1)

287

* Restricted Cubic Splines Fit - Plot of ("Jacknife Residuals versus Predicted")

Restricted Cubic Splines Model

Jacknife Residuals v Predicted

Studentized residuals Studentized

-1 -2 60 80 100 120 140 160 Linear prediction

* Restricted Cubic Splines Fit - Plot of ("Cook Distances")

Restricted Cubic Splines Model

Cook Distances

Cook's D Cook's

.2 0

0 100 200 300 subj_id

* Note: Spline prediction looks good. Biggest cook’s distance is still less than 1

288

* Compare Linear (1) and Splines (2) side-by-side

(1) (2) serum_25d serum_25d

bmi 0.673 142.2 (0.749) (1645.9)

bmi2 -6.855 (84.49)

bmi3 0.110 (1.442)

knot_bmip25 0.0448 (2.704)

knot_bmip50 -0.453 (2.048)

knot_bmip75 0.411 (0.842)

Constant 63.81*** -908.7 (17.36) (10655.8)

Observations 174 174 R-squared 0.005 0.011 Adjusted R-squared -0.001 -0.025 rmse 32.31 32.69

Standard errors in parentheses * p<0.05, ** p<0.01, *** p<0.001

* Partial F-test of Departure from Linearity

( 1) bmi = 0 ( 2) bmi2 = 0 ( 3) bmi3 = 0 ( 4) knot_bmip25 = 0 ( 5) knot_bmip50 = 0 ( 6) knot_bmip75 = 0

F( 6, 167) = 0.31 Prob > F = 0.9312 * Note: Departure from linearity is NOT significant. Proceed with the linear model

289

*Y =Serum 25-OHD; X = Waist Circumference * Lowess Regression Curve and Linear Prediction Fit for the Association between Serum 25-OHD Levels and Waist Circumference

Lowess smoother

200

150

100

Serum 25-OHD (nmol/L) 0

60 80 100 120 Waist Circumference (cm)

serum_25d lowess: serum_25d Fitted values

bandwidth = .3

Lowess smoother

200

150

100

Serum 25-OHD (nmol/L) 0

60 80 100 120 Waist Circumference (cm)

serum_25d lowess: serum_25d 95% CI Fitted values

bandwidth = .3

* Define knots - Lacking specific knots – Define knots using evenly spaced percentiles Binom. Interp. Variable Obs Percentile Centile [95% Conf. Interval]

wst_cm 270 0 58.4 58.4 58.4* 25 71.1 68.6 72.4 50 76.2 74.93828 77.5 75 83.35 80 85.1 100 114.3 114.3 114.3*

* Lower (upper) confidence limit held at minimum (maximum) of sample

* Note: Above suggests knots at p25=71.1, p50=76.2 and p75=83.35

290

* LINEAR MODEL – Fit

Source SS df MS Number of obs = 174 F( 1, 172) = 0.33 Model 344.80124 1 344.80124 Prob > F = 0.5668 Residual 180059.011 172 1046.85472 R-squared = 0.0019 Adj R-squared = -0.0039 Total 180403.813 173 1042.7966 Root MSE = 32.355

serum_25d Coef. Std. Err. t P>|t| [95% Conf. Interval]

wst_cm .1593932 .2777336 0.57 0.567 -.388812 .7075983 _cons 66.70997 21.97659 3.04 0.003 23.33143 110.0885

* LINEAR MODEL – Diagnostics

Shapiro-Wilk W test for normal data

Variable Obs W V z Prob>z

ehat_linea~t 174 0.94873 6.778 4.372 0.00001

* Normality of residuals check is significant (null rejected)

* Linear Model Fit – Plot of ("Normality of Residuals")

Linear Model

Normality of Residuals

1.00

0.75

0.50

0.25

Normal F[(ehat_linear_wst-m)/s] Normal 0.00 0.00 0.25 0.50 0.75 1.00 Empirical P[i] = i/(N+1)

291

* Linear Model Fit – Plot of ("Jacknife Residuals versus Predicted")

Linear Model

Jacknife Residuals v Predicted

Studentized residuals Studentized

-1 -2 76 78 80 82 84 86 Linear prediction

* Linear Model Fit – Plot of ("Cook Distances")

Linear Model

Cook Distances

Cook's D Cook's

.2 0

0 100 200 300 subj_id

* Note: Linear prediction looks good. Biggest cook’s distance is less than 1

292

*RESTRICTED CUBIC SPLINES – Fit

Source SS df MS Number of obs = 174 F( 6, 167) = 0.44 Model 2787.26946 6 464.54491 Prob > F = 0.8535 Residual 177616.543 167 1063.57211 R-squared = 0.0155 Adj R-squared = -0.0199 Total 180403.813 173 1042.7966 Root MSE = 32.612

serum_25d Coef. Std. Err. t P>|t| [95% Conf. Interval]

wst_cm -2501.276 1822.423 -1.37 0.172 -6099.233 1096.68 wst_cm2 36.32253 26.36053 1.38 0.170 -15.7203 88.36536 wst_cm3 -.1755514 .1269376 -1.38 0.169 -.4261605 .0750578 knot_wst_cmp25 .2840046 .1951372 1.46 0.147 -.101249 .6692582 knot_wst_cmp50 -.1413803 .0981395 -1.44 0.152 -.3351343 .0523737 knot_wst_cmp75 .0378234 .0305493 1.24 0.217 -.0224892 .098136 _cons 57404.21 41944.12 1.37 0.173 -25404.84 140213.3

* RESTRICTED CUBIC SPLINES – Diagnostics

Shapiro-Wilk W test for normal data

Variable Obs W V z Prob>z

ehat_splin~t 174 0.94623 7.109 4.481 0.00000

* Normality of residuals check is significant (null rejected)

* Restricted Cubic Splines Fit - Plot of ("Normality of Residuals")

Restricted Cubic Splines Model

Normality of Residuals

1.00

0.75

0.50

0.25

Normal F[(ehat_spline_wst-m)/s] Normal 0.00 0.00 0.25 0.50 0.75 1.00 Empirical P[i] = i/(N+1)

293

* Restricted Cubic Splines Fit - Plot of ("Jacknife Residuals versus Predicted")

Restricted Cubic Splines Model

Jacknife Residuals v Predicted

Studentized residuals Studentized

-1 -2 50 100 150 200 250 Linear prediction

* Restricted Cubic Splines Fit - Plot of ("Cook Distances")

Restricted Cubic Splines Model

Cook Distances

Cook's D Cook's .8

.2 0

0 100 200 300 subj_id

* Note: Spline prediction looks good. Cook’s distance is > 1 for one observation

294

* Compare Linear (1) and Splines (2) side-by-side

(1) (2) serum_25d serum_25d

wst_cm 0.159 -2501.3 (0.278) (1822.4)

wst_cm2 36.32 (26.36)

wst_cm3 -0.176 (0.127)

knot_wst_cmp25 0.284 (0.195)

knot_wst_cmp50 -0.141 (0.0981)

knot_wst_cmp75 0.0378 (0.0305)

Constant 66.71** 57404.2 (21.98) (41944.1)

Observations 174 174 R-squared 0.002 0.015 Adjusted R-squared -0.004 -0.020 rmse 32.36 32.61

Standard errors in parentheses * p<0.05, ** p<0.01, *** p<0.001

* Partial F-test of Departure from Linearity

( 1) wst_cm = 0 ( 2) wst_cm2 = 0 ( 3) wst_cm3 = 0 ( 4) knot_wst_cmp25 = 0 ( 5) knot_wst_cmp50 = 0 ( 6) knot_wst_cmp75 = 0

F( 6, 167) = 0.44 Prob > F = 0.8535

* Note: Departure from linearity is NOT significant. Proceed with the linear model

295

*Y =Serum 25-OHD; X = Arm Fat Percentage * Lowess Regression Curve and Linear Prediction Fit for the Association between Serum 25-OHD Levels and Arm Fat Percentage

Lowess smoother

200

150

100

Serum 25-OHD (nmol/L) 0

10 20 30 40 50 Arm Fat Percentage(%)

serum_25d lowess: serum_25d Fitted values

bandwidth = .3

Lowess smoother

200

150

100

Serum 25-OHD (nmol/L) 0

10 20 30 40 50 Arm Fat Percentage(%)

serum_25d lowess: serum_25d 95% CI Fitted values

bandwidth = .3

* Define knots - Lacking specific knots – Define knots using evenly spaced percentiles Binom. Interp. Variable Obs Percentile Centile [95% Conf. Interval]

arm_fat_pe~t 270 0 8.2 8.2 8.2* 25 24.475 22.65607 25.51843 50 29.6 28.09141 31.66172 75 38 36.58157 39.47197 100 52.5 52.5 52.5*

* Lower (upper) confidence limit held at minimum (maximum) of sample * Note: Above suggests knots at p25=24.48, p50=29.6 and p75=38

296

* LINEAR MODEL – Fit

Source SS df MS Number of obs = 174 F( 1, 172) = 0.04 Model 39.3269831 1 39.3269831 Prob > F = 0.8467 Residual 180364.486 172 1048.63073 R-squared = 0.0002 Adj R-squared = -0.0056 Total 180403.813 173 1042.7966 Root MSE = 32.383

serum_25d Coef. Std. Err. t P>|t| [95% Conf. Interval]

arm_fat_percent .0514622 .2657383 0.19 0.847 -.4730659 .5759902 _cons 77.70931 8.294727 9.37 0.000 61.33674 94.08187

* LINEAR MODEL – Diagnostics Shapiro-Wilk W test for normal data

Variable Obs W V z Prob>z

ehat_linea~f 174 0.94866 6.788 4.375 0.00001

* Normality of residuals check is significant (null rejected)

* Linear Model Fit – Plot of ("Normality of Residuals")

Linear Model

Normality of Residuals

1.00

0.75

0.50

0.25

Normal F[(ehat_linear_af-m)/s] Normal 0.00 0.00 0.25 0.50 0.75 1.00 Empirical P[i] = i/(N+1)

297

* Linear Model Fit – Plot of ("Jacknife Residuals versus Predicted")

Linear Model

Jacknife Residuals v Predicted

Studentized residuals Studentized

-1 -2 78 78.5 79 79.5 80 80.5 Linear prediction

* Linear Model Fit – Plot of ("Cook Distances")

Linear Model

Cook Distances

Cook's D Cook's

.2 0

0 100 200 300 subj_id

* Note: Linear prediction looks good. Biggest cook’s distance is less than 1

298

*RESTRICTED CUBIC SPLINES – Fit

Source SS df MS Number of obs = 174 F( 6, 167) = 1.35 Model 8323.46199 6 1387.24366 Prob > F = 0.2394 Residual 172080.351 167 1030.42126 R-squared = 0.0461 Adj R-squared = 0.0119 Total 180403.813 173 1042.7966 Root MSE = 32.1

serum_25d Coef. Std. Err. t P>|t| [95% Conf. Interval]

arm_fat_percent 48.4344 22.85006 2.12 0.036 3.322193 93.5466 arm_fat_percent2 -2.551403 1.21711 -2.10 0.038 -4.954308 -.1484977 arm_fat_percent3 .0420874 .0207216 2.03 0.044 .0011774 .0829974 knot_arm_fat_percentp25 -.0774471 .0613727 -1.26 0.209 -.1986134 .0437193 knot_arm_fat_percentp50 .0235405 .0676138 0.35 0.728 -.1099474 .1570284 knot_arm_fat_percentp75 .0351601 .0516609 0.68 0.497 -.0668324 .1371526 _cons -202.7319 136.3584 -1.49 0.139 -471.9404 66.47647

* RESTRICTED CUBIC SPLINES – Diagnostics

Shapiro-Wilk W test for normal data

Variable Obs W V z Prob>z

ehat_splin~f 174 0.95212 6.330 4.216 0.00001

* Normality of residuals check is significant (null rejected)

* Restricted Cubic Splines Fit - Plot of ("Normality of Residuals")

Restricted Cubic Splines Model

Normality of Residuals

1.00

0.75

0.50

0.25

Normal F[(ehat_spline_af-m)/s] Normal 0.00 0.00 0.25 0.50 0.75 1.00 Empirical P[i] = i/(N+1)

299

* Restricted Cubic Splines Fit - Plot of ("Jacknife Residuals versus Predicted")

Restricted Cubic Splines Model

Jacknife Residuals v Predicted

Studentized residuals Studentized

-1 -2 50 60 70 80 90 Linear prediction

* Restricted Cubic Splines Fit - Plot of ("Cook Distances")

Restricted Cubic Splines Model

Cook Distances

Cook's D Cook's

.2 0

0 100 200 300 subj_id

* Note: Spline prediction looks good. Biggest cook’s distance is still less than 1

300

* Compare Linear (1) and Splines (2) side-by-side

(1) (2) serum_25d serum_25d

arm_fat_percent 0.0515 48.43* (0.266) (22.85)

arm_fat_percent2 -2.551* (1.217)

arm_fat_percent3 0.0421* (0.0207)

knot_arm_fat_perc~25 -0.0774 (0.0614)

knot_arm_fat_perc~50 0.0235 (0.0676)

knot_arm_fat_perc~75 0.0352 (0.0517)

Constant 77.71*** -202.7 (8.295) (136.4)

Observations 174 174 R-squared 0.000 0.046 Adjusted R-squared -0.006 0.012 rmse 32.38 32.10

Standard errors in parentheses * p<0.05, ** p<0.01, *** p<0.001

* Partial F-test of Departure from Linearity

( 1) arm_fat_percent = 0 ( 2) arm_fat_percent2 = 0 ( 3) arm_fat_percent3 = 0 ( 4) knot_arm_fat_percentp25 = 0 ( 5) knot_arm_fat_percentp50 = 0 ( 6) knot_arm_fat_percentp75 = 0

F( 6, 167) = 1.35 Prob > F = 0.2394 * Note: Departure from linearity is NOT significant. Proceed with the linear model

301

*Y =Serum 25-OHD; X = Leg Fat Percentage * Lowess Regression Curve and Linear Prediction Fit for the Association between Serum 25-OHD Levels and Leg Fat Percentage

Lowess smoother

200

150

100

Serum 25-OHD (nmol/L) 0

10 20 30 40 50 Leg Fat Percentage(%)

serum_25d lowess: serum_25d Fitted values

bandwidth = .3

Lowess smoother

200

150

100

Serum 25-OHD (nmol/L) 0

10 20 30 40 50 Leg Fat Percentage(%)

serum_25d lowess: serum_25d 95% CI Fitted values

bandwidth = .3

* Define knots - Lacking specific knots – Define knots using evenly spaced percentiles Binom. Interp. Variable Obs Percentile Centile [95% Conf. Interval]

leg_fat_pe~t 270 0 8.7 8.7 8.7* 25 30.175 28.82803 31.5 50 35.1 34.3 36.2 75 40.6 39.86315 41.2 100 54.2 54.2 54.2*

* Lower (upper) confidence limit held at minimum (maximum) of sample

* Note: Above suggests knots at p25=30.18, p50=35.1 and p75=40.6

302

* LINEAR MODEL – Fit

Source SS df MS Number of obs = 174 F( 1, 172) = 0.05 Model 50.3299067 1 50.3299067 Prob > F = 0.8268 Residual 180353.483 172 1048.56676 R-squared = 0.0003 Adj R-squared = -0.0055 Total 180403.813 173 1042.7966 Root MSE = 32.382

serum_25d Coef. Std. Err. t P>|t| [95% Conf. Interval]

leg_fat_percent .0698886 .3190005 0.22 0.827 -.5597713 .6995485 _cons 76.735 11.71084 6.55 0.000 53.61954 99.85046

* LINEAR MODEL – Diagnostics

Shapiro-Wilk W test for normal data

Variable Obs W V z Prob>z

ehat_linea~f 174 0.94908 6.732 4.356 0.00001

* Normality of residuals check is significant (null rejected)

* Linear Model Fit – Plot of ("Normality of Residuals")

Linear Model

Normality of Residuals

1.00

0.75

0.50

0.25

Normal F[(ehat_linear_lf-m)/s] Normal 0.00 0.00 0.25 0.50 0.75 1.00 Empirical P[i] = i/(N+1)

303

* Linear Model Fit – Plot of ("Jacknife Residuals versus Predicted")

Linear Model

Jacknife Residuals v Predicted

Studentized residuals Studentized

-1 -2 77 78 79 80 81 Linear prediction

* Linear Model Fit – Plot of ("Cook Distances")

