UNIVERSITY of CALIFORNIA, IRVINE Dietary Patterns and Cardiometabolic Disease Risk DISSERTATION Submitted in Partial Satisfactio

Home , Dietary supplement, Dieting, MyPyramid

UNIVERSITY OF CALIFORNIA, IRVINE

Dietary Patterns and Cardiometabolic Disease Risk

DISSERTATION

submitted in partial satisfaction of the requirements for the degree of

DOCTOR OF PHILOSOPHY

in Epidemiology

Kristin Michelle Hirahatake

Dissertation Committee: Assistant Professor Andrew Odegaard, Chair Assistant Professor Luohua Jiang Adjunct Professor Nathan Wong

2018

TABLE OF CONTENTS

Page

LIST OF FIGURES iv

LIST OF TABLES v

ACKNOWLEDGEMENTS vii

CURRICULUM VITAE viii

ABSTRACT ix

INTRODUCTION 1

CHAPTER 1: The Natural History of Type 2 Diabetes 3 Normal Glucose Homeostasis 3 Pathogenesis of T2D 4 Epidemiology 9 Risk Factors 10 Diagnosis 13 Treatment 18 Prevention 22 Complications 24

CHAPTER 2: Dietary Patterns and Type 2 Diabetes Risk 26 Overview of Dietary Indices 29 Literature Review of Studies: Dietary Patterns and T2D Risk 32 Overview of Beverages and T2D Risk 36 Literature Review of Studies: Beverage Consumption and T2D Risk 38

CHAPTER 3: Type 2 Diabetes and Cardiovascular Disease 43 Natural History of Cardiovascular Disease 43 T2D and Cardiovascular Disease Risk 46 Dietary Patterns and Cardiovascular Disease Risk in T2D 51

CHAPTER 4: Populations Under Study 70 Coronary Artery Risk Development in Young Adults (CARDIA) 70 Women’s Health Initiative (WHI) 75

CHAPTER 5: Variations on Dietary Patterns and Risk of Incident Type 2 Diabetes 85 Mellitus in Young Adults: the CARDIA Study Introduction 86 Methods 87 Results 94

ii Discussion 98 Tables 104

CHAPTER 6: Artificially-sweetened and Sugar-sweetened Beverages and Risk of 119 Incident Type 2 Diabetes in Young Adults: the CARDIA Study Introduction 121 Methods 122 Results 128 Discussion 131 Tables 137

CHAPTER 7: Diet Quality and Cardiovascular Disease Risk in Postmenopausal 143 Women with Type 2 Diabetes: The Women’s Health Initiative Introduction 145 Methods 146 Results 153 Discussion 155 Tables 162

CHAPTER 8: Summary and Conclusions 176

REFERENCES 177

iii LIST OF FIGURES

Page

Figure 1.1 Normal Glucose Homeostasis 8

Figure 1.2 The Ominous Octet 8

Figure 3.1 Atherosclerosis Progression 45

iv LIST OF TABLES

Page

Table 1.1 Type 2 Diabetes Risk Factors 12

Table 1.2 Criteria for Type 2 Diabetes Diagnosis 15

Table 1.3 Categories of Prediabetes 17

Table 1.4 Type 2 Diabetes Medical Nutrition Therapy 20

Table 1.5 Glucose-lowering Medications for Type 2 Diabetes 21

Table 2.1 Cohort Studies of Dietary Patterns and Type 2 Diabetes 33

Table 2.2 Cohort Studies of Artificially Sweetened Beverages and Type 2 38

Diabetes

Table 4.1 Women’s Health Initiative Inclusion and Exclusion Criteria 77

Table 4.2 Women’s Health Initiative Participant Baseline Characteristics 80

Table 5.1 Baseline Characteristics of CARDIA Participants by Cumulative 104

Average Dietary Pattern Score

Table 5.2 Pearson’s Correlation Between Dietary Pattern Scores 108

Table 5.3 Association between Cumulative Average Dietary Patterns and 109

30-year Diabetes Risk in the CARDIA Study

Table 5.1S Description of Dietary Pattern Scores 110

Table 5.2S Description of CARDIA A Priori Diet Quality Score 111

Table 5.3S Nutrient Characteristics by 2015 DGA Diet Score Quartile 115

Table 5.4S Nutrient Characteristics by Paleolithic Score Quartile 116

Table 5.5S Nutrient Characteristics by Low Carbohydrate Score Quartile 117

Table 5.6S Nutrient Characteristics by Empty Calorie Score Quartile 118

v Table 5.7S Association Between Year 20 Percent Empty Calories and T2D 118

Risk in the CARDIA Study

Table 6.1 Baseline Characteristics of CARDIA Participants by Quintile 137

of Sweetened Beverage Intake

Table 6.2 Association Between Cumulative Average Sweetened Beverage 139

Intake and T2D Risk in the CARDIA Study

Table 6.3 Association Between Beverage Consumption Patterns Over 20 140

Years and T2D Risk in the CARDIA Study

Table 6.1S Characteristics of CARDIA Participants According to Quintile 141

of Total Sweetened Beverage Consumption

Table 6.2S Association of Frequency of Sweetened Beverage Consumption 142

and T2D Risk in the CARDIA Study

Table 7.1 Baseline Characteristics of WHI Participants with Prevalent T2D 162

by Dietary Pattern Score

Table 7.2 Pearson’s Correlation Between Dietary Pattern Scores 169

Table 7.3 Association Between Dietary Patterns and Incident CVD, Coronary 170

Heart Disease, and Stroke in Women with T2D in the WHI

Table 7.1S Nutrient Characteristics by DASH Diet Score Quintile 172

Table 7.2S Nutrient Characteristics by Mediterranean Diet Score Quintile 173

Table 7.3S Nutrient Characteristics by ADA Diet Score Quintile 174

Table 7.4S Nutrient Characteristics by Paleolithic Diet Score Quintile 175

vi ACKNOWLEDGEMENTS

I would like to express my most sincere appreciation to my committee chair,

Professor Andrew Odegaard, for his mentorship throughout my time as a graduate student. He continually challenged and encouraged me as I worked to master epidemiologic methods and refine my research and analytical skills. I am grateful for his willingness to help me pursue my goals and for his continued support along this journey. I would also like to thank my other dissertation committee members,

Professors Luohua Jiang and Nathan Wong, for their guidance and helpful commentary on my manuscripts. Lastly, I could not have completed this dissertation and graduate program without the ongoing support of my loving husband, Chris, and my wonderful family and friends.

vii CURRICULUM VITAE

Kristin Michelle Hirahatake

2008 B.S. Human Biology, University of California, San Diego

2010-2011 Teaching Assistant, University of California, Davis

2011 M.S. Nutritional Biology, University of California, Davis

2011-2012 Dietetic Internship, Veteran’s Affairs San Diego Healthcare

System, La Jolla

2016-2017 Teaching Assistant, University of California, Irvine

2018 PhD Epidemiology, University of California, Irvine

FIELD OF STUDY

Dietary Patterns and Cardiometabolic Disease Risk

PUBLICATIONS

Meissen, J. K., Hirahatake, K. M., Adams, S. H., & Fiehn, O. (2015). Temporal metabolomic responses of cultured HepG2 liver cells to high fructose and high glucose exposures. Metabolomics, 11(3), 707-721.

Hirahatake, K. M., Slavin, J. L., Maki, K. C., & Adams, S. H. (2014). Associations between dairy foods, diabetes, and metabolic health: potential mechanisms and future directions. Metabolism, 63(5), 618-627.

Hirahatake, K.M., Meissen, J.K., Feihn, O., Adams, S.H. Comparative Effects of Fructose and Glucose on Lipogenic Gene Expression and Intermediary Metabolism in HepG2 Liver Cells. PLoS ONE 6 (11): e26583.doi:10.1371/journal.pone.0026583.

viii ABSTRACT OF THE DISSERTATION

Dietary Patterns and Cardiometabolic Disease Risk By

Kristin Michelle Hirahatake

Doctor of Philosophy in Epidemiology

University of California, Irvine, 2018

Assistant Professor Andrew Odegaard, Chair

Diet was recently identified as the largest contributing risk factor across the leading causes of death in the United States. The nature of the diet-disease relationship based on the current body of evidence, however, remains inconclusive. The objective of this dissertation is to further explore the diet-disease relationship, specifically focusing on type 2 diabetes mellitus

(T2D) and cardiovascular disease (CVD).

This dissertation consists of three separate research projects aiming to investigate the association between 1) contemporary national dietary guidelines and popular adapted dietary trends and 2) sweetened beverage consumption and T2D risk in young adults and 3) diet quality and CVD risk in postmenopausal women with prevalent T2D. In the first and second projects, prospective analyses of 4,627 young adults from the Coronary Artery Risk Development in

Young Adults (CARDIA) study with repeated dietary histories were conducted. The first project examined the association between a priori dietary pattern scores created to reflect the 2015

Dietary Guidelines for Americans Scientific Report, a modern-day Paleolithic diet, low carbohydrate diet, CARDIA A Priori Diet Quality Score and diet high in overall consumption of empty calories and T2D risk in young adults over time. The second project examined the association between artificially-sweetened beverage, sugar-sweetened beverage, and total sweetened beverage consumption and T2D risk in young adults in the CARDIA cohort. For the

ix third project, a prospective analysis of 5,809 postmenopausal women with prevalent T2D from the Women’s Health Initiative was performed to study the relationship between diet quality as measured by four dietary pattern scores, an alternate Mediterranean (aMed), Dietary Approaches to Stop Hypertension (DASH), Paleolithic, and the American Diabetes Association (ADA) recommendations, and CVD risk over time. Multivariable Cox proportional hazards regression was used to characterize the prospective associations with incident T2D for the first two projects and incident CVD for the third project.

Dietary pattern scores reflecting contemporary recommendations and popular patterns were not associated with T2D risk while the A Priori Diet Quality score, which largely aligns with the 2015 DGA, was strongly inversely associated with lower 30-year T2D risk in the

CARDIA cohort. Additionally, long-term total-sweetened beverage consumption was associated with an increased risk of T2D, independent of diet quality, lifestyle factors, and BMI in this population; suggesting that whether or not they are causally related to T2D, higher intake of these beverages serves as a strong predictor of higher T2D risk in young adults. The aMed,

DASH and ADA dietary patterns, emphasizing fruits, vegetables, whole grains, nuts/seeds, legumes and a high unsaturated:saturated fat ratio, were associated with a lower risk of incident

CVD in postmenopausal women with T2D.

x INTRODUCTION

Rising healthcare costs and the growing public health burden of cardiometabolic diseases necessitate a shift in healthcare delivery with more of an emphasis placed on interventions that target disease prevention for the population at large. The vast majority of financial and clinical resources are currently devoted to secondary and tertiary prevention of cardiometabolic conditions such as obesity, type 2 diabetes (T2D), and cardiovascular disease (CVD). Lifestyle factors, such as diet and physical activity, represent modifiable risk factors that are practical targets of prevention efforts for large populations. Diet was recently identified as the number one risk factor associated with the top 10 causes of death in the United States 1 and is an exposure that all individuals encounter repeatedly throughout the lifespan. Although a robust body of evidence has examined several aspects of the relationship between diet and cardiometabolic conditions, the results have often been inconsistent. One reason for these inconclusive findings may be that the reductionist approach of studying specific nutrients or foods in isolation is inadequate to capture the more complex relationship between diet and disease risk. Dietary patterns have been increasingly recognized as the optimal approach to study diet-disease relationships because they are better suited to account for the synergistic effects of total diet on health, since individuals are not exposed to single nutrients or foods,2 and can be effectively translated into nutrition guidelines that may be more easily understood and implemented by the general population.3 Traditional approaches to study the relationship between diet and disease risk also do not shed light on how overall diet quality may associate with other cardiometabolic risk factors. Furthermore, much of the existing research on the diet-disease relationship is based on observational data with measures of diet at only one point in time, which does not account for how changes in diet quality or the length of exposure to a certain dietary pattern may influence

1 disease risk. Many of the longitudinal studies available have used older participants or individuals with multiple comorbidities who are already at increased risk for T2D or CVD, and therefore may not effectively account for how diet influences risk earlier in life, during potentially sensitive periods. The very description of T2D and CVD as chronic diseases suggests their etiology spans longer periods of time than more acute conditions. At the same time, research related to the relationship between diet and chronic disease is limited in certain high risk populations, such as postmenopausal women. Dietary intervention trials are expensive and can be burdensome for both researchers and participants, and are therefore often limited in duration and their ability to capture the true effect of diet as an exposure over the lifespan. To address these knowledge gaps, I developed the following specific aims for my dissertation research:

Aim 1: Examine the association between dietary pattern scores created to reflect the 2015

Dietary Guidelines for Americans Scientific Report, a modern-day Paleolithic diet, low carbohydrate diet, CARDIA A Priori diet quality score and diet high in overall consumption of empty calories and T2D risk in young adults over time.

Aim 2: Examine the association between artificially-sweetened beverage, sugar-sweetened beverage, and total sweetened beverage consumption and T2D risk in young adults in the

CARDIA cohort over 30 years.

Aim 3: Examine the relationship between diet quality, as measured by aMed, DASH, ADA and

Paleolithic a priori dietary pattern scores, and CVD risk over time in postmenopausal women with T2D from the Women’s Health Initiative.

2 CHAPTER 1 The Natural History of Type 2 Diabetes Mellitus

Type 2 diabetes mellitus (T2D) is a complex metabolic disorder of energy substrate homeostasis. Defective insulin secretion and action lead to multiple metabolic abnormalities, including hyperglycemia due to impaired insulin-stimulated glucose uptake and uncontrolled hepatic glucose production, and dyslipidemia, the perturbed homeostasis of fatty acids, triglycerides and lipoproteins. The economic and public health burdens of this disease are rapidly rising worldwide, and efforts to improve both prevention and management are of critical importance.

Normal Glucose Homeostasis

Under normal physiological conditions, blood glucose homeostasis is maintained through an intricate neurohormonal feedback loop, as shown in Figure 1.1.4 In the basal or fasted state, glucose is supplied to the body through endogenous glucose production. The liver produces 85% of basal glucose through glycogenolysis and gluconeogenesis, while the remaining 15% is produced by the kidneys. During this phase, glucose disposal mainly occurs in insulin- independent tissues; approximately 50% by the brain and 25% by the splanchnic area (liver and gastrointestinal tissues). The remaining 25% is used by insulin-dependent tissues; primarily muscle and to a lesser extent adipose tissue.5 During this state, glucagon, a hormone produced by the alpha-cells of the pancreas, helps to regulate and maintain hepatic glucose production (HGP) to match glucose disposal rates.5,6 Following the ingestion of a carbohydrate-containing meal, insulin is secreted by the pancreatic β cells in response to increased postprandial plasma glucose concentrations and incretin hormones secreted from the digestive tract. Circulating insulin inhibits glucagon secretion and stimulates glucose uptake by the splanchnic and peripheral tissues.5 In the liver, these actions results in the suppression of endogenous glucose production

3 while increasing hepatic glucose uptake and storage through glycogen synthesis.7 The majority of postprandial glucose disposal occurs in muscle (80-85%), with adipose tissue accounting for only a small percentage (4-5%).5 In muscle and adipose tissue, insulin activates complex post- receptor signal transduction pathways that recruit glucose transport proteins to the cell surface to facilitate glucose uptake from the blood stream.8 Insulin signaling also regulates the metabolism of other energy substrates. In muscle, insulin promotes amino acid uptake and inhibits protein breakdown, and to a lesser extent stimulates new protein synthesis.9,10 In adipose tissue, insulin inhibits lipolysis and promotes the synthesis and storage of triglycerides.11–13

Pathogenesis of Type 2 Diabetes

Type 2 diabetes is a complex metabolic disorder characterized by perturbations in energy substrate metabolism and glucose homeostasis resulting from impaired insulin secretion and action.5,14 The natural history of T2D has been well studied and characterized in many diverse ethnic populations.5,15 A combination of genetic and lifestyle factors, mainly obesity and physical inactivity, lead to the development of insulin resistance in susceptible individuals.15–18

This is manifested in the liver by an overproduction of glucose during the basal state and impaired suppression of HGP in response to insulin the postprandial state.19–21 In muscle tissue, impaired intracellular signaling in response to insulin decreases glucose uptake following the consumption of a carbohydrate-containing meal, resulting in postprandial hyperglycemia.21,22 In response to these perturbations, the β cells of the pancreas must increase the amount of insulin secreted to overcome the insulin resistance and maintain normal glucose tolerance.22

Over time, β cell dysfunction leads to inadequate in insulin production and secretion to offset the level insulin resistance. Increasing age, genetic predisposition, insulin resistance, lipotoxicity (discussed below) and glucotoxicity have all been shown to contribute to β cell

4 failure.22 Although the exact mechanism remains unknown, insulin resistance is thought to play a role in β cell dysfunction as a result of the increased demand it places on the pancreas to upregulate insulin secretion. In addition, glucotoxicity, defined as chronically elevated plasma glucose concentrations, also impairs β cell function and insulin secretion.22 While evidence suggests that the relative contributions of insulin resistance and β cell failure to the development of T2D likely differs among different ethnic groups, the onset and rate of β cell failure determines the progression of hyperglycemia.23 Ultimately, this leads to a rise in both postprandial and fasting plasma glucose concentrations and the onset of overt diabetes.22 Over the past decade, a large body of in vitro and in vivo research has demonstrated that these abnormalities are even more intricate and complex than originally thought. As shown in Figure

1.2, in addition to insulin sensitive tissues and pancreatic β cells, multiple other organ systems including the gastrointestinal (GI) tract, α cells of the pancreas, kidney and brain are involved in the pathogenesis of T2D.22

Although adipose tissue uses only a small percentage of total postprandial blood glucose, it plays an important role in overall glucose homeostasis through its regulation of lipolysis and the release of adipocytokines that directly influence the insulin sensitivity of muscle and liver.5

Alterations in adipocyte metabolism and fat topography have been implicated in the pathogenesis of glucose intolerance in T2D.22 Insulin resistant adipocytes do not respond to the normal antilipolytic effect of insulin, resulting in extended elevations in plasma free fatty acid (FFA) concentration.20,24 Chronically increased plasma FFA levels lead to a series of metabolic abnormalities collectively referred to as lipotoxicity, which includes stimulation of hepatic gluconeogenesis, hepatic and muscle insulin resistance, and impaired insulin secretion.22 Insulin resistant adipocytes also have a reduced capacity to store triglycerides. As a result, excess lipid is

5 deposited into muscle, liver and β cells, often as toxic metabolites such as fatty acyl-coenzyme-

A, diacylglycerol and ceramide as opposed to inert triglycerides, thereby exacerbating muscle and hepatic insulin resistance and impairing insulin secretion.22 Furthermore, metabolically impaired adipocytes produce excessive amounts of adipocytokines associated with insulin resistance and inflammation and inadequate amounts of insulin-sensitizing cytokines such as adiponectin.25,26 Clearly, adipocyte insulin resistance and metabolic alterations play a significant role in the development of T2D.

Functional alterations in several other organs involved in glucose metabolism and homeostasis have also been implicated in the development of glucose intolerance. It has been known for some time that glucagon production by pancreatic α cells during the fasted state is elevated in individuals with T2D. The resulting increase in basal HGP is a major contributor to the fasting hyperglycemia seen in T2D.27,28 Furthermore, glucagon secretion is inadequately suppressed in the postprandial state in individuals with T2D.29 The tissues of the GI tract have also been shown to play an important role in the pathophysiology of T2D. As mentioned above and depicted in Figure 1, the small intestine secretes two important incretin hormones, glucagon- like peptide 1 (GLP-1) and gastric inhibitory polypeptide (GIP), to augment insulin secretion in response to ingested glucose.30 In individuals with impaired glucose tolerance, there is a deficiency in GLP-1 secretion along with resistance to its stimulatory effect on insulin secretion and inhibitory effect on glucagon secretion, resulting in impaired HGP suppression in the postprandial state.31,32 GIP secretion, on the other hand, is increased in individuals with T2D but

β cells are resistant to its stimulatory effect on insulin secretion.33,34 In addition to a small contribution to basal glucose production through gluconeogenesis, under normal conditions the kidneys reabsorb filtered glucose through transporters in the proximal tubule, preventing it from

6 being lost in the urine.35 Recent evidence suggests that the maximal renal tubular reabsorptive capacity for glucose is increased in diabetes, inhibiting the ability of the kidneys to excrete excess glucose in the urine to correct hyperglycemia.22

Lastly, although glucose uptake by the brain is an insulin-independent process, animal models have suggested that insulin acts as a strong appetite suppressant on the brain36,37 and affects central regulation of HGP and muscle glucose uptake.38,39 Two areas of the hypothalamus containing structures that play an important role in appetite regulation, the lower posterior hypothalamus (ventromedial nuclei) and upper posterior hypothalamus (paraventricular nuclei), normally show inhibition on imaging tests following glucose ingestion. However, in obese insulin-resistant individuals the magnitude of the inhibitory response appears to be diminished following glucose ingestion.40 It has therefore been postulated that the impaired appetite regulation and increased food intake seen in obese individuals, despite the presence of hyperinsulinemia, reflects insulin resistance in the brain and may further exacerbate the metabolic dysregulation seen in T2D .22 Such extensive metabolic disarray poses a clear challenge to the prevention and treatment of T2D. Further research is still needed to inform a multidimensional approach for T2D risk assessment, prevention, and treatment. Not surprisingly, a combination of medication and lifestyle interventions are often required to optimize blood glucose management and comorbidity risk reduction in individuals with T2D, discussed in further detail below.

Figure 1.1. Normal Glucose Homeostasis Fasted: In the fasted state, blood glucose concentration is maintained by endogenous glucose production, mainly from hepatic glycogenolysis and gluconeogenesis, and use by insulin-independent tissues, such as the brain. Insulin- sensitive tissues, such as muscle and adipose, use non-glucose nutrients (e.g., non-esterified fatty acids from lipolysis) for energy in this state. Fed: In the fed state, blood glucose concentrations rise due to absorption in the digestive tract, which stimulates insulin secretion by the β cells and suppresses glucagon secretion from α cells. The release of incretin hormones and glucagon-like peptide-1 (GLP-1) further increase glucose-stimulated insulin secretion and glucose-suppression of glucagon secretion. Insulin suppresses endogenous glucose production and promotes glucose uptake by insulin- sensitive peripheral tissues, such as heart, skeletal muscle, and adipose tissue to increase the rate of glucose disposal. Additionally, insulin stimulates amino acid and fatty acid uptake, suppresses adipose tissue lipolysis and promotes anabolic metabolism. Under normal physiological conditions, this process brings plasma glucose concentrations close to fasting levels within 2 hours of ingesting a carbohydrate load. *Figure from Nolan CJ, Damm P, Prentki M. Lancet. 2011;378(9786):169-181. **Text adapted from Nolan, et al. 2011 and Kahn, et al. 2014

Figure 1.2. The Ominous Octet The development of glucose intolerance in individuals with T2D is more complex than originally thought. In addition to the core pathophysiologic defects of muscle and liver insulin resistance and β cell failure, alpha cells of the pancreas, adipocytes, the GI tract, kidney, and brain also play a role in T2D pathogenesis. *Adapted from DeFronzo RA. Diabetes. 2009;58:773-795.

8 Epidemiology

Diabetes represents a major economic and public health burden. Annually, an estimated

$245 billion is spent on total medical costs and lost work and wages for individuals with diagnosed diabetes.41 According to the most recent report from the Centers for Disease Control

(CDC), 29.1 million people in the United States have diabetes, with T2D representing 90-95% of these cases. Diabetes disproportionately affects individuals by age, gender, and ethnicity.

According to U.S. census data, 25.9% of individuals 65 years of age and older have diabetes, whereas only 16.2% of individuals aged 45-64 and 4.1% of individuals aged 20-44 have diabetes. Diabetes prevalence is also slightly higher in men than women, 13.6% and 11.2% respectively. Between 2010 and 2012, the age-adjusted rate of diabetes diagnosis in people aged

20 years and older was 7.6% in non-Hispanic whites, 9.0% in Asian Americans, 12.8% in

Hispanics, 13.2% in non-Hispanic blacks, and 15.9% in American Indian/Alaska Natives.42

Globally, an estimated 422 million adults were living with diabetes in 2014. According to a World Health Organization (WHO) report, the age standardized global prevalence of diabetes in the adult population has nearly double since 1980, from 4.7% to 8.5% in 2014. Diabetes prevalence has been steadily increasing over the past three decades in high-income countries, while a more rapid increase in prevalence has been seen in low- and middle-income countries.

Substantial economic loss is experienced both on the individual level and by health systems and national economies through direct medical costs and loss of work and wages. In 2012, 1.5 million deaths were attributed to diabetes plus an additional 2.2 million deaths from the increased risk for cardiovascular and other diseases associated with diabetes.43 Despite growth in the understanding of the etiology of diabetes and advancements in treatment and management, diabetes and its complications continue to pose a major economic and public health burden

9 worldwide. Analysis from the CDC projects that if current trends continue, the number of

Americans with diabetes will double or triple by the year 2050, with as many as 1 in 3 adults in the U.S. having diabetes.42

Risk Factors

Type 2 diabetes is a highly heritable condition, but the rapid rise in prevalence over the past several decades cannot be accounted for by genetic risk alone. The etiology of the abnormalities present in T2D is complex and heterogeneous, involving both genetic and environmental factors. While multiple molecular and metabolic mechanisms have been postulated, the exact pathogenesis of T2D remains unclear and likely differs considerably among individuals.44 Table 1.1 identifies the factors currently known to increase T2D risk. As a greater understanding of the role of environmental factors in the development of T2D has emerged, headway has also been made in understanding how genetic factors contribute risk. For some time, the high concordance rates of T2D in monozygotic twins45 and substantially increased risk in individuals with affected first-degree relatives46 have suggested a highly heritable component of diabetes, currently estimated to be greater than 50%.47 A substantial body of evidence also shows different risk and prevalence rates of diabetes by race/ethnicity. Asians, Hispanics, and

African American individuals have a higher risk for developing T2D than non-Hispanic whites.48,49

Advances in analytical technology, such as genome-wide association studies, have identified genes linked with T2D risk, including 53 gene loci related to insulin and blood glucose concentration regulation and 50 loci associated with specifically with T2D. Interestingly, only 33 of these loci are shared between the risk factors and T2D outcome.50,51 The majority of the T2D risk loci identified have been linked with β cell function, and fewer with obesity and insulin

10 resistance.52 Despite such advances, these susceptibility loci are only able to explain about 10% of the heritability of T2D and the use of genotype risk scores are only slightly more sensitive than traditional clinical risk factors for the prediction of T2D.53–55

Multiple environmental factors have also been identified as strong contributing risk factors for T2D, with the predominant being obesity. Obesity, especially central or abdominal obesity, and weight gain are associated with insulin resistance, β cell dysfunction and T2D.44,56–58

Obesity itself has a heterogeneous etiology, influenced by both genetic determinants and lifestyle factors such as diet and physical activity.59–61 There are several proposed mechanisms through which obesity negatively impacts blood glucose regulation. Adipose tissue functions as an endocrine organ and the dysfunctional adipocytes associated with obesity release excessive amounts of insulin resistance-inducing, pro-inflammatory adipocytokines into circulation. This chronically elevated cytokine secretion results in systemic inflammation, which negatively affects liver and muscle metabolism and may also contribute to β cell dysfunction.25,26

Furthermore, as described above, the inability to suppress lipolysis in adipose tissue secondary to insulin resistance increases plasma free fatty acids and results in lipotoxicity.22 Several large, well-conducted intervention trials have demonstrated that even modest weight loss in overweight/obese individuals effectively reduces the risk of developing T2D.62,63

A combination of dietary factors including increased calorie intake, decreased energy expenditure and diet quality or dietary pattern have also been identified as important contributors to obesity, insulin resistance, β cell dysfunction and glucose intolerance and will be discussed in detail in a later chapter. In general, dietary patterns high in red and processed meats, sugar- sweetened beverages, and refined grains increase the risk of T2D, whereas consumption of a diet high in fruits, vegetables, nuts, and whole grains is associated with a reduced risk.64–69

11 In addition to diet, physical activity is strongly associated with obesity and T2D risk. A large body of research has provided evidence for an inverse relationship between regular moderate physical activity and reduced T2D risk,70 whereas a sedentary lifestyle with low physical activity reduces energy expenditure, may promote weight gain, and is associated with reduced insulin sensitivity.71,72 Other factors such as environmental toxins and contaminants,73–75 intestinal microbiota composition,76 medical conditions including gestational diabetes,77,78 cardiovascular disease,79,80 and depression81 and socioeconomic status82–84 have also been shown to contribute to T2D risk. A detailed review of the mechanisms of these relationships is beyond the scope of this dissertation. Current areas of research related to T2D risk and etiology include gene-environment interactions and the identification of metabolic substrate biomarkers using metabolomics and lipidomics that may serve as targets for prevention and/or monitoring.85

Table 1.1. Type 2 Diabetes Mellitus Risk Factors Clinical Factors • Family history of T2D • Ethnicity (Asian, Hispanic/Latino, African American, Alaska Native, American Indian, Pacific Islander) • Obesity • Body composition (central (abdominal) adiposity) Lifestyle Factors • Physical inactivity/sedentary lifestyle • Smoking • Sleep duration (short ≥5-6 hours/day & long >8-9 hours/day) Dietary Factors Increased Risk: • Western diet (high in red meat, processed meat, high fat dairy products, sweets, and desserts) • Sugar-sweetened beverages • Vitamin D deficiency Reduced Risk: • Mediterranean diet (high in fruits, vegetables, nuts, whole grains, and olive oil) • Dairy products • Nuts • Whole grains/cereal fiber Environmental Exposures • Inorganic arsenic in drinking water • Bisphenol A • Organophosphate & chlorinated pesticides Medical Factors • Gestational diabetes • Cardiovascular disease • Hyperuricemia

12 • Polycystic Ovarian Syndrome • Metabolic Syndrome • Depression Demographic Factors • Low socioeconomic status

Diagnosis

The current American Diabetes Association (ADA) criteria for diagnosing T2D are shown in Table 1.2. A clinical diabetes diagnosis may be based on plasma glucose criteria, either the fasting plasma glucose (FPG) or the 2-hour plasma glucose (2-h PG) value after a 75-gram oral glucose tolerance test (OGTT), or hemoglobin A1C (HbA1C) criteria. The diagnostic threshold of these tests is historically based on the shape of the risk curve between glucose levels and specific microvascular complications of diabetes.86

The FPG test is performed by obtaining a blood sample after a minimum 8 hour overnight fast from all food and drink (except water). In the absence of diabetes, fasting plasma glucose concentrations normally range from 60-99 mg/dl. A FPG  126 mg/dl on at least 2 separate occasions is used to support a clinical diagnosis of diabetes. Interestingly, the cut point for this test was modified in 1997 by the Expert Committee on the Diagnosis and Classification of Diabetes Mellitus when evidence from several epidemiological studies of retinopathy and glycemia showed that the original FPG cut point of 140 mg/dl identified much fewer cases of diabetes than the 2-h PG cut point.87 The FPG test is convenient with a simple blood draw and inexpensive assay, however many intra- and inter-individual factors can contribute to variability in FPG test results. An individual’s fasting plasma glucose concentrations fluctuate from day to day, and may be further altered by food intake, prolonged fasting, exercise, illness, acute stress and time of day.88 According to the Third National Health and Nutrition Examination Survey

(NHANES III) only 70.4% of people with FPG ≥126 mg/dl on the first test met the FPG criteria

13 when analysis was repeated 2 weeks later.89 Furthermore, differences in sample processing and analysis between laboratories can introduce substantial variability to results.88

The OGTT assesses the body’s ability to metabolize glucose. To carry out the test, a fasting individual (usually for at least 8 hours) consumes a solution containing 75 grams of glucose, blood glucose concentrations are then assayed from samples collected before drinking the solution and 2 hours after ingestion. Elevated postprandial plasma glucose, which primarily reflects decreased insulin sensitivity at the peripheral insulin-sensitive tissues usually occurs before fasting glucose increases. The OGTT is therefore considered a more sensitive indicator of diabetes risk and an earlier marker of impaired glucose homeostasis.88 However, the utility of this test is limited by poor reproducibility and even greater intra-individual variability than the

FPG89, high patient burden, cost and influence of numerous medical conditions and medications on the results.88

The HbA1C test is a measure of glycated hemoglobin that serves as a marker of estimated average plasma glucose concentrations over the preceding 2-3 months. Test results are reported as a percentage, reflecting the percent of hemoglobin in the blood sample with glucose bound to it; and validation efforts suggest that results are accurate within 0.5%. This test was added to the diagnostic panel for T2D in 2009.90 Unlike direct plasma glucose measurements, there is evidence to suggest that glycated hemoglobin is associated with an increased risk of both microvascular and macrovascular complications.91,92 The HbA1C test has several additional advantages over the FPG and OGTT, such as greater convenience in that it does not require fasting, greater pre-analytical stability, and less fluctuation during stress and illness.93 However, the accuracy of the test results may be limited by an individual’s race/ethnicity and medical conditions such as anemia, hemoglobinopathies, liver and kidney disease.90

14 Recent efforts to validate the effectiveness of these diagnostic tools have found an imperfect concordance between HbA1C and plasma glucose criteria, in part reflecting inter- individual and inter-ethnic differences in HbA1C formation. NHANES data has shown that the current HbA1C cut point of ≥6.5% identifies one-third fewer cases of undiagnosed diabetes than the FPG cut point.94 Correlations between FPG and HbA1C have ranged from 0.47-0.82 and between 2-h PG and HbA1C 0.49-0.78 and in general, FPG and HbA1C have been found to be less sensitive, but more specific than the 2-h PG for diagnosis diabetes.95 These discrepancies may be in part explained by the heterogeneous nature of T2D and the fact that the tests measure different components of blood glucose homeostasis. The ADA recommends that in the absence of unequivocal hyperglycemia with acute metabolic decompensation, results should be confirmed by repeat testing on a separate occasion.

Table 1.2. Criteria for the Diagnosis of Type 2 Diabetes Mellitus HbA1C ≥6.5% Performed in a laboratory using a method that is NGSP certified and standardized to the DCCT assay. OR FPG ≥126 mg/dl (7.0 mmol/L) Fasting is defined as no caloric intake for at least 8 hours. OR 2-h PG ≥200 mg/dl (11.1 mmol/L) during an OGTT The test should be performed as described by the WHO, using a glucose load containing the equivalent of 75 g anhydrous glucose dissolved in water. OR A random plasma glucose ≥200 mg/dl (11.1 mmol/L) In an individual with classic symptoms of hyperglycemia or hyperglycemic crisis.

Although not routinely done in the clinical setting, insulin sensitivity and β-cell function are often included in research related to diabetes and blood sugar control. Methods commonly used to obtain these measurements include the hyperinsulinemic-euglycemic clamp and homeostasis model assessment (HOMA). The hyperinsulinemic-euglycemic clamp provides information specific for tissue sensitivity to insulin. With this technique, an individual’s plasma insulin concentration is raised to approximately 100 U/ml above basal levels via peripheral

15 intravenous insulin infusion, glucose (usually in the form of a 20% solution) is then administered at a variable rate to maintain plasma glucose concentrations at a normal level (euglycemia).

Blood samples are typically collected every 5 minutes for analysis. Under these conditions, the glucose infusion rate is assumed to be equal to the whole-body glucose disposal rate (GDR) and reflects the amount of exogenous glucose required to fully compensate for the hyperinsulinemia.96,97 GDR is expressed as a function of metabolic body size, such as body weight, body surface area, or fat free mass. A GDR less than 4.7 mg/kg/min is generally interpreted as insulin resistance.98 This method has long been considered the gold standard for measuring insulin sensitivity, however the technique is demanding and expensive and therefore not feasible for most large-scale studies. In contrast, the HOMA, which can be used to assess both insulin resistance and β-cell function, is a much simpler method and therefore may be more appropriate for use in large epidemiological studies when rapid or repeated measurements of insulin sensitivity are needed. The HOMA method uses an idealized mathematical model to compute homeostatic insulin and glucose concentrations calibrated to give normal β-cell function of 100% and normal insulin resistance of 1. A single measurement of individual’s fasting plasma glucose and insulin concentrations can then be compared to the model’s predictions and an index of insulin resistance can be obtained. HOMA estimates have also been shown to correlate with euglycemic hyperinsulinemic clamp readings, however the clamp techniques may be more precise for small-scale use.99

Pre-diabetes is the term used to describe the intermediate stages of altered glucose metabolism when blood glucose levels are above normal ranges, but not elevated enough to meet the criteria for diabetes. The criteria for pre-diabetes diagnosis are shown in Table 1.3. This early risk-state can be further characterized by two different types of alterations in glucose

16 homeostasis: impaired glucose tolerance (IGT) and impaired fasting glucose (IFG).100 These early metabolic abnormalities differ both in the site of insulin resistance and pattern of insulin secretion. Individuals with isolated IFG predominantly have hepatic insulin resistance and normal muscle insulin sensitivity, while individuals with isolated IGT have normal to slightly reduced hepatic insulin sensitivity and moderate to severe muscle insulin resistance. In addition, those with IFG have a reduced early-phase (first 30 minutes) insulin secretory response to oral glucose, but a normal late-phase (60-120 minutes) insulin response during an oral glucose tolerance test (OGTT), resulting in a normal 2-hour PG. Individuals with IGT on the other hand, have a decreased early-phase insulin response to an oral glucose load and severe reduction in late-phase insulin secretion. For individuals with IFG, this combination results in excessive fasting hepatic glucose production, accounting for fasting hyperglycemia, and an excessive early rise in plasma glucose during the first hour of the OGTT with a return to preload blood glucose values during the second hour due to preserved late insulin secretion and normal muscle insulin sensitivity, whereas isolated IGT results in prolonged hyperglycemia after the oral glucose load.101

Table 1.3. Categories for Increased Risk for Diabetes (Prediabetes)86 HbA1C 5.7–6.4% OR FPG 100 mg/dL (5.6 mmol/L) to 125 mg/dL (6.9 mmol/L) (IFG) OR 2-h PG in the 75-g OGTT 140 mg/dL (7.8 mmol/L) to 199 mg/dL (11.0 mmol/L) (IGT) For all three tests, risk is continuous, extending below the lower limit of the range and becoming disproportionately greater at higher ends of the range.

The prevalence of these two intermediate states of blood glucose dysregulation varies widely across the U.S. and individuals with prediabetes often have a combination of IGT/IFG and exhibit characteristics of both. Although it may take many years to progress from

17 prediabetes to diabetes, current estimates indicate that up to 70% of individuals with IGT and/or

IFG eventually develop diabetes.101 IGT and IFG form a continuum in which the magnitude of reduction in β cell function establishes the degree of increase in plasma glucose. Insulin resistance is already well established if IGT is present and further progressive deterioration of β cell function accounts for the evolving natural history of the disease, leading to further elevation in glycemia and eventually resulting in T2D.85

Treatment

Treatment of T2D typically involves several components. According to the American

Diabetes Association’s 2017 Standards of Medical Care in Diabetes, lifestyle management is a fundamental part of diabetes treatment and includes diabetes self-management education

(DSME), diabetes self-management support (DSMS), physical activity, nutrition therapy, smoking cessation counseling, and psychosocial care.102 DSME and DSMS programs are designed to facilitate the knowledge, skills and abilities necessary for individuals to optimize diabetes self-care, while taking into account the individual’s needs, goals and life experiences.

The goal is to support informed decision making, self-care behaviors, problem solving, and active collaboration with the medical team to improve clinical outcomes, health status, and quality of life for individuals with diabetes.103

Physical activity is another important component of glycemic control.104 Adults with T2D are recommended to engage in at least 150 minutes of moderate to vigorous physical activity per week, spread over at least 3 days per week, with no more than 2 consecutive days without activity. The current guidelines also recommend 2-3 sessions per week of resistance exercise on nonconsecutive days. Adults with T2D are further advised to decrease the daily amount of time spent in sedentary behavior by interrupting prolonged periods of sitting every 30 minutes.102

18 Individualized nutrition therapy is another integral part of T2D management. The American

Diabetes Association (ADA) recommends that all individuals with diabetes receive individualized medical nutrition therapy (MNT) for education on healthful eating patterns and meal planning.102 The current evidence-based nutrition therapy guidelines are summarized below in Table 1.4.

In some cases, individuals can manage diabetes with strict adherence to diet and physical activity guidelines, but in most cases pharmacological therapy is also required. There are several classes of glucose-lowering medications available for treatment that may be used alone, or in conjunction with additional hypoglycemic medications and/or insulin injections. Metformin, a biguanide, is the preferred initial medication used in diabetes treatment due to its low risk of hypoglycemia, effect on CVD risk reduction, high A1C efficacy and low cost.102 Other classes of diabetes medication that may be used in combination with metformin to achieve glycemic control include meglitinides, sulfonylureas, thiazolidinediones (TZDs), alpha-glucosidase inhibitors, dipeptidyl peptidase (DPP)-4 inhibitors, sodium-glucose co-transporter 2 (SGLT2) inhibitors, glucagon-like peptide (GLP)-1 receptor agonists, bile acid sequestrants, amylin mimetics, and dopamine-2 agonists (Table 1.5). Given the complexity of T2D, using a combination of medications that act on different target systems is often the optimal way to reach glycemic targets. Long term, many individuals with T2D require insulin therapy for blood glucose management due to the progressive nature of the disease.102 According to a recent CDC report, 50% of individuals with diabetes use oral medication alone, 17.8% use only insulin, and

13% use a combination of oral medications and insulin to achieve their glycemic targets.105

While medication regimens often change overtime, lifestyle modification and diabetes self- management education are critical components throughout course of the disease.102

19 Table 1.4. Evidence-based Type 2 Diabetes Medical Nutrition Therapy Recommendations102 Topic Recommendations Energy balance • Modest weight loss (5% initial body weight) achievable by the combination of reduction of calorie intake and lifestyle modification benefits overweight or obese adults with type 2 diabetes and those with prediabetes

Eating patterns & macronutrient distribution • There is no single ideal dietary distribution of calories among carbohydrates, fats, and proteins for people with diabetes; macronutrient distribution should be individualized while keeping total calorie and metabolic goals in mind. • A variety of eating patterns are acceptable for the management of type 2 diabetes and prediabetes including Mediterranean, DASH, and plant-based diets. • Carbohydrate intake from whole grains, vegetables, fruits, legumes, and dairy products, with an emphasis on foods higher in fiber and lower in glycemic load, should be advised over other sources, especially those containing sugars. • Individuals with diabetes and those at risk should avoid sugar-sweetened beverages to control weight and reduce their risk for CVD and fatty liver and should minimize the consumption of foods with added sugar that have the capacity to displace healthier, more nutrient-dense food choices

Protein • In individuals with type 2 diabetes, ingested protein appears to increase insulin response without increasing plasma glucose concentrations. Thus, carbohydrate sources high in protein should not be used to treat or prevent hypoglycemia

Dietary Fat • Data on the ideal total dietary fat content for people with diabetes are inconclusive, an eating plan emphasizing elements of a Mediterranean-style diet rich in monounsaturated fats may improve glucose metabolism and lower CVD risk and can be an effective alternative to a diet low in total fat but relatively high in carbohydrates • Eating foods rich in long-chain omega-3 fatty acids, such as fatty fish (EPA and DHA) and nuts and seeds (ALA) is recommended to prevent or treat CVD; however, evidence does not support a beneficial role for omega-3 dietary supplements

Alcohol • Adults with diabetes who drink alcohol should

20 do so in moderation (no more than 1 drink/day for adult women and no more than 2 drinks/day for adult men) • Alcohol consumption may place people with diabetes at increased risk for hypoglycemia, especially if taking insulin or insulin secretagogues

Micronutrients and supplements • There is no clear evidence that dietary supplementation with vitamins, minerals, herbs, or spices can improve outcomes in people with diabetes who do not have underlying deficiencies, and there may be safety concerns regarding the long-term use of antioxidant supplements such as vitamins E and C and carotene

Nonnutritive sweeteners • The use of nonnutritive sweeteners has the potential to reduce overall calorie and carbohydrate intake if substituted for caloric sweeteners and without compensation by intake of additional calories from other food sources • Nonnutritive sweeteners are generally safe to use within the defined acceptable daily intake levels.

Table 1.5. Properties of Glucose-lowering Medications Available in the U.S. for Treatment of Type 2 Diabetes102 Class Compound(s) Primary Physiological Action(s) Biguanides Metformin ↓ Hepatic glucose production

Sulfonylureas Glyburide ↑ Insulin secretion Glipizide Glimepiride Meglitinides Repaglinide ↑ Insulin secretion Nateglinide TZDs Pioglitazone ↑ Insulin sensitivity Rosiglitazone α-Glucosidase inhibitors Acarbose Slows intestinal carbohydrate Miglitol digestion/absorption

DPP-4 Inhibitors Sitagliptin ↑ Insulin secretion (glucose Saxagliptin dependent) Linagliptin Alogliptin ↓ Glucagon secretion (glucose dependent)

Bile acid sequestrants Colesevelam ? ↓ Hepatic glucose production

? ↑ Incretin levels

21 Dopamine-2 agonists Bromocriptine Modulates hypothalamic regulation of metabolism

↑ Insulin sensitivity

SGLT2 inhibitors Canagliflozin Blocks glucose reabsorption by the kidneys, increasing glycosuria Dapagliflozin

Empagliflozin

GLP-1 receptor agonists Exenatide Insulin secretion (glucose Liraglutide dependent)

Albiglutide ↓ Glucagon secretion (glucose dependent) Lixisenatide Slows gastric emptying Dulaglutide ↑ Satiety

Amylin mimetics Pramlintide ↓ Glucagon secretion

Slows gastric emptying

↑ Satiety

Prevention A substantial body of evidence exists to support the role of lifestyle modification in the prevention of T2D in at risk populations. Intervention trials have clearly demonstrated that lifestyle changes to achieve a healthy body weight and support moderate physical activity are effective to prevent or delay the development of T2D. The early T2D prevention trials were relatively small studies, whose interventions differed in structure, duration and intensity.

Although the findings must be interpreted with caution, results suggested that a lifestyle intervention consisting of diet and exercise treatment could reduce the risk of progression to T2D in individuals with impaired glucose tolerance.106,107

Since then, larger randomized controlled trials (RCT) have provided strong support of the effectiveness of lifestyle interventions for T2D risk reduction. In 1986, 110,660 men and women

22 from 33 health clinics in Da Qing, China were screened, with 577 classified as having IGT and enrolled into a RCT. The trial consisted of three intervention groups plus a control group.

Participants were randomized to either receive individual and group counseling on a healthy diet for weight maintenance or loss (if BMI ≥ 25), how to increase leisure time physical activity, or a combination of both. Those randomized to the control group were provided with written material containing general information about diabetes and IGT. After 6 years, participants in all three intervention arms had a reduced risk of developing T2D compared to controls; specifically, a relative risk reduction of 31% in the diet alone group, 46% in the exercise alone group, and 42% in the diet/exercise group.108

An even stronger relationship between lifestyle change and T2D risk was found in the

Finnish Diabetes Prevention Study (DPS), where 522 overweight men and women with IGT were randomized to an intensive lifestyle intervention or control group at one of five participating centers. The intervention consisted of individualized dietary counseling with a nutritionist, physical training sessions, and advice to increase overall physical activity. The explicit intervention goals were weight loss, reduced total fat and saturated fat intake, and increased physical activity and dietary fiber intake. After approximately 4 years of active intervention, the trial was ended early due to the significantly lower rate of diabetes in the intervention group. A 58% risk reduction was observed for the intervention group, and the authors found that the magnitude of risk reduction was related to the number of study goals achieved.62 Extended follow up publications by the investigators have shown a sustained risk reduction in intervention subjects beyond the active intervention period.109,110

The largest of these intervention trials, the Diabetes Prevention Program (DPP), began in

1996 and was carried out at 27 centers across the U.S. This multicenter RCT enrolled 3,234

23 subjects with IGT and elevated fasting glucose to compare the efficacy and safety of three interventions for T2D prevention: a lifestyle modification program, metformin (850 mg twice daily) and placebo. The goal of the DPP intensive lifestyle modification program was to achieve a minimum 7% weight loss through a healthy low calorie, low fat diet and moderate physical activity. The intervention was delivered on a one-to-one basis through a 16-lesson curriculum during the first 6 months, followed by ongoing individual and group sessions for the remainder of the study. After the 2.8-year intervention, both metformin and lifestyle modification were effective for T2D prevention and improved glucose tolerance compared with placebo.

Interestingly, the risk reduction with lifestyle modification was even higher than that of metformin, 58% compared to 31%.63 Clearly, these robust findings support the role of lifestyle modification in the prevention of T2D in high risk individuals and as a target for population health.

Complications

Chronic hyperglycemia and other metabolic abnormalities present in T2D can lead to serious long term complications that affect many different tissues and organ systems. Individuals with diabetes have an increased risk for developing both microvascular and macrovascular comorbidities.111 The most common microvascular complications are diabetic retinopathy, potentially leading to blindness; nephropathy, which may result in end-stage renal disease; and neuropathy, and the risk appears to be further elevated in individuals with poorly controlled blood glucose levels.112 Compared to individuals without T2D, those with T2D also have a higher risk for experiencing macrovascular events, including stroke, coronary heart disease

(CHD), peripheral artery disease (PAD), cardiomyopathy, and congestive heart failure (CHF) than individuals without diabetes, and the pathologies are often more severe.113 Although the

24 underlying mechanism is not entirely clear, diabetic dyslipidemia (characterized by elevated fasting and postprandial triglycerides, low high-density lipoprotein (HDL) cholesterol, elevated low-density lipoprotein (LDL) cholesterol, and the predominance of small dense LDL particles) and diabetic arteriopathy (endothelial dysfunction, hypercoagulability, changes in blood flow and platelet abnormalities) are likely the main contributors.113,114 The relationship between T2D and cardiovascular disease will be discussed in further detail in the next chapter.

Diabetes is a complex, heterogeneous disease placing tremendous burden on healthcare systems worldwide. Efforts to further elucidate the complex mechanisms underlying this disease are needed to for effective prevention and treatment interventions. In the meantime, targeting at- risk individuals with validated measures for risk reduction and disease management are of critical importance.

25 CHAPTER 2

Dietary Patterns and Type 2 Diabetes Risk

As discussed in the previous chapter, intensive lifestyle interventions can effectively prevent or delay the onset of T2D in high risk individuals. However, such interventions are often not feasibly translated to the population at large. Diet was recently identified as the number one risk factor associated with the top 10 causes of death in the United States1 and is an exposure that all individuals encounter throughout the lifespan. Understanding the role of individual nutrients, foods, and dietary patterns in the prevention of T2D has therefore been the focus of a large body of research. A review of the literature related to specific nutrients and T2D risk is beyond the scope of this dissertation. Many prospective studies have provided evidence that the intake of some individual foods or food groups are associated with T2D risk. For instance, whole grain intake has been consistently shown to have an inverse relationship with T2D risk,115 while high intake of processed grains, such as white rice, is associated with increased T2D risk.116 Two meta-analyses found that total fruit and vegetable intake was not associated with T2D risk, but greater intake of green leafy vegetables was associated with lower risk.117,118 In addition, three large prospective cohort studies provided evidence that consumption of specific whole fruits such as blueberries, grapes, and apples was significantly associated with reduced T2D risk.119

Observational evidence suggests that the quality of dietary carbohydrates may be a more important determinant of T2D risk than carbohydrate quantity, as multiple studies have shown that a diet rich in fiber, especially cereal fiber, is inversely associated with T2D risk.120–122

Greater consumption of dairy products, especially yogurt, has been associated with a moderate reduction in T2D risk.123 Regular consumption of red and processed meats, such as bacon, sausage, and hot dogs is strongly associated with higher T2D risk.124 Fish and seafood

26 consumption was not significantly associated with diabetes risk in a recent meta-analysis of prospective cohort studies, but when the authors stratified the analysis by geographic region, higher fish/seafood consumption was associated with a higher risk of T2D in North America and

Europe, but a lower risk in Asia.125 Although the exact reason for this discrepancy is unknown, the authors postulated that differences in the type of fish consumed, cooking methods, or levels of exposure to pollutants may explain the differences. Greater intake of nuts, despite their high fat content, are associated with a reduced T2D risk.126,127 While it was believed that higher total fat intake contributed to an increased risk for diabetes by promoting insulin resistance and weight gain, results from human metabolic studies do not support this hypothesis128 and observational studies have found no association between total fat intake and T2D risk.120 Furthermore, findings from the Women’s Health Initiative showed no increased in T2D incidence among women who consumed a low-fat diet compared to the control group.129 Like carbohydrates, these findings have supported the idea that the quality of dietary fat is likely more important than total fat intake in relation to cardiometabolic disease risk. For example, greater intake of omega-6 polyunsaturated fatty acids (PUFA), especially when replacing dietary saturated fat, is associated with lower risk of developing diabetes.120,130 However, the relationship between omega-3 PUFA and T2D risk remains inconsistent.125

While this substantial body of evidence highlighting the potential role of specific dietary factors in T2D prevention has provided useful insight to guide nutrition recommendations, the reductionist approach of studying specific nutrients or foods in isolation is inadequate to capture the more complex relationship between diet and disease risk. Free living individuals eat meals with complex combinations of nutrients that are likely to have interactive or synergistic relationships. 131 Several statistical challenges also arise when studying nutrients in isolation, the

27 high level of inter-correlation among some nutrients makes it difficult to examine their separate effects and the effect of a single nutrient may be too small to detect, while the cumulative effect of multiple nutrients included in a dietary pattern may be sufficiently large for detection.131

Further, much of the existing research on the diet-diabetes relationship is based on observational data with measures of diet at only one point in time, which does not account for how changes in diet quality or the length of exposure to a certain dietary pattern may influence disease risk.

Another important limitation of this approach is the lack of precision in the calculation of specific nutrients from the dietary data available in observational studies.132 Dietary patterns have been increasingly recognized as the optimal approach to study diet-disease relationships because they are better suited to account for the complex and synergistic effects of total diet on health, since individuals are not exposed to single nutrients or foods.131,2,133 Furthermore, dietary patterns, more so than specific nutrients or even individual foods, can be effectively translated into nutrition guidelines that may be more easily understood and implemented by the general population and across a range of cultural dietary practices.3

Four methods are commonly used for dietary patterns analysis: factor analysis, cluster analysis, reduced rank regression and dietary indices. Factor analysis is a multivariate statistical technique that uses data collected from FFQs or dietary records to identify common underlying factors or patterns of food consumption. Specific food items in the dataset are then grouped based on the degree to which they are correlated with one another. A summary score is derived for each pattern and used either in correlation or regression analysis to examine the relationship between eating patterns and specific outcomes of interest.131 Cluster analysis is another multivariate method that characterizes dietary patterns by aggregating individuals into relatively homogenous subgroups (clusters) with similar diets. Individuals may be classified into clusters

28 based on the frequency of food consumed, the percentage of energy contributed by each food or food group, the average grams of food intakes, standardized nutrient intakes, or a combination of dietary and biochemical measures. Further analysis, such as comparing dietary profiles across clusters, is then done to interpret the patterns identified.131 Unlike the a posteroriori methods that derive eating patterns through statistical modeling of empirical data, dietary indices are constructed a priori on the basis of prevailing dietary recommendations. Dietary indices provide a summary score of the degree to which an individual’s diet reflects specific guidelines or recommendations, with higher scores indicating higher adherence to the pattern of interest.131

Reduced rank regression includes aspects of both types of analyses by using a linear combination of dietary data as predictor variables to explain as much variation in a set of disease-specific response variables as possible.134

Dietary Indices

One of the first standardized dietary indices was the HEI, developed in 1995 to measure diet quality based on adherence to the U.S. Dietary Guidelines for Americans (DGA). This 100- point score awards points for dietary diversity; higher intakes of grains, vegetables, fruit and milk; and lower intakes of meat, total fat, saturated fat, cholesterol, and sodium. The HEI was later updated to the HEI-2005 to reflect the 2005 DGA135 and then HEI-2010 following the publication of the 2010 DGA.136 The HEI-2010 includes 12 components: 9 adequacy components (dietary components to increase) and 3 moderation components (dietary components whose consumption should be decreased). The 9 adequate components include total fruit, whole fruit, total vegetables, greens and beans, whole grains, dairy, total protein foods, seafood and plant proteins and fatty acids. The remaining 3 components are refined grains, sodium and empty calories (energy from solid fats, alcohol and added sugars). Components in the HEI-2010 are

29 energy-adjusted on a density bases (per 1000 Calories/day). For adequate components, a maximum score (5 or 10 depending on the component) is given for reported intake greater than or equal to the amount recommended by the 2010 DGA, where a higher score reflects higher consumption. For moderation components, a maximum score (10 or 20) is given for reported intake less than or equal to the recommended amount, such that a higher score indicates lower consumption. For all components, higher scores reflect better diet quality. The scores of the 12 components are then summed to yield a total score, with a maximum value of 100.

