Development of Cardiovascular Disease Risk Score by Adding a Body Composition Assessment

Material Information

Development of Cardiovascular Disease Risk Score by Adding a Body Composition Assessment
Jo, Ara
University of Florida
Publication Date:

Thesis/Dissertation Information

Doctorate ( Ph.D.)
Degree Grantor:
University of Florida
Degree Disciplines:
Health Services Research
Health Services Research, Management, and Policy
Committee Chair:
Committee Co-Chair:
Committee Members:


Subjects / Keywords:


General Note:
Cardiovascular disease has been a top leading cause of mortality in the US. To decrease cardiovascular disease(CVD) prevalence and mortality, the risk score has been developed since late 1990. It is affordable and effective tool to estimate future CVD risk in clinical setting. However despite threaten of obesity, a few risk score has adopted BMI into the model. Also, recent studies found that BMI may cause misclassification of people who fall in normal weight and moderate overweight range. Therefore, the purpose of this study was to develop CVD risk score by adding a body composition assessment for adults aged over 44 and older. This study used the multi ethnic study of atherosclerosis (MESA) dataset which takes advantage of collecting asymptomatic population with diversity of race/ethnicity (n=5,483). It was divided into two datasets of equal size randomly, one used for creating the risk score and another used for validating the model. As a result, the risk model included eight classical CVD risk factors including age, sex, systolic blood pressure, hypertensive medication, diabetes, total cholesterol, high-density lipoprotein cholesterol (HDL-C) and smoking and quartile of waist-to-hip ratio (WHR) as the optimal body composition assessment. By using cox proportional hazards regression with 10 years of follow-up period, non-invasive and affordable risk score was derived and risk scores has own established based on hazard ratios(AUC, 0.763). This risk score was validated (AUC, 0.746). Particularly, the model showed better predictive accuracy in normal weight and overweight population (AUC, 0.771). This risk score may contribute to improving doctor's decision making and providing appropriate treatment particularly for population who may be misclassified as healthy or unhealthy due to BMI.

Record Information

Source Institution:
Rights Management:
All applicable rights reserved by the source institution and holding location.
Embargo Date:


This item has the following downloads:

Full Text




2 2017 Ara Jo


3 To my mom, dad, brother and grand father


4 ACKNOWLEDGMENTS I would like to appreciate all of my committee members for their support and sincere mentorship during my PhD journey. Firstly, I would like to thank Dr. Mainous, my committee chair, for his mentorship, training, and support. I always think that I am lucky to have you as my mentor and I would have never finished my dissertation without your tremendous support and advice. I would also like to recognize Dr. Marlow, who always educated me particularly in methods and skills and provided positive support through out my research and dissertation. I want to thank Dr. Anton who inspired and guided me in multidisciplinary research. I would also like to appreciate Dr. Beau De Rochars, who guided my dissertation and gave me an opportunity to work on a valuable internati onal research project. A special thank you to Rebecca Tanner, my research my research and dissertation. I also thank the wonderful staff within the Department of Health S ervices Research, Management and Policy at the University of Florida who always took care of my complicated situations and dealt with my endless questions. I also would like to express my sincere gratitude to Dr. H arman, Dr. Duncan, Dr. Harle, and Dr. Hal l. When I started in the program, their research and teachings always nurtured my knowledge and encouraged me to pursue this path. Without their lessons, I could not start studying this area successfully. I am appreciative of their endless dedication to me ntorship, always making me feel valued, and inspiring me to become better. I would also like to thank my colleagues in the HSRMP Program. Thank you for your sincere trust in me and support through this academic journey. A special thanks to Shenae K. Samuel s, my dissertation mate. Without her, I would not have finished my


5 dissertation on time. I will never forget our bagel time. In addition, thank you to all of my friends at the University of Florida, for your support and time spent with me. I look forward t o working with all of you in future career. I also would like to extend many thanks to all my friends in South Korea and AIU friends who pushed me to pursu e my stud ies in the U nited S tates I especially want to thank Bonsang Han and Sunyoung Park, my best friends in South Korea, for their encouragement, trust, and support. I could not have completed this long journey without your support and trust. Finally, I want to thank my family and my aunt for their unconditional love, support, and pati ence during this long and trying journey. To Mom, NG Kim and Dad, HC Jo, I truly appreciate your patience, support, and love. I also want to thank my brother WS Jo for supporting and making me smile. A special thanks to my aunt, SG Kim. This dissertation c ould be completed without your tremendous encouragement and trust in me. I really appreciate all of my family for their continued support.


6 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ .. 4 LIST OF TABLES ................................ ................................ ................................ ............ 8 LIST OF FIGURES ................................ ................................ ................................ .......... 9 LIST OF ABBREVIATIONS ................................ ................................ ........................... 10 ABSTRACT ................................ ................................ ................................ ................... 12 CHAPTER 1 INTRODUCTION ................................ ................................ ................................ .... 14 2 LITERATURE REVIEW ................................ ................................ .......................... 16 Cardiovascular Disease (CVD) ................................ ................................ ............... 16 Risk Factors ................................ ................................ ................................ ..... 17 Demographics ................................ ................................ ............................ 17 Cholesterol ................................ ................................ ................................ 19 High b lood pressure. ................................ ................................ .................. 19 Diabetes. ................................ ................................ ................................ .... 19 Smoking ................................ ................................ ................................ ..... 20 Obesity ................................ ................................ ................................ ............. 20 Misclassification of BMI ................................ ................................ ........................... 23 Body Composition Assessments ................................ ................................ ............ 26 Waist Circumference (WC) ................................ ................................ ............... 2 8 Waist to Hip Ratio (WHR) ................................ ................................ ................ 29 Waist to Height Ratio (WtoHR) ................................ ................................ ........ 30 CVD Risk Score ................................ ................................ ................................ ...... 30 Framingham Risk Score (FRS) ................................ ................................ ........ 32 Adding a New Indicator ................................ ................................ .................... 33 Specific Aims ................................ ................................ ................................ .......... 34 Aim 1. Select an O ptimal Body Composition A ssessment ............................... 34 A im 2. Develop the Final Risk Score ................................ ................................ 35 Aim 3. Test a Validation of Developed Model and Improvement of Predictive Accuracy in Norma l Weight and Overweight Populations ............................. 35 3 METHODS ................................ ................................ ................................ .............. 36 Data ................................ ................................ ................................ ........................ 36 Participants ................................ ................................ ................................ ............. 38 Outcomes ................................ ................................ ................................ ............... 38 Independent Variables ................................ ................................ ............................ 39