Linear Model

Cook Distances

Cook's D Cook's

.2 0

0 100 200 300 subj_id

* Note: Linear prediction looks good. Biggest cook’s distance is less than 1

304

*RESTRICTED CUBIC SPLINES – Fit

Source SS df MS Number of obs = 174 F( 6, 167) = 0.82 Model 5167.39672 6 861.232786 Prob > F = 0.5553 Residual 175236.416 167 1049.31986 R-squared = 0.0286 Adj R-squared = -0.0063 Total 180403.813 173 1042.7966 Root MSE = 32.393

serum_25d Coef. Std. Err. t P>|t| [95% Conf. Interval]

leg_fat_percent 7.315349 21.07453 0.35 0.729 -34.29149 48.92219 leg_fat_percent2 -.2746332 .9679993 -0.28 0.777 -2.185726 1.63646 leg_fat_percent3 .0030541 .0139906 0.22 0.827 -.0245672 .0306754 knot_leg_fat_percentp25 .019731 .0641618 0.31 0.759 -.1069417 .1464037 knot_leg_fat_percentp50 -.0521618 .105227 -0.50 0.621 -.2599083 .1555847 knot_leg_fat_percentp75 .0133673 .0986049 0.14 0.892 -.1813055 .2080401 _cons 22.07104 143.3152 0.15 0.878 -260.872 305.0141

* RESTRICTED CUBIC SPLINES – Diagnostics . Shapiro-Wilk W test for normal data

Variable Obs W V z Prob>z

ehat_splin~f 174 0.95324 6.182 4.162 0.00002

* Normality of residuals check is significant (null rejected)

* Restricted Cubic Splines Fit - Plot of ("Normality of Residuals")

Restricted Cubic Splines Model

Normality of Residuals

1.00

0.75

0.50

0.25

Normal F[(ehat_spline_lf-m)/s] Normal 0.00 0.00 0.25 0.50 0.75 1.00 Empirical P[i] = i/(N+1)

305

* Restricted Cubic Splines Fit - Plot of ("Jacknife Residuals versus Predicted")

Restricted Cubic Splines Model

Jacknife Residuals v Predicted

Studentized residuals Studentized

-1 -2

40 50 60 70 80 90 Linear prediction

* Restricted Cubic Splines Fit - Plot of ("Cook Distances")

Restricted Cubic Splines Model

Cook Distances

Cook's D Cook's

1 0

0 100 200 300 subj_id

* Note: Spline prediction looks good. Cook’s distance > 1 for one observation*

306

* Compare Linear (1) and Splines (2) side-by-side

(1) (2) serum_25d serum_25d

leg_fat_percent 0.0699 7.315 (0.319) (21.07)

leg_fat_percent2 -0.275 (0.968)

leg_fat_percent3 0.00305 (0.0140)

knot_leg_fat_perc~25 0.0197 (0.0642)

knot_leg_fat_perc~50 -0.0522 (0.105)

knot_leg_fat_perc~75 0.0134 (0.0986)

Constant 76.73*** 22.07 (11.71) (143.3)

Observations 174 174 R-squared 0.000 0.029 Adjusted R-squared -0.006 -0.006 rmse 32.38 32.39

Standard errors in parentheses * p<0.05, ** p<0.01, *** p<0.001

* Partial F-test of Departure from Linearity

( 1) leg_fat_percent = 0 ( 2) leg_fat_percent2 = 0 ( 3) leg_fat_percent3 = 0 ( 4) knot_leg_fat_percentp25 = 0 ( 5) knot_leg_fat_percentp50 = 0 ( 6) knot_leg_fat_percentp75 = 0

F( 6, 167) = 0.82 Prob > F = 0.5553 * Note: Departure from linearity is NOT significant. Proceed with the linear model

307

*Y =Serum 25-OHD; X = Trunk Fat Percentage * Lowess Regression Curve and Linear Prediction Fit for the Association between Serum 25-OHD Levels and Trunk Fat Percentage

Lowess smoother

200

150

100

Serum 25-OHD (nmol/L) 0

10 20 30 40 50 60 Trunk Fat Percentage (%)

serum_25d lowess: serum_25d Fitted values

bandwidth = .3

Lowess smoother

200

150

100

Serum 25-OHD (nmol/L) 0

10 20 30 40 50 60 Trunk Fat Percentage (%)

serum_25d lowess: serum_25d 95% CI Fitted values

bandwidth = .3

* Define knots - Lacking specific knots – Define knots using evenly spaced percentiles

Binom. Interp. Variable Obs Percentile Centile [95% Conf. Interval]

trunk_fat_~t 270 0 11.7 11.7 11.7* 25 25.275 23.4 26.41843 50 31 30.13828 32.22344 75 38.525 35.8 40.3 100 60.7 60.7 60.7*

* Lower (upper) confidence limit held at minimum (maximum) of sample * Note: Above suggests knots at p25=25.28, p50=31 and p75=38.53

308

* LINEAR MODEL – Fit

Source SS df MS Number of obs = 174 F( 1, 172) = 0.80 Model 830.965283 1 830.965283 Prob > F = 0.3736 Residual 179572.847 172 1044.02818 R-squared = 0.0046 Adj R-squared = -0.0012 Total 180403.813 173 1042.7966 Root MSE = 32.311

serum_25d Coef. Std. Err. t P>|t| [95% Conf. Interval]

trunk_fat_percent .224919 .2521104 0.89 0.374 -.2727097 .7225477 _cons 72.08415 8.390594 8.59 0.000 55.52235 88.64594

* LINEAR MODEL – Diagnostics Shapiro-Wilk W test for normal data

Variable Obs W V z Prob>z

ehat_linea~f 174 0.94909 6.731 4.356 0.00001

* Normality of residuals check is significant (null rejected)

* Linear Model Fit – Plot of ("Normality of Residuals")

Linear Model

Normality of Residuals

1.00

0.75

0.50

0.25

Normal F[(ehat_linear_tf-m)/s] Normal 0.00 0.00 0.25 0.50 0.75 1.00 Empirical P[i] = i/(N+1)

309

* Linear Model Fit – Plot of ("Jacknife Residuals versus Predicted")

Linear Model

Jacknife Residuals v Predicted

Studentized residuals Studentized

-1 -2 75 80 85 Linear prediction

* Linear Model Fit – Plot of ("Cook Distances")

Linear Model

Cook Distances

Cook's D Cook's

.2 0

0 100 200 300 subj_id

* Note: Linear prediction looks good. Biggest cook’s distance is less than 1

310

*RESTRICTED CUBIC SPLINES – Fit

Source SS df MS Number of obs = 174 F( 6, 167) = 2.01 Model 12163.0274 6 2027.17124 Prob > F = 0.0667 Residual 168240.785 167 1007.42985 R-squared = 0.0674 Adj R-squared = 0.0339 Total 180403.813 173 1042.7966 Root MSE = 31.74

serum_25d Coef. Std. Err. t P>|t| [95% Conf. Interval]

trunk_fat_percent -99.27567 45.27739 -2.19 0.030 -188.6655 -9.885824 trunk_fat_percent2 4.680598 2.161648 2.17 0.032 .4129187 8.948277 trunk_fat_percent3 -.0719512 .0333991 -2.15 0.033 -.13789 -.0060124 knot_trunk_fat_percentp25 .1641801 .0724989 2.26 0.025 .0210475 .3073126 knot_trunk_fat_percentp50 -.1470753 .0667792 -2.20 0.029 -.2789156 -.015235 knot_trunk_fat_percentp75 .0690709 .0410088 1.68 0.094 -.0118916 .1500335 _cons 761.2512 305.5506 2.49 0.014 158.0114 1364.491

* RESTRICTED CUBIC SPLINES – Diagnostics Shapiro-Wilk W test for normal data

Variable Obs W V z Prob>z

ehat_splin~f 174 0.97659 3.095 2.581 0.00492

* Normality of residuals check is significant (null rejected)*

* Restricted Cubic Splines Fit - Plot of ("Normality of Residuals")

Restricted Cubic Splines Model

Normality of Residuals

1.00

0.75

0.50

0.25

Normal F[(ehat_spline_tf-m)/s] Normal 0.00 0.00 0.25 0.50 0.75 1.00 Empirical P[i] = i/(N+1)

311

* Restricted Cubic Splines Fit - Plot of ("Jacknife Residuals versus Predicted")

Restricted Cubic Splines Model

Jacknife Residuals v Predicted

Studentized residuals Studentized

-2 -4 60 80 100 120 Linear prediction

* Restricted Cubic Splines Fit - Plot of ("Cook Distances")

Restricted Cubic Splines Model

Cook Distances

Cook's D Cook's

.2 0

0 100 200 300 subj_id

* Note: Spline prediction looks good. Biggest cook’s distance is still less than 1

312

* Compare Linear (1) and Splines (2) side-by-side

(1) (2) serum_25d serum_25d

trunk_fat_percent 0.225 -99.28* (0.252) (45.28)

trunk_fat_percent2 4.681* (2.162)

trunk_fat_percent3 -0.0720* (0.0334)

knot_trunk_fat_pe~25 0.164* (0.0725)

knot_trunk_fat_pe~50 -0.147* (0.0668)

knot_trunk_fat_pe~75 0.0691 (0.0410)

Constant 72.08*** 761.3* (8.391) (305.6)

Observations 174 174 R-squared 0.005 0.067 Adjusted R-squared -0.001 0.034 rmse 32.31 31.74

Standard errors in parentheses * p<0.05, ** p<0.01, *** p<0.001

* Partial F-test of Departure from Linearity

( 1) trunk_fat_percent = 0 ( 2) trunk_fat_percent2 = 0 ( 3) trunk_fat_percent3 = 0 ( 4) knot_trunk_fat_percentp25 = 0 ( 5) knot_trunk_fat_percentp50 = 0 ( 6) knot_trunk_fat_percentp75 = 0

F( 6, 167) = 2.01 Prob > F = 0.0667 * Note: Departure from linearity is marginally significant. Proceed with the linear model

313

*Y =Serum 25-OHD; X = Android Fat Percentage * Lowess Regression Curve and Linear Prediction Fit for the Association between Serum 25-OHD Levels and Android Fat Percentage

Lowess smoother

200

150

100

Serum 25-OHD (nmol/L) 0

10 20 30 40 50 60 Android Fat Percentage(%)

serum_25d lowess: serum_25d Fitted values

bandwidth = .3

Lowess smoother

200

150

100

Serum 25-OHD (nmol/L) 0

10 20 30 40 50 60 Android Fat Percentage(%)

serum_25d lowess: serum_25d 95% CI Fitted values

bandwidth = .3

* Define knots - Lacking specific knots – Define knots using evenly spaced percentiles Binom. Interp. Variable Obs Percentile Centile [95% Conf. Interval]

android_fa~t 270 0 12.5 12.5 12.5* 25 26.475 25.6 28.11843 50 33.8 32.3 36.2 75 41.825 39.76315 44.37197 100 62.9 62.9 62.9*

* Lower (upper) confidence limit held at minimum (maximum) of sample

* Note: Above suggests knots at p25=26.48, p50=33.8 and p75=41.83

314

* LINEAR MODEL – Fit

Source SS df MS Number of obs = 174 F( 1, 172) = 0.70 Model 736.082288 1 736.082288 Prob > F = 0.4024 Residual 179667.73 172 1044.57983 R-squared = 0.0041 Adj R-squared = -0.0017 Total 180403.813 173 1042.7966 Root MSE = 32.32

serum_25d Coef. Std. Err. t P>|t| [95% Conf. Interval]

android_fat_percent .1941041 .231229 0.84 0.402 -.2623078 .6505159 _cons 72.50559 8.392461 8.64 0.000 55.94011 89.07107

* LINEAR MODEL – Diagnostics Shapiro-Wilk W test for normal data

Variable Obs W V z Prob>z

ehat_linea~f 174 0.94868 6.785 4.374 0.00001

* Normality of residuals check is significant (null rejected)

* Linear Model Fit – Plot of ("Normality of Residuals")

Linear Model

Normality of Residuals

1.00

0.75

0.50

0.25

Normal F[(ehat_linear_anf-m)/s] Normal 0.00 0.00 0.25 0.50 0.75 1.00 Empirical P[i] = i/(N+1)

315

* Linear Model Fit – Plot of ("Jacknife Residuals versus Predicted")

Linear Model

Jacknife Residuals v Predicted

Studentized residuals Studentized

-1 -2 75 80 85 Linear prediction

* Linear Model Fit – Plot of ("Cook Distances")

Linear Model

Cook Distances

Cook's D Cook's

.2 0

0 100 200 300 subj_id

* Note: Linear prediction looks good. Biggest cook’s distance is less than 1

316

*RESTRICTED CUBIC SPLINES – Fit

Source SS df MS Number of obs = 174 F( 6, 167) = 1.63 Model 9991.28454 6 1665.21409 Prob > F = 0.1412 Residual 170412.528 167 1020.4343 R-squared = 0.0554 Adj R-squared = 0.0214 Total 180403.813 173 1042.7966 Root MSE = 31.944

serum_25d Coef. Std. Err. t P>|t| [95% Conf. Interval]

android_fat_percent -91.38856 42.97005 -2.13 0.035 -176.2231 -6.554043 android_fat_percent2 4.205491 1.947363 2.16 0.032 .3608691 8.050112 android_fat_percent3 -.0626731 .0285218 -2.20 0.029 -.1189828 -.0063633 knot_android_fat_percentp25 .1350824 .0548391 2.46 0.015 .0268152 .2433496 knot_android_fat_percentp50 -.1298347 .0500871 -2.59 0.010 -.2287202 -.0309492 knot_android_fat_percentp75 .0859876 .0390651 2.20 0.029 .0088624 .1631127 _cons 718.6123 304.954 2.36 0.020 116.5504 1320.674

* RESTRICTED CUBIC SPLINES – Diagnostics Shapiro-Wilk W test for normal data

Variable Obs W V z Prob>z

ehat_splin~f 174 0.96636 4.447 3.409 0.00033

* Normality of residuals check is significant (null rejected)

* Restricted Cubic Splines Fit - Plot of ("Normality of Residuals")

Restricted Cubic Splines Model

Normality of Residuals

1.00

0.75

0.50

0.25

Normal F[(ehat_spline_anf-m)/s] Normal 0.00 0.00 0.25 0.50 0.75 1.00 Empirical P[i] = i/(N+1)

317

* Restricted Cubic Splines Fit - Plot of ("Jacknife Residuals versus Predicted")

Restricted Cubic Splines Model

Jacknife Residuals v Predicted

Studentized residuals Studentized

-2 -4 60 80 100 120 140 Linear prediction

* Restricted Cubic Splines Fit - Plot of ("Cook Distances")

Restricted Cubic Splines Model

Cook Distances

Cook's D Cook's

.2 0

0 100 200 300 subj_id

* Note: Spline prediction looks good. Biggest cook’s distance is still less than 1

318

* Compare Linear (1) and Splines (2) side-by-side

(1) (2) serum_25d serum_25d

android_fat_percent 0.194 -91.39* (0.231) (42.97)

android_fat_percent2 4.205* (1.947)

android_fat_percent3 -0.0627* (0.0285)

knot_android_fat_~25 0.135* (0.0548)

knot_android_fat_~50 -0.130* (0.0501)

knot_android_fat_~75 0.0860* (0.0391)

Constant 72.51*** 718.6* (8.392) (305.0)

Observations 174 174 R-squared 0.004 0.055 Adjusted R-squared -0.002 0.021 rmse 32.32 31.94

Standard errors in parentheses * p<0.05, ** p<0.01, *** p<0.001

* Partial F-test of Departure from Linearity

( 1) android_fat_percent = 0 ( 2) android_fat_percent2 = 0 ( 3) android_fat_percent3 = 0 ( 4) knot_android_fat_percentp25 = 0 ( 5) knot_android_fat_percentp50 = 0 ( 6) knot_android_fat_percentp75 = 0

F( 6, 167) = 1.63 Prob > F = 0.1412

* Note: Departure from linearity is NOT significant. Proceed with the linear model

319

*Y =Serum 25-OHD; X = Gynoid Fat Percentage * Lowess Regression Curve and Linear Prediction Fit for the Association between Serum 25-OHD Levels and Gynoid Fat Percentage

Lowess smoother

200

150

100

Serum 25-OHD (nmol/L) 0

20 30 40 50 60 Gynoid Fat Percentage(%)

serum_25d lowess: serum_25d Fitted values

bandwidth = .3

Lowess smoother

200

150

100

Serum 25-OHD (nmol/L) 0

20 30 40 50 60 Gynoid Fat Percentage(%)

serum_25d lowess: serum_25d 95% CI Fitted values

bandwidth = .3

* Define knots - Lacking specific knots – Define knots using evenly spaced percentiles

Variable Obs Percentile Centile [95% Conf. Interval]

gynoid_fat~t 270 0 16.1 16.1 16.1* 25 36.275 35 37.2 50 40.95 39.67656 41.9 75 45.2 44.38158 45.97197 100 59.9 59.9 59.9*