In an attempt to improve the original HEI, McCullough, et al.137 created the 9-component

Alternate Healthy Eating Index (AHEI), designed to target food choices and macronutrient sources consistently associated with reduced chronic disease risk in clinical and epidemiological studies. Scoring for the AHEI was based on intake levels of fruits, vegetables, the ratio of white

(seafood and poultry) to red meat, trans-fat, the ratio of polyunsaturated to saturated fat, cereal fiber, nuts and soy, moderate alcohol consumption (0.5-1.5 servings/day), and long-term multivitamin use (<5 or 5+ years).137 The AHEI was also updated in 2005 and again in 2010 to reflect the most current scientific evidence related to the diet-chronic disease relationship. The

AHEI-2010 score is based on criteria for intake of fruits, vegetables, whole grains, sugar sweetened beverages and fruit juice, nuts and legumes, red and processed meats, alcohol, sodium, trans-fat, long chain (n-3) fatty acids and percent of energy from polyunsaturated fatty acids. Components are scored from 0 (worst adherence to recommended intake levels) to 10

(best adherence) and then summed, with a total possible AHEI-2010 score ranging from 0-

110.138

Early ecologic evidence suggesting potential beneficial health effects of a Mediterranean- type dietary pattern139 eventually led to the creation of a Mediterranean dietary index.

30 The Mediterranean diet was first studied using a dietary pattern score in a prospective epidemiologic study by Trichopoulou, et al.140 This score was based on a traditional

Mediterranean diet and consisted of 8 components: high monounsatured:saturated fat ratio; moderate alcohol consumption; high consumption of legumes, cereals (including bread and potatoes), fruits, and vegetables; and low consumption of meat and meat products, milk and dairy products. Intakes above the study population median received 1 point and all other intakes received 0 points. For meat and dairy consumption, intake less than the median received 1 point.

Several versions of the original Mediterranean diet score have been created for use in non-

Mediterranean populations and as updated evidence has become available.141–143

In addition to the Mediterranean dietary index, the Dietary Approaches to Stop

Hypertension (DASH) diet score is commonly used to assess diet quality in epidemiologic studies. The DASH score is based on the DASH diet recommendations for intake of 8 food components: high intake of fruits, vegetables, nuts and legumes, low-fat dairy products, and whole grains and low intake of sodium, sweetened beverages, and red and processed meats.144

To calculate a DASH diet score, study participants are first divided into quintiles for each component based on their intake ranking from FFQ data. Components whose intake is emphasized in the DASH diet (fruits, vegetables, nuts and legumes, low-fat dairy products and whole grains) are scored directly from the quintile ranking, such that quintile 1 is assigned 1 point and quintile 5 is assigned 5 points. Components with low recommended intake (sodium, sweetened beverages, red and processed meats) are scored in reverse; quintile 1 is assigned 5 points and quintile 5 is assigned 1 point.145 Each component score is then summed to obtain an overall DASH score for participants, with a range of 8-40.

Literature Review: Dietary Patterns and T2D Prospective Cohort Studies

31 Numerous prospective cohort studies have investigated the relationship between dietary patterns and T2D risk. Recently, the use of a priori diet pattern scores as a measure of diet quality has begun to replace the other data-driven methods described above. Reasons for this may be that the results of data-driven approaches do not always correspond with established recommendations and identified patterns are difficult to compare across studies because the outcomes are dependent on the specific population under study.146 Several meta-analyses have found that despite dietary pattern variation between populations, common characteristics between dietary patterns obtained by data-driven methods are associated with T2D risk. Dietary patterns consisting of healthy foods with higher energy contributions from whole grains, fruits and vegetables are associated with a reduced risk of T2D, while dietary patterns characterized by unhealthy food choices and higher energy contributions from red or processed meats, high-fat dairy, refined grains, fried foods and sweets may increase the risk of developing T2D.147–149

The dietary indices most widely used to assess diet quality today are the Mediterranean,

DASH, HEI and AHEI diet scores. A recent meta-analysis showed that individuals in the upper quantiles of these dietary patterns scores have 20-23% lower risk of developing T2D compared to those in the lowest quantiles.149 A summary of the large, well conducted prospective cohort studies examining the relationship between diet quality, as measured by an a priori diet pattern score, and T2D risk with sample populations of at least 10,000 participants, is shown in Table

2.1.

Evidently, multiple dietary patterns convey protection against the development of T2D.

However, the findings are limited by the use of older populations, single dietary assessment, short follow up periods, and in several cases, self-reported T2D and study populations of health professionals and university graduates, who may be influenced by the healthy-user bias and are

32 not likely to represent the general population. Further research that includes contemporary and understudied dietary patterns and uniformly defined T2D in young, healthy population-based cohorts is needed assess whether a critical window of exposure may exist and further postulate how different dietary approaches may reduce T2D risk.

Table 2.1. Summary of selected prospective cohort studies using a priori diet indices to examine the relationship between dietary patterns and incident type 2 diabetes mellitus. Author, Year Study Population Dietary Assessment Outcome Methods Results Martinez-Gonzales, SUN project -Validated semi- -Baseline T2D -Incidence rate ratios -After median follow et al., 200869 (Seguimiento quantitative FFQ defined as a self- estimated using up of 4.4 years, 33 Universidad de with 136 items to reported medical Poisson regression incident cases of Navarra) cohort, assess usual intake diagnosis or use of models with robust T2D were confirmed started in 1999 over the past 12 insulin/oral standard errors -High adherence to -13,380 former months antidiabetic -Multivariate models the diet was students of the -Sum of the medication adjusted for known associated with an University of consumption of -Incident cases self- T2D risk factors: 83% relative Navarra, registered similar individual reported on follow sex, age, education, reduction in the risk nurses from Spanish food items used to up questionnaires, total energy intake, of developing provinces, and estimate overall adjudicated using BMI, physical diabetes (incidence university graduates consumption of food additional activity, sedentary rate ratio 0.17, 95% from other groups questionnaires to time, family history CI 0.04-0.75). associations -Mediterranean diet obtain details of the of diabetes, HTN, - As a continuous -Free of T2D and adherence assessed diagnosis, then and smoking variable, an increase CVD at baseline using the score reviewed by a panel -Participants were in score of 2 points -Average age 37.8 created by of physicians to grouped into 3 was associated with years Trichopoulou, et al. classify the diagnosis categories of an incidence rate as incident T2D or Mediterranean diet ratio of 0.65 (95% CI not according to adherence: lowest 0.44-0.95)

ADA guidelines. (score < 3), moderate

(3-6), and high (7-9)

InterAct Consortium, Subcohort of the -Country-specific Incident cases of -Cox proportional -Higher rMED score 2011150 European validated dietary T2D ascertained hazards regression, was associated with Prospective questionnaires to using at least two modified for the a lower risk of Investigation into assess usual food multiple sources of case-cohort design developing T2D. Cancer and Nutrition intake evidence including according to the -Individuals in the (EPIC) cohort -Estimated self-report, linkage Prentice method higher rMED score recruited between individual nutrient to primary or -Multivariate model range were 12% less 1992-2000, aged 25- intakes derived from secondary care adjusted for sex, likely to develop 70 (average 53 at standardized EPIC registers, drug BMI, educational T2D than those in enrollment) nutrient database registers, hospital level, smoking with low rMED - 11,994 individuals -Relative admissions, and status, physical scores (adjusted HR with incident T2D Mediterranean diet mortality data activity, and total 0.88, 95% CI 0.79– -15,798 randomly score (rMED) energy intake. 0.97) selected individuals -rMED scores - Alcohol, meat, and from the EPIC classified into olive oil components cohort categories of accounted for most adherence: low (0-6), of the observed medium (7-10), and association high (11-18)

33 Rossi, et al., 2013151 -Greek cohort of the -Baseline validated -Incident T2D -Cox proportional -Significant inverse EPIC study semi-quantitative identified through hazards regression association of -Participants with FFQ to assess usual medical records, models used to increased adherence preexisting diabetes, intake of 150 discharge diagnosis assess the to MDS with T2D, CVD, or cancer at foods/beverages over or death certificates, relationship of T2D HR 0.88 (95% CI baseline were the past 1 year first-reported T2D- with MDS 0.78-0.99) for those excluded -Nutrient, ethanol specific medication -Multivariate models with a MDS ≥ 6 -Final sample 22,295 and energy intakes use or self-reported adjusted for the compared with ≤ 3 participants calculated using a medical diagnosis of potential -HRs were 0.94 -Average age 50 food composition T2D during follow- confounders: age, (95% CI 0.63, 1.40) years database modified to up sex, education, BMI, for BMI accommodate the -About 60% of all physical activity, <25 kg/m2 and 0.87 Greek diet diabetic cases were WHR, total energy (95% CI 0.77, 0.98) -Mediterranean diet verified by medical intake for BMI score (MDS) records ≥25 kg/m2 (p for -Range 0-9 heterogeneity 0.868) Fung, et al. 2007152 -Subcohort of -Validated 116-item -Incident T2D -Relative risk for -Inverse association women aged 38-63 FFQ administered in obtained by self- T2D calculated using between AHEI score from the Nurses’ 1984 (considered report in biennial Cox proportional and T2D Health Study (NHS), baseline for this questionnaires hazards models -Relative risk (RR) followed from 1984- study), and in 1986, -Diagnoses -Cumulative average comparing top to 2002 1990, 1994, and confirmed with of AHEI scores from bottom quintiles was -Women with a 1998 supplementary repeated FFQs used 0.64 ([95% CI 0.58– history of cancer, -Designed to questionnaire mailed in the models 0.71] Ptrend < 0.0001) cardiovascular measure average to assess symptoms, -Examined change in -In stratified disease, and diabetes food intake over the diagnostic tests, and score in 4, 6-8, and analyses, AHEI at baseline excluded past 1 year treatment 10-12 years score was associated -Final study -Total energy and -Criteria for associated with T2D with reduced T2D population 80,029 nutrient intake was classification risk risk only in non- women calculated by consistent with -Models adjusted for smokers, women summing up energy National Diabetes age, diabetes family without hypertension or nutrients from all Data Group and history, smoking, and those with foods 1997 ADA criteria postmenopausal normal blood -9 component AHEI -Medical records hormone use, energy cholesterol levels score reviewed by a intake, leisure time -Women whose -Range 0-87.5 blinded physical activity and score improved over endocrinologist BMI any time period had confirmed 61 of 62 -Additional adjusted lower risk than reports (98%) for waist-to-hip ratio women whose low -Deaths reported by when available to score did not change family members, minimize postal service, or confounding by National Death Index adiposity

Chiuve, et al., -71,495 women aged -Validated FFQs -Incident T2D was -Cumulative mean of -The HEI 2005 and 2012138 30-55 years at completed in the self-reported on diet scores used to AHEI 2010 were baseline from the NHS in 1984, 1986, follow-up incorporate repeated inversely associated NHS who completed and then every 4 questionnaires, measures with risk of major an extensive FFQ in years through 2006 confirmed by a -Cox proportional chronic disease 1984 -Validated FFQ validated hazards regression among men and -41,029 men enrolled completed in the supplemental used to estimate RR women, and in a in the Health HPFS every 4 years questionnaire and for disease by pooled analysis of Professionals from 1986-2006 review of medical quintile of diet score, both cohorts Follow-up Study -Nutrient intake records by study lowest quintile was -High adherence to (HPFS), age 40-75 at calculated by investigators blinded the reference group the 2005 DGA was baseline who multiplying the to participants’ risk -Multivariate models associated with an completed a frequency of intake status adjusted for potential 18% reduced risk of diet/medical history for each food by its confounders: age, T2D in the pooled questionnaire in nutrient content and energy intake, analysis (HR 0.82, 1986 summing nutrient physical activity, 95% CI 0.76-0.89) -Exclusion criteria contributions across BMI, family medical -Participants in the included a history of all food items history, vitamin E highest quintile of CVD, diabetes, or -HEI 2005 supplementation, AHEI 2010 score cancer at baseline -AHEI 2010 history of HTN and had a 33% reduced elevated cholesterol risk of developing T2D (HR 0.67, 95% CI 0.61-0.74)

Qiao, et al. 2013153 -WHI cohort of -Diet history self- -Prevalent diabetes -Cox proportional -Most women did

34 postmenopausal reported at baseline obtained by self- hazards models were not meet the AHEI women; began in using a validated report at baseline used to estimate the dietary goals 1992 with an average FFQ with 122 -Incident treated hazard ratios (HR) -Women in the of 7.2 years of composite and single diabetes ascertained incident diabetes highest quintile of follow up food items to assess at each annual or -Lowest AHEI score AHEI were 24% less -After exclusion for average daily semi-annual contact, quintile was the likely to develop missing data and nutrient intake over defined as a self- reference group T2D than those in prevalent T2D cases, the past 3 months report of physician- -Fully adjusted the lowest (HR 0.74, 154,493 women -University of diagnosed diabetes model included age, 95%CI 0.70-0.82) were included in the Minnesota Nutrition treated with oral race/ethnicity, BMI, -This association was final analyses Data System for medication or education, cigarette observed in Whites Research, 2005 used insulin, from self- smoking, waist/hip (HR 0.72, 95%CI to estimate nutrient report on follow-up ratio, physical 0.68-0.82) and intake questionnaires activity, daily energy Hispanics (HR 0.68, -Diet quality -Accuracy of self- intake, family history 95%CI 0.46-0.99), measured by AHEI reported diabetes in of diabetes, study but not Blacks (HR score WHI was validated arm, and hormone 0.85, 95%CI 0.69- using medication and therapy use 1.05) and Asians laboratory data -Sensitivity analyses (HR 0.88, 95%CI by race/ethnicity 0.57-1.38) de Koning, et al., Subcohort of the -Validated HPFS -Self-reported cases -The associated -Participants in the 2011154 prospective HPFS, FFQs were of T2D were between quintile of highest quintiles of enrollment began in administered every 4 confirmed when at diet score and T2D DASH, aMed, and 1986 with years to assess usual least one of the risk was tested using aHEI had a 25% (HR completion of dietary intake symptoms, positive Cox proportional 0.75, CI 0.65-0.85), questionnaires to -HEI-2005 diagnostic glucose hazards models with 25% (HR 0.75, CI assess health and -AHEI tests, and medication time-varying 0.66-0.86), and 23% lifestyle every 2-4 -51-point use were reported on covariates (HR 0.77, CI 0.67- years through 2006 Recommended Food a supplementary -Diet quality was 0.88) lower risk of -41,615 men free of Score (RFS) questionnaire calculated as the developing T2D T2D, CVD, cancer -aMed -Glucose criteria cumulative average during the follow-up or with implausible -DASH were from the at each time point period, respectively energy intake at National Diabetes -Multivariate models -Only the DASH baseline Data Group (cases were adjusted for score remained prior to 1998) and age, smoking, inversely associated the American physical activity, with T2D risk after Diabetes Association coffee intake, family mutual adjustment (cases after 1998) history of T2D, and for other significant -In a validation total energy intake scores study, 97% of cases -The lowest quintile - Overweight or were confirmed by of each diet score obese men had medical record served as the greater reductions in review reference group absolute risk of T2D compared with normal weight

Jacobs, et al., 2015155 Hawaii component -Self-administered -Incident T2D -Association -Significant inverse of the Multiethnic quantitative FFQ on identified through a between the 4 diet associations between Cohort (MEC) usual food intake follow-up indices and risk of higher DASH index -Participants of over the past 12 questionnaire (1997- T2D was assessed scores and risk of primarily white, months collected at 2003), medication with Cox T2D in white men Japanese-American baseline inventory (2003- proportional hazards and women, as well and Native Hawaiian -Validated and 2006) or linkage models, follow-up as in Japanese- ancestry calibrated in each with health insurance time used as the time American women -Aged 45-75 years at ethnic-sex group plans (2007) metric and Native Hawaiian baseline, 1993-1996 -32 food groups -Models stratified by men -Exclusions: created from age and adjusted for -Higher adherence to prevalent diabetes, MyPyramid ethnicity (white = the AHEI-2010 and questionable diabetes Equivalent Database ref), physical aMED diet was status or (MPED) activity, smoking related to a 13–28% invalid/missing -HEI-2010 status, education, lower risk of type 2 covariate data -AHEI-2010 total energy intake, diabetes in white -89,185 participants -aMed and BMI participants but not -DASH -Diet scores group in other ethnic into quintiles, lowest groups was the reference Cespedes, et al. -Women’s Health -Diet history self- -Prevalent diabetes -Multivariable- -Better quality diets 2016156 Initiative (WHI) reported at baseline obtained by self- adjusted Cox were associated with cohort, consisting of using a validated report at baseline proportional hazards lower risk of T2D

35 postmenopausal FFQ with 122 -Only incident models used to regardless of the women, aged 50-79 composite and single treated diabetes was calculate hazards dietary index recruited into one or food items ascertained, defined ratios and 95% -A 1-SD increase in more clinical trial or -University of as a self-report of confidence intervals score on any index an observational Minnesota Nutrition physician-diagnosed -Diet indices was associated with study between 1993- Data System for diabetes treated with modeled a 10%–14% lower 1998 Research, 2005 used oral medication or categorically as T2D risk (p< 0.001) -Women who to estimate nutrient insulin, from self- quintiles and -Adjustment for BMI participated in the intake report on follow-up continuously, per category attenuated WHI dietary -Food groups created questionnaires 10% increment in the associations modification trial from MPED -Accuracy of self- theoretical score and (5%–10% lower were excluded -Diet quality reported diabetes in per 1 standard risk; P < 0.001) -Study population measured by 4 WHI was validated deviation (SD) -Comparing the top included 101,504 indices: aMED, using medication and -Models adjusted for and bottom quintiles, women without T2D, DASH, HEI-2010, laboratory data age, educational higher indices were a baseline FFQ and AHEI-2010) attainment, associated with a complete pertinent race/ethnicity, lower relative risk of data smoking status, T2D between 26- family history of 36% diabetes, hormone therapy, total daily energy intake, physical activity quintile, and study arm Tonstad et al., Subcohort of the Vegetarian diets: -T2D was self- -Chi-square to -Vegan (OR 0.38, 2013157 Adventist Health -Vegan consumed no reported on bi-annual compare the number 95% CI 0.24–0.62), Study-2, a animal products hospitalization of newT2D cases by lacto-ovo (OR 0.62, prospective cohort -Lacto-ovo history follow-up category of 95% CI 0.50–0.76), study conducted vegetarian ate dairy questionnaires, vegetarian diet and semi-vegetarian among Adventist products and/or eggs administered every 2 (vegan, lacto ovo, (OR 0.49, 95% CI church members in ≥1 time/month but years after baseline pesco or semi- 0.31–0.76) diets the US and Canada, no fish or meat -Validated diagnoses vegetarian) or non- were associated with started in 2002 - Pesco-vegetarian in a subgroup of the vegetarian diet a lower incidence of -Participants at least ate fish ≥1 cohort via phone -Multiple logistic T2D 30 years of age were time/month and dairy interview regression to -Pesco-vegetarian recruited through products and/or eggs estimate odds ratios diets were not churches but no red meat or for incident T2D, associated with -15,200 men and poultry -Semi- using the non- lower incident 26,187 women were vegetarians vegetarian diet as the diabetes compared to free of diabetes at consumed dairy reference group non-vegetarian baseline and had products and/or eggs -Confounders complete pertinent and (red meat and included age, gender, data poultry ≥1 ethnicity (Black -41,387 included in time/month and <1 versus non-Black), final analyses, 17.3% time/week) education, income, black -Non-vegetarians television watching, consumed animal hours of sleep, products (red meat, alcohol consumption, poultry, fish, eggs, smoking and milk and dairy physical activity products >1 time/week)

Beverage Consumption and Type 2 Diabetes Risk

In addition to dietary patterns, beverage consumption habits are also known to influence

T2D risk. The rise in obesity and type 2 diabetes (T2D) prevalence in the United States over the last several decades has closely paralleled the rise in both artificially-sweetened (ASB) and sugar-sweetened beverage (SSB) consumption.158,159 A robust body of research has

36 demonstrated an association between SSB consumption and increased cardiometabolic risk, including T2D, primarily through weight gain and metabolic effects.160–167 In a recent meta- analysis by Imamura, et al.163 higher consumption of SSBs was associated with a greater incidence of T2D by 18% (95% CI 9-28%) and 13% per serving/day (95% CI 6-21%), before and after adjustment for adiposity. Additionally, the authors estimated that of 20.9 million incident cases of T2D predicted to occur over 10 years in the US, 1.8 million would be attributable to the consumption of SSBs, a population attributable fraction of 8.7% (95% CI 3.9-

12.9%).

ASBs on the other hand, are often promoted as a healthy alternative to sugar sweetened beverages, but the relationship between ASB consumption T2D risk is less clear. A small body of evidence has begun to emerge that suggests a positive association between ASB consumption and cardiometabolic risk168–171 and even T2D,172 however the results have thus far been inconsistent and no clear biological mechanism has been agreed upon by the scientific community. Initially, it was thought that the observed association between ASB intake and cardiometabolic risk was primarily due to residual confounding by other dietary behaviors, lifestyle factors, or demographic characteristics.173,174 It has been suggested that artificial sweeteners in foods and beverages may increase the desire for and consumption of sugar- sweetened, energy-dense foods and beverages,175 or may disrupt the ability to accurately estimate energy intake and remaining energy needs.176 In either case, ASB consumption could result in excessive calorie intake, increased body weight, and consequentially metabolic dysfunction. In animal models, it has been shown that the dissociation between sweet taste cues and calories associated with non-nutritive sweetener intake can lead to a decreased ability of sweet tastes to initiate physiological responses that aid in energy balance regulation.177 The 2015 review by

37 Imamura, et al. found that higher consumption of ASB was associated with an 8% (95% CI 2-

15%) increased risk of developing T2D after adjustment for adiposity. However, publication bias and residual confounding were indicated in the studies on ASB and T2D risk under review.163

Table 2.2 provides a comprehensive overview of the large prospective cohort studies that include an analysis of the association between ASB consumption and incident T2D.

Table 2.2. Summary of selected prospective cohort studies examining the relationship between artificially- sweetened beverage consumption and incident type 2 diabetes.

Author, Year Study Population Beverage Outcome Methods Results Consumption Palmer, et al., -Black Women's -Data on food and -Incident cases of -Age-stratified Cox -During the period 2008178 Health Study beverage intake T2D obtained from proportional hazard from 2001 through (BWHS), a obtained via a self- reported models used to 2005, the IRR for ≥1 nationwide ongoing modified, 68-item diabetes on any of calculate incidence diet soft drink per follow-up study of version of the short- the follow-up rate ratios (IRRs) day relative to less African American form Block–National questionnaires -Multivariable than 1 drink per women Cancer Institute FFQ -The accuracy of models included age, month was 1.06 -Women aged 21-69 -The 2001 FFQ self-reported questionnaire cycle, (95% CI, 0.83-1.36) years old were specifically included diabetes was family history of enrolled in the questions on diet assessed among a diabetes, cigarette BWHS starting 1995 soda consumption random sample of smoking, physical - Women with participants, with the activity, years of diabetes, GDM, diagnosis confirmed education, glycemic CVD, cancer, or for 94% of the index of the diet, and pregnant at baseline, women intake of coffee, red <30 years at the end meat, processed meat, of follow-up, or with and cereal fiber incomplete data were excluded -Final study sample of 43,960 women

38 Nettleton, et al., -Multi-Ethnic Study -Diet assessed at the -T2D defined as self- -Cox proportional -Daily consumers of 2009170 of Atherosclerosis baseline (2000-2002) reported T2D fasting hazards models used diet soda had a 67% (MESA) via FFQ glucose >126 mg/dl to calculate HR for elevated risk of T2D -Population-based -Diet soda intake at any examination, T2D compared with non- study of 6,814 quantified from an or use of -Multivariable model consumers with Caucasian, African item listing “diet soft hypoglycemic adjusted for baseline adjustment for American, Hispanic, drinks, unsweetened medication age, sex, demographics and and Chinese adults, mineral water” -Incident cases race/ethnicity, lifestyle factors (HR aged 45–84 years -Intake of diet soda included individuals examination site, 1.67, 95% CI 1.27- characterized as without T2D at energy intake, 2.20) rare/never, > baseline who met attained education, -Adjustment for rare/never but <1 any one of the three time spent in inactive other dietary factors serving/week, >1 criteria above at and active pursuits did not markedly serving/week but <1 follow-up during, smoking change risk serving/day, and ≥1 examinations status, and regular estimates serving/day dietary supplement -Adjustment for use baseline differences -Also adjusted for in waist various dietary circumference factors, baseline slightly attenuated waist circumference HRs (HR 1.40, 95% BMI, and change in CI 1.06-1.84) waist circumference -Results similar after or body weight adjustment for change in WC or body weight de Koning, et al., -40,389 healthy -Intake of ASB was -Incident T2D was -Cox proportional -Intake of ASBs was 2011179 participants from the assessed using a 131- self-reported on hazard models with significantly HPFS item validated semi- follow-up time-varying associated with T2D -Enrollment of men quantitative FFQ questionnaires covariates in the age-adjusted aged 40-75 began in sent to participants -Verified by a mailed -Covariates: smoking, analysis (HR 1.91; 1986 every 4 years supplementary physical activity, 95% CI: 1.72- -Questionnaires -ASBs were defined questionnaire, based alcohol intake, family 2.11; P for trend < mailed every other as caffeinated, on National Diabetes history of T2D, high 0.01) but not in the year to participants caffeine-free, and Data Group criteria baseline triglycerides multivariate- to assess health noncarbonated low- and ADA criteria for and blood pressure, adjusted analysis status and lifestyle calorie beverages cases reported after diuretic use, (HR1.09; 95% CI: factors through 2006 -Beverage intake was 1998 multivitamin intake 0.98, 1.21; P for -Exclusions included grouped into from 1988 onward, trend = 0.13) prevalent CVD, quartiles diet quality (aHEI diabetes, cancer or quintiles), recalled implausible energy weight change intakes between 1981 and 1986, adherence to a low-calorie diet in 1994, total energy intake, and BMI

39 Pan, et al., 2012180 -82,902 participants -Usual beverage -Incident diabetes -4 year lagged -Higher ASB from the Nurses’ intake over the diagnoses were self- analysis consumption was Health Study (NHS) previous year was reported on biennial -Time-dependent Cox associated with an II free of cancer, assessed by a 133- questionnaires proportional hazards increased risk of CVD, and diabetes item FFQ was -New cases were models to estimate developing T2D with a complete FFQ administered in confirmed through the RR for T2D when beverage and pertinent data at 1991, 1995, 1999, supplementary associated with ASB intake was used as a baseline (1991) and 2003 questionnaires that consumption continuous variable -ASB consumption asked about diabetes - Multivariable (p trend 0.04) categories: ≤ diagnosis and models adjusted for -Frequency of ASB 1cup/week, 2-4 treatment, based on age, race, family consumption did not cups/week, 5-7 National Diabetes history of diabetes, significantly affect cups/week, 2-3 Data Group and BMI, smoking status, T2D risk in the cups/day, ≥ 4 ADA criteria alcohol intake, multivariate cups/day; one cup physical activity, categorical analysis equivalent to 240 ml postmenopausal (~8.12 oz.) status, menopausal hormone use, oral contraceptive use, and overall dietary quality represented by the AHEI InterAct -Case–cohort study -Beverage -Ascertainment of -Cox proportional -High consumers of Consortium, 2013166 of 11,684 incident consumption over incident type 2 hazards regression, artificially T2D cases and a the previous year diabetes involved a modified for the sweetened soft stratified subcohort was estimated using review of the case−cohort design drinks had an of 15, 374 country-specific existing EPIC according to the increased risk of participants selected validated dietary datasets at each Prentice method developing T2D from 8 European questionnaires center using multiple -Models adjusted for compared to low cohorts participating administered once at sources of evidence: adjusted for sex, consumers (adjusted in the EPIC study baseline self-report, linkage smoking status, HR 1.93, 95% CI -Participants with -Included sugar- to primary-care alcohol consumption, 1.47, 2.54; p for diabetes or CVD at sweetened and registers, secondary- educational level and trend <0.0001) baseline were artificially- care registers, physical activity. -The association was excluded sweetened soft drink medication use (drug -Sugar-sweetened attenuated and -Different types of consumption registers), hospital and artificially became statistically soft drinks could be -Soft drinks assessed admissions and sweetened soft drinks not significant when distinguished in all as categorical mortality data were also mutually BMI was included EPIC centers except variables: <1 adjusted and also in the model (HR Italy, Spain and glass/month, 1–4 adjusted for juice and 1.13, 95% CI 0.85, Umeå (excluded glasses/month, >1–6 nectar consumption 1.52; p for trend from analysis) glasses/week, ≥1 -Sensitivity analyses 0.24) glass/day; one glass included adjustment equivalent to 250 g for total energy (~8.8 oz.) intake and BMI

40 Fagherazzi, et al., -The E3N study is a -Usual diet over the -One set of incident -Cox multivariate -ASB consumption 2013181 French prospective previous year was cases of diabetes regression models was positively cohort study of assessed using a came from self- with age as the associated with 98,995 female validated 208-item reported either timescale were used - greater risk of T2D - teachers initiated in diet-history diabetes, a diabetes Models adjusted for High consumption 1990 questionnaire in diet plan, the use of years of education, of ASBs (>603 -The E3N is the 1993, structured diabetic drugs, or a smoking status, mL/week) was French component of according to the hospitalization for physical activity, associated with the EPIC study French meal pattern diabetes in ≥1 of the hypertension, significant greater - 66,118 women -Women could 8 questionnaires sent hypercholesterolemia risk of diabetes (HR were included in the specify if their soda up until July 2005 , use of hormone 2.21, 95% CI 1.56- final analysis, after or fruit drinks were -Another set of cases replacement therapy, 3.14) compared with exclusions for sugar or artificially was identified family history of that for non- incomplete dietary sweetened exclusively from a diabetes, self- consumers data, prevalent - Beverage drug-reimbursement reported use of -Results were diabetes, or extreme consumption (in file, without any antidiabetic drugs, similar when the energy intakes mL/week) was previous report of alcohol intake, first 5 years of categorized into 5 diabetes omega-3 fatty acid follow-up were categories that -A total of 3496 intake, carbohydrate excluded corresponded to diabetes cases intake, total energy -The association quartiles, plus a non- diagnosed until June intake, coffee, fruit remained consumer category 2007 were validated and vegetables and significant, but was in the E3N cohort processed-meat attenuated with consumption and further adjustment dietary pattern for BMI (HR 1.68, (Western or 95% CI 1.19-2.39) Mediterranean)

Sakurai, et al., -Study participants -Diet soft drink -Fasting plasma -Cox proportional -Diet soda 2014182 were employees of a consumption over glucose and HbA1c hazards models consumption was factory that produces the past 1 month was were measured -Possible significantly zippers and assessed using a self- during the annual confounders: age, associated with the aluminum sashes in administered medical BMI, family history incident risk of the Toyama validated Japanese examinations of diabetes, smoking diabetes (P for trend Prefecture, Japan diet history -T2D diagnosis status, alcohol = 0.013) who underwent questionnaire (DHQ) based on ADA and consumption, -Compared to testing for diabetes -Diet soft drink the JDS criteria of at habitual exercise, rare/never during annual consumption was least one of the HTN, dyslipidemia, consumers, those medical converted from mL following: fasting dietary intervention who consumed >1 examinations into servings, one PG ≥ 126 mg/dl; for chronic disease, serving/week to <1 between 2003-2010 serving = 237 mL or HbA1c ≥ 6.5%; or total energy intake, serving/day had a -2,037 participants, 8 oz. treatment with dietary fiber intake, 70% increased risk average age 46.7 -Categorized as insulin or oral and adjustment for for incident T2D years, were included rare/never, more than hypoglycemic consumption of SSB, (HR 1.71, 95% CI in the final analysis rare/never but <1 medication fruit juice, vegetable 1.11–2.63) after exclusions serving/week, >1 juice, and coffee serving/week but <1 serving/day, and ≥1 serving/day

Although non-nutritive sweeteners are approved by the FDA, long term studies that adequately assess potential health risks associated with consumption over time are still needed.

Understanding the potential impact of ASB consumption on T2D risk is critical for accurate health messaging and nutrition education since ASBs are often marketed as healthy alternatives to caloric sugar-sweetened beverages. Current epidemiologic evidence is limited by dietary assessment at only one point in time as well as the use of older subjects. As a chronic disease,

41 diabetes takes many years to develop, thus studying the effects of dietary components on risk in older adults may miss a critical risk window. Furthermore, understanding whether the association of ASB with T2D risk is merely a byproduct of higher consumption by individuals at higher risk for T2D or if a plausible biological mechanism could explain this relationship would greatly contribute to this area of research.

42 CHAPTER 3

Type 2 Diabetes and Cardiovascular Disease

The Natural History of Cardiovascular Disease

Atherosclerotic cardiovascular disease (ASCVD) is the leading cause of morbidity and mortality for individuals with diabetes and the largest contributor to the direct and indirect costs of the disease.183,184 ASCVD is defined as acute coronary syndromes, a history of MI, stable or unstable angina, coronary or other arterial revascularization, stroke, transient ischemic attack, or peripheral arterial disease presumed to be of atherosclerotic origin.185 In the U.S., death rates from CVD are 1.7 times higher among individuals with diabetes than those without, resulting primarily from an increased risk for stroke and MI.41

Atherosclerosis is defined as a progressive disease of large and medium-sized arteries characterized by endothelial dysfunction, vascular inflammation and accumulation of lipids, cholesterol, calcium and cellular debris to form a plaque within the intima of the blood vessel wall.186 A depiction of this process is presented in Figure 3.1. In genetically susceptible individuals, the first stage of atherogenic plaque development is the formation of a fatty streak on the inner surface of an artery in response to endothelial injury. Injury to the arterial endothelium from exposure to physical forces (e.g., shear stress) or chemical irritants and toxins (e.g., increased reactive oxygen species from cigarette smoking, elevated circulating LDL levels and hyperglycemia) leads to inflammatory activation in vascular cells and endothelial dysfunction.187

In endothelial cells, inflammation results in disruption of the permeability barrier, increased production of inflammatory cytokines that further increase permeability, increased production of leukocytes adhesion molecules, decreased production of vasodilatory molecules and decreased production of antithrombic molecules. In smooth muscles cells, inflammatory activation causes

43 increased production of inflammatory cytokines, increased extracellular matrix synthesis, and increased migration and proliferation into the arterial subintima.188,189

As a result of increased endothelial permeability, lipoprotein particles, mainly LDL, enter the arterial walls where they become trapped and form what is known as a fatty streak. Over time, accumulation and chemical modification (e.g., oxidation and glycation) of LDL particles in the endothelial wall has a pro-inflammatory effect on the vessel, leading to the recruitment of leukocytes, mainly monocytes and to a lesser extent T-lymphocytes.187–189 In this inflammatory state, monocytes penetrate the arterial walls and differentiate into phagocytic macrophages that increase the uptake of modified LDL particles into the arterial wall to eventually form a foam cell. Foam cells, along with activated platelets and endothelial cells, release cytokines and growth factors that stimulate the migration and proliferation of smooth muscle cells, exacerbating the inflammatory state and leading to the formation of a fibrofatty lesion. As fibrosis continues, smooth muscle cell death leads to the formation of a fibrous capsule surrounding a lipid-rich core. In later stages of atherosclerosis, plaque progression and thickening of the arterial wall can lead to significant narrowing of the lumen of the artery, limiting blood flow.187

Over time, hemodynamic stresses and degradation of the cellular matrix increase the risk of the plaque’s fibrous cap rupturing. Depending on the location of the plaque in the body, a rupture plaque may result in the formation of a thrombus (blood clot) on the plaque’s surface or a piece(s) of the plaque may break off and be carried by the bloodstream until it gets stuck in a different part of the artery. Either scenario can result in a partial or complete blockage of the artery, leading to severe complications or death.187

44 Atherosclerotic plaques can form in different types of arteries, including those in the heart, brain, pelvis, legs, arms or kidneys, giving rise to the different types of vascular disease.

Coronary heart disease (CHD) is caused by plaque formation in the arteries in or leading to the heart. A common symptom associated with CHD is angina (chest pain) from reduced blood flow in arteries supplying blood to the heart. Carotid artery disease results from plaque formation in the neck arteries that supply blood to the brain while peripheral artery disease (PAD) results from plaque in the arteries of the extremities, especially the legs. Plaque buildup in the renal arteries is one of the causes of chronic kidney disease.190 The main acute clinical complications of atherosclerosis are MI and stroke, caused by a blockage in an artery of the heart or brain, respectively, and account for the majority of deaths from CVD.191

Multiple factors that may contribute to inflammation and endothelial damage/dysfunction have been identified, including age, male gender, smoking, elevated plasma cholesterol, HTN, hyperglycemia, and some inflammatory markers.187 Elevated LDL cholesterol has been shown to drive the development of atherosclerosis independently of other risk factors.192 Although not entirely understood, it is currently believed that the other risk factors may accelerate the disease process by increasing the atherogenicity of LDL particles by altering the size, number and composition or by increasing the susceptibility of the arterial wall through increased permeability, glycation, or inflammation.187

Figure 3.1. Atherosclerosis Progression In response to vascular injury, inflammation increases arterial wall permeability, allowing lipids to accumulate in the intima forming a fatty streak. Monocytes recruited to the site of injury penetrate arterial walls, differentiating into macrophages and engulfing LDL cholesterol to form foam cells. Over time additional lipids, smooth muscle cells, connective tissue and other cellular debris accumulate in the artery wall. The resulting plaque contains a large lipid core with a thin fibrous covering. The artery expands to accommodate the growing plaque and is vulnerable to rupture and thrombosis. *Adapted from Pepine CJ. Am J Cardiol. 1998; 82(suppl.10A):23S-27S

T2D & Cardiovascular Disease Risk

Individuals with diabetes have a greater burden of the major atherogenic risk factors than individuals without diabetes, primarily hypertension (HTN), dyslipidemia, and obesity.193 Other risk factors that may directly contribute to ASCVD development in individuals with T2D are increased oxidative stress, increased coagulability, endothelial dysfunction and autonomic neuropathy.183,194 A robust body of evidence supports the benefit of managing individual CVD risk factors to prevent or delay ASCVD in individuals with T2D, with the greatest protection conferred when multiple risk factors are addressed simultaneously.184

Obesity

As discussed in the previous chapter, obesity and insulin resistance are associated with the overexpression of multiple adipocytokines, including tumor necrosis factor- α, interleukin

(IL)-1, IL-6, leptin, resistin, monocyte chemoattractant protein-1, plasminogen activator inhibitor-1, fibrinogen and angiotensin.195 High circulating levels of these cytokines leads to

46 increased inflammation and dyslipidemia, which play a key role in the development of endothelial dysfunction and atherosclerosis.183 The chronic, systemic low-grade inflammation characteristic of obesity has therefore been proposed as one of the mechanism linking T2D and

CVD risk.196 In addition, individuals with T2D exhibit decreased production of the adipocytokine adiponectin, which helps to protect against atherosclerosis by increasing nitric oxide production, reducing the expression of adhesion molecules and inhibiting the oxidation of low-density lipoprotein (LDL) cholesterol.195,197 Weight loss is therefore an important component CVD risk reduction for overweight and obese individuals with T2D.

Hypertension

Hypertension is one of the most common comorbidities of diabetes and a major risk factor for both ASCVD and the microvascular complication previously discussed, especially diabetic nephropathy.198 The loss of arterial elasticity and increased blood pressure associated with HTN create physical force against artery walls, leading to corrosion and inflammation over time.199 Recent evidence suggests that common pathways, including obesity, inflammation, oxidative stress, and insulin resistance may underlie the development of both HTN and T2D, and thereby CVD risk.200 Furthermore, the progression of diabetic nephropathy in individuals with poorly controlled diabetes often results in chronic activation of the renin-angiotensin aldosterone system (RAAS), leading to hypertension and nephrotic syndrome, characterized by proteinuria, a hypercoagulable state, and hypertriglyceridemia.193

Dyslipidemia

Individuals with T2D are at an increased risk for developing dyslipidemia compared to individuals without T2D.201 Diabetic dyslipidemia is characterized by a high concentrations of plasma triglyceride concentrations and small dense LDL-cholesterol particles and low HDL

47 cholesterol concentrations.202,203 One mechanism thought to underlie this risk is the increased release of free fatty acids into circulation by insulin resistance adipocytes, which promotes triglyceride production, secretion of apolipoprotein B and very low-density lipoprotein (VLDL) cholesterol.204–206 Dyslipidemia may occur in T2D through other mechanisms as well. For instance, insulin resistance and inflammation are associated with low HDL cholesterol levels207,208 while hyperglycemia may negatively alter LDL and VLDL cholesterol through modifications such as increased glycosylation and oxidation.193 Evidently, targeting CVD risk factors is a vital component of T2D management.

Prevention

The abundance of evidence linking T2D and CVD has informed several well-conducted studies that examined the efficacy of treating CVD risk factors in individuals with T2D. Two landmark intervention trials, Action to Control Cardiovascular Risk in Diabetes (ACCORD) and

Action in Diabetes and Vascular Disease: Preterax and Diamicron MR Controlled Evaluation–

Blood Pressure (ADVANCE- BP), examined the effect of tight blood pressure control with medication on CVD risk in subjects with T2D. While the ACCORD study did not find a benefit of intensive blood pressure treatment for the primary outcome (nonfatal MI, nonfatal stroke, and cardiovascular death) compared with standard treatment, follow-up analysis showed a strong interaction between glycemic control and blood pressure control with a significant risk reduction in subjects with intensive blood pressure and intensive glycemic control.209,210 In the

ADVANCE-BP study, the active blood pressure intervention arm did have a reduction in the risk of both the primary composite endpoint (major macrovascular or microvascular event) and death from any cause or cardiovascular causes.211 Well-controlled studies of lifestyle modification in the treatment and management of HTN in individuals with T2D are lacking. Current

48 recommendations for T2D rely on findings from the Dietary Approaches to Stop Hypertension

(DASH) study that demonstrated the antihypertensive effect of a healthy dietary pattern, comparable in magnitude to pharmacologic therapy.212 A lifestyle intervention to reduce excess body weight, lower sodium intake, increase consumption of fruit, vegetables, and low-fat dairy products, limit alcohol consumption and increase activity levels is encouraged even in individuals with T2D who have mildly elevated blood pressure as it may also benefit glycemia and lipid control.184

A strong evidence base does exist, however, for the effectiveness of lifestyle modification for ASCVD risk reduction through lipid management in individuals with T2D. Current recommendations include a diet focusing on weight loss (if needed), reduced saturated fat, trans- fat and cholesterol intake, increased plant stanol/sterol, omega-3 fatty acids and viscous fiber intake and increased physical activity to improve lipid profiles in individuals with T2D.184 As with HTN, simultaneously optimizing glycemic control in individuals with diabetic dyslipidemia is the most beneficial approach. Pharmacologic interventions using high-intensity statin therapy are recommended in additional to lifestyle therapy for individuals of all ages with diabetes and

ASCVD.184 An intensive lifestyle intervention that focuses on weight loss through decreased calorie intake and increased physical activity, as done in the Action for Health in Diabetes (Look

AHEAD) trial may also be considered for improving glucose control and some ASCVD risk factors.213

Only a small body of clinical trial data has addressed the primary prevention of CVD in individuals with T2D. Despite the proven effectiveness of lifestyle modification in the reduction of ASCVD risk factors and intermediate markers of risk,214,215 the evidence surrounding the association with clinical outcomes remains inconclusive. Although the Look AHEAD

49 intervention resulted in greater weight loss, glycemic control and improvements in CVD risk factors in the intervention group, no association was found between an intensive lifestyle intervention focusing on weight loss and composite of death from cardiovascular causes, nonfatal MI, nonfatal stroke or hospitalization for angina in overweight or obese subjects with

T2D.213 Furthermore, follow-up studies from the China Da Qing trial showed no difference in

CVD event rates or mortality between the intervention and control groups after 20 years of follow-up216, but did find a significantly risk reduction in the intervention group at the 23 year follow-up.217 Similarly, results from a 10 year follow-up analysis of the DPS showed no difference in cardiovascular morbidity between the lifestyle intervention group and controls.218

Both authors note limited power to detect small differences between the trial arms given the small sample sizes of the follow-up cohorts, and the reports have received criticism since the original trials ended early, which may introduce bias for long term analyses.

On the other hand, the smaller Steno-2 randomized trial, found a reduction in CVD events in subjects with T2D and microalbuminuria who received an 8-year intervention aimed at multiple risk factors. In this study, 80 subjects were randomized to receive conventional treatment and 80 received intensive treatment with stepwise implementation of behavior modification and pharmacologic therapy that targeted hyperglycemia, HTN, dyslipidemia, and microalbuminuria, along with secondary prevention of cardiovascular disease with aspirin.

Subjects receiving the intensive treatment had a 47% reduced risk of CVD, defined as a composite of death from cardiovascular causes, nonfatal myocardial infarction, coronary-artery bypass grafting, percutaneous coronary intervention, nonfatal stroke, amputation as a result of ischemia, or vascular surgery for peripheral atherosclerotic artery disease.219

50 Findings from the Japanese Diabetes Complication Study, a nationwide multicenter RCT, found that individuals with T2D in the lifestyle intervention group who received education on dietary habits and physical activity, and adherence to treatment by telephone counseling and outpatient clinic visits, had a lower risk of stroke after 8 years of follow-up than individuals in the control group receiving standard treatment. Interestingly, most classic cardiovascular risk factors, including body weight, glycemia, lipids and blood pressure, as well as the incidence of

CHD, retinopathy and nephropathy did not differ between the control and intervention groups.220

Further research is needed to elucidate the utility of lifestyle modification for the prevention or delay of CVD events in individuals with T2D.

Dietary Patterns and CVD Risk in Individuals with T2D

Although diet is widely accepted as an integral component of T2D management, strong evidence for the optimal dietary approach for glycemic control and cardiovascular disease risk reduction in individuals with T2D is lacking. Despite efforts to examine this relationship through intervention and observational studies, the findings remain inconclusive. Evidence-based recommendations for nutrition therapy for individuals with T2D even vary among the prominent institutions with respect to percent of total calories from carbohydrates, protein, fat, glycemic index (GI), and fiber.102,221,222 Current guidelines suggest that multiple dietary approaches may be effective for T2D management, but the evidence to support these claims, especially related to

CVD risk, is moderate at best.102,222 The reasons for this knowledge gap likely stem from the lack of quality studies available223 and the complex heterogeneous nature of T2D.

Intervention Studies

Many intervention trials have been conducted to examine the effect of different types of diets on T2D management, but fewer have focused specifically on cardiovascular risk in

51 populations with T2D. Systematic reviews and meta-analyses have shown that low glycemic index (GI), high fiber, and Mediterranean diets improve blood glucose management in individuals with T2D.224–226 While most studies focus on reductions in HbA1C and/or weight loss as primary outcomes, some have also included CVD risk factors such as blood pressure and lipids as secondary outcomes of interest. Early intervention trials were often limited by small sample size, lack of control groups, and short follow-up time and as a result produced inconsistent findings and few meaningful conclusions.223,227 Over the past decade, several well conducted RCT’s have suggested that low carbohydrate, high fiber, low GI, Mediterranean, and vegetarian diets are effective in improving various markers of cardiovascular risk in individuals with T2D.228 However, the interpretation and application of these findings remains limited by study design and the lack of hard clinical endpoint outcomes.

Carbohydrate restriction has been shown to be beneficial for diabetes management through improved glycemic control and weight loss, with additional beneficial effects on certain lipids for CVD risk reduction.229,230 A recent meta-analysis by Ajala et al. found that dietary interventions with low-carbohydrate diets, compared to low-fat or low-GI diets over a 6- to 24- month period, led to significant improvements in the lipid profiles of individuals with T2D, with a 10% increase in HDL, 1% reduction in LDL, and 9% reduction in triglycerides.228 However, an earlier review published by Wheeler et al. showed that while low carbohydrate diets improved markers of glycemic control, they had no significant effects on lipids compared with control diets.231 Several limitations must be kept in mind when interpreting these findings. The most striking is the lack of a standard definition of “low carbohydrate” across intervention studies.

Differences between low carbohydrate treatments among studies are quite large, with the percent of total calories from carbohydrates ranging from 13-45% in the carbohydrate-restricted study

52 arms.232–236 This makes the application of findings from meta-analyses ambiguous and impractical in the clinical setting. Furthermore, limited treatment adherence, as is common with long term dietary intervention trials, impacts the strength of the associated that can be detected.232,234 Although these studies were done exclusively in individuals with T2D, the populations under study still had notable differences. Findings from an intervention in adults with T2D managed by diet alone233 may not be reproducible in individuals requiring pharmacological interventions for glycemic management. Additionally, some studies included only overweight or obese adults with T2D232,234–236 while others did not exclude or include participants based on BMI.233,237 Studies using a standardized low-carbohydrate intervention in different populations of individuals with T2D are needed for a more conclusive understanding of one dietary approach that may improve comorbidity risk.

In addition to carbohydrate content, carbohydrate quality has also been proposed to benefit glycemic control and cardiovascular risk in individuals with T2D. Two common approaches focusing on carbohydrate quality are high fiber diets and low GI diets. In the late

1970’s Jenkins, et al. conducted several controlled intervention trials that demonstrated that unabsorbable carbohydrates, such as gums and viscous fibers, slow carbohydrate absorption, decrease the postprandial rise in serum glucose and insulin, and also effectively lower cholesterol in individuals with T2D.238–242 While the effect of whole grains and dietary fiber on glycemic control in individuals with T2D remains inconclusive,231,243 a substantial body of evidence has supported the role of dietary fiber in CVD risk reduction through its beneficial effects on lipoproteins and blood pressure in both individuals with and without T2D.244–250

In 2004, Anderson, et al. conducted a systematic review of studies that examined the effects of both dietary carbohydrate and fiber content in individuals with T2D. Of the 24 studies

53 that met the inclusion criteria, only 13 were randomized controlled trials and included in the subsequent meta-analysis. The levels of carbohydrate intake were classified as high-carbohydrate

(60% of daily energy intake), moderate (30-59.9% energy), and low (30% or less of energy).

Fiber levels were classified as high-fiber (20g/1000 Calories), moderate (10-19.9g/1000

Calories), and low (10g/1000 calories). The results showed that in participants with T2D, moderate-carbohydrate, high-fiber diets compared to moderate-carbohydrate, low-fiber diets were associated with significantly lower total and LDL cholesterol, and triglycerides. High- carbohydrate, high-fiber diets provided even greater benefits on glycemia and reduction in lipid values for individuals with T2D when compared to low or moderate carbohydrate, low or moderate fiber diets, however these diets were also lower in saturated fat and cholesterol than the control diets.251 The studies included in this meta-analysis were limited by small sample sizes of

13-16 participants on average and short treatment periods of 1 week to 3 months. Further, not all studies provided data on body weight change during the intervention, which is known to impact serum glucose and lipoprotein levels.252 Although the results of the meta-analysis were significant, not all studies included produced positive findings. For example, a randomized crossover study of 23 individuals by Jenkins, et al. showed no difference in body weight, fasting blood glucose, HbA1c, serum lipids, blood pressure or a number of other CVD risk factors between high cereal fiber (additional 19 g/day) and low cereal fiber (4 g/day) treatments. The authors noted that the lack of effect may have been due to the short treatment duration or the fact that insoluble fibers, such as wheat bran, do not appear to have the same effect on conventional markers of glycemia and CVD risk factors as soluble fibers.253

Several years after the meta-analysis was published, a larger randomized parallel study by Jenkins, et al. demonstrated a beneficial effect of a high cereal fiber diet on glycemic control,

54 as measured by HbA1c reduction, but no change in serum lipids. In this study, 210 participants with T2D were randomized to receive dietary advice on a high cereal fiber or low GI diet for 6 months. Participants randomized to the high cereal fiber diet were advised to choose the “brown” or whole grain option of breads, cereals and other carbohydrates. As the high cereal fiber diets contained both soluble and insoluble fibers, the significant, yet modest reduction in HbA1c (-

0.18%; 95% CI -0.29 to -0.07%) is consistent with earlier findings.254 Based on the available evidence of the general health benefits of fiber, the current ADA guidelines recommend individuals with diabetes follow the same guidelines as the general public for fiber intake of about 25 g/day for adult women and 38 g/day for adult men, by consuming at least half of all grains as whole grains.255

Though sometimes criticized for being less easily translated into nutrition guidelines than a high fiber diet, low GI diets have also been shown to improve glycemic response and cardiovascular risk in individuals with T2D. Pioneering work in this area by Jenkins, et al. using controlled feeding interventions in the early 1980’s first shed light on the fact that different carbohydrate food sources affect glycemic response differently.256–258 The GI is a method of ranking foods according their glycemic effect on the body, defined as the area under the 2-hour blood glucose response curve after the ingestion of 50 grams of carbohydrate. The GI of a specific food is determined by dividing its area under the curve by that of a standard (glucose or white bread) and then multiplying by 100.256,259 The GI is difficult to translate into direct guidelines, however, because the glycemic response of a meal depends not only on the GI, but also the type and quality of fat, protein and carbohydrate included in the meal.260 Nonetheless, in a randomized crossover study, Wolver, et al. found that consumption of a diet with low GI carbohydrate foods compared with higher GI carbohydrates significantly decreased postprandial

55 blood glucose response and lowered fasting serum fructosamine and cholesterol levels in individuals with T2D.261

In the 2008 randomized, parallel study by Jenkins, et al., 6-month dietary treatment with a low GI diet resulted in a significant increase in HDL cholesterol levels of 1.7 mg/dl (95% CI

0.8-2.6 mg/dl) and lower HbA1c levels compared with a high cereal fiber diet.254 In light of contradictions regarding the benefit of fruit for individuals with diabetes due to sugar content, a secondary analysis of this data was conducted to examine the relationship between fruits based on GI and glycemic response. Results showed that an increase in low GI fruit consumption was associated with lower HbA1c, systolic blood pressure and CHD risk (based on the Framingham cardiovascular risk equation) and higher HDL cholesterol.262 Similarly, an RCT that included

121 participants with T2D compared the effect of consumption of legumes, which are very low

GI foods, to insoluble fiber from whole wheat consumption on glycemic control and CHD risk score. Relative to the high wheat fiber diet, the low GI legume diet resulted in greater reduction in HbA1c values and lower CHD risk.263

The meta-analysis by Ajala et al., found that compared with other diets, low GI diets beneficially increased HDL, with no significant reduction in LDL or triglycerides in individuals with T2D.228 An earlier meta-analysis by Anderson et al. showed that compared to high GI diets, low GI diets were associated with a 6% decrease in serum triglycerides individuals with T2D, and although total and LDL cholesterol levels were lower, the difference was not statistically significant.251 Although 9 of the 10 studies included in this analysis were RCTs, they were small

(average 14 subjects), of short duration (average 33 days), and some included individuals with type 1 diabetes and children, making the extrapolation of the findings to the generally older adult population with T2D questionable. Furthermore, like intervention trials using low carbohydrate

56 diets, conclusions from trials using low GI diets are limited by the variable definitions of “low

GI” in the intervention arms. Separating the independent effects of GI from those of dietary fiber on glycemic control and other outcomes is also an important challenge and limitation of many of these studies.

One specific dietary pattern that has been well defined and extensively studied is the

Mediterranean diet; characterized by high intake of fruits, vegetables, whole grains, nuts, seeds, legumes and olive oil; moderate intake of dairy productions (mainly cheese and yogurt), fish, poultry and wine; and low intake of red and processed meats.264 A number of well-conducted intervention trials have demonstrated the effectiveness of a Mediterranean diet for improved glycemic control and CVD risk reduction in high risk individuals.265,266 The high omega-3 and monounsaturated fat content, high fiber and antioxidant vitamins of a Mediterranean diet compared to the standard low-fat ADA diet, represents a plausible biological mechanism for this relationship.267–270 High MUFA intake in particular has been shown to improve insulin sensitivity,271 lead to more favorable lipid profiles,272,273 and reduce postprandial glucose concentrations.274

The Lyon Diet Heart Study was an early landmark intervention trial to study the cardioprotective effects of a Mediterranean diet. This randomized, single-blind, secondary prevention trial was designed to test whether a Mediterranean type diet compared with a prudent

Western type diet reduced the recurrence of cardiovascular complications following a first MI.267

Between 1988-1992, 605 eligible patients who survived a first MI were randomized to either the experimental group and instructed to follow a Mediterranean type diet or the control group. The control group was not given specific dietary instructions, but advised to follow a prudent diet by their physicians. Primary outcomes included cardiovascular and nonfatal acute MI and secondary

57 outcomes included unstable angina, stroke, heart failure, pulmonary or peripheral embolism.267

An intermediate analysis showed a strong protective effect of the Mediterranean diet and the trial was stopped early. Interestingly, serum lipids, blood pressure and BMI remained similar between the two groups, but the event rates were significantly different. After 27 months, 59 primary and major secondary events had occurred in the control group, compared with 14 in the experimental group (adjusted HR 0.24, 95% CI 0.13-0.44). When all primary and secondary endpoints were combined, a total of 104 events were documented in the control compared with 68 events in the experimental group (adjusted HR 0.63, 95% CI 0.46-0.87).267 An extended follow up analysis showed that the protective effect of the Mediterranean type diet persisted for up to 4 years.275

The results also showed that traditional risk factors, such as high blood pressure, total and LDL cholesterol, were independent and joint predictors of event recurrence, suggesting that a cardioprotective diet may represent a separate component of risk reduction, to be used in conjunction with other interventions (e.g., pharmacological) aimed at modifying known risk factors. The Lyon Diet Heart Study did not identify subjects with T2D within the study population. Although individuals with T2D are at high risk of developing CVD, their risk is different than someone who has already had a clinical CVD event.