7 Risk Factors ................................ ................................ ................................ ..... 39 Body Compositions ................................ ................................ .......................... 40 Statistical Analysis ................................ ................................ ................................ .. 45 AIM 1. Select an O ptimal Body Composition A ssessment .............................. 45 AIM 2. Develop the Final Risk Score ................................ ............................... 48 AIM 3. Test a Validation of Developed Model and Improvement of Predictive Accuracy in Normal Weight and Overweight Population .............. 48 4 RESULTS ................................ ................................ ................................ ............... 50 Descriptive Analyses ................................ ................................ ............................... 50 Development of Risk Score ................................ ................................ .................... 52 Aim 1. Select an O ptimal Body Composition A ssessment ............................... 52 Aim 2. Develop the Final Risk Score ................................ ................................ 54 Aim 3. Test a Validation of the Developed Model and Improvement of Predictive Accuracy in Normal Weight and Overweight Populations ............. 54 5 DISCUSSION ................................ ................................ ................................ ......... 66 Discussion ................................ ................................ ................................ .............. 66 Body Composition Assessment ................................ ................................ ........ 66 Risk Factors ................................ ................................ ................................ ..... 70 Risk Score ................................ ................................ ................................ ........ 73 Implication ................................ ................................ ................................ ............... 79 6 CONCLUSION ................................ ................................ ................................ ........ 81 APPENDIX A FLOW CHART OF STUDY POPULATION ................................ ............................. 82 B AREA UNDER THE CURVE OF THE FINAL MODEL ................................ ............ 83 C AREA UNDER THE CURVE OF THE VALIDATION TEST ................................ .... 84 D AREA UNDER THE CURVE IN NORMAL WEIGHT AND OVERWEIGHT POPULATION ................................ ................................ ................................ ......... 85 LIST OF REFERENCES ................................ ................................ ............................... 86 BIOGRAPHICAL SKETCH ................................ ................................ .......................... 101


8 LIST OF TABLES Table page 3 1 Description of Operation for CVD Risk Factors ................................ .................. 42 3 2 Description of Sex Specific Cutoffs of Body Composition Assessments ............ 43 3 3 Description of Classification of Integrated Body Composition Assessments Combining with BMI ................................ ................................ ............................ 44 4 1 Comparison of Characteristics between Group 1 and Group 2 .......................... 56 4 2 Baseline Profile of CVD Risk Factors of Study Population from Group 1 ........... 57 4 3 Baseline Characteristics of Body Composition of Group 1 ................................ 58 4 4 Results of Propo rtionality Assumption Test with Body Composition Assessments and the Time to the First CVD Event in Group 1 .......................... 60 4 5 Results of Comparison of CVD Risk Models Using Cox Proportional Hazard Regressions ................................ ................................ ................................ ........ 61 4 6 Scoring on the Final Risk Model ................................ ................................ ......... 63 4 7 Summary of Body Mass Index Misclassification in US Adults aged over 44 and Older Using National Health and Nutrition Examination Survey (NHA NES), 1999 2006 ................................ ................................ ....................... 64 4 8 Summary of a Validation Test in Group 2 and Subpopulation of Normal weight and Overweight population only in Group 2 ................................ ............ 65


9 LIST OF FIGURES Figure page A 1 Flow Chart of the Multi Ethnic Study of Atherosclerosis (MESA) with Exclusion Criteria Presenting the Final Sample Size ................................ .......... 82 B 1 Ar eas Under the Curve (AUC) comparing the predictive ability of the reference model compared to the models with the quartile of WHR ................... 83 C 1 Area Under the Curve (AUC) of the Validation Test in Group 2. ........................ 84 D 1 Area Under the Curve (AUC) of the Final Model in Nor mal Weight and Overweight Population in Group 2. ................................ ................................ ..... 85


10 LIST OF ABBREVIATIONS AHA American Heart Association ACC American College of Cardiology ADA American Diabetes Association AUC Area Under Curve ASCVD Atherosclerotic cardiovascular disease ASSIGN ASsessing cardiovascular risk using SIGN guidelines BIA Bioelectrical Impedance Analysis BMI Body Mass Index CACS Coronary Artery Calcium Score CDC Center for Dis ease and Prevention Controls CRF CardioRespiratory Fitness CVD Cardiovascular Disease CHD Coronary Heart Disease CT Computed Tomography scan DXA Dual energy X ray Absorptiometry scan FHS Framingham Heart Study FHS OS Framingham Heart Study Offspring Study HDL C High Density Lipoprotein Cholesterol IL 1 Interleukin 1 MESA Multi Ethnic Study of Atherosclerosis MRI Magnetic Resonance Imaging NFL National Football League NHLBI National Heart, Lung and Blood Institute NHANES National Health and Nutrition Examin ation Survey


11 NIDDK National Institute for Diabetes, Digestive, and Kidney Diseases NWO Normal Weight Obesity ROC Receiver Operating Characteristics SAD Sagittal Abdominal Diameter TNF Tumor Necrosis Factor TOS The Obesity Society USPSTF United States Preventive Services Task Force WC Waist Circumference WHR Waist to Hip Ratio WtoHR Waist to Height Ratio WHO World Health Organization %BF Percentage of Body Fat