* Lower (upper) confidence limit held at minimum (maximum) of sample

* Note: Above suggests knots at p25=36.28, p50=40.95 and p75=45.2

320

* LINEAR MODEL – Fit

Source SS df MS Number of obs = 174 F( 1, 172) = 0.03 Model 33.0079556 1 33.0079556 Prob > F = 0.8594 Residual 180370.805 172 1048.66747 R-squared = 0.0002 Adj R-squared = -0.0056 Total 180403.813 173 1042.7966 Root MSE = 32.383

serum_25d Coef. Std. Err. t P>|t| [95% Conf. Interval]

gynoid_fat_percent .0634391 .3575743 0.18 0.859 -.6423598 .7692379 _cons 76.6577 14.78117 5.19 0.000 47.48185 105.8335

* LINEAR MODEL – Diagnostics Shapiro-Wilk W test for normal data

Variable Obs W V z Prob>z

ehat_linea~f 174 0.94927 6.707 4.348 0.00001

* Normality of residuals check is significant (null rejected)

* Linear Model Fit – Plot of ("Normality of Residuals")

Linear Model

Normality of Residuals

1.00

0.75

0.50

0.25

Normal F[(ehat_linear_gyf-m)/s] Normal 0.00 0.00 0.25 0.50 0.75 1.00 Empirical P[i] = i/(N+1)

321

* Linear Model Fit – Plot of ("Jacknife Residuals versus Predicted")

Linear Model

Jacknife Residuals v Predicted

Studentized residuals Studentized

-1 -2 77 78 79 80 81 Linear prediction

* Linear Model Fit – Plot of ("Cook Distances")

Linear Model

Cook Distances

Cook's D Cook's

.2 0

0 100 200 300 subj_id

* Note: Linear prediction looks good. Biggest cook’s distance is less than 1

322

*RESTRICTED CUBIC SPLINES – Fit Source SS df MS Number of obs = 174 F( 6, 167) = 0.95 Model 5939.91318 6 989.985531 Prob > F = 0.4626 Residual 174463.899 167 1044.69401 R-squared = 0.0329 Adj R-squared = -0.0018 Total 180403.813 173 1042.7966 Root MSE = 32.322

serum_25d Coef. Std. Err. t P>|t| [95% Conf. Interval]

gynoid_fat_percent 3.642658 40.26293 0.09 0.928 -75.84728 83.13259 gynoid_fat_percent2 -.0228067 1.405261 -0.02 0.987 -2.797172 2.751558 gynoid_fat_percent3 -.0009727 .0158987 -0.06 0.951 -.032361 .0304157 knot_gynoid_fat_percentp25 .0672372 .0796772 0.84 0.400 -.0900672 .2245416 knot_gynoid_fat_percentp50 -.16901 .1459351 -1.16 0.248 -.4571254 .1191054 knot_gynoid_fat_percentp75 .1124944 .1147833 0.98 0.328 -.1141191 .3391078 _cons 21.45657 373.3076 0.06 0.954 -715.5538 758.4669

* RESTRICTED CUBIC SPLINES – Diagnostics Shapiro-Wilk W test for normal data

Variable Obs W V z Prob>z

ehat_splin~f 174 0.95981 5.314 3.816 0.00007

* Normality of residuals check is significant (null rejected)

* Restricted Cubic Splines Fit - Plot of ("Normality of Residuals")

Restricted Cubic Splines Model

Normality of Residuals

1.00

0.75

0.50

0.25

Normal F[(ehat_spline_gyf-m)/s] Normal 0.00 0.00 0.25 0.50 0.75 1.00 Empirical P[i] = i/(N+1)

323

* Restricted Cubic Splines Fit - Plot of ("Jacknife Residuals versus Predicted")

Restricted Cubic Splines Model

Jacknife Residuals v Predicted

Studentized residuals Studentized

-1 -2 40 50 60 70 80 90 Linear prediction

* Restricted Cubic Splines Fit - Plot of ("Cook Distances")

Restricted Cubic Splines Model

Cook Distances

Cook's D Cook's

1 0

0 100 200 300 subj_id

* Note: Spline prediction generally looks good. Cook’s distance > 1 for two observations

324

* Compare Linear (1) and Splines (2) side-by-side

(1) (2) serum_25d serum_25d

gynoid_fat_percent 0.0634 3.643 (0.358) (40.26)

gynoid_fat_percent2 -0.0228 (1.405)

gynoid_fat_percent3 -0.000973 (0.0159)

knot_gynoid_fat_p~25 0.0672 (0.0797)

knot_gynoid_fat_p~50 -0.169 (0.146)

knot_gynoid_fat_p~75 0.112 (0.115)

Constant 76.66*** 21.46 (14.78) (373.3)

Observations 174 174 R-squared 0.000 0.033 Adjusted R-squared -0.006 -0.002 rmse 32.38 32.32

Standard errors in parentheses * p<0.05, ** p<0.01, *** p<0.001

* Partial F-test of Departure from Linearity

( 1) gynoid_fat_percent = 0 ( 2) gynoid_fat_percent2 = 0 ( 3) gynoid_fat_percent3 = 0 ( 4) knot_gynoid_fat_percentp25 = 0 ( 5) knot_gynoid_fat_percentp50 = 0 ( 6) knot_gynoid_fat_percentp75 = 0

F( 6, 167) = 0.95 Prob > F = 0.4626

* Note: Departure from linearity is NOT significant. Proceed with the linear model

325

*Y =Serum 25-OHD; X = Android to Gynoid Fat Ratio * Lowess Regression Curve and Linear Prediction Fit for the Association between Serum 25-OHD Levels and Android to Gynoid Fat Ratio

Lowess smoother

200

150

100

Serum 25-OHD (nmol/L) 0

.5 1 1.5 2 Android to Gynoid Fat Percentage (ratio)

serum_25d lowess: serum_25d Fitted values

bandwidth = .3

Lowess smoother

200

150

100

Serum 25-OHD (nmol/L) 0

.5 1 1.5 2 Android to Gynoid Fat Percentage (ratio)

serum_25d lowess: serum_25d 95% CI Fitted values

bandwidth = .3

* Define knots - Lacking specific knots – Define knots using evenly spaced percentiles Binom. Interp. Variable Obs Percentile Centile [95% Conf. Interval]

ag_ratio 270 0 .4048913 .4048913 .4048913* 25 .721889 .7014008 .7700477 50 .8389978 .8179142 .8581198 75 .9528462 .9280353 .9741237 100 1.776397 1.776397 1.776397*

* Lower (upper) confidence limit held at minimum (maximum) of sample * Note: Above suggests knots at p25=0.72, p50=0.84 and p75=0.95

326

* LINEAR MODEL – Fit Source SS df MS Number of obs = 174 F( 1, 172) = 0.78 Model 813.680118 1 813.680118 Prob > F = 0.3786 Residual 179590.132 172 1044.12868 R-squared = 0.0045 Adj R-squared = -0.0013 Total 180403.813 173 1042.7966 Root MSE = 32.313

serum_25d Coef. Std. Err. t P>|t| [95% Conf. Interval]

ag_ratio 12.568 14.23693 0.88 0.379 -15.5336 40.66961 _cons 68.69532 12.1976 5.63 0.000 44.61906 92.77159

* LINEAR MODEL – Diagnostics Shapiro-Wilk W test for normal data

Variable Obs W V z Prob>z

ehat_linea~r 174 0.94686 7.026 4.454 0.00000

* Normality of residuals check is significant (null rejected)

* Linear Model Fit – Plot of ("Normality of Residuals")

Linear Model

Normality of Residuals

1.00

0.75

0.50

0.25

Normal F[(ehat_linear_agr-m)/s] Normal 0.00 0.00 0.25 0.50 0.75 1.00 Empirical P[i] = i/(N+1)

327

* Linear Model Fit – Plot of ("Jacknife Residuals versus Predicted")

Linear Model

Jacknife Residuals v Predicted

Studentized residuals Studentized

-1 -2 75 80 85 90 Linear prediction

* Linear Model Fit – Plot of ("Cook Distances")

Linear Model

Cook Distances

Cook's D Cook's

.2 0

0 100 200 300 subj_id

* Note: Linear prediction looks good. Biggest cook’s distance is less than 1

328

*RESTRICTED CUBIC SPLINES – Fit Source SS df MS Number of obs = 174 F( 6, 167) = 1.29 Model 8008.40907 6 1334.73485 Prob > F = 0.2630 Residual 172395.403 167 1032.30781 R-squared = 0.0444 Adj R-squared = 0.0101 Total 180403.813 173 1042.7966 Root MSE = 32.13

serum_25d Coef. Std. Err. t P>|t| [95% Conf. Interval]

ag_ratio 1434.181 3864.864 0.37 0.711 -6196.108 9064.47 ag_ratio2 -2873.935 6209.942 -0.46 0.644 -15134.04 9386.173 ag_ratio3 1788.044 3269.968 0.55 0.585 -4667.759 8243.846 knot_ag_ratiop25 -9161.747 7917.894 -1.16 0.249 -24793.82 6470.321 knot_ag_ratiop50 16732.21 9261.342 1.81 0.073 -1552.194 35016.61 knot_ag_ratiop75 -11034.34 5248.039 -2.10 0.037 -21395.39 -673.288 _cons -137.8501 786.9147 -0.18 0.861 -1691.433 1415.733

* RESTRICTED CUBIC SPLINES – Diagnostics Shapiro-Wilk W test for normal data

Variable Obs W V z Prob>z

ehat_splin~r 174 0.95249 6.282 4.198 0.00001 * Normality of residuals check is significant (null rejected)

* Restricted Cubic Splines Fit - Plot of ("Normality of Residuals")

Restricted Cubic Splines Model

Normality of Residuals

1.00

0.75

0.50

0.25

Normal F[(ehat_spline_agr-m)/s] Normal 0.00 0.00 0.25 0.50 0.75 1.00 Empirical P[i] = i/(N+1)

329

* Restricted Cubic Splines Fit - Plot of ("Jacknife Residuals versus Predicted")

Restricted Cubic Splines Model

Jacknife Residuals v Predicted

Studentized residuals Studentized

-1 -2

60 80 100 120 140 Linear prediction

* Restricted Cubic Splines Fit - Plot of ("Cook Distances")

Restricted Cubic Splines Model

Cook Distances

1708.085

Cook's D Cook's 0

0 100 200 300 subj_id

* Note: Spline prediction generally appears good. Cook’s distance > 1 for one observation

330

* Compare Linear (1) and Splines (2) side-by-side

(1) (2) serum_25d serum_25d

ag_ratio 12.57 1434.2 (14.24) (3864.9)

ag_ratio2 -2873.9 (6209.9)

ag_ratio3 1788.0 (3270.0)

knot_ag_ratiop25 -9161.7 (7917.9)

knot_ag_ratiop50 16732.2 (9261.3)

knot_ag_ratiop75 -11034.3* (5248.0)

Constant 68.70*** -137.9 (12.20) (786.9)

Observations 174 174 R-squared 0.005 0.044 Adjusted R-squared -0.001 0.010 rmse 32.31 32.13

* Partial F-test of Departure from Linearity

( 1) ag_ratio = 0 ( 2) ag_ratio2 = 0 ( 3) ag_ratio3 = 0 ( 4) knot_ag_ratiop25 = 0 ( 5) knot_ag_ratiop50 = 0 ( 6) knot_ag_ratiop75 = 0

F( 6, 167) = 1.29 Prob > F = 0.2630 * Note: Departure from linearity is NOT significant. Proceed with the linear model

331

APPENDIX E LINEARITY AND SPLINE MODEL FIT FOR SERUM 25-OHD LEVELS *Y =Inflammation (hs-CRP); X = Serum 25-OHD * Lowess Regression Curve and Linear Prediction Fit for the Association between Inflammation and Serum 25-OHD Levels

Lowess smoother

loghs-C ReactiveProtein(mg/L) -2

0 50 100 150 200 Serum 25-OHD (nmol/L)

ths-crp_mgl lowess: ths-crp_mgl Fitted values

bandwidth = .3

Lowess smoother

loghs-C ReactiveProtein(mg/L) -2

0 50 100 150 200 Serum 25-OHD (nmol/L)

ths-crp_mgl lowess: ths-crp_mgl 95% CI Fitted values

bandwidth = .3

* Lacking specific knots – Define knots using evenly spaced percentiles Binom. Interp. Variable Obs Percentile Centile [95% Conf. Interval]

serum_25d 174 0 21.7 21.7 21.7* 25 55.275 51.33308 60.6 50 72.7 68.11087 77.24457 75 97.3 90.27447 108 100 174.3 174.3 174.3*

* Lower (upper) confidence limit held at minimum (maximum) of sample

332

* Note: Above suggests knots at p25=55.28, p50=72.7 and p75=97.3 * LINEAR MODEL – Fit

Source SS df MS Number of obs = 168 F( 1, 166) = 1.62 Model 3.94999093 1 3.94999093 Prob > F = 0.2055 Residual 405.907787 166 2.44522763 R-squared = 0.0096 Adj R-squared = 0.0037 Total 409.857778 167 2.45423819 Root MSE = 1.5637

tcrp_mgl Coef. Std. Err. t P>|t| [95% Conf. Interval]

serum_25d .0047232 .0037162 1.27 0.206 -.0026139 .0120603 _cons -.4428095 .3188431 -1.39 0.167 -1.07232 .1867008

* LINEAR MODEL – Diagnostics

Shapiro-Wilk W test for normal data

Variable Obs W V z Prob>z

ehat_lin~25d 168 0.96310 4.733 3.545 0.00020

* Normality of residuals check is significant (null rejected)

* Linear Model Fit – Plot of ("Normality of Residuals")

Linear Model

Normality of Residuals

1.00

0.75

0.50

0.25

Normal F[(ehat_linear_25d-m)/s] Normal 0.00 0.00 0.25 0.50 0.75 1.00 Empirical P[i] = i/(N+1)

333

* Linear Model Fit – Plot of ("Jacknife Residuals versus Predicted")

Linear Model

Jacknife Residuals v Predicted

Studentized residuals Studentized

-1 -2 -.4 -.2 0 .2 .4 Linear prediction

* Linear Model Fit – Plot of ("Cook Distances")

Linear Model

Cook Distances

Cook's D Cook's

.2 0

0 100 200 300 subj_id

334

* Note: Linear prediction looks good. Biggest cook’s distance is less than 1

*RESTRICTED CUBIC SPLINES – Fit

Source SS df MS Number of obs = 168 F( 6, 161) = 1.13 Model 16.5985128 6 2.7664188 Prob > F = 0.3458 Residual 393.259265 161 2.44260413 R-squared = 0.0405 Adj R-squared = 0.0047 Total 409.857778 167 2.45423819 Root MSE = 1.5629

tcrp_mgl Coef. Std. Err. t P>|t| [95% Conf. Interval]

serum_25d .8396773 .6082801 1.38 0.169 -.3615592 2.040914 serum_25d2 -.0191812 .0136405 -1.41 0.162 -.0461186 .0077562 serum_25d3 .0001393 .0000973 1.43 0.154 -.0000528 .0003315 knot_serum_25dp25 -.0002547 .0001712 -1.49 0.139 -.0005928 .0000834 knot_serum_25dp50 .0001467 .0001098 1.34 0.183 -.0000702 .0003637 knot_serum_25dp75 -.0000269 .0000445 -0.60 0.547 -.0001149 .0000611 _cons -11.75217 8.533572 -1.38 0.170 -28.60434 5.099999

* RESTRICTED CUBIC SPLINES – Diagnostics

Shapiro-Wilk W test for normal data

Variable Obs W V z Prob>z

ehat_spl~25d 168 0.96700 4.233 3.290 0.00050

* Normality of residuals check is significant (null rejected)*

* Restricted Cubic Splines Fit - Plot of ("Normality of Residuals")

Restricted Cubic Splines Model

Normality of Residuals

1.00

0.75

0.50

0.25

Normal F[(ehat_spline_25d-m)/s] Normal 0.00 0.00 0.25 0.50 0.75 1.00 Empirical P[i] = i/(N+1)

335

* Restricted Cubic Splines Fit - Plot of ("Jacknife Residuals versus Predicted")

Restricted Cubic Splines Model

Jacknife Residuals v Predicted

-1

Studentized residuals Studentized -2

-1 0 1 2 Linear prediction

* Restricted Cubic Splines Fit - Plot of ("Cook Distances")

Restricted Cubic Splines Model

Cook Distances

Cook's D Cook's

.2 0

0 100 200 300 subj_id

* Note: Spline prediction looks good. One observation has Cook’s distance > 1

336

* Compare Linear (1) and Splines (2) side-by-side

(1) (2) tcrp_mgl tcrp_mgl

serum_25d 0.00472 0.840 (0.00372) (0.608)

serum_25d2 -0.0192 (0.0136)

serum_25d3 0.000139 (0.0000973)

knot_serum_25dp25 -0.000255 (0.000171)

knot_serum_25dp50 0.000147 (0.000110)

knot_serum_25dp75 -0.0000269 (0.0000445)

Constant -0.443 -11.75 (0.319) (8.534)

Observations 168 168 R-squared 0.010 0.040 Adjusted R-squared 0.004 0.005 rmse 1.564 1.563

Standard errors in parentheses * p<0.05, ** p<0.01, *** p<0.001

* Partial F-test of Departure from Linearity

( 1) serum_25d = 0 ( 2) serum_25d2 = 0 ( 3) serum_25d3 = 0 ( 4) knot_serum_25dp25 = 0 ( 5) knot_serum_25dp50 = 0 ( 6) knot_serum_25dp75 = 0

F( 6, 161) = 1.13 Prob > F = 0.3458

* Note: Departure from linearity is NOT significant. Proceed with the linear model.