The largest intervention trial to investigate the effects of a Mediterranean diet was the

PREDIMED (Prevención con Dieta Mediterránea) study. The PREDIMED was a parallel-group, multicenter, randomized, controlled 4-year clinical trial designed to assess the effects of the

Mediterranean diet on the primary prevention of CVD.276 Beginning in 2003, 7,447 high risk individuals were randomized to one of three arms, each receiving quarterly education about their prescribed diet. Eligible participants had T2D or at least three of the following major CVD risk factors: smoking, hypertension, elevated LDL cholesterol levels, low HDL cholesterol levels,

58 overweight or obesity, or a family history of premature coronary heart disease. The two intervention arms were a Mediterranean diet plus provision of either 1 liter/week of extra-virgin olive oil (EVOO) or 30 grams/day mixed nuts and a low-fat diet served as the control group. All diets were ad libitum and did not include an increase in physical activity as part of the intervention.277 The primary endpoint of this study was a composite of MI, stroke, and death from cardiovascular causes. Secondary endpoints included stroke, MI, death from cardiovascular causes and all-cause mortality. Although the study was not done exclusively in individuals with

T2D, approximately 50% (3,614) of the study population had T2D at baseline. Participants were overweight and obese adults, with an average age of 67, the majority of whom had dyslipidemia and HTN.277 After 1 year of follow up, analyses of intermediate markers of cardiovascular risk showed positive effects of both Mediterranean diets on blood pressure, lipid profiles, lipoprotein particles, inflammation, oxidative stress and the expression of pro-atherogenic genes involved in vascular events and thrombosis.278 At the end of the study (mean follow-up 4.8 years), 96 primary outcome events had occurred in the group assigned to a Mediterranean diet with EVOO

(p = 0.009), 83 in the group assigned to a Mediterranean diet with nuts (p = 0.02), and 109 in the control group. Multivariate analyses showed a protective effect of both Mediterranean diets compared with the control diet for the primary endpoint (EVOO HR 0.70, 95% CI 0.54-0.92; nuts HR 0.72, 95% CI 0.54-0.96).277 No separate analyses were done for individuals with T2D.

However, like the Lyon Diet Heart Study, the use of clinical events as the main outcomes as opposed to CVD risk factors, provides a strong evidence base for the efficacy of the

Mediterranean diet for risk reduction in high risk populations.

Several RCTs have investigated the relationship between a Mediterranean diet and cardiovascular risk in individuals with T2D. A randomized trial by Esposito et al. compared the

59 effect of a Mediterranean versus low-fat diet on the need for antihyperglycemic medication in overweight patients with newly diagnosed T2D and included markers of cardiovascular risk as a secondary outcome.279 Participants were randomized to either a low-carbohydrate Mediterranean type diet (<50% daily calories from carbohydrates, n=108) or to a low-fat diet based on the

American Heart Association Guidelines (<30% of daily calories from fat, n = 107). Both groups received dietary advice from study dietitians in monthly sessions during the first year and bimonthly sessions for the remaining 3 years.279 Throughout the trials, participants assigned to the Mediterranean diet group showed significantly greater increased in HDL cholesterol levels and decreased in triglyceride levels than those in the low-fat group. Total cholesterol levels decreased more in the Mediterranean diet group, but this difference was only statistically significant during the first 2 years. Both systolic and diastolic blood pressure decreased more in the Mediterranean diet group, but the between-group differences were no longer significant by the end of the study.279 The patients included in this study represent a unique group individuals with T2D, as they were newly diagnosed and not taking anti-hyperglycemic medications.

Elhayany et al. conducted a randomized three arm intervention trial to compare the effects of a low carbohydrate Mediterranean (35% calories from carbohydrate, 45% calories from fat (50% MUFA), traditional Mediterranean, and 2003 ADA diet on health parameters in individuals with T2D over a 1 year period. Participants received dietary counseling by the same dietitian every 2 weeks the study and were encouraged to engage in 30–45 minutes of aerobic activity at least 3 days a week.235At the end of the study period, participants in both

Mediterranean diet groups had greater reductions in triglycerides compared with the ADA diet group. Only subjects in the low carbohydrate Mediterranean arm had significantly increased

60 HDL levels compared to ADA diet, and LDL cholesterol levels were reduced in all three groups.235

Toobert et al. randomized 279 postmenopausal women with T2D to a comprehensive lifestyle self-management program that included a Mediterranean low-saturated fat diet or a usual care control group. Women in the intervention group showed significant improvements in

HbA1c, BMI, plasma free fatty acids and quality of life after 6 months of follow-up. Although favorable patterns in lipids and blood pressure were also seen in the intervention group, they did not reach statistical significance.280 As seen for low carbohydrate diets, differences in dietary interventions, in terms of diet composition and delivery of dietary advice, are limiting factors in this body of research but may explain some inconsistencies in the findings. Despite a marked improvement of traditional cardiovascular risk factors, the findings of these studies remain limited by the lack of hard clinical outcome data.

A few small intervention trials have examined the relationship between vegan or vegetarian diets and CVD risk in individuals with T2D. A randomized intervention trial by

Barnard et al. compared the effect of a low-fat vegan diet (∼10% of energy from fat, 15% protein, and 75% carbohydrate) with a standard ADA diet (15–20% protein, <7% saturated fat,

60–70% carbohydrate and monounsaturated fats, and cholesterol ≤200 mg/day) on glycemic control and cardiovascular risk factors in 99 overweight adults with T2D. After 22 months, those in the vegan diet arm had significantly lower total cholesterol, LDL cholesterol and HbA1c levels.281 To compare the effects of a calorie-restricted vegetarian and conventional diabetic diet on insulin resistance and oxidative stress, Kahleova et al. randomized 74 adults with T2D to one of two parallel arms in a 24-week study. Participants in the vegetarian diet group showed increased insulin sensitivity and greater reduction in markers of oxidative stress compared to the

61 control group. In this case, greater weight loss, specifically reduction in visceral fat, was strongly correlated with these changes and thought to be the underlying mechanism.282 While these studies suggest that vegan and vegetarian diets may be beneficial for reducing cardiovascular risk in individuals with T2D, the findings are far from conclusive.

Observational Studies

Evidence-based nutrition recommendations that support glycemic control, a healthy body weight, and optimal lipid profiles to prevent the chronic comorbidities of diabetes is paramount for comprehensive diabetes management. While the association of a healthy diet pattern with reduced risk of CVD has been clearly demonstrated in the general population,145,283,284 evidence from observational studies supporting this relationship in individuals with T2D is very limited and to date no prospective cohort studies have been conducted in this population.

A cross-sectional study conducted by Lim et al.285 was one of the first observational studies to investigate the association of dietary patterns with cardiovascular risk in adults with

T2D. Using data from the 4th Korean National Health and Nutrition Examination Survey

(KNHANES), 2007-2008 the authors studied the relationship between dietary patterns and blood lipid profiles in Korean adults with T2D. KNHANES is a nationwide survey conducted by the

Korea Centers for Disease Control and Prevention that includes general and health-related questionnaires, a health examination and dietary survey. The 2007-2008 surveys were completed by 14,338 individuals. Respondents older than 30 years of age, who completed the 24-hour recall in the diet survey and had data from the health examination with anthropometric measurements, were eligible for inclusion in the study. To identify individuals with T2D, the ADA diagnosis definition of fasting plasma glucose ≥ 126 mg/dL and/or using anti-diabetic medication and/or

62 self-reported physician diagnosed diabetes was used. A total of 680 subjects were included in the final analyses.

For dietary pattern analysis, data from 24-hour recalls collected as part of the survey was used to create 22 common food groups based on classification in the Korean Nutrient Database to serve as food input variables for factor analysis. Four major dietary patterns were identified using principal component analysis and factor scores were obtained for each participant: the

“Meat & Bread & Alcohol” pattern characterized by high consumption of breads, sugars, meats, oil, beverage, and alcohols; the “Noodle & Seafood” pattern, characterized by high consumption of noodles, kimchi, fish, and seaweed; the “Rice & Vegetable” pattern represented the highest loading of rice and considerable high loading of vegetables and egg, although also appeared in other patterns; and the “Korean Healthy” pattern, characterized by high consumption of whole grains, legumes, nuts, vegetables, mushrooms, and fruit. Based on their factor score, participants were categorized into four groups with quartiles, such that subjects in the highest quartile of a pattern reported a diet that most closely reflected that dietary pattern.

In the analysis, categorical variables were compared using chi-square tests and continuous variables were compared using generalized linear models, both based on quartile of dietary pattern scores. Covariates that were considered as potential confounders included age, sex, BMI, energy intake, educational level, household income, smoking, physical activity, duration of diabetes, and diabetes treatment in all models. The authors found that BMI, waist circumference, fasting, glucose and HbA1C were not associated with any of the dietary patterns after adjusted for all potential confounders. The highest quartile of “Bread & Meat & Alcohol” pattern had significantly higher total and HDL serum cholesterol levels than those in the lowest quartile, while those in the highest quartile of the “Korean Healthy” pattern lower total

63 cholesterol than those in the lowest quartile, LDL cholesterol levels were not included in the reported results. Serum triglyceride levels were also significantly decreased in the “Korean

Healthy” pattern. No association was found with dietary pattern and blood pressure. Thus, the authors concluded that dietary patterns were associated with blood lipid profiles among Korean adults with T2D. Particularly, that a diet high in grains, legumes, nuts, vegetables, and fruits is associated with favorable lipid profiles. Strengths of this study include the use of a large, nationwide study sample and dietary pattern analysis, which better captures the complexity of the diet-disease relationship than the use of single nutrients or foods. However, the findings may be limited by misclassification of the subjects by the use of a single measurement of fasting blood glucose and self-reported treatment. In addition, causal inferences cannot be made from cross- sectional studies because a temporal relationship between diet intake and blood lipids is not available.

In 2016, Osonoi et al.286 conducted a cross-sectional study to assess the relationship between dietary patterns and CVD risk factors in Japanese men and women with T2D and no history of CVD. Subjects receiving outpatient treatment for T2D were screened and recruited from three clinics in three Japanese cities. A total of 1,032 consecutive subjects were screened,

906 met the inclusion criteria and 736 agreed to enroll in the study. Eligible participant were patients with T2D between the ages of 25-70 years who consented to participate in the study.

The exclusion criteria included: type 1 or secondary diabetes, presence of a severe infectious disease, recent surgery, or trauma, history of MI, angina pectoris, cerebral stroke or infarction, chronic renal failure requiring dialysis, liver failure, heart failure, active cancer, currently pregnant or breastfeeding. Data was collected using validated self-administered questionnaires and fasting blood samples were obtained at the outpatient clinics.

64 Dietary intake during the preceding month was assessed with the validated, self- administered, Brief Diet History Questionnaire, a structured questionnaire that includes questions about the frequency of consumption of 56 selected food and beverage items with specified serving sizes. Factor analysis identified six dietary patterns. Factor 1, labeled “Seaweeds,

Vegetables, Soy products and Mushrooms” was characterized by high intake of seaweed, vegetables, soy products and mushrooms. Factor 2, which was characterized by high intake of fish, potatoes, meat, fats and oils, was labeled “Fish and Meat” pattern. Factor 3 was high in noodle and soup consumption and named “Noodle and Soup” pattern. Factor 4 was characterized by meat, fats and oils, seasonings and eggs, and was named “Meat, Fats and Oils, Seasonings and Eggs” pattern. Factor 5 was characterized by sweet, fruit and dairy products and was labeled

“Fruit, Dairy products and Sweets” pattern. Factor, characterized by high consumption of rice and miso soups, and was named “Rice and Miso soups” pattern. Diet pattern scores were calculated for all patients and ranked into quintiles for each dietary pattern. Trends across quintiles were analyzed using linear regression for continuous variables and logistic regression for categorical variables. The fully adjusted model included age, gender, BMI, Morning Evening

Questionnaire score, Pittsburg Sleep Quality Index, Beck Depression inventory-II, physical activity and smoking as covariates.

The authors found that among the six dietary patterns identified, patients in the highest quintile of the “Seaweeds, Vegetables, Soy products and Mushrooms” had lower use of diabetes medications and healthier lifestyles, compared to those in the lowest quintile. A diet reflecting the “Noodle Soup” pattern was associated with higher BMI and triglyceride levels, as well as lower renal function. The “Fruit, Dairy products and Sweets” pattern was associated with lower blood pressure, albuminuria, and brachial-ankle pulse wave velocity. No significant association

65 was found between the remaining three dietary patterns and CVD risk factors. The authors therefore concluded that dietary patterns could be a potentially important therapeutic target to achieve metabolic control and prevent the onset of CVD in patients with T2D. However, this study is limited by the use of self-reported data that may be influenced by recall and/or social desirability bias, a diet history that included only 56 food items and may have left out others associated with CVD risk, lack of hard clinical endpoints, and inability to establish a temporal relationship between diet and CVD risk due to the cross-sectional study design.

Shadman et al.287 also conducted a cross-sectional study to investigate the relationship between dietary patterns and cardiometabolic risk factors in individuals with T2D. In this study, the association of dietary patterns with glycemic control, blood pressure, lipid profile and anthropometric measurements were analyzed in 525 Iranian adults with T2D. Participants were selected from patients followed for at least two years by the Diabetes and Metabolic Disease

Clinic of Tehran University of Medical Sciences. Patients aged 35-65 years old, who were diagnosed with diabetes after the age of 30 at least 5 years prior were eligible for enrollment in the study. Exclusion criteria included insulin use, medication change during the past year, history of MI, angina pectoris, stroke, chronic inflammation, major illness, alcohol consumption, smoking, vegetarian diet and pregnancy. Dietary intake was assessed using a validated 168-item

FFQ administered in-person by a trained registered dietitian at the clinic. Twenty-two food groups were then created based on nutrient content for diet pattern analysis. Fasting blood samples were collected for biochemical analysis, blood pressure and anthropometric data was collected at the clinic, and physical activity was self-reported. For the analysis, univariate and multivariate linear regression models were used to assess the association of dietary pattern adherence with cardiovascular risk factors. Covariates included as potential confounders

66 included age, sex, diabetes, duration, type and drug dosage (such as hypoglycemic, antihypertensive and cardiovascular drugs), calorie, dietary protein and fat intake.

Using principal component analysis, the authors identified three main dietary patterns: a

Western pattern, characterized by high intake of sweets, fast food, carbonated drinks, red meat, mayonnaise, nuts, refined grains, potato and visceral meat; an Asian pattern, high in vegetables, fish, poultry, egg, nuts and low-fat dairy; and a Traditional pattern, rich in high-fat dairy, oils, whole grains, vegetables and fruits. Participants were ranked into tertiles for adherence to each of the dietary patterns, with the highest tertile representing a diet most adherent to the dietary pattern. The Western dietary pattern was positively associated with fasting serum glucose, total cholesterol, and LDL cholesterol; however, after adjustment for confounders only the association between serum total cholesterol and Western like pattern remained significant. No associations between cardiovascular risk factors and Asian or Traditional diet patterns were seen. The authors note that the lack of association detected may have resulted from extensive adjustment of potentially important mediating factors in the statistical models, such as dietary protein and fat intake.

The only observational study to examine the relationship between a specific dietary pattern and cardiovascular risk in individuals with T2D was conducted by Ciccarone, et al. in

2003.288 Using a nested case-control design, the authors investigated the association between a

Mediterranean dietary pattern and PAD in Italian individuals with T2D. Cases and controls were selected from a cohort of 944 individuals with T2D who were consecutively seen at the Diabetes

Care Center of the General Hospital of Pescara from 1996-2000 and free of any major concomitant disease. From the participating patients, 144 cases of PAD were clinically identified and confirmed. Of the remaining 588 patients who were free of macrovascular complications,

67 288 age and sex frequency-matched controls were selected. All patient information was collected by a standardized questionnaire and trained monitors reviewed medical records and used standardized criteria to confirm or reject the diagnosis of all reported cardiovascular events.

Dietary information was obtained from a validated semi-quantitative 24-item FFQ, which included information on the frequency of consumption and portion size of each item. A

Mediterranean dietary score was calculated by assigning a score of 1 to any food

(frequency/amount) item with adequate supporting evidence for a beneficial effect on coronary artery disease risk, and a score of 0 for items with potentially harmful effects. The total of all single item scores were summed to produce a final score for each individual, with higher final scores indicative of healthier eating habits. The validity of the score was assessed by comparing the results obtained using the FFQ with a 7-day food record for a subsample of participants. All cases were age- and sex-matched with two control patients. Dietary pattern scores were grouped into 3 categories based on tertiles of distribution of the diet score within controls: 0-8 points, 9-

10 points, and ≥ 11. For the analysis, means for continuous variables were compared by analysis of variance and odds ratios were calculated by multivariate conditional to matching logistic regression. The multivariate models were adjusted for duration of diabetes, smoking habits, history of hyperlipidemia, hypertension, BMI and physical activity. The results identified dietary pattern score, diabetes duration and hypertension as variables significantly and independently associated with risk of PAD. The highest tertile of dietary pattern was associated with 56% reduced risk of PAD compared to the lowest tertile (OR 0.44, 95% CI 0.24-0.83).

The existing evidence on dietary patterns and CVD risk in individuals with T2D limited by the use of single measurements of dietary intake at only one point in time, which does not capture the dynamics of dietary habits over time, and the lack of clinical cardiovascular event

68 outcomes. Another significant limitation to these studies is the lack of a prospective design, resulting in limited ability to infer a causal association due to the lack of established temporal relationships in cross-sectional studies. Therefore, large, prospective, population studies evaluating the relationship between well-characterized healthy eating patterns and cardiovascular disease risk, as measured by clinical endpoints, in individuals with T2D are necessary.

Considering the growing diabetes epidemic, this evidence is urgently needed to inform accurate dietary guidelines as a critical component of T2D management and comorbidity risk reduction.

69 CHAPTER 4

Study Populations

Coronary Artery Risk Development in Young Adults Study

Study Design

The Coronary Artery Risk Development in Young Adults (CARDIA) prospective cohort study was initiated in 1984 to examine the development and determinants of clinical and subclinical cardiovascular disease and their risk factors during young adulthood. The study aimed to recruit a sample representative of the U.S. population of black and white, male and female individuals aged 18-30 years that could be stratified to achieve nearly equal numbers of blacks and whites, males and females, ages less than 25 years and greater than or equal to 25 years, and education levels less than or equal to high school and greater than high school. Study participation was limited to the two largest racial groups in the U.S. to produce statistically reliable results for males and females at different ages. Following one year of protocol development and planning, 5,115 back and white women and men, aged 18-30 years, were enrolled in the study. Participants were selected for equal distribution of race, gender, education, and age at each of the four study centers: Birmingham, Alabama; Chicago, Illinois; Minneapolis,

Minnesota; and Oakland, California. Recruitment was done primarily via telephone, expect in some areas of Minneapolis where this was not feasible. In such cases, participants were recruited by door-to-door recruiters. Individuals were eligible for study participation if they were between the ages of 18 and 30 years at the time of the initial telephone recruitment interview, and completed the initial examination prior to their 31st birthday, were white or black, had a permanent residence in a target urban area, and were free of a long-term disease or disability that

70 could interfere substantially with any part of the examination. Women were also excluded for being pregnant or up to 3 months postpartum.289

Following a brief telephone screening questionnaire to determine eligibility, willing and eligible individuals attended a CARDIA initial examination at one of the 4 study centers. The initial examination included standardized measurements by trained and certified CARDIA staff of major risk factors and assessments of psychosocial, dietary, and exercise-related characteristics that might influence them, or that might be independent risk factors. Participants were instructed to come to the examination in the morning after a minimum 12-hour fast and to abstain from smoking or heavy physical activity for 2 hours prior to the examination. The initial examination included a greeting and informed consent, followed by blood pressure measurements and the completion of a sociodemographic questionnaire to collect information on residence, education, occupation, marital and family background. Participants then had blood drawn for clinical chemistries, hematology counts, lipid, lipoprotein and apolipoprotein measurements, insulin, and cotinine levels. Participants completed self-administered medical and psychosocial questionnaires that were sometimes followed by an interviewer questionnaire to collect more detailed information about a reported illness, medication, and smoking habits.

Participants completed pulmonary function tests, followed by a diet interview. An extensive interviewer-administered FFQ was collected using food models, measuring cups, and spoons to assist with portion size estimation. The FFQ, described in detail below, was designed to reflect the previous one month’s usual intake. Finally, body weight and height were measured with participants wearing light clothing and no shoes before participants completed a treadmill test, with physician clearance.289

71 Follow-up examinations were conducted during 1987-1988 (Year 2), 1990-1991 (Year

5), 1992-1993 (Year 7), 1995-1996 (Year 10), 2000-2001 (Year 15), 2005-2006 (Year 20), and

2010-2011 (Year 25), and 2015-2016 (Year 30). Retention rates among consenting subjects were high at each follow-up examination (90%, 86%, 81%, 79%, 74%, 72%, and 72%, respectively).

Initial analyses of the original cohort showed that compared to national samples at the time, smoking was less prevalent and weight tended to be greater in CARDIA participants. Cholesterol levels were representative, but lower blood pressures in CARDIA were likely due, in part, to differences in measurement methods. The study authors also noted several differences in risk factor levels by demographic subgroup, including a higher BMI among black than white women and a much higher prevalence of cigarette smoking among participants with a high school education or less than among those with more education.290

Dietary Intake Assessment

The CARDIA dietary history was developed and validated specifically for the study to provide accurate and reliable quantitative data on habitual individual nutrient intakes. Dietary intake was assessed at years 0, 7, and 20 using an interviewer-administered method that included a short questionnaire about general dietary practices and a comprehensive FFQ about typical food intake, using the previous month as a reference for recall. The FFQ referenced 1,609 separate food items within 100 food categories in years 0 and 7, and several thousand separate food items in year 20. Follow-up questions for selected foods concerned serving size, frequency of consumption, and common additions to these foods. Provision was made for reporting foods not found in the food frequency list. For example, a participant who reported usually eating bread received follow-up open ended questions about the type of bread (e.g., white, whole wheat/mixed grains, rye) and, for each bread type, the frequency of consumption (as servings per

72 day, week, or month). Cue cards were used to prompt responses and plastic food models assisted with estimations of usual amounts consumed. The interview took approximately 45 minutes to complete. Diet history data was coded by the University of Minnesota Nutrition Coordinating

Center (NCC) and foods were placed into 166 food groups using the food grouping system developed by the NCC. Food group intake was calculated as the total number of servings per day. 291

For this analysis, the creation of dietary pattern scores will be modeled after the a priori

Diet Quality Score created in previous CARDIA studies.292,293 From the 166 food groups created by the NCC system, I created 44 condensed food groups based on similar nutrient characteristics, hypothesized health effects, and comparability to food groups defined in previous studies.292–296

(Supplemental Table 1) The study population will be ranked into quintiles of intake for each of the 43 food groups. The dietary pattern scores will then be created by classifying each food group as beneficial (+), adverse (-), or neutral (0) in terms of the recommendations of the specific diet as depicted in Supplemental Table 2. The categorization will be based on published dietary guidelines for each diet pattern examined, and thus scoring of individual diet data will reflect compliance with the pre-specified and defined dietary pattern. The dietary pattern score for each participant will be the sum of category scores 0 to 4 for intake of items considered beneficial by the specific dietary pattern (with the highest quintile receiving a score of 4 and the lowest a score of 0) plus scores in reverse order (4 to 0) for adverse foods (with the lowest quintile of intake receiving a score of 4 and the highest a score of 0). Food categories recommended to be consumed in moderation will be scored 0-2-4-2-0, with moderate intake receiving the highest score. Neutral foods will not be included in the final score.

73 The reliability and comparative validity of the CARDIA dietary history was evaluated a study by Liu, et al. using a sample of 30 white men, 33 white women, 33 black men, and 32 black women aged 18 to 35 years, in four regions reflecting the race, sex, age and geographical distribution of the CARDIA participants. Reliability was tested using two diet history interviews administered one month apart by trained nutritionists. To test the comparative validity of the method, seven telephone-assessed 24-hour dietary recalls were randomly scheduled during a 28- day period and followed by a dietary history interview. The CARDIA Nutrition Working Group selected calories, total fat, saturated fat, polyunsaturated fat, monounsaturated fat, dietary cholesterol, protein, carbohydrate, alcohol, potassium, calcium, and vitamin A for nutrient comparison. The results showed that mean nutrient values from the first history tended to be higher than those obtained from the second, but the differences for most nutrients were significant for blacks but not for whites. Correlations for log-transformed and calorie-adjusted nutrient values from the histories were generally to be in the range of 0.5-0.8 for whites and 0.3-

0.7 for blacks. The average nutrient values estimated from the dietary histories were higher than those estimated from the average of the 24-hour recalls, with the differences being larger in blacks than in whites. The authors concluded that the CARDIA diet history is a reliable and valid dietary survey method for collecting habitual dietary intake information in whites, but noted that the results were less consistent in blacks.297 As these findings raise questions about reporting differences in blacks and whites, all analyses will be stratified by race as a sensitivity analysis.

Diabetes Mellitus Assessment

Diabetes mellitus was identified as a CARDIA study morbidity endpoint and event identification occurred during examinations and annual follow-up of the cohort. Morbidity events were adjudicated by the CARDIA Endpoints Surveillance and Adjudication

74 Subcommittee through review of medical records related to all relevant hospitalizations and outpatient care. Diabetes, however, was adjudicated as an endpoint only if it was the primary reason for a hospitalization. Diabetes status was assessed at baseline and study years 7, 10, 15,

20, 25 and 30. Participants were classified as having diabetes if any of the following criteria was met: fasting glucose ≥ 7 mmol/l; use of medications for diabetes treatment; 2 h GTT ≥ 11.1 mmol/l or HbA1c ≥ 6.5% (48 mmol/mol IFCC). Serum glucose concentrations were measured using the hexokinase method. CARDIA did not differentiate between type 1 and type 2 diabetes mellitus; however, it is likely that most incident cases identified during follow-up are T2D given the age of the cohort. To avoid misclassification bias, sensitivity analyses will be performed excluding subjects who simultaneously report initiation of insulin treatment the same follow-up year as their diabetes diagnosis and those who develop diabetes before age 30, as this cutoff has also been used a surrogate to screen for type 1 diabetes.

Women’s Health Initiative

Study Design

The Women’s Health Initiative (WHI) is a prospective national health cohort study designed to identify strategies for the prevention of heart disease, breast and colorectal cancer, and osteoporotic fractures in postmenopausal women. The study was conducted by 40 Clinical

Centers in 24 states across the U.S. and the District of Columbia. Recruitment of postmenopausal women between the ages of 50-79 years began in 1993 and continued through 1998. The WHI prioritized enrolling women of racial/ethnic minority groups in the same proportion as found in the general population, according to the 1990 census.298 At the local level, Clinical Centers primarily used mass mailings to identify potential participants for screening, supplemented by community presentations, local newspaper ads and articles, public service announcements (TV

75 and radio), and health fairs. National recruitment activities by the National Institutes of Health, the study clinical coordinating center and various study-wide committees, included centrally produced recruitment and screening materials, central workshops and support for local recruitment staff, a national public awareness campaign, and a toll-free recruitment telephone line. Recruitment materials were available in English and Spanish.299 Women who were interested in participating in the study were first screened by phone to confirm basic eligibility, followed by up to three screening visits to complete physical measurements (height, weight blood pressure, heart rate, waist and hip circumference), collection of blood specimens, a medication and supplement inventory, and questionnaires on demographic characteristics, medical history, family history, lifestyle and behavioral factors, and quality of life.298

At the end of the recruitment period, a total of 161, 808 women had joined the WHI, with approximately 17% representing racial/ethnic minorities. The WHI had two major parts, a set of three randomized controlled Clinical Trials (CT) and an Observational Study (OS). Eligible women could be randomized into one, two, or three of the CT components, which included the

Hormone Therapy (HT) Trials, Dietary Modification (DM) Trial, and Calcium and Vitamin D

(CaD) Trial. Briefly, the HT trial examined the effects of estrogen plus progestin or estrogen alone on the prevention of CHD and osteoporotic fractures, and risk for breast cancer,300 the DM trial evaluated the effect of a low-fat dietary pattern on the prevention of breast and colorectal cancers and CHD,301 and the CaD trial evaluated the effect of calcium and vitamin D supplementation of the prevention of osteoporotic fractures and colorectal cancer.302 Women who were ineligible or unwilling to join the CT were invited to join the OS. The OS was designed to examine the relationship between lifestyle, health and risk factors and disease

76 outcomes.303 A total of 68,132 postmenopausal women were enrolled in the CT and 93,676 in the OS. Detailed inclusion and exclusion criteria for the WHI study arms are shown in Table 8.

Table 4.1 Women’s Health Initiative Inclusion and Exclusion Criteria (adapted from Hays, et al. Ann Epidemiol. 2003;13:18-77)299 Component Inclusion Criteria Exclusion Criteria Clinical Trial and Observational 50–79 years of age Postmenopausal Competing risk: Study If age 55, no menstrual period for at Any medical condition with least 6 months predicted survival of 3 y If age 50–54, no menstrual period Adherence or retention reasons: for at least 12 months Ability and Alcohol or drug dependency willingness to provide written Mental illness, including severe informed consent (component depression specific) Intention to reside in area Dementia for at least 3 years Active participation in other randomized intervention trial

Clinical Trial Competing risk: Any invasive cancer previous 10 y Breast cancer at any time Mammogram or CBE findings suspicious of breast cancer MI in previous 6 months Stroke or TIA in past 6 months Chronic hepatitis or severe cirrhosis

Safety reasons: Severe hypertension (systolic BP 200 mm Hg or diastolic BP 105 mm Hg) Severely underweight (BMI 18 kg/m2) Hematocrit 32% Platelets 75,000 cells/ml Current use of oral daily corticosteroids

Adherence or retention reasons: Unwilling to participate in baseline or follow-up examination components Dietary Modification Adherence or retention reasons: Special dietary requirements incompatible with the intervention (e.g., celiac sprue) On a diabetic or low salt diet Gastrointestinal conditions contraindicating a high fiber diet Type 1 diabetes Colorectal cancer at any time Routinely eat 10 meals per week prepared out of the home Unable to keep a 4-day food record FFQ percent calories from fat 32%

77 FFQ energy intakes 600 or 5000 kcal Previous bilateral prophylactic mastectomy Postmenopausal Hormone Therapy Safety reasons: Endometrial cancer at any time Endometrial hyperplasia Malignant melanoma at any time History of pulmonary embolism or deep vein thrombosis Previous osteoporosis-related fracture being treated with hormones History of bleeding disorder requiring transfusion History of hypertriglyceridemia Currently on anticoagulants Currently on tamoxifen Abnormalities in baseline pap smear, pelvic exam, or pelvic ultrasound Adherence or retention reasons: Severe menopausal symptoms that would make placebo treatment intolerable Inadequate adherence to placebo run-in Unable or unwilling to discontinue use of PHT or testosterone Refusal to have baseline endometrial aspiration Calcium and VitD Safety reasons: History of renal calculi or hypercalcemia Current use of oral corticosteroids or calcitriol Intention to continue taking 600 IUs of vitamin D per day

Participants in the OS study were followed for 8 to12 years, while the CT closed out between 2004 and 2005. Overall, study participants were followed for an average of 9 years.

Two WHI Extension Studies continued to follow consenting participants over two additional five-year periods, from 2005-2010 and 2010-2015, and a third Extension Study is currently in progress through 2020. Outcomes are collected in the Extension Studies by mailed questionnaires filled out by the remaining participants.

78 Throughout the study, data was collected through self-administered forms, in-person interviews and clinical measures by trained, certified Clinical Center staff. All clinical outcomes were identified by self-report on a routine medical history follow-up form administered semi- annually to CT participants and annually to OS participants. Participants who reported a potential

WHI-defined outcome were contacted by mail or phone and asked to fill out a detailed medical history update form to provide more specific information on the diagnosis for use in the adjudication process. All clinical outcomes, except diabetes mellitus requiring medication

(obtained by self-report only), were adjudicated locally or centrally. Local adjudication was done a physician at the local Clinical Center confirming or denying a potential clinical outcome.

Depending on the type of outcome, WHI Clinical Centers were also asked to send locally adjudicated case packets to the Clinical Coordinating Center for central adjudication. For

Cardiovascular and death outcomes, centrally-trained cardiovascular physician and neurologist adjudicators reviewed and adjudicated the cases. Primary cancers, including breast, colon, rectum, endometrium, and ovary, were coded centrally by tumor registry coders. All hip fractures were centrally adjudicated at the WHI Bone Density Center.304

79 Table 4.2 Baseline Characteristics of Women’s Health Initiative Final Enrollment Participants (adapted from Hays, et al. Ann Epidemiol. 2003;13:18-77) 299

Type 2 Diabetes Assessment

Diabetes status in the WHI was assessed by self-report at baseline and throughout the duration of follow-up. At baseline, women were asked if a doctor had ever told them they had

‘sugar diabetes or high blood sugar’ when they were not pregnant. At this time, women were also asked about treatment with insulin or oral anti-diabetic medications. Baseline diabetes was defined as an affirmative answer to the above question or the reported use of a medication to

80 treat diabetes. Participants were asked to bring all current prescription medications in the original containers to the baseline visit to create the medication inventory.

In validation studies, diabetes self-reports were compared with medication inventories and fasting glucose levels at baseline. Baseline prevalent medication-treated diabetes was defined as self-report of physician diagnosis and treatment with insulin or oral anti-diabetic medication. Self-report treated diabetes was found to be concordant with medication inventories in 79% of CT and 77% of OS participants and over 99.5% of women without a self-report of treated diabetes had no diabetes medication in their baseline medication inventory. Furthermore,

3% of women who did not report diabetes had fasting blood glucose > 126 mg/dl, 88% of whom subsequently reported treated diabetes during 7 years of follow up.305 The authors of the validation study noted that medication inventories are subject to error and may be incomplete, or fasting glucose may have been lowered by treatment, underestimating concordance. Oral glucose tolerance tests were not performed and fasting glucose was only measured in a small random sample of 5,884 women. A total of 9,618 cases of diabetes were identified at baseline.

Dietary Intake Assessment

The main dietary assessment tool used in the WHI was a FFQ designed by an ad hoc dietary assessment working group composed of WHI scientists and based on instruments used in the WHI feasibility studies and the original National Cancer Institute/Block FFQ. The WHI FFQ was designed to reflect regional and ethnic eating patterns in the U.S. The final questionnaire included many examples of southern foods (e.g., okra was added to the food item “summer squash, such as zucchini”), 12 foods to reflect Hispanic eating patterns, and Indian fry bread for

Native Americans. The FFQ consisted of three sections: 122 composite and single food line items, 19 adjustment questions related to type of fat intake, and 4 summary questions asking

81 about the usual intake of fruits, vegetables and added fats. The main section of 122 foods/food groups included questions on frequency of intake and portion size. Frequency choices ranged from “never or less than once per month” to “2+ per day”. Portion sizes were small, medium, or large compared to the stated medium portion size. The adjustment questions were designed to allow for more refined analysis of fat intake by asking questions about food preparation practices and types of added fats used. This component was especially important for the dietary modification clinical trial, where it was used for screening purposes and to evaluate the effectiveness of the low-fat dietary intervention. The time reference for all questions in the FFQ was “in the last 3 months”.

Questionnaires were collected at baseline for all subjects and then at specified follow up visits on a rotating basis for a cross-sectional subsample of the cohort. All DM trial participants completed an FFQ at baseline, before randomization, and at year 1. The FFQ was self- administered, with instructions limited to directions and examples printed on the questionnaire with an additional page of portion size pictures. The WHI dataset includes 32 additional variables for food group measures derived from the USDA MyPyramid Equivalents Database 2.0

(MPEDs),306 which have been used extensively for dietary analysis in this cohort through the creation of a priori diet pattern scores.

A validation study by Patterson, et al.307 was conducted to assess the measurement characteristics of the WHI FFQ compared to a short-term dietary and recording methods. This was done by evaluating the agreement of intakes of 30 nutrients estimated using the WHI FFQ with corresponding estimates from a 4-day food record and four 24-hour recalls. For this analysis, the study Clinical Centers recruited 169 women screened for participation in the WHI in 1995, 125 of whom completed the full dietary assessment. After excluding women with

82 implausible estimated energy intakes (< 600 or > 3,500), the final study sample included data from 113 women. Women participating in the study were mailed a second FFQ after their first clinic visit. Over a 6 to 10-week period between the first and second clinic visits, the women completed four 24-hour recalls. These were unannounced and obtained over the phone by trained

WHI staff. After the second clinic visit, participants were asked to keep a 4-day food record on alternating days to include both weekday and weekend days. The WHI FFQ nutrient database was derived from the University of Minnesota Nutrition Coding Center nutrient database

(Nutrition Coordinating Center, Minneapolis, MN).

Overall, nutrient estimates from the FFQ were similar but slightly lower than those from the 24-hour recalls and 4-day food records. The mean nutrient intakes estimated by the FFQ were within 10% for 22 out of 30 nutrients estimated from the dietary recall and 21 of 30 for the food record. The authors reported a high test-retest reliability, although the data was not shown. The intraclass correlation coefficients ranged from 0.67 for retinol, vitamin C, and vitamin B12 to

0.82 for fiber, 0.84 for magnesium and 0.92 for alcohol, with a mean correlation coefficient of

0.76.

Cardiovascular Disease Assessment

Cardiovascular disease outcomes were ascertained in the WHI through medical update questionnaires completed by participants every 6-12 months and digital electrocardiograms

(ECGs) acquired every 3 years (for CT participants only) and analyzed by a core laboratory. As discussed above, self-reported cardiovascular diagnoses were reviewed and adjudicated by central physician adjudicators or trained local adjudicators, all blinded to treatment assignment.

Cases were chosen for central adjudication by the WHI outcomes database or by a request for clarification of an unclear case from a local adjudicator. Centrally adjudicated outcomes included

83 any hospitalized MI, angina, CHF and non-hospitalized coronary revascularization, or coronary death.

The specific cardiovascular diagnoses monitored during the WHI included myocardial infarction (MI) (acute or silent, fatal and nonfatal), sudden unexplained death, rapid unexplained death, other cardiovascular death, congestive heart failure (CHF) (fatal and nonfatal) requiring hospitalization, angina pectoris requiring hospitalization, coronary artery bypass surgery

(CABG), other coronary arterial surgery or procedure, including angioplasty (PTCA), coronary stent placement or atherectomy, stroke requiring hospitalization (fatal and nonfatal, hemorrhagic ischemic, or unknown), transient ischemic attack (TIA) requiring hospitalization, carotid artery disease requiring hospitalization, peripheral artery disease requiring hospitalization, and other cardiovascular events associated with hospitalization. The criteria for coronary heart disease

(CHD), as defined in WHI included MI, coronary death, hospitalized CHF, hospitalized angina pectoris, and coronary revascularization. CHD diagnoses were based on a reported hospitalization with appropriate documentation for CHD, a report of a death possibly due to heart disease with appropriate documentation or (for CT participants only) changes in ECGs at the clinic visits at baseline and/or 3, 6, and 9 and at close-out.

All hospitalized cardiovascular events required a discharge summary, hospital face sheet with ICD-9-CM codes, and/or physician attestation sheet with ICD-9-CN codes for adjudication.

Other documents, reports, and clinical data were also requested to confirm specific events, such as MI, CHF, stroke, and coronary revascularization. All fatal events required a death certificate for adjudication. In addition to a death certificate, in-hospital deaths also required a hospital discharge/death summary and autopsy report, if available. Out of hospital deaths required a coroner or medical examiner’s report and autopsy report, if available.308

84 CHAPTER 5 Variations on Dietary Patterns and Risk of Incident Type 2 Diabetes Mellitus in Young Adults: the CARDIA Study

Background Diet is a well-established risk factor for type 2 diabetes (T2D), yet evidence for the role of dietary patterns reflecting contemporary recommendations and trends and T2D risk is limited.

Objective To examine the association between dietary pattern scores created to reflect the 2015

Dietary Guidelines for Americans (DGA) Scientific Report, a modern-day Paleolithic diet, low carbohydrate diet, and diet high in overall consumption of empty calories (foods high in calories and low in nutrients) as well as the original CARDIA A Priori Diet Quality score and T2D risk over time.

Design Prospective analysis of 4,627 young adult black and white men and women from the

Coronary Artery Risk Development in Young Adults (CARDIA) study with repeated dietary histories. Using multivariable Cox proportional hazards regression models, I examined the relationship between cumulative average dietary pattern scores and incident T2D.

Results During the 30-year follow-up period, 627 (13.6%) incident cases of T2D occurred.

There was no association between any of the contemporary dietary pattern scores (2015 DGA,

Paleolithic, low carbohydrate or empty calorie index) and T2D risk. Participants in the highest quartile of A Priori Diet Quality score had a 46% lower risk of T2D compared with those in the lowest quartile in the fully adjusted multivariable model (HR 0.54, 95% CI 0.40-0.74; p trend

<0.0001).

Conclusions Dietary pattern scores reflecting contemporary recommendations and popular patterns were not associated with T2D risk while the A Priori Diet Quality score, which largely

85 aligns with the 2015 DGA, was strongly inversely associated with lower 30-year T2D risk in the

CARDIA cohort.

Introduction

Type 2 diabetes (T2D) is reaching epidemic levels, resulting in major economic and public health consequences.43 If current trends continue, it is projected that the prevalence of

T2D in the United States will increase from approximately 1 in 8 to 1 in 3 adults by the year

2050.41 Dietary intake is a major, modifiable risk factor for T2D and the current body of evidence suggests dietary patterns with higher intakes of fruits, vegetables, whole grains, legumes, nuts and seeds, modest amounts of some animal-based foods and low amounts of refined grains, sugar-sweetened beverages, and processed meats leads to lower T2D risk.149,309

An important yet lacking component of this evidence base is an examination of the role of the most recent findings of the 2015 Dietary Guidelines Advisory Committee as a dietary pattern score in relation to T2D risk, as these recommendations reflect the amalgamation of the evidence base summarized above and constitute the basis for dietary related recommendations and policy in the U.S. Additionally, there is limited evidence on the relationship between other dietary patterns that have been increasingly adopted by Americans and T2D risk, such as a modern-day Paleolithic dietary pattern, low carbohydrate diet, and the role of overall intake of empty calories (energy-dense foods and beverages with minimal essential nutrient contributions).

To address this gap, I examined the association of dietary pattern scores constructed to reflect the 2015 Dietary Guidelines for Americans Scientific Report recommendations, a modern-day Paleolithic dietary pattern, low carbohydrate dietary pattern, and overall consumption of foods high in empty calories with risk of T2D in a cohort of young black and white men and women with repeated measures of diet over 30 years. For context, the relationship

86 between the A Priori CARDIA Diet Quality Score and T2D risk was also examined, as this score largely aligns with the 2015 DGA and has been shown to inversely associate with cardiovascular risk.292,310–312

Methods

Study Population

The Coronary Artery Risk Development in Young Adults (CARDIA) study is a prospective, multi-center cohort study designed to investigate the development and determinants of cardiovascular disease and its associated risk factors in young adults. Briefly, 5,115 black and white men and women, age 18-30 years were recruited between 1985-1986 from four U.S. cities:

Birmingham, Alabama; Chicago, Illinois; Minneapolis, Minnesota, and Oakland, California.

Participant enrollment targeted balance among age, race, sex and educational attainment. The initial examination included standardized measures of known cardiovascular risk factors as well as psychosocial, dietary, and exercise-related characteristics. Reexamination occurred 2, 5, 7, 10,

15, 20, 25, and 30 years after baseline, with retention of 91%, 86%, 81%, 79%, 74%, 72% and

72% of the surviving cohort, respectively. The CARDIA study was approved by the institutional review board at each clinical center, and informed consent was obtained from all participants prior to enrollment.289

Participants who had a diagnosis of diabetes at baseline (n = 34), were missing baseline diabetes status (n = 82) or baseline dietary data (n = 4) were excluded from this analysis.

Individuals who only participated in the baseline visit and thus had no follow-up data were also excluded (n = 153). In addition, individuals who reported extreme energy intakes (<600 kcal/day or >6,000 kcal/day for women (n = 99) and <800 kcal/day or >8,000 kcal/day for men (n = 116)) were also excluded. The rate of missing data for other pertinent covariates was low (<1%) and

87 therefore missing values were imputed by sex and race subgroup to the median value for continuous variables (BMI, alcohol intake, physical activity), and the most frequent categorical value for multichotomous variables (smoking status). This approach was validated by also performing a complete case analysis, and results were similar. The final study sample for this analysis included 4,627 black and white young adult men and women.

Dietary Intake Assessment

Dietary intake was assessed at study years 0, 7, and 20 by participant self-report to the interviewer-administered validated CARDIA Diet History, as previously described.291,297 Briefly, interviewers asked study participants open-ended questions about dietary consumption in the past month within 100 food categories, referencing 1,609 separate food items in years 0 and 7 and several thousand separate food items in year 20. Follow-up questions for selected foods concerned serving size, frequency of consumption, and common additions to these foods.

Provision was made for reporting foods not found in the food frequency list. Diet history data was then coded by the University of Minnesota Nutrition Coordinating Center (NCC) and foods were placed into 166 food groups using the food grouping system developed by the NCC. Food group intake was calculated as the total number of servings per day. 291

For this analysis, the creation of dietary pattern scores was modeled after the A Priori

Diet Quality Score used in previous CARDIA studies.292,293 From the 166 food groups created by the NCC system, 44 condensed food groups were created based on similar nutrient characteristics and comparability to food groups defined in previous studies.292–296 The study population was then ranked into quintiles of intake for each of the 44 food groups and dietary pattern scores were created by classifying each food group as beneficial (+), adverse (-), moderate (+/-) or neutral (0) in terms of the recommendations of the specific diet, as presented in

88 Supplemental Table 1 for reference. Food groups with a large percentage of non-consumers were divided into 5 levels of intake by grouping subjects who did not report any intake and ranking the remaining subjects into quartiles to allow for consistent scoring. The individual dietary pattern scores were the sum of food group scores 0 to 4 for intake of items considered beneficial by the specific dietary pattern (with the highest quintile receiving a score of 4 and the lowest a score of 0) plus scores in reverse order (4 to 0) for adverse foods. Food groups recommended to be consumed in moderation were scored 0-2-4-2-0, with moderate intake receiving the highest score. Neutral foods were not included in the final score. Higher scores reflect higher dietary alignment with the pre-specified patterns.

The original A Priori Diet Quality Score that has historically been utilized in CARDIA dietary studies was defined by previous investigators by classifying foods in terms of hypothesized health effects and disease relations to characterize higher diet quality. As shown in

Supplemental Table 2, the score contained 20 beneficial, 13 adverse and 13 neutral components, with a theoretical maximum score of 132.292

The DGA are published every 5 years following an extensive review of the most current scientific evidence to inform policy and the development of recommendations for individuals in the U.S. on healthy dietary practices. The 2015 DGA Scientific Report concluded that characteristics of dietary patterns associated with positive health outcomes are higher intake of vegetables, fruits, whole grains, low- or non-fat dairy, seafood, legumes, and nuts; moderate intake of alcohol (among adults); lower intake of red and processed meat, and low intake of sugar-sweetened foods and drinks, and refined grains.313 Foods and food groups not explicitly mentioned in the Advisory Committee’s consensus statement were not included in the score.

89 Thus, the 2015 DGA dietary pattern score contained 14 beneficial, 12 adverse, 1 moderate and

17 neutral categories, with a maximum theoretical score of 108.

The evolution of the modern-day Paleolithic dietary pattern was based on foods assumed to have been available to humans prior to the establishment of agriculture, mainly wild-animal and uncultivated-plant sources of foods, including meats, fish, vegetables, roots, eggs and nuts and excluding grains, legumes, dairy, salt, refined sugar and processed oils.314 The Paleolithic score was modeled after indices used in previous studies315,316 with minor adjustments, and contained 14 beneficial, 21 adverse, 2 moderate and 7 neutral categories, with a maximum theoretical score of 148.

Observational studies on low carbohydrate dietary patterns have commonly used diet scores based on percentages of energy intake from carbohydrate, protein, and fat calculated from food frequency questionnaires (FFQs). To utilize a whole-food based approach to study this dietary pattern, the score was created based on guidelines used in previous low carbohydrate intervention trials.317,318 In general, these guidelines targeted low carbohydrate intake by providing lists of foods to consume (meats, fatty fish and seafood, eggs, low-carbohydrate vegetables, full-fat dairy, nuts and seeds) and those to avoid (sugar, grains, high-carbohydrate vegetables, fruits, and legumes) to reduce dietary carbohydrate intake to 20-50 grams/day. 319

The low carbohydrate (LC) diet score contained 17 beneficial, 18 adverse and 9 neutral categories, with a maximum theoretical score of 140.

According to the 2015 DGA Scientific Report, empty calorie foods are those high in calories from added sugars, solid fats, and refined starch that contribute few or no nutrients to the diet. Such foods have been identified as the leading contributors to excess calorie intake in the

U.S.313 To evaluate the relationship between the consumption of empty calorie (EC) foods,

90 whose intake is discouraged by the dietary patterns described above, and T2D risk, a score was calculated based on intake from 9 food groups considered to be calorie-dense and nutrient-poor, with a maximum theoretical score of 36. Intake from these food groups was measured as servings per 1,000 calories. For comparison purposes, a score was also calculated using unadjusted intake levels from the 9 food groups.

To capture the relationship between long term dietary intake and T2D risk over time, I calculated a cumulative average dietary pattern score for each participant using dietary data from years 0, 7 and 20. For individuals who had not developed diabetes through Y7, the average of dietary pattern scores from Y0 and Y7 was used to examine T2D risk. For individuals with repeated measures of diet who had no documented diabetes through Y20, the average of dietary pattern scores at Y0, Y7, and Y20 was used. Cumulative average scores for participants who did not develop diabetes during follow-up included all available dietary data prior to censoring.

Cumulative averages were calculated based on available data, individuals without repeated measures of diet were assigned their baseline dietary pattern score.

Incident Type 2 Diabetes Mellitus

Diabetes status was assessed clinically at examination years 0, 7, 10, 15, 20, 25 and 30.

All blood samples were drawn and processed according to standard procedures and serum glucose was assayed using the hexokinase method at a central laboratory.289 Diabetes was defined as measured fasting glucose  7 mmol/l (126 mg/dl), 2-hour oral glucose tolerance test 

11.1 mmol/l (available years 10, 20 and 25), hemoglobin A1c  6.5% (available year 25), or self- report of oral hypoglycemic medication(s) or insulin use. CARDIA did not differentiate between type 1 and type 2 diabetes mellitus; however, it is likely that most incident cases identified

91 during follow-up are T2D given the age of the cohort. To avoid potential misclassification, a sensitivity analysis was performed excluding subjects who developed diabetes before age 30.

Covariates

At the CARDIA baseline and follow-up examinations, participants completed self- administered questionnaires to collect information on sociodemographic, psychosocial, and medical background. Some of these questionnaires were followed up with interviewer- administered questions to obtain more detailed information about illnesses, medication, smoking habits, alcohol use, and life events.289 Physical activity was assessed using the CARDIA physical activity questionnaire, a validated interviewer-based self-report of duration and intensity of participation in 13 categories of exercise over the past year.320 Physical activity was reported in exercise units (EU), where 300 EU is approximately equal to 150 minutes of moderate-intensity physical activity per week or 30 minutes of moderate-intensity activity 5 days/week.321 Total energy intake (kilocalories) was calculated from the CARDIA Diet History. An alternate

Mediterranean (aMed) diet scores was calculated, as previously described,141 to account for other dietary components in the EC score analysis. Body weight was measured with light clothing to the nearest 0.09 kg and height was measured without shoes to the nearest 0.5 cm. Body mass index (BMI) was calculated from these measurements as weight in kilograms divided by height in meters squared.289,290

Statistical Analysis

Participant’s demographic, lifestyle, and clinical characteristics were described for the study sample by quartile of cumulative average dietary pattern score, using means with standard deviation for continuous variables and frequencies with percentages for categorical variables. To compare characteristics between quartiles, analysis of variance and chi-squared tests were

92 performed for continuous variables and categorical variables, respectively. Survival analysis using multivariable Cox proportional hazards models was used to estimate the hazard ratios

(HRs) and corresponding 95% confidence intervals (95% CI) for incident diabetes during follow- up. Separate models were fit for the 2015 DGA, Paleolithic, LC, EC and CARDIA A Priori dietary pattern scores. Once a participant developed documented diabetes at a CARDIA follow- up examination, they were considered to have an event and were subsequently censored. Follow- up time was calculated as the time (years) from baseline to the examination visit where diabetes was identified or until the last examination (due to death, loss to follow-up or end of cohort surveillance), whichever came first.

For the analyses, participants were ranked into quartiles of cumulative average dietary pattern score, with the lowest quartile for each score serving as the reference group.

Multivariable models adjusted for preselected potential demographic and lifestyle confounders were utilized. The base model adjusted for age, race, sex, and CARDIA center. Model 2 included Model 1 covariates plus potential sociodemographic and lifestyle confounders

(smoking, education, estimated total energy intake and physical activity). Smoking status and education were treated as time-varying covariates in the model, and the cumulative average of

Y0, Y7, and Y20 data was used to account for repeated measures of energy intake and physical activity. A third model was adjusted for Model 2 covariates plus cumulative average BMI from years 0, 7, and 20 (or until censored) as a potential mediator. Since family history of diabetes was missing for 676 participants (15%), the analysis was repeated using both models in the subset of the population who had information on family history of diabetes (n = 3,951). Prior to adjustment for BMI, EC models were additionally adjusted for cumulative average aMed diet score to account for overall diet quality as the scoring does not include foods in the EC index.

93 Dietary patterns scores were also examined individually as continuous variables to test for linear trend, with the HRs calculated per standard deviation of the score.

Potential effect measure modification by sex, race, BMI, smoking, and family history was tested by performing analyses stratified on these covariates and including an interaction term for the variable of interest and each dietary pattern score separately. As a secondary analysis to the

EC score, the relationship between quartiles of percent of total calories from added sugar and saturated fat intake from the Y20 dietary assessment in a subgroup of individuals with available dietary data who were free of T2D at Y20 (n = 2,489) was also examined. To avoid potential misclassification, a sensitivity analysis was performed excluding subjects who develop diabetes before age 30 (n=8). The proportional hazards assumption was tested by including an interaction term with log (base-e) transformed time for each covariate. There was no evidence that the assumption was violated in any of the models. All analyses were performed using SAS 9.4 (SAS

Institute, Inc.).

Results

Participant characteristics by quartiles of the dietary pattern scores are presented in Table

1. Participants with higher cumulative average 2015 DGA scores were older, more educated and physically active, consumed less alcohol, had lower mean BMI and blood pressure, higher HDL- cholesterol, were less likely to smoke and have a family history of diabetes than individuals with lower scores; and a greater proportion were female and white. Individuals with higher Paleolithic scores were older, more educated, consumed less alcohol, had lower blood pressure, higher

HDL-cholesterol and were less likely to smoke than those with lower scores. Like the 2015 DGA score, a greater proportion of these individuals were also female and white. Individuals with higher LC scores were more educated and physically active, consumed more alcohol, and had

94 higher HDL-cholesterol than individuals with lower scores. A greater proportion of these individuals were female and current or former smokers. Individuals with higher EC scores were younger, less physically active, consumed less alcohol, had lower HDL than those with lower EC scores, and a greater proportion were black and male. Individuals with higher A Priori Diet

Quality scores had similar characteristics as those in the upper quarterlies of the 2015 DGA score, except that they consumed more alcohol at baseline than those in the lower quartiles. Y0,

Y7, and Y20 cumulative average BMI increased, while average physical activity decreased across all quartiles of the 4 dietary pattern scores. Trends for these variables remained consistent between quartiles for each dietary pattern.