12 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy DEVELOPMENT OF C ARDIOVASCULAR DISEASE RISK SCORE BY ADDING A BODY COMPOSITION ASSESSMENT By Ara Jo December 2017 Chair: Arch G. Mainous III Major: Health Services Research Cardiovascular disease has been a top leading cause of mortality in the U nited S tates (US) In order to assess cardiovascular disease (CVD) risk, the risk score was developed in the late 1990s to decrease CVD prevalence and mortality. It is an affordable and effective tool to estimate future CVD risk in clinical settings. However, to take into account obesity epidemic, some risk scores h ave adopted Body Mass Index (BMI) into their model. Recent studies have found that BMI may cause misclassification of people who fall in the normal weight and moderate overweight range. Therefore, the purpose of this study was to develop a CVD risk score b y adding a body composition assessment for adults ages 44 and older. This study used the M ulti E thnic Study of A therosclerosis (MESA) dataset, collecting asymptomatic population data with diverse races and ethnicities (n=5,483). It was randomly divided in to two datasets of equal size. One data set was used to create the risk score, and the other used to validate the model. As a result, the risk model included eight classical CVD risk factors: age, sex, systolic blood pressure, hypertensive medication, diabe tes, total cholesterol, high density lipoprotein cholesterol


13 (HDL C), and smoking. The quartile of waist to hip ratio (WHR) was used as the optimal body composition assessment. By using Cox proportional hazards regression with a 10 year follow up period, n on invasive and affordable risk scores were derived with moderately good performance. Scores of risk factors have been established based on hazard ratios. This risk score was validated (AUC, 0.746). Specifically, the model showed better predictive accuracy in the normal weight and the overweight population (AUC, 0.771). This developed CVD risk score may contribute to improving health care decision making when providing treatment, particularly for populations who may be misclassified as healthy or unhealthy as a result of their BMI score. Moreover, given the estimated potential CVD risk, it may assist in effectively managing the heath of t his population.


14 CHAPTER 1 INTRODUCTION Cardiovascular disease (CVD) has been a top leading cause of death and a primary reason for high economic burden in the United States (US) for decades. Several risk factors such as hypertension, diabetes and smoking for developing CVD have been identifie d. Ob esity in particular is considered a primary risk factor associated with CVD due to sub stantial prevalence and the strong association of chronic diseases Obesity is typically defined by body mass index (BMI, kg/m 2 ), which is a simple equation of weight divided by height squared. However, since body weight reflects several body compositions such as fat, bone mass, muscle mass or body water, BMI alone may not accurately assess the true amount of body fat. Thus, the use of BMI alone may misclassify population s at low risk as being intermediate risk in reality such as normal weight with excessive fat or overweight with lower fat. Misclassification disrupts accurate CVD risk estimation. For instance, people with normal weight who have excessive fat are prone to suffer from metabolic syndrome, prediabetes or diabe tes. To improve predictive accuracy of risk estimation, a variety of body composition factors (i.e., waist circumference and grip strength) has paid more attention. B ody composition assessments ar e a simple non invasive and lower cost method to measure b ody fat. These assessments have been shown to play important roles in stratifying population s who fall into the borderline of normal weight and obesity. This in turn allows for improv ements in population stratification with enhanced predictive accuracy. R i sk score is a practical tool to estimate the potential risk of CVD development. It provides objective evidence of current health states to patients and it enhances health


15 care provider s decision making in clinical setting s Developed in 1970, t he Framing ham risk score (FRS) is the initial CVD risk score. Current risk scores include several common demographic s (i.e., age and sex) and biomarkers (i.e., systolic blood pressure, total cholesterol and diabetes) O besity related factors such as BMI or waist cir cumference are not commonly included in risk scores M any researchers have attempted to include specific biomarkers to the FRS ; however, there is limited evidence that supports including body composition factors in risk estimation. Therefore, the proposed study aim s to identify the optimal body composition assessment in CVD risk estimation and develop CVD risk score with a chosen body composition assessment. The goals of the proposed study are to: 1) determine an optimal body composition assessment to predi ct CVD risk ; 2) develop the CVD risk score by adding a selected body composition assessment ; and 3) test validation of the risk model, and to examine improvement of predictive accuracy in normal weight and overweight population s who would be at intermediat e risk of CVD.


16 CHAPTER 2 LITERATURE REVIEW Cardiovascular Disease (CVD) C ardiovascular disease (CVD) is defined as disorders of the heart and blood vessels that can cause a heart attack or stroke (A merican H eart A ssociation 2014) Specifically, A therosclerosis a blood clot or plaque that blocks the blood flow to the heart or to the brain can damage the brain cells or the heart muscle which may result in death. Coronary h eart d isease (CHD), arrhythmia, heart attack and stroke are all forms of CVD CVD has been a top leading cause of death in the United States since 1969 (Ma, Ward, Siegel, & Jemal, 2015) While its rate decreased to about 65.7% between 2011 and 2013, a recent study reported the total number of CVD related deaths increased in 2011 to 2014 (Ma et al., 2015; Sidney et al., 2016) T he prevalence of CVD was projected to reach more than 40% of the US population by 2030 ( American Heart Association 2016; Heidenreich et al., 2011) Furthermore, CVD was associated with a vast economic burden. Total direct medical cost s of CVD we re expected to double to $818.1 billion and the total costs of CVD wa s projected to reach $1 trillion in 2030 (Heidenreich et al., 2011) With healthcare having reached $3.2 trillion in the US in 2015, the growth of CVD costs will continue to be a burden for the US economy. To combat the increased threat of CVD in public health, the American Heart G the number of CVD and stroke deaths by 20 percent by the year 2020 (Roger et al., 2011) AHA recommend s seven effective approaches to improve cardiovascular health. The approaches for CVD prevention include d metabolic


17 biomarkers such as no rmal range of blood pressure (below 120/80mmHg), cholesterol (below of 170mg/dL), weight (below of 25kg/m 2 ), blood glucose (less than 100mg/dL) ; having a non smoking status (never smoked or having quit for 12 months or longer) ; lifestyle interventions such as regular physical activity (150 minutes per week for moderate intensity activity or 60 minutes per week for vigorous intensity activity) ; and healthy diet (eat 4 to 5 components of nutrition) (Lloyd Jones et al., 2010) T hese recommendations and approac hes were established based on common CVD risk factors. Risk factors for CVD, such as metabolic biomarkers and demographics were identified using the Framingham Heart Study (FHS) (Black, 1992; Kennel, O'Agustino, & Betanger, 1968) Starting in the late 1970s, FHS recruited 5 127 participants age s 30 to 62 in Framingham, Massachusetts in 1948 R health every two years for more than 30 years (Black, 1992) Using epidemiological results, this resea rch study contributed identifying several CVD risk factors. To become CVD risk factors, they must meet criteria. Specifically, these risk factors should be statistically associated with CVD regardless of age, gender or race. The associations also should be prove n by experimental studies with biological sense and the factors must be a contributor of increased risk of developing CVD (Black, 1992) Therefore, s everal major risk factors were selected and have been widely used to estimate risk of CVD. The f ollow ing section will present a current list of risk factors. Risk Factors Demographics Demographics are non modifiable factors. CVD risk increases with ag e More than a third of CVD mortality wa s attributed to older adults over 65 years of age (Alan S.