337

BIBLIOGRAPHY

Abbas S, Linseisen J, Slanger T, Kropp S, Mutschelknauss EJ, Flesch-Janys D, Chang- Claude J. Serum 25-hydroxyvitamin D and risk of postmenopausal breast cancer, results of a large case-control study. Carcinogenesis 2008; 29: 93-9.

Adams CF. Nutritive value of American foods. Agricultural handbook no .456. Washington, DC: US Government Printing office, 1975.

Ainsworth BE, Haskell WL, Leon AS, Jacobs DR jr, Montoye HJ, Sallis JF, Paffenbarger RS, jr. Compendium of physical activities: classification of energy costs of human physical activities. Med Sci Sports Exerc 1993; 25: 1575-81.

Akashi K, Shibuya T, Harada M, Takamatsu Y, Uike N, Eto T, Niho Y. Interleukin 4 suppresses the spontaneous growth of chronic myelomonocytic leukemia cells. J Clin Invest 1991; 88:223-30.

Amer M, Qayyum R. Relation between serum 25-hydroxyvitamin D and C-reactive protein in asymptomatic adults (from the continuous National Health and Nutrition Examination Survey 2001 to 2006). Am J Cardiol 2012; 109: 226-30.

American Council on Exercise. What are the guidelines for percentage of body fat loss, 2009? Available online: http://www.acefitness.org/blog/112/what-are-the guidelines-for-percentage-of-body-fat (accessed 30th March, 2013).

Andersson J, Libby P, Hansson GK. Adaptive immunity and atherosclerosis. Clin Immunol 2010; 134: 33-46.

Arunabh S, Pollack S, Yeh J, Aloia JF. Body fat content and 25-hydroxyvitamin D levels in healthy women. J Clin Endocrinol Metab 2003; 88: 157-61.

Atkinson J, Epand RF, Epand RM. Tocopherols and tocotrienols in membranes: a critical review. Free Radic Biol Med 2008; 44: 739-64.

Aviram M. Review of human studies on oxidative damage and antioxidant protection related to cardiovascular diseases. Free Radic Res 2000; 33: 85S-97S.

Azzi A. Molecular mechanism of α-tocopherol action. Free Radic Biol Med. 2007; 43: 16-21.

Badolato R, Oppenheim JJ. Role of cytokines, acute-phase proteins, and chemokines in the progression of rheumatoidarthritis. Semin Arthritis Rheum 1996; 26: 526-38.

338

Bahl R, Bhandari N, Hambidge KM, Bhan MK. Plasma zinc as a predictor of diarrheal and respiratory morbidity in children in an urban slum setting. Am J Clin Nutr 1998; 68: 414S-17S.

Bailey RL, Dodd KW, Goldman JA, Gahche JJ, Dwyer JT, Moshfegh AJ, Sempos CT, Picciano MF. Estimation of total usual calcium and vitamin D intakes in the United States. J Nutr 2010; 140: 817-22.

Ball K, Owen N, Salmon Jo, Bauman A, Gore C. Associations of physical activity with body weight and fat in men and women. Int J Obes 2001; 25: 914-19.

Banerjee P, Chatterjee M. Anti-proliferative role of vitamin D and its analog-a brief overview. Mol Cell Biochem 2003; 253: 247-54.

Barker T, Martins TB, Hill HR, Kjeldsberg CR, Henriksen VT, Dixon BM, Schneider ED, Dern A, Weaver LK. Different doses of supplemental vitamin D maintain interleukin-5 without altering skeletal muscle strength: a randomized, double- blind, placebo-controlled study in vitamin D sufficient adults. Nutr Metab (Lond) 2012; 9:16.

Barker T, Martins TB, Hill HR, Kjeldsberg CR, Dixon BM, Schneider ED, Henriksen VT, Weaver LK. Circulating pro-inflammatory cytokines are elevated and peak power output correlates with 25-hydroxyvitamin D in vitamin D insufficient adults. Eur J Appl Physiol 2013; 113: 1523-34.

Barnes MS, Horigan G, Cashman KD, Hill TR, Forsythe LK, Lucey AJ, McSorley EM, Kiely M, Bonham MP, Magee PJ, Strain JJ, Wallace JM. Maintenance of wintertime vitamin D status with cholecalciferol supplementation is not associated with alterations in serum cytokine concentrations among apparently healthy younger or older adults. J Nutr 2011; 141: 476-81.

Barrett-Connor E, Chang JC, Edelstein SL. Coffee-associated osteoporosis offset by daily milk consumption. JAMA 1994; 271: 280-83.

Barton, BE IL-6: insights into novel biological activities. Clin Immunol Immunopathol 1997; 85: 16-20.

Barton, BE, Shortall, J, Jackson, JV Interleukins 6 and 11 protect mice from mortality in a staphylococcal enterotoxin-induced toxic shock model. Infect Immun 1996; 64: 714-718.

339

Bastard JP, Maachi M, Lagathu C, Kim MJ, Caron M, Vidal H, Capeau J, Feve B. Recent advances in the relationship between obesity, inflammation, and insulin resistance. Eur Cytokine Netw 2006: 17: 4-12.

Bazzoni F, Beutler B. The tumor necrosis factor ligand and receptor families. N Engl J Med 1996; 334: 1717-25.

Beck FW, Prasad AS, Kaplan J, Fitzgerald JT, Brewer GJ. Changes in cytokine production and T cell subpopulations in experimentally induced zinc-deficient humans. Am J Physiol 1997; 272: E1002-E1007.

Beilfuss J, Berg V, Sneve M, Jorde R, Kamycheva E. Effects of a 1- year supplementation with cholecalciferol on interleukin-6, tumor necrosis factor- alpha and insulin resistance in overweight and obese subjects. Cytokine 2012; 60: 870-4.

Bell NH, Epstein S, Greene A, Shary J, Oexmann MJ, Shaw S. Evidence for alteration of the vitamin D-endocrine system in obese subjects. J Clin Invest 1985; 76: 370-3.

Bellia A, Garcovich C, D'Adamo M, Lombardo M, Tesauro M, Donadel G, Gentileschi P, Lauro D, Federici M, Lauro R, Sbraccia P. Serum 25- hydroxyvitaminD levels are inversely associated with systemic inflammation in se vere obese subjects. Intern Emerg Med. Published online: 25 March 2011.

Berg KM. Association between alcohol consumption and both osteoporotic fracture and bone density. Newton, MA: Cahners Publishing Company. 2008.

Berger A. Bone mineral density scans. BMJ 2002; 325: 484.

Bergman BC, Cornier MA, Horton TJ, Bessesen DH. Effects of fasting on insulin action and glucose kinetics in lean and obese men and women. Am J Physiol Endocrinol Metab 2007; 293: E1103-11.

Bertone-Johnson ER, Chocano-Bedoya PO, Zagarins SE, Micka AE, Ronnenberg AG.

Dietary vitamin D intake, 25-hydroxyvitamin D3 levels and prementural syndrome in a college-aged population. J Steroid Biochem Mol Biol 2010; 121: 434-37.

Beydoun MA, Boueiz A, Shroff MR, Beydoun HA, Wang Y, Zonderman AB. Associations among 25-hydroxyvitamin D, diet quality, and metabolic disturbance differ by adiposity in adults in the United States. J Clin Endocrinol Metab 2010; 95: 3814-27.

340

Binkley N, Krueger D, Cowgill CS, Plum L, Lake E, Hansen KE, DeLuca HF, Drezner MK. Assay variation confounds the diagnosis of hypovitaminosis D: a call for standardization. J Clin Endocrinol Metab 2004; 89: 3152-57.

Bischoff-Ferrari HA, Borchers M, Gudat F, Durmuller U, Stahelin HB, Dick W. Vitamin D receptor expression in human muscle tissue decreases with age. J Bone Miner Res 2004; 19:265-9.

Bischoff-Ferrari HA, Giovannucci E, Willett WC, Dietrich T, Dawson-Hughes B. Estimation of optimal serum concentrations of 25-hydroxyvitamin D for multiple health outcomes. Am J Clin Nutr 2006; 84: 18-28.

Black RE. Zinc deficiency, infectious disease and mortality in the developing world. J Nutr 2003; 133: 1485S-9S.

Bolanowski M, Nilsson BE. Assessment of human body composition using dual-energy x-ray absorptiometry and bioelectrical impedance analysis. Med Sci Monit 2001; 7: 1029-33.

Bosse Y, Lemire M, Poon AH, Daley D, He JQ, Sandford A, White JH, James AL, Musk AW, Palmer LJ, Raby BA, Weiss ST, Kozyrskyj AL, Becker A, Hudson TJ,Laprise C. Asthma and genes encoding components of the vitamin D pathway. Respir Res 2009; 10: 98.

Brandtzaeg P, Osnes L, Ovstebo R, Joo GB, Westvik AB, Kierulf P. Net inflammatory capacity of human septic shock plasma evaluated by a monocyte-based target cell assay: identification of interleukin-10 as a major functional deactivator of human monocytes. J Exp Med 1996; 184: 51-60.

Brannon PM, Yetley EA, Bailey RL, Picciano MF. Overview of the conference “Vitamin D and health in the 21st century: an update”. Am J Clin Nutr 2008; 88:483-90S.

Breslau NA. Calcium, estrogen, and progestin in the treatment of osteoporosis. Rheum Dis Clin North Am 1994; 20: 691-716.

Briefel RR. Secular trends in dietary intake in the United States. Annu Rev Nutr 2004; 24: 401-31.

341

Brigelius-Flohe R, Davies KJ . Is vitamin E an antioxidant, a regulator of signal transduction and gene expression, or a 'junk' food? Comments on the two accompanying papers: Molecular mechanism of alpha-tocopherol action" by A. Azzi and Vitamin E, antioxidant and nothing more" by Traber M. and Atkinson J. Free Radic Biol Med 2007; 43: 2-3.

Brigelius-Flohe R, Traber MG. Vitamin E: function and metabolism. FASEB J 1999; 13: 1145-55.

Bringhurst FR, Demay MB, Kronenberg HM. Hormones and disorders of mineral metabolism. In: Goldman L, Ausiello D, eds. Cecil Medicine. 23rd ed. Philadelphia, Pa: Saunders Elsevier; 2007: chap 28.

Bronner F. Recent developments in intestinal calcium absorption. Nutr Rev. 2009; 67: 109-13.

Brooks WA, Santosham M, Naheed A, Goswami D, Wahed MA, Diener-West M, Faruque AS, Black RE. Effect of weekly zinc supplements on incidence of pneumonia and diarrhea in children younger than 2 years in an urban, low-income population in Bangladesh: randomized controlled trial. Lancet 2005; 366: 999- 1004.

Brot C, Jorgensen NR, Sørensen OH. The influence of smoking on vitamin D status and calcium metabolism. Eur J Clin Nutr 1999; 53: 920-6.

Brown MA, Hural J. Functions of IL-4 and control of its expression. Crit Rev Immunol 1997; 17: 1-32.

Browning LM, Hsieh SD, Ashwell M. A systematic review of waist-to-height ratio as a screening tool for the prediction of cardiovascular disease and diabetes: 0·5 could be a suitable global boundary value. Nutr Res Rev 2010; 23: 247-69.

Buchanan JR, Santen R, Cauffman S, Cavaliere A, Greer RB, Demers LM. The effect of endogenous estrogen fluctuation on metabolism of 25-hydroxyvitamin D. Calcif Tissue Int 1986; 39: 139-44.

Buchowski MS, Semenya J, Johnson AO. Dietary calcium intake in lactose maldigesting intolerant and tolerant African-American Women. J Am Coll Nutr 2002; 21: 47- 54.

Calvo MS, Eastell R, Offord KP, Bergstralh EJ, Burritt MF. Circadian variation ionized calcium and intact parathyroid hormone: Evidence of sex differnces in calcium homeostatis. J Clin Endocrinol 1991; 72: 69-76.

342

Calvo MS, Whiting SJ, Barton CN. Vitamin D fortification in the United States and Canada: current status and data needs. Am J Clin Nutr 2004; 80: 1710S-6S.

Calvo MS. Dietary phosphorus, calcium metabolism and bone. J Nutr 1993; 123: 1627- 33.

Canning MO, Grotenhuis K, de Wit H, Ruwhof C, Drexhage HA. 1-alpha,25-

Dihydroxyvitamin D3 (1,25(OH)(2)D(3)) hampers the maturation of fully active immature dendritic cells from monocytes. Eur J Endocrinol 2001; 145: 351-7.

Caron-Jobin M, Morisset AS, Tremblay A, Huot C, Legare D, Tchernof A. Elevated serum 25-OHD concentrations, vitamin D, and calcium intakes are associated with reduced adipocyte size in women. Obesity 2011; 19: 1335-41.

Carr AC, Frei B. Toward a new recommended dietary allowance for vitamin C based on antioxidant and health effects in humans. Am J Clin Nutr 1999; 69: 1086-107.

Carrillo AE, Flynn MG, Pinkston C, Markofski MM, Jiang Y, Donkin SS, Teegarden D. Vitamin D supplementation during exercise training does not alter inflammatory biomarkers in overweight and obese subjects. Eur J Appl Physiol 2012; 112: 3045-52.

Carson RT, Vignali DA. Simultaneous quantitation of 15 cytokines using a multiplexed flow cytometric assay. J Immunol Methods 1999; 227: 41-52.

Carter GD. 25-hydroxyvitamin D assays: the quest for accuracy. Clin Chem 2009; 55: 1300-02.

CDC NCHS. National Health and Nutrition Examination Survey (NHANES) data for ages 1 - 70 years from 2001 - 2006. Center for Disease Control and Prevention. Overweight and Obesity Defining Overweight and Obesity. Available at http://www.nhlbi.nih.gov/health/health-topics/topics/obe/diagnosis.html.

Chacko SA, Song Y, Manson JE, Van Horn L, Eaton C, Martin LW, McTiernan A, Curb JD, Wylie-Rosett J, Phillips LS, Plodkowski RA, Liu S. Serum 25- hydroxyvitamin D concentrations in relation to cardiometabolic risk factors and metabolic syndrome in postmenopausal women. Am J Clin Nutr 2011; 94: 209- 17.

Charles P, Eriksen EF, Hasling C, Sondegard K, Mosekilde L. Dermal, intestinal, and renal obligatory losses of calcium: Relation to skeletal loss. Am J Clin Nutr 1991; 54: 266S-73S.

343

Chattipakorn N, Settakorn J, Petsophonsakul P, Suwannahoi P, Mahakranukrauh P, Srichairatanakool S, Chattipakorn SC. Cardiac mortality is associated with low levels of omega-3 and omega-6 fatty acids in the heart of cadavers with a history of coronary heart disease. Nutr Res 2009; 29: 696-704.

Chattopadhyay N, Mithal A, Brown EM. Physiology and Pathophysiology of Mineral Ion Metabolism. Endocrine Reviews 1996; 17: 289-307.

Chen TC, Chimeh F, Lu Z, Mathieu J, Person KS, Zhang A, Kohn N, Martinello S, Berkowitz R, Holick MF. Factors that influence the cutaneous synthesis and dietary sources of vitamin D. Arch Biochem Biophys 2007; 460: 213-7.

Chen TC, Turner AK, Holick MF. Methods for the determination of the circulating concentration of 25-hydroxyvitamin D. J Nutr Biochem 1990; 1: 315-9.

Cheng S, Lyytikainen A, Kroger H, Lamberg-Allardt C, Alen M, Koistinen A, Wang QJ, Suuriniemi M, Suominen H, Mahonen A, Nicholson PH, Ivaska KK, Korpela R, Ohlsson C, Vaananen KH, Tylavsky F. Effects of calcium, dairy product, and vitamin D supplementation on bone mass accrual and body composition in 10-12- y-old girls: a 2-y randomized trial. Am J Clin Nutr 2005; 82: 1115-26.