The 2015 DGA and A Priori Diet Quality scores as well as Paleolithic and LC scores were strongly positively correlated, as shown in Table 2. The 2015 DGA and A Priori scores were also positively correlated with the Paleolithic and LC scores. The EC score was negatively correlated with all other dietary pattern scores under study. The descriptive dietary characteristics by quartile of dietary pattern score are presented in Supplemental Tables 3-6.

A total of 627 incident cases of diabetes occurred during the 30-year follow-up period

(mean (SD) 24.3 (8.6) years). The results for the cumulative average 2015 DGA dietary pattern score and risk for T2D are presented in Table 3. In model 1, participants in the highest quartile of 2015 DGA dietary pattern score had a 38% lower risk of developing T2D compared with those in the lowest quartile (HR 0.62, 95% CI 0.47-0.82; p trend 0.001). However, additional adjustment for education, smoking, physical activity and energy intake attenuated this association (Model 2 HR 0.86, 95% CI 0.64-1.16). Education levels, specifically, were strongly associated with T2D and largely responsible for the attenuation of the association observed in model 2; and there was evidence that BMI may act as a mediator of this association over time.

95 The Paleolithic score displayed a suggestive positive association with T2D risk when comparing the highest quartile with lowest quartile in the model adjusting for basic demographic characteristics, education, smoking status, physical activity and total energy intake, but the 95%

CI limits conclusive inference (Model 2 HR 1.28, 95% CI 0.98-1.66). Similarly, the LC score was suggestively positively associated with T2D risk in the upper quartiles compared with the lowest quartile, independent of education, smoking, physical activity, and energy intake (HRQ3

1.25, 95% CI 1.00-1.56; HRQ4 1.21, 95% CI 0.95-1.53).

No association was found between the cumulative average EC dietary pattern score and

T2D risk in this population. This finding was consistent using both the EC score calculated per

1,000 calories of total energy intake and unadjusted total servings of foods high in empty calories

(data not presented). In a secondary analysis, higher percentage of total calories from added sugar and saturated fat intake at Y20 was associated with an increased risk for T2D in the upper quartiles compared with the lowest (HRQ3 1.58, 95% CI 1.02-2.45; HRQ4 1.57, 95% CI 1.01-

2.43; p trend 0.004), but the association was attenuated by further adjustment for lifestyle factors, diet quality and BMI (Supplemental Table 7). In a sensitivity analysis of a subgroup of the study population with family history of diabetes data, inclusion of this variable in the multivariable models did not materially alter the estimates, although they were less precise (data not presented).

The CARDIA A Priori Diet Quality score was strongly inversely associated with T2D risk. Individuals in the highest quartile of cumulative average A Priori Diet Quality score had a

46% lower risk of T2D during the follow-up period compared with those in the lowest quartile in the fully adjusted multivariable model (HR 0.54, 95% CI 0.40-0.74; p trend <0.0001). When

96 examining the association by standard deviation of the score in model 3 there was a 23% lower risk of T2D over time (HR 0.77, 95% CI 0.69-0.87).

Sensitivity analyses provided evidence for potential effect measure modification by smoking status in the 2015 DGA (p interaction = 0.007) and A Priori Diet Quality (p interaction

= 0.004) analyses. When the 2015 DGA analysis was stratified on smoking status (current vs non-smokers) with HRs calculated per standard deviation of diet score, a modest but imprecise inverse association was observed between the 2015 DGA score and T2D risk in non-smokers

(n=3,494) (HR 0.89, 95% CI 0.78-1.02; p trend = 0.09) and no association was seen in current smokers (n=1133) (HR 1.09, 95% CI 0.89-1.34; p trend = 0.40). For the A Priori score, stratified analysis demonstrated a strong inverse association in non-smokers consistent with the main study findings (HR 0.73, 95% CI 0.64-0.83) and no association in current smokers (HR 0.90, 95% CI

0.74-1.11). Sample size limitations prevented us from investigating these stratified group associations by quartile of dietary pattern score, which would provide the basis to make stronger conclusions related to how smoking may modify the diet-T2D association. There was no evidence that the association between cumulative average Paleolithic, LC, or EC dietary pattern scores and T2D risk differed by smoking status. Stratified analyses as well as formal tests for interaction between dietary pattern scores and race, sex, cumulative average BMI, and family history of diabetes provided no evidence of effect measure modification by these factors. In addition, results of a sensitivity analysis to address potential misclassification, by excluding incident T2D cases documented before age 30 (n=8), were not materially different from the results of the main analysis.

97 Discussion

The results of this study demonstrate that both contemporary dietary recommendations

(2015 DGA) and popular adapted dietary patterns (Paleolithic, LC, and EC) were not associated with risk for T2D; although the direction of the point estimates for the 2015 DGA score were inverse with T2D risk and the direction of the point estimates for the Paleolithic and LC scores were positive with T2D risk. The A Priori Diet Quality score, which was highly correlated with and scored similarly to the 2015 DGA score, was strongly inversely associated with risk for

T2D, independent of lifestyle factors and BMI.

Formal tests for interaction and stratification suggested that the relationship between the

2015 DGA and A Priori Diet Quality scores and T2D risk may differ by smoking status. When the fully-adjusted, multivariable analysis was stratified on smoking status, a suggestive inverse association was seen between the 2015 DGA score and T2D risk in non-smokers, while no association was detected in current smokers. The A Priori analysis provided evidence for a strong inverse association in non-smokers but no association between diet quality and T2D in current-smokers. Smoking has been identified as a strong, independent risk factor for diabetes322 and previous studies of diet and diabetes risk suggest that the protective association of a healthy dietary pattern may be blunted by smoking status,323–325 which is consistent with the findings.

This is the first prospective observational study to examine the relationship between a modern-day Paleolithic dietary pattern and incident T2D. Previously, several small intervention trials have demonstrated that short-term consumption of a Paleolithic diet improves cardiometabolic risk factors, such as blood pressure, plasma insulin during an OGTT, serum lipids, body weight and weight circumference.326–328 However, total energy intake decreased substantially in the Paleolithic groups during several of the interventions, potentially

98 confounding the observed results. Proponents of the Paleolithic diet assert that humans are genetically adapted to foods assumed to have been available prior to changes in the food supply resulting from the establishment of agriculture, and that the incompatibility of core metabolic and physiological processes with these dietary changes may lead to chronic diseases such as obesity, T2D and CVD.314,329 However, I found no association between consumption of a modern-day Paleolithic informed dietary pattern over time and T2D risk in the analysis. Whole grains, legumes, and dairy products, all known to have a protective relationship with T2D risk,330–332 are notably absent from a Paleolithic dietary pattern, which may also explain the lack of an association. More research is needed to clarify the relationship between a modern-day

Paleolithic dietary pattern and chronic disease risk, as well as the sustainability and nutritional adequacy of long-term consumption of a Paleolithic-type diet, in the context of the modern food supply.

No association was found between a low carbohydrate dietary pattern score and T2D risk over time. Although some evidence from randomized controlled trials has shown an inverse relationship between low carbohydrate diets and T2D risk factors,317,318 these studies targeted and achieved significantly lower carbohydrate intake levels than were observed in this study population or in previous observational studies,333–335 suggesting that an extreme low carbohydrate dietary pattern is essentially not consumed by free-living populations; and previous reports and summaries largely suggest that this type of dietary pattern is not a sustainable behavior over time.336,337 The lack of an association observed between a low carbohydrate dietary pattern and T2D risk in the CARDIA cohort are consistent with the results of the study by Halton, et al. that examined the association between a low carbohydrate diet score and risk of

T2D in the Nurse’s Health Study (NHS) cohort. Based on results of their fully adjusted

99 multivariable model, the authors found that diets lower in carbohydrate and higher in fat and protein were not associated with T2D risk when comparing women in the tenth decile of low- carbohydrate score with women in the first decile (RR 0.90, 95% CI 0.78-1.04).333 On the other hand, de Koning, et al. found that a diet score representing a low-carbohydrate diet high in total fat and protein was positively associated with T2D risk in the Health Professionals Follow-Up

Study (HPFS) when comparing those in the highest quintile of the score with those in the lowest

(HR 1.31, 95% CI 1.14-1.49).334 In both the NHS and HPFS, additional analyses suggested a potential protective role of low-carbohydrate diets high in plant-based protein and fat. A notable difference between my approach to study a low carbohydrate dietary pattern compared with the approach used in these studies is the use of calculated macronutrient intakes as opposed to whole foods in creating the dietary pattern score. This is an important consideration since low carbohydrate diets are often characterized by an increase in animal protein, total and saturated fat intake and a decrease in whole grains, cereal fiber, and fruit intake, which contradict the evidence current recommendations of the 2015 DGA and American Diabetes Association for

T2D prevention are based upon.102

The nature of the results for the 2015 DGA score suggest that the diet-diabetes relationship is mediated in part by BMI, but in this population the diet-diabetes relationship was largely explained by educational attainment over time. To my knowledge, this is the first prospective observational study to examine the relationship between a dietary pattern score based on the findings of the 2015 DGA Scientific Report and incident T2D. It should be noted that several potentially important and highly consumed foods and food groups, such as poultry, fried foods, coffee and tea, eggs, oil and potatoes were not included in the score since the overall summary statement of the report does not make specific recommendations about these foods.

100 However, the report does state that strong evidence shows that it is not necessary to eliminate food groups or conform to a single dietary pattern to achieve a healthy diet.313 Previous observational studies examining dietary pattern scores created to reflect earlier versions of the

DGA have also produced inconclusive results with regard to adherence to the federal dietary guidelines and T2D risk. For instance, Zamora, et al. found no association between the 2005 Diet

Quality Index, created to reflect the key dietary recommendations of the 2005 DGA, and 20-year

T2D risk in CARDIA, although the score was associated with favorable changes in HDL and blood pressure.338 Several prospective studies have also produced null findings when examining the relationship between Healthy Eating Index (HEI) scores, reflecting both the 2005 and 2010

DGAs, and T2D risk.154,155 In addition, the Alternate Healthy Eating Index (AHEI) for 2005 and

2010, which reflect the DGA and also include foods and nutrients indicated by scientific literature to be predictive of chronic diseases,137 have been associated with a decreased risk for

T2D in some study populations,138,154,155 but not others.339 The development of a dietary pattern score to reflect the most current evidence-based nutrition recommendations is needed to serve as a standard by which to compare associations between other dietary patterns and health outcomes and to further evaluate the effectiveness of the DGA in other populations.

The null findings from the EC score and strong inverse association observed between the

CARDIA A Priori score and T2D risk highlight the importance of overall diet quality in relation to T2D risk. The EC score, in which a higher score represented intake of foods high in calories from added fats, sugars and refined grains, was not associated with T2D risk in this population.

Although one can infer that high intake of energy-dense, nutrient poor foods negatively scored by the other dietary patterns studied are not beneficial to overall health, these results suggest there is more to the diet-disease relationship than simply avoiding empty calories, and that both

101 foods included and excluded from a dietary pattern are important in relation to T2D risk. This is exemplified by the results of the A Priori score analysis, which showed a reduced risk of T2D with higher intake of fruits, vegetables, low-fat dairy, fish, poultry, legumes, nuts and seeds, whole grains, alcohol, coffee and tea and lower intake of red and processed meat, fried foods, sugar-sweetened foods and beverages, and whole-fat dairy. Findings from previous CARDIA studies have shown an inverse association between the A Priori score and markers of oxidative stress292 and endothelial dysfunction311 and positive association with cardiorespiratory fitness.310

Although the 2015 DGA and A Priori scores were highly correlated, differences between these scoring indices are important to consider as imprecision surroundings these components due to the limitations of self-reported dietary data may have impacted the strength of the association detectable in this cohort. For instance, alcohol was scored moderately in the 2015 DGA score and positively in the A Priori score. Foods scored positively in the A Priori score, yet not included in the 2015 DGA score, included oil, poultry, coffee and tea. In addition, the A Priori score negatively scored butter, fried foods, and whole fat dairy, which were not included in the

2015 DGA score; while fruit juice, margarine and refined grains were negatively scored in the

2015 DGA score but not included in the A Priori score. The similarity of the scores, however, including positive scoring of fruits, vegetables, avocado, nuts and seeds, legumes, fish, low-fat dairy and whole grains and negative scoring of red and processed meats, sugar-sweetened foods and beverages, fried potatoes and salty snacks suggest that a dietary pattern reflective of the 2015

DGA Scientific Report is potentially protective against long-term T2D risk. Further research is needed to confirm this observation, especially in high risk individuals such as current smokers.

The interpretation of these findings is strengthened by the design of the CARDIA study, with repeated measures of diet and other important covariates and clinically ascertained T2D

102 outcomes. The use of repeated measures of dietary intake reduces random measurement error and provides a robust assessment of long term dietary intake and T2D risk. The age of the

CARDIA cohort also offered a unique perspective on the diet-diabetes relationship in young adulthood. Furthermore, the use of dietary pattern scores to study diet-disease relationships may be more effectively translated into population-based recommendations.131,2,133

There are also several limitations to the study worth noting. Although I accounted for many confounders in the models, residual confounding likely occurs in all diet-disease observational studies. The use of self-reported dietary data is also subject to recall and other biases that may alter estimates. The validity and reliability of the CARDIA Diet History has been demonstrated, but nutrient and energy estimates were found to have larger variability among blacks than whites.297,340 While there was adequate power to detect an association between dietary intake and T2D, a larger sample size could improve the precision of the estimates given these and other limitations of observational diet-disease research.132

In conclusion, contemporary dietary recommendations (2015 DGA) and popular adapted dietary patterns (Paleolithic, LC, and EC) were not associated with risk for T2D. However, the

CARDIA A Priori Diet Quality score, which was highly correlated with the 2015 DGAs and similarly scored, was strongly associated with a lower risk of T2D in this population. This finding suggests that the recommendations of the 2015 DGA Scientific Report may generally characterize a dietary pattern that is protective against T2D, but detection of this association may have been limited in this study due to the scoring of the metric and the size of the cohort. Future research should examine these contemporary dietary pattern scores with diabetes risk in larger cohorts to inform public health nutrition recommendations and the role of popular contemporary diet trends in the epidemic of T2D in the United States and globally.

103 Tables

TABLE 5.1a Baseline characteristics of CARDIA participants according to Y0, Y7, and Y20 cumulative average 2015 Dietary Guidelines for Americans scientific report diet score quartile. Quartile of DGA 2015 Score Characteristic a Q1 Q2 Q3 Q4 P d N 1164 1137 1166 1160 DGA 2015 Score* 38.4 (4.0) 47.0 (2.0) 54.3 (2.3) 65.1 (5.2) <0.0001 Age (years) 23.6 (3.7) 24.6 (3.7) 25.3 (3.5) 25.9 (3.1) <0.0001 Race (% white) 24.4 40.8 58.0 79.6 <0.0001 Sex (% male) 74.7 59.1 43.1 20.0 <0.0001 Education (years) 13.1 (4.7) 13.6 (4.1) 14.2 (3.3) 15.1 (3.3) <0.0001 Smoking Status (%) Never 55.7 56.6 60.1 58.6 Former 6.3 10.7 15.7 21.5 <0.0001 Current 38.1 32.6 24.2 19.9 Alcohol Intake (ml/day) 13.6 (23.1) 11.8 (21.1) 10.1 (15.4) 10.2 (15.3) <0.0001 Physical Activity 383.1 (292.5) 374.2 (280.9) 424.8 (293.9) 472.2 (283.7) <0.0001 (EU/week)b Cumulative Average 340.4 (242.3) 336.2 (231.4) 383.1 (239.2) 426.9 (228.1) <0.0001 Physical Activity Body Mass Index (kg/m2) 24.4 (5.3) 25.1 (5.4) 24.8 (5.0) 23.5 (4.0) <0.0001 Cumulative Average BMI 26.4 (5.7) 27.1 (5.7) 26.7 (5.5) 25.0 (4.7) <0.0001 Family History DM (%)c 18.2 19.2 15.7 12.1 <0.0001 Fasting Glucose (mg/dl) 81.6 (11.4) 81.5 (8.3) 82.0 (8.4) 82.0 (7.4) 0.38 Insulin (µU/ml) 11.5 (8.2) 11.9 (9.4) 10.6 (7.3) 9.3 (6.2) <0.0001 Blood Pressure (mmHg) Systolic 111.7 (10.7) 111.2 (10.8) 110.3 (10.9) 108.1 (10.7) <0.0001 Diastolic 68.7 (10.0) 68.8 (9.8) 68.8 (9.6) 67.9 (8.8) 0.07 Triglycerides (mg/dl) 74.4 (47.6) 74.6 (50.5) 72.7 (40.9) 69.7 (53.1) 0.05 Cholesterol (mg/dl) Total Cholesterol 174.6 (34.5) 177.3 (33.3) 179.6 (33.8) 176.4 (31.8) 0.003 LDL-Cholesterol 108.1 (32.3) 110.4 (30.9) 112.0 (31.9) 107.0 (29.3) 0.0005 HDL-Cholesterol 51.6 (13.3) 52.0 (12.6) 53.1 (12.8) 55.5 (13.3) <0.0001 Abbreviations: BMI, body mass index; BP, blood pressure; CARDIA, Coronary Artery Risk Development in Young Adults; DM, diabetes mellitus; HDL, high-density lipoprotein; LDL, low-density lipoprotein a Unadjusted mean (SD) for all characteristics unless noted as percentage. b EU = Exercise units; physical activity score derived from the CARDIA physical activity history, where 300 EU is approximately equal to 150 minutes of moderate-intensity physical activity per week. c Family history data unavailable for 627 subjects. d P-value is analysis of variance (age, education, physical activity, alcohol, BMI, glucose, BP, lipids) or Chi-square t-test (race, gender, smoking, family history) of association between dietary pattern quartile and characteristic. *Mean (SD) 2015 DGA score for the study population was 51.2 (10.5) with a range of 20-86.5.

104 TABLE 5.1b Baseline characteristics of CARDIA participants according to Y0, Y7, and Y20 cumulative average Paleolithic diet score quartile. Quartile of Paleolithic Score Characteristic a Q1 Q2 Q3 Q4 P d N 1159 1190 1124 1154 Paleolithic Score* 62.6 (4.0) 70.4 (1.7) 75.8 (1.5) 83.4 (4.2) <0.0001 Age (years) 23.7 (3.7) 24.6 (3.6) 25.2 (3.5) 25.9 (3.2) <0.0001 Race (% white) 40.6 45.4 54.8 62.7 <0.0001 Sex (% male) 64.6 47.5 39.1 28.9 <0.0001 Education (years) 13.2 (2.0) 13.7 (2.1) 14.1 (2.2) 14.6 (2.3) <0.0001 Smoking Status (%) Never 57.8 58.7 56.8 57.7 Former 9.7 11.0 15.6 18.1 <0.0001 Current 32.5 30.3 27.7 24.2 Alcohol Intake (ml/day) 13.4 (22.3) 11.2 (18.2) 10.6 (18.5) 10.3 (16.7) 0.0003 Physical Activity 439.3 (305.3) 390.4 (279.8) 410.3 (294.7) 415.7 (279.5) 0.0008 (EU/week)b Cumulative Average 394.5 (251.2) 346.8 (229.1) 369.3 (239.2) 377.2 (230.3) <0.0001 Physical Activity Body Mass Index (kg/m2) 24.3 (4.9) 24.4 (5.1) 24.4 (4.8) 24.7 (5.1) 0.19 Cumulative Average BMI 26.2 (5.4) 26.3 (5.5) 26.3 (5.3) 26.4 (5.7) 0.80 Family History DM (%)c 17.0 14.9 16.6 16.1 0.62 Fasting Glucose (mg/dl) 81.8 (8.7) 82.0 (10.7) 81.7 (8.3) 81.6 (7.9) 0.69 Insulin (µU/ml) 10.9 (7.8) 11.2 (8.5) 10.7 (8.1) 10.4 (7.2) 0.10 Blood Pressure (mmHg) Systolic 111.7 (10.7) 110.8 (10.6) 109.7 (10.8) 109.0 (11.1) <0.0001 Diastolic 68.9 (10.0) 68.6 (9.6) 68.2 (9.4) 68.5 (9.2) 0.45 Triglycerides (mg/dl) 73.8 (46.7) 72.8 (43.6) 73.0 (49.0) 71.6 (53.3) 0.74 Cholesterol (mg/dl) Total Cholesterol 173.2 (32.8) 176.3 (33.9) 177.5 (33.5) 180.8 (33.0) <0.0001 LDL-Cholesterol 107.2 (30.9) 109.0 (31.5) 109.9 (31.1) 111.5 (31.0) 0.01 HDL-Cholesterol 51.3 (12.7) 52.7 (12.6) 53.1 (13.0) 55.1 (13.9) <0.0001 Abbreviations: BMI, body mass index; BP, blood pressure; CARDIA, Coronary Artery Risk Development in Young Adults; DM, diabetes mellitus; HDL, high-density lipoprotein; LDL, low-density lipoprotein a Unadjusted mean (SD) for all characteristics unless noted as percentage. b EU = Exercise units; physical activity score derived from the CARDIA physical activity history, where 300 EU is approximately equal to 150 minutes of moderate-intensity physical activity per week. c Family history data unavailable for 627 subjects. d P-value is analysis of variance (age, education, physical activity, alcohol, BMI, glucose, BP, lipids) or Chi-square t-test (race, gender, smoking, family history) of association between dietary pattern quartile and characteristic. *Mean (SD) Paleolithic score for the study population was 73.0 (8.2) with a range of 44-105

105 TABLE 5.1c Baseline characteristics of CARDIA participants according to Y0, Y7, and Y20 cumulative average Low Carbohydrate diet score quartile. Quartile of LC Score Characteristic a Q1 Q2 Q3 Q4 P d N 1171 1105 1208 1143 Low Carbohydrate Score* 59.4 (3.4) 65.7 (1.3) 70.1 (1.4) 77.2 (3.7) <0.0001 Age (years) 24.1 (3.7) 24.6 (3.7) 24.9 (3.6) 25.7 (3.3) <0.0001 Race (% white) 52.4 49.0 46.0 56.0 <0.0001 Sex (% male) 53.0 44.6 40.9 41.9 <0.0001 Education (years) 13.6 (2.1) 13.7 (2.2) 13.9 (2.1) 14.4 (2.3) <0.0001 Smoking Status (%) Never 62.6 60.6 58.0 49.8 Former 10.1 12.6 13.3 18.4 <0.0001 Current 27.3 26.8 28.7 31.9 Alcohol Intake (ml/day) 9.8 (17.8) 11.2 (19.4) 10.6 (18.6) 14.1 (20.4) <0.0001 Physical Activity 409.7 (290.5) 406.6 (289.0) 407.5 (286.2) 431.5 (295.5) 0.13 (EU/week)b Cumulative Average 371.5 (242.4) 360.6 (235.4) 366.9 (236.2) 388.2 (237.7) 0.04 Physical Activity Body Mass Index (kg/m2) 24.1 (4.8) 24.3 (5.0) 24.7 (4.9) 24.8 (5.1) 0.003 Cumulative Average BMI 25.9 (5.2) 26.2 (5.5) 26.7 (5.5) 26.4 (5.7) 0.002 Family History DM (%)c 15.4 15.8 17.1 16.2 0.75 Fasting Glucose (mg/dl) 81.9 (10.9) 81.5 (8.7) 81.6 (9.2) 82.1 (7.9) 0.31 Insulin (µU/ml) 10.7 (7.4) 10.5 (7.0) 11.3 (9.0) 10.7 (8.0) 0.06 Blood Pressure (mmHg) Systolic 110.6 (10.5) 109.8 (11.0) 110.4 (10.9) 110.4 (11.1) 0.41 Diastolic 68.8 (9.7) 68.3 (9.5) 68.4 (9.4) 68.7 (9.7) 0.42 Triglycerides (mg/dl) 73.7 (48.7) 72.6 (41.4) 72.2 (48.2) 72.9 (53.6) 0.88 Cholesterol (mg/dl) Total Cholesterol 172.4 (32.5) 178.2 (33.7) 178.6 (33.2) 178.7 (33.8) <0.0001 LDL-Cholesterol 106.8 (30.6) 110.8 (31.4) 110.3 (31.2) 109.8 (31.3) 0.008 HDL-Cholesterol 51.0 (12.4) 52.8 (13.1) 54.0 (12.9) 54.5 (13.7) <0.0001 Abbreviations: BMI, body mass index; BP, blood pressure; CARDIA, Coronary Artery Risk Development in Young Adults; DM, diabetes mellitus; HDL, high-density lipoprotein; LC; low carbohydrate; LDL, low-density lipoprotein a Unadjusted mean (SD) for all characteristics unless noted as percentage. b EU = Exercise units; physical activity score derived from the CARDIA physical activity history, where 300 EU is approximately equal to 150 minutes of moderate-intensity physical activity per week. c Family history data unavailable for 627 subjects. d P-value is analysis of variance (age, education, physical activity, alcohol, BMI, glucose, BP, lipids) or Chi-square t-test (race, gender, smoking, family history) of association between dietary pattern quartile and characteristic. *Mean (SD) LC score for the study population was 68.1 (7.0) with a range of 40-94.

106 TABLE 5.1d Baseline characteristics of CARDIA participants according to Y0, Y7, and Y20 cumulative average Empty Calorie score quartile. Quartile of Empty Calorie Score Characteristic a Q1 Q2 Q3 Q4 P d N 1168 1157 1157 1145 Empty Calorie score* 13.2 (2.1) 17.4 (0.8) 20.1 (0.8) 23.7 (1.8) <0.0001 Mediterranean Diet score 4.2 (1.6) 4.2 (1.5) 4.1 (1.5) 4.2 (1.4) 0.02 Age (years) 25.5 (3.4) 25.1 (3.5) 24.6 (3.7) 24.2 (3.7) <0.0001 Race (% white) 66.4 55.6 43.3 37.7 <0.0001 Sex (% male) 39.4 40.3 46.9 54.0 <0.0001 Education (years) 14.2 (2.3) 14.2 (4.9) 13.8 (4.1) 13.8 (4.1) 0.002 Smoking Status (%) Never 51.5 55.7 60.5 63.5 Former 17.2 14.4 12.1 10.5 <0.0001 Current 31.3 30.0 27.4 26.0 Alcohol Intake (ml/day) 15.0 (22.9) 11.7 (19.4) 10.1 (18.3) 8.7 (14.2) <0.0001 Physical Activity 433.3 (285.4) 407.0 (287.7) 400.6 (285.1) 414.0 (302.5) 0.04 (EU/week)b Cumulative Average 387.4 (232.6) 366.9 (229.8) 358.4 (234.4) 374.4 (254.4) 0.03 Physical Activity Body Mass Index (kg/m2) 24.3 (4.5) 24.4 (5.1) 24.7 (5.2) 24.5 (5.1) 0.23 Cumulative Average BMI 26.0 (5.2) 26.2 (5.6) 26.6 (5.8) 26.4 (5.5) 0.08 Family History DM (%)c 14.8 15.7 16.9 17.3 0.42 Fasting Glucose (mg/dl) 82.3 (8.1) 81.4 (8.2) 81.8 (10.9) 81.6 (8.5) 0.10 Insulin (µU/ml) 10.0 (6.9) 10.4 (7.9) 11.3 (8.2) 11.5 (8.4) <0.0001 Blood Pressure (mmHg) Systolic 109.7 (11.2) 110.0 (10.9) 110.8 (10.9) 110.8 (10.4) 0.04 Diastolic 68.4 (9.5) 68.7 (9.2) 68.4 (9.9) 68.8 (9.7) 0.64 Triglycerides (mg/dl) 74.2 (53.4) 71.6 (42.9) 72.4 (45.9) 73.0 (50.0) 0.64 Cholesterol (mg/dl) Total Cholesterol 177.3 (33.2) 178.2 (32.7) 176.1 (33.5) 176.2 (34.2) 0.36 LDL-Cholesterol 107.9 (20.7) 110.2 (30.9) 109.3 (31.3) 110.3 (31.7) 0.21 HDL-Cholesterol 54.8 (13.9) 53.8 (12.7) 52.2 (12.7) 51.4 (12.7) <0.0001 Abbreviations: BMI, body mass index; BP, blood pressure; CARDIA, Coronary Artery Risk Development in Young Adults; DM, diabetes mellitus; HDL, high-density lipoprotein; LDL, low-density lipoprotein a Unadjusted mean (SD) for all characteristics unless noted as percentage. b EU = Exercise units; physical activity score derived from the CARDIA physical activity history, where 300 EU is approximately equal to 150 minutes of moderate-intensity physical activity per week. c Family history data unavailable for 627 subjects. d P-value is analysis of variance (age, education, physical activity, alcohol, BMI, glucose, BP, lipids) or Chi-square t-test (race, gender, smoking, family history) of association between dietary pattern quartile and characteristic. *Mean (SD) empty calorie score for the study population was 18.6 (4.1) with a range of 4-31.

107 Table 5.2 Pearson’s correlations between dietary pattern scores. A priori DGA 2015 Paleolithic LCHF Empty Calorie A priori 1.00 0.89 0.57 0.37 -0.44 DGA 2015 0.89 1.00 0.66 0.44 -0.53 Paleolithic 0.57 0.66 1.00 0.72 -0.63 LCHF 0.37 0.44 0.72 1.00 -0.58 Empty Calorie -0.44 -0.53 -0.63 -0.58 1.00 P < 0.0001 for all correlation

108 Table 5.3 Association between Y0, Y7, and Y20 cumulative average dietary patterns and 30-year diabetes risk in young adult men and women from the CARDIA study, 1985-2015. Dietary Pattern Score Quartile

2015 DGA Score Q1 Q2 Q3 Q4 Per SD P trend n T2D/person- 203/26,721 174/27,268 147/28,770 103/29,649 years Model 1a 1.0 (ref) 0.90 (0.73-1.11) 0.78 (0.62-0.98) 0.62 (0.47-0.82) 0.82 (0.74-0.90) 0.0001 Model 2b 1.0 (ref) 1.00 (0.81-1.23) 0.98 (0.77-1.25) 0.86 (0.64-1.16) 0.94 (0.84-1.05) 0.24 Model 3c 1.0 (ref) 0.97 (0.79-1.20) 0.94 (0.74-1.20) 0.92 (0.68-1.23) 0.95 (0.85-1.06) 0.35 Paleolithic Score n T2D/person- 170/27,410 159/29,518 150/27,649 148/27,831 years Model 1a 1.0 (ref) 0.89 (0.71-1.11) 0.96 (0.76-1.20) 1.01 (0.80-1.29) 0.99(0.91-1.09) 0.90 Model 2b 1.0 (ref) 0.99 (0.79-1.24) 1.13 (0.88-1.44) 1.28 (0.98-1.66) 1.09(0.98-1.10) 0.10 Model 3c 1.0 (ref) 0.98 (0.78-1.23) 1.12 (0.88-1.43) 1.22 (0.94-1.59) 1.07(0.96-1.18) 0.22 LC Score n T2D/person- 150/28,510 142/27,718 185/29,394 150/26,786 years Model 1a 1.0 (ref) 0.96 (0.76-1.21) 1.17 (0.94-1.45) 1.12 (0.89-1.42) 1.05(0.96-1.14) 0.26

Model 2b 1.0 (ref) 0.99 (0.78-1.25) 1.25 (1.00-1.56) 1.21 (0.95-1.53) 1.08(0.99-1.18) 0.09

Model 3c 1.0 (ref) 0.94 (0.75-1.19) 1.18 (0.95-1.48) 1.12 (0.89-1.42) 1.06(0.97-1.16) 0.18

A Priori Score n diabetes / 211/24,785 181/28,389 148/28,716 87/30,518 person-years Model 1a 1.0 (ref) 0.73 (0.60-0.90) 0.63 (0.51-0.80) 0.39 (0.29-0.52) 0.68 (0.62-0.76) <0.0001

Model 2b 1.0 (ref) 0.79 (0.64-0.97) 0.76 (0.60-0.96) 0.51 (0.37-0.70) 0.76 (0.68-0.85) <0.0001

Model 3c 1.0 (ref) 0.77 (0.62-0.94) 0.76 (0.60-0.97) 0.54 (0.40-0.74) 0.77 (0.69-0.87) <0.0001 EC Score n T2D/person- 146/28,408 139/28,788 167/27,742 175/27,470 years Model 1a 1.0 (ref) 0.84 (0.66-1.06) 0.97 (0.77-1.21) 0.98 (0.78-1.23) 1.02 (0.94-1.11) 0.57 Model 2b 1.0 (ref) 0.85 (0.67-1.07) 0.97 (0.77-1.22) 0.96 (0.76-1.21) 1.01 (0.93-1.10) 0.78 Model 3d 1.0 (ref) 0.85 (0.67-1.08) 0.96 (0.77-1.21) 0.95 (0.75-1.20) 1.01 (0.93-1.10) 0.85 Model 4e 1.0 (ref) 0.84 (0.66-1.06) 0.92 (0.73-1.15) 0.93 (0.74-1.18) 1.01 (0.92-1.10) 0.89 Abbreviations: aMed, alternate Mediterranean diet; BMI, body mass index; CARDIA, Coronary Artery Risk Development in Young Adults; CI, confidence interval; DGA, Dietary Guidelines for Americans; EC, empty calorie; HR, hazard ratio; LC, low carbohydrate; SD, standard deviation. a HRs (95% CI) derived from Cox proportional hazards models adjusted for potential confounding by age, race, sex, and CARDIA center.

109 b HRs (95% CI) derived from Cox proportional hazards models adjusted for Model 1 covariates plus repeated measures of education and smoking status, cumulative average physical activity, and cumulative average estimated total energy intake. c HRs (95% CI) derived from Cox proportional hazards models adjusted for Model 2 covariates and cumulative average BMI. d HRs (95% CI) derived from Cox proportional hazards models adjusted for EC Model 2 covariates and cumulative average Mediterranean diet score. e HRs (95% CI) derived from Cox proportional hazards models adjusted for EC Model 3 covariates and cumulative average BMI.

110 Supplemental Table 5.1 Description of the dietary pattern scores. The dietary pattern score for each participant was the sum of category scores 0 to 4 for intake of items considered beneficial by the specific dietary pattern plus scores in reverse order (4 to 0) for adverse foods. Food items with moderate recommended consumption (+/-) were scored 0-2-4-2- 0, with moderate intake receiving the highest score. Neutral foods were not included in the final score. CARDIA Food 2015 Groups DGA Paleo LC EC Index Avocado + + + 0 Beans + - - 0 Alcohol +/- - - 0 Butter 0 - + 0 Chocolate 0 - - + Coffee/Tea 0 0 0 0 Artificially Sweetened Beverages 0 0 0 0 Dairy Dessert - - - + Eggs 0 + + 0 Fish + + + 0 Fried foods 0 - - 0 Fried potatoes - - - + Fruit + +/- - 0 Fruit juice - - - + Grain dessert - - - + Green vegetables + + + 0 Lean fish + + + 0 Lean red meat - + + 0 Low fat dairy + - - 0 Margarine - - - + Meal replacement 0 - 0 0 Oil 0 - + 0 Organ meat 0 + + 0 Other vegetables + + + 0 Pickled foods 0 - 0 0 Potatoes 0 +/- - 0 Poultry 0 + + 0 Processed meat - - + 0 Regular red meat - + + 0 Refined grains - - - 0 Salty snacks - - - +

111 Sauces 0 0 0 0 Seeds, nuts + + + 0 Shellfish + + + 0 SSBs - - - + Soups 0 0 0 0 Soy products + 0 0 0 Sugar substitutes 0 0 0 0 Sweet extra - - - + Tomato + + - 0 Whole fat dairy 0 - + 0 Whole grains + - - 0 Yellow vegetables + + + 0 Nonalcoholic beer 0 0 0 0 Total Possible Score 108 148 140 36

112 Supplemental Table 5.2 Description of the A Priori Diet Quality Score (Sijtsma FP, Meyer KA, et al. Diet quality and markers of endothelial function: the CARDIA study. Nutr Metab Cardiovasc Dis. 2014 Jun;24(6):632-8.) CARDIA Food Groups A Priori Diet Quality Score Avocado + Beans + Beer + Coffee + Fish + Fruit + Dark green vegetables + Lean fish + Low fat dairy + Liquor + Oil + Other vegetables + Poultry + Seeds, nuts + Soy products + Tea + Tomato + Whole grains + Wine + Yellow Vegetables + Butter - Fried foods - Fried potatoes - Grain dessert - Organ meat - Processed meat - Regular red meat - Salty snacks - Sauces - Soft drink - Sweet breads - Sweet extras - Whole fat dairy - Chocolate 0 Diet soft drink 0 Eggs 0 Fruit juice 0

113 Lean red meat 0 Margarine 0 Meal replacement 0 Pickled foods 0 Potatoes 0 Refined grains 0 Shellfish 0 Soups 0

114

Supplemental Table 5.3 Y0, Y7 and Y20 cumulative average nutrient and food group characteristics of the 2015 Dietary Guidelines for Americans dietary pattern score by quartile. Quartile of 2015 DGA Score Nutrient Intakea Q1 Q2 Q3 Q4 Total Energy (Calories) 3235.9 (1193.9) 2781.6 (1159.3) 2470.7 (979.0) 2338.0 (835.7) Carbohydrate (g) 116.5 (15.8) 116.0 (15.9) 117.7 (16.1) 122.6 (17.0) % Energy from Carbohydrate 46.6 (6.3) 46.2 (6.4) 47.0 (6.4) 49.0 (6.8) Fiber (g) 3.6 (1.5) 4.6 (1.7) 5.9 (2.3) 7.8 (2.8) Protein (g) 34.5 (4.9) 36.5 (5.0) 38.2 (5.4) 39.1 (6.2) % Energy from Protein 13.8 (2.0) 14.6 (2.0) 15.3 (2.2) 15.6 (2.5) Total Fat (g) 41.9 (5.2) 42.0 (5.5) 40.6 (5.6) 38.6 (6.4) % Energy from Fat 37.8 (4.7) 37.8 (5.0) 36.6 (5.1) 34.8 (5.7) Saturated Fat (g) 15.3 (2.5) 15.1 (2.7) 14.4 (2.7) 13.3 (3.0) Monounsaturated Fat (g) 15.9 (2.1) 15.7 (2.3) 15.0 (2.4) 14.2 (2.9) Polyunsaturated Fat (g) 7.7 (1.9) 8.1 (2.1) 8.1 (2.0) 8.2 (2.3) Dietary Cholesterol (mg) 160.2 (47.3) 157.4 (50.2) 147.9 (49.6) 129.8 (45.8) Sodium (mg) 1449.1 (211.7) 1498.9 (228.0) 1528.1 (260.6) 1534.0 (259.9) Alcohol (g) 3.6 (5.1) 3.5 (5.1) 3.8 (4.6) 4.2 (4.5) Food Group Intakeb Fruit 0.9 (0.7) 1.1 (0.8) 1.3 (0.8) 1.7 (0.8) Vegetables 1.0 (0.4) 1.3 (0.5) 1.6 (0.6) 2.4 (1.1) Grains 2.9 (0.7) 2.8 (0.7) 2.8 (0.7) 2.8 (0.7) Dairy 1.0 (0.6) 1.2 (0.8) 1.4 (0.9) 1.4 (0.8) Meat 2.0 (0.7) 2.0 (1.1) 1.8 (0.8) 1.4 (1.0) Fish 0.2 (0.2) 0.3 (0.4) 0.4 (0.4) 0.6 (0.5) Nuts 0.1 (0.2) 0.3 (0.3) 0.3 (0.4) 0.4 (0.4) Legumes 0.07 (0.1) 0.08 (0.1) 0.1 (0.1) 0.2 (0.3) a Estimated daily nutrient and food group intakes reported as quartile mean (SD); values for all nutrients (except Total Energy) are per 1,000 Calories. b Units = serving per 1,000 Calories per day

115 Supplemental Table 5.4 Y0, Y7 and Y20 cumulative average nutrient and food group characteristics of the Paleolithic dietary pattern score by quartile Quartile of Paleolithic Score Nutrient Intakea Q1 Q2 Q3 Q4 Total Energy (Calories) 3502.8 (1200.9) 2788.4 (970.9) 2432.9 (899.0) 2088.0 (778.3) Carbohydrate (g) 118.5 (14.4) 118.3 (16.0) 118.2 (16.6) 117.5 (18.4) % Energy from Carbohydrate 47.4 (5.7) 47.3 (6.4) 47.3 (6.6) 47.0 (7.4) Fiber (g) 4.3 (1.9) 5.1 (2.2) 5.8 (2.7) 6.8 (3.0) Protein (g) 34.5 (4.8) 35.9 (4.8) 37.6 (5.2) 40.4 (6.5) % Energy from Protein 13.8 (1.9) 14.3 (1.9) 15.1 (2.1) 16.2 (2.6) Total Fat (g) 41.7 (5.1) 41.2 (5.5) 40.5 (5.9) 39.7 (6.7) % Energy from Fat 37.6 (4.6) 37.1 (4.9) 36.5 (5.3) 35.7 (6.0) Saturated Fat (g) 15.5 (2.5) 14.8 (2.6) 14.3 (2.8) 13.5 (3.0) Monounsaturated Fat (g) 15.6 (2.1) 15.3 (2.3) 15.0 (2.5) 14.8 (3.0) Polyunsaturated Fat (g) 7.6 (1.8) 8.0 (2.0) 8.1 (2.1) 8.3 (2.5) Dietary Cholesterol (mg) 147.7 (43.5) 149.7 (47.5) 148.0 (51.6) 149.7 (55.5) Sodium (mg) 1476.9 (207.1) 1487.7 (222.6) 1516.2 (253.7) 1530.5 (280.6) Alcohol (g) 3.3 (4.1) 3.5 (4.6) 3.9 (5.1) 4.4 (5.4) Food Group Intakeb Fruit 1.0 (0.7) 1.1 (0.7) 1.3 (0.8) 1.6 (0.9) Vegetables 1.1 (0.4) 1.3 (0.5) 1.6 (0.7) 2.3 (1.1) Grains 2.9 (0.6) 2.9 (0.7) 2.8 (0.7) 2.7 (0.7) Dairy 1.2 (0.7) 1.2 (0.7) 1.3 (0.9) 1.2 (0.9) Meat 1.7 (0.6) 1.8 (0.7) 1.8 (0.8) 1.9 (1.4) Fish 0.2 (0.2) 0.3 (0.3) 0.4 (0.4) 0.6 (0.6) Nuts 0.2 (0.3) 0.3 (0.3) 0.3 (0.4) 0.4 (0.4) Legumes 0.1 (0.1) 0.1 (0.2) 0.1 (0.2) 0.1 (0.2) a Estimated daily nutrient and food group intakes reported as quartile mean (SD); values for all nutrients (except Total Energy) are per 1,000 Calories. b Units = serving per 1,000 Calories per day

116 Supplemental Table 5.5 Y0, Y7 and Y20 cumulative average nutrient and food group characteristics of the Low Carbohydrate dietary pattern score by quartile Quartile of Low Carbohydrate Score Nutrient Intakea Q1 Q2 Q3 Q4 Total Energy (Calories) 3007.5 (1147.3) 2709.3 (1110.7) 2569.7 (1037.9) 2539.2 (1068.0) Carbohydrate (g) 124.6 (15.3) 120.5 (15.1) 117.4 (15.3) 109.9 (16.3) % Energy from Carbohydrate 49.8 (6.1) 48.2 (6.0) 46.9 (6.1) 44.0 (6.5) Fiber (g) 5.0 (2.4) 5.4 (2.5) 5.6 (2.7) 5.9 (2.9) Protein (g) 34.7 (5.2) 36.0 (4.9) 37.6 (5.3) 40.0 (6.3) % Energy from Protein 13.9 (2.1) 14.4 (2.0) 15.0 (2.1) 16.0 (2.5) Total Fat (g) 39.6 (5.4) 40.3 (5.7) 41.0 (5.8) 42.3 (6.2) % Energy from Fat 35.6 (4.9) 36.3 (5.1) 36.9 (5.2) 38.0 (5.6) Saturated Fat (g) 14.3 (2.6) 14.4 (2.7) 14.6 (2.9) 14.8 (3.1) Monounsaturated Fat (g) 14.8 (2.4) 15.0 (2.4) 15.2 (2.5) 15.7 (2.7) Polyunsaturated Fat (g) 7.6 (1.7) 7.9 (2.1) 8.0 (1.9) 8.6 (2.5) Dietary Cholesterol (mg) 133.4 (41.7) 143.3 (44.0) 153.8 (51.1) 164.6 (55.2) Sodium (mg) 1471.2 (222.9) 1484.6 (226.8) 1500.4 (239.1) 1554.4 (273.3) Alcohol (g) 2.8 (3.8) 3.5 (4.2) 3.7 (4.7) 5.1 (6.0) Food Group Intakeb Fruit 1.1 (0.8) 1.2 (0.8) 1.3 (0.8) 1.3 (0.8) Vegetables 1.2 (0.6) 1.4 (0.7) 1.6 (0.8) 2.0 (1.1) Grains 3.0 (0.7) 2.9 (0.7) 2.8 (0.7) 2.6 (0.7) Dairy 1.2 (0.7) 1.2 (0.7) 1.3 (1.0) 1.2 (0.8) Meat 1.6 (0.7) 1.7 (0.7) 1.8 (0.8) 2.1 (1.3) Fish 0.2 (0.3) 0.3 (0.3) 0.4 (0.4) 0.6 (0.6) Nuts 0.2 (0.3) 0.3 (0.3) 0.3 (0.4) 0.3 (0.4) Legumes 0.1 (0.2) 0.1 (0.1) 0.1 (0.1) 0.1 (0.2) a Estimated daily nutrient and food group intakes reported as quartile mean (SD); values for all nutrients (except Total Energy) are per 1,000 Calories. b Units = serving per 1,000 Calories per day

117 Supplemental Table 5.6 Y0, Y7 and Y20 cumulative average nutrient and food group characteristics of Empty Calorie dietary pattern score by quartile Quartile of Empty Calorie Score Nutrient Intakea Q1 Q2 Q3 Q4 Total Energy (Calories) 2348.5 (957.9) 2566.9 (997.7) 2731.5 (1075.2) 3186.8 (1208.3) Carbohydrate (g) 115.6 (19.8) 117.5 (16.3) 118.8 (15.0) 120.6 (13.4) % Energy from Carbohydrate 46.2 (7.9) 47.0 (6.5) 47.5 (6.0) 48.3 (5.4) Fiber (g) 6.5 (3.3) 5.7 (2.5) 5.1 (2.2) 4.6 (2.0) Protein (g) 39.8 (6.7) 37.7 (5.2) 36.2 (4.8) 34.6 (4.4) % Energy Intake from Protein 15.9 (2.7) 15.1 (2.1) 14.5 (1.9) 13.9 (1.8) Total Fat (g) 39.5 (7.1) 40.8 (5.8) 41.3 (5.4) 41.6 (4.7) % Energy from Fat 35.6 (6.3) 36.7 (5.3) 37.1 (4.8) 37.4 (4.2) Saturated Fat (g) 13.9 (3.4) 14.5 (2.8) 14.8 (2.7) 14.9 (2.3) Monounsaturated Fat (g) 14.6 (3.0) 15.1 (2.5) 15.4 (2.3) 15.6 (2.0) Polyunsaturated Fat (g) 8.0 (2.6) 8.1 (2.1) 8.0 (2.0) 8.0 (1.8) Dietary Cholesterol (mg) 149.0 (61.3) 148.7 (48.8) 149.8 (45.9) 147.7 (40.1) Sodium (mg) 1556.7 (289.8) 1512.7 (244.7) 1482.1 (224.3) 1457.8 (191.9) Alcohol (g) 5.7 (6.6) 3.8 (4.3) 3.1 (3.9) 2.4 (3.0) Food Group Intakeb Fruit 1.3 (0.9) 1.3 (0.9) 1.2 (0.7) 1.1 (0.7) Vegetables 2.0 (1.2) 1.6 (0.8) 1.4 (0.7) 1.3 (0.5) Grains 2.8 (0.8) 2.8 (0.7) 2.8 (0.6) 2.8 (0.6) Dairy 1.3 (0.9) 1.3 (0.8) 1.2 (0.8) 1.1 (0.6) Meat 1.9 (1.4) 1.8 (0.8) 1.8 (0.7) 1.8 (0.6) Fish 0.5 (0.5) 0.4 (0.4) 0.4 (0.4) 0.3 (0.3) Nuts 0.3 (0.4) 0.3 (0.3) 0.3 (0.4) 0.2 (0.3) Legumes 0.2 (0.3) 0.1 (0.1) 0.1 (0.1) 0.09 (0.1) a Estimated daily nutrient and food group intakes reported as quartile mean (SD); values for all nutrients (except Total Energy) are per 1,000 Calories. b Units = serving per 1,000 Calories per day

Supplemental Table 5.7 Association between percent total calories from added sugar and saturated fat at Y20 and diabetes risk in young adult men and women from the CARDIA study. Quartile of % Total Calories from Added Sugar + Saturated Fat Q1 Q2 Q3 Q4 Per SD P trend Mean % total 16.1 (2.6) 21.4 (1.2) 25.6 (1.4) 34.6 (6.4) calories (SD) n T2D/n 32/624 50/625 58/625 61/624 Model 1a 1.0 (ref) 1.49(0.95-2.32) 1.58(1.02-2.45) 1.57(1.01-2.43) 1.21(1.06-1.37) 0.004 Model 2b 1.0 (ref) 1.45(0.93-2.27) 1.46(1.95-2.27) 1.27(0.81-1.98) 1.11(0.98-1.26) 0.11 Model 3c 1.0 (ref) 1.39(0.89-2.17) 1.35(0.87-2.11) 1.11(0.69-1.76) 1.07(0.94-1.23) 0.31 Model 4d 1.0 (ref) 1.25(0.79-1.96) 1.30(0.83-2.02) 1.03(0.65-1.65) 1.05(0.91-1.21) 0.48 a HRs (95% CI) derived from Cox proportional hazards models adjusted for potential confounding by age, race, sex, and CARDIA center. b HRs (95% CI) derived from Cox proportional hazards models adjusted for Model 1 covariates plus Y20 measures of education and smoking status, Y20 physical activity and Y20 total energy intake. c HRs (95% CI) derived from Cox proportional hazards models adjusted for EC Model 2 covariates and Y20 alternate Mediterranean diet score. d HRs (95% CI) derived from Cox proportional hazards models adjusted for EC Model 3 covariates and Y20 BMI.

118 CHAPTER 6 Artificially-sweetened and Sugar-sweetened Beverages and Risk of Incident Type 2 Diabetes in Young Adults: the CARDIA Study

Background Epidemiological evidence has demonstrated a positive association between artificially-sweetened (ASB) and sugar-sweetened (SSB) beverage consumption and T2D risk.

However, data to support this relationship in young adults is limited.

Objective To examine the association between ASB, SSB, and total sweetened beverage (ASB +

SSB) consumption and T2D risk in young adults over time.

Design Prospective analysis of 4,627 young adult black and white men and women from the

Coronary Artery Risk Development in Young Adults (CARDIA) study with repeated dietary histories. Cox proportional hazards regression models were used to examine the relationship between cumulative average sweetened beverage intake and incident T2D.

Results During the 30-year follow-up period, 627 (13.6%) incident cases of T2D occurred. ASB consumption was associated with a 48% increased risk of T2D in the highest quintile of intake compared with the lowest (HR 1.48, 95% CI 1.11-1.97), but adjustment for dieting behavior and cumulative average BMI attenuated this association (HR 1.23, 95% CI .93-1.64). For SSBs, the quintile analysis showed a positive, yet imprecise association between SSB intake and risk for

T2D in the highest quintile compared with the lowest quintile (HR 1.35, 95% CI 1.00-1.83).

Adjustment for diet quality attenuated this relationship and further adjustment for BMI mediated this relationship over time (HR 1.14, 95% CI 0.84-1.55). When examining cumulative average

SSB intake as a continuous variable, there was a modest positive association between higher

SSB intake and T2D risk per standard deviation of SSB intake (1.6 servings/day), independent of overall diet quality and cumulative average BMI (HR 1.10, 95% CI 1.01-1.20). Participants in

119 the highest quintile of total sweetened beverage intake had a 49% increased risk of T2D compared with those in the lowest quintile in the fully adjusted model, independent of dieting behavior and BMI (HR 1.49, 95% CI 1.06-1.84).

Conclusions In CARDIA, the observed association between ASB consumption and T2D risk was largely explained by dieting behavior and BMI over time. Categorical analysis of SSB intake was suggestive of a moderate positive association that was attenuated by diet quality and

BMI, whereas analysis of SSB intake on a continuous level was positively associated, independent of the aforementioned confounder and mediator. Long-term total-sweetened beverage consumption was associated with an increased risk of T2D, independent of diet quality, lifestyle factors, and BMI in this population; suggesting that whether or not they are causally related to T2D, higher intake of these beverages serves as a strong predictor of higher T2D risk in young adults.

120 Introduction

For several decades, the rise in obesity and type 2 diabetes (T2D) prevalence in the

United States closely paralleled the rise in both artificially-sweetened (ASB) and sugar- sweetened beverage (SSB) consumption.341 Consequently, lowering SSB intake has become an important focus of public health nutrition and recent reports indicate that SSB intake in the U.S. is declining.342 Consumption of ASBs however, often marketed as healthy alternatives to SSBs, has continued to increase343 along with obesity and T2D levels.344 Notably, recent reports have also indicated that intake trends of both ASBs and SSBs vary among different age and sociodemographic groups.342,343

Although the body of evidence supports a positive association between SSB consumption and obesity and T2D risk,163,345,346 there are methodological limitations of the intervention studies and the heterogeneity of observational studies suggest potential publication bias.162,172 On the other hand, despite the promotion of ASBs as a healthy sugar-free, low-calorie alternative to

SSBs,347,348 their role in weight management remains inconclusive and evidence on the metabolic and health effects of routine ASB consumption over time is limited.349,350 Furthermore, evidence from prospective observational studies suggests a positive association between ASB consumption and long-term cardiometabolic risk, but the results thus far have been inconsistent and publication bias and residual confounding have been implicated here as well.163,349,350

Overall, the observational studies on this topic have largely examined middle-aged and older adults and evidence is lacking for how sweetened beverage consumption habits relate to T2D risk beginning in early adulthood.

The Coronary Artery Risk Development in Yong Adults (CARDIA) study is well suited to further inform this topic as it offers a young-adult population aged 18-30 years with repeated

121 measures of beverage intake, body weight, and dieting behavior over a 30-year follow-up period.

I therefore examined the relationships between ASB, SSB, and total sweetened beverage (ASB and SSB intake combined) consumption over time and T2D risk in young adult men and women from the CARDIA study.

Methods

Study Population

Birmingham, Alabama; Chicago, Illinois; Minneapolis, Minnesota, and Oakland, California.

15, 20, 25, and 30 years after baseline, with retention of 91%, 86%, 81%, 79%, 74%, 72% and

Participants with a diagnosis of diabetes at baseline (n = 34), missing baseline diabetes status (n = 82) or baseline dietary data (n = 4) were excluded from this analysis. Individuals who only participated in the baseline visit and thus had no follow-up data were excluded (n = 153). In addition, individuals with extreme energy intakes (<600 kcal/day or >6,000 kcal/day for women (n = 99) and <800 kcal/day or >8,000 kcal/day for men (n = 116)) were also excluded.

122 The rate of missing data for other pertinent covariates was low, (<1%), and therefore missing values were imputed by sex and race subgroup to the median value for continuous variables

(BMI, alcohol intake, physical activity), and the most frequent categorical value for multichotomous variables (smoking status). The final study sample for this analysis included

4,672 black and white young adult men and women.

Beverage Consumption

Dietary intake was assessed at baseline (year 0), years 7 and 20 using the CARDIA Diet

History, an interviewer-administered, validated dietary history consisting of a short questionnaire on general dietary practices followed by a comprehensive questionnaire about typical intake of foods using the previous one month as a reference for recall.340 Diet history data was then coded by the University of Minnesota Nutrition Coordinating Center (NCC) and categorized into 166 food groups. As done in previous CARDIA studies,294,351 I further collapsed these groups to define SSBs as sugar-sweetened soft drinks and fruit drinks and ASBs as soft drinks and fruit drinks sweetened with non-nutritive (non-caloric) sweeteners, both measured as the total number of servings per day. I also calculated combined total sweetened beverage intake as the sum of

SSB and ASB intake at each time point to account for cross-over from consumption of SSBs to

ASBs or vice versa that could occur over time, as simply adjusting for intake of the other beverage type may not completely account for the impact on T2D risk.