18 Go et al., 2014) O lder adults we re also more likely to develop chronic diseases which may result in increased risk of CV D (Alan S Go et al., 2014; Menke, Casagrande, Geiss, & Cowie, 2015) S ince nothing can delay aging to prevent CVD it needs to pay more atte ntion to healthy aging with diet and maintaining fitness. Gender is another predictor of CVD. Incident rates of CHD in men w ere three times higher than in women and CHD mortality among men was five times greater compared to women (Jousil ahti, Vartiainen, Tuomilehto, & Puska, 1999) According to Jousilahti and colleagues, most risk factors were observed lower risk for women whereas high smoking rates was shown in men (Jousilahti et al., 1999) Moreover, research indicated that estrogen th e primary female sex hormone may play a protective role and thus gender difference s might be a strong predictor of CVD (Stampfer & Colditz, 1991) Lastly, race and ethnicity were also considered to be a major risk factor of CVD. In 2010, African American m en had the highest CVD death rates (369.2 per 100,000 person) followed closely by African American women (260.5 per 100,000 person) (Alan S. Go et al., 2014) A mong non Hispanic African American adults age s 20 and older, 4 6 % of men and 48% of women had C VD This same group saw 46,081 male and 47,130 female CVD related deaths in 2011 ( American Heart Association / American Stroke Association 2015) The p rimary reason for CVD health disparities wa s the high prevalence of hypertension and diabetes as risk fa ctors. Forty four percent of African American adults have hypertension in 2010 and almost 22% ha d diabetes in 2012 (Alan S Go et al., 2014; Menke et al., 2015)


19 Cholesterol High cholesterol i s the total cholesterol greater or equal to 240mg/dL High d ensity l ipoprotein c holesterol (HDL C) of lower than 40mg/dL is known to be one of primary predictors of CVD ( National Heart, Lung, and Blood Institute 2002) Hyperlipidemia is a form of high blood lipids over the body and fat type lipids may accumulate i n the vessels, blocking blood circulation. Approximately 31.9 million US adults had higher total cholesterol in 2010 ( Go et al., 2014) Moreover, lower leve l of HDL C wa s associated with higher CVD mortality (Wilson, Abbott, & Castelli, 1988) High b lood pressure High blood pressure is an independent risk factor for CVD. H i gh blood pressure w a s determined by measured blood pressure that is greater or equal to 140mmHg of systolic blood pressure or greater or equal to 90mmHg of diastolic blood pressure (Chobanian et al., 2003; Alan S Go et al., 2014) Over 70% of people who had CVD was attributab le to high blood pressure ( Go et al., 2014) and more than 28% of US adults had hypertension in 2010 (Egan, Zhao, & Axon, 2010; Yoon, Burt, Louis, & Carroll, 2012) Although over half of people with high blood pressure are controlled, it is still ma jor contributor of CVD (Egan et al., 2010) Diabetes Diabetes is defined as a group of metabolic diseases cause by abnormal glucose level of greater or equal to 6.5 % of HbA1c level or greater or equal to 126 mg/d L of fasting plasma glucose level (A meric an D iabetes A ssociation 2010) While diabetes w a s independently associated with complications such as nephropathy, retinopathy or peripheral neuropathy, many epidemiologic al studies found that diabetes w a s highly associated with CVD and CVD mortality ( A merican D iabetes A ssociation 2010; Hu et


20 al., 2001; Wei, Gaskill, Haffner, & Stern, 1998) The mortality rate of patients with diabetes was 4.7 times greater than patients without diabetes (Wei et al., 1998) The prevalence of diabetes had steadily increa sed among US adults, to a rate of 12.3% in 2012, and as such, should be considered a major risk factor for CVD (Menke et al., 2015) As a result, AHA included diabetes as a major risk factor for CVD in its clinical guideline (Goff et al., 2014) Smoking S moking is a modifiable behavioral risk factor of CVD. According to the Centers for Disease Control and Prevention (CDC), smoking accounted for 32.7% of deaths from CVD among adults over 35 years of age in the U S (Centers for Disease & Prevention, 2008) Th e relationship between smoking and CVD is that chemicals in tobacco may cause blood vessels to swell or be inflamed; these damaged vessels may result in CV D (CDC, 2014) Obesity Obesity is a strong predictor of CVD Obesity is defined as excessive or abnormal body fat accumulation over the body ( World Health Organization 2016) It is typically measured by Body Mass Index (BMI, kg/m 2 ). The World Health Organization (WHO) established clinical BMI cut points of 18.5 kg/m 2 for norm al, 25 kg/m 2 for overweight and 30 kg/m 2 for obesity in order to classify population s based on BMI association s and mortality ( World Health Organization 1995) As BMI and mortality present J shaped association, obesity itself is associated with higher mo rtality rate (Adams et al., 2006) Obesity wa s caused by a combination of genetic, community environmental and behavioral factors. For genetic factors, a specific gene defect can result in the disruption of physiological pathways to control the development of obesity,