Cheng S, Massaro JM, Fox CS, Larson MG, Keyes MJ, McCabe EL, Robins SJ, O’Donnell CJ, Hoffmann U, Jacques PF, Booth SL, Vasan RS, Wolf M, Wang TJ. Adiposity, cardiometabolic risk, and vitamin D status: the Framingham Heart Study. Diabetes 2010; 59: 242-8.

Cheung BM, Ong KL, Tso AW, Leung RY, Xu A, Cherny SS, Sham PC, Lam TH, Lam KS. C-reactive protein as a predictor of hypertension in the Hong Kong Cardiovascular Risk Factor Prevalence Study (CRISPS) cohort. J Hum Hypertens 2012; 26: 108-16.

Chonchol M, Scragg R. 25-Hydroxyvitamin D, insulin resistance, and kidney function in the Third National Health and Nutrition Examination Survey. Kidney Int 2007; 71: 134-9.

Cigolini M, Iagulli MP, Miconi V, Galiotto M, Lombardi S, Targher G. Serum 25 hydroxyvitamin D3 concentrations and prevalence of cardiovascular disease among type 2 diabetic patients. Diabetes Care 2006; 29: 722-4.

Clark VR, Greenberg B, Harris TS, Carson BL. Body mass index and waist circumference predictors of cardiovascular risk in African Americans. Ethn Dis 2012; 22: 162-7.

344

Clarke CJP, Hales A, Hunt A, Foxwell BM. IL-10 mediated suppression of TNF-α production is independent of its ability to inhibit NF-κB activity. Eur J Immunol. 1998; 28: 1719-26.

Clary JP, Martin AD, Drinkwater DT, Marfell-Jones MJ. The skinfold: Myth and reality. J Sports Sci 1987; 5:3-33.

Clemens TL, Henderson SL, Adams JS, Holick MF. Increased skin pigments reduces the

capacity of skin to synthesize vitamin D3. Lancet 1982; 1: 74-6.

Committee to Review Dietary Reference Intakes for Vitamin D and Calcium, Food and Nutrition Board, Institute of Medicine. Dietary Reference Intakes for Calcium and Vitamin D. Washington, DC: National Academy Press, 2010.

Consumer and Food Economic Institute. Composition of foods. Agricultural handbook no.8. Washington, Dc: US Government Printing office, 1989.

Cranney A, Horsley T, O'Donnell S, Weiler H, Puil L, Ooi D, Atkinson S, Ward L, Moher D, Hanley D, Fang M, Yazdi F, Garritty C, Sampson M, Barrowman N,Tsertsvadze A, Mamaladze V. Effectiveness and safety of vitamin D in relation to bone health. Evid Rep Technol Assess 2007; 158: 1-235.

Damania B, Blackbourn DJ. Innate barriers to viral infection. Future Microbiol 2012 ;7:815-22.

Danesh J, Wheeler JG, Hirschfield GM, Eda S, Eiriksdottir G, Rumley A, Lowe GD, Pepys MB, Gudnason V. C-reactive protein and other circulating markers of inflammation in the prediction of coronary heart disease. N Engl J Med 2004; 350: 1387-97.

Daniel CR, McCullough ML, Patel RC, Jacobs EJ, Flanders WD, Thun MJ, Calle EE. Dietary intake of omega-6 and omega-3 fatty acids and risk of colorectal cancer in a prospective cohort of U.S. men and women. Cancer Epidemiol Biomarkers Prev 2009; 18: 516-25.

Davies KM, Heaney RP, Recker RR, Lappe JM, Barger-Lux MJ, Rafferty K, Hinders S. Calcium Intake and Body Weight. J Clin Endocrinol Metab 2000; 85: 4635-8. de Waal Malefyt, R, Figdor, CG, Huijbens R, Mohan-Peterson S, Bennett B, Culpepper J, Dang W, Zurawski G, de Vries JE. Effects of IL-13 on phenotype, cytokine production, and cytotoxic function of human monocytes. J Immunol 1993; 151: 6370-81.

345

Dehghan A, Kardys I, de Maat MP, Uitterlinden AG, Sijbrands EJ, Bootsma AH, Stijnen T, Hofman A, Schram MT, Witteman JC. Genetic variation, C-reactive protein levels, and incidence of diabetes. Diabetes 2007; 56: 872-78.

DeLuca HF. Overview of the general physiologic features and functions of vitamin D. Am J Clin Nutr 2004; 1689S-96S.

Despres JP, Lemieux I. Abdominal obesity and metabolic syndrome. Nature 2006; 444: 881-7.

Devaraj S, Hugou I, Jialal I. Alpha-Tocopherol decreases CD36 expression in human monocyte-derived macrophages. J Lipid Res 2001; 42: 521-7.

Dewailly E, Blanchet C, Lemieux S, Sauve L, Gingras S, Ayotte P, Holub BJ.n-3 fatty acids and cardiovascular disease risk factors among the Inuit of Nunavik. Am J Clin Nutr 2001; 74: 464-73.

Dinarello CA. Proinflammatory Cytokines. Chest 2000; 118: 503-8.

Ding C, Parameswaran V, Blizzard L, Burgess J, Jones G. Not a simple fat soluble vitamin: changes in serum 25-(OH)D levels are predicted by adiposity and adipocytokines in older adults. J Intern Med 2010; 268: 501-10.

Dorgan JF, Klifa C, Shepherd JA, Egleston BL, Kwiterovich PO, Himes JH, Gabriel K, Horn L, Snetselaar LG, Stevens VJ, Barton BA, Robson AM, Lasser NL, Deshmukh S, Hylton NM. Height, adiposity and body fat distribution and breast density in young women. Breast Cancer Res 2012; 14: R107.

Dorn JM, Hovey K, Muti P, Freudenheim JL, Russell M, Nochajski TH, Trevisan M. Alcohol Drinking Patterns Differentially Affect Central Adiposity as Measured by Abdominal Height in Women and Men. J Nutr 2003; 133: 2655-62.

Du Clos TW. C-reactive protein as a regulator of autoimmunity and inflammation. Arthritis Rheum 2003; 48: 1475-77.

Du H, Bennett D, Li L, Whitlock G, Guo Y, Collins R, Chen J, Bian Z, Hong LS, Feng S, Chen X, Chen L, Zhou R, Mao E, Peto R, Chen Z; China Kadoorie Biobank Collaborative Group. Physical activity and sedentary leisure time and their associations with BMI, waist circumference, and percentage body fat in 0.5 million adults: the China Kadoorie Biobank study. Am J Clin Nutr. 2013; 97: 487- 96.

Dunton GF, Berrigan D, Ballard-Barbash R, Graubard B, Atienza AA. Joint associations of physical activity and sedentary behaviors with body mass index: results from a time use survey of US adults. Int J Obes (Lond). 2009; 33: 1427-36.

346

Elkind MS, Cheng J, Boden-Albala B, Rundek T, Thomas J, Chen H, Rabbani LE, Sacco RL. Tumor necrosis factor receptor levels are associated with carotid atherosclerosis. Stroke 2002; 33: 31-7.

Ellis KJ. Human body composition: In vivo methods. Physiol Rev 2000; 80: 649-80.

Fatourou EM, Koskinas JS. Adaptive immunity in hepatocellular carcinoma: prognostic and therapeutic implications. Expert Rev Anticancer Ther 2009; 9: 1499-510.

Feldmann M, Brennan FM, Maini RN. Role of cytokines in rheumatoid arthritis. Annu Rev Immunol 1996; 14: 397-440.

Feskanich D, Rimm EB, Giovannucci EL, Colditz GA, Stampfer MJ. Reproducibility and validity of food intake measurements from a semi-quantitative food frequency questionnaire. J Am Diet Assoc 1993; 93: 790-6.

Fiorito LM, Mitchell DC, Smiciklas-Wright H, Birch LL. Girls’ calcium intake is associated with bone mineral content during childhood. J Nutr 2006; 36: 1281-6.

Flegal KM, Carroll MD, Kit BK, Ogden CL. Prevalence of obesity and trends in the distribution of body mass index among US adults, 1999-2010. JAMA 2012; 307: 491-7.

Flegal KM, Shepherd JA, Looker AC, Graubard BI, Borrud LG, Ogden CL, Harris TB, Everhart JE, Schenker N. Comparisons of percentage body fat, body mass index, waist circumference, and waist-stature ratio in adults. Am J Clin Nutr 2009; 89: 500-8.

Fleming KH, Yanovski JT. Composition of calcium in the US: food sources and intake levels. J Nutr 1994; 124: 1426S-30S.

Food and Nutrition Board (FNB) / Institute of Medicine (IOM). Dietary Reference Intakes for Calcium and Vitamin D (2010). 2101 Constitution Avenue, N.W. Washington, DC 20418. National Academy Press.

Food and Nutrition Board, Institute of Medicine. Dietary Reference Intakes. Washington, DC, National Academy Press, 1997.

Ford ES, Ajani UA, McGuire LC, Liu S. Concentrations of serum vitamin D and the metabolic syndrome among U.S. adults. Diabetes Care 2005; 28: 1228-30.

Forrest KYZ, Stuhldreher WL. Prevalence and correlates of vitamin D deficiency in US adults. Nutrition Research 2011; 31: 48-54.

347

Fotuhi M, Mohassel P, Yaffe K. Fish consumption, long-chain omega-3 fatty acids and risk of cognitive decline or Alzheimer disease: a complex association. Nat Clin Pract Neurol 2009; 5: 140-52.

Freeman DJ, Norrie J, Caslake MJ, Gaw A, Ford I, Lowe GD O'Reilly DS, Packard CJ, Sattar N. C-reactive protein is an independent predictor of risk for the development of diabetes in the West of Scotland Coronary Prevention Study. Diabetes 2002; 51: 1596–600.

Frei B, England L, Ames BN. Ascorbate is an outstanding antioxidant in human blood plasma. Proc Natl Acad Sci USA .1989; 86: 6377-81.

Fried SK, Bunkin DA, Greenberg AS. Omental and subcutaneous adipose tissues of obese subjects release interleukin-6: depot difference and regulation by glucocorticoid. J Clin Endocrinol Meta. 1998; 83: 847-50.

Gaffney-Stromberg E, Sun BH, Cucchi Ce, Simpson CA, Gundberg C, Kerstetter JE, Insogna KL. The effects of dietary protein on intestinal calcium absorption in rats. Endocrinology 2010; 151: 1071-78.

Galic S, Oakhill JS, Steinberg GR. Adipose tissue as an endocrine organ. Mol Cell Endocrinol 2010; 316:129-39.

Gallagher JC, Riggs BL, Deluca HF. Effect of estrogen on calcium absorption and serum vitamin D metabolites in postmenopausal osteoporosis. J Clin Endocrinol Metab 1980; 51: 1359-64.

Gannage-Yared MH, Azoury M, Mansour I, Baddoura R, Halaby G, Naaman R. Effects of a short-term calcium and vitamin D treatment on serum cytokines, bone markers, insulin and lipid concentrations in healthy post-menopausal women. J Endocrinol Invest 2003; 26:748-53.

García-Bailo B, Josse AR, Jamnik J, Badawi A, El-Sohemy A. Positive Association Between 25-Hydroxyvitamin D and C-Reactive Protein is Confounded by Hormonal Contraceptive Use. J Womens Health 2013; 22:417-25.

Gepner AD, Ramamurthy R, Krueger DC, Korcarz CE, Binkley N, Stein LH. A prospective randomized trial of the effects of vitamin D supplementation on cardiovascular disease risk. PLoS One 2012; 7: e36617.

Gerard D, Bryns C, Marchant A, Abramowicz D, Vandenabeele P, Delvaux A, Fiers W, Goldman M, Velu T. Interleukin 10 reduces the release of tumor necrosis factor and prevents lethality in experimental endotoxemia. J Exp Med 1993; 177: 547- 50.

348

Gershoff SN. Vitamin C (ascorbic acid): new roles, new requirements? Nutr Rev 1993;51:313-26.

Giovannucci E, Liu Y, Hollis BW, Rimm EB. 25-Hydroxyvitamin D and Risk of Myocardial Infarction in Men; A Prospective Study. Arch Intern Med 2008; 168: 1174-80.

Goodman GE, Thornquist MD, Balmes J, Cullen MR, Meyskens FL, Omenn GS, Valanis B, Williams JH Jr. The beta-carotene and retinol efficacy trial: incidence of lung cancer and cardiovascular disease mortality during 6-year follow-up after stopping β-carotene and retinol supplements. J Natl Cancer Inst 2004; 96: 1743- 50.

Goodyear-Bruch C, Pierce JD. Oxidative stress in critically ill patients. AJCC 2002; 11: 543–51.

Grant WB, Holick MF. Benefits and requirements of vitamin D for optimal health: A review. Altern Med Rev 2005; 10: 94-111.

Greenland P, Alpert JS, Beller GA, Benjamin EJ, Budoff MJ, Fayad ZA, Foster E, Hlatky MA, Hodgson JM, Kushner FG, Lauer MS, Shaw LJ, Smith SC Jr, Taylor AJ, Weintraub WS, Wenger NK, Jacobs AK. ACCF/AHA guideline for assessment of cardiovascular risk in asymptomatic adul ts: a report of the American College of Cardiology Foundation/American Heart Association Task Force on Practice Guidelines. Circulation 2010; 122: e584-636.

Gropper SS, Smith JL, Groff JL. Advanced nutrition and human metabolism. 5th Edition. Wadsworth Publishing, Belmont. CA, USA. February 14, 2008.

Ha E, Zemel MB. Functional properties of whey, whey components, and essential amino acids: mechanisms underlying health benefits for active people (review). J Nutr Biochem 2003; 14: 251-8.

Haarbo J, Gotfredsen A, Hassager C, Christiansen C. Validation of body composition by dual energy X-ray absorptiometry (DEXA). Clin Physiol 1991; 11: 331-41.

Haderslev KV, Haderslev PH, Staun M. Accuracy of body composition measurement by dual energy x-ray absorptiometry in underweight patients with chronic intestinal disease and in lean subjects. Dyn Med 2005; 4: 1.

349

Haderslev KV, Staun M. Comparison of dual-energy X-ray absorptiometry to four other methods to determine body composition in underweight patients with chronic gastrointestinal disease. Metabolism 2000; 49: 360-6.

Hanley DA, Davison KS. Vitamin D insufficiency in North America. J Nutr 2005; 135: 332-7.

Harris SS, Dawson-Hughes B. The association of oral contraceptive use with plasma 25- hydroxyvitamin D levels. J Am Coll Nutr 1998; 17: 282-4.

Hayes DP. Vitamin D and ageing. Biogerontology 2010; 11:1-16.

Heaney RP, Rafferty K. Carbonated beverages and urinary calcium excretion. Am J Clin Nutr 2001; 74: 343-7.

Heaney RP, Recker RR, Stegman MR, Moy AJ. Calcium absorption in women: relationships to calcium intake, estrogen status, and age. J Bone Miner Res 1989; 4: 469-75.

Heaney RP. Bone mass, nutrition, and other lifestyle factors. Nutr Rev 1996; 54: 3S-10S.

Heaney RP. Functional indices of vitamin D status and ramifications of vitamin D deficiency. Am J Clin Nutr 2004; 80: 1706-9S.

Herrera E, Barbas C. Vitamin E: Action, metabolism and perspectives. J Physiol Biochem 2001; 57: 43–56.

Heyneman CA. Zinc deficiency and taste disorders. Ann Pharmacother 1996; 30: 186-7.

Hirsch PE, Peng TC. Effects of alcohol on calcium homeostasis and bone. In: Anderson J, Garner S, eds. Calcium and Phosphorus in Health and Disease. Boca Raton, FL: CRC Press, 1996: 289-300.

Hitchon C, El-Gabalawy H. Oxidation in rheumatoid arthritis. Arthritis Res Ther.2004; 6: 265-78.

Holick MF, Binkley NC, Bischoff-Ferrari HA, Gordon CM, Hanley DA, Heaney RP, Murad MH, Weaver CM. Evaluation, treatment, and prevention of vitamin D deficiency: an Endocrine Society clinical practice guideline. J Clin Endocrinol Metab 2011; 96: 1911-30.

Holick MF, Chen TC. Vitamin D deficiency, a worldwide problem with health consequences. Am J Clin Nutr 2008; 87: 1080-6S.

350

Holick MF, Garabedian M. Vitamin D: photobiology, metabolism, mechanism of action, and clinical applications. In: Favus MJ, editor. Primer on the Metabolic Bone Diseases and Disorders of Mineral Metabolism. Sixth Edition. American Society for Bone and Mineral Research; Washington, DC: 2006. pp. 129-37.