To examine the relationship between beverage intake over time and T2D risk, I calculated a cumulative average value of ASB, SSB, and total sweetened beverage consumption for each participant using dietary data from years 0, 7, and 20. For individuals who had not developed diabetes through Y7, the cumulative average beverage intake from Y0 and Y7 was used. For individuals who had no documented diabetes through Y20, the cumulative average of

123 beverage intake at Y0, Y7, and Y20 was used. Cumulative averages for participants who did not develop diabetes during follow-up included all available beverage intake data prior to censoring.

Cumulative averages were calculated based on available data, individuals without repeated measures of diet were assigned their baseline beverage intake level.

Incident Type 2 Diabetes Mellitus

Diabetes status was assessed at examination years 0, 7, 10, 15, 20, 25 and 30. All blood samples were drawn and processed according to standard procedures and serum glucose was assayed using the hexokinase method at a central laboratory.289 Diabetes was defined as measured fasting glucose  7 mmol/l (126 mg/dl), 2-hour oral glucose tolerance test  11.1 mmol/l (available Y10, Y20, and Y25), hemoglobin A1c  6.5% (Y25), or self-report of oral hypoglycemic medication(s) or insulin use. CARDIA did not differentiate between type 1 and type 2 diabetes mellitus; however, it is likely that most incident cases identified during follow-up are T2D given the age of the cohort.

Covariates

124 energy intake (kilocalories) was calculated from the CARDIA Diet History. An alternate

Mediterranean (aMed) diet score was calculated using methods described in Fung et al.141. The aMed diet score was utilized because it does not include ASB or SSB consumption, yet provides an account of overall diet quality including alcohol intake as a covariate. Body weight was measured with light clothing to the nearest 0.09 kg and height was measured without shoes to the nearest 0.5 cm. Body mass index (BMI) was calculated from these measurements as weight in kilograms divided by height in meters squared. Dieting behavior was assessed as part of a weight history questionnaire at baseline and on medical history questionnaires at each follow-up year.

Specifically, participants were asked, “Have you ever been on a weight reducing diet?”, and “If yes: are you on such a diet now?” (yes/no).

Statistical Analysis

I examined differences in participants’ sociodemographic and clinical characteristics by cumulative average frequency of intake of ASBs and SSBs intake using 2-sided t tests and chi- square tests for continuous and categorical variables, respectively. Survival analysis using multivariable Cox proportional hazards models was used to estimate the hazard ratios (HRs) and corresponding 95% confidence intervals (95% CI) for incident diabetes during follow-up.

Separate models were fit for ASBs, SSBs, and total sweetened beverages. Once a participant developed documented diabetes at a CARDIA follow-up examination, they were considered to have an event and were subsequently censored. Follow-up time was calculated as the time from baseline to the examination visit where diabetes was identified or until the last examination where diabetes status was ascertained (due to death, loss to follow-up or end of cohort surveillance), whichever came first.

125 For the analysis, participants were ranked into quintiles of beverage intake. Since many participants were ASB non-consumers, non-consumers were coded as 0 and divided consumers into quartiles to ensure variability across 5 levels of consumption. The lowest quintile served as the reference group for the analysis. In a sensitivity analysis, 5 beverage intake frequency categories were created that allowed for cut points with an adequate number of subjects and alignment with common levels and thresholds of intake. For ASBs, participants were categorized according to frequency of consumption as 0/never, any to 3 servings/week, 4-6 servings/week, 1-

2 servings/day, and >2 servings/day. For the SSB and total sweetened beverage analyses, individuals who reported consumption of less than 1 serving per week were grouped with non- consumers for statistical stability in comparisons because of the small number of non-consumers.

Using this approach, never consumers served as the reference group for the ASB analysis and never/infrequent consumers served as the reference group for the SSB analysis.

Multivariable models adjusted for preselected sociodemographic and lifestyle related confounders were used. The first model was adjusted for age, race, sex, education, CARDIA center, smoking, cumulative average energy intake, and cumulative average physical activity.

The second model was adjusted for all Model 1 covariates plus cumulative average aMed score and to a third model I further adjusted for cumulative average BMI as a potential mediator. For the ASB analysis, I first adjusted the basic model for participant report of being on a weight loss diet at the time of dietary history assessment at year 0, 7, 20 prior to adjustment for diet quality and BMI. Cumulative averages were calculated for covariates in the same way as beverage intake, using data from Y0, Y7, and Y20 or until censoring. Education and smoking status were treated as repeated measures in the models. All ASB models were adjusted for cumulative average SSB intake, and vice-versa for SSB models. Beverage consumption was also modeled as

126 a continuous variable and hazard ratios were calculated per standard deviation of beverage intake.

It should be noted that use of cumulative average beverage intake may mask important changes in consumption patterns over time on T2D risk. To address this, a sensitivity analysis was performed using participants who were free of diabetes at Y20 and had Y20 dietary data (n

= 2,498). For these individuals, the change in beverage intake (servings/day) was calculated between the cumulative average of Y0 and Y7 and consumption at Y20 to create 4 consumption patterns over time: no-low cumulative average intake, stable intake (within ½ serving/day of baseline), increased intake greater than ½ serving/day, decreased intake greater than ½ serving/day. For the ASB analysis, non-consumers served as the reference group. Individuals with no intake to <1 serving/week served as the reference group for the SSB analysis. For statistical purposes, individuals with a stable cumulative average intake of ≤ ½ serving/day were included in the reference group for analysis of total sweetened beverage consumption patterns.

In this study sample, family history of diabetes was missing for 676 participants (15%), thus the main analysis was repeated in the subset of the population who had information on family history of diabetes (n = 3,951). In addition, I examined the data for effect modification by sex, race, BMI, smoking, and family history by including an interaction term for the variable of interest and each dietary pattern score separately and by performing stratified analyses. To avoid potential misclassification, a sensitivity analysis was performed excluding subjects who developed diabetes before age 30 (n=8). The proportional hazards assumption was tested by including an interaction term with log (base-e) transformed time for each covariate. There was no evidence that the models violated this assumption. All analyses were performed using SAS

9.4 (SAS Institute, Inc.), with a significance level of 0.05 for 2-sided tests.

127 Results

Mean (SD) cumulative average of Y0, Y7 and Y20 ASB and SSB intake was 0.47 (1.0) servings/day and 1.33 (1.6) servings/day, respectively, in the study population. Participant characteristics are presented in Table 1 according to quintile of ASB and SSB intake. Participant characteristics by quintile of total sweetened beverage intake are shown in Supplemental Table

1. Compared with non-consumers, participants who consumed ASBs were older and had lower cumulative average estimated energy intake. With higher ASB intake a greater proportion of participants were white and female, had higher education, physical activity levels, and cumulative average BMI. With higher SSB intake, a greater proportion of participants were male, black, current smokers, less educated, had higher cumulative average estimated energy intake, lower aMed diet quality scores, higher cumulative average BMI and consumed more alcohol. These individuals were also younger than those reporting less frequent SSB intake.

A total of 627 incident cases of T2D (13.6%) were documented during the follow-up period (mean 24.3 years). ASB consumption was positively associated risk of T2D, independent of dieting behavior and overall diet quality, as presented in Table 2. Participants in the upper quintile of ASB intake had a 48% increased risk for T2D compared with non-consumers (Q1)

(HR 1.48, 95% CI 1.11-1.97; p trend 0.003). This association was attenuated by adjustment for cumulative average BMI (HR 1.23, 95% CI 0.93-1.64; p trend = 0.21). Analysis per standard deviation of ASB intake mirrored these results.

As presented in Table 2, higher intake of SSBs was modestly associated with T2D risk in the quintile analysis using Model 1 (HR 1.35, 95% CI 1.00-1.83 for the highest vs lowest quintile). However, adjustment for diet quality (Model 2 HR 1.30, 95% CI 0.95-1.76) and BMI

(Model 3 HR 1.14, 95% CI 0.84-1.55) attenuated the association. Alternatively, the analysis

128 modeling cumulative average SSB intake as a continuous variable per standard deviation of intake (1.6 servings/day) showed a positive association between higher SSB intake and T2D risk, independent of overall diet quality and cumulative average BMI (Model 3 HR(per SD) 1.10, 95%

CI 1.01-1.20; p trend 0.04).

Higher frequency of total sweetened beverage intake was associated with an increased risk for T2D, independent of overall diet quality, energy intake, and BMI. Individuals in the highest quintile of intake had a 49% increased risk of T2D during follow-up compared with those in the lowest quintile (HR 1.49, 95% CI 1.06-1.84; p trend 0.02). Modeling cumulative average sweetened beverage intake as a continuous variable, higher intake was associated with a

10% increased risk of T2D per standard deviation of intake in the fully adjusted multivariable model (HR 1.10, CI 1.02-1.19). In the secondary analysis using ASB, SSB, and total sweetened beverage intake frequency categories, the findings were similar to the main results, as presented in Supplemental Table 2.

Results of the sub-group analysis in the 2,498 individuals who had not developed T2D and had dietary data at Y20 are presented in Table 3. Compared with ASB non-consumers, T2D risk appeared elevated in other categories of consumption dynamics, but the confidence intervals of the estimates were imprecise, limiting further inference. In the analysis of SSB consumption dynamics, an increase in SSB consumption of > ½ serving/day over time was associated with an increased risk of T2D compared with individuals with a cumulative average SSB consumption of none to < 1 serving/week in Model 1 (HR 1.85, 95% CI 1.01-3.67), but further adjustment for diet quality and BMI attenuated these results. A suggestive positive, yet imprecise, association was also observed between decreased SSB consumption and T2D risk. There was no difference in risk between the SSB control group and stable SSB intake over time. On average, individuals

129 whose intake was stable over time consumed less than 1 SSB serving/day, with a mean (SD) intake of 0.73 (0.69) servings/day for Y0/7 cumulative average and 0.61 (0.75) servings/day at

Y20, whereas those whose intake both increased and decreased consumed higher amounts than the other groups. Compared with non-consumers and individuals consuming less than ½ serving/day, those with stable intake of total sweetened beverages (1.34-1.38 servings/day) as well as increased or decreased intake greater than ½ serving/day appeared to have an increased risk for T2D over time, though the estimates lacked precision.

Stratified analyses as well as formal tests for interaction between frequency of ASB intake and cumulative average BMI, race, sex, smoking status and family history of diabetes provided no evidence of effect measure modification by these factors. Results of these analyses for SSB consumption suggested potential effect measure modification by race (p interaction =

0.01) and BMI (p interaction = 0.03). In an analysis stratified by race, examining SSB intake as a continuous variable and HR calculated per standard deviation of SSB intake, SSB intake was associated with an increased risk of T2D in whites (HR 1.23, 95% CI 1.05-1.44) but not blacks

(HR 1.09, 95% CI 0.98-1.20) in Model 2. Stratification on BMI category (normal (BMI <24.9), overweight (BMI 25.0-29.9), obese (BMI ≥30)) suggested that the association may be stronger with lower BMI (HRNormal 1.17, 95% CI (0.99-1.38); HROverweight 1.12, 95% CI 0.96-1.31;

HRObese 1.10, 95% CI 0.96-1.26). Sample size limitations prevented us from investigating these stratified group associations by quintile of SSB intake and thus limit the interpretation of the whether the relationship differs between groups. The sensitivity analyses provided no evidence for effect measure modification of the association between SSB intake and T2D risk by sex, smoking status, or family history of diabetes. The main study findings were not impacted by a

130 sensitivity analysis to address potential misclassification by the exclusion of the 8 participants who developed diabetes before age 30.

Discussion

In this longitudinal analysis of young adults in the CARDIA cohort, a strong, positive association was observed between ASB intake and T2D risk in individuals in the highest quintile of consumption compared with those in the lowest (non-consumers), independent of total energy intake, overall diet quality and other lifestyle factors. However, dieting behavior slightly attenuated this association and the association was further attenuated by adjustment for cumulative average BMI. A positive, yet imprecise association was observed between SSBs and

T2D risk in the highest quintile of intake compared with the lowest, but adjustment for diet quality and BMI completely attenuated this relationship. Analysis of SSB intake as a continuous variable, however, showed that higher intake over time was associated with a 10% increased risk per SD of intake, independent of energy intake, diet quality, and BMI. In the analysis examining the cumulative average consumption of both sweetened beverages (ASB + SSB) in quintiles I observed a 49% increased risk for incident T2D even after adjustment for BMI in those with the highest level of intake.

Previous studies have found that overweight and obese individuals report higher consumption of ASBs than leaner individuals,347,352 and those who consume ASBs often do so in an attempt to lose weight or because of poor health.353 In CARDIA, adjustment for BMI largely explains the association between ASB intake and T2D risk. While consumption of ASBs was associated with an increased risk of metabolic syndrome and T2D in the Multi-Ethnic Study of

Atherosclerosis,170 Atherosclerosis Risk in Communities,354 and Framingham offspring174 prospective cohort studies, the findings align with those of more recent studies by de Koning et

131 al.,179 and Romaguera, et al.,355 which found associations between ASB consumption and increased risk of T2D were attenuated after adjustment for measures of body weight. In the

Health Professionals Follow-Up Study, de Koning, et al. found that the observed association between ASB consumption and increased risk of T2D was attenuated and no longer significant after adjustment for BMI and measures of previous weight change, dieting, and total energy intake (HR 1.09, 95% CI 0.98-1.21 for the top vs bottom quartile of intake; p trend 0.13).179 In the European Prospective Investigation into Cancer and Nutrition (EPIC) study, the association between one 12-ounce daily increment in ASB consumption and increased T2D risk was no longer significant after adjustment for energy intake and BMI (HR 1.11, 95% CI 0.95-1.31).355

As noted in several recent systematic reviews, such findings suggest that reverse causality due to higher consumption of ASBs by overweight and obese individuals or residual confounding from clustering of lifestyle factors associated with consumption, such as diet quality, may have biased the ASB-cardiometabolic risk seen in previous studies.163 In addition, publication bias that favors studies reporting a positive association between ASBs and T2D has been identified in this body of research.350

Few intervention studies have directly examined the effects of ASB consumption on cardiometabolic parameters in human adults. Randomized controlled trials (RCTs) investigating the effect of ASBs on weight management have thus far provided inconsistent results. For instance, small RCTs with short interventions and follow-up periods have shown a null or modest effect of ASBs on reduction in body weight, waist circumference, and BMI; whereas

ASBs have consistently been associated with weight loss in long-term RCTs (greater than 12 months) with industry sponsorship.349,350 On the other hand, large prospective observational

132 studies have found positive associations between higher ASB intake and BMI as well as higher risk of obesity, hypertension, metabolic syndrome, T2D, stroke and cardiovascular events.350

Several mechanisms have been proposed to explain the potential role of artificial sweeteners in metabolic dysregulation;177,356 including alterations of the composition and function of gut microbiota.357,358 Results from an intervention trial by Pepino, et al. showed that compared with water, sucralose (a non-nutritive sweetener) intake affected glycemic response to an oral glucose load in obese subjects as evidenced by an increase in peak plasma glucose concentrations and glucose-stimulated insulin response in a modified 5-hour oral glucose tolerance test.359 Although observational studies have shown a positive association between

ASBs and glucose levels,163,360 intervention studies in lean, healthy adults have not replicated these findings.356,361 It has also been suggested that consumption of artificial sweeteners in foods and beverages over time may alter taste preferences and diet quality by increasing preference for sweet tasting foods, increase appetite, change gut hormone secretion and lead to excess caloric intake.341,349,362 In animal models, it has been shown that chronic long-term exposure to artificial sweeteners leads to increased food consumption, weight gain, and adiposity.363 However, studies of artificial sweetener intake in humans have failed to consistently replicate these findings364,365 and an extensive review by Mattes and Popkin concluded that the evidence currently available has not universally demonstrated an increase in appetite or energy intake with artificial sweetener use in humans.366 The lack of consistent evidence to support the use of artificial sweeteners for weight loss and preventing metabolic abnormalities coupled with observational evidence suggesting a positive association between routine ASB intake and cardiometabolic risk highlights the need for more research to assess the safety of long-term consumption, inform

133 accurate public health messaging, and further elucidate the potential mechanisms through which

ASBs may affect T2D risk.

The results from the SSB analysis are consistent with previous research demonstrating a link between SSB consumption and T2D risk after adjustment for confounders such as diet quality and total energy intake. Although weight gain attributed to excess calorie intake from

SSB consumption is accepted to mediate the association between SSBs and T2D,162,164 the results suggest that other etiological factors also contribute to this relationship as the results for both

SSB (as a continuous variable) and total sweetened beverage intake were only partially attenuated by adjustment for cumulative average BMI. Further, stratified analysis of SSB intake by BMI category suggested a stronger positive association with T2D risk in individuals with a cumulative average BMI less than 25 kg/m2 than in overweight and obese participants. SSB consumption has frequently been linked to weight gain and T2D risk in observational studies162,172,346,367 and several long-term intervention studies have demonstrated that SSB consumption is associated with positive energy balance.368–371 However, the results of short-term intervention studies on this topic have provided inconsistent results372 and evidence for whether or not decreased consumption would impact obesity prevalence and T2D risk remains inconclusive.373 For instance, a well-conducted randomized controlled trial by Ebbeling, et al. demonstrated that replacement of SSBs by non-caloric beverages was effective at reducing body weight only for adolescents in the upper tertile of baseline BMI.374 In this study, inference about the effect of changing SSB consumption patterns over time on T2D risk could not be made due to sample size limitations and variability in SSB intake. In CARDIA, individuals whose intake increased or decreased greater than ½ serving per day appeared to be at increased risk for T2D compared with never/rare consumers, although the estimates lacked precision. Further

134 exploration and clarification of the nature of the mechanisms responsible for the relationship between SSBs and obesity and diabetes warrants investigation.

The interpretation of these findings is strengthened by the design of the CARDIA study, with repeated measures of beverage intake to examine long term exposure and assess changes in consumption patterns in relation to T2D risk. All beverage intake data used in the analysis was collected prior to diabetes diagnosis, reducing the likelihood that recall bias strongly impacted the results. Additionally, I was able to adjust for repeated measures of important factors such as total energy intake, diet quality, and BMI as well as assess potential confounding by dieting behavior. The CARDIA cohort is younger than populations used in previous studies of beverage consumption and T2D risk, thus offering a unique assessment of this relationship.

I also noted several limitations to this study. Although I accounted for many confounders in the models, residual confounding likely occurs in all diet-disease observational studies. The use of self-reported dietary data is also subject to recall and other biases that may alter estimates.

The validity and reliability of the CARDIA Diet History has been demonstrated, however nutrient and energy estimates were found to have larger variability among blacks than whites.297,340 This may in part explain why the stratified analysis showed a positive association between SSB intake and T2D risk in whites, but not blacks. While the study was adequately powered to detect an association between beverage intake and T2D, a larger sample size could improve the precision of the estimates given the limitations of self-reported dietary data, particularly for the sensitivity analysis of beverage intake patterns over time.

In conclusion, in the CARDIA cohort of young adult men and women, long-term total sweetened beverage consumption was associated with an increased risk of T2D, independent of diet quality, lifestyle factors, and cumulative average BMI. When examined as a continuous

135 variable, SSB intake over time was also positively associated with T2D risk in this population.

The observed association between long-term ASB consumption and T2D risk was explained by dieting behavior and cumulative average BMI. Although the imprecision of the results precludes strong inference as to the impact of sweetened beverage consumption dynamics over time in this population, there was suggestive evidence that higher total sweetened beverage intake, whether stable, increasing or decreasing over time, was predictive of T2D risk relative to consistent no- low intake. These findings suggest that whether or not they are causally related to T2D, higher intake of sweetened beverages serves as a strong predictor of higher T2D risk in young adults.

Further research is needed to fully elucidate the etiological relationship between SSB consumption and T2D risk. In addition, more evidence from experimental and prospective observational studies is necessary to assess the long-term impact of ASBs on cardiometabolic health to inform accurate public health messaging.

136 Tables

TABLE 6.1a Characteristics of CARDIA participants according to Y0, Y7, and Y20 cumulative average quintile of ASB consumption. Quintile of ASB Consumption Characteristic a Q1 Q2 Q3 Q4 Q5 P d N 2323 575 571 582 576 ASB Intake 0.0 (0.0) 0.07 (0.04) 0.30 (0.09) 0.76 (0.19) 2.62 (1.62) <0.0001 (servings/day) Energy Intake 2921.2 2503.0 2534.7 2476.9 2444.5 <0.0001 (kcal/day) (1192.1) (950.8) (1043.7) (938.6) (928.5) aMed diet score 4.0 (1.5) 4.3 (1.4) 4.5 (1.5) 4.4 (1.4) 4.1 (1.5) <0.0001 SSB Intake 1.8 (1.9) 1.1 (1.2) 1.0 (1.1) 0.8 (1.0) 0.7 (1.0) <0.0001 (servings/day) Baseline Age (years) 24.6 (3.7) 25.0 (3.6) 25.0 (3.4) 25.3 (3.4) 25.2 (3.4) <0.0001 Race (% white) 35.3 49.9 58.5 70.6 86.8 <0.0001 Sex (% male) 51.7 41.6 37.5 36.6 37.9 <0.0001 Education (years) 13.8 (2.4) 15.1 (2.4) 15.0 (2.5) 15.5 (2.5) 15.8 (2.6) <0.0001 Smoking Status (%) Never 54.1 64.2 61.7 61.9 58.3 Former 11.6 15.3 14.4 17.7 14.6 <0.0001 Current 34.3 20.5 24.0 20.5 27.1 Physical Activity 360.0 374.3 391.6 373.5 395.6 0.004 (EU/week)b (247.9) (229.2) (239.7) (219.0) (220.3) Baseline Alcohol Intake 11.9 (20.9) 8.6 (14.5) 10.3 (16.2) 10.9 (15.4) 13.7 (21.2) <0.0001 (ml/day) BMI (kg/m2) 25.9 (5.5) 26.2 (5.4) 26.9 (5.7) 26.6 (5.3) 27.1 (5.4) <0.0001 Family Hx DM (%)c 16.8 15.8 13.9 15.8 16.6 0.63 Abbreviations: ASB, artificially-sweetened beverage; BMI, body mass index; CARDIA, Coronary Artery Risk Development in Young Adults; DM, diabetes mellitus; SSB, sugar-sweetened beverage. a Unadjusted cumulative average mean (SD) for all characteristics unless noted as baseline value or percentage. b EU = Exercise units; physical activity score derived from the CARDIA physical activity history, where 300 EU is approximately equal to 150 minutes of moderate-intensity physical activity per week. c Family history data unavailable for 676 subjects. d P-value is analysis of variance (age, education, physical activity, alcohol, BMI, glucose, BP, lipids) or Chi-square t-test (race, gender, smoking, family history) of association between dietary pattern quartile and characteristic.

137 TABLE 6.1b Characteristics of CARDIA participants according to Y0, Y7, and Y20 cumulative average quintile of SSB consumption. Quintile of SSB Consumption Characteristic a Q1 Q2 Q3 Q4 Q5 P d N 925 926 925 926 925 SSB Intake 0.07 (0.07) 0.39 (0.11) 0.85 (0.16) 1.56 (0.27) 3.79 (1.94) <0.0001 (servings/day) Energy Intake 2174.8 2371.3 2583.3 2949.8 3452.4 <0.0001 (kcal/day) (789.6) (890.6) (978.5) (1099.8) (1232.3) aMed diet score 4.6 (1.5) 4.5 (1.5) 4.1 (1.4) 4.0 (1.5) 3.8 (1.4) <0.0001 ASB Intake 1.0 (1.5) 0.5 (0.9) 0.4 (1.0) 0.3 (0.7) 0.2 (0.7) <0.0001 (servings/day) Baseline Age (years) 25.8 (3.3) 25.1 (3.5) 24.8 (3.5) 24.2 (3.7) 24.2 (3.7) <0.0001 Race (% white) 81.1 60.2 47.6 34.3 30.9 <0.0001 Sex (% male) 30.2 39.9 46.1 55.0 54.4 <0.0001 Education (years) 14.0 (4.4) 14.8 (3.9) 14.9 (2.2) 14.1 (3.7) 13.0 (4.2) <0.0001 Smoking Status (%) Never 60.1 60.8 61.7 56.1 50.2 Former 19.2 17.8 11.6 10.3 8.9 <0.0001 Current 20.7 21.4 26.7 33.7 41.0 Physical Activity 399.5 367.0 360.3 368.8 363.4 0.003 (EU/week)b (225.9) (223.5) (229.1) (258.2) (250.3) Baseline Alcohol Intake 10.2 (14.4) 10.1 (17.3) 11.1 (19.0) 12.4 (19.6) 13.3 (23.8) 0.0005 (ml/day) BMI (kg/m2) 25.3 (5.1) 26.1 (5.1) 26.6 (5.7) 26.5 (5.7) 26.9 (6.0) <0.0001 Family Hx DM (%)c 14.9 15.0 16.7 17.4 16.9 0.38 Abbreviations: ASB, artificially-sweetened beverage; BMI, body mass index; CARDIA, Coronary Artery Risk Development in Young Adults; DM, diabetes mellitus; SSB, sugar-sweetened beverage. a Unadjusted cumulative average mean (SD) for all characteristics unless noted as baseline value or percentage. b EU = Exercise units; physical activity score derived from the CARDIA physical activity history, where 300 EU is approximately equal to 150 minutes of moderate-intensity physical activity per week. c Family history data unavailable for 676 subjects. d P-value is analysis of variance (age, education, physical activity, alcohol, BMI, glucose, BP, lipids) or Chi-square t-test (race, gender, smoking, family history) of association between dietary pattern quartile and characteristic.

138 Table 6.2 Association between Y0, Y7, and Y20 cumulative average ASB, SSB, and total sweetened beverage intake and diabetes risk in young adult men and women from the CARDIA study, 1985-2015. Quintile of Beverage Intake ASBs Q1 Q2 Q3 Q4 Q5 By SDf P trend ASB Intakea 0 (0) 0.07 (0.04) 0.30 (0.09) 0.76 (0.19) 2.62 (1.62) (servings/day) n diabetes / 336/53,104 73/15,083 76/14,845 69/15,078 73/14,298 person-years Model 1b HR 0.97 (0.75- 1.11 (0.86- 1.16 (0.88- 1.54 (1.16- 1.13 (1.04- 1.0 (ref) 0.002 (95% CI) 1.25) 1.43) 1.52) 2.04) 1.21) Model 2c HR 0.96 (0.74- 1.08 (0.84- 1.13 (0.86- 1.49 (1.12- 1.12 (1.04- 1.0 (ref) 0.004 (95% CI) 1.24) 1.40) 1.49) 1.98) 1.20) Model 3d HR 0.96 (0.74- 1.09 (0.84- 1.14 (0.86- 1.48 (1.11- 1.12 (1.03- 1.0 (ref) 0.006 (95% CI) 1.25) 1.42) 1.50) 1.97) 1.20) Model 4e HR 0.94 (0.72- 0.97 (0.75- 1.02 (0.78- 1.23 (0.93- 1.06 (0.98- 1.0 (ref) 0.16 (95% CI) 1.21) 1.25) 1.34) 1.64) 1.14) SSBs SSB Intakea 0.07 (0.07) 0.39 (0.11) 0.85 (0.16) 1.56 (0.27) 3.79 (1.94) (servings/day) n diabetes / 89/22,955 107/23,697 113/23,177 148/21,961 170/20,618 person-years Model 1b HR 1.01 (0.76- 0.97 (0.72- 1.18 (0.87- 1.35 (1.00- 1.13 (1.04- 1.0 (ref) 0.003 (95% CI) 1.35) 1.29) 1.57) 1.83) 1.23) Model 2d HR 1.00 (0.75- 0.95 (0.71- 1.15 (0.86- 1.30 (0.95- 1.12 (1.03- 1.0 (ref) 0.01 (95% CI) 1.34) 1.27) 1.54) 1.76) 1.22) Model 3e HR 0.94 (0.71- 0.88 (0.66- 1.04 (0.78- 1.14 (0.84- 1.10 (1.01- 1.0 (ref) 0.04 (95% CI) 1.26) 1.17) 1.40) 1.55) 1.20) Sweetened

Beverages Total Intakea 0.27 (0.16) 0.78 (0.14) 1.31 (0.17) 2.11 (0.31) 4.52 (1.96) (servings/day) n diabetes / 92/22,886 104/23,399 119/22,753 141/22,373 171/20,997 person-years Model 1b HR 1.04 (0.78- 1.16 (0.88- 1.28 (0.97- 1.74 (1.32- 1.16 (1.08- 1.0 (ref) <0.0001 (95% CI) 1.38) 1.52) 1.68) 2.28) 1.25) Model 2d HR 1.03 (0.76- 1.14 (0.86- 1.25 (0.95- 1.68 (1.27- 1.15 (1.06- 1.0 (ref) 0.0005 (95% CI) 1.36) 1.50) 1.65) 2.22) 1.25) Model 3e HR 1.01 (0.76- 1.03 (0.78- 1.09 (0.83- 1.49 (1.06- 1.10 (1.02- 1.0 (ref) 0.02 (95% CI) 1.34) 1.36) 1.44) 1.84) 1.20) Abbreviations: ASB, artificially-sweetened beverage; CARDIA, Coronary Artery Risk Development in Young Adults; CI, confidence interval; HR, hazard ratio; SD, standard deviation; SSB, sugar-sweetened beverage. a Y0, Y7, and Y20 cumulative average mean (SD) beverage intake. b Multivariable model adjusted for potential confounding by age, race, sex, CARDIA center, repeated measures of education, cumulative average energy intake (Y0, 7, and 20), physical activity and repeated measures of smoking status. c ASB Model 2 adjusted for Model 1 covariates plus dieting behavior (weight reducing diet Y/N). d Multivariable models adjusted for Model 1 covariates plus cumulative average Mediterranean diet score. e Multivriable models adjusted for Model 2 covariates plus, cumulative average BMI as a potential mediator. f HR (95% CI) calculated per standard deviation of beverage intake. *ASB models additionally adjusted for SSB intake and dieting behavior (Models 3-4); SSB models adjusted for ASB intake; total sweetened beverage models adjusted for dieting behavior.

139 Table 6.3 Association between sweetened beverage consumption patterns over 20 years and diabetes risk in young adult men and women from the CARDIA study, 1985-2015. 20-Year Beverage Intake Pattern ASBs None Stablee Increasef Decreaseg n diabetes / n 91/1029 52/724 32/390 26/355 Y0/7 Intakea 0.0 (0.0) 0.32 (0.47) 0.79 (1.21) 1.93 (1.76) (servings/day) Y20 Intakea 0.0 (0.0) 0.25 (0.48) 3.02 (2.59) 0.43 (1.0) (servings/day) Model 1b HR 1.0 (ref) 1.19 (0.83-1.71) 1.46 (0.95-2.24) 1.41 (0.88-2.26) (95% CI) Model 2c HR 1.0 (ref) 1.23 (0.86-1.77) 1.45 (0.94-2.24) 1.43 (0.89-2.29) (95% CI) Model 3d HR 1.0 (ref) 1.16 (0.81-1.66) 1.27 (0.82-1.95) 1.24 (0.77-1.99) (95% CI) SSBs None - <1/weekh Stable Increase Decrease n diabetes / n 18/428 39/716 42/349 102/1005 Y0/7 Intakea 0.06 (0.06) 0.73 (0.69) 1.26 (1.35) 2.19 (1.76) (servings/day) Y20 Intakea 0.03 (0.07) 0.61 (0.75) 3.22 (2.35) 0.48 (0.77) (servings/day) Model 1b HR 1.0 (ref) 1.07 (0.60-1.91) 1.85 (1.01-3.67) 1.62 (0.94-2.78) (95% CI) Model 2c HR 1.0 (ref) 1.02 (0.57-1.82) 1.69 (0.92-3.10) 1.50 (0.87-2.59) (95% CI) Model 3d HR 1.0 (ref) 0.98 (0.55-1.76) 1.49 (0.81-2.73) 1.41 (0.82-2.42) (95% CI) Total SBs None - ≤ ½/dayi Stable Increase Decrease n diabetes / n 16/381 37/482 49/523 99/1112 Y0/7 Intakea 0.24 (0.14) 1.38 (0.92) 2.39 (1.76) 1.89 (1.45) (servings/day) Y20 Intakea 0.17 (0.18) 1.34 (0.96) 3.81 (2.70) 0.78 (1.02) (servings/day) Model 1b HR 1.0 (ref) 1.51 (0.84-2.73) 1.71 (0.96-3.04) 1.59 (0.93-2.72) (95% CI) Model 2c HR 1.0 (ref) 1.41 (0.78-2.56) 1.57 (0.88-2.81) 1.50 (0.87-2.58) (95% CI) Model 3d HR 1.0 (ref) 1.26 (0.70-2.29) 1.30 (0.73-2.32) 1.30 (0.76-2.23) (95% CI) Abbreviations: aMed, alternate Mediterranean diet; ASB, artificially-sweetened beverage; CARDIA, Coronary Artery Risk Development in Young Adults; CI, confidence interval; HR, hazard ratio; SSB, sugar-sweetened beverage. a Unadjusted mean (standard deviation) b Model 1 adjusted for potential confounding by age, race, sex, CARDIA center, repeated measures of education, cumulative average energy intake (Y0, 7, and 20), physical activity and repeated measures of smoking status. c Model 2 adjusted for Model 1 covariates plus cumulative average Mediterranean diet score. d Model 3 adjusted for Model 2 covariates plus, cumulative average BMI as a potential mediator. e Stable intake reflects a change in beverage intake of ≤ 0.5 servings/day between the cumulative average of Y0/Y7 and Y20. f Increased intake reflects an increase in beverage intake of > 0.5 servings/day between the cumulative average of Y0/Y7 and Y20.

140 g Decreased intake reflects a decreased in beverage intake of > 0.5 servings/day between the cumulative average of Y0/Y7 and Y20. h Reference group includes non-consumers and individuals with <1 serving/week cumulative average SSB intake. i Reference group includes non-consumers and individuals with <0.5 serving/day cumulative average SSB intake. *ASB models additionally adjusted for SSB intake and dieting behavior; SSB models adjusted for ASB intake.

Supplemental Table 6.1 Characteristics of CARDIA participants according to Y0, Y7, and Y20 cumulative average quintile of total sweetened beverage consumption. Quintile of ASB Consumption Characteristic a Q1 Q2 Q3 Q4 Q5 P d N 925 927 924 926 925 ASB Intake 0.06 (0.1) 0.2 (0.3) 0.3 (0.5) 0.5 (0.7) 1.2 (1.9) <0.0001 (servings/day) SSB Intake 0.2 (0.2) 0.6 (0.3) 1.0 (0.5) 1.6 (0.8) 3.3 (2.4) <0.0001 (servings/day) Energy Intake 2278.3 2420.2 2666.4 2882.5 3284.5 <0.0001 (kcal/day) (863.8) (951.9) (1028.0) (1126.6) (1232.0) aMed diet score 4.6 (1.6) 4.3 (1.5) 4.2 (1.5) 4.0 (1.5) 3.8 (1.5) <0.0001 Baseline Age (years) 25.4 (3.5) 25.1 (3.5) 24.7 (3.6) 24.6 (3.6) 24.4 (3.6) <0.0001 Race (% white) 62.0 52.6 48.3 42.8 48.4 <0.0001 Sex (% male) 38.9 42.9 44.1 49.0 51.6 <0.0001 Education (years) 15.2 (2.5) 15.0 (2.6) 14.6 (2.5) 14.2 (2.5) 13.9 (2.5) <0.0001 Smoking Status (%) Never 58.3 63.5 59.7 56.6 50.7 Former 19.2 13.6 13.1 11.0 10.8 <0.0001 Current 22.5 22.9 27.2 32.4 38.5 Physical Activity 375.8 366.2 374.0 374.9 368.1 0.88 (EU/week)b (223.9) (230.8) (241.4) (251.8) (241.9) Baseline Alcohol Intake 9.3 (15.6) 10.0 (17.5) 12.1 (18.8) 11.4 (19.4) 14.3 (23.1) <0.0001 (ml/day) BMI (kg/m2) 25.1 (4.9) 26.0 (5.1) 26.3 (5.6) 26.9 (5.8) 27.1 (5.9) <0.0001 Family Hx DM (%)c 13.7 16.7 15.8 16.2 18.5 0.14 Abbreviations: ASB, artificially-sweetened beverage; BMI, body mass index; CARDIA, Coronary Artery Risk Development in Young Adults; DM, diabetes mellitus; SSB, sugar-sweetened beverage. a Unadjusted cumulative average mean (SD) for all characteristics unless noted as baseline value or percentage. b EU = Exercise units; physical activity score derived from the CARDIA physical activity history, where 300 EU is approximately equal to 150 minutes of moderate-intensity physical activity per week. c Family history data unavailable for 676 subjects. d P-value is analysis of variance (age, education, physical activity, alcohol, BMI, glucose, BP, lipids) or Chi-square t-test (race, gender, smoking, family history) of association between dietary pattern quartile and characteristic.

141 Supplemental Table 6.2 Association between Y0, Y7, and Y20 ASB, SSB, and total sweetened beverage intake and diabetes risk in young adult men and women from the CARDIA study, 1985-2015. Frequency of Beverage Intake (servings)

ASBs 0/Never Any-3/week 4-6/week 1-2/day >2/day n diabetes / 336/53,104 141/28,228 67/14,063 40/9,444 43/7,569 person-years Model 1a HR 1.0 (ref) 1.03(0.84-1.26) 1.18(0.90-1.55) 1.19(0.84-1.68) 1.72(1.23-2.42) (95% CI) Model 2d HR 1.0 (ref) 1.01(0.83-1.25) 1.15(0.87-1.52) 1.15(0.81-1.64) 1.67(1.19-2.36) (95% CI) Model 3b HR 1.0 (ref) 1.04(0.85-1.27) 1.19(0.90-1.57) 1.19(0.84-1.69) 1.70(1.21-2.40) (95% CI) Model 4c HR 1.0 (ref) 0.94(0.77-1.15) 1.01(0.77-1.33) 1.01(0.71-1.42) 1.25(0.88-1.76) (95% CI) SSBs 0-<1/week 1-3/week 4-6/week 1-2/day >2/day n diabetes / 71/18,471 79/18,797 137/26,947 158/26,196 182/21,997 person-years Model 1a HR 1.0 (ref) 0.98(0.71-1.35) 1.03(0.76-1.38) 1.07(0.79-1.45) 1.33(0.97-1.84) (95% CI) Model 2b HR 1.0 (ref) 0.97(0.70-1.35) 1.01(0.75-1.36) 1.04(0.77-1.42) 1.28(0.92-1.77) (95% CI) Model 3c HR 1.0 (ref) 0.93(0.67-1.29) 0.96(0.71-1.29) 0.95(0.70-1.30) 1.14(0.82-1.58) (95% CI) Sweetened 0-<1/week 1-3/week 4-6/week 1-2/day >2/day Beverages n diabetes / 18/5,501 50/12,134 125/27,399 182/33,808 252/33,566 person-years Model 1a HR 1.0 (ref) 1.12(0.65-1.92) 1.21(0.74-1.20) 1.31(0.80-2.13) 1.76(1.08-2.87) (95% CI) Model 2b HR 1.0 (ref) 1.11(0.65-1.91) 1.20(0.73-1.97) 1.28(0.79-2.09) 1.70(1.04-2.78) (95% CI) Model 3c HR 1.0 (ref) 0.99(0.58-1.70) 1.05(0.64-1.73) 1.03(0.63-1.69) 1.28(0.78-2.09) (95% CI) Abbreviations: aMed, alternate Mediterranean diet; ASB, artificially-sweetened beverage; CARDIA, Coronary Artery Risk Development in Young Adults; CI, confidence interval; HR, hazard ratio; SD, standard deviation; SSB, sugar-sweetened beverage. a Model 1 adjusted for potential confounding by age, race, sex, CARDIA center, repeated measures of education, cumulative average energy intake (Y0, 7, and 20), physical activity and repeated measures of smoking status. b Model 2 adjusted for Model 1 covariates plus cumulative average Mediterranean diet score. c Model 3 adjusted for Model 2 covariates plus, cumulative average BMI as a potential mediator. d ASB Model 2 adjusted for Model 1 covariates plus dieting behavior. *ASB models additionally adjusted for SSB intake and dieting behavior; SSB models adjusted for ASB intake.

142 CHAPTER 7 Diet Quality and Cardiovascular Disease Risk in Postmenopausal Women

with Type 2 Diabetes: The Women’s Health Initiative

Background Dietary patterns have been linked with cardiovascular disease (CVD) risk in the general population, but evidence for the relationship between diet quality and CVD risk in individuals with diabetes is limited in older adults.

Objective To determine the relationship between diet quality and CVD risk in a population with prevalent type 2 diabetes (T2D).

Design Prospective population-based cohort study of 5,809 postmenopausal women aged 50-79 years, of diverse backgrounds and ethnicities, who reported having T2D at baseline in 1993 from the Women’s Health Initiative (WHI) clinical trials and observational study. Dietary intake was assessed in the WHI using a validated food frequency questionnaire (FFQ) designed to estimate usual intake over the past 3 months. Diet quality was defined using alternate Mediterranean

(aMed), Dietary Approach to Stopping Hypertension (DASH), Paleolithic (Paleo) and American

Diabetes Association (ADA) a priori dietary pattern scores calculated from the FFQ data.

Incident CVD, defined as fatal or nonfatal coronary heart disease (CHD), coronary revascularization, peripheral artery disease (PAD), congestive heart failure (CHF), or stroke.

CHD and stroke were also analyzed individually as separate outcomes.

Results During a mean 12.4 years of follow up, 1,454 (25%) CVD, 635 (10.9%) CHD and 372

(6.4%) stroke cases were documented. Women with higher aMed, DASH, and ADA dietary pattern scores had a lower risk of CVD compared to women with lower scores (Q5 v Q1)

(HRaMed 0.77, 95% CI 0.64-0.92; HRDASH 0.69, 95% CI 0.58-0.83; HRADA 0.83, 95% CI 0.69-

0.99). A strong inverse relationship was also observed between aMed, DASH, and ADA scores

143 and risk for CHD and stroke. No association was observed between the Paleo dietary pattern score and CVD risk.

Conclusions Dietary patterns that emphasize higher intake of fruits, vegetables, whole grains, nuts/seeds, legumes, a high unsaturated:saturated fat ratio as well as lower intake of red and processed meats, added sugars and sodium are associated with lower risk of CVD in postmenopausal women with T2D. Replicating these findings in other populations could inform more targeted interventions for T2D treatment and management.

144 Introduction

Diabetes is a major public health burden that affects 30.3 million (1 in 8) adults in the

U.S. If trends continue, it is projected that as many as 1 in 3 U.S. adults will have diabetes by the year 2050.42,375 Diabetes is a major risk factor for cardiovascular disease (CVD) as adults with type 2 diabetes (T2D) are at least two times more likely to develop CVD than adults without diabetes, and greater than 70% of individuals with diabetes will die of CVD.41 People with T2D tend to harbor major risk factors for CVD such as dyslipidemia, hypertension, and obesity, in addition to poor glycemic control.376–379 These aforementioned risk factors have demonstrated links with dietary intake.380–382 However, little is known about how dietary patterns may impact risk for CVD outcomes among individuals with T2D.

Evidence for the relationship between diet and CVD risk in people with T2D is sparse.

One small randomized intervention study conducted in overweight individuals with T2D showed modest improvement in serum lipids with diabetic and Mediterranean diets,235 and a sub-group analysis of the Prevención con Dieta Mediterránea (PREDIMED) study demonstrated the benefit of a Mediterranean diet on CVD risk in individuals with T2D.277 Moreover, there is a lack of observational evidence linking dietary patterns to risk for CVD outcomes in populations with

T2D. To address this gap, I examined the relationship between diet quality, as measured by four a priori dietary pattern scores at baseline, and CVD risk over time in postmenopausal women with T2D from the Women’s Health Initiative (WHI), a population-based prospective cohort study.

Methods

Study Population

145 The WHI is a longitudinal national health study of 161,808 postmenopausal women aged

50-79 years at enrollment from 1993 to 1998. It was conducted at 40 clinical centers across the

U.S. The study enrolled 68,132 women into one or more overlapping randomized clinical trials

(CT) testing the health effects of hormone therapy, dietary modification, and/or calcium and vitamin D supplementation and 93,676 women into an observational study (OS). Eligibility criteria included the ability to complete study visits based on local residency and expected survival (free of serious health conditions, e.g., class IV congestive heart failure, obstructive lung disease requiring supplemental oxygen, or severe chronic liver or kidney disease) for at least 3 years. The WHI targeted enrollment of 20% minority women including African American,

Hispanic (Mexican American, Puerto-Rican Hispanic, and Cuban Hispanic), Asian/Pacific

Islander (Japanese, Chinese, Hawaiian), and Native American women to reflect the general population of postmenopausal women in the U.S. at the time. The main study ended in 2005, with an average follow-up length of 9 years. Two 5-year WHI extension studies (2005-2010 and

2010-2015) continued follow-up for all consenting women. Institutional review boards at all participating institutions approved the study protocols and procedures and all participants provided written informed consent.298

This analysis included women from the CT and OS arms of the WHI who reported having T2D at baseline. At baseline, participants were asked if a physician had ever told them they had ‘sugar diabetes or high blood sugar’ when they were not pregnant, and about treatment with insulin or oral anti-diabetic medications. Baseline diabetes was defined as an affirmative answer to the above question or reported use of medication to treat diabetes. A validation study of the accuracy of self-reported diabetes in the WHI based on medication and laboratory data found the self-report to be valid and reliable.383 A total of 9,618 women out of 161,808 reported

146 having diabetes at baseline. Since the WHI did not differentiate type 1 from type 2 diabetes, women who reported an age of diabetes diagnosis less than 21 years (n = 140) or who were missing data on age at diagnosis (n=69) were excluded from the analysis in an attempt to select only women with prevalent T2D. Additionally, women who reported a history of CVD or stroke at baseline (n = 3,286) or who reported extreme energy intake using the upper and lower 1% of the sample distribution as cutoffs (<465 or >3,931 kcal/day) (n = 314) were excluded. The final study sample included 5,809 postmenopausal women from the WHI cohort.

Assessment of Diet Quality

The main diet assessment tool used in the WHI was a validated food frequency questionnaire (FFQ) based on instruments used in the WHI feasibility studies and the original

National Cancer Institute/Block FFQ.384 The FFQ was designed to capture usual intake over the past 3 months and consisted of three sections: 122 composite and single food line items, which included questions on the frequency of consumption and portion sizes, 19 adjustment questions related to type of fat intake, and 4 summary questions asking about usual intake of fruits, vegetables and added fats. Questionnaires were collected at baseline for all subjects and then at specified follow-up visits on a rotating basis for a cross-sectional subsample of the cohort each year. Only baseline dietary data were used for this analysis due to potential biases and sample size limitations for the follow-up years. A nutrient database consisting of 30 nutrients was derived from the WHI FFQ using the University of Minnesota Nutrition Coding Center nutrient database (Nutrition Coordinating Center, Minneapolis, MN).385 Additional food group measures were also derived from the WHI FFQ using the USDA MyPyramid Equivalents Database 2.0

(MPEDs).306

147 The relationship between diet quality and CVD risk was measured using four a priori dietary pattern scores. Higher scores indicate a diet more closely aligned with the specified dietary pattern. Specifically, the alternate Mediterranean (aMed) and Dietary Approach to Stop

Hypertension (DASH) diets, as well as two dietary patterns hypothesized to be relevant to glycemic control and diabetes management: the Paleolithic (Paleo) diet and the American

Diabetes Association (ADA) dietary recommendations. The aMed and DASH diet scores were calculated as previously described for WHI analyses.145,386 The aMed diet score is a variation of the original Mediterranean score that has been slightly modified based on dietary patterns and eating behaviors that have been consistently associated with lower risks of chronic disease in clinical and epidemiological studies.387 In brief, the aMed score assigns 1 point for intake above the cohort-specific median for fruits, vegetables, legumes, nuts, whole-grain products, fish, and

MUFA:SFA fat ratio and 1 point for intake below the median for red and processed meats.

Alcohol intake of 5-15 grams/day receives 1 point. Individual food group scores are summed for the total aMed score, with a range of 0-9. The DASH score awards points for high intake of fruits, vegetables, nuts and legumes, low-fat dairy products, and whole grains according to quintile rankings, where an individual in the highest quintile of intake for a food group receives 5 points, whereas someone in the lowest quintile receives 1. Intake of sodium, sweetened beverages, and red and processed meats are reversely scored (5-1), with participants in lower quintiles of intake receiving higher points. Individual food groups scores are summed, with a score range of 8-40.

To calculate a Paleo diet score, each participant was assigned a quintile rank for 13 food categories relevant to the underlying theoretical construct that is the basis for this diet score.314

The composite score was calculated like the DASH score with quintile rankings. As done in

148 previous studies,388,389 higher scores were given for higher intakes of foods that are considered characteristics of a Paleo diet (vegetables, fruit, lean meats (including lean beef, pork, veal, lamb, game and poultry), fish, nuts, and green leafy vegetables (to represent intake of calcium from sources other than dairy foods)) and for low or no consumption of foods that are not considered characteristic of a Paleo diet (processed meats, dairy, starches, empty calories, alcohol, sodium, and sugar-sweetened beverages), with a score range of 13-65.

The ADA dietary pattern score was developed according to the ADA dietary recommendations with evidence ratings from A to B, as previously described.390 This score is a sum of the scores for 10 components, 7 beneficial and 3 adverse for control of diabetes. A score of 1 is given for reporting greater than the median intake of beneficial components (whole grains, fruits, vegetables, low-fat dairy, plant protein (soy products, legumes, seed and nuts), protein foods (meat, poultry, seafood, and eggs), fatty acid ratio (PUFA+MUFA)/SFA)) and less than the median intake of adverse components (oils, empty energy (solid fat and sugar), sodium). The

ADA score ranges from 0-10.

Cardiovascular Disease Ascertainment

The primary outcome for this analysis is incident CVD, defined as the first occurrence of an acute myocardial infarction (MI) requiring overnight hospitalization, silent MI determined from serial ECGs, coronary heart disease (CHD) death, coronary revascularization, peripheral artery disease (PAD), congestive heart failure (CHF) or ischemic or hemorrhagic stroke. CHD

(defined in WHI as the first occurrence of a clinical MI, definite silent MI or a death due to definite or possible CHD) and stroke (ischemic and hemorrhagic) were also analyzed individually as separate outcomes. Cardiovascular disease outcomes were ascertained in the WHI through self-reported medical update questionnaires completed by participants every 6-12

149 months, depending on study assignment, and digital electrocardiograms (ECGs) acquired every 3 years (for CT participants only) and analyzed by a core laboratory. Medical records for all overnight hospitalizations and outpatient coronary revascularization procedures were reviewed by central physician adjudicators (for CHD death) or trained local adjudicators (for all other coronary end points), all blinded to treatment assignment.304,391,392

Covariates

At baseline, participants completed interview and self-administered standardized questionnaires to ascertain medical history, demographic and health behavior information, including age, race/ethnicity, education level, annual income, marital status, smoking history, and physical activity. Using the validated WHI physical activity questionnaire, participants reported the frequency, duration, and intensity of recreational physical activity, including walking, mild, moderate, and strenuous activity.393 From these data, metabolic equivalents of physical activities in MET-hours/week (kcal/week/kg) were computed, as previously described.394

In the medical history questionnaire, women were asked (yes/no), “Has a doctor told you that you have or have you had high cholesterol requiring pills?” Baseline hypertensive status was self-selected as “never hypertensive”, “untreated hypertensive”, or “treated hypertensive”.

Subjects were also asked, “Has a doctor ever told you that you had heart problems, problems with your blood circulation, or blood clots?” Previous validation studies have found self-report of CVD at baseline in the WHI to be reliable.395,396

At the baseline clinic visit, trained and certified WHI clinical staff measured height, weight, and blood pressure using standardized procedures. Weight was measured using a calibrated balance-beam scale and height using a fixed stadiometer. From these measurements,

150 body mass index (BMI) was calculated as weight in kilograms divided by height in meters squared. Blood pressure was measured twice after a 5-minute rest period using a conventional mercury sphygmomanometer and appropriately sized cuffs. Baseline systolic and diastolic blood pressure were recorded as the average of the two measurements.397

Statistical Analysis

Baseline characteristics were described by overall study population and quintile of dietary pattern score using means with standard deviation for continuous variables and frequencies with percentages for categorical variables. To compare baseline characteristics, chi- squared tests were used for categorical variables and analysis of variance for continuous variables. Participants had low rates of missing data for pertinent covariates (< 4%). Missing values were imputed by age and race sub-group to the median value for continuous variables

(BMI and physical activity), “no” for dichotomous variables (insulin use and high cholesterol), and the most frequent categorical value for polychotomous variables (race, education, income, marriage, alcohol intake, smoking, and hypertensive status).398 A complete case analysis was done to validate this approach and produced results similar to the main findings (data not shown).

Cox proportional hazards models were used to evaluate the association between diet quality and incident CVD. Separate models were fit for each dietary pattern score to estimate the hazard ratios (HRs) of incident CVD. The time to event was measured as the number of days since enrollment to the first occurrence of a cardiovascular event. Otherwise, participants were censored at the time of a woman’s last documented follow-up contact, whether due to loss of follow-up, death (non-cardiovascular), or end of study. For the analyses, participants were ranked into quintiles of dietary pattern score, with the lowest quintile of each score serving as the

151 reference group. Two multivariate models adjusted for preselected potential confounders were used. Model 1 was adjusted for age, race/ethnicity, education, income, marital status, physical activity, cigarette smoking, BMI, geographical region and WHI study arm. Model 2 was additionally adjusted for age at diabetes diagnosis, energy intake, insulin use, blood pressure and high cholesterol. Laboratory triglyceride and cholesterol measurements were collected on only

6% of CT participants and 1% of OS participants; thus, measured blood pressure and reported hypercholesterolemia status were used to account for baseline CVD risk.

For the DASH and ADA dietary pattern score analyses, both models were also adjusted for alcohol intake since alcohol is not part of the score. To assess the nature of the dietary pattern-CVD risk relationship, the score was entered as a continuous variable in the model to test for trend. The same analyses were conducted for the major types of CVD in the cohort, incident

CHD and stroke, to inform the interpretation of the results. The proportional hazards assumption was tested by including an interaction term with log (base-e) transformed time for each covariate. In the CVD models, there was evidence that the covariates smoking and blood pressure violated the proportional hazards assumption. In the stroke models, the covariates age at diabetes onset, income, and hormone therapy study participation violated the assumption. To address this, the time-dependent interaction terms were added to the fully adjusted models for the non-proportional covariates. No violations were found in the CHD analysis.

Analyses were conducted to test for potential effect measure modification by race, BMI, duration of diabetes, insulin use and smoking. This was done by separately including a term for the interaction between the variable of interest and each dietary pattern score in the fully adjusted models as well as stratified analyses, when subgroups contained adequate sample sizes. To address potential misclassification of T2D with a cutoff of 21 years of age at diagnosis, a

152 sensitivity analysis was conducted by repeating all the analyses using only women who reported an age at diabetes diagnosis ≥ 30 years. Another sensitivity analysis excluding incident cases of

CVD within the first 3 years of follow-up was also done to address possible reverse causation.

CHD and stroke analyses were replicated using less stringent exclusion criteria, excluding only women with a self-report history of MI, coronary artery bypass graft (CABG), or percutaneous transluminal coronary angioplasty (PTCA) in the CHD analysis (n = 8,154) and stroke or transient ischemic attack (TIA) in the stroke analysis (n = 8,463). All analyses were performed using SAS 9.4 (SAS Institute Inc., Cary, NC).

Results The 5,809 women in this sample had a mean (SD) age of 64.0 (6.9) years and mean BMI of 31.9 (6.8) kg/m2. Approximately 33.1% of the women were from racial/ethnic groups other than White, with Blacks representing the largest of these groups (20.6%). Only 6.5% of the women were current smokers at baseline, 10.2% reported untreated hypertension while 48.9% reported treated hypertension, and 21.4% reported a history of hypercholesterolemia requiring medication.

Comparison among quintiles of dietary pattern scores showed significant differences in baseline characteristics as presented in Table 1. Women in the upper quintiles of the DASH, aMed and ADA dietary pattern scores were older, more likely to be White or Asian/Pacific

Islander, had higher income, education, and physical activity levels, were more likely to be in the

OS arm, less likely to be current smokers and had lower mean BMI compared with those in the lowest quintile of these scores. There was no significant difference among the quintiles for insulin use, hypertensive status or hypercholesterolemia. Women in the upper quintile of the

Paleo dietary pattern score were older, more like to be Black or Asian/Pacific Islander, had higher education, income and physical activity levels, were more likely to use insulin and have a

153 history of hypercholesterolemia, less likely to be current smokers and had a lower mean BMI.

The dietary patterns under study had modest, positive correlations as shown in Table 2.