21 which in turn, may results in obesity Community environment also cause d the development of obesity. For example, a lack of access to a physical activity related facility such as a gym and an increasing amount of ti me spent in a car may induc e a sedentary lifestyle, which wa s associated with high prevalence of obesity (Frank, Andresen, & Schmid, 2004; Gordon Larsen, Nelson, Page, & Popkin, 2006) Lastly, a lack of physical activity or excessive fat intake may disrupt gy balance. When energy intake wa s greater than energy consumption, extra fat accumulates in the organs and can lead to common chronic diseases (Heymsfield & Wadden, 2017; Larsen et al., 2014) Obesity wa s closely linked to high mortality and se veral chronic diseases such as hypertension, diabetes, dyslipidemia, nonalcoholic fatty liver disease, chronic kidney disease (Adams et al., 2006; Black, 1992; Heymsfield & Wadden, 2017; Jensen et al., 2014) Individuals who were obese showed a greater risk of developing hypertension as a result of the increases in blood pressure and volum e (Lavie, Milani, & Ventura, 2009; Wilson, D'agostino, Sullivan, Parise, & Kannel, 2002) They also had seven times higher odds of developing diabetes and two times higher odds of developing high cholesterol compared to people with normal weight (Mokdad et al., 2003) The prevalence of o besity continues to increase. T he most recent trend study reporte d that prevalence of obesity has doubled since 1980 with a pr evalence rate of 37.7% in men and 40.4% in women in 2014 (Flegal, Kruszon Moran, Carroll, Fryar, & Ogden, 2016) In an attempt t o seize th is trend, the United States Preventive Services Task Force (USPSTF) recommends adults who hav e a BMI of 30kg/m 2 or gre ater be provided with counseling or behavioral interventions for weight loss (U.S. Preventive Services Task Force., June,


22 2012) Contrastingly, recent studies have found protective effect of obesity and worse prognosis of heart attack in population with no rmal weight. This arise s from the use of BMI only when defining obesity. Physiological mechanism between body fat and CVD Abnormal or excessive body fat accumulation is a primary cause of metabolic disorders. The physiological system store s nutrient surplus as a form of fat tissue around the abdomen, organs, skeletal muscle or tissues (Heymsfield & Wadden, 2017) This fat tissue controlled amount of fat according to energy balance needs (Tchkonia et al., 2013) Excessive body fat accumulatio n called to obesity occurs when energy intakes exceed ed energy consumption Especially, sedentary lifestyl e has been epidemic and people we re less likely to consume their energy. Cumulative body fat elevates proinflammatory cytokines such as tumor necrosi s factor (TNF) 1 (Heymsfield & Wadden, 2017) They we re produced by fat tissue and play roles in regulating infection, immune response and inflammation (Dinarello, 2000) Increase of proinflammatory cytokines disrupts impaired insulin signaling and elevated insulin resistance (Heymsfield & Wadden, 2017) This abnormal process eventually may cause diabetes and CVD. Body fat depots we re associated with CVD risk factors. It is well documented that abdominal obesity played a significant r ole in the increase of insulin resistance and CVD (Barreira et al., 2012) In addition, to date, visceral fat which is observed underneath subcutaneous fat and around organs has been paid attention as a risk factor of CVD. One study found that visceral f at located around the organs can cause inflammation around organs and can consequently cause hypertension and insulin resistance


23 (Fontana, Eagon, Trujillo, Scherer, & Klein, 2007) Particularly, since visceral fat stores lipids, increased visceral fat wa s associated with insulin resistance (Katsuki et al., 2003) In addition, accumulation of excessive visceral fat around kidney increased abdominal pressure and it can cause hypertension (Hall et al., 2010) Likewise, the primary cause of obesity induced CVD risk factors and CVD was body fat instead of body weight. Therefore, accurate measurement of body fat needs to be considered when predicting CVD risk. Misclassification of BMI BMI is widely used to def ine excessive body fat or obesity. Preliminary studies have shown that BMI wa s highly correlated to percentage of body fat (%BF) and cardiovascular mortality (Gmez Ambrosi et al., 2012; Ortega, Lavie, & Blair, 2016) B ody weight includes the weight of or gans, bones, waters, muscles and fats and thus the use of BMI only may misclassify the population placed at borderline between the normal and overweight range. B ody weight cannot measure the pure fat that determine s obesity. A n early report of a WHO expert committee previously addressed concerns of the use of BMI only when defining obesit y ( World Health Organization 1995) The report noted that clinical cut offs were arbitrarily chosen based on association between BMI and mortality ( World Health Organization 1995) Therefore, BMI is not a perfect proxy of obesity measurement. Other studies have found that prevalence of obesity measured by BMI alone was different from prevalence of true obesity defined by body fat percentage In one study, o ne thi rd of men were misclassified as obes e due to different contributors of body weight such as muscle mass and muscle density (Pasco et al., 2014) Similarly, Gomez Ambrosi and colleagues found in their study that the majority of people categorized as overweig ht based on BMI w ere actually obese when


24 using body fat percent (Gmez Ambrosi et al., 2012) Additional studies have found using BMI alone misclassified population s at risk of mobility impairments, cardiometabolic diseases and CVD (Peterson, Al Snih, Sto ddard, Shekar, & Hurvitz, 2014; Silventoinen, Magnusson, Tynelius, Batty, & Rasmussen, 2009) According to Peterson and colleague s about 17% of individuals with high body fat percent who w ere misclassified as normal BMI ( usually referred as low risk ) showed higher prevalence of metabolic syndrome (Peterson et al., 2014) For example, while most Asians we re s hown to be normal weight, they we re more likely to have diabetes and CVD (Kim et al., 2014; Raji, Seely, Arky, & Simonson, 2001) Multiple studies suggest ed lower cut offs of BMI for the Asian population (Barba, Cavalli Sforza, Cutter, & Darnton Hill, 2004) These arguments indicate d that the use of BMI only to define obesity may not be the best proxy in predicting CVD risk A n example of BMI miscla ssification can be seen in professional athletes. One study demonstrat ing CVD risks in National Football League (NFL) players found that professional football players had lower prevalence of CVD risk factors (i.e., impaired fasting glucose, smoking, and dy slipidemia) compared to healthy men of similar age ; however, the study found that more than half of the professional football players were obese based on BMI alone (Tucker et al., 2009) In addition, as opposed to conventional deteriorated effect of obesity, the o besity paradox explained the ironic protective effect of obesity defined by BMI. The o besity paradox explained better prognosis or lower mortality rate of overweight or obesity pa tients with known CVD compared to individuals who have normal BMI (Asgari et al., 2016; Curtis et al., 2005; Lavie et al., 2016; Oreopoulos et al., 2008) Whi le one meta analysis argued that the