Holick MF, Matsuoka LY, Wortsman J. Age, vitamin D, and solar ultraviolet. Lancet 1989; 2: 1104-5.

Holick MF. Resurrection of vitamin D deficiency and rickets. J Clin Invest 2006; 116: 2062-72.

Holick MF. Sunlight dilemma: risk of skin cancer or bone disease and muscle weakness. Lancet 2001; 357: 4-6.

Holick MF. The use and interpretation of assays for vitamin D and its metabolites. J Nutr 1990; 120: 1464-9.

Holick MF. The Vitamin D Solution: A 3-Step Strategy to Cure Our Most Common Health Problems. Hudson Street Press, 1st Edition, 2010.

Holick MF. Vitamin D deficiency. N Engl J Med 2007; 357: 266-81.

Holick MF. Vitamin D status: measurement, interpretation and clinical application. Ann Epidemiol 2009; 19: 73-8.

Holick MF. Vitamin D: new horizons for the 21st century. Am J Clin Nutr 1994; 60: 619- 30.

Hollis BW, Wagner CL. Assesssment of dietary vitamin D requirements during pregnancy and lactation. Am J Clin Nutr 2004; 79: 717-26.

Hollis BW, Wagner CL. Vitamin D requirements during lactation: high dose maternal supplementation as therapy to prevent hypovitaminosis D for both the mother and the nursing infant. Am J Clin Nutr 2004; 80: 1752S-58S.

Hollis BW. Circulating 25-hydroxyvitamin D levels indicative of vitamin D sufficiency: implications for establishing a new effective dietary recommendation for vitamin D. J Nutr 2005; 135: 317-22.

Hollis BW. Editorial: the determination of circulating 25-hydroxyvitamin D: no easy task. J Clin Endocrinol Metab 2004; 89: 3149-51.

Hollis, B. Comparison of commercially available (125) I-based RIA method for the determination of circulating 25-hydroxyvitamin D. Clin Chem 2000; 46: 1657- 61.

351

Hotamisligil GS, Shargill NS, Spiegelman BM. Adipose expression of tumor necrosis factor-alpha: direct role in obesity-linked insulin resistance. Science 1993; 259: 87-91.

Hsieh SD, Muto T. The superiority of waist-to-height ratio as an anthropometric index to evaluate clustering of coronary risk factors among non-obese men and women. Prev Med 2005; 40: 216-20.

Hsieh SD, Yoshinaga H, Muto T. Waist-to-height ratio, a simple and practical index for assessing central fat distribution and metabolic risk in Japanese men and women. Int J Obes Relat Metab Disor. 2003; 27: 610-6.

Ilich JZ, Brownbill RA, Tamborini L, Crncevic-Orlic Z: To drink or not to drink: how are alcohol, caffeine and past smoking related to bone mineral density in elderly women? J Am Coll Nutr 2002; 21: 536-44.

Institute of Medicine, Food and Nutrition Board. Dietary Reference Intakes for Vitamin A, Vitamin K, Arsenic, Boron, Chromium, Copper, Iodine, Iron, Manganese, Molybdenum, Nickel, Silicon, Vanadium, and Zinc. Washington, DC: National Academy Press, 2001.

Institute of Medicine. Food and Nutrition Board. Dietary Reference Intakes for Vitamin C, Vitamin E, Selenium, and Carotenoids. Washington, DC: National Academy Press, 2000.

International Diabetes Federation. The IDF Consensus Worldwide Definition of the Metabolic Syndrome (2006), Avenue Emile De Mot 19, B-1000 Brussels, Belgium. Available online: http//www.idf.org/webdata/docs/IDF_Meta_def_final.pdf (accessed 30th March, 2013).

Jackson AS, Pollack ML. Practical assessment of body composition. Physician and Sportsmed 1985; 13: 76-90.

Jacob RA, Sotoudeh G. Vitamin C function and status in chronic disease. Nutr Clin Care 2002; 5: 66-74.

Jacqmain M, Doucet E, Despres JP, Bouchard C, Tremblay A. Calcium intake, body composition, and lipoprotein-lipid concentrations in adults. Am J Clin Nutr 2003; 77: 1448-52.

Janeway CA, Travers P, Walport M, Shlomchik MJ. Immunology biology, the immune system in health and disease. 6th Edition. Garland Science Publishing, 270 Madison Avenue, NewYork. 2005.

352

Jansen JH, Wientjens GJ, Fibbe WE, Willemze R, Kluin-Nelemans HC. Inhibition of human macrophage colony formation by interleukin 4. J Exp Med 1989; 170:577- 82.

Jialal I, Devaraj S, Venugopal. C-reactive protein: risk marker or mediator in atherothrombosis? Hypertension 2004; 44: 1-6.

Jialal I, Devaraj S, Kaul N. The effect of alpha-tocopherol on monocyte proatherogenic activity. J Nutr 2001; 131: 389S-94S.

Johnson EJ, Russell RM. Beta-Carotene. In: Coates PM, Betz JM, Blackman MR, et al., eds. Encyclopedia of Dietary Supplements. 2nd ed. London and New York: Informa Healthcare; 2010: 115-20.

Johnson F, Giulivi C. Superoxide dismutases and their impact upon human health. Mol Aspects Med 2005; 26: 340–52.

Jones G. Pharmacokinetics of vitamin D toxicity. Am J Clin Nutr 2008; 88: 582S-6S.

Jorde R, Sneve M, Torjesen PA, Figenschau Y, Goransson LG, Omdal R. No effect of supplementation with cholecalciferol on cytokines and markers of inf lammation in overweight and obese subjects. Cytokine 2010; 50: 175-80.

Jungert A, Roth HJ, Neuhauser-Berthold M. Serum 25-hydroxyvitamin D3 and body composition in an elderly cohort from Germany: a cross-sectional study. Nutr Metab 2012; 9: 42.

Kadam N, Chiplonkar S, Fischer PR, Khadilkar V. School-based Calcium vitamin D with micronutrient supplementation enhances bone mass in underprivileged Indian premenarchal girls. Bone 2012; 51: 1-7.

Kaisho T. Elucidating the mechanism behind immunity using dendritic cells. Riken Research, 2007;2(8). Available at http://www.rikenresearch.riken.jp/eng/frontline/5028.

Karakia H, Ozaki H, Hori M, Mitsui-saito M, Amano KI, Harada KI, Miyamoto S, Nakazawa H, Won KJ, Sato K. Calcium movements, distribution, and functions in smooth muscle. Pharmacol Rev 1997; 49: 2157-230.

Kerstetter JE, O'Brien KO, Caseria DM, Wall DE, Insogna KL. The impact of dietary protein on calcium absorption and kinetic measures of bone turnover in women. J Clin Endocrinol Metab 2005; 90: 26-31.

353

Khadilkar A, Kadam N, Chiplonkar S, Fischer PR, Khadilkar V. School-based Calcium- vitamin D with micronutrient supplementation enhances bone mass in underprivileged Indian premenarchal girls. Bone 2012; 51: 1-7.

Khoo AL, Chai LY, Koenen HJ, Sweep FC, Joosten I, Netea MG, van der Ven AJ. Regulation of cytokine responses by seasonality of vitamin D status in healthy individuals. Clin Exp Immunol 2011; 164:72-9.

Kim K, Valentine RJ, Shin Y, Gong K. Associations of visceral adiposity and exercise participation with C-reactive protein, insulin resistance, and endothelia dysfunction in Korean healthy adults. Metabolism 2008; 57: 1181-9.

Kimm SYS, Glynn NW, Aston CE, Poehlman ET, Daniels SR. Effects of Race, Cigarette Smoking, and Use of Contraceptive Medications on Resting Energy Expenditure in Young Women. Am J Epidemiol 2001; 154: 718-24.

Kong H, Qian YS, Tang XF, Zhang J, Gao PJ, Zhang Y, Zhu DL. C-reactive protein (CRP) gene polymorphisms, CRP levels and risk of incident essential hypertension: findings from an observational cohort of Han Chinese. Hypertens Res 2012; 35: 1019-23.

Konradsen S, Ag H, Lindberg F, Hexeberg S, Jorde R. Serum 125- dihydroxy vitamin D is inversely associated with body mass index. Eur J Nutr 2008; 47: 87-91.

Kremer JM. N-3 fatty acid supplements in rheumatoid arthritis. Am J Clin Nutr 2000; 71: 349S-51S.

Kremer R, Campbell PP, Reinhardt T, Gilsanz V. Vitamin D status and its relationship to body fat, final height, and peak bone mass in young women. J Clin Endocrinol Metab 2009; 94: 67-73.

Kris-Etherton P, Eckel RH, Howard BV, St. Jeor S, Bazzare TL. AHA Science Advisory: Lyon Diet Heart Study. benefits of a mediterranean-style, National cholesterol education program/American Heart Association step in dietary pattern on cardiovascular disease. Circulation 2001; 103: 1823.

Krishnan AV, Feldman D. Mechanisms of the anti-cancer and anti-inflammatory actions of vitamin D. Annu Rev Pharmacol Toxicol 2011; 51: 311-36.

Laitinen K, Valimaki M, Lamberg-Allardt C, Kivisaari L, Lalla M, Karkkainen M, Ylikahri R. Deranged vitamin D metabolism but normal bone mineral density in Finnish noncirrhotic male alcoholics. Alcohol Clin Exp Res 1990; 14: 551-6.

354

Lanou AJ, Berkow SE, Barnard ND. Calcium, dairy Products, and Bone Health in Children and Young Adults: A reevaluation of the evidence. Pediatrics 2005;115:736-43.

Lau DC, Dhillon B, Yan H, Szmitko PE, Verma S. Adipokines: molecular links between obesity and atheroslcerosis. Am J Physiol Heart Circ Physiol 2005; 288: 2031- 41.

Lee CM, Huxley RR, Wildman RP, Woodward M. Indices of abdominal obesity are better discriminators of cardiovascular risk factors than BMI: a meta-analysis. J Clin Epidemiol 2008; 61: 646-53.

Lee JH, O'Keefe JH, Lavie CJ; Harris WS. Omega-3 fatty acids: cardiovascular benefits, sources and sustainability. Nat Rev Cardiol 2009; 6: 753-8.

Leonard WJ. W.E. Paul (Ed.), Fundamental Immunology (fourth ed.), Lippincott-Raven Publishers, Philadelphia (1999), pp. 741-74.

Li Y, Schellhorn HE. New developments and novel therapeutic perspectives for vitamin C. J Nutr 2007; 137: 2171-84.

Liel Y, Ulmer E, Shary J, Hollis BW, Bell NH. Low circulating vitamin D in obesity. Calcif Tissue Int 1988; 43: 199-201.

Lin YC, Lyle RM, McCabe LD, McCabe GP, Weaver CM, Teegarden D. Dairy calcium is related to changes in body composition during a two-year exercise intervention in young women. J Am Coll Nutr 2000; 19: 754-60.

Linos E, Keiser E, Kanzler M, Sainani KL, Lee W, Vittinghoff E, Chren MM, Tang JY. Sun protective behaviors and vitamin D levels in the US population: NHANES 2003-2006. Cancer Causes Control 2012; 23: 133-40.

Looker AC, Pfeiffer CM, Lacher DA, Schleicher RL, Picciano MF, Yetley EA. Serum 25-hydroxyvitamin D status of the U.S. population: 1988–1994 compared with 2000-2004. Am J Clin Nutr 2008; 88: 1519-27.

Loos RJF, Rankinen T, Leon AS, Skinner JS, Wilmore JH, Rao DC, Bouchard C. Calcium Intake Is Associated with Adiposity in Black and White Men and White Women of the HERITAGE Family Study. J Nutr 2004; 134: 1772-8.

Lopez-Garcia E, Schulze MB, Meigs JB, Manson JE, Rifai N, Stampfer MJ, Willett WC, Hu FB. Consumption of trans fatty acids is related to plasma biomarkers of inflammation and endothelial dysfunction. J Nutr 2005; 135: 562-6.

355

Lovejoy JC, Champagne CM, Smith SR, de Jonge L, Xie H. Ethnic differences in dietary intakes, physical activity, and energy expenditure in middle-aged, premenopausal women: the Healthy Transitions Study. Am J Clin Nutr 2001; 74: 90-5.

Lumb GA, Mawer EB, Stanbury SW. The apparent vitamin D resistance of chronic renal failure: a study of the physiology of vitamin D in man. Am J Med 1971; 50: 421- 44.

Maetani M, Maskarinec G, Franke AA, Cooney RV. Association of leptin, 25- hydroxyvitamin D, and parathyroid hormone in women. Nutr Cancer 2009; 61: 225-31.

Malham M, Peter Jorgensen S, Lauridsen AL, Ott P, Glerup H, Dahlerup JF. The effect of a single oral megadose of vitamin D provided as either ergocalciferol (D₂) or cholecalciferol (D₃) in alcoholic liver cirrhosis. Eur J Gastroenterol Hepatol 2012; 24: 172-8.

Mao XQ, Kawai M, Yamashita T, Enomoto T, Dake Y, Sasaki S, Kataoka Y, Fukuzumi T, Endo K, Sano H, Aoki T, Kurimoto F, Adra CN, Shirakawa T, Hopkin JM. Imbalance production between interleukin-1beta (IL-1ß) and IL-1 receptor antagonist (IL-1ra) in bronchial asthma. Biochem Biophys Res Commun 2000; 276: 607-62.

Marantes I, Achenbach SJ, Atkinson EJ, Khosla S, Melton LJ 3rd, Amin S. Is vitamin D a determinant of muscle mass and strength? J Bone Miner Res 2011; 26: 2860- 71.

Marchant A, Bruyns C, Vandenabeele P, Ducarme M, Gerard C, Delvaux A, De Groote D, Abramowicz D, Velu T, Goldman M. Interleukin-10 controls interferon-γ and tumor necrosis factor production during experimental endotoxemia. Eur J Immunol 1994; 24: 1167-71.

Maret W, Sandstead HH. Zinc requirements and the risks and benefits of zinc supplementation. J Trace Elem Med Biol 2006; 20: 3-18.

Marnell L, Mold C, Du Clos T. C-reactive protein: receptors and role in inflammation. Clin Immunol 2005; 117: 104-11.

Martin AD, Drinkwater DT. Variability in the measures of body fat, assumptions or technique? Sports Med 1991; 11:277-88.

356

Martins D, Wolf M, Pan D, Zadshir A, Tareen N, Thadhani R, Felsenfeld A, Levine B, Mehrotra R, Norris K. Prevalence of cardiovascular risk factors and the serum levels of 25-hydroxyvitamin D in the United States: data from the Third National Health and Nutrition Examination Survey. Arch Intern Med 2007; 167: 1159-65.

Marques-Vidal P, Bochud M, Bastardot F, Lüscher T, Ferrero F, Gaspoz JM, Paccaud F, Urwyler A, von Känel R, Hock C, Waeber G, Preisig M, Vollenweider P. Association between inflammatory and obesity markers in a Swiss population- based sample (CoLaus Study). Obes Facts 2012; 5: 734-44.

Massey LK, Whiting SJ. Caffeine, urinary calcium, calcium metabolism, and bone. J Nutr 1993; 123: 1611-4.

Mattila PH, Piironen VI, Uusi-Rauva EJ, Koivistoinen PE. Vitamin D contents in edible mushrooms. J Agric Food Chem 1994; 42: 2449-53.

Mauriege P, Galitzky J, Berlan M, Lafontan M. Heterogeneous distribution of beta and alpha-2 adrenoceptor binding sites in human fat cells from various fat deposits: functional consequences. Eur J Clin Invest 1987; 17: 156-65.

Mawer EB, Backhouse J, Holman CA, Lumb GA, Stanbury SW. The distribution and storage of vitamin D and its metabolites in human tissues. Clin Sci 1972; 43: 414- 31.

Mayne, ST. Antioxidant nutrients and chronic disease: use of biomarkers of exposure and oxidative stress status in epidemiologic research. J Nutr 2003; 133: 933S-40S.

McGregor GP, Biesalski HK. Rationale and impact of vitamin C in clinical nutrition. Curr Opin Clin Nutr Metab Care 2006; 9: 697-703.

Melamed ML, Michos ED, Post W, Astor B. 25-hydroxyl Vitamin D Levels and the Risk of Mortality in the General Population. Arch Intern Med 2008; 168: 1629-37.

Menant JC, Close JC, Delbaere K, Sturnieks DL, Trollor J, Sachdev PS, Brodaty H, Lord SR. Relationships between serum vitamin D levels, neuromuscular and neuropsychological function and falls in older men and women. Osteoporos Int 2012; 23: 981-9.