Descriptive characteristics of estimated nutrient and food group intakes by quintile of dietary pattern score are presented in Supplemental Tables 1-4.

A total of 1,454 (25.0%) incident cases of CVD were documented during the follow-up period (mean (SD) 10.9 (5.9) years). Table 3 presents the results of the analyses of the association between the DASH, aMed, ADA and Paleo dietary patterns and CVD risk. Higher

DASH, aMed and ADA diet scores were all associated with a reduced risk of CVD over the follow-up time. In both multivariate models, compared with women in the lowest quintiles, women in the highest quintile of the DASH and ADA scores had a reduced risk of CVD, with adjusted HRs of 0.69 (95% CI 0.58-0.83) and 0.83 (95% CI 0.69-0.99), respectively. Compared with the lowest quintile, women in the upper two quintiles of the aMed score had a 19-23% reduced risk of developing CVD during follow-up (4th quintile HR 0.81; 95% CI 0.68-0.97; 5th quintile HR 0.77; 95% CI 0.64-0.92). There was no association between the Paleo diet score and

CVD risk (HR 0.90; 95% CI 0.74-1.08, comparing Q5 v. Q1). Results of a supplemental analysis examining the association by standard deviation of the respective scores showed a 10% risk reduction for every standard deviation increase of the DASH score (HR 0.90; 95% CI 0.85-0.95),

7% risk reduction for the aMed score (HR 0.93; 95% CI 0.87-0.98) and 8% risk reduction for the

ADA score (HR 0.92; 95% CI 0.87-0.98). The Paleo diet score was not associated with total

CVD in this analysis (HR 0.96, 95% CI 0.91-1.03).

Table 3 also presents the results for CHD and stroke as separate outcomes. During the follow-up period, 635 (10.9%) women experienced incident CHD and 372 (6.4%) women had a stroke. A strong inverse relationship was found between higher DASH, aMed and ADA-based

154 diet scores and risk for CHD and stroke. Compared with women in the lowest quintile, women in the highest quintile of the DASH, aMed, and ADA scores had a 25%, 31% and 27% reduced risk of incident CHD and 44%, 33%, and 35% reduced risk of stroke, respectively. Similar to the findings for total CVD, risk for CHD and stroke did not differ among quintiles of the Paleolithic score (CHD HR 1.04; 95% CI 0.78-1.39 and stroke HR 0.84; 95% CI 0.58-1.2 when comparing the highest quintile with the lowest). Results for both CHD and stroke were similar when the analysis was replicated using a larger sample of women where the exclusion criteria included only a history of CHD or stroke, respectively, instead of total CVD at baseline.

To address potential reverse causality, I performed a sensitivity analysis excluding incident CVD cases documented within the first three years of follow-up. The strength and direction of the results were not materially different from the results of the main analysis. An additional sensitivity analysis was performed to address misclassification bias by including only women who reported an age of onset of diabetes ≥ 30 years, instead of the original cutoff of 21 years. These results were also similar in nature to those of the main analysis. Stratified analyses, as well as formal tests for interaction, did not show any effect measure modification for dietary pattern scores by race, BMI, duration of diabetes, insulin use, or smoking status.

Discussion

The study results demonstrate that higher aMed, DASH, and ADA dietary pattern diet quality scores are associated with lower risk of incident CVD over an average of 12.4 years of follow-up in postmenopausal women with T2D. These relationships were consistent for both

CHD and stroke separately as outcomes. Contrary to the study hypothesis, no association was observed between a Paleo diet score and CVD risk.

155 There were also important nuances in the individual pattern-CVD associations to note.

The association between the aMed score and CVD risk was monotonic in nature. The DASH diet displayed a strong inverse association with CVD risk between Q5 and Q1, but the risks of Q2-Q4 were similar to each other. The risks of CVD were higher in the 2nd and 3rd quintiles of the ADA diet score compared with the lowest quintile, while lower risks were observed for the 4th and 5th quintiles, suggesting residual confounding may be present. Future studies in other populations with large sample sizes and more complete measures of potential confounders are warranted to elucidate whether the relationship between these diet pattern and CVD risk are indeed nonlinear.

The common thread among diet quality scores and CVD risk in this population is an emphasis on higher intake of fruits, vegetables, whole grains, nuts/seeds, legumes, a high unsaturated:saturated fat ratio and lower intake of red and processed meats, added sugars and sodium. These findings are consistent with previous investigations of diet-CVD risk in the general population without diabetes.145,283,399 However, the evidence examining this relationship in populations with T2D is lacking, save for several cross-sectional studies which report an association between healthful dietary patterns and improved CVD risk factors in populations of

Korean,285 Japanese,286 and Iranian287 adults with T2D. In 2011, a cross-sectional study by

Huffman, et al.400 found an inverse association between diet quality, measured by the Alternative

Healthy Eating Index (AHEI), and calculated 10-year CHD risk in a sample of Cuban Americans with T2D (p = 0.001). Using a nested case-control study design, Ciccarone, et al.288 found a reduced risk of PAD among Italian adults with T2D in the upper tertile of Mediterranean diet score compared with those in the lowest tertile (OR 0.44; 95% CI 0.24-0.83). Evidence from intervention studies is limited to a randomized trial by Elhayany, et al.235 that compared the effects of a low-carbohydrate Mediterranean, traditional Mediterranean and ADA diet on CVD

156 risk factors and glycemic control specifically in a population with T2D. The authors found that an intensive 12-month dietary intervention in a community-based setting was effective for weight loss and reducing hemoglobin A1c, triglyceride, total- and LDL-cholesterol levels in all dietary groups. An increase in HDL-cholesterol levels was observed only in those randomized to the low-carbohydrate Mediterranean diet235.

In addition to the paucity of studies on dietary patterns and CVD risk in individuals with

T2D, a major limitation to the current body of evidence is the reliance on clinical CVD risk factors instead of hard endpoints as outcomes. Several well-conducted intervention trials have demonstrated the effectiveness of a Mediterranean diet for improved glycemic control and CVD risk reduction using clinical outcomes in high risk individuals.399 For instance, the Lyon Diet

Heart Study found a reduced risk for the recurrence of CVD complications following a first MI in individuals randomized to a Mediterranean diet compared with a prudent Western-type diet

(adjusted HR 0.63; 95% CI 0.46-0.87).267 The PREDIMED study tested the effect of a

Mediterranean diet supplemented with nuts or extra virgin olive oil (EVOO) compared with a standard low-fat diet on the primary prevention of CVD in a sample of 7,447 high-risk individuals, approximately 50% (3,614) of whom had T2D at baseline.276 After 1 year of follow up, both Mediterranean diets showed positive effects on intermediate markers of CVD risk, including blood pressure, lipid profiles, lipoprotein particles, inflammation, oxidative stress and the expression of pro-atherogenic genes involved in vascular events and thrombosis.278 After approximately 5 years of follow-up, compared with the control diet, the Mediterranean diets showed a protective effect for major CVD events (EVOO HR 0.70, 95% CI 0.54-0.92; nuts HR

0.72, 95% CI 0.54-0.96) in multivariate analyses. Results of a sub-group analysis including only subjects with T2D were consistent with the main study findings (HR 0.71; 95% CI 0.53-0.96).277

157 Although these studies were not done exclusively in populations with T2D, the use of clinical events as the main outcomes, as opposed to CVD risk factors, provides a strong evidence base for the efficacy of the Mediterranean diet for risk reduction in high-risk populations.

Contrary to my hypothesis, the Paleo diet, which emphasizes higher intake of lean meats, fish, vegetables, eggs and nuts and deemphasizes grains, legumes, diary, salt, refined sugar and processed oils314 was not associated with CVD risk in this population. Although a Paleo diet was not a mainstream dietary trend in the mid 1990’s, the calculated scores ranged from 19-60 and followed a normal distribution. Proponents of the Paleo diet assert that humans are genetically adapted to foods assumed to have been available prior to the establishment of agriculture, mainly wild-animal and uncultivated-plant sources of foods, and that the incompatibility of core metabolic and physiological processes with the dietary changes introduced by modern food- producing practices may lead to chronic diseases such as obesity, T2D and CVD.329 To date, limited research on the Paleo diet has been reported in scientific literature. Results from a pooled cross-sectional study by Whalen, et al. showed that a Paleo -type dietary pattern may be associated with lower levels of systemic inflammation and oxidative stress in healthy adults389 and several pilot trials have provided evidence for improved metabolic and physiologic parameters with consumption of a Paleo diet in humans.326–328 The results of several small intervention trials have suggested a beneficial effect of a Paleo diet on glycemic control and

CVD risk in individuals with T2D.315,401,402 However, prospective observational evidence from larger cohorts to support these findings is lacking. In addition, the Paleo diets used in previous studies often had lower total energy, carbohydrate, glycemic index, and dietary glycemic load, among another nutrient differences, compared with control diets; limiting any inference about the potential biological mechanisms underlying the observed associations.

158 Unlike the other dietary pattern scores used in the present study, intake of whole grains, a constituent with noted positive health benefits,403,404 was unnecessary to achieve a high Paleo score. Previous studies have demonstrated the important role of carbohydrate quality in clinical risk reduction for individuals with T2D. Early work by Jenkins, et al. first identified specific components of dietary carbohydrates that can affect cardiometabolic parameters.238,239 In several randomized controlled trials, he later demonstrated the beneficial role of dietary fiber and glycemic index (GI) in diabetes management and CVD risk reduction in individuals with

T2D.262,263 Collectively, however, studies of dietary carbohydrate and fiber intake and CVD risk in individuals with T2D have produced inconclusive results.228,231 The precise nature of the relationship between dietary carbohydrate quality and quantity and CVD risk in individuals with

T2D remains unclear. With the addition of these findings, this body of evidence suggests that dietary carbohydrate quantity and quality may be important for both glycemic management and

CVD risk in diabetes since CHO exclusion is also demonstrably difficult to adhere to long term336 and further research is needed to clarify this association.

To my knowledge, this is the first prospective observational study to analyze the relationship between diet quality and clinical CVD endpoints in a population of individuals with

T2D. The inference of the results is strengthened by the rigorous study design, the racially diverse well-characterized cohort, and the adjudicated outcomes. There are several limitations to consider. First, measurement error and other potential biases might be present in self-reported diet, which has been demonstrated to most often lead to non-differential misclassification between diet and disease endpoints.132 In addition, I was unable to account for the potential effect of diet quality changes over time because only a small number of women had follow-up diet data. The women in this study were, on average, obese (based on mean BMI classification).

159 Both generalized and food-specific underreporting has been clearly demonstrated in obese populations under study.405 Since mean BMI tended to decrease across quintiles with increasing diet quality, reporting errors may have introduced systematic errors that led to an underestimation of the true effect size. Another limitation is the reliance on self-report for some of the lifestyle data and medical history such as smoking, physical activity, diabetes and cardiovascular history, which may have introduced misclassification and residual confounding in the analyses. Although diabetes status was validated, all prevalent cases may not have been captured without additional blood glucose measurements. Additionally, hemoglobin A1c was not measured in WHI so I was unable to adjust for baseline glycemic control. To address this, I included insulin use as a proxy for disease severity in the fully adjusted models. Insulin use did not differ substantially among dietary pattern score quintiles, suggesting that the diet quality score-CVD relationship was likely not confounded by underlying glycemic control. To account for baseline CVD risk, I used self-reported history of hypercholesterolemia as a proxy for clinical measurements of blood lipids, as these were measured in only a small sub-set of the study population. Since it is plausible that these clinical risk factors may also represent mediators of the observed associations, the availability of precisely measured clinical values may have improved estimates and provided further insight into the nature of the diet-CVD relationship. Since this analysis only included post-menopausal women with T2D, the findings may not be generalizable to men or younger adult populations with T2D. Finally, it is not possible to control for all confounding factors in observational studies, thus undetected residual confounding may have influenced the results.

In conclusion, dietary patterns emphasizing fruits, vegetables, whole grains, nuts/seeds, legumes, a high unsaturated:saturated fat ratio and low intake of red and processed meats, added

160 sugars and sodium was associated with a substantially reduced risk for CVD in postmenopausal women with T2D. Replication of this topic in other cohorts will provide further insight.

161 Tables Table 7.1a Baseline characteristics of WHI participants with type 2 diabetes according to DASH diet score quintile Quintile of DASH Diet Score P Characteristica Q1 Q2 Q3 Q4 Q5 valueb Number of participants 1257 1234 969 1305 1044 DASH diet score 18.9 (2.1) 23.1 (0.8) 25.5 (0.5) 27.9 (0.8) 31.7 (1.7) <0.0001 Age (years) 62.0 (6.7) 63.4 (6.7) 64.6 (6.9) 64.7 (6.9) 65.5 (6.6) <0.0001 Race (%) American Indian/Alaska Native 36 (2.9) 15 (1.2) 8 (0.8) 9 (0.7) 2 (0.2) Asian/Pacific Islander 40 (3.2) 59 (4.8) 41 (4.2) 78 (6.0) 50 (4.8) Black 383 (30.5) 279 (22.6) 197 (20.3) 201 (15.4) 134 (12.8) <0.0001 Hispanic/Latino 136 (10.8) 117 (9.5) 44 (4.5) 64 (4.9) 31 (3.0) White 648 (51.6) 742 (60.1) 669 (69.0) 934 (71.6) 813 (77.9) Other 14 (1.1) 22 (1.8) 10 (1.0) 19 (1.5) 14 (1.3) Education (%) < High school 230 (18.3) 131 (10.6) 82 (8.5) 95 (7.3) 48 (4.6) High school/GED 309 (24.6) 296 (24.0) 184 (19.0) 252 (19.3) 169 (16.2) <0.0001 > High school, < 4 y college 491 (39.1) 520 (42.1) 431 (44.5) 532 (40.8) 380 (36.4) ≥ 4 y college 227 (18.1) 287 (23.3) 272 (28.1) 426 (32.6) 447 (42.8) Smoking status (%) Never 655 (52.1) 687 (55.7) 528 (54.5) 703 (53.9) 573 (54.9) Past 469 (37.3) 450 (36.5) 395 (40.8) 540 (41.4) 433 (41.5) <0.0001 Current 133 (10.6) 97 (7.9) 46 (4.8) 62 (4.8) 38 (3.6) Income ($) (%) <10,000 185 (14.7) 106 (8.6) 71 (7.3) 80 (6.1) 54 (5.2) 10-34,999 589 (46.9) 558 (45.2) 468 (48.3) 567 (43.5) 462 (44.3) 35-74,999 355 (28.2) 416 (33.7) 317 (32.7) 482 (36.9) 382 (36.6) <0.0001 ≥75,000 64 (5.1) 103 (8.4) 87 (9.0) 135 (10.3) 119 (11.4) Unknown 46 (5.1) 51 (4.1) 26 (2.7) 41 (3.1) 27 (2.6) Marital Status (%) Never married 62 (4.9) 57 (4.6) 50 (5.7) 50 (3.8) 74 (7.1) Divorced/separated 253 (20.1) 199 (16.1) 172 (17.8) 190 (14.6) 182 (17.4) 0.0007 Widowed 254 (20.2) 279 (22.6) 288 (23.5) 293 (22.5) 212 (20.3) Married/committed 688 (54.7) 699 (56.7) 519 (53.6) 772 (59.2) 576 (55.2) relationship Clinical Trial Participation (%) 325 (25.9) 257 (20.8) 194 (20.0) 264 (20.2) 220 (21.1) 0.002 HRT 478 (38.0) 459 (37.2) 355 (36.6) 368 (28.2) 199 (19.1) <0.0001 DM 324 (25.9) 317 (25.7) 235 (24.3) 311 (23.8) 199 (19.1) 0.001 CAD Observational Study (%) 581 (46.2) 607 (49.2) 504 (52.0) 736 (56.4) 662 (63.4) <0.0001 BMI (kg/m2) 33.6 (6.9) 32.4 (6.7) 31.9 (6.1) 31.3 (6.6) 30.3 (6.9) <0.0001 Alcohol (servings)c (%) Non-drinker 231 (18.4) 216 (17.5) 157 (16.2) 222 (17.0) 169 (16.2) Past-drinker 488 (38.8) 464 (37.6) 376 (38.8) 475 (36.4) 408 (39.1) < 1/month 186 (14.8) 193 (15.6) 126 (13.0) 168 (12.9) 127 (12.2) 0.10 < 1/week 185 (14.7) 200 (16.2) 155 (16.0) 234 (17.9) 182 (17.4) 1-6/week 116 (9.2) 114 (9.2) 108 (11.2) 153 (11.7) 126 (12.1) ≥ 7/week 51 (4.1) 47 (3.8) 47 (4.9) 53 (4.1) 32 (3.1) Physical Activity 5.9 (9.0) 8.6 (11.8) 9.5 (11.9) 10.8 (12.4) 13.9 (14.0) <0.0001 (MET-hours/week) Age at T2D diagnosis (%)

162 21-39 years 132 (10.5) 113 (9.2) 78 (8.1) 102 (7.8) 66 (6.3) 40-59 years 770 (61.3) 697 (56.5) 513 (52.9) 678 (52.0) 506 (48.5) <0.0001 ≥ 60 years 355 (28.2) 424 (34.4) 378 (39.0) 525 (40.2) 472 (45.2) Insulin Shots (% ever used) 326 (25.9) 311 (25.2) 224 (23.1) 316 (24.2) 241 (23.1) 0.42 Hypertension (%) Never hypertensive 499 (39.7) 500 (40.5) 374 (38.6) 556 (42.6) 449 (43.0) Untreated hypertensive 126 (10.0) 114 (9.2) 104 (10.7) 132 (10.1) 115 (11.0) 0.26 Treated hypertensive 632 (50.3) 620 (50.2) 491 (50.7) 617 (47.3) 480 (46.0) Hx hypercholesterolemia 256 (20.4) 260 (21.1) 210 (21.7) 286 (21.9) 229 (21.9) 0.86 requiring medication (%) a Data are unadjusted means (SD) unless noted as percentage, then frequency and corresponding percentage based on the number of women in each quintile are shown. b P value is based on Chi-squared for categorical variables and analysis of variance for continuous variables. c Number of servings per week of beer, wine and/or liquor based on a medium serving size of 12 oz beer, 6 oz wine and 1.5 oz liquor. Abbreviations: GED, general education development; HRT, hormone replacement therapy; DM, dietary modification trial; CAD, calcium and vitamin D trial; BMI, body mass index (calculated as weight in kilograms divided by height in meters squared); MET, metabolic equivalents; MET- hours/week (kcal/week per kilogram body weight), energy expenditure from recreational activity (includes walking, mild, moderate and strenuous physical activity).

163

Table 7.1b Baseline characteristics of WHI participants with type 2 diabetes according to Mediterranean (aMed) diet score quintile Quintile of aMed Diet Score P Characteristica Q1 Q2 Q3 Q4 Q5 valueb Number of participants 1202 1141 1211 1106 1149 aMed diet score 1.6 (0.6) 3.0 (0.0) 4.0 (0.0) 5.0 (0.0) 6.5 (0.7) <0.0001 Age (years) 63.3 (6.9) 64.0 (7.0) 64.2 (6.8) 64.0 (6.9) 64.4 (6.8) 0.0009 Race (%) American Indian/Alaska Native 30 (2.5) 15 (1.3) 10 (0.8) 13 (1.2) 2 (0.2) Asian/Pacific Islander 17 (1.4) 33 (2.9) 52 (4.3) 73 (6.6) 93 (8.1) Black 265 (22.2) 264 (23.1) 239 (19.7) 208 (18.8) 218 (19.0) <0.0001 Hispanic/Latino 130 (10.8) 90 (7.9) 78 (6.4) 53 (4.8) 41 (3.6) White 743 (61.8) 727 (63.7) 820 (67.7) 741 (67.0) 775 (67.5) Other 17 (1.4) 12 (1.1) 12 (1.0) 18 (1.6) 20 (1.7) Education (%) < High school 180 (15.0) 136 (11.9) 120 (9.9) 86 (7.8) 64 (5.6) High school/GED 311 (25.9) 279 (24.5) 245 (20.2) 207 (18.7) 168 (14.6) <0.0001 > High school, < 4 y college 496 (41.3) 471 (41.3) 487 (40.2) 440 (39.8) 460 (40.0) ≥ 4 y college 215 (17.9) 255 (22.4) 359 (29.6) 373 (33.7) 457 (39.8) Smoking status (%) Never 652 (54.2) 623 (54.6) 631 (52.11) 618 (55.9) 622 (54.1) Past 443 (36.9) 431 (37.78) 500 (41.3) 437 (39.5) 476 (41.4) <0.0001 Current 107 (8.9) 87 (7.6) 80 (6.6) 51 (4.6) 51 (4.4) Income ($) (%) <10,000 150 (12.5) 103 (9.0) 101 (8.3) 78 (7.1) 64 (5.6) 10-34,999 581 (48.3) 570 (50.0) 539 (44.5) 496 (44.9) 458 (39.9) 35-74,999 351 (29.2) 340 (29.8) 417 (34.4) 386 (34.9) 458 (39.9) <0.0001 ≥75,000 69 (5.7) 79 (6.9) 120 (9.9) 110 (10.0) 130 (11.3) Unknown 51 (4.2) 49 (4.3) 34 (2.8) 36 (3.3) 39 (3.4) Marital Status (%)

Never married 76 (6.6) 62 (5.2) 43 (3.8) 50 (4.2) 62 (5.6) Divorced/separated 179 (15.6) 0.009 215 (17.9) 209 (18.3) 190 (15.7) 203 (18.4) Widowed 235 (20.5) 285 (23.7) 258 (22.6) 273 (22.5) 215 (19.4) Married/committed 659 (57.4) 640 (53.2) 631 (55.3) 698 (57.6) 626 (56.6) relationship Clinical Trial Participation

(%) 292 (24.3) 257 (22.5) 240 (19.8) 230 (20.8) 241 (21.0) 0.07 HRT 421 (35.0) 368 (32.3) 399 (33.0) 332 (30.0) 339 (29.5) 0.03 DM 295 (24.5) 285 (25.0) 288 (23.8) 247 (22.3) 271 (23.6) 0.63 CAD Observational Study (%) 591 (49.2) 612 (53.6) 641 (52.9) 617 (55.8) 629 (54.7) 0.017 BMI (kg/m2) 32.8 (6.6) 32.3 (6.7) 31.8 (6.8) 31.7 (6.9) 31.0 (6.7) <0.0001 Alcohol (servings)c (%) Non-drinker 223 (18.6) 211 (18.5) 203 (16.8) 179 (16.2) 179 (15.6) Past-drinker 502 (41.8) 448 (39.3) 457 (37.7) 415 (37.5) 389 (33.9) < 1/month 189 (15.7) 155 (13.6) 174 (14.4) 141 (12.8) 141 (12.3) <0.0001 < 1/week 167 (13.9) 194 (17.0) 206 (17.0) 189 (17.1) 200 (17.4) 1-6/week 75 (6.2) 94 (8.2) 120 (9.9) 126 (11.4) 202 (17.6) ≥ 7/week 46 (3.8) 39 (3.4) 51 (4.2) 56 (5.1) 38 (3.3) Physical Activity 6.8 (9.5) 8.2 (11.3) 10.1 (13.3) 10.9 (13.1) 12.1 (12.4) <0.0001 (MET-hours/week) Age at T2D diagnosis (%) 21-39 years 106 (8.8) 108 (9.5) 110 (9.1) 88 (8.0) 79 (6.9)

164 40-59 years 709 (59.0) 607 (53.2) 658 (54.3) 588 (53.2) 602 (52.4) 0.002 ≥ 60 years 387 (32.2) 426 (37.3) 443 (36.6) 430 (38.9) 468 (40.7) Insulin Shots (% ever used) 309 (25.7) 292 (25.6) 299 (24.7) 247 (22.3) 271 (23.6) 0.29 Hypertension (%) Never hypertensive 470 (39.1) 462 (40.5) 514 (42.4) 455 (41.1) 477 (41.5) Untreated hypertensive 133 (11.1) 112 (9.8) 123 (10.2) 109 (9.9) 114 (9.9) 0.85 Treated hypertensive 599 (49.8) 567 (46.7) 574 (47.4) 542 (49.0) 558 (48.6) Hx hypercholesterolemia 236 (19.6) 258 (22.6) 266 (22.0) 240 (21.7) 241 (21.0) 0.46 requiring medication (%) a Data are unadjusted means (SD) unless noted as percentage, then frequency and corresponding percentage based on the number of women in each quintile are shown. b P value is based on Chi-squared for categorical variables and analysis of variance for continuous variables. c Number of servings per week of beer, wine and/or liquor based on a medium serving size of 12 oz beer, 6 oz wine and 1.5 oz liquor. Abbreviations: GED, general education development; HRT, hormone replacement therapy; DM, dietary modification trial; CAD, calcium and vitamin D trial; BMI, body mass index (calculated as weight in kilograms divided by height in meters squared); MET, metabolic equivalents; MET- hours/week (kcal/week per kilogram body weight), energy expenditure from recreational activity (includes walking, mild, moderate and strenuous physical activity).

165 Table 7.1c Baseline characteristics of WHI participants with type 2 diabetes according to American Diabetes Association (ADA) diet score quintile Quintile of ADA Diet Score Characteristica Q1 Q2 Q3 Q4 Q5 P valueb Number of participants 1042 1208 1351 1146 1062 ADA diet score 2.6 (0.6) 4.0 (0.0) 5.0 (0.0) 6.0 (0.0) 7.4 (0.6) <0.0001 Age (years) 63.1 (6.9) 63.1 (6.7) 64.2 (6.9) 64.4 (6.8) 65.2 (6.8) <0.0001 Race (%) American Indian/Alaska Native 24 (2.3) 21 (1.7) 10 (0.7) 11 (1.0) 4 (0.4) Asian/Pacific Islander 24 (2.4) 41 (3.4) 67 (5.0) 57 (5.0) 78 (7.3) Black 253 (24.3) 267 (22.1) 295 (21.8) 215 (18.8) 164 (15.4) <0.0001 Hispanic/Latino 88 (8.5) 120 (9.9) 92 (6.8) 66 (5.8) 26 (2.5) White 640 (61.4) 739 (61.2) 869 (64.3) 781 (68.2) 777 (73.2) Other 12 (1.2) 20 (1.7) 18 (1.3) 16 (1.4) 13 (1.2) Education (%) < High school 147 (14.1) 154 (12.8) 141 (10.4) 75 (6.5) 69 (6.5) High school/GED 272 (26.1) 269 (22.3) 273 (20.2) 203 (17.7) 193 (18.2) <0.0001 > High school, < 4 y college 414 (39.7) 504 (41.7) 560 (41.5) 483 (42.2) 393 (37.0) ≥ 4 y college 209 (20.1) 281 (23.3) 377 (27.9) 385 (33.6) 407 (38.3) Smoking status (%) Never 522 (50.1) 637 (52.7) 750 (55.5) 645 (56.3) 592 (55.7) Past 410 (39.4) 469 (38.8) 523 (38.7) 462 (40.3) 423 (39.8) <0.0001 Current 110 (10.6) 102 (8.4) 78 (5.8) 39 (3.4) 47 (4.4) Income ($) (%) <10,000 117 (11.2) 123 (10.2) 121 (9.0) 70 (6.1) 65 (6.1) 10-34,999 505 (48.5) 559 (46.3) 607 (44.9) 515 (44.9) 458 (43.1) 35-74,999 316 (30.3) 385 (31.9) 444 (32.9) 432 (37.7) 375 (35.3) <0.0001 ≥75,000 67 (6.4) 90 (7.5) 124 (9.2) 96 (8.4) 131 (12.3) Unknown 37 (3.6) 51 (4.2) 55 (4.1) 33 (2.9) 33 (3.1) Marital Status (%)

Never married 47 (4.5) 55 (4.6) 70 (5.2) 53 (4.6) 68 (6.4) Divorced/separated 203 (19.5) 222 (18.4) 228 (16.9) 170 (14.4) 173 (16.3) Widowed 0.07 242 (23.2) 256 (21.2) 277 (20.5) 260 (22.7) 231 (21.8) Married/committed 550 (52.8) 675 (55.9) 776 (57.4) 663 (57.9) 590 (55.6) relationship Clinical Trial Participation

(%) 230 (22.1) 262 (21.7) 293 (21.7) 238 (20.77) 237 (22.3) 0.92 HRT 420 (59.7) 442 (36.6) 468 (34.6) 316 (27.57) 213 (20.1) <0.0001 DM 267 (25.6) 304 (25.2) 337 (24.9) 270 (23.56) 208 (19.6) 0.006 CAD Observational Study (%) 476 (45.7) 609 (50.4) 700 (51.8) 656 (57.24) 649 (61.1) <0.0001 BMI (kg/m2) 33.3 (7.2) 32.6 (6.5) 31.9 (6.6) 31.5 (6.7) 30.4 (6.6) <0.0001 Alcohol (servings)c (%) 257 (19.0) Non-drinker 163 (15.6) 190 (15.7) 202 (17.6) 183 (17.2) 491 (36.3) Past-drinker 422 (40.5) 459 (38.0) 445 (38.8) 394 (37.1) 175 (13.0) < 1/month 149 (14.3) 181 (15.0) 150 (13.1) 145 (13.7) 0.04 231 (17.1) < 1/week 158 (15.2) 199 (16.5) 197 (17.2) 171 (16.1) 154 (11.4) 1-6/week 90 (8.6) 134 (11.1) 111 (9.7) 128 (12.1) 43 (3.2) ≥ 7/week 60 (5.8) 45 (3.7) 41 (3.6) 41 (3.9) Physical Activity 6.5 (9.5) 8.2 (11.3) 9.3 (11.9) 11.3 (13.4) 12.8 (13.2) <0.0001 (MET-hours/week) Age at T2D diagnosis (%) 21-39 years 98 (9.4) 118 (9.8) 113 (8.4) 88 (7.7) 74 (7.0) 40-59 years 605 (58.1) 698 (57.8) 729 (54.0) 594 (51.8) 538 (50.7) <0.0001

166 ≥ 60 years 339 (32.5) 392 (32.5) 509 (37.7) 464 (40.5) 450 (42.4) Insulin Shots (% ever used) 249 (23.9) 306 (25.3) 326 (24.1) 284 (24.8) 253 (23.8) 0.90 Hypertension (%) Never hypertensive 397 (38.1) 490 (40.6) 553 (40.9) 483 (42.2) 455 (42.8) Untreated hypertensive 99 (9.5) 128 (10.6) 139 (10.3) 108 (9.4) 117 (11.0) 0.28 Treated hypertensive 546 (52.4) 590 (48.8) 659 (48.8) 555 (48.4) 490 (46.1) Hx hypercholesterolemia 193 (18.5) 258 (21.4) 293 (21.7) 243 (21.2) 254 (23.9) 0.06 requiring medication (%) a Data are unadjusted means (SD) unless noted as percentage, then frequency and corresponding percentage based on the number of women in each quintile are shown. b P value is based on Chi-squared for categorical variables and analysis of variance for continuous variables. c Number of servings per week of beer, wine and/or liquor based on a medium serving size of 12 oz beer, 6 oz wine and 1.5 oz liquor. Abbreviations: GED, general education development; HRT, hormone replacement therapy; DM, dietary modification trial; CAD, calcium and vitamin D trial; BMI, body mass index (calculated as weight in kilograms divided by height in meters squared); MET, metabolic equivalents; MET- hours/week (kcal/week per kilogram body weight), energy expenditure from recreational activity (includes walking, mild, moderate and strenuous physical activity).

167 Table 7.1d Baseline characteristics of WHI participants with type 2 diabetes according to Paleolithic diet score quintile Quintile of Paleo Diet Score P Characteristica Q1 Q2 Q3 Q4 Q5 valueb Number of participants 1237 1031 1123 1323 1095 Paleo diet score 32.0 (2.8) 37.1 (0.8) 40.0 (0.8) 43.5 (1.1) 48.6 (2.5) <0.0001 Age (years) 62.8 (6.8) 63.3 (7.0) 64.1 (6.9) 64.7 (6.8) 64.9 (6.6) <0.0001 Race (%) American Indian/Alaska Native 29 (2.3) 14 (1.4) 6 (0.5) 10 (0.8) 11 (1.0) Asian/Pacific Islander 17 (1.4) 35 (3.4) 56 (5.0) 76 (5.7) 84 (7.7) Black 222 (18.0) 184 (17.9) 226 (20.1) 283 (21.4) 279 (25.5) <0.0001 Hispanic/Latino 133 (10.8) 65 (6.3) 83 (7.4) 72 (5.4) 39 (3.6) White 824 (66.6) 722 (70.0) 735 (65.5) 860 (65.0) 665 (60.7) Other 12 (1.0) 11 (1.1) 17 (1.5) 22 (1.7) 17 (1.6) Education (%) < High school 172 (13.9) 95 (9.2) 98 (8.7) 127 (9.6) 94 (8.6) High school/GED 288 (23.3) 220 (21.3) 248 (22.1) 274 (20.7) 180 (16.4) <0.0001 > High school, < 4 y college 509 (41.2) 452 (43.8) 437 (38.9) 534 (40.4) 222 (38.5) ≥ 4 y college 268 (21.7) 264 (25.6) 340 (30.3) 388 (29.3) 399 (36.4) Smoking status (%) Never 619 (50.0) 535 (51.9) 612 (54.5) 769 (58.1) 611 (55.8) Past 503 (40.7) 430 (41.7) 441 (39.3) 485 (36.7) 428 (39.1) <0.0001 Current 115 (9.3) 66 (6.4) 70 (6.2) 69 (5.2) 56 (5.1) Income ($) (%) <10,000 117 (11.2) 123 (10.2) 121 (9.0) 70 (6.1) 65 (6.1) 10-34,999 505 (48.5) 559 (46.3) 607 (44.9) 515 (44.9) 458 (43.1) 35-74,999 316 (30.3) 385 (31.9) 444 (32.9) 432 (37.7) 375 (35.3) 0.0009 ≥75,000 67 (6.4) 90 (7.5) 124 (9.2) 96 (8.4) 131 (12.3) Unknown 37 (3.6) 51 (4.2) 55 (4.1) 33 (2.9) 33 (3.1) Marital Status (%)

Never married 68 (6.2) 58 (4.7) 48 (4.7) 61 (5.4) 58 (4.4) Divorced/separated 181 (16.5) 0.12 243 (19.6) 178 (17.3) 187 (16.7) 207 (15.7) Widowed 245 (22.4) 267 (21.6) 206 (20.0) 236 (21.0) 312 (23.6) Married/committed 601 (54.9) 669 (54.1) 599 (58.1) 639 (56.9) 746 (56.4) relationship Clinical Trial Participation

(%) 216 (19.8) 0.11 289 (23.4) 220 (21.3) 263 (23.4) 272 (20.6) HRT 248 (22.7) <0.0001 484 (39.1) 360 (34.9) 395 (35.2) 372 (28.1) DM 210 (19.2) <0.0001 337 (27.2) 242 (23.5) 295 (26.3) 302 (22.8) CAD Observational Study (%) 577 (46.7) 535 (51.9) 552 (49.2) 757 (57.2) 669 (61.1) <0.0001 BMI (kg/m2) 33.3 (6.9) 32.5 (6.7) 31.7 (6.5) 31.3 (6.6) 30.8 (6.9) <0.0001 Alcohol (servings)c (%) Non-drinker 166 (13.4) 149 (14.5) 194 (17.3) 267 (20.2) 219 (20.0) Past-drinker 390 (31.5) 366 (35.5) 402 (35.8) 536 (40.5) 517 (47.2) < 1/month 181 (14.6) 153 (14.8) 154 (13.7) 152 (11.5) 160 (14.6) <0.0001 < 1/week 255 (20.6) 184 (17.9) 191 (17.0) 218 (16.5) 108 (9.9) 1-6/week 174 (14.1) 128 (12.4) 134 (11.9) 123 (9.3) 58 (5.3) ≥ 7/week 71 (5.7) 51 (4.5) 48 (4.3) 27 (2.0) 33 (3.0) Physical Activity 7.3 (10.8) 8.3 (11.2) 9.7 (12.3) 9.7 (11.3) 13.2 (14.2) <0.0001 (MET-hours/week) Age at T2D diagnosis (%) 21-39 years 117 (9.5) 90 (8.7) 98 (8.7) 100 (7.6) 86 (7.6) 40-59 years 700 (56.6) 578 (56.1) 603 (53.7) 700 (52.9) 583 (53.2) 0.10

168 ≥ 60 years 420 (34.0) 363 (35.2) 422 (37.6) 523 (39.5) 426 (38.9) Insulin Shots (% ever used) 251 (20.3) 253 (24.5) 300 (26.7) 331 (25.0) 283 (25.8) 0.003 Hypertension (%) Never hypertensive 501 (40.5) 420 (40.7) 480 (42.7) 519 (39.2) 458 (41.8) Untreated hypertensive 126 (10.2) 110 (10.7) 114 (10.2) 133 (10.1) 108 (9.9) 0.83 Treated hypertensive 610 (49.3) 501 (48.6) 529 (47.1) 671 (50.7) 529 (48.3) Hx hypercholesterolemia 216 (17.5) 226 (21.9) 236 (21.0) 313 (23.66) 250 (22.8) 0.002 requiring medication (%) a Data are unadjusted means (SD) unless noted as percentage, then frequency and corresponding percentage based on the number of women in each quintile are shown. b P value is based on Chi-squared for categorical variables and analysis of variance for continuous variables. c Number of servings per week of beer, wine and/or liquor based on a medium serving size of 12 oz beer, 6 oz wine and 1.5 oz liquor. Abbreviations: GED, general education development; HRT, hormone replacement therapy; DM, dietary modification trial; CAD, calcium and vitamin D trial; BMI, body mass index (calculated as weight in kilograms divided by height in meters squared); MET, metabolic equivalents; MET- hours/week (kcal/week per kilogram body weight), energy expenditure from recreational activity (includes walking, mild, moderate and strenuous physical activity).

Table 7.2 Pearson’s correlations between dietary pattern scores. DASH aMed ADA Paleo DASH 1.00 0.63 0.70 0.50 aMed 0.63 1.00 0.64 0.33 ADA 0.70 0.64 1.00 0.54 Paleo 0.50 0.33 0.54 1.00 P < 0.0001 for all correlations.

169 Table 7.3 Association between dietary patterns and incident cardiovascular disease, coronary heart disease, and stroke risk in postmenopausal women with T2D from WHI, 1993-2015. Dietary Pattern Quintile DASH Q1 Q2 Q3 Q4 Q5 P trend n CVD / person 340/4529908 298/4831603 243/383925 341/5180100 232/4442774 years Model 1 HR 1.0 0.87(0.74-1.01) 0.85(0.72-1.10) 0.90(0.77-1.06) 0.70(0.59-0.84) <0.001 (95% CI) [reference] Model 2 HR 1.0 0.85(0.72-0.99) 0.85(0.72-1.01) 0.87(0.75-1.02) 0.69(0.58-0.83) <0.001 (95% CI) [reference] n CHD / person 324/5145663 307/5287395 301/4245559 257/5730241 265/4819620 years Model 1 HR 1.0 0.93(0.73-1.18) 0.88(0.68-1.14) 0.91(0.71-1.16) 0.75(0.57-0.98) 0.02 (95% CI) [reference] Model 2 HR 1.0 0.93(0.73-1.19) 0.88(0.68-1.14) 0.89(0.69-1.13) 0.75(0.57-0.98) 0.01 (95% CI) [reference] n Stroke / 94/5191879 67/5363202 54/4300558 100/5759430 57/4886419 person years Model 1 HR 1.0 0.69(0.50-0.95) 0.64(0.45-0.90) 0.90(0.67-1.22) 0.57(0.40-0.80) 0.03 (95% CI) [reference] Model 2 HR 1.0 0.67(0.49-0.93) 0.63(0.45-0.90) 0.89(0.66-1.21) 0.56(0.40-0.80) 0.03 (95% CI) [reference] aMed n CVD / person 324/4322465 307/4425918 301/4787578 257/4438426 265/4849323 years Model 1 HR 1.0 0.93(0.79-1.09) 0.90(0.77-1.05) 0.84(0.71-0.99) 0.82(0.69-0.97) 0.02 (95% CI) [reference] Model 2 HR 1.0 0.90(0.77-1.05) 0.86(0.73-1.01) 0.81(0.68-0.97) 0.77(0.64-0.92) 0.005 (95% CI) [reference] n CHD / person 150/4823154 124/4967735 136/5267909 106/4883322 119/5286358 years Model 1 HR 1.0 0.78(0.61-0.99) 0.86(0.68-1.09) 0.72(0.56-0.93) 0.78(0.61-0.99) 0.07 (95% CI) [reference] Model 2 HR 1.0 0.75(0.59-0.95) 0.81(0.63-1.02) 0.68(0.52-0.88) 0.69(0.53-0.91) 0.01 (95% CI) [reference] n Stroke / 85/4894920 79/4961501 74/5396884 70/4911566 64/5336617 person years Model 1 HR 1.0 0.89(0.65-1.21) 0.79(0.57-1.08) 0.83(0.60-1.14) 0.70(0.50-0.97) 0.02 (95% CI) [reference] Model 2 HR 1.0 0.88(0.65-1.20) 0.78(0.56-1.07) 0.82(0.58-1.14) 0.67(0.47-0.96) 0.01 (95% CI) [reference] ADA n CVD / person 259/3850707 344/4586097 345/5243180 271/4688623 235/4455103 years Model 1 HR 1.0 1.20(1.02-1.41) 1.02(0.87-1.20) 0.92(0.77-1.10) 0.84(0.70-1.01) 0.004 (95% CI) [reference] Model 2 HR 1.0 1.19(1.01-1.40) 1.02(0.87-1.20) 0.90(0.76-1.08) 0.83(0.69-0.99) 0.003 (95% CI) [reference] n CHD / person 120/4257868 162/5155297 139/5816819 116/5146065 98/4852429 years Model 1 HR 1.0 1.22(0.96-1.55) 0.89(0.69-1.14) 0.85(0.66-1.10) 0.74(0.56-0.97) 0.003 (95% CI) [reference] Model 2 HR 1.0 1.23(0.97-1.56) 0.88(0.68-1.13) 0.83(0.64-1.08) 0.73(0.55-0.96) 0.002 (95% CI) [reference]

170 n Stroke / 70/4321480 81/5251815 90/5825006 76/5179326 55/4923861 person years Model 1 HR 1.0 0.98(0.71-1.35) 0.91(0.66-1.25) 0.86(0.62-1.21) 0.65(0.45-0.93) 0.01 (95% CI) [reference] Model 2 HR 1.0 0.96(0.70-1.33) 0.92(0.67-1.26) 0.86(0.62-1.21) 0.65(0.45-0.93) 0.01 (95% CI) [reference] Paleo n CVD / person 306/4807665 283/4069294 299/4335000 315/5326313 251/4285438 years Model 1 HR 1.0 1.15(0.88-1.50) 1.13(0.86-1.47) 0.92(0.70-1.22) 0.89(0.65-1.23) 0.35 (95% CI) [reference] Model 2 HR 1.0 1.07(0.91-1.27) 1.06(0.89-1.25) 0.89(0.75-1.06) 0.90(0.74-1.08) 0.12 (95% CI) [reference] n CHD / person 127/5360103 127/4508421 135/4776909 136/5867494 110/4715551 years Model 1 HR 1.0 1.22(0.95-1.56) 1.21(0.95-1.55) 0.99(0.77-1.27) 1.02(0.79-1.33) 0.61 (95% CI) [reference] Model 2 HR 1.0 1.18(0.92-1.52) 1.22(0.95-1.57) 1.01(0.78-1.31) 1.04(0.78-1.39) 0.80 (95% CI) [reference] n Stroke / 79/5406310 76/4560649 80/4839861 69/5969754 68/4724914 person years Model 1 HR 1.0 1.15(0.84-1.58) 1.11(0.81-1.52) 0.75(0.54-1.05) 0.94(0.67-1.32) 0.21 (95% CI) [reference] Model 2 HR 1.0 1.08(0.78-1.48) 1.04(0.75-1.44) 0.69(0.49-0.97) 0.84(0.58-1.21) 0.07 (95% CI) [reference] Abbreviations: HR, hazard ratio; CI, confidence interval; CVD, cardiovascular disease (defined as the first occurrence of an acute myocardial infarction (MI) requiring overnight hospitalization, silent MI determined from serial ECGs, coronary heart disease (CHD) death, coronary revascularization, peripheral artery disease (PAD), congestive heart failure (CHF) or ischemic or hemorrhagic stroke); CHD, coronary heart disease (first occurrence of a clinical MI, definite silent MI or a death due to definite or possible CHD). a Model 1 HRs estimated from Cox proportional hazards models adjusting for age, race, education, income, marital status, physical activity, smoking, BMI, WHI study arm, and geographical region. b Model 2 HRs estimated from Cox proportional hazards models adjusting for Model 1 covariates + age at T2D diagnosis, energy intake, insulin use, blood pressure, and history of high cholesterol. *DASH and ADA models additionally adjusted for alcohol **CVD models additionally adjusted for smoking*time, systolic BP*time, diastolic BP*time. ***Stroke models additionally adjusted for income*time, age at T2D onset*time, HRT CT arm*time.

171 Supplemental Tables

Supplemental Table 1. Nutrient and food group characteristics of the DASH dietary pattern score by quintile of WHI participants with type 2 diabetes Quintile of DASH Diet Score Nutrient Intakea Q1 Q2 Q3 Q4 Q5 1622.5 1542.6 1613.1 1652.9 1665.3 Total Energy (Calories) (709.5) (705.9) (663.2) (662.3) (569.6) Carbohydrate (g) 106.3 (20.8) 114.2 (21.0) 119.6 (20.7) 127.4 (20.2) 139.1 (20.3) Total Sugar (g) 47.4 (20.4) 49.5 (18.4) 53.5 (17.3) 58.6 (16.5) 67.0 (15.9) Fiber (g) 7.4 (2.1) 9.3 (2.6) 10.5 (3.3) 11.9 (3.5) 14.0 (3.7) Protein (g) 43.2 (9.6) 44.2 (8.7) 44.8 (8.5) 45.2 (8.3) 45.3 (7.4) Total Fat (g) 44.8 (8.0) 41.4 (8.3) 39.1 (8.4) 36.0 (8.2) 31.7 (8.2) Saturated Fat (g) 14.8 (3.3) 13.5 (3.3) 12.7 (3.4) 11.6 (3.2) 9.9 (2.9) Monounsaturated Fat (g) 17.2 (3.3) 15.9 (3.6) 14.8 (3.6) 13.6 (3.5) 11.9 (3.5) Polyunsaturated Fat (g) 9.0 (2.8) 8.7 (2.7) 8.3 (2.6) 7.8 (2.3) 7.3 (2.3) Dietary Cholesterol (mg) 187.3 (83.9) 168.3 (73.1) 152.8 (58.4) 142.6 (59.6) 118.3 (49.5) 1732.5 1788.1 1789.8 1798.6 1793.4 Sodium (mg) (304.8) (3.09.7) (323.9) (299.0) (293.9) Alcohol (g) 1.1 (4.0) 1.2 (4.2) 1.4 (4.7) 1.2 (3.7) 1.1 (3.6) Food Group Intakea Added Sugarb 5.5 (3.7) 4.5 (2.7) 4.3 (2.2) 4.3 (2.1) 4.3 (1.8) Total Dairyc 0.7 (0.5) 0.9 (0.5) 1.0 (0.7) 1.1 (0.6) 1.3 (0.6) Total Fruitc 0.6 (0.5) 0.9 (0.7) 1.0 (0.8) 1.3 (0.8) 1.6 (0.7) Total Vegetablesc 0.8 (0.3) 0.9 (0.4) 1.0 (0.5) 1.1 (0.5) 1.3 (0.5) Starchy Vegetablesc 0.2 (0.1) 0.2 (0.1) 0.2 (0.1) 0.2 (0.1) 0.2 (0.1) Dark Green Vegetablesc 0.04 (0.05) 0.05 (0.06) 0.06 (0.09) 0.08 (0.1) 0.1 (0.1) Total Grainsd 3.0 (1.0) 3.2 (1.1) 3.2 (1.1) 3.2 (1.0) 3.2 (1.0) Whole Grainsd 0.4 (0.4) 0.6 (0.5) 0.8 (0.6) 0.9 (0.6) 1.1 (0.6) Legumesc 0.04 (0.06) 0.05 (0.06) 0.05 (0.06) 0.06 (0.07) 0.07 (0.08) Nuts/Seedse 0.1 (0.2) 0.2 (0.3) 0.2 (0.3) 0.3 (0.4) 0.4 (0.4) Meatf 1.5 (0.9) 1.2 (0.7) 1.1 (0.7) 0.9 (0.6) 0.6 (0.5) Poultryg 0.7 (0.6) 0.7 (0.6) 0.7 (0.6) 0.7 (0.6) 0.7 (0.6) Fishg 0.4 (0.4) 0.4 (0.4) 0.4 (0.4) 0.4 (0.4) 0.4 (0.4) a Estimated daily nutrient and food group intakes reported as quintile mean (SD); values for all nutrients (except Total Energy) and food groups are per 1,000 Calories. b Teaspoon equivalents c Cup equivalents d Ounce equivalents e Ounce equivalents of lean meat f Ounces of cooked lean meat from beef, pork, lamb, game and veal g Ounces of cooked poultry h Ounces of cooked fish

172

Supplemental Table 2. Nutrient and food group characteristics of the alternate Mediterranean (aMed) dietary pattern score by quintile of WHI participants with type 2 diabetes Quintile of aMed Diet Score Nutrient Intakea Q1 Q2 Q3 Q4 Q5 1330.3 1441.2 1619.6 1784.4 1935.2 Total Energy (Calories) (598.0) (584.3) (641.6) (688.4) (562.6) Carbohydrate (g) 112.1 (23.3) 118.0 (23.5) 121.7 923.2) 124.2 (22.4) 128.7 (21.2) Total Sugar (g) 51.8 (21.0) 54.1 (19.7) 55.1 (19.6) 56.0 (18.3) 57.7 (16.2) Fiber (g) 8.4 (3.1) 9.8 (3.5) 10.6 (3.6) 11.5 (3.8) 12.5 (3.6) Protein (g) 44.4 (9.7) 44.1 (8.9) 44.4 (8.5) 45.0 (8.2) 44.6 (7.4) Total Fat (g) 42.1 (9.1) 40.0 (9.5) 38.6 (9.2) 37.2 (9.1) 35.8 (8.6) Saturated Fat (g) 14.7 (3.7) 13.2 (3.5) 12.4 (3.4) 11.7 (3.2) 10.8 (3.0) Monounsaturated Fat (g) 15.8 (3.7) 15.3 (4.1) 14.7 (4.0) 14.2 (3.9) 13.9 (3.7) Polyunsaturated Fat (g) 8.1 (2.6) 8.3 (2.9) 8.3 (2.7) 8.2 (2.6) 8.3 (2.3) Dietary Cholesterol (mg) 181.5 (84.9) 164.0 (70.1) 151.2 (68.5) 146.2 (64.8) 131.0 (49.1) 1711.3 1745.7 1786.5 1822.6 1836.5 Sodium (mg) (316.7) (315.4) (309.3) (303.8) (268.1) Alcohol (g) 1.1 (4.3) 1.2 (4.7) 1.1 (3.4) 1.4 (4.6) 1.2 (3.0) Food Group Intakea Added Sugarb 5.1 (3.5) 4.8 (2.7) 4.5 (2.6) 4.3 (2.3) 4.3 (1.8) Total Dairyc 1.0 (0.7) 1.0 (0.6) 1.0 (0.6) 1.0 (0.6) 1.0 (0.6) Total Fruitc 0.8 (0.7) 1.0 (0.8) 1.1 (0.8) 1.2 (0.8) 1.2 (0.6) Total Vegetablesc 0.8 (0.4) 0.9 (0.5) 1.0 (0.5) 1.1 (0.5) 1.2 (0.5) Starchy Vegetablesc 0.2 (0.1) 0.2 (0.1) 0.2 (0.1) 0.2 (0.1) 0.2 (0.1) Dark Green Vegetablesc 0.05 (0.06) 0.05 (0.8) 0.06 (0.08) 0.08 (0.1) 0.08 (0.1) Total Grainsd 2.9 (1.0) 3.1 (1.1) 3.2 (1.1) 3.2 (1.0) 3.3 (1.0) Whole Grainsd 0.5 (0.5) 0.7 (0.6) 0.8 (0.6) 0.9 (0.6) 1.0 (0.6) Legumesc 0.04 (0.05) 0.05 (0.07) 0.05 (0.07) 0.06 (0.07) 0.07 (0.07) Nuts/Seedse 0.1 (0.2) 0.2 (0.3) 0.2 (0.4) 0.3 (0.4) 0.4 (0.4) Meatf 1.3 (0.9) 1.1 (0.8) 1.0 (0.7) 1.0 (0.7) 0.8 (0.6) Poultryg 0.7 (0.6) 0.7 (0.6) 0.7 (0.6) 0.7 (0.6) 0.7 (0.5) Fishg 0.3 (0.3) 0.3 (0.3) 0.4 (0.4) 0.5 (0.4) 0.5 (0.4) a Estimated daily nutrient and food group intakes reported as quintile mean (SD); values for all nutrients (except Total Energy) and food groups are per 1,000 Calories. b Teaspoon equivalents c Cup equivalents d Ounce equivalents e Ounce equivalents of lean meat f Ounces of cooked lean meat from beef, pork, lamb, game and veal g Ounces of cooked poultry h Ounces of cooked fish

173 Supplemental Table 3. Nutrient and food group characteristics of the American Diabetes Association (ADA) recommendations dietary pattern score by quintile of WHI participants with type 2 diabetes Quintile of ADA Diet Score Nutrient Intakea Q1 Q2 Q3 Q4 Q5 1647.9 1563.2 1630.8 1696.9 1552.2 Total Energy (Calories) (633.4) (720.6) (109.2) (698.6) (527.4) Carbohydrate (g) 107.5 (20.6) 114.5 (21.2) 120.8 (22.3) 125.8 (22.4) 136.0 (20.6) Total Sugar (g) 48.1 (18.8) 50.7 (19.0) 54.3 (19.0) 58.2 (18.3) 63.6 (16.6) Fiber (g) 7.5 (2.3) 9.1 (3.0) 10.5 (3.4) 11.7 (3.5) 13.7 (3.6) Protein (g) 40.4 (7.9) 43.5 (8.5) 44.5 (8.3) 46.2 (8.0) 48.0 (8.3) Total Fat (g) 45.6 (8.0) 41.6 (8.2) 38.9 (8.4) 36.3 (8.4) 31.4 (7.7) Saturated Fat (g) 15.8 (3.4) 13.7 (3.2) 12.5 (3.0) 11.4 (2.9) 9.5 (2.5) Monounsaturated Fat (g) 17.2 (3.4) 15.8 (3.5) 14.9 (3.7) 13.9 (3.7) 12.0 (3.4) Polyunsaturated Fat (g) 8.9 (2.8) 8.7 (2.9) 8.3 (2.6) 8.0 (2.5) 7.3 (2.1) Dietary Cholesterol (mg) 176.1 (81.8) 167.8 (75.5) 153.7 (65.6) 146.5 (64.5) 130.7 (55.4) 1695.3 1761.9 1786.5 1815.2 1835.54 Sodium (mg) (303.6) (288.6) (308.3) (309.6) (306.5) Alcohol (g) 1.4 (4.6) 1.2 (4.2) 1.1 (3.7) 1.1 (4.0) 1.2 (3.8) Food Group Intakea Added Sugarb 5.5 (3.5) 4.8 (2.8) 4.4 (2.4) 4.3 (2.2) 4.1 (1.8) Total Dairyc 0.9 (0.6) 0.9 (0.6) 1.0 (0.6) 1.1 (0.6) 1.1 (0.6) Total Fruitc 0.6 (0.4) 0.8 (0.7) 1.1 (0.8) 1.2 (0.8) 1.5 (0.7) Total Vegetablesc 0.7 (0.3) 0.9 (0.4) 1.0 (0.5) 1.1 (0.5) 1.3 (0.5) Starchy Vegetablesc 0.2 (0.1) 0.2 (0.1) 0.2 (0.1) 0.2 (0.1) 0.2 (0.1) Dark Green Vegetablesc 0.04 (0.05) 0.05 (0.07) 0.06 (0.08) 0.07 (0.09) 0.1 (0.1) Total Grainsd 3.0 (1.0) 3.1 (1.10) 3.2 (1.1) 3.1 (1.0) 3.2 (1.0) Whole Grainsd 0.5 (0.4) 0.6 (0.5) 0.8 (0.6) 0.9 (0.6) 1.1 (0.6) Legumesc 0.04 (0.05) 0.05 (0.07) 0.05 (0.07) 0.06 (0.07) 0.06 (0.08) Nuts/Seedse 0.2 (0.3) 0.2 (0.3) 0.2 (0.4) 0.3 (0.4) 0.3 (0.4) Meatf 1.1 (0.8) 1.1 (0.8) 1.1 (0.8) 1.0 (0.7) 0.9 (0.7) Poultryg 0.5 (0.4) 0.7 (0.6) 0.7 (0.6) 0.7 (0.6) 0.8 (0.7) Fishg 0.3 (0.3) 0.4 (0.3) 0.4 (0.3) 0.4 (0.4) 0.5 (0.5) a Estimated daily nutrient and food group intakes reported as quintile mean (SD); values for all nutrients (except Total Energy) and food groups are per 1,000 Calories. b Teaspoon equivalents c Cup equivalents d Ounce equivalents e Ounce equivalents of lean meat f Ounces of cooked lean meat from beef, pork, lamb, game and veal g Ounces of cooked poultry h Ounces of cooked fish

174 Supplemental Table 7.4 Nutrient and food group characteristics of the Paleolithic (Paleo) dietary pattern score by quintile of WHI participants with type 2 diabetes Quintile of Paleo Diet Score Nutrient Intakea Q1 Q2 Q3 Q4 Q5 2054.9 1798.6 1578.3 1422.7 1233.6 Total Energy (Calories) (660.0) (667.0) (658.7) (575.0) (469.1) Carbohydrate (g) 113.5 (19.9) 116.2 (20.8) 118.1 (22.1) 123.8 (23.5) 132.7 (25.6) Total Sugar (g) 51.4 (18.7) 51.6 (17.7) 52.8 (18.7) 56.0 (18.7) 62.9 (19.6) Fiber (g) 7.9 (2.3) 9.0 (2.5) 10.0 (2.8) 11.5 (3.3) 14.2 (4.2) Protein (g) 40.9 (7.4) 44.2 (7.9) 44.7 (8.1) 46.0 (8.5) 47.0 (9.6) Total Fat (g) 42.8 (8.1) 40.4 (8.9) 39.6 (8.7) 37.2 (9.3) 33. 8 (9.1) Saturated Fat (g) 14.7 (3.4) 13.4 (3.4) 12.9 (3.3) 11.7 (3.3) 10.2 (3.1) Monounsaturated Fat (g) 16.2 (3.4) 15.4 (3.7) 15.1 (3.7) 14.2 (3.9) 12.8 (4.1) Polyunsaturated Fat (g) 8.5 (2.6) 8.4 (2.6) 8.3 (2.6) 8.2 (2.9) 7.9 (2.5) Dietary Cholesterol (mg) 157.1 (62.5) 157.8 (67.2) 158.2 (67.4) 154.5 (75.7) 147.4 (78.9) 1731.7 1780.0 1777.0 1815.22 1793 .3 Sodium (mg) (296.3) (274.5) (302.1) (320.6) (327.9) Alcohol (g) 1.4 (4.2) 1.3 (3.8) 1.4 (4.6) 1.0 (3.8) 0.9 (3.7) Food Group Intakea Added Sugarb 5.8 (3.4) 4.9 (2.6) 4.4 (2.4) 4.1 (2.2) 3.7 (1.8) Total Dairyc 1.1 (0.6) 1.1 (0.7) 1.0 (0.6) 1.0 (0.6) 0.9 (0.6) Total Fruitc 0.6 (0.4) 0.8 (0.4) 1.0 (0.6) 1.2 (0.7) 1.8 (1.0) Total Vegetablesc 0.7 (0.3) 0.8 (0.3) 0.9 (0.4) 1.2 (0.4) 1.5 (0.6) Starchy Vegetablesc 0.2 (0.1) 0.2 (0.1) 0.2 (0.1) 0.2 (0.1) 0.2 (0.1) Dark Green Vegetablesc 0.02 (0.03) 0.03 (0.03) 0.05 (0.05) 0.08 (0.08) 0.1 (0.1) Total Grainsd 3.3 (1.1) 3.3 (1.0) 3.2 (1.0) 3.1 (1.0) 2.8 (1.0) Whole Grainsd 0.7 (0.6) 0.8 (0.6) 0.8 (0.6) 0.8 (0.6) 0.8 (0.6) Legumesc 0.05 (0.07) 0.05 (0.06) 0.06 (0.07) 0.06 (0.07) 0.06 (0.08) Nuts/Seedse 0.2 (0.3) 0.2 (0.3) 0.2 (0.3) 0.3 (0.4) 0.3 (0.5) Meatf 1.0 (0.7) 1.1 (0.7) 1.1 (0.7) 1.1 (0.8) 1.0 (0.8) Poultryg 0.5 (0.4) 0.6 (0.5) 0.7 (0.5) 0.7 (0.6) 0.9 (0.8) Fishg 0.2 (0.2) 0.3 (0.3) 0.4 (0.3) 0.5 (0.4) 0.6 (0.5) a Estimated daily nutrient and food group intakes reported as quintile mean (SD); values for all nutrients (except Total Energy) and food groups are per 1,000 Calories. b Teaspoon equivalents c Cup equivalents d Ounce equivalents e Ounce equivalents of lean meat f Ounces of cooked lean meat from beef, pork, lamb, game and veal g Ounces of cooked poultry h Ounces of cooked fish

175 CHAPTER 8 Summary and Conclusions Diet has been identified as the largest contributing risk factor across the leading causes of death in the United States. The nature of the diet-disease relationship based on the current body of evidence, however, remains inconclusive, largely due to the historical reductionist approach of studying specific foods and nutrients in isolation, which may not capture the more complex relationship between diet as a whole and disease risk. The objective of this dissertation was to further explore the diet-disease relationship using a priori dietary pattern scores, specifically focusing on type 2 diabetes mellitus (T2D) and cardiovascular disease (CVD) risk over time.