25 protective effect of obesity attenuated with mortality (Wang et al., 2015) other studies still advocated for the benefits of obesity among individuals with heart failure or coronary artery disease (Curtis et al., 2005; Oreopoulos et al., 2008; Romero Corral et al., 2006; Shakiba, Soori, Mansournia, Nazari, & Salimi 2016) Meanwhile regardless of having any type of CVD, people with normal weight abdominal obesity had twice higher mortality than people with overweight or obesity (K. R. Sahakyan et al., 2015) This inverse relationship also can be found in inpatient mortality where patients who we re obese showed a slightly longer length of stay (Fonarow et al., 2007) Obesity Paradox and Normal Weight Obesity Prognosis of patients with overweight classification or obesity is also associated with the obesity paradox. Patients receiving h emodialysis showed improved survival over 5 years (Kalantar Zadeh et al., 2010) and patients who received percutaneous coronary interventions with corona ry artery disease had greater 5 year survival compared to normal weight p atients (Hastie et al., 2010) In addition, being metabolically healthy or not may play an important role in increased CVD risk regardless of being overwe ight or obese. Individuals who we re metabolically healthy we re individual whose metabolic biomarkers such as glucose level, blood pressure, cholesterol level and insulin resistance we re placed within normal range. A growing body of evidence has shown that people with normal weight who we re metabolically unhealthy had greater risk of CVD than people with overweig ht or obesity who we re metabolically healthy (Artham, Lavie, Milani, & Ventura, 2008; Voulgari et al., 2011) This evidence indicate s obesity defined by BMI may require a question of independent risk factor of CVD risk assessment.


26 N ormal weight obesity ( NWO) refers to individuals who fall in the normal range of BMI (18.5kg/m 2 25kg/m 2 ) and who ha ve excessive body fat (Romero Corral et al., 2009) Preliminary studies found that people with NWO showed higher prevalence of metabolic syndrome and higher mortal ity rate (Batsis et al., 2013; Romero Corral et al., 2009; Shea, King, Yi, Gulliver, & Sun, 2012) A bout 18% of the p opulation with normal weight had prediabetes and its trend of prediabetes in normal weight ha d significantly increased (Mainous, Tanner, Jo, & Anton, 2016) Furthermore, adults with normal BMI who had sedentary lifestyle we re more likely to have prediabetes or diabetes (Mainous, Tanner, Anton, Jo, & Luetke, 2017) Despite the risk of misclassification of the use of BMI shown by several stud ies, people at normal BMI we re usually neglected in chronic disease screening guideline s offered by the USPSTF. For instance, diabetes screening is recommended people aged between 40 to 70 years old who are overweight or obesity defined by BMI ( U.S. Preventive Services Task Force 201 5) T he screening guideline recommends providing intensive behavioral counseling to adults who are overweight or obese measured by BMI and who have CVD ( U.S. Preventive Services Task Force 2014) This indicate s that people who a re place d at borderline of BMI and who a re at high risk of any type of chronic disease may miss a n opportunity to receive appropriate healthcare recommendations to prevent the onset of chronic diseases. Therefore, the use of BMI alone may miss people at low risk who would be at intermediate high risk of CVD dev elopment Body Composition Assessments A b ody composition assessment is used to quantify current body composition that affects physical function s and health states. It has played an important role particular ly in sport s medicine, because some weight sensitive sports such as boxing or


27 rhythmic gymnastics requires to frequently monitor ing body weight and subtle change s in body mass which can result in a competitive advantage (Ackland et al., 2012) Moreover, the assessment can provide essential guideline s for training and nutrition to athletes due to the strong correlation to body composition (Nana, Slater, Hopkins, & Burke, 2013) M onitoring body composition allows for population stratification of those at risk of disease Recent studies have shown that body composition assessments contributed to identifying population s at high risk for chronic diseases (Mainous, Tanner, Anton, & Jo, 2015, 2016; Mainous, Tanner, Jo, et al., 2016) Among population s with n ormal BMI, lower g rip strength w a s associated with hypertension, prediabetes and diabetes (Mainous et al., 2015; Mainous, Tanner, Anton, et al., 2016) This indicates that body composition assessment s may stratify population s at risk who are typically neglected in health examination. Body composition assessment s are a convenient and non invasive measurement in risk assessment. W hile its cost s may vary depending on type of techniques i t is relatively inexpensive and can be ea sily us ed in any healthcare s etting To date, there is BMI is widely used to identify obese population s in several risk assessments, it is not included in current CVD risk assessment. Wilson and colleagues have shown th at BMI was a significant factor in CVD risk predictio n (Wilson et al., 2008) However, as aforementioned, BMI only assessments may misclassify people who fall into the borderline between normal and overweight BMI who may be at risk for CVD As substitutes, diverse body compositions have been considered as a risk factor for several chronic diseases. However body composition


28 assessment s ha ve been controversial in identifying people at high risk of developing CVD and ha ve not been widely utilized with respect to risk estimation in routine clinical setting s Numerous body composition assessments have been proposed in clinical setting s Some require simple measurements using a scale or tapeline, whereas others demand expertise and a high technique laboratory machine such as c omputed t omography (CT) scan s, d ual energy x ray a bsorptiometry (DXA) scan s, b ioelectrical i mpedance a nalysis (BIA) or m agnetic r esonance i maging (MRI). Despite their accuracy in measurement, these machines are extremely expensive for the purpose of measur ing body composition and estimating CVD risk in asymptomatic population s. Furthermore, a lack of health insurance coverage may be a barrier in utilizing these machines. The proposed study will propose t he use of inexpensive and non invasive body composition assessment s The f ollowing se cti on will provide details about the different type s of body composition assessments Waist Circumference (WC) Waist circumference (WC) is a strong indicator of abdominal obesity (Pouliot et al., 1994; Karine R Sahakyan et al., 2015) It w a s highly correlated with body fat mass and percentage of body fat (Barreira et al., 2012) Research has shown that waist circumference wa s significantly associated with CVD incidence and mortalit y, although the results were varied between sex (Aune et al., 2016; De Koning, Merchant, Pogue, & Anand, 2007) Specifically, for every 1cm increase d in WC there wa s an associated 2% increase in the relative risk of CVD (De Koning et al., 2007) Several s tudies have reported that WC wa s a better predictor for mortality and CVD risk factors than BMI (Kartheuser et al., 2013) Since 1988, the mean WC for men went from 96.0cm t o