Meydani SN, Barnett JB, Dallal GE, Fine BC, Jacques PF, Leka LS, Hamer DH. Serum zinc and pneumonia in nursing home elderly. Am J Clin Nutr 2007; 86: 1167-73.

Millen AE, Bodnar LM. Vitamin D assessment in population-based studies: a review of the issues. Am J Clin Nutr 2008; 87: 1102S-5S.

357

Miller GD, Jarvis JK, McBean LD. Handbook of dairy foods and nutrition. Third Edition (2007). McBean CRC Press. Boca Raton, FL.

Mohamed-Ali V, Goodrick S, Rawesh A, Katz DR, Miles JM, Yudkin JS, Klein S, Coppack SW. Subcutaneous adipose tissue releases interleukin-6, but not tumor necrosis factor-alpha, in vivo. J Clin Endocrinol Metab 1997; 82: 4196– 200.

Moore CE, Murphy MM, Holick MF. Vitamin D intakes by children and adults in the United States differ among ethnic groups. J Nutr 2005; 135: 2478-85.

Moriseset AS, Huot C, Legare D. Tcernof A. Circulating IL-6 concentrations and abdominal adipocyte is isoproterenol-stimulated lipolysis in women. Obesity 2008; 16: 1487-92.

Morris K, Wang Y, Kim S, Moustaid-Moussa N. Dietary and hormonal regulation of the mammalian fatty acid synthase gene. Moustaid-Moussa, N. Berdanier, C. D. eds. Nutrient-Gene Interactions in Health and Disease, 2001. CRC Press Boca Raton, FL.

Moschonis G, Tanagra S, Koutsikas K, Nikolaidou A, Androutsos O, Manios Y. Association between serum 25-hydroxyvitamin D levels and body composition in postmenopausal women: the Postmenopausal Health Study. Menopause 2009; 16:701-7.

Muhlhofer A, Mrosek S, Schlegel B, Trommer W, Rozario F, Bohles H, Schremmer D, Zoller WG, Biesalski HK. High-dose intravenous vitamin C is not associated with an increase of pro-oxidative biomarkers. Eur J Clin Nutr 2004; 58: 1151-8.

Mullally, M, Meisel, H, Fitzgerald, R. Antiotensin-I-converting enzyme inhibitory activities of gastric and pancreatic proteinase digests of whey proteins. Int Dairy J 1997; 2: 299-303.

National Heart, Lung, and Blood Institute and the North American Association for the Study of Obesity. The Practical Guide: Identification, Evaluation, and Treatment of Overweight and Obesity in Adults (2000). NIH Publication Number 00-4084. Available online: http://www.nhlbi.nih.gov/guidelines/obesity/prctgd_c.pdf (accessed 30th March, 2013).

National Institute of Standards and Technology. NIST releases vitamin D standard reference material, 2009. Available at http://ts.nist.gov/measurementservices/referencematerials.

358

Need AG, Morris HA, Horowitz M, Nordin C. Effects of skin thickness, age, body fat, and sunlight on serum 25-hydroxyvitamin D. Am J Clin Nutr 1993; 58: 882-5.

Nguyen TV, Center JR, Eisman JA. Osteoporosis in Elderly Men and Women: Effects of Dietary Calcium, Physical Activity, and Body Mass Index. J Bone Miner Res 2000; 15: 322-31.

Nielsen S, Guo Z, Johnson CM, Hensrud DD, Jensen MD. Splanchnic lipolysis in human obesity. J Clin Invest 2004; 113: 1582-8.

Norman AW, Bouillon R, Whiting SJ, Vieth R, Lips P. 13th Workshop consensus for the vitamin D nutritional guidelines. J Steroid Biochem Mol Bio 2007; 103:204-5.

Norman AW, Henry HH. Vitamin D. In: Bowman BA, Russell RM, eds. Present Knowledge in Nutrition, 9th ed. Washington DC: ILSI Press, 2006.

Obarzanek E, Schreiber GB, Crawford PB, Goldman SR, Barrier PM, Frederick MM, Lakatos E. Energy intake and physical activity in relation to indexes of body fat: the National Heart, Lung, and Blood Institute Growth and Health Study. Am J Clin Nutr 1994; 60: 15-22.

O'Dell JR. Anticytokine therapy-a new era in the treatment of rheumatoid arthritis? N Engl J Med 1999; 340: 310-12.

Okamoto M, Misunobu F, Ashida K, Mifune T, Hosaki Y, Tsugeno H, Harada S, Tanizaki Y. Effects of dietary supplementation with n-3 fatty acids compared with n-6 fatty acids on bronchial asthma. Int Med 2000; 39: 107-111.

Onimaru H, Ballanyi K, Homma I. Contribution of Ca2+ dependent conductances to membrane potential fluctuations of medullary respiratoryneurons of newborn rats in vitro. J Physiol 2003; 552: 727-41.

Opal SM, DePalo VA. Anti-inflammatory cytokines. Chest 2000; 117: 1162-72.

Opal, SM, Wherry, JC, Grint, P Interleukin-10: potential benefits and possible risks in clinical infectious diseases. Clin Infect Dis 1998; 27: 1497-507.

Ovesen L, Brot C, Jakobsen J. Food contents and biological activity of 25- hydroxyvitamin D: a vitamin D metabolite to be reckoned with? Ann Nutr Metab 2003; 47: 107-13.

Ozawa H, Aiba S, Nakagawa, Tagami H. Interferon-gamma and interleukin-10 inhibit antigen presentation by Langerhans cells for T helper type 1 cells by suppressing their CD80 (B7-1) expression. Eur J Immunol 1996; 26:648-52.

359

Packard RR, Lichtman AH, Libby P. Innate and adaptive immunity in atherosclerosis. Semin Immunopathol 2009; 31: 5-22.

Paolo JA, Lee LS, Wu B, Banks PA, Steen H, Conwell DL. Cytokine profiling of pancreatic fluid using the ePFT collection method in tandem with a multiplexed microarray assay. J Immunol Methods 2011; 369: 98-107.

Parikh SJ, Yanovski JA. Calcium intake and adiposity. Am J Clin Nutr 2003; 77: 281-7.

Parikh SJ, Edelman M, Uwaifo GI, Freedman RJ, Semega-Janneh M, Reynolds J, Yanovski JA. The relationship between obesity and serum 1,25-dihydroxy Vitamin D concentrations in healthy adults. J Clin Endocrinol Metab 2004; 89: 1196-9.

Park HS, Park JY, Yu R. Relationship of obesity and visceral adiposity with serum concentrations of CRP, TNF-α and IL-6. Diabetes Res Clin Pract. 2005; 69: 29- 35.

Paul, WE Interleukin-4: a prototypic immune-regulatory lymphokine. Blood 1991; 77: 1859-70.

Pepys MB, Hirschfield GM. C-reactive protein: a critical update. J Clin Invest 2003; 111: 1805-12.

Phillips KM, Ruggio DM, Horst RL, Minor B, Simon RR, Feeney MJ, Byrdwell WC, Haytowitz DB. Vitamin D and sterol composition of 10 types of mushrooms from retail suppliers in the United States. J Agric Food Chem 2011; 59: 7841-53.

Pickup J, Mattock M. Activation of the innate immune system as a predictor of cardiovascular mortality in Type 2 diabetes mellitus. Diabet Med 2003; 20: 723-6.

Pietrobon D, DiVirgilio F, Pozzan T. Structural and functional aspects of calcium homeostasis in eukaryotic cells. Eur J Biochem 1990; 193: 599-622.

Pihlanto-Leppala, A, Koskinen, P, Piilola, K, Tupasela, T, Korhonen, H. Antiotensin I- converting enzyme inhibitory properties of whey protein digests: concentration and characterization of active peptides. J. Dairy Res 2000; 96: 53-64.

Pischon T, Hankinson SE, Hotamisligil GS, Rifai N, Willett WC, Rimm EB. Habitual dietary intake of n-3 and n-6 fatty acids in relation to inflammatory markers among US men and women. Circulation 2003; 108: 155-60.

Pittas AG, Dawson-Hughes B, Li T, Van Dam RM, Willett WC, Manson JE, Hu FB. Vitamin D and calcium intake in relation to type 2 diabetes in women. Diabetes Care 2006; 29: 650-6.

360

Poole KES, Loveridge N, Barker PJ, Halsall DJ, Rose C, Reeve J, Warburton EA. Reduced vitamin D in acute stroke. Stroke 2006; 37: 243-245.

Pradhan AD, Manson JE, Rifai N, Buring JE, Ridker PM. C-reactive protein, interleukin 6, and risk of developing type 2 diabetes mellitus. JAMA 2001; 286: 327-34.

Prasad AS. Effects of zinc deficiency on Th1 and Th2 cytokine shifts. J Infect Dis 2000; 182: 62S-68S.

Prasad AS. Zinc: an overview. Nutrition 1995; 11: 93-99.

Preedy VR, Watson RR, Sherma Z. Dietary Components and Immune Function (Nutrition and Health). Totowa, NJ: Humana Press, 2010.

Ramagopalan SV, Heger A, Berlanga AJ, Maugeri NJ, Lincoln MR, Burrell A, Handunnetthi L, Handel AE, Disanto G, Orton SM, Watson CT, Morahan JM,Giovannoni G, Ponting CP, Ebers GC, Knight JC. A ChIP-seq defined genome-wide map of vitamin D receptor binding: associations with disease and evolution. Genome Res 2010; 20:1352-60.

Reboul E, Richelle M, Perrot E, Desmoulins-Malezet C, Pirisi V, Borel P. Bioaccessibility of carotenoids and vitamin E from their main dietary sources. J Agric Food Chem 2006; 54: 8749-55.

Rebuffe-Scrive M, Andersson B, Olbe L, Bjorntorp P. Metabolism of adipose tissue in intra-abdominal depots of non-obese men and women. Metabolism 1989; 38: 453- 8.

Ridker PM, Buring JE, Shih J, Matias M, Hennekens CH. Prospective study of c-reactive protein and the risk of future cardiovascular events among apparently healthy women. Circulation 1998; 98: 731-33.

Ridker PM, Rifai N, Stampfer MJ, Hennekens CH. Plasma concentration of interleukin-6 and the risk of future myocardial infarction among apparently healthy men. Circulation 2000; 101: 1767-72.

Ridker PM. High-sensitivity C-reactive protein: potential adjunct for global risk assessment in the primary prevention of cardiovascular disease. Circulation 2001; 103: 1813-8.

Rimm EB, Giovannucci EL, Stampfer MJ, Colditz GA, Litin LB, Willett WC. Reproducibility and validity of an expanded semi-quantitative food frequency questionnaire among male health professional. Am J Epidemiol 1992; 135: 1114- 26; discussion 1127-36.

361

Rink L, Gabriel P. Zinc and the immune system. Proc Nutr Soc 2000; 59: 541-52.

Rizzoli R, Bianchi ML, Garabedian M, McKay HA, Moreno LA. Maximizing bone mineral mass gain during growth for the prevention of fractures in the adolescents and the elderly. Bon. 2010; 46: 294-305.

Roche AF, Sievogel RM, Chumlea WC, Webb P. Grading body fatness from limited anthropometric data. Am J Clin Nutr 1981; 34: 2831-8.

Rodriguez-Rodriguez E, Navia B, Lopez-Sobaler AM, Ortega RM. Vitamin D in overweight/obese women and its relationship with dietetic and anthropometric variables. Obesity 2009; 17: 778-82.

Rolls BJ, Morris EL, Roe LS. Portion size of food affects energy intake in normal-weight and overweight men and women. Am J Clin Nutr 2002; 76: 1207-13.

Rosenstreich SJ, Rich C, Volwiler W. Deposition in and release of vitamin D3 from body fat: evidence for a storage site in the rat. J Clin Invest 1971; 50: 679-87.

Ross A. Vitamin A and Carotenoids. In: Shils M, Shike M, Ross A, Caballero B, Cousins R, eds. Modern Nutrition in Health and Disease. 10th ed. Baltimore, MD: Lippincott Williams & Wilkins; 2006. pp 351-75.

Ross CA. Vitamin A. In: Coates PM, Betz JM, Blackman MR, et al., eds. Encyclopedia of Dietary Supplements. 2nd ed. London and New York: Informa Healthcare; 2010. pp 778-91.

Rost NS, Wolf PA, Kase CS, Kelly-Hayes M, Silbershatz H, Massaro JM, D'Agostino RB, Franzblau C, Wilson PW. Plasma concentration of C-reactive protein and risk of ischemic stroke and transient ischemic attack: the Framingham study. Stroke 2001; 32: 2575-9.

Ruggiero C, Lattanzio F, Lauretani F, Gasperini B, Andres-Lacueva C, Cherubini A. Omega-3 polyunsaturated fatty acids and immune-mediated diseases: inflammatory bowel disease and rheumatoid arthritis. Curr Pharm Des 2009; 15: 4135-48.

Sagawa K, Mochizuki M, Sugita S, Nagai K, Sudo T, Itoh K. Suppression by IL-10 and IL-4 of cytokine production induced by two-way autologous mixed lymphocyte reaction. Cytokine 1996; 8:501-6.

Saggese G, Igli Baroncelli G. Nutritional aspects of calcium and vitamin D from infancy to adolescence. Ann Ist Super Sanita 1995; 31: 461-79.

362

Salekzamani S, Neyestani TR, Alavi-Majd H, Houshiarrad A, Kalayi A, Shariatzadeh N, Gharavi A. Is vitamin D status a determining factor for metabolic syndrome? A case-control study. Diabetes Metab Syndr Obes 2011; 4: 205-12.

Salvini S, Hunter DJ .Sampson L, Stampfer MJ, Colditz GA, Rosner B, Willet WC. Food-based validation of a dietary questionnaire: the effects of week-to-week variation in food consumption. Int J Epidemiol 1989; 18: 858-67.

Sampson HW. Alcohol and other factors affecting osteoporosis risk in women. Alcohol Res Health 2002; 26: 292-8.

Satoh K, Sakagami H. Effect of metal ions on radical intensity and cytotoxic activity of ascorbate. Anticancer Research 1997; 17: 1125-9.

Schleithoff SS, Zittermann A, Tenderich G, Berthold HK, Stehle P, Koerfer R. Vitamin D supplementation improves cytokine profiles in patients with congestive heart failure: a double-blind, randomized, placebo-controlled trial. Am J Clin Nutr 2006; 83: 754-9.

Schmidt MI, Duncan BB, Sharrett AR, Lindberg G, Savage PJ, Offenbacher S, Azambuja MI, Tracy RP, Heiss G. Markers of inflammation and prediction of diabetes mellitus in adults (Atherosclerosis Risk in Communities study): a cohort study. Lancet 1999; 353: 1649-52.

Schneider HJ, Friedrich N, Klotsche J, Pieper L, Nauck M, John U, Dorr M, Felix S, Lehnert H, Pittrow D, Silber S, Volzke H, Stalla GK, Wallaschofski H, Wittchen HU. The predictive value of different measures of obesity for incident cardiovascular events and mortality. J Clin Endocrinol Metab 2010; 95: 1777-85.

Schwartz B, Smirnoff P, Shany S, Liel Y. Estrogen controls expression and bioresponse of 1,25dihydroxyvitamin D recepto rs in the rat colon. Mol Cell Biochem 2000; 203: 87-93.

Scragg R, Sowers MF, Bell C. Serum 25-hydroxyvitamin D, diabetes, and ethnicity in the Third National Health and Nutrition Examination Survey. Diabetes Care 2004; 27: 2813-8.

Scragg R, Sowers MF, Bell C. Serum 25-hydroxyvitamin D, ethnicity, and blood pressure in the Third National Health and Nutrition Examination Survey. Am J Hypertens 2007; 20: 713-9.

Sen CK, Khanna S, Roy S. Tocotrienols: Vitamin E beyond tocopherols. Life Sciences 2006; 78: 2088-98.

363

Sesso HD, Buring JE, Christen WG, Kurth T, Belanger C, MacFadyen J, Bubes V, Manson JE, Glynn RJ, Gaziano JM. Vitamins E and C in the Prevention of Cardiovascular Disease in Men: The Physicians' Health Study II Randomized Trial. JAMA 2008; 300: 2123-33.

Shab-Bidar S, Neyestani TR, Djazayery A, Eshraghian MR, Houshiarrad A, Kalayi A, Shariatzadeh N, Khalaji N, Gharavi A. Improvement of vitamin D status resulted in amelioration of biomarkers of systemic inflammation in the subjects with type 2 diabetes. Diabetes Metab Res Rev 2012; 28: 424-30.

Shankar AH, Prasad AS. Zinc and immune function: the biological basis of altered resistance to infection. Am J Clin Nutr 1998; 68: 447S-63S.