This dissertation consisted of three separate research projects to investigate the association between 1) contemporary national dietary guidelines and popular adapted dietary trends and 2) sweetened beverage consumption and T2D risk in young adults and 3) diet quality and CVD risk in postmenopausal women with prevalent T2D. Results of the analyses showed that a dietary pattern characterized by higher intake of fruits, vegetables, low-fat dairy, fish, poultry, legumes, nuts/seeds, whole grains, alcohol, coffee/tea and lower intake of red/processed meat, fried foods, sugar-sweetened foods and beverages, and whole fat dairy is associated with a lower risk of T2D in young adults. In addition, higher sweetened beverage intake serves as a strong predictor of higher T2D risk in young adults. Finally, a dietary pattern emphasizing fruits, vegetables, whole grains, nuts/seeds, legumes and a high unsaturated:saturated fat ratio is associated with a lower risk of incident CVD in postmenopausal women with T2D. In conclusion, diet quality is associated with cardiometabolic disease risk throughout adulthood and regardless of underlying disease risk. Future research should replicate these findings in other populations to provide further insight to inform chronic disease prevention efforts at the population level.

176 REFERENCES

1. Marczak L, O’Rourke K, Shepard D. When and Why People Die in the United States, 1990-2013. JAMA. 2016;315(3):241. doi:10.1001/jama.2015.17599. 2. Jacobs DR, Steffen LM, Sabaté J, Rajaram S. Nutrients, foods, and dietary patterns as exposures in research: a framework for food synergy. Am J Clin Nutr. 2003;78:508-513. 3. Health NRC (US) C on D and. Diet and Health. National Academies Press (US); 1989. doi:10.17226/1222. 4. Kahn SE, Prigeon RL, McCulloch DK, et al. Quantification of the relationship between insulin sensitivity and beta-cell function in human subjects. Evidence for a hyperbolic function. Diabetes. 1993;42(11):1663-1672. http://www.ncbi.nlm.nih.gov/pubmed/8405710. Accessed April 20, 2017. 5. DeFronzo RA. Pathogenesis of type 2 diabetes mellitus. Med Clin North Am. 2004;88(4):787-835. doi:10.1016/j.mcna.2004.04.013. 6. Butler PC, Rizza RA. Contribution to Postprandial Hyperglycemia and Effect on Initial Splanchnic Glucose Clearance of Hepatic Glucose Cycling in Glucose-Intolerant or NIDDM Patients. Diabetes. 1991;40(1). http://diabetes.diabetesjournals.org/content/40/1/73. Accessed April 24, 2017. 7. Ramnanan CJ, Edgerton DS, Rivera N, et al. Molecular characterization of insulin- mediated suppression of hepatic glucose production in vivo. Diabetes. 2010;59(6):1302- 1311. doi:10.2337/db09-1625. 8. Kahn BB. Lilly lecture 1995. Glucose transport: pivotal step in insulin action. Diabetes. 1996;45(11):1644-1654. http://www.ncbi.nlm.nih.gov/pubmed/8866574. Accessed April 20, 2017. 9. Jefferson LS. Lilly Lecture 1979: role of insulin in the regulation of protein synthesis. Diabetes. 1980;29(6):487-496. http://www.ncbi.nlm.nih.gov/pubmed/6991336. Accessed April 20, 2017. 10. Flakoll PJ, Kulaylat M, Frexes-Steed M, et al. Amino acids augment insulin’s suppression of whole body proteolysis. Am J Physiol. 1989;257(6 Pt 1):E839-47. http://www.ncbi.nlm.nih.gov/pubmed/2692456. Accessed April 20, 2017. 11. Strålfors P, Björgell P, Belfrage P. Hormonal regulation of hormone-sensitive lipase in intact adipocytes: identification of phosphorylated sites and effects on the phosphorylation

177 by lipolytic hormones and insulin. Proc Natl Acad Sci U S A. 1984;81(11):3317-3321. http://www.ncbi.nlm.nih.gov/pubmed/6374655. Accessed April 20, 2017. 12. Fielding BA, Frayn KN. Lipoprotein lipase and the disposition of dietary fatty acids. Br J Nutr. 1998;80(6):495-502. http://www.ncbi.nlm.nih.gov/pubmed/10211047. Accessed April 20, 2017. 13. Lass A, Zimmermann R, Oberer M, Zechner R. Lipolysis – A highly regulated multi- enzyme complex mediates the catabolism of cellular fat stores. Prog Lipid Res. 2011;50(1):14-27. doi:10.1016/j.plipres.2010.10.004. 14. DeFronzo RA. The Triumvirate: β-Cell, Muscle, Liver: A Collusion Responsible for NIDDM. Diabetes. 1988;37(6). http://diabetes.diabetesjournals.org/content/37/6/667. Accessed April 24, 2017. 15. DeFronzo R. Pathogenesis of type 2 diabetes: Metabolic and molecular implications for identifying diabetes genes. Diabetes Rev. 1997;5(3):177-269. https://uthscsa.influuent.utsystem.edu/en/publications/pathogenesis-of-type-2-diabetes- metabolic-and-molecular-implicati. Accessed April 24, 2017. 16. Groop L, Lyssenko V. Genes and type 2 diabetes mellitus. Curr Diab Rep. 2008;8(3):192- 197. http://www.ncbi.nlm.nih.gov/pubmed/18625115. Accessed April 24, 2017. 17. James WPT. The fundamental drivers of the obesity epidemic. Obes Rev. 2008;9(s1):6-13. doi:10.1111/j.1467-789X.2007.00432.x. 18. Nolan CJ, Damm P, Prentki M. Type 2 diabetes across generations: from pathophysiology to prevention and management. Lancet. 2011;378(9786):169-181. doi:10.1016/S0140- 6736(11)60614-4. 19. DeFronzo RA, Ferrannini E, Simonson DC. Fasting hyperglycemia in non-insulin- dependent diabetes mellitus: Contributions of excessive hepatic glucose production and impaired tissue glucose uptake. Metabolism. 1989;38(4):387-395. doi:10.1016/0026- 0495(89)90129-7. 20. Groop LC, Bonadonna RC, DelPrato S, et al. Glucose and free fatty acid metabolism in non-insulin-dependent diabetes mellitus. Evidence for multiple sites of insulin resistance. J Clin Invest. 1989;84(1):205-213. doi:10.1172/JCI114142. 21. Ferrannini E, Simonson DC, Katz LD, et al. The disposal of an oral glucose load in patients with non-insulin-dependent diabetes. Metabolism. 1988;37(1):79-85.

178 http://www.ncbi.nlm.nih.gov/pubmed/3275860. Accessed April 24, 2017. 22. DeFronzo RA. From the Triumvirate to the Ominous Octet: A New Paradigm for the Treatment of Type 2 Diabetes Mellitus. Diabetes. 2009;58(4). http://diabetes.diabetesjournals.org/content/58/4/773. Accessed April 21, 2017. 23. Abdul-Ghani MA, Matsuda M, Sabbah M, et al. The relative contributions of insulin resistance and beta cell failure to the transition from normal to impaired glucose tolerance varies in different ethnic groups. Diabetes Metab Syndr Clin Res Rev. 2007;1(2):105-112. doi:10.1016/j.dsx.2007.02.004. 24. GROOP LC, SALORANTA C, SHANK M, BONADONNA RC, FERRANNINI E, DEFRONZO RA. The Role of Free Fatty Acid Metabolism in the Pathogenesis of Insulin Resistance in Obesity and Noninsulin-Dependent Diabetes Mellitus*. J Clin Endocrinol Metab. 1991;72(1):96-107. doi:10.1210/jcem-72-1-96. 25. Bays H, Mandarino L, DeFronzo RA. Role of the Adipocyte, Free Fatty Acids, and Ectopic Fat in Pathogenesis of Type 2 Diabetes Mellitus: Peroxisomal Proliferator- Activated Receptor Agonists Provide a Rational Therapeutic Approach. J Clin Endocrinol Metab. 2004;89(2):463-478. doi:10.1210/jc.2003-030723. 26. Bays HE, González-Campoy JM, Bray GA, et al. Pathogenic potential of adipose tissue and metabolic consequences of adipocyte hypertrophy and increased visceral adiposity. Expert Rev Cardiovasc Ther. 2008;6(3):343-368. doi:10.1586/14779072.6.3.343. 27. Unger RH, Aguilar-Parada E, Müller WA, Eisentraut AM. Studies of pancreatic alpha cell function in normal and diabetic subjects. J Clin Invest. 1970;49(4):837-848. doi:10.1172/JCI106297. 28. Baron AD, Schaeffer L, Shragg P, Kolterman OG. Role of hyperglucagonemia in maintenance of increased rates of hepatic glucose output in type II diabetics. Diabetes. 1987;36(3):274-283. http://www.ncbi.nlm.nih.gov/pubmed/2879757. Accessed April 24, 2017. 29. Meier JJ, Deacon CF, Schmidt WE, Holst JJ, Nauck MA. Suppression of glucagon secretion is lower after oral glucose administration than during intravenous glucose administration in human subjects. Diabetologia. 2007;50(4):806-813. doi:10.1007/s00125-007-0598-z. 30. Drucker DJ. The biology of incretin hormones. Cell Metab. 2006;3(3):153-165.

179 doi:10.1016/j.cmet.2006.01.004. 31. Drucker DJ, Nauck MA. The incretin system: glucagon-like peptide-1 receptor agonists and dipeptidyl peptidase-4 inhibitors in type 2 diabetes. Lancet. 2006;368(9548):1696- 1705. doi:10.1016/S0140-6736(06)69705-5. 32. Meier JJ, Nauck MA. Incretins and the development of type 2 diabetes. Curr Diab Rep. 2006;6(3):194-201. doi:10.1007/s11892-006-0034-7. 33. Meier JJ, Hücking K, Holst JJ, Deacon CF, Schmiegel WH, Nauck MA. Reduced insulinotropic effect of gastric inhibitory polypeptide in first-degree relatives of patients with type 2 diabetes. Diabetes. 2001;50(11):2497-2504. http://www.ncbi.nlm.nih.gov/pubmed/11679427. Accessed April 24, 2017. 34. Jones IR, Owens DR, Luzio S, Williams S, Hayes TM. The glucose dependent insulinotropic polypeptide response to oral glucose and mixed meals is increased in patients with type 2 (non-insulin-dependent) diabetes mellitus. Diabetologia. 1989;32(9):668-677. http://www.ncbi.nlm.nih.gov/pubmed/2676668. Accessed April 24, 2017. 35. Abdul-Ghani M, DeFronzo R. Inhibition of Renal Glucose Reabsorption: A Novel Strategy for Achieving Glucose Control in Type 2 Diabetes Mellitus. Endocr Pract. 2008;14(6):782-790. doi:10.4158/EP.14.6.782. 36. Porte D. Central Regulation of Energy Homeostasis. Diabetes. 2006;55(Supplement 2). http://diabetes.diabetesjournals.org/content/55/Supplement_2/S155. Accessed April 24, 2017. 37. Plum L, Belgardt BF, Brüning JC. Central insulin action in energy and glucose homeostasis. J Clin Invest. 2006;116(7):1761-1766. doi:10.1172/JCI29063. 38. Obici S, Feng Z, Tan J, Liu L, Karkanias G, Rossetti L. Central melanocortin receptors regulate insulin action. J Clin Invest. 2001;108(7):1079-1085. doi:10.1172/JCI12954. 39. Obici S, Feng Z, Karkanias G, Baskin DG, Rossetti L. Decreasing hypothalamic insulin receptors causes hyperphagia and insulin resistance in rats. Nat Neurosci. 2002;5(6):566- 572. doi:10.1038/nn861. 40. Matsuda M, Liu Y, Mahankali S, et al. Altered hypothalamic function in response to glucose ingestion in obese humans. Diabetes. 1999;48(9):1801-1806. http://www.ncbi.nlm.nih.gov/pubmed/10480611. Accessed April 24, 2017.

180 41. Centers for Disease Control and Prevention. National Diabetes Statistics Report: Estimates of Diabetes and Its Burden in the United States, 2014. Atlanta; 2014. https://www.cdc.gov/diabetes/data/statistics/2014statisticsreport.html. 42. Centers for Disease Control and Prevention. National Diabetes Statistics Report, 2017. Atlanta, GA; 2017. 43. World Health Organization. Global Report on Diabetes. Geneva; 2016. 44. Muoio DM, Newgard CB. Mechanisms of disease: Molecular and metabolic mechanisms of insulin resistance and β-cell failure in type 2 diabetes. Nat Rev Mol Cell Biol. 2008;9(3):193-205. doi:10.1038/nrm2327. 45. Poulsen P, Kyvik KO, Vaag A, Beck-Nielsen H. Heritability of type II (non-insulin- dependent) diabetes mellitus and abnormal glucose tolerance--a population-based twin study. Diabetologia. 1999;42(2):139-145. http://www.ncbi.nlm.nih.gov/pubmed/10064092. Accessed April 27, 2017. 46. Meigs JB, Cupples LA, Wilson PW. Parental transmission of type 2 diabetes: the Framingham Offspring Study. Diabetes. 2000;49(12):2201-2207. http://www.ncbi.nlm.nih.gov/pubmed/11118026. Accessed April 27, 2017. 47. Herder C, Roden M. Genetics of type 2 diabetes: pathophysiologic and clinical relevance. Eur J Clin Invest. 2011;41(6):679-692. doi:10.1111/j.1365-2362.2010.02454.x. 48. Narayan KMV, Boyle JP, Thompson TJ, Sorensen SW, Williamson DF. Lifetime Risk for Diabetes Mellitus in the United States. JAMA. 2003;290(14):1884. doi:10.1001/jama.290.14.1884. 49. Shai I, Jiang R, Manson JE, et al. Ethnicity, Obesity, and Risk of Type 2 Diabetes in Women: A 20-year follow-up study. Diabetes Care. 2006;29(7):1585-1590. doi:10.2337/dc06-0057. 50. Scott RA, Lagou V, Welch RP, et al. Large-scale association analyses identify new loci influencing glycemic traits and provide insight into the underlying biological pathways. Nat Genet. 2012;44(9):991-1005. doi:10.1038/ng.2385. 51. Morris AP, Voight BF, Teslovich TM, et al. Large-scale association analysis provides insights into the genetic architecture and pathophysiology of type 2 diabetes. Nat Genet. 2012;44(9):981-990. doi:10.1038/ng.2383. 52. Feero WG, Guttmacher AE, McCarthy MI. Genomics, Type 2 Diabetes, and Obesity. N

181 Engl J Med. 2010;363(24):2339-2350. doi:10.1056/NEJMra0906948. 53. Lyssenko V, Jonsson A, Almgren P, et al. Clinical Risk Factors, DNA Variants, and the Development of Type 2 Diabetes. N Engl J Med. 2008;359(21):2220-2232. doi:10.1056/NEJMoa0801869. 54. Meigs JB, Shrader P, Sullivan LM, et al. Genotype Score in Addition to Common Risk Factors for Prediction of Type 2 Diabetes. N Engl J Med. 2008;359(21):2208-2219. doi:10.1056/NEJMoa0804742. 55. Ali O. Genetics of type 2 diabetes. World J Diabetes. 2013;4(4):114-123. doi:10.4239/wjd.v4.i4.114. 56. Cerf ME. Beta cell dysfunction and insulin resistance. Front Endocrinol (Lausanne). 2013;4:37. doi:10.3389/fendo.2013.00037. 57. Chan JM, Rimm EB, Colditz GA, Stampfer MJ, Willett WC. Obesity, fat distribution, and weight gain as risk factors for clinical diabetes in men. Diabetes Care. 1994;17(9):961- 969. http://www.ncbi.nlm.nih.gov/pubmed/7988316. Accessed April 29, 2017. 58. Biggs ML, Mukamal KJ, Luchsinger JA, et al. Association Between Adiposity in Midlife and Older Age and Risk of Diabetes in Older Adults. JAMA. 2010;303(24):2504. doi:10.1001/jama.2010.843. 59. Chaput J-P, Pérusse L, Després J-P, Tremblay A, Bouchard C. Findings from the Quebec Family Study on the Etiology of Obesity: Genetics and Environmental Highlights. Curr Obes Rep. 2014;3(1):54-66. doi:10.1007/s13679-013-0086-3. 60. Serra-Majem L, Bautista-Castaño I. Etiology of obesity: two "key issues" and other emerging factors. Nutr Hosp. 2013;28(5):32-4332. http://www.nutricionhospitalaria.com/pdf/6916.pdf. Accessed April 27, 2017. 61. Atkinson RL. Current status of the field of obesity. Trends Endocrinol Metab. 2014;25(6):283-284. doi:10.1016/j.tem.2014.03.003. 62. Tuomilehto J, Lindstrom J, Eriksson JG, et al. Prevention of type 2 diabetes mellitus by changes in lifestyle among subjects with impaired glucose tolerance. N Engl J Med. 2001;344(18):1343-1350. doi:10.1056/NEJM200105033441801. 63. Diabetes Prevention Program Research Group. Reduction in the Incidence of Type 2 Diabetes with Lifestyle Intervention or Metformin. N Engl J Med. 2002;346(6):393-403. doi:10.1056/NEJMoa012512.

182 64. van Dam RM, Rimm EB, Willett WC, Stampfer MJ, Hu FB. Dietary patterns and risk for type 2 diabetes mellitus in U.S. men. Ann Intern Med. 2002;136(3):201-209. http://www.ncbi.nlm.nih.gov/pubmed/11827496. 65. van Dam RM, Willett WC, Rimm EB, Stampfer MJ, Hu FB. Dietary fat and meat intake in relation to risk of type 2 diabetes in men. Diabetes Care. 2002;25(3):417-424. http://www.ncbi.nlm.nih.gov/pubmed/11874924. 66. Pan A, Sun Q, Bernstein AM, Manson JE, Willett WC, Hu FB. Changes in Red Meat Consumption and Subsequent Risk of Type 2 Diabetes Mellitus. JAMA Intern Med. 2013;173(14):1328. doi:10.1001/jamainternmed.2013.6633. 67. Ley SH, Hamdy O, Mohan V, Hu FB. Prevention and Management of Type 2 Diabetes: Dietary Components and Nutritional Strategies. Lancet (London, England). 2014;383(9933):1999. doi:10.1016/s0140-6736(14)60613-9. 68. Hu FB, van Dam RM, Liu S. Diet and risk of Type II diabetes: the role of types of fat and carbohydrate. Diabetologia. 2001;44(7):805-817. doi:10.1007/s001250100547. 69. Martínez-González MÁ, Fuente-Arrillaga C de la, Nunez-Cordoba JM, et al. Adherence to Mediterranean diet and risk of developing diabetes: prospective cohort study. BMJ. 2008;336(7657). http://www.bmj.com/content/336/7657/1348. Accessed April 29, 2017. 70. Aune D, Norat T, Leitzmann M, Tonstad S, Vatten LJ. Physical activity and the risk of type 2 diabetes: a systematic review and dose–response meta-analysis. Eur J Epidemiol. 2015;30(7):529-542. doi:10.1007/s10654-015-0056-z. 71. Hamilton MT, Hamilton DG, Zderic TW. Sedentary Behavior as a Mediator of Type 2 Diabetes. In: Medicine and Sport Science. Vol 60. ; 2014:11-26. doi:10.1159/000357332. 72. Wareham N, Brage S, Franks P, Abbot R. Physical Activity and Insulin Resistance. In: Insulin Resistance: Insulin Action and Its Disturbances in Disease. John Wiley & Sons, Ltd; 2005:317-360. 73. Carpenter DO. Environmental contaminants as risk factors for developing diabetes. Rev Environ Health. 23(1):59-74. http://www.ncbi.nlm.nih.gov/pubmed/18557598. Accessed April 29, 2017. 74. Patel CJ, Bhattacharya J, Butte AJ, Zeggini E, Freathy R. An Environment-Wide Association Study (EWAS) on Type 2 Diabetes Mellitus. Zhang B, ed. PLoS One. 2010;5(5):e10746. doi:10.1371/journal.pone.0010746.

183 75. Lee D-H, Lind PM, Jacobs DR, Salihovic S, van Bavel B, Lind L. Polychlorinated Biphenyls and Organochlorine Pesticides in Plasma Predict Development of Type 2 Diabetes in the Elderly. Diabetes Care. 2011;34(8):1778-1784. doi:10.2337/dc10-2116. 76. Diamant M, Blaak EE, de Vos WM. Do nutrient-gut-microbiota interactions play a role in human obesity, insulin resistance and type 2 diabetes? Obes Rev. 2011;12(4):272-281. doi:10.1111/j.1467-789X.2010.00797.x. 77. Feig DS, Zinman B, Wang X, Hux JE. Risk of development of diabetes mellitus after diagnosis of gestational diabetes. CMAJ. 2008;179(3):229-234. doi:10.1503/cmaj.080012. 78. Kim C, Newton KM, Knopp RH. Gestational diabetes and the incidence of type 2 diabetes: a systematic review. Diabetes Care. 2002;25(10):1862-1868. http://www.ncbi.nlm.nih.gov/pubmed/12351492. Accessed April 29, 2017. 79. Mozaffarian D, Marfisi R, Levantesi G, et al. Incidence of new-onset diabetes and impaired fasting glucose in patients with recent myocardial infarction and the effect of clinical and lifestyle risk factors. Lancet. 2007;370(9588):667-675. doi:10.1016/S0140- 6736(07)61343-9. 80. Tenenbaum A, Motro M, Fisman EZ, et al. Functional class in patients with heart failure is associated with the development of diabetes. Am J Med. 2003;114(4):271-275. http://www.ncbi.nlm.nih.gov/pubmed/12681453. Accessed April 29, 2017. 81. Mezuk B, Eaton WW, Albrecht S, Golden SH. Depression and type 2 diabetes over the lifespan: a meta-analysis. Diabetes Care. 2008;31(12):2383-2390. doi:10.2337/dc08- 0985. 82. Krishnan S, Cozier YC, Rosenberg L, Palmer JR. Socioeconomic Status and Incidence of Type 2 Diabetes: Results From the Black Women’s Health Study. Am J Epidemiol. 2010;171(5):564-570. doi:10.1093/aje/kwp443. 83. Lee TC, Glynn RJ, Peña JM, et al. Socioeconomic Status and Incident Type 2 Diabetes Mellitus: Data from the Women’s Health Study. Herder C, ed. PLoS One. 2011;6(12):e27670. doi:10.1371/journal.pone.0027670. 84. Stringhini S, Zaninotto P, Kumari M, Kivimäki M, Batty GD. Lifecourse socioeconomic status and type 2 diabetes: the role of chronic inflammation in the English Longitudinal Study of Ageing. Sci Rep. 2016;6:24780. doi:10.1038/srep24780. 85. Kahn SE, Cooper ME, Del Prato S. Pathophysiology and treatment of type 2 diabetes:

184 perspectives on the past, present, and future. Lancet. 2014;383(9922):1068-1083. doi:10.1016/S0140-6736(13)62154-6. 86. American Diabetes Association. Classification and Diagnosis of Diabetes. Diabetes Care. 2015;38(Supplement 1):S8-S16. 87. American Diabetes Association. Diagnosis and Classification of Diabetes Mellitus. Diabetes Care. 2013;36(1):S67-S74. doi:10.2337/dc13-S067. 88. Sacks DB. A1C versus glucose testing: a comparison. Diabetes Care. 2011;34(2):518- 523. doi:10.2337/dc10-1546. 89. Selvin E, Crainiceanu CM, Brancati FL, Coresh J. Short-term Variability in Measures of Glycemia and Implications for the Classification of Diabetes. Arch Intern Med. 2007;167(14):1545. doi:10.1001/archinte.167.14.1545. 90. International Expert Committee. International Expert Committee report on the role of the A1C assay in the diagnosis of diabetes. Diabetes Care. 2009;32(7):1327-1334. 91. Khaw K-T, Wareham N, Bingham S, Luben R, Welch A, Day N. Association of Hemoglobin A_1c with Cardiovascular Disease and Mortality in Adults: The European Prospective Investigation into Cancer in Norfolk. Ann Intern Med. 2004;141(6):413. doi:10.7326/0003-4819-141-6-200409210-00006. 92. Selvin E, Steffes MW, Zhu H, et al. Glycated Hemoglobin, Diabetes, and Cardiovascular Risk in Nondiabetic Adults. N Engl J Med. 2010;362(9):800-811. doi:10.1056/NEJMoa0908359. 93. Classification and Diagnosis of Diabetes. Diabetes Care. 2014;38(Supplement 1). 94. Picón MJ, Murri M, Muñoz A, Fernández-García JC, Gomez-Huelgas R, Tinahones FJ. Hemoglobin A1c versus oral glucose tolerance test in postpartum diabetes screening. Diabetes Care. 2012;35(8):1648-1653. doi:10.2337/dc11-2111. 95. Metcalf PA, Kyle C, Kenealy T, Sundborn G, Jackson RT, Metcalf P. Comparison of Fasting Glucose, and HbA1c for Use in Screening for Type 2 Diabetes. Diabetes Res Metab. 2016;1(1). http://crescopublications.org/drmoa/DRMOA. Accessed February 8, 2017. 96. DeFronzo R, Tobin J, Andres R. Glucose clamp technique: a method for quantifying insulin secretion and resistance. Am J Physiol - Gastrointest Liver Physiol. 1979;237(3). 97. Tam CS, Xie W, Johnson WD, Cefalu WT, Redman LM, Ravussin E. Defining Insulin

185 Resistance From Hyperinsulinemic-Euglycemic Clamps. Diabetes Care. 2012;35(7). 98. Bergman RN, Finegood DT, Ader M. Assessment of insulin sensitivity in vivo. Endocr Rev. 1985;6(1):45-86. doi:10.1210/edrv-6-1-45. 99. Bergman RN, Finegood DT, Kahn SE. The evolution of β-cell dysfunction and insulin resistance in type 2 diabetes. Eur J Clin Investig Eur J Clin Invest. 2002;32(32):35-45. 100. Diagnosis and Classification of Diabetes Mellitus. Diabetes Care. 2014;37(Supplement 1). doi:10.2337/dc14-S081. 101. Nathan DM, Davidson MB, DeFronzo RA, et al. Impaired Fasting Glucose and Impaired Glucose Tolerance. Diabetes Care. 2007;30(3). http://care.diabetesjournals.org/content/30/3/753.full#ref-4. Accessed April 5, 2017. 102. American Diabetes Association Standards of Medical Care in Diabetes - 2017. Diabetes Care. 2017;40(Supplement 1). 103. Powers MA, Bardsley J, Cypress M, et al. Diabetes Self-management Education and Support in Type 2 Diabetes: A Joint Position Statement of the American Diabetes Association, the American Association of Diabetes Educators, and the Academy of Nutrition and Dietetics. Diabetes Care. 2015. http://care.diabetesjournals.org/content/early/2015/06/02/dc15-0730. Accessed April 25, 2017. 104. Boulé NG, Haddad E, Kenny GP, Wells GA, Sigal RJ. Effects of Exercise on Glycemic Control and Body Mass in Type 2 Diabetes Mellitus. JAMA. 2001;286(10):1218. doi:10.1001/jama.286.10.1218. 105. Centers for Disease Control and Prevention. Age-Adjusted Percentage of Adults with Diabetes Using Diabetes Medication, by Type of Medication, United States, 1997-2011.; 2013. 106. Sartor G, Scherstén B, Carlström S, Melander A, Nordén Å, Persson G. Ten-year Follow- up of Subjects with Impaired Glucose Tolerance: Prevention of Diabetes by Tolbutamide and Diet Regulation. Diabetes. 1980;29(1). http://diabetes.diabetesjournals.org/content/29/1/41. Accessed April 13, 2017. 107. Eriksson KF, Lindgärde F. Prevention of type 2 (non-insulin-dependent) diabetes mellitus by diet and physical exercise. The 6-year Malmö feasibility study. Diabetologia. 1991;34(12):891-898. http://www.ncbi.nlm.nih.gov/pubmed/1778354. Accessed April 13,

186 2017. 108. Pan X-R, Li G-W, Hu Y-H, et al. Effects of Diet and Exercise in Preventing NIDDM in People With Impaired Glucose Tolerance: The Da Qing IGT and Diabetes Study. Diabetes Care. 1997;20(4). http://care.diabetesjournals.org/content/20/4/537. Accessed April 13, 2017. 109. Lindström J, Ilanne-Parikka P, Peltonen M, et al. Sustained reduction in the incidence of type 2 diabetes by lifestyle intervention: follow-up of the Finnish Diabetes Prevention Study. Lancet (London, England). 2006;368(9548):1673-1679. doi:10.1016/S0140- 6736(06)69701-8. 110. Lindstrom J, Peltonen M, Eriksson JG, et al. Improved lifestyle and decreased diabetes risk over 13 years: long-term follow-up of the randomised Finnish Diabetes Prevention Study (DPS). Diabetologia. 2013;56(2):284-293. doi:10.1007/s00125-012-2752-5. 111. Goldman L, Schafer A. Cecil Medicine (24th Edition). 24th ed. Philadephia: Saunders, Elsevier Inc.; 2011. 112. Stratton IM, Adler AI, Neil HAW, et al. Association of glycaemia with macrovascular and microvascular complications of type 2 diabetes (UKPDS 35): prospective observational study. BMJ. 2000;321(7258). 113. Vinik A, Flemmer M. Diabetes and macrovascular disease. J Diabetes Complications. 2002;16(3):235-245. doi:10.1016/S1056-8727(01)00212-4. 114. Wu L, Parhofer KG. Diabetic dyslipidemia. Metabolism. 2014;63(12):1469-1479. doi:10.1016/j.metabol.2014.08.010. 115. Aune D, Norat T, Romundstad P, Vatten LJ. Whole grain and refined grain consumption and the risk of type 2 diabetes: a systematic review and dose–response meta-analysis of cohort studies. Eur J Epidemiol. 2013;28(11):845-858. doi:10.1007/s10654-013-9852-5. 116. Hu EA, Pan A, Malik V, Sun Q. White rice consumption and risk of type 2 diabetes: meta-analysis and systematic review. BMJ. 2012;344. http://www.bmj.com/content/344/bmj.e1454. Accessed May 19, 2017. 117. Carter P, Gray LJ, Troughton J, Khunti K, Davies MJ. Fruit and vegetable intake and incidence of type 2 diabetes mellitus: systematic review and meta-analysis. BMJ. 2010;341. http://www.bmj.com/content/341/bmj.c4229. Accessed May 19, 2017. 118. Cooper AJ, Forouhi NG, Ye Z, et al. Fruit and vegetable intake and type 2 diabetes: EPIC-

187 InterAct prospective study and meta-analysis. Eur J Clin Nutr. 2012;66(10):1082-1092. doi:10.1038/ejcn.2012.85. 119. Muraki I, Imamura F, Manson JE, et al. Fruit consumption and risk of type 2 diabetes: results from three prospective longitudinal cohort studies. BMJ. 2013;347. http://www.bmj.com/content/347/bmj.f5001. Accessed May 19, 2017. 120. Hu FB, Liu S, van Dam RM. Diet and risk of Type II diabetes: the role of types of fat and carbohydrate. Diabetologia. 2001;44(7):805-817. doi:10.1007/s001250100547. 121. Hauner H, Bechthold A, Boeing H, et al. Evidence-Based Guideline of the German Nutrition Society: Carbohydrate Intake and Prevention of Nutrition-Related Diseases. Ann Nutr Metab. 2012;60(s1):1-58. doi:10.1159/000335326. 122. Schulze MB, Schulz M, Heidemann C, Schienkiewitz A, Hoffmann K, Boeing H. Fiber and magnesium intake and incidence of type 2 diabetes: a prospective study and meta- analysis. Arch Intern Med. 2007;167. doi:10.1001/archinte.167.9.956. 123. Tong X, Dong J-Y, Wu Z-W, Li W, Qin L-Q. Dairy consumption and risk of type 2 diabetes mellitus: a meta-analysis of cohort studies. Eur J Clin Nutr. 2011;65(9):1027- 1031. doi:10.1038/ejcn.2011.62. 124. Pan A, Sun Q, Bernstein AM, et al. Red meat consumption and risk of type 2 diabetes: 3 cohorts of US adults and an updated meta-analysis. Am J Clin Nutr. 2011;94(4):1088- 1096. doi:10.3945/ajcn.111.018978. 125. Wu JHY, Micha R, Imamura F, et al. Omega-3 fatty acids and incident type 2 diabetes: a systematic review and meta-analysis. Br J Nutr. 2012;107(S2):S214-S227. doi:10.1017/S0007114512001602. 126. Jiang R, Manson JE, Stampfer MJ, Liu S, Willett WC, Hu FB. Nut and Peanut Butter Consumption and Risk of Type 2 Diabetes in Women. JAMA. 2002;288(20):2554. doi:10.1001/jama.288.20.2554. 127. Pan A, Sun Q, Manson JE, Willett WC, Hu FB. Walnut Consumption Is Associated with Lower Risk of Type 2 Diabetes in Women. J Nutr. 2013;143(4):512-518. doi:10.3945/jn.112.172171. 128. RISERUS U, WILLETT W, HU F. Dietary fats and prevention of type 2 diabetes. Prog Lipid Res. 2009;48(1):44-51. doi:10.1016/j.plipres.2008.10.002. 129. Tinker LF, Bonds DE, Margolis KL, et al. Low-Fat Dietary Pattern and Risk of Treated

188 Diabetes Mellitus in Postmenopausal Women<subtitle>The Women’s Health Initiative Randomized Controlled Dietary Modification Trial</subtitle> Arch Intern Med. 2008;168(14):1500. doi:10.1001/archinte.168.14.1500. 130. Salmeron J, Hu FB, Manson JE, et al. Dietary fat intake and risk of type 2 diabetes in women. Am J Clin Nutr. 2001;73(6):1019-1026. http://www.ncbi.nlm.nih.gov/pubmed/11382654. 131. Hu FB. Dietary pattern analysis: a new direction in nutritional epidemiology. Curr Opin Lipidol. 2002;13(1):3-9. doi:10.1097/00041433-200202000-00002. 132. Willett WC. Nutritional Epidemiology. 3rd Editio. New York: Oxford University Press; 2013. 133. Jacobs DR, Tapsell LC. Food, not nutrients, is the fundamental unit in nutrition. Nutr Rev. 2007;65(10):439-450. http://www.ncbi.nlm.nih.gov/pubmed/17972438. Accessed May 19, 2017. 134. Hoffmann K, Schulze MB, Schienkiewitz A, Nöthlings U, Boeing H. Application of a New Statistical Method to Derive Dietary Patterns in Nutritional Epidemiology. Am J Epidemiol. 2004;159(10):935-944. doi:10.1093/aje/kwh134. 135. Guenther PM, Reedy J, Krebs-Smith SM. Development of the Healthy Eating Index-2005. J Am Diet Assoc. 2008;108(11):1896-1901. doi:10.1016/j.jada.2008.08.016. 136. Guenther PM, Casavale KO, Reedy J, et al. Update of the Healthy Eating Index: HEI- 2010. J Acad Nutr Diet. 2013;113(4):569-580. doi:10.1016/j.jand.2012.12.016. 137. McCullough ML, Feskanich D, Stampfer MJ, et al. Diet quality and major chronic disease risk in men and women: moving toward improved dietary guidance. Am J Clin Nutr. 2002;76(6):1261-1271. http://www.ncbi.nlm.nih.gov/pubmed/12450892. Accessed June 12, 2017. 138. Chiuve SE, Fung TT, Rimm EB, et al. Alternative Dietary Indices Both Strongly Predict Risk of Chronic Disease. J Nutr. 2012;142(6):1009-1018. doi:10.3945/jn.111.157222. 139. Keys A. Seven Countries. A Multivariate Analysis of Death and Coronary Heart Disease. Cambridge : Harvard University Press; 1980. https://www.cabdirect.org/cabdirect/abstract/19841460563. Accessed July 6, 2017. 140. Trichopoulou A, Kouris-Blazos A, Wahlqvist ML, et al. Diet and overall survival in elderly people. BMJ. 1995;311(7018). http://www.bmj.com/content/311/7018/1457.

189 Accessed July 6, 2017. 141. Fung TT, McCullough ML, Newby PK, et al. Diet-quality scores and plasma concentrations of markers of inflammation and endothelial dysfunction. Am J Clin Nutr. 2005;82(1):163-173. http://www.ncbi.nlm.nih.gov/pubmed/16002815. 142. Buckland G, Agudo A, Lujan L, et al. Adherence to a Mediterranean diet and risk of gastric adenocarcinoma within the European Prospective Investigation into Cancer and Nutrition (EPIC) cohort study. Am J Clin Nutr. 2010;91(2):381-390. doi:10.3945/ajcn.2009.28209. 143. Rumawas ME, Dwyer JT, McKeown NM, Meigs JB, Rogers G, Jacques PF. The development of the Mediterranean-style dietary pattern score and its application to the American diet in the Framingham Offspring Cohort. J Nutr. 2009;139(6):1150-1156. doi:10.3945/jn.108.103424. 144. U.S. Department of Health and Human Service. Your Guide to Lowering Your Blood Pressure with DASH.; 2006. doi:10.1037/e566492010-001. 145. Fung TT, Chiuve SE, McCullough ML, Rexrode KM, Logroscino G, Hu FB. Adherence to a DASH-style diet and risk of coronary heart disease and stroke in women. Arch Intern Med. 2008;168(7):713-720. doi:10.1001/archinte.168.7.713. 146. Jacques PF, Tucker KL. Are dietary patterns useful for understanding the role of diet in chronic disease? Am J Clin Nutr. 2001;73(1):1-2. http://www.ncbi.nlm.nih.gov/pubmed/11124739. 147. Esposito K, Kastorini C-M, Panagiotakos DB, Giugliano D. Prevention of Type 2 Diabetes by Dietary Patterns: A Systematic Review of Prospective Studies and Meta- Analysis. Metab Syndr Relat Disord. 2010;8(6):471-476. doi:10.1089/met.2010.0009. 148. Alhazmi A, Stojanovski E, McEvoy M, Garg ML. The association between dietary patterns and type 2 diabetes: a systematic review and meta-analysis of cohort studies. J Hum Nutr Diet. 2014;27(3):251-260. doi:10.1111/jhn.12139. 149. Jannasch F, Kröger J, Schulze MB. Dietary Patterns and Type 2 Diabetes: A Systematic Literature Review and Meta-Analysis of Prospective Studies. J Nutr. 2017;147(6):1174- 1182. doi:10.3945/jn.116.242552. 150. The Inter Act Consortium. Mediterranean Diet and Type 2 Diabetes Risk in the European Prospective Investigation Into Cancer and Nutrition (EPIC) Study THE INTERACT

190 CONSORTIUM*. Diabetes Care. 2011;34(9):1913-1918. doi:10.2337/dc11-0891. 151. Rossi M, Turati F, Lagiou P, et al. Mediterranean diet and glycaemic load in relation to incidence of type 2 diabetes: results from the Greek cohort of the population-based European Prospective Investigation into Cancer and Nutrition (EPIC). Diabetologia. 2013;56(11):2405-2413. doi:10.1007/s00125-013-3013-y. 152. Fung TT, McCullough M, van Dam RM, Hu FB. A Prospective Study of Overall Diet Quality and Risk of Type 2 Diabetes in Women. Diabetes Care. 2007;30(7):1753-1757. doi:10.2337/dc06-2581. 153. Qiao Y, Tinker L, Olendzki BC, et al. Racial/ethnic disparities in association between dietary quality and incident diabetes in postmenopausal women in the United States: the Women’s Health Initiative 1993-2005. Ethn Health. 2014;19(3):328-347. doi:10.1080/13557858.2013.797322. 154. de Koning L, Chiuve SE, Fung TT, Willett WC, Rimm EB, Hu FB. Diet-Quality Scores and the Risk of Type 2 Diabetes in Men. Diabetes Care. 2011;34(5):1150-1156. doi:10.2337/dc10-2352. 155. Jacobs S, Harmon BE, Boushey CJ, et al. A priori-defined diet quality indexes and risk of type 2 diabetes: the Multiethnic Cohort. Diabetologia. 2015;58(1):98-112. doi:10.1007/s00125-014-3404-8. 156. Cespedes EM, Hu FB, Tinker L, et al. Multiple Healthful Dietary Patterns and Type 2 Diabetes in the Women’s Health Initiative. Am J Epidemiol. 2016;183(7):622-633. doi:10.1093/aje/kwv241. 157. Tonstad S, Stewart K, Oda K, Batech M, Herring RP, Fraser GE. Vegetarian diets and incidence of diabetes in the Adventist Health Study-2. Nutr Metab Cardiovasc Dis. 2013;23(4):292-299. doi:10.1016/j.numecd.2011.07.004. 158. Duffey KJ, Popkin BM. Shifts in patterns and consumption of beverages between 1965 and 2002. Obes (Silver Spring). 2007;15(11):2739-2747. doi:10.1038/oby.2007.326. 159. Menke A, Casagrande S, Geiss L, Cowie CC. Prevalence of and Trends in Diabetes Among Adults in the United States, 1988-2012. JAMA. 2015;314(10):1021-1029. doi:10.1001/jama.2015.10029. 160. Basu S, McKee M, Galea G, Stuckler D. Relationship of Soft Drink Consumption to Global Overweight, Obesity, and Diabetes: A Cross-National Analysis of 75 Countries.

191 Am J Public Health. 2013;103(11):2071-2077. doi:10.2105/AJPH.2012.300974. 161. Malik VS, Popkin BM, Bray GA, Despres JP, Willett WC, Hu FB. Sugar-sweetened beverages and risk of metabolic syndrome and type 2 diabetes: a meta-analysis. Diabetes Care. 2010;33(11):2477-2483. doi:10.2337/dc10-1079. 162. Greenwood DC, Threapleton DE, Evans CEL, et al. Association between sugar-sweetened and artificially sweetened soft drinks and type 2 diabetes: systematic review and dose– response meta-analysis of prospective studies. Br J Nutr. 2014;112(5):725-734. doi:10.1017/S0007114514001329. 163. Imamura F, O’Connor L, Ye Z, et al. Consumption of sugar sweetened beverages, artificially sweetened beverages, and fruit juice and incidence of type 2 diabetes: systematic review, meta-analysis, and estimation of population attributable fraction. BMJ. 2015;351:h3576. doi:10.1136/bmj.h3576. 164. Malik VS, Hu FB. Sweeteners and Risk of Obesity and Type 2 Diabetes: The Role of Sugar-Sweetened Beverages. Curr Diab Rep. 2012. doi:10.1007/s11892-012-0259-6. 165. Maersk M, Belza A, Stodkilde-Jorgensen H, et al. Sucrose-sweetened beverages increase fat storage in the liver, muscle, and visceral fat depot: a 6-mo randomized intervention study. Am J Clin Nutr. 2012;95(2):283-289. doi:10.3945/ajcn.111.022533. 166. InterAct C, Romaguera D, Norat T, et al. Consumption of sweet beverages and type 2 diabetes incidence in European adults: results from EPIC-InterAct. Diabetologia. 2013;56(7):1520-1530. doi:10.1007/s00125-013-2899-8. 167. Hu FB, Malik VS. Sugar-sweetened beverages and risk of obesity and type 2 diabetes: epidemiologic evidence. Physiol Behav. 2010;100(1):47-54. doi:10.1016/j.physbeh.2010.01.036. 168. Dhingra R, Sullivan L, Jacques PF, et al. Soft drink consumption and risk of developing cardiometabolic risk factors and the metabolic syndrome in middle-aged adults in the community. Circulation. 2007;116(5):480-488. doi:10.1161/CIRCULATIONAHA.107.689935. 169. Lutsey PL, Steffen LM, Stevens J. Dietary intake and the development of the metabolic syndrome: the atherosclerosis risk in communities study. Circulation. 2008;117. doi:10.1161/circulationaha.107.716159. 170. Nettleton JA, Lutsey PL, Wang Y, Lima JA, Michos ED, Jacobs Jr. DR. Diet soda intake

192 and risk of incident metabolic syndrome and type 2 diabetes in the Multi-Ethnic Study of Atherosclerosis (MESA). Diabetes Care. 2009;32(4):688-694. doi:10.2337/dc08-1799. 171. Duffey KJ, Steffen LM, Horn L Van, Jacobs DR, Popkin BM. Dietary patterns matter: diet beverages and cardiometabolic risks in the longitudinal Coronary Artery Risk Development in Young Adults (CARDIA) Study 1–3. Am J Clin Nutr. 2012;95:909-915. doi:10.3945/ajcn.111.026682. 172. Pereira MA. Sugar-sweetened and artificially-sweetened beverages in relation to obesity risk. Adv Nutr. 2014;5(6):797-808. doi:10.3945/an.114.007062. 173. Lutsey PL, Steffen LM, Stevens J. Dietary intake and the development of the metabolic syndrome: the Atherosclerosis Risk in Communities study. Circulation. 2008;117(6):754- 761. doi:10.1161/CIRCULATIONAHA.107.716159. 174. Dhingra R, Sullivan L, Jacques PF, et al. Soft Drink Consumption and Risk of Developing Cardiometabolic Risk Factors and the Metabolic Syndrome in Middle-Aged Adults in the Community. Circulation. 2007;116(5):480-488. doi:10.1161/CIRCULATIONAHA.107.689935. 175. Davidson TL, Swithers SE. A Pavlovian approach to the problem of obesity. Int J Obes. 2004;28(7):933-935. doi:10.1038/sj.ijo.0802660. 176. Swithers SE, Davidson TL. A role for sweet taste: Calorie predictive relations in energy regulation by rats. Behav Neurosci. 2008;122(1):161-173. doi:10.1037/0735- 7044.122.1.161. 177. Swithers SE, Martin AA, Davidson TL. High-intensity sweeteners and energy balance. Physiol Behav. 2010;100(1):55-62. doi:10.1016/j.physbeh.2009.12.021. 178. Palmer JR, Boggs DA, Krishnan S, Hu FB, Singer M, Rosenberg L. Sugar-Sweetened Beverages and Incidence of Type 2 Diabetes Mellitus in African American Women. Arch Intern Med. 2008;168(14):1487. doi:10.1001/archinte.168.14.1487. 179. de Koning L, Malik VS, Rimm EB, Willett WC, Hu FB. Sugar-sweetened and artificially sweetened beverage consumption and risk of type 2 diabetes in men. Am J Clin Nutr. 2011;93(6):1321-1327. doi:10.3945/ajcn.110.007922. 180. Pan A, Malik VS, Schulze MB, Manson JE, Willett WC, Hu FB. Plain-water intake and risk of type 2 diabetes in young and middle-aged women. Am J Clin Nutr. 2012;95(6):1454-1460. doi:10.3945/ajcn.111.032698.

193 181. Fagherazzi G, Vilier A, Saes Sartorelli D, Lajous M, Balkau B, Clavel-Chapelon F. Consumption of artificially and sugar-sweetened beverages and incident type 2 diabetes in the Etude Epidemiologique aupres des femmes de la Mutuelle Generale de l’Education Nationale-European Prospective Investigation into Cancer and Nutrition cohort. Am J Clin Nutr. 2013;97(3):517-523. doi:10.3945/ajcn.112.050997. 182. Sakurai M, Nakamura K, Miura K, et al. Sugar-sweetened beverage and diet soda consumption and the 7-year risk for type 2 diabetes mellitus in middle-aged Japanese men. Eur J Nutr. 2014;53(4):1137-1138. doi:10.1007/s00394-014-0681-4. 183. Matheus AS de M, Tannus LRM, Cobas RA, Palma CCS, Negrato CA, Gomes M de B. Impact of Diabetes on Cardiovascular Disease: An Update. Int J Hypertens. 2013;2013:1- 15. doi:10.1155/2013/653789. 184. American Diabetes Association. Cardiovascular disease and risk management. Sec. 9. In Standards of Medical Care in Diabetes-2017. Diabetes Care. 2017;40(Suppl. 1):S75-S87. 185. Goff DC, Lloyd-Jones DM, Bennett G, et al. 2013 ACC/AHA Guideline on the Assessment of Cardiovascular Risk. Circulation. 2014;129(25 suppl 2). http://circ.ahajournals.org/content/129/25_suppl_2/S49. Accessed April 29, 2017. 186. Yelle D. Atherosclerosis. In: Chaundhry S, Wong E, eds. McMaster Pathophysiology Review. ; 2012. 187. Falk E. Pathogenesis of Atherosclerosis. J Am Coll Cardiol. 2006;47(8):C7-C12. doi:10.1016/j.jacc.2005.09.068. 188. Libby P. Inflammation in atherosclerosis. Nature. 2002;420(6917):868-874. doi:10.1038/nature01323. 189. Hansson GK. Inflammation, Atherosclerosis, and Coronary Artery Disease. N Engl J Med. 2005;352(16):1685-1695. doi:10.1056/NEJMra043430. 190. National Heart, Lung and BI. What is Atherosclerosis? NIH. https://www.nhlbi.nih.gov/health/health-topics/topics/atherosclerosis. Published 2016. Accessed April 30, 2017. 191. World Health Organization. Global Status Report on Noncommunicable Diseases 2010. Geneva; 2011. 192. Glass CK, Witztum JL. Atherosclerosis: The Road Ahead. Cell. 2001;104(4):503-516. doi:10.1016/S0092-8674(01)00238-0.

194 193. Leon BM, Maddox TM. Diabetes and cardiovascular disease: Epidemiology, biological mechanisms, treatment recommendations and future research. World J Diabetes. 2015;6(613):1246-1258. doi:10.4239/wjd.v6.i13.1246. 194. Eckel RH, Wassef M, Chait A, et al. Prevention Conference VI: Diabetes and Cardiovascular Disease. Circulation. 2002;105(18). http://circ.ahajournals.org/content/105/18/e138.long. Accessed April 29, 2017. 195. Shoelson SE, Lee J, Goldfine AB. Inflammation and insulin resistance. J Clin Invest. 2006;116(7):1793-1801. doi:10.1172/JCI29069. 196. Duncan BB, Schmidt MI, Pankow JS, et al. Low-grade systemic inflammation and the development of type 2 diabetes: the atherosclerosis risk in communities study. Diabetes. 2003;52(7):1799-1805. http://www.ncbi.nlm.nih.gov/pubmed/12829649. Accessed April 29, 2017. 197. Ferrarezi DAF, Cheurfa N, Reis AF, Fumeron F, Velho G. Adiponectin gene and cardiovascular risk in type 2 diabetic patients: a review of evidences. Arq Bras Endocrinol Metabol. 2007;51(2):153-159. doi:10.1590/S0004-27302007000200003. 198. The Diabetes Control and Complications Trial Research Group. The Effect of Intensive Treatment of Diabetes on the Development and Progression of Long-Term Complications in Insulin-Dependent Diabetes Mellitus. N Engl J Med. 1993;329(14):977-986. doi:10.1056/NEJM199309303291401. 199. Ram C V. Hypertension and atherosclerosis. Prim Care. 1991;18(3):559-575. http://www.ncbi.nlm.nih.gov/pubmed/1946788. Accessed May 1, 2017. 200. Cheung BMY, Li C. Diabetes and hypertension: is there a common metabolic pathway? Curr Atheroscler Rep. 2012;14(2):160-166. doi:10.1007/s11883-012-0227-2. 201. Kannel WB. Lipids, diabetes, and coronary heart disease: insights from the Framingham Study. Am Heart J. 1985;110(5):1100-1107. http://www.ncbi.nlm.nih.gov/pubmed/4061265. Accessed April 29, 2017. 202. Goldberg IJ. Diabetic Dyslipidemia: Causes and Consequences. J Clin Endocrinol Metab. 2001;86(3):965-971. doi:10.1210/jcem.86.3.7304. 203. Mooradian AD. Dyslipidemia in type 2 diabetes mellitus. Nat Clin Pract Endocrinol Metab. 2009;5(3):150-159. doi:10.1038/ncpendmet1066. 204. Taskinen M-R. Diabetic dyslipidaemia: from basic research to clinical practice*.