29 100.4cm and from 89.0cm to 94.0cm for women (Li, Ford, McGuire, & Mokdad, 2007) At the same time, the prevalence of abdominal obesity significantly increased to 42.2% in men and 61.3% in women (Li et al., 2007) A mong several indices of a bdominal obesity, WC w a s the preeminent indicator of body fat mass (Barreira et al., 2012) WC is a simple and portable measurement of abdominal obesity by using a line tape. An examiner mark s the midpoint between the lower ribs and the iliac crest and the line tape is brought around the waist following a normal exhale (Sampaio, Simes, Assis, & Ramos, 2007) This method may be prone to measurement error depending on able to estimate whole body fat Waist to Hip Ratio (WHR) Waist to h ip r atio (WHR) is also a common indicator of the relative accumulation of ab dominal fat (Consultation, 2008) In fact, unlike WC, WHR has a higher correlation to fat distribution and is relatively lower correlated with body fat mass compared to WC (Barreira et al., 2012) However, WHR wa s better suited in identifying individuals at higher risk of developing obesity associated health complications (Smith, 2016) study confirmed that WHR was a stronger predictor of CVD events compared to WC (De Koning et al., 2007) These findings can be seen in the case o f the NFL players Whereas most players were place d within the obesity range as defined by BMI, the ir mean WHR was lower than cut off point for obesity; this may explain lower risk of mortality and favorable risk factors (Tucker et al., 2009) WHR is compu ted by WC and h ip circumference. Hip circumference is measured by using the line tape similar to measuring the WC. An examiner measure s at the widest circumference point around the hip (Sampaio et al., 2007) The WHO provide d


30 clinical cut off points of 0.90 in men and 0.85 in women ( World Health Organization 2008) Waist to Height Ratio (WtoHR) Waist to h eight r atio (WtoHR) is a n indicator of abdominal obesity (Ashwell & Gibson, 2016) It play ed a role in predicting diabetes (MacKay, Haffner, Wage nknecht, D'agostino, & Hanley, 2009) and it contribute d to identifying early health risk s among people who are considered normal weight (Ashwell & Gibson, 2016) I t was also associated with higher risk of cardiometabolic risk factors (Ashwell & Gibson, 201 6) Specifically people with normal weight with an unhealthy WtoHR were observed as having a high risk of prediabetes (Mainous, Tanner, Jo, et al., 2016) Since WtoHR w a s highly correlated with body fat mass and percentage body fat, it may alter BMI (Barreira et al., 2012) A meta study found that WtoHR wa s a better indicator of cardio metabolic risk factors than BMI or WC (Ashwell & Gibson, 2016) Consistent cut off points have not been established in clinical setting s CVD Risk Score CVD r isk score is a tool used to estimate future risk of CVD within 10 years or a lifetime. The g iven score s allow patients to know their current CVD risk status Using these risk scores, h ealthcare providers, in particular physicians, are able to recommend timely and appropriate health care services and can provide health management services to patients for the next step in screening (Koopman & Mainous, 2008) In clinical setting s risk assessment s showed effectiveness in reducing CVD risk and prescribed statin wa s dramatically increased (Artac, Dalton, Majeed, Car, & Millett, 2013) Another advantage of CVD risk scores is that it can lead to improve d patient physician communication and the score s may lead to better clinical decision making by


31 physicians (Koopman & M ainous, 2008) In Europe, CVD guideline s have recommend the use of risk score model s in health examination s as a prevention strategy since 1994 (Pyrl, De Backer, Graham, Poole Wilson, & Wood, 1994) Risk score played a role in stratifying patients into different risk levels (Koopman & Mainous, 2008) H ealth care providers can take advantage of patient classification s in determining the level of treatment and identify priority group s who need preventive services Several CVD risk evaluations have been de veloped for different cohort population s. For example, the Framingham r isk s core (FRS) was developed for non Hispanic Whites (Lloyd Jones et al., 2004) the Reynolds risk score for women (Ridker, Buring, Rifai, & Cook, 2007) the QRISK for general populati on in England (Hippisley Cox et al., 2007) the Assessing cardiovascular risk using SIGN guidelines (ASSIGN score) for Scottish populations (Tunstall Pedoe, Woodward, Tavendale, A'brook, & McCluskey, 1997) the SCORE model for the general population in Eur ope (Conroy et al., 2003) and the Atherosclerotic cardiovascular disease (ASCVD) risk score used for US populations (Goff et al., 2014) T he FRS was the original CVD score and most assessments developed derived from the FRS. Thus risk scores share d common factors such as demographics (i.e., age and sex) and metabolic biomarkers (i.e., systolic blood pressure, HDL C, and diabetes), while a few variables differ For instance, QRISK and ASSIGN score s include d social deprivation factor s in order to redu ce health inequality in CVD risk (Hippisley Cox et al., 2007; Tunstall Pedoe et al., 1997) The FRS and the ASCVD focus ed on metabolic biomarkers (Goff et al., 2014; Lloyd Jones et al., 2004) The ASCVD included race to take into account genetic effect and it also consider ed several biomarkers such as C reactive protein, apolipoprotein B,


32 family history in population classification. (Lloyd Jones et al., 2010) Finally, the Re s on women (Ridker et al., 2007) Existing scores have proven their validation by applying their measures to other population s and by comparing the other scores. For example QRISK developed in England was showed valid CVD score s among UK primary care population and it compared scores with the FRS and the ASSIGN (Hippisley Cox et al., 2007) The ASCVD was verified among a contemporary US population (Muntner et al., 2014) To derive estimated risk score, these evaluations assign a specific value on each risk factor based on statistical proba bility and estimate s the sum of these scores (Sullivan, Massaro, & D'Agostino, 2004; Wilson et al., 1998) Chapter 2 will articulate how to calculate the risk score. Framingham Risk Score (FRS) FRS derived from the Framingham Heart Study in 19 98 and it wa s the first measurement of CVD risk estimation among adults (Wilson et al., 1998) This score included eight traditional risk factors : age, sex, diabetes, smoking status, measured systolic blood pressure, history of diagnosed hypertension, total cholestero l and HDL cholesterol The FRS was initially created for non Hispanic White populations, and as such was considered to be less valid in predictive power in regards to race and ethnicity. However recent studies have verified the validity and reliability of the score in diverse population s (D'Agostino Sr, Grundy, Sullivan, & Wilson, 2001) with men only in the US (J. Gander, 2014) with men only in the UK (P. Brindle et al., 2003) older population s (Rodondi et al., 2012) and populations with different socioeconomic status (P. M. Brindle et al., 2005) Specifically, the biggest concern that the FRS derived from white