Sheikh MS, Santa Ana CA, Nicar MJ, Schiller LR, Fordtran JS. Gastrointestinal absorption of calcium from milk and calcium salts. N Engl J Med 1987; 317: 532- 36.

Shi H, Norman AW, Okamura WH, Anintida S, Zemel MB. 1α, 25-dihydroxyvitamin

D3 inhibits uncoupling protein 2 expression in human adipocytes. FASEB J 2002; 16: 1808-10.

Shi H, Norman AW, Okamura WH, Sen A, Zemel MB. 1, 25-Dihydroxyvitamin D3 Modulates Human Adipocyte Metabolism via Nongenomic Action. FASEB J 2001; 15: 2751-3.

Shields M, Tremblay MS, Connor Gorber S, Janssen I. Abdominal obesity and cardiovascular disease risk factors within body mass index categories. Health Rep 2012; 23: 7-15.

Shoji T, Shinohara K, Kimoto E, Emoto M, Tahara H, Koyama H, Inaba M, Fukumoto S, Ishimura E, Miki T, Tabata T, Nishizawa Y. Lower risk for cardiovascular mortality in oral 1alpha-hydroxy vitamin D3 users in a haemodialysis population. Nephrol Dial Transplant 2004; 19: 179-84.

Sies H. Oxidative Stress: Oxidants and antioxidants. Exp Physiol 1997; 82: 291-5.

Siffrin V, Brandt AU, Herz J, Zipp F. New insights into adaptive immunity in chronic neuroinflammation. Adv Immunol 2007; 96: 1-40.

Simopoulos, A. The importance of the ratio of omega-6/omega-3 essential fatty acids. Biomed Pharmacother 2002; 56: 365-79.

Singh U, Devaraj S, Jialal I. Vitamin E, oxidative stress, and inflammation. Annu Rev Nutr 2005; 25: 151-74.

364

Solomons NW. Mild human zinc deficiency produces an imbalance between cell- mediated and humoral immunity. Nutr Rev 1998; 56: 27-8.

Solomons NW. Vitamin A. In: Bowman B, Russell R, eds. Present Knowledge in Nutrition. 9th ed. Washington, DC: International Life Sciences Institute; 2006.

Stephensen CB. Vitamin A, infection, and immune function. Annu Rev Nutr 2001; 21: 167-92.

Stipanuk MH. Biochemical, physiological, molecular aspects of human nutrition. 2nd Edition. Saunders Elsevier Inc, 2006.

Stubbs JR, Idiculla A, Slusser J, Menard R, Quarles LD. Cholecalciferol supplementation alters calcitriol-responsive monocyte proteins and decreases inflammatory cytokines in ESRD. J Am Soc Nephrol 2010; 21: 353-61.

Svendsen OL, Haarbo J, Hassager C, Christiansen C. Accuracy of measurements of body composition by dual-energy x-ray absorptiometry in vivo. Am J Clin Nutr 1993; 57: 605-8.

Svendsen OL, Hassager C, Skodt V, Christiansen C. Impact of soft tissue on in vivo accuracy of bone mineral measurements in the spine, hip, and forearm: A human cadaver study. J Bone Miner Res 1995; 10: 868-73.

Swardfager W, Herrmann N, Cornish S, Mazereeuw G, Marzolini S, Sham L, Lanctot KL. Exercise intervention and inflammatory markers in coronary artery disease: a meta-analysis. Am Heart J 2012; 163: 666-76.

Tak PP, Zvaifler NJ, Green DR, Firestein GS. Rheumatoid arthritis and p53: how oxidative stress might alter the course of inflammatory diseases. Immunol Today 2000; 21: 78-82.

Tangpricha V, Pearce EN, Chen TC, Holick MF. Vitamin D Insufficiency among Free- Living Healthy Young Adults Am J Med. 2002; 112: 659-62.

Tayal V, Kalra BS. Cytokines and anti-cytokines as therapeutics: an update. Eur J Pharmacol 2008; 579: 1-12.

Taylor RW, Keil D, Gold EJ, Williams SM, Goulding A. Body mass index, waist girth, and waist-to-hip ratio as indexes of total and regional adiposity in women, evaluation using receiver operating characteristic curve. Am J Clin Nutr 1988; 67: 44-9.

365

te Velde AA, Huijbens RJF, de Vries JE, Figdor CG. Interleukin-4 (IL-4) inhibits secretion of IL-1β, tumor necrosis factor α, and human IL-6 by human monocytes. Blood 1990; 76: 1392-7.

Tepper RI, Coffman RL, Leder P. An eosinophil-dependent mechanism of the antitumor effect of IL-4. Science 1992; 257: 548-51.

Terry P, Lichtenstein P, Feychting M, Ahlbom A, Wolk A. Fatty fish consumption and risk of prostate cancer. Lancet 2001; 357: 1764-66.

Thomas SR, Stocker R. Molecular action of vitamin E in lipoprotein oxidation: implications for atherosclerosis. Free Radic Biol Med 2000; 28: 1795-805.

Tidwell DK, Valliant MW. Higher amounts of body fat are associated with inadequate intakes of calcium and vitamin D in African American women. Nutr Res 2011; 31: 527-36.

Toi M, Bicknell R, Harris AL. Inhibition of colon and breast carcinoma cell growth by interleukin-4. Cancer Res.1992; 52: 275-9.

Toi M, Harris AL, Bicknell R. Interleukin-4 is a potent mitogen for capillary endothelium. Biochem Biophys Res Commun 1991; 174: 1287-93.

Tothill P, Avenell A, Love J, Reid DM. Comparisons between Hologic, Lunar and Norland dual-energy X-ray absorptiometers and other techniques used for whole- body soft tissue measurements . Eur J Clin Nutr 1994; 48: 781-94.

Traber M G, Atkinson J. Vitamin E, antioxidant and nothing more Free Radic Biol Med 2007; 43: 4-15.

Traber MG. Vitamin E regulatory mechanisms. Annu Rev Nutr 2007; 27: 347-62.

Traber MG. Vitamin E. In: Shils ME, Shike M, Ross AC, Caballero B, Cousins R, eds. Modern Nutrition in Health and Disease. 10th ed. Baltimore, MD: Lippincott Williams & Wilkins, 2006. pp 396-411.

Turner RT. Skeletal response to alcohol. Alcohol Clin Exp Res 2000; 24: 1693-701.

U.S. Department of Agriculture, Agricultural Research Service. 2011. USDA National Nutrient Database for Standard Reference, Release 24. Nutrient Data Laboratory Home Page,http://www.ars.usda.gov/ba/bhnrc/ndl.

Uchitel OD, Protti DA, Sanchez V, Cherksey BD,Sugimori M, llinas R.P-type voltage- dependent calcium channel mediates presynaptic calcium influx and transmitter release in mammalian synapses. Proc NatI Acad Sci 1992; 89: 3330-33.

366

Valko M, Leibfritz D, Moncol J, Cronin M, Mazur M, Telser J. Free radicals and antioxidants in normal physiological functions and human disease. Int J Biochem Cell Biol 2007; 39: 44-84.

Van der Wielen RPJ, Lowik MRH, van den Berg H, de Groot LC, Haller J, Moreiras O, van Staveren WA. Serum vitamin D concentrations among elderly people in Europe. Lancet 1995; 346: 207-10.

Van Dullemen HM, Van Deventer SJH, Hommes DW, Bijl HA, Jansen J, Tytgat GNJ, Woody J. Treatment of Crohn's disease with anti-tumor necrosis factor chimeric monoclonal antibody (cA2). Gastroenterology 1995; 109: 129-35.

Vannier E, Miller MC, Dinarello, CA. Coordinated anti-inflammatory effects of interleukin4: interleukin 4 suppresses interleukin 1 production but up-regulates gene expression and synthesis of interleukin 1 receptor antagonist. Proc Natl Acad Sci USA 1992; 89: 4076-80.

Varghese B, Swamy S, Srilakshmi MA, Santhosh MJ, Shetty GG, Varghese K, Patil CB, Iyengar SS. Visceral adiposity in young patients with coronary artery disease-a case control study. Indian Heart J 2012; 64: 284-9.

Verhagen H, Buijsse B, Jansen E, Bueno-de-Mesquita B. The state of antioxidant affairs. Nutr Today 2006; 41: 244-50.

Vieth R. Why the optimal requirement for vitamin D3 is probably higher than what is officially recommended for adults. J Steroid Biochem Mol Biol 2004; 89: 575-9.

Vilarrasa N, Maravall J, Estepa A, Sanchez R, Masdevall C, Navarro MA, Alia P, Soler J, Gomez JM. Low 25-hydroxyvitamin D concentrations in obese women: their clinical significance and relationship with anthropometric and body composition variables. J Endocrinol Invest 2007; 30: 653-8.

Vozarova B, Weyer C, Hanson K, Tataranni PA, Bogardus C, Pratley RE. Circulating interleukin-6 in relation to adiposity, insulin action, and insulin secretion. Obes Res 2001; 9 :414-7.

Wang Y, DeLuca HF. Is the vitamin d receptor found in muscle? Endocrinology 2011; 152:354-63.

Wang P, Wu P, Siegel ML, Egan RW, Billah MM. Interleukin (IL)-10 inhibits nuclear factor κB (NF-κB) activation in human monocytes: IL-10 and IL-4 suppress cytokine synthesis by different mechanisms. J Biol Chem 1995; 270: 9558-63.

Wang X, Quinn PJ. Vitamin E and its function in membranes. Prog Lipid Res 1999; 38: 309-36.

367

Wang Y, Zhu J, DeLuca HF. Where is the vitamin D receptor? Arch Biochem Biophys 2012; 523: 123-33.

Weaver CM, Heaney RP, Martin BR, Fitzsimmons ML. Human calcium absorption from whole-wheat products. J Nutr 1991; 121: 1769-75.

Weaver CM, Heaney RP. Isotopic exchange of ingested calcium between labeled sources: evidence that ingested calcium does not form a common absorptive pool. Calcif Tissue Int 1991; 49: 244-7.

Weaver CM, Plawecki KL. Dietary calcium adequacy of a vegetarian diet. Am J Clin Nutr 1994; 59: 1238S-41S.

Weaver CM, Proulx WR, Heaney RP. Choices for achieving adequate dietary calcium with a vegetarian diet. Am J Clin Nutr 1999; 70: 543S-8S.

Webb AR, DeCosta BR, Holick MF. Sunlight regulates the cutaneous production of vitamin D3 by causing its photodegradation. J Clin Endocrinol Metab 1989; 68: 882-7.

Weinstein M, Babyn P, Zlotkin S. An orange a day keeps the doctor away: scurvy in the year 2000. Pediatrics 2001; 108: E55.

Weinstock-Guttman B, Baier M, Park Y Feichter J, Lee-Kwen P, Gallagher E, Venkatraman J, Meksawan K, Deinehert S, Pendergast D, Awad AB, Ramanathan M, Munschauer F, Rudick R. Low fat dietary intervention with omega-3 fatty acid supplementation in multiple sclerosis patients. Prostaglandins Leukot Essent Fatty Acids 2005; 73: 397-404.

Wellens RI, Roche AF, Khamis HJ, Jackson AS, Pollock ML, Siervogel RM. Relationships between the body mass index and body composition. Obes Res 1996; 4: 35-44.

Wharton B, Bishop N. Rickets. Lancet 2003; 362: 1389-400.

Willet WC, Sampson L, Stampler MJ, Rosner B, Bain C, Witschi J, Hennekens CH, Speizer FE. Reproducibility and validity of a semi-quantitative food frequency questionnaire. Am J Epidemiol 1985; 122, 51-65.

Willett W. Nutritional epidemiology. 2nd Edition. New York, New York: Oxford University Press, 1998. pp74-156.

Williams JK. Rationale for new oral contraceptive dosing. Int J Fertil Womens Med 2004; 49: 30-5.

368

Wintergerst ES, Maggini S, Hornig DH. Contribution of selected vitamins and trace elements to immune function. Ann Nutr Metab 2007; 51: 301-23.

Wolf AM, Hunter DJ, Colditz GA, Manson JE, Stamoer MJ, Corsano KA, Rosner B, Kriska A, Willet WC. Reproducibility and validity of a self-administered physical activity questionnairee. Int J Epidemiol 1994: 23: 991-9.

Wolpowitz D, Gilchrest BA. The vitamin D questions: how much do you need and how should you get it? J Am Acad Dermatol 2006; 54: 301-17.

Wood RJ, Tchack L, Taparia S. 1,25-Dihydroxyvitamin D3 increases the expression of the CaT1 epithelial calcium channel in the Caco-2human intestinal cell line. BMC Physiol 2001; 1:11.

World Health Organization (2008). Available at http://www.who.int/mediacentre/factsheets/fs311/en/

World Health Organization. Preventing and Managing the Global Epidemic. WHO, Geneva, 2000. WHO Obesity Technical Report Series, Obesity; Number 894.

Wortsman J, Matsuoka LY, Chen TC, Lu Z, Holick MF. Decreased bioavailability of vitamin D in obesity. Am J Clin Nutr 2000; 72: 690-3.

Wysolmerski JJ, Insogna KL. The parathyroid glands, hypercalcemia, and hypocalcemia. In: Goldman L, Ausiello D, eds. Cecil Medicine. 23rd Edition. Philadelphia, Pa: Saunders Elsevier; 2007: chap 266.

Xu H, Barnes TG, Yang Q, Tan G, Yang D, Chou CJ, Sole J, Nichols A, Ross JS, Tartaglia LA, Chen H. Chronic inflammation in fat plays a crucial role in the development of obesity-related insulin resistance. J Clin Invest 2003; 112: 1821- 30.

Yudkin JS, Kumari M, Humphries SE, Mohamed-Ali V. Inflammation, obesity, stress and coronary heart disease: is interleukin-6 the link? Atherosclerosis 2000; 148: 209-14.

Zaman MA, Oparil S, Calhoun DA. Drugs targeting the renin-angiotensin-aldosterone system. Nat Rev Drug Discov 2002; 621-636.

Zelko I, Mariani T, Folz R. Superoxide dismutase multigene family: a comparison of the Cu-Zn-SOD (SOD1), Mn-SOD (SOD2), and EC-SOD (SOD3) gene structures, evolution, and expression. Free Radic Biol Med 2002; 33: 337-49.

369

Zemel MB, Richards J, Milstead A, Campbell P. Effects of calcium and dairy on body composition and weight loss in African American adults. Obes Res 2005; 13: 1218-25.

Zemel MB, Shi H, Greer B, Dirienzo D, Zemel PC. Regulation of adiposity by dietary calcium. FASEB J 2000; 14: 1132-8.

Zemel MB, Warren T, Milstead A, Morris K, Campbell P. Calcium and dairy acceleration of weight and fat loss during energy restriction in obese adults. Obes Res 2004; 12: 582-90.

Zemel MB, Zemel PC, Bryg R, Sowers, JR. Dietary Calcium-Induced Regression of Left Ventricular Hypertrophy. Am J Hypertens 1990; 3: 458-63.

Zemel MB. Dietary Calcium, Calcitrophic Hormones and Hypertension. Nutr Metab Cardiovasc Dis. 1994; 4:224-8.

Zemel MB. Mechanisms of dairy modulation of Adiposity. J Nutr 2003; 133: 252-6S.

Zemel MB. Role of dietary calcium and dairy products in modulating adiposity. Lipids 2003; 38: 139-46.

Zemel, MB. Calcium Modulation of Hypertension and Obesity: Mechanisms and Implications. J Am Coll Nutr 2001; 20: 428S-35S.

Zhu S, Wang Z, Heshka S, Heo M, Faith MS, Heymsfield SB. Waist circumference and obesity-associated risk factors among whites in the third National Health and Nutrition Examination Survey: clinical action thresholds. Am J Clin Nutr 2002; 76: 743-9.

Zhu Y, Mahon BD, Froicu M, Cantorna MT. Calcium and 1alpha, 25-dihydroxyvitamin D3 target the TNF-α pathway to suppress experimental inflammatory bowel disease. Eur J Immunol 2005; 35: 217-24.

Ziegler EE, Fomon SJ. Lactose enhances mineral absorption in infancy. J Pediatr Gastr Nutr 1983; 2: 288-94.

Zingg JM, Azzi A. Non-antioxidant activities of vitamin E. Curr Med Chem 2004; 11: 1113-33.

Zittermann A, Frisch S, Berthold HK, Gotting C, Kuhn J, Kleesiek K, Stehle P, Koertke H, Koerfer R. Vitamin D supplementation enhances the beneficial effects of weight loss on cardiovascular disease risk markers. Am J Clin Nutr 2009; 89: 1321-7.

370

Zurawski G, de Vries JE. Interleukin 13, an interleukin-4-like cytokine that acts on monocytes and B cells, but not on T cells. Immunol Today 1994; 15: 19-26.

371