195 Diabetologia. 2003;46(6):733-749. doi:10.1007/s00125-003-1111-y. 205. Krauss RM, Siri PW. Dyslipidemia in type 2 diabetes. Med Clin North Am. 2004;88(4):897-909. doi:10.1016/j.mcna.2004.04.004. 206. Chahil TJ, Ginsberg HN. Diabetic Dyslipidemia. Endocrinol Metab Clin North Am. 2006;35(3):491-510. doi:10.1016/j.ecl.2006.06.002. 207. Mooradian AD, Albert SG, Haas MJ. Low serum high-density lipoprotein cholesterol in obese subjects with normal serum triglycerides: the role of insulin resistance and inflammatory cytokines. Diabetes, Obes Metab. 2007;9(3):441-443. doi:10.1111/j.1463- 1326.2006.00636.x. 208. Schofield JD, Liu Y, Rao-Balakrishna P, Malik RA, Soran H. Diabetes Dyslipidemia. Diabetes Ther. 2016;7(2):203-219. doi:10.1007/s13300-016-0167-x. 209. The ACCORD Study Group. Effects of Intensive Blood-Pressure Control in Type 2 Diabetes Mellitus. N Engl J Med. 2010;362(17):1575-1585. doi:10.1056/NEJMoa1001286. 210. Margolis KL, O’Connor PJ, Morgan TM, et al. Outcomes of Combined Cardiovascular Risk Factor Management Strategies in Type 2 Diabetes: The ACCORD Randomized Trial. Diabetes Care. 2014;37(6). http://care.diabetesjournals.org/content/37/6/1721.long. Accessed April 13, 2017. 211. Patel A, ADVANCE Collaborative Group, MacMahon S, et al. Effects of a fixed combination of perindopril and indapamide on macrovascular and microvascular outcomes in patients with type 2 diabetes mellitus (the ADVANCE trial): a randomised controlled trial. Lancet. 2007;370(9590):829-840. doi:10.1016/S0140-6736(07)61303-8. 212. Sacks FM, Svetkey LP, Vollmer WM, et al. Effects on Blood Pressure of Reduced Dietary Sodium and the Dietary Approaches to Stop Hypertension (DASH) Diet. N Engl J Med. 2001;344(1):3-10. doi:10.1056/NEJM200101043440101. 213. The Look AHEAD Research Group. Cardiovascular Effects of Intensive Lifestyle Intervention in Type 2 Diabetes. N Engl J Med. 2013;369(2):145-154. doi:10.1056/NEJMoa1212914. 214. Expert Panel on Detection, Evaluation, and Treatment of High Blood Cholesterol in Adults E and T of HBC in A. Executive Summary of the Third Report of the National Cholesterol Education Program (NCEP) Expert Panel on Detection, Evaluation, and

196 Treatment of High Blood Cholesterol in Adults (Adult Treatment Panel III). JAMA J Am Med Assoc. 2001;285(19):2486-2497. doi:10.1001/jama.285.19.2486. 215. Fox CS, Golden SH, Anderson C, et al. Update on Prevention of Cardiovascular Disease in Adults With Type 2 Diabetes Mellitus in Light of Recent Evidence: A Scientific Statement From the American Heart Association and the American Diabetes Association. Diabetes Care. 2015. http://care.diabetesjournals.org/content/early/2015/08/05/dci15- 0012.1. Accessed April 13, 2017. 216. Li G, Zhang P, Wang J, et al. The long-term effect of lifestyle interventions to prevent diabetes in the China Da Qing Diabetes Prevention Study: a 20-year follow-up study. Lancet. 2008;371(9626):1783-1789. doi:10.1016/S0140-6736(08)60766-7. 217. Li G, Zhang P, Wang J, et al. Cardiovascular mortality, all-cause mortality, and diabetes incidence after lifestyle intervention for people with impaired glucose tolerance in the Da Qing Diabetes Prevention Study: a 23-year follow-up study. Lancet Diabetes Endocrinol. 2014;2(6):474-480. doi:10.1016/S2213-8587(14)70057-9. 218. Uusitupa M, Peltonen M, Lindström J, et al. Ten-Year Mortality and Cardiovascular Morbidity in the Finnish Diabetes Prevention Study—Secondary Analysis of the Randomized Trial. Sorensen TIA, ed. PLoS One. 2009;4(5):e5656. doi:10.1371/journal.pone.0005656. 219. Gæde P, Vedel P, Larsen N, Jensen GVH, Parving H-H, Pedersen O. Multifactorial Intervention and Cardiovascular Disease in Patients with Type 2 Diabetes. N Engl J Med. 2003;348(5):383-393. doi:10.1056/NEJMoa021778. 220. Sone H, Tanaka S, Iimuro S, Oida K, Yamasaki Y. Long-term lifestyle intervention lowers the incidence of stroke in Japanese patients with type 2 diabetes: a nationwide multicentre randomised controlled trial (the Japan Diabetes Complications Study). Diabetologia. 2010;53. doi:10.1007/s00125-009-1622-2. 221. Mann JI, De Leeuw I, Hermansen K, et al. Evidence-based nutritional approaches to the treatment and prevention of diabetes mellitus. Nutr Metab Cardiovasc Dis. 2004;14(6):373-394. http://www.ncbi.nlm.nih.gov/pubmed/15853122. Accessed May 8, 2017. 222. Dworatzek PD, Arcudi K, Gougeon R, Husein N, Sievenpiper JL, Williams SL. Nutrition Therapy. Can J Diabetes. 2013;37:S45-S55. doi:10.1016/j.jcjd.2013.01.019.

197 223. Moore H, Summerbell C, Hooper L, et al. Dietary advice for treatment of type 2 diabetes mellitus in adults. In: Moore H, ed. Cochrane Database of Systematic Reviews. Chichester, UK: John Wiley & Sons, Ltd; 2004:CD004097. doi:10.1002/14651858.CD004097.pub3. 224. Wolfram T, Ismail-Beigi F. Efficacy of High-Fiber Diets in The Management of Type 2 Diabetes Mellitus. Endocr Pract. 2011;17(1):132-142. doi:10.4158/EP10204.RA. 225. Serra-Majem L, Roman B, Estruch R. Scientific evidence of interventions using the Mediterranean diet: a systematic review. Nutr Rev. 2006;64(2 Pt 2):S27-47. http://www.ncbi.nlm.nih.gov/pubmed/16532897. Accessed May 15, 2017. 226. Brand-miller J, Hayne S, Petocz P, Colagiuri S. Low–Glycemic Index Diets in the Management of Diabetes A meta-analysis of randomized controlled trials. http://care.diabetesjournals.org/content/diacare/26/8/2261.full.pdf. Accessed May 15, 2017. 227. Dyson PA. A review of low and reduced carbohydrate diets and weight loss in type 2 diabetes. J Hum Nutr Diet. 2008;21(6):530-538. doi:10.1111/j.1365-277X.2008.00896.x. 228. Ajala O, English P, Pinkney J. Systematic review and meta-analysis of different dietary approaches to the management of type 2 diabetes. Am J Clin Nutr. 2013;97(3):505-516. doi:10.3945/ajcn.112.042457. 229. Arora SK, McFarlane SI. The case for low carbohydrate diets in diabetes management. Nutr Metab (Lond). 2005;2(1):16. doi:10.1186/1743-7075-2-16. 230. Nordmann AJ, Nordmann A, Briel M, et al. Effects of Low-Carbohydrate vs Low-Fat Diets on Weight Loss and Cardiovascular Risk Factors. Arch Intern Med. 2006;166(3):285. doi:10.1001/archinte.166.3.285. 231. Wheeler ML, Dunbar SA, Jaacks LM, et al. Macronutrients, Food Groups, and Eating Patterns in the Management of Diabetes: A systematic review of the literature, 2010. Diabetes Care. 2012;35(2):434-445. doi:10.2337/dc11-2216. 232. Westman EC, Yancy WS, Mavropoulos JC, Marquart M, McDuffie JR, McDuffie JR. The effect of a low-carbohydrate, ketogenic diet versus a low-glycemic index diet on glycemic control in type 2 diabetes mellitus. Nutr Metab (Lond). 2008;5:36. doi:10.1186/1743- 7075-5-36. 233. Wolever TMS, Gibbs AL, Mehling C, et al. The Canadian Trial of Carbohydrates in

198 Diabetes (CCD), a 1-y controlled trial of low-glycemic-index dietary carbohydrate in type 2 diabetes: no effect on glycated hemoglobin but reduction in C-reactive protein. Am J Clin Nutr. 2008;87(1):114-125. http://www.ncbi.nlm.nih.gov/pubmed/18175744. Accessed May 4, 2017. 234. Davis NJ, Tomuta N, Schechter C, et al. Comparative Study of the Effects of a 1-Year Dietary Intervention of a Low-Carbohydrate Diet Versus a Low-Fat Diet on Weight and Glycemic Control in Type 2 Diabetes. Diabetes Care. 2009;32(7):1147-1152. doi:10.2337/dc08-2108. 235. Elhayany A, Lustman A, Abel R, Attal-Singer J, Vinker S. A low carbohydrate Mediterranean diet improves cardiovascular risk factors and diabetes control among overweight patients with type 2 diabetes mellitus: a 1-year prospective randomized intervention study. Diabetes, Obes Metab. 2010;12(3):204-209. doi:10.1111/j.1463- 1326.2009.01151.x. 236. Iqbal N, Vetter ML, Moore RH, et al. Effects of a Low-intensity Intervention That Prescribed a Low-carbohydrate vs. a Low-fat Diet in Obese, Diabetic Participants. Obesity. 2010;18(9):1733-1738. doi:10.1038/oby.2009.460. 237. Haimoto H, Iwata M, Wakai K, Umegaki H. Long-term effects of a diet loosely restricting carbohydrates on HbA1c levels, BMI and tapering of sulfonylureas in type 2 diabetes: A 2-year follow-up study. Diabetes Res Clin Pract. 2008;79(2):350-356. doi:10.1016/j.diabres.2007.09.009. 238. Jenkins D., Newton C, Leeds A, Cummings J. EFFECT OF PECTIN, GUAR GUM, AND WHEAT FIBRE ON SERUM-CHOLESTEROL. Lancet. 1975;305(7916):1116-1117. doi:10.1016/S0140-6736(75)92503-9. 239. Jenkins DA, Leeds A, Wolever TS, et al. UNABSORBABLE CARBOHYDRATES AND DIABETES: DECREASED POST-PRANDIAL HYPERGLYCÆMIA. Lancet. 1976;308(7978):172-174. doi:10.1016/S0140-6736(76)92346-1. 240. JENKINS DJA, LEEDS AR, GASSULL MA, COCHET B, ALBERTI KGMM. Decrease in Postprandial Insulin and Glucose Concentrations by Guar and Pectin. Ann Intern Med. 1977;86(1):20. doi:10.7326/0003-4819-86-1-20. 241. Jenkins DJ, Wolever TM, Nineham R, et al. Guar crispbread in the diabetic diet. Br Med J. 1978;2(6154):1744-1746. http://www.ncbi.nlm.nih.gov/pubmed/737474. Accessed July

199 18, 2017. 242. Jenkins DJ, Leeds AR, Slavin B, Mann J, Jepson EM. Dietary fiber and blood lipids: reduction of serum cholesterol in type II hyperlipidemia by guar gum. Am J Clin Nutr. 1979;32(1):16-18. http://www.ncbi.nlm.nih.gov/pubmed/216259. Accessed July 18, 2017. 243. Franz MJ, Powers MA, Leontos C, et al. The Evidence for Medical Nutrition Therapy for Type 1 and Type 2 Diabetes in Adults. J Am Diet Assoc. 2010;110(12):1852-1889. doi:10.1016/j.jada.2010.09.014. 244. Anderson JW. Dietary fibre, complex carbohydrate and coronary artery disease. Can J Cardiol. 1995;11 Suppl G:55G-62G. http://www.ncbi.nlm.nih.gov/pubmed/7585294. Accessed July 20, 2017. 245. Anderson JW, Hanna TJ. Impact of nondigestible carbohydrates on serum lipoproteins and risk for cardiovascular disease. J Nutr. 1999;129(7 Suppl):1457S-66S. http://www.ncbi.nlm.nih.gov/pubmed/10395621. Accessed July 20, 2017. 246. Anderson JW, Allgood LD, Turner J, Oeltgen PR, Daggy BP. Effects of psyllium on glucose and serum lipid responses in men with type 2 diabetes and hypercholesterolemia. Am J Clin Nutr. 1999;70(4):466-473. http://www.ncbi.nlm.nih.gov/pubmed/10500014. Accessed July 20, 2017. 247. Anderson JW. Dietary fiber prevents carbohydrate-induced hypertriglyceridemia. Curr Atheroscler Rep. 2000;2(6):536-541. doi:10.1007/s11883-000-0055-7. 248. Anderson JW, Allgood LD, Lawrence A, et al. Cholesterol-lowering effects of psyllium intake adjunctive to diet therapy in men and women with hypercholesterolemia: meta- analysis of 8 controlled trials. Am J Clin Nutr. 2000;71(2):472-479. http://www.ncbi.nlm.nih.gov/pubmed/10648260. Accessed July 20, 2017. 249. Brown L, Rosner B, Willett WW, Sacks FM. Cholesterol-lowering effects of dietary fiber: a meta-analysis. Am J Clin Nutr. 1999;69(1):30-42. http://www.ncbi.nlm.nih.gov/pubmed/9925120. Accessed July 20, 2017. 250. Chandalia M, Garg A, Lutjohann D, von Bergmann K, Grundy SM, Brinkley LJ. Beneficial Effects of High Dietary Fiber Intake in Patients with Type 2 Diabetes Mellitus. N Engl J Med. 2000;342(19):1392-1398. doi:10.1056/NEJM200005113421903. 251. Anderson JW, Randles KM, Kendall CWC, Jenkins DJA. Carbohydrate and fiber recommendations for individuals with diabetes: a quantitative assessment and meta-

200 analysis of the evidence. J Am Coll Nutr. 2004;23(1):5-17. http://www.ncbi.nlm.nih.gov/pubmed/14963049. Accessed May 13, 2017. 252. Wing RR, Lang W, Wadden TA, et al. Benefits of Modest Weight Loss in Improving Cardiovascular Risk Factors in Overweight and Obese Individuals With Type 2 Diabetes. Diabetes Care. 2011;34(7). http://care.diabetesjournals.org/content/34/7/1481.short. Accessed July 21, 2017. 253. Jenkins DJA, Kendall CWC, Augustin LSA, et al. Effect of wheat bran on glycemic control and risk factors for cardiovascular disease in type 2 diabetes. Diabetes Care. 2002;25(9):1522-1528. http://www.ncbi.nlm.nih.gov/pubmed/12196421. Accessed July 20, 2017. 254. Jenkins DJA, Kendall CWC, McKeown-Eyssen G, et al. Effect of a Low–Glycemic Index or a High–Cereal Fiber Diet on Type 2 Diabetes. JAMA. 2008;300(23):2742. doi:10.1001/jama.2008.808. 255. Evert AB, Boucher JL, Cypress M, et al. Nutrition Therapy Recommendations for the Management of Adults With Diabetes. Diabetes Care. 2013;36(11). http://care.diabetesjournals.org/content/36/11/3821. Accessed July 21, 2017. 256. Jenkins DJ, Wolever TM, Taylor RH, et al. Glycemic index of foods: a physiological basis for carbohydrate exchange. Am J Clin Nutr. 1981;34(3):362-366. http://www.ncbi.nlm.nih.gov/pubmed/6259925. Accessed May 12, 2017. 257. Jenkins DJA, Wolever TMS, Jenkins AL, et al. The Glycaemic Index of Foods Tested in Diabetic Patients: A New Basis for Carbohydrate Exchange Favouring the Use of Legumes. Diabetologia. 1983;24:257-264. https://link.springer.com/content/pdf/10.1007%2FBF00282710.pdf. Accessed July 18, 2017. 258. Jenkins DJ, Wolever TM, Buckley G, et al. Low-glycemic-index starchy foods in the diabetic diet. Am J Clin Nutr. 1988;48(2):248-254. http://www.ncbi.nlm.nih.gov/pubmed/3407604. Accessed July 18, 2017. 259. Brouns F, Bjorck I, Frayn KN, et al. Glycaemic index methodology. Nutr Res Rev. 2005;18(1):145. doi:10.1079/NRR2005100. 260. Augustin LSA, Kendall CWC, Jenkins DJA, et al. Glycemic index, glycemic load and glycemic response: An International Scientific Consensus Summit from the International

201 Carbohydrate Quality Consortium (ICQC). Nutr Metab Cardiovasc Dis. 2015;25(9):795- 815. doi:10.1016/j.numecd.2015.05.005. 261. Wolever TM, Jenkins DJ, Vuksan V, et al. Beneficial effect of a low glycaemic index diet in type 2 diabetes. Diabet Med. 1992;9(5):451-458. http://www.ncbi.nlm.nih.gov/pubmed/1611833. Accessed July 18, 2017. 262. Jenkins DJA, Srichaikul K, Kendall CWC, et al. The relation of low glycaemic index fruit consumption to glycaemic control and risk factors for coronary heart disease in type 2 diabetes. Diabetologia. 2011;54(2):271-279. doi:10.1007/s00125-010-1927-1. 263. Jenkins DJA, Kendall CWC, Augustin LSA, et al. Effect of Legumes as Part of a Low Glycemic Index Diet on Glycemic Control and Cardiovascular Risk Factors in Type 2 Diabetes Mellitus. Arch Intern Med. 2012;172(21):1653. doi:10.1001/2013.jamainternmed.70. 264. Willett WC, Sacks F, Trichopoulou A, et al. Mediterranean diet pyramid: a cultural model for healthy eating. Am J Clin Nutr. 1995;61(6 Suppl):1402S-1406S. http://www.ncbi.nlm.nih.gov/pubmed/7754995. Accessed May 9, 2017. 265. Esposito K, Maiorino MI, Ceriello A, Giugliano D. Prevention and control of type 2 diabetes by Mediterranean diet: A systematic review. Diabetes Res Clin Pract. 2010;89(2):97-102. doi:10.1016/j.diabres.2010.04.019. 266. Kastorini C-M, Milionis HJ, Esposito K, Giugliano D, Goudevenos JA, Panagiotakos DB. The Effect of Mediterranean Diet on Metabolic Syndrome and its Components: A Meta- Analysis of 50 Studies and 534,906 Individuals. J Am Coll Cardiol. 2011;57(11):1299- 1313. doi:10.1016/j.jacc.2010.09.073. 267. de Lorgeril M, Salen P, Martin J-L, et al. Effect of a mediterranean type of diet on the rate of cardiovascular complications in patients with coronary artery disease insights into the cardioprotective effect of certain nutriments. J Am Coll Cardiol. 1996;28(5):1103-1108. doi:10.1016/S0735-1097(96)00280-X. 268. Rimm EB, Ascherio A, Giovannucci E, Spiegelman D, Stampfer MJ, Willett WC. Vegetable, fruit, and cereal fiber intake and risk of coronary heart disease among men. JAMA. 1996;275(6):447-451. http://www.ncbi.nlm.nih.gov/pubmed/8627965. Accessed May 12, 2017. 269. Gillman MW, Cupples LA, Gagnon D, et al. Protective Effect of Fruits and Vegetables on

202 Development of Stroke in Men. JAMA J Am Med Assoc. 1995;273(14):1113. doi:10.1001/jama.1995.03520380049034. 270. Kris-Etherton PM, Harris WS, Appel LJ. Fish Consumption, Fish Oil, Omega-3 Fatty Acids, and Cardiovascular Disease. Circulation. 2002;106(21). http://circ.ahajournals.org/content/106/21/2747.full. Accessed May 12, 2017. 271. Esposito K, Marfella R, Ciotola M, et al. Effect of a Mediterranean-Style Diet on Endothelial Dysfunction and Markers of Vascular Inflammation in the Metabolic Syndrome. JAMA. 2004;292(12):1440. doi:10.1001/jama.292.12.1440. 272. Rodriguez-Villar C, Perez-Heras A, Mercade I, Casals E, Ros E. Comparison of a high- carbohydrate and a high-monounsaturated fat, olive oil-rich diet on the susceptibility of LDL to oxidative modification in subjects with Type 2 diabetes mellitus. Diabet Med. 2004;21(2):142-149. doi:10.1046/j.1464-5491.2003.01086.x. 273. Garg A, Bantle JP, Henry RR, et al. Effects of varying carbohydrate content of diet in patients with non-insulin-dependent diabetes mellitus. JAMA. 1994;271(18):1421-1428. http://www.ncbi.nlm.nih.gov/pubmed/7848401. Accessed May 9, 2017. 274. O’Keefe JH, Gheewala NM, O’Keefe JO. Dietary Strategies for Improving Post-Prandial Glucose, Lipids, Inflammation, and Cardiovascular Health. J Am Coll Cardiol. 2008;51(3):249-255. doi:10.1016/j.jacc.2007.10.016. 275. de Lorgeril M, Salen P, Martin JL, Monjaud I, Delaye J, Mamelle N. Mediterranean diet, traditional risk factors, and the rate of cardiovascular complications after myocardial infarction: final report of the Lyon Diet Heart Study. Circulation. 1999;99(6):779-785. http://www.ncbi.nlm.nih.gov/pubmed/9989963. 276. Estruch R, Martínez-González MÁ, Corella D, et al. Effects of a Mediterranean-Style Diet on Cardiovascular Risk Factors: a Randomized Trial. Ann Intern Med. 2006;145(1):1. doi:10.7326/0003-4819-145-1-200607040-00004. 277. Estruch R, Ros E, Salas-Salvadó J, et al. Primary Prevention of Cardiovascular Disease with a Mediterranean Diet. N Engl J Med. 2013;368(14):1279-1290. doi:10.1056/NEJMoa1200303. 278. Ros E, Martinez-Gonzalez MA, Estruch R, et al. Mediterranean Diet and Cardiovascular Health: Teachings of the PREDIMED Study. Adv Nutr An Int Rev J. 2014;5(3):330S- 336S. doi:10.3945/an.113.005389.

203 279. Esposito K, Maiorino MI, Ciotola M, et al. Effects of a Mediterranean-style diet on the need for antihyperglycemic drug therapy in patients with newly diagnosed type 2 diabetes: a randomized trial. Ann Intern Med. 2009;151(5):306-314. http://www.ncbi.nlm.nih.gov/pubmed/19721018. Accessed May 9, 2017. 280. Toobert DJ, Glasgow RE, Strycker LA, et al. Biologic and Quality-of-Life Outcomes From the Mediterranean Lifestyle Program. Diabetes Care. 2003;26(8). http://care.diabetesjournals.org/content/26/8/2288.long. Accessed May 9, 2017. 281. Barnard ND, Cohen J, Jenkins DJA, et al. A Low-Fat Vegan Diet Improves Glycemic Control and Cardiovascular Risk Factors in a Randomized Clinical Trial in Individuals With Type 2 Diabetes. Diabetes Care. 2006;29(8). http://care.diabetesjournals.org/content/29/8/1777. Accessed May 12, 2017. 282. Kahleova H, Matoulek M, Malinska H, et al. Vegetarian diet improves insulin resistance and oxidative stress markers more than conventional diet in subjects with Type 2 diabetes. Diabet Med. 2011;28(5):549-559. doi:10.1111/j.1464-5491.2010.03209.x. 283. Hu FB, Rimm EB, Stampfer MJ, Ascherio A, Spiegelman D, Willett WC. Prospective study of major dietary patterns and risk of coronary heart disease in men. Am J Clin Nutr. 2000;72(4):912-921. http://www.ncbi.nlm.nih.gov/pubmed/11010931. 284. Sofi F, Cesari F, Abbate R, Gensini GF, Casini A. Adherence to Mediterranean diet and health status: meta-analysis. BMJ. 2008;337. http://www.bmj.com/content/337/bmj.a1344.short. Accessed May 18, 2017. 285. Lim JH, Lee Y-S, Chang HC, Moon MK, Song Y. Association between dietary patterns and blood lipid profiles in Korean adults with type 2 diabetes. J Korean Med Sci. 2011;26(9):1201-1208. doi:10.3346/jkms.2011.26.9.1201. 286. Osonoi Y, Mita T, Osonoi T, et al. Relationship between dietary patterns and risk factors for cardiovascular disease in patients with type 2 diabetes mellitus: a cross-sectional study. Nutr J. 2016;15(1):1-11. doi:10.1186/s12937-016-0132-6. 287. Shadman Z, Akhoundan M, Poorsoltan N, et al. Association of Major Dietary Patterns with Cardio-metabolic Risk Factors in Type 2 Diabetic Patients. Iran J Public Health. 2016;45(11):1491-1501. http://www.ncbi.nlm.nih.gov/pubmed/28032067. Accessed May 15, 2017. 288. Ciccarone E, Di Castelnuovo A, Salcuni M, et al. A high-score Mediterranean dietary

204 pattern is associated with a reduced risk of peripheral arterial disease in Italian patients with Type 2 diabetes. J Thromb Haemost. 2003;1(8):1744-1752. doi:10.1046/j.1538- 7836.2003.00323.x. 289. Friedman GD, Cutter GR, Donahue RP, et al. CARDIA: study design, recruitment, and some characteristics of the examined subjects. J Clin Epidemiol. 1988;41(11):1105-1116. http://www.ncbi.nlm.nih.gov/pubmed/3204420. 290. Cutter GR, Burke GL, Dyer AR, et al. Cardiovascular Risk Factors in Young Adults The CARDIA Baseline Monograph. 291. McDonald A, Van Horn L, Slattery M, et al. The CARDIA dietary history: development, implementation, and evaluation. J Am Diet Assoc. 1991;91(9):1104-1112. http://www.ncbi.nlm.nih.gov/pubmed/1918764. 292. Meyer KA, Sijtsma FP, Nettleton JA, et al. Dietary patterns are associated with plasma F(2)-isoprostanes in an observational cohort study of adults. Free Radic Biol Med. 2013;57:201-209. doi:10.1016/j.freeradbiomed.2012.08.574. 293. Sijtsma FP, Meyer KA, Steffen LM, et al. Longitudinal trends in diet and effects of sex, race, and education on dietary quality score change: the Coronary Artery Risk Development in Young Adults study. Am J Clin Nutr. 2012;95(3):580-586. doi:10.3945/ajcn.111.020719. 294. Jacobs Jr. DR, Sluik D, Rokling-Andersen MH, Anderssen SA, Drevon CA. Association of 1-y changes in diet pattern with cardiovascular disease risk factors and adipokines: results from the 1-y randomized Oslo Diet and Exercise Study. Am J Clin Nutr. 2009;89(2):509-517. doi:10.3945/ajcn.2008.26371. 295. Lockheart MS, Steffen LM, Rebnord HM, et al. Dietary patterns, food groups and myocardial infarction: a case-control study. Br J Nutr. 2007;98(2):380-387. doi:10.1017/S0007114507701654. 296. Nettleton JA, Schulze MB, Jiang R, Jenny NS, Burke GL, Jacobs Jr. DR. A priori-defined dietary patterns and markers of cardiovascular disease risk in the Multi-Ethnic Study of Atherosclerosis (MESA). Am J Clin Nutr. 2008;88(1):185-194. http://www.ncbi.nlm.nih.gov/pubmed/18614740. 297. Liu K, Slattery M, Jacobs D, et al. A study of the reliability and comparative validity of the cardia dietary history. Ethn Dis. 1994;4(1):15-27.

205 http://www.ncbi.nlm.nih.gov/pubmed/7742729. Accessed October 4, 2017. 298. The Women’s Health Initiative Study Group. Design of the Women’s Health Initiative clinical trial and observational study. The Women’s Health Initiative Study Group. Control Clin Trials. 1998;19(1):61-109. http://www.ncbi.nlm.nih.gov/pubmed/9492970. Accessed July 26, 2017. 299. Hays J, Hunt JR, Hubbell FA, et al. The Women’s Health Initiative Recruitment Methods and Results. Ann Epidemiol. 2003;13:18-77. doi:10.1016/S1047-2797(03)00042-5. 300. Stefanick ML, Cochrane BB, Hsia J, Barad DH, Liu JH, Johnson SR. The Women’s Health Initiative Postmenopausal Hormone Trials: Overview and Baseline Characteristics of Participants. Ann Epidemiol. 2003;13:78-86. doi:10.1016/S1047-2797(03)00045-0. 301. Ritenbaugh C, Patterson RE, Chlebowski RT, et al. The Women’s Health Initiative Dietary Modification trial: overview and baseline characteristics of participants. Ann Epidemiol. 2003;13(9 Suppl):S87-97. http://www.ncbi.nlm.nih.gov/pubmed/14575941. 302. Jackson RD, Lacroix AZ, Cauley JA, Mcgowan J. The Women’s Health Initiative Calcium–Vitamin D Trial: Overview and Baseline Characteristics of Participants. Ann Epidemiol. 2003;13:98-106. doi:10.1016/S1047-2797(03)00046-2. 303. Langer RD, White E, Lewis CE, Kotchen JM, Hendrix SL, Trevisan M. The Women’s Health Initiative Observational Study: Baseline Characteristics of Participants and Reliability of Baseline Measures. Ann Epidemiol. 2003;13:107-121. doi:10.1016/S1047- 2797(03)00047-4. 304. Curb JD, McTiernan A, Heckbert SR, et al. Outcomes ascertainment and adjudication methods in the Women’s Health Initiative. Ann Epidemiol. 2003;13(9 Suppl):S122-8. http://www.ncbi.nlm.nih.gov/pubmed/14575944. 305. Margolis KL, Lihong Qi L, Brzyski R, et al. Validity of diabetes self-reports in the Women’s Health Initiative: comparison with medication inventories and fasting glucose measurements. Clin Trials. 2008;5(3):240-247. doi:10.1177/1740774508091749. 306. Bowman SA, Friday JE, Moshfegh AJ. MyPyramid Equivalents Database, 2.0 for USDA Survey Foods, 2003-2004: Documentation and User Guide. 2008. http://www.ars.usda.gov/ba/bhnrc/fsrg. Accessed December 8, 2016. 307. Patterson RE, Kristal AR, Tinker LF, Carter RA, Bolton MP, Agurs-Collins T. Measurement characteristics of the Women’s Health Initiative food frequency

206 questionnaire. Ann Epidemiol. 1999;9(3):178-187. http://www.ncbi.nlm.nih.gov/pubmed/10192650. Accessed December 5, 2016. 308. WHI Manuals: Volume 8. Section 5 - Outcome Classifications: Cardiovascular Disease. 2002:5-1-5-15. https://biolincc.nhlbi.nih.gov/static/studies/whict/hidden/doc/whi/outcome/5.pdf. 309. Alhazmi A, Stojanovski E, McEvoy M, Garg ML. The association between dietary patterns and type 2 diabetes: a systematic review and meta-analysis of cohort studies. J Hum Nutr Diet. 2014;27(3):251-260. doi:10.1111/jhn.12139. 310. Shikany JM, Jacobs DR, Lewis CE, et al. Associations between food groups, dietary patterns, and cardiorespiratory fitness in the Coronary Artery Risk Development in Young Adults study. Am J Clin Nutr. 2013;98(6):1402-1409. doi:10.3945/ajcn.113.058826. 311. Sijtsma FPC, Meyer KA, Steffen LM, et al. Diet quality and markers of endothelial function: the CARDIA study. Nutr Metab Cardiovasc Dis. 2014;24(6):632-638. doi:10.1016/j.numecd.2013.12.010. 312. Buijsse B, Jacobs DR, Steffen LM, Kromhout D, Gross MD. Plasma Ascorbic Acid, A Priori Diet Quality Score, and Incident Hypertension: A Prospective Cohort Study. Singh PK, ed. PLoS One. 2015;10(12):e0144920. doi:10.1371/journal.pone.0144920. 313. Dietary Guidelines Advisory Committee. Scientific Report of the 2015 Dietary Guidelines Advisory Committee. Washingt USDA US Dep Heal Hum Serv. 2015;53(9):1689-1699. doi:10.1017/CBO9781107415324.004. 314. Eaton SB, Konner M. Paleolithic Nutrition - A Consideration of Its Nature and Current Implications. N Engl J Med. 1985;312(5):283-289. doi:10.1056/NEJM198501313120505. 315. Jönsson T, Granfeldt Y, Ahrén B, et al. Beneficial effects of a Paleolithic diet on cardiovascular risk factors in type 2 diabetes: a randomized cross-over pilot study. Cardiovasc Diabetol. 2009;8(1):35. doi:10.1186/1475-2840-8-35. 316. Whalen KA, McCullough M, Flanders WD, Hartman TJ, Judd S, Bostick RM. Paleolithic and Mediterranean diet pattern scores and risk of incident, sporadic colorectal adenomas. Am J Epidemiol. 2014;180(11):1088-1097. doi:10.1093/aje/kwu235. 317. Samaha FF, Iqbal N, Seshadri P, et al. A low-carbohydrate as compared with a low-fat diet in severe obesity. N Engl J Med. 2003;348(21):2074-2081. doi:10.1056/NEJMoa022637.

207 318. McAuley KA, Hopkins CM, Smith KJ, et al. Comparison of high-fat and high-protein diets with a high-carbohydrate diet in insulin-resistant obese women. Diabetologia. 2005;48(1):8-16. doi:10.1007/s00125-004-1603-4. 319. Atkins RC. Dr. Atkins’ New Diet Revolution. New York: Avon Books; 1992. 320. Jacobs DR Haskell WL, Pirie P, Sidney S HLP. Validity and reliability of a short physical activity history: CARDIA and the Minnesota Heart Health Program. J Cardiopulm Rehabil. 1989;9:12. 321. Parker ED, Schmitz KH, Jacobs DR, Dengel DR, Schreiner PJ, Schreiner PJ. Physical activity in young adults and incident hypertension over 15 years of follow-up: the CARDIA study. Am J Public Health. 2007;97(4):703-709. doi:10.2105/AJPH.2004.055889. 322. Willi C, Bodenmann P, Ghali WA, Faris PD, Cornuz J. Active Smoking and the Risk of Type 2 Diabetes. JAMA. 2007;298(22):2654. doi:10.1001/jama.298.22.2654. 323. Sargeant LA, Khaw K-T, Bingham S, et al. Cigarette smoking and glycaemia: the EPIC- Norfolk Study. Int J Epidemiol. 2001;30(3):547-554. doi:10.1093/ije/30.3.547. 324. Odegaard AO, Koh WP, Butler LM, et al. Dietary patterns and incident type 2 diabetes in chinese men and women: the singapore chinese health study. Diabetes Care. 2011;34(4):880-885. doi:10.2337/dc10-2350. 325. Hozawa A, Jacobs Jr. DR, Steffes MW, Gross MD, Steffen LM, Lee DH. Associations of serum carotenoid concentrations with the development of diabetes and with insulin concentration: interaction with smoking: the Coronary Artery Risk Development in Young Adults (CARDIA) Study. Am J Epidemiol. 2006;163(10):929-937. doi:10.1093/aje/kwj136. 326. Frassetto LA, Schloetter M, Mietus-Synder M, Morris RC, Sebastian A. Metabolic and physiologic improvements from consuming a paleolithic, hunter-gatherer type diet. Eur J Clin Nutr. 2009;63(8):947-955. doi:10.1038/ejcn.2009.4. 327. Österdahl M, Kocturk T, Koochek A, Wändell PE. Effects of a short-term intervention with a paleolithic diet in healthy volunteers. Eur J Clin Nutr. 2008;62(5):682-685. doi:10.1038/sj.ejcn.1602790. 328. Boers I, Muskiet FA, Berkelaar E, et al. Favourable effects of consuming a Palaeolithic- type diet on characteristics of the metabolic syndrome: a randomized controlled pilot-

208 study. Lipids Health Dis. 2014;13(1):160. doi:10.1186/1476-511X-13-160. 329. Eaton SB, Cordain L, Lindeberg S. Evolutionary Health Promotion: A Consideration of Common Counterarguments. Prev Med (Baltim). 2002;34(2):119-123. doi:10.1006/pmed.2001.0966. 330. Hallfrisch J, Facn, Behall KM. Mechanisms of the effects of grains on insulin and glucose responses. J Am Coll Nutr. 2000;19(3 Suppl):320S-325S. http://www.ncbi.nlm.nih.gov/pubmed/10875604. Accessed January 3, 2018. 331. Higgins JA. Whole grains, legumes, and the subsequent meal effect: implications for blood glucose control and the role of fermentation. J Nutr Metab. 2012;2012:829238. doi:10.1155/2012/829238. 332. Bjørnshave A, Hermansen K. Effects of dairy protein and fat on the metabolic syndrome and type 2 diabetes. Rev Diabet Stud. 2014;11(2):153-166. doi:10.1900/RDS.2014.11.153. 333. Halton TL, Liu S, Manson JE, Hu FB. Low-carbohydrate-diet score and risk of type 2 diabetes in women. Am J Clin Nutr. 2008;87(2):339-346. http://www.ncbi.nlm.nih.gov/pubmed/18258623. Accessed December 1, 2017. 334. de Koning L, Fung TT, Liao X, et al. Low-carbohydrate diet scores and risk of type 2 diabetes in men. Am J Clin Nutr. 2011;93(4):844-850. doi:10.3945/ajcn.110.004333. 335. Nanri A, Mizoue T, Kurotani K, et al. Low-Carbohydrate Diet and Type 2 Diabetes Risk in Japanese Men and Women: The Japan Public Health Center-Based Prospective Study. Song Y, ed. PLoS One. 2015;10(2):e0118377. doi:10.1371/journal.pone.0118377. 336. Tobias DK, Chen M, Manson JE, Ludwig DS, Willett W, Hu FB. Effect of low-fat diet interventions versus other diet interventions on long-term weight change in adults: a systematic review and meta-analysis. Lancet Diabetes Endocrinol. 2015;3(12):968-979. doi:10.1016/S2213-8587(15)00367-8. 337. Freedhoff Y, Hall KD. Weight loss diet studies: we need help not hype. Lancet. 2016;388(10047):849-851. doi:10.1016/S0140-6736(16)31338-1. 338. Zamora D, Gordon-Larsen P, He K, Jacobs Jr. DR, Shikany JM, Popkin BM. Are the 2005 Dietary Guidelines for Americans Associated With reduced risk of type 2 diabetes and cardiometabolic risk factors? Twenty-year findings from the CARDIA study. Diabetes Care. 2011;34(5):1183-1185. doi:10.2337/dc10-2041. 339. InterAct Consortium. Adherence to predefined dietary patterns and incident type 2

209 diabetes in European populations: EPIC-InterAct Study. Diabetologia. 2014;57(2):321- 333. doi:10.1007/s00125-013-3092-9. 340. McDonald A, Van Horn L, Slattery M, et al. The CARDIA dietary history: development, implementation, and evaluation. J Am Diet Assoc. 1991;91(9):1104-1112. http://www.ncbi.nlm.nih.gov/pubmed/1918764. Accessed December 5, 2016. 341. Swithers SE. Artificial sweeteners produce the counterintuitive effect of inducing metabolic derangements. Trends Endocrinol Metab. 2013;24(9):431-441. doi:10.1016/j.tem.2013.05.005. 342. Bleich SN, Vercammen KA, Koma JW, Li Z. Trends in Beverage Consumption Among Children and Adults, 2003-2014. Obesity. November 2017. doi:10.1002/oby.22056. 343. Sylvetsky AC, Rother KI. Trends in the consumption of low-calorie sweeteners. Physiol Behav. 2016;164(Pt B):446-450. doi:10.1016/j.physbeh.2016.03.030. 344. National Center for Health Statistics. Health, United States, 2016: With Chartbook on Long-Term Trends in Health. Hyattsville, MD; 2017. 345. Malik VS, Popkin BM, Bray GA, Després J-P, Willett WC, Hu FB. Sugar-Sweetened Beverages and Risk of Metabolic Syndrome and Type 2 Diabetes. Diabetes Care. 2010;33(11). 346. Te Morenga L, Mallard S, Mann J. Dietary sugars and body weight: systematic review and meta-analyses of randomised controlled trials and cohort studies. BMJ. 2012;346:e7492. doi:10.1136/BMJ.E7492. 347. Gardner C, Wylie-Rosett J, Gidding SS, et al. Nonnutritive sweeteners: current use and health perspectives: a scientific statement from the American Heart Association and the American Diabetes Association. Diabetes Care. 2012;35(8):1798-1808. doi:10.2337/dc12-9002. 348. Fitch C, Keim K. Position of the academy of nutrition and dietetics: use of nutritive and nonnutritive sweeteners. J Acad Nutr Diet. 2012;112:739-758. 349. Borges MC, Louzada ML, de Sá TH, et al. Artificially Sweetened Beverages and the Response to the Global Obesity Crisis. PLoS Med. 2017;14(1):e1002195. doi:10.1371/journal.pmed.1002195. 350. Azad MB, Abou-Setta AM, Chauhan BF, et al. Nonnutritive sweeteners and cardiometabolic health: a systematic review and meta-analysis of randomized controlled

210 trials and prospective cohort studies. CMAJ. 2017;189(28):E929-E939. doi:10.1503/cmaj.161390. 351. Duffey KJ, Steffen LM, Van Horn L, Jacobs Jr. DR, Popkin BM. Dietary patterns matter: diet beverages and cardiometabolic risks in the longitudinal Coronary Artery Risk Development in Young Adults (CARDIA) Study. Am J Clin Nutr. 2012;95(4):909-915. doi:10.3945/ajcn.111.026682. 352. Bleich SN, Wolfson JA, Vine S, Wang YC. Diet-Beverage Consumption and Caloric Intake Among US Adults, Overall and by Body Weight. Am J Public Health. 2014;104(3):e72-e78. doi:10.2105/AJPH.2013.301556. 353. Elfhag K, Tynelius P, Rasmussen F. Sugar-sweetened and artificially sweetened soft drinks in association to restrained, external and emotional eating. Physiol Behav. 2007;91(2-3):191-195. doi:10.1016/j.physbeh.2007.02.005. 354. Lutsey PL, Steffen LM, Stevens J. Dietary Intake and the Development of the Metabolic Syndrome: The Atherosclerosis Risk in Communities Study. Circulation. 2008;117(6):754-761. doi:10.1161/CIRCULATIONAHA.107.716159. 355. InterAct Consortium, Romaguera D, Norat T, et al. Consumption of sweet beverages and type 2 diabetes incidence in European adults: results from EPIC-InterAct. Diabetologia. 2013;56(7):1520-1530. doi:10.1007/s00125-013-2899-8. 356. Pepino MY. Metabolic effects of non-nutritive sweeteners. Physiol Behav. 2015;152(Pt B):450-455. doi:10.1016/j.physbeh.2015.06.024. 357. Suez J, Korem T, Zilberman-Schapira G, Segal E, Elinav E. Non-caloric artificial sweeteners and the microbiome: findings and challenges. Gut Microbes. 2015;6(2):149- 155. doi:10.1080/19490976.2015.1017700. 358. Suez J, Korem T, Zeevi D, et al. Artificial sweeteners induce glucose intolerance by altering the gut microbiota. Nature. 2014;514(7521):181-186. doi:10.1038/nature13793. 359. Pepino MY, Tiemann CD, Patterson BW, Wice BM, Klein S. Sucralose affects glycemic and hormonal responses to an oral glucose load. Diabetes Care. 2013;36(9):2530-2535. doi:10.2337/dc12-2221. 360. Miller PE, Perez V. Low-calorie sweeteners and body weight and composition: a meta- analysis of randomized controlled trials and prospective cohort studies. Am J Clin Nutr. 2014;100(3):765-777. doi:10.3945/ajcn.113.082826.

211 361. Wiebe N, Padwal R, Field C, Marks S, Jacobs R, Tonelli M. A systematic review on the effect of sweeteners on glycemic response and clinically relevant outcomes. BMC Med. 2011;9(1):123. doi:10.1186/1741-7015-9-123. 362. Mattes RD, Popkin BM. Nonnutritive sweetener consumption in humans: effects on appetite and food intake and their putative mechanisms. Am J Clin Nutr. 2009;89(1):1-14. doi:10.3945/ajcn.2008.26792. 363. Fowler SPG. Low-calorie sweetener use and energy balance: Results from experimental studies in animals, and large-scale prospective studies in humans. Physiol Behav. 2016;164(Pt B):517-523. doi:10.1016/j.physbeh.2016.04.047. 364. Rogers PJ, Hogenkamp PS, de Graaf C, et al. Does low-energy sweetener consumption affect energy intake and body weight? A systematic review, including meta-analyses, of the evidence from human and animal studies. Int J Obes (Lond). 2016;40(3):381-394. doi:10.1038/ijo.2015.177. 365. Bellisle F, Drewnowski A. Intense sweeteners, energy intake and the control of body weight. Eur J Clin Nutr. 2007;61(6):691-700. doi:10.1038/sj.ejcn.1602649. 366. Mattes RD, Popkin BM. Nonnutritive sweetener consumption in humans: effects on appetite and food intake and their putative mechanisms. Am J Clin Nutr. 2009;89(1):1-14. doi:10.3945/ajcn.2008.26792. 367. Imamura F, O’Connor L, Ye Z, et al. Consumption of sugar sweetened beverages, artificially sweetened beverages, and fruit juice and incidence of type 2 diabetes: systematic review, meta-analysis, and estimation of population attributable fraction. BMJ. 2015;351:h3576. http://www.ncbi.nlm.nih.gov/pubmed/26199070. Accessed July 10, 2017. 368. DiMeglio DP, Mattes RD. Liquid versus solid carbohydrate: effects on food intake and body weight. Int J Obes Relat Metab Disord. 2000;24(6):794-800. http://www.ncbi.nlm.nih.gov/pubmed/10878689. 369. Raben A, Vasilaras TH, Møller AC, Astrup A. Sucrose compared with artificial sweeteners: different effects on ad libitum food intake and body weight after 10 wk of supplementation in overweight subjects. Am J Clin Nutr. 2002;76(4):721-729. http://www.ncbi.nlm.nih.gov/pubmed/12324283. Accessed January 10, 2018. 370. Tordoff MG, Alleva AM. Effect of drinking soda sweetened with aspartame or high-

212 fructose corn syrup on food intake and body weight. Am J Clin Nutr. 1990;51(6):963-969. http://www.ncbi.nlm.nih.gov/pubmed/2349932. Accessed January 10, 2018. 371. Van Wymelbeke V, Béridot-Thérond M-E, de La Guéronnière V, Fantino M. Influence of repeated consumption of beverages containing sucrose or intense sweeteners on food intake. Eur J Clin Nutr. 2004;58(1):154-161. doi:10.1038/sj.ejcn.1601762. 372. Vartanian LR, Schwartz MB, Brownell KD. Effects of soft drink consumption on nutrition and health: a systematic review and meta-analysis. Am J Public Health. 2007;97(4):667- 675. doi:10.2105/AJPH.2005.083782. 373. Mattes RD, Shikany JM, Kaiser KA, Allison DB. Nutritively sweetened beverage consumption and body weight: a systematic review and meta-analysis of randomized experiments. Obes Rev. 2011;12(5):346-365. doi:10.1111/j.1467-789X.2010.00755.x. 374. Ebbeling CB, Feldman HA, Osganian SK, Chomitz VR, Ellenbogen SJ, Ludwig DS. Effects of Decreasing Sugar-Sweetened Beverage Consumption on Body Weight in Adolescents: A Randomized, Controlled Pilot Study. Pediatrics. 2006;117(3):673-680. doi:10.1542/peds.2005-0983. 375. Boyle JP, Thompson TJ, Gregg EW, Barker LE, Williamson DF. Projection of the year 2050 burden of diabetes in the US adult population: dynamic modeling of incidence, mortality, and prediabetes prevalence. Popul Health Metr. 2010;8:29. doi:10.1186/1478- 7954-8-29. 376. Lehto S, Ronnemaa T, Haffner SM, Pyorala K, Kallio V, Laakso M. Dyslipidemia and hyperglycemia predict coronary heart disease events in middle-aged patients with NIDDM. Diabetes. 1997;46(8):1354-1359. http://www.ncbi.nlm.nih.gov/pubmed/9231662. 377. Turner RC, Millns H, Neil HA, et al. Risk factors for coronary artery disease in non- insulin dependent diabetes mellitus: United Kingdom Prospective Diabetes Study (UKPDS: 23). BMJ. 1998;316(7134):823-828. http://www.ncbi.nlm.nih.gov/pubmed/9549452. 378. Juutilainen A, Lehto S, Ronnemaa T, Pyorala K, Laakso M. Type 2 diabetes as a “coronary heart disease equivalent”: an 18-year prospective population-based study in Finnish subjects. Diabetes Care. 2005;28(12):2901-2907. http://www.ncbi.nlm.nih.gov/pubmed/16306552.

213 379. Schramm TK, Gislason GH, Kober L, et al. Diabetes patients requiring glucose-lowering therapy and nondiabetics with a prior myocardial infarction carry the same cardiovascular risk: a population study of 3.3 million people. Circulation. 2008;117(15):1945-1954. doi:10.1161/CIRCULATIONAHA.107.720847. 380. Lichtenstein AH, Appel LJ, Brands M, et al. Diet and Lifestyle Recommendations Revision 2006: A Scientific Statement From the American Heart Association Nutrition Committee. Circulation. 2006;114(1):82-96. doi:10.1161/CIRCULATIONAHA.106.176158. 381. Papakonstantinou E, Panagiotakos DB, Pitsavos C, Chrysohoou C, Zampelas A. Food Group Consumption and Glycemic Control in People With and Without Type 2 Diabetes The ATTICA study. Diabetes Care. 2005;28(10):2539-2540. http://care.diabetesjournals.org/content/diacare/28/10/2539.full.pdf. Accessed November 18, 2017. 382. Mozaffarian D, Hao T, Rimm EB, Willett WC, Hu FB. Changes in Diet and Lifestyle and Long-Term Weight Gain in Women and Men. N Engl J Med. 2011;364(25):2392-2404. doi:10.1056/NEJMoa1014296. 383. Margolis KL, Lihong Q, Brzyski R, et al. Validity of diabetes self-reports in the Women’s Health Initiative: comparison with medication inventories and fasting glucose measurements. Clin Trials. 2008;5(3):240-247. doi:10.1177/1740774508091749. 384. Patterson RE, Kristal AR, Tinker LF, Carter RA, Bolton MP, Agurs-Collins T. Measurement characteristics of the Women’s Health Initiative food frequency questionnaire. Ann Epidemiol. 1999;9(3):178-187. http://www.ncbi.nlm.nih.gov/pubmed/10192650. 385. Schakel SF, Sievert YA, Buzzard IM. Sources of data for developing and maintaining a nutrient database. J Am Diet Assoc. 1988;88(10):1268-1271. http://www.ncbi.nlm.nih.gov/pubmed/3171020. Accessed September 5, 2017. 386. Fung TT, Rexrode KM, Mantzoros CS, Manson JE, Willett WC, Hu FB. Mediterranean diet and incidence of and mortality from coronary heart disease and stroke in women. Circulation. 2009;119(8):1093-1100. doi:10.1161/CIRCULATIONAHA.108.816736. 387. Fung TT, Hu FB, McCullough ML, Newby PK, Willett WC, Holmes MD. Diet quality is associated with the risk of estrogen receptor-negative breast cancer in postmenopausal

214 women. J Nutr. 2006;136(2):466-472. http://www.ncbi.nlm.nih.gov/pubmed/16424129. Accessed December 5, 2016. 388. Whalen KA, McCullough M, Flanders WD, Hartman TJ, Judd S, Bostick RM. Paleolithic and Mediterranean diet pattern scores and risk of incident, sporadic colorectal adenomas. Am J Epidemiol. 2014;180(11):1088-1097. doi:10.1093/aje/kwu235. 389. Whalen KA, McCullough ML, Flanders WD, Hartman TJ, Judd S, Bostick RM. Paleolithic and Mediterranean Diet Pattern Scores Are Inversely Associated with Biomarkers of Inflammation and Oxidative Balance in Adults. J Nutr. 2016;146(6):1217- 1226. doi:10.3945/jn.115.224048. 390. Wu PY, Huang CL, Lei WS, Yang SH. Alternative health eating index and the Dietary Guidelines from American Diabetes Association both may reduce the risk of cardiovascular disease in type 2 diabetes patients. J Hum Nutr Diet. 2016;29(3):363-373. doi:10.1111/jhn.12317. 391. Rossouw JE, Anderson GL, Prentice RL, et al. Risks and benefits of estrogen plus progestin in healthy postmenopausal women: principal results From the Women’s Health Initiative randomized controlled trial. JAMA. 2002;288(3):321-333. http://www.ncbi.nlm.nih.gov/pubmed/12117397. Accessed December 9, 2016. 392. Manson JE, Hsia J, Johnson KC, et al. Estrogen plus Progestin and the Risk of Coronary Heart Disease. N Engl J Med. 2003;349(6):523-534. doi:10.1056/NEJMoa030808. 393. Meyer A-M, Evenson KR, Morimoto L, Siscovick D, White E. Test-retest reliability of the Women’s Health Initiative physical activity questionnaire. Med Sci Sports Exerc. 2009;41(3):530-538. doi:10.1249/MSS.0b013e31818ace55. 394. McTiernan A, Kooperberg C, White E, et al. Recreational Physical Activity and the Risk of Breast Cancer in Postmenopausal Women. JAMA. 2003;290(10):1331. doi:10.1001/jama.290.10.1331. 395. Heckbert SR, Kooperberg C, Safford MM, et al. Comparison of Self-Report, Hospital Discharge Codes, and Adjudication of Cardiovascular Events in the Women’s Health Initiative. Am J Epidemiol. 2004;160(12):1152-1158. doi:10.1093/aje/kwh314. 396. Conen D, Ridker PM, Mora S, Buring JE, Glynn RJ. Blood pressure and risk of developing type 2 diabetes mellitus: The Women’s Health Study. Eur Heart J. 2007;28(23):2937-2943. doi:10.1093/eurheartj/ehm400.

215 397. Anderson GL, Manson J, Wallace R, et al. Implementation of the Women’s Health Initiative Study Design. Ann Epidemiol. 2003;13:5-17. doi:10.1016/S1047- 2797(03)00043-7. 398. Harrell F. Regression Modeling Strategies: With Application to Linear Models, Logistic Regression, and Survival Analysis. New York: Springer-Verlag; 2001. 399. Martinez-Gonzalez MA, Bes-Rastrollo M. Dietary patterns, Mediterranean diet, and cardiovascular disease. Curr Opin Lipidol. 2014;25(1):20-26. doi:10.1097/MOL.0000000000000044. 400. Huffman FG, Zarini GG, McNamara E, Nagarajan A. The Healthy Eating Index and the Alternate Healthy Eating Index as predictors of 10-year CHD risk in Cuban Americans with and without type 2 diabetes. Public Heal Nutr. 2011;14(11):2006-2014. doi:10.1017/S1368980011001054. 401. Lindeberg S, Jönsson T, Granfeldt Y, et al. A Palaeolithic diet improves glucose tolerance more than a Mediterranean-like diet in individuals with ischaemic heart disease. Diabetologia. 2007;50(9):1795-1807. doi:10.1007/s00125-007-0716-y. 402. Masharani U, Sherchan P, Schloetter M, et al. Metabolic and physiologic effects from consuming a hunter-gatherer (Paleolithic)-type diet in type 2 diabetes. Eur J Clin Nutr. 2015;69(8):944-948. doi:10.1038/ejcn.2015.39. 403. Liu S, Stampfer MJ, Hu FB, et al. Whole-grain consumption and risk of coronary heart disease: results from the Nurses’ Health Study. Am J Clin Nutr. 1999;70(3):412-419. http://www.ncbi.nlm.nih.gov/pubmed/10479204. Accessed November 28, 2017. 404. Jensen MK, Koh-Banerjee P, Hu FB, et al. Intakes of whole grains, bran, and germ and the risk of coronary heart disease in men. Am J Clin Nutr. 2004;80(6):1492-1499. http://www.ncbi.nlm.nih.gov/pubmed/15585760. Accessed November 28, 2017. 405. Lissner L. Measuring food intake in studies of obesity. Public Health Nutr. 2002;5(6a):889-892. doi:10.1079/PHN2002388.

216