33 dominant population was solved with good performance of the model in multi ethnic g roups (D'Agostino Sr et al., 2001) In addition, C index which measures discrimination of risk score s among older adults between the ages of 70 and 79 was not significantly different from the original FRS (p=.54) and thus the FRS was valid in older adul ts as well (Rodondi et al., 2012) However the FRS has limitation s in stratifying population s who fall into the blind spot of risk because of an absence of obesity related factor s in the model. Adding a New Indicator Adding new indicators to the risk as sessment might contribute to improving accuracy of risk estimation. Numerous studies have suggested adding novel biomarkers such as C reactive protein, c oronary a rtery c alcium s core (CACS), ankle brachial index and genetic markers to the risk model (Albert Glynn, & Ridker, 2003; Collaboration, 2008; Greenland, LaBree, Azen, Doherty, & Detrano, 2004) Including ankle brachial index to the FRS reclassified 19% of men and 36% of women into risk categories (Collaboration, 2008) and combined models with CACS an d the FRS showed improvement of predictive risk among asymptomatic populations (Greenland et al., 2004) Similarly, adding innovative markers to existing models may allow for predict ion with enhanced accuracy and may contribute to better patient classification Adding many risk factors to the model, however, is not necessary to enhance predictive accuracy. There are two aspects of predictive accuracy T he first is a reliable prediction and the second is discrimination (Harrell, Lee, Califf, Pryor & Rosati, 1984) As several factors w e re involved in the model, there might be a complex effect of interaction s among these factors. It may not be able to reproduce a reliable prediction. Furthermore, the ability of a model classify patients by the outco me is called as


34 discrimination (Harrell et al., 1984) However if there wa s an interaction effect, the quality of a model may be deteriorated. According to the parsimonious principle least number of factors that a re strongly associated with the outcomes would be appropriate to include in the model (Ridker et al., 2007) As previously mentioned, a dding new innovative biomarkers to existing models has been shown to predict CVD risk with similar predictive ability (Albert et al., 2003; Collaboration, 2008; Greenland et al., 2004) Unfortunately, despite the importance of body composition to date, few studies include body composition component in CVD risk model. Furthermore, there is no gold standard estimator and existing risk scores do not reflect obesity r elated mechanism by excluding body composition assessment s Therefore, this proposed study will use the FRS to develop risk score s with improved predictive accuracy. Specific Aims T he goals of the proposed study we re: 1) to determine the optimal body composition assessment to predict CVD risk, 2) to develop CVD risk score by assigning numeric value to each variables, and 3) to test validation of the developed CVD risk model and improvement of predictive accuracy in normal weight and overweight populations. Aim 1. Select an O ptimal Body Composition A ssessment The proposed study identifies an optimal body composition assessment to predict CVD risk Candidates with body composition assessments were chosen based on pre liminary studies (Ashwell & Gibson, 2016; De Koning et al., 2007) Each body composition is evaluated to meet the assumption of survival analysis and multicollinearity tests. S everal models including each qualified body composition


35 assessment from assumpti on tests and multicollinearity tests are specified After comparing specified models in regards to predictive accuracy, the final model with the highest capability of prediction was chosen. H 1 : The model with an optimal body composition assessment will pr edi ct CVD risk with better predictive accuracy compared to the reference model without a body composition assessment. Aim 2. Develop the F inal R isk S core Once the best fit model with the optimal body composition assessment was identified numerical values will be assigned to each level of variables according to the Hazard Ratio (Mainous et al., 2007) Aim 3. Test a V alidation of D eveloped M odel and I mprovement of P redictive A ccuracy in N ormal W e ight and O verweight P opulation s The risk sc ore tests two hypotheses. Prior to the test of the first hypothesis, the study report prevalence of BMI misclassification among normal weight and overweight population s derived from a national representative dataset. This result gives a robust rationale fo r focusing on normal weight and overweight population s The National Health and Nutrition Examination Survey (NHANES) for years between 1999 and 2006, was used. This was f ollow ed by a validation test of the developed model with the selected body composition assessment. In addition, the improvement of predictive accuracy of the final model was examined in individuals who are normal weight and overweight by using the second group fr om the Multi Ethnic Study of Atherosclerosis (MESA). H 2 : The developed model will be validated in different populations. H 3 : The developed model will show greater predictive accuracy in normal weight and overweight populations compared to the overall pop ulation


36 CHAPTER 3 METHODS Data Th is study used the Multi Ethnic Study of Atherosclerosis (MESA) to develop a CVD risk score. The MESA was designed to identify a variety of risk factors as well as subclinical diseases defined as noninvasively detected diseases by the National Heart, Lung and Blood Institution ( NHLBI ) in July of 2000 (Bild et al., 2002) The MESA is capable of investigating pathophysiology of CVD and subclinical CVD progress such as conditions of the aorta and coronary arter i es and thus can be used in identifying new CVD risk factors and assessing CVD risk as a prevention strategy (Bild et al., 2002) The MES A derived from a population based cohort sample composed of 6,814 men and women equally between the ages of 44 and 84 who were free of CVD at baseline. Participants were recruited from six US field centers beginning in 2000, and risk factors and subclinica l disease indicators were measured repeatedly for five times with follow up periods of up to 12 years (Bild et al., 2002) A strength o f this cohort is in its racial and ethnic diversity The population of the cohort was approximately 38 % White, 28 % Africa n American, 22 % Hispanic and 12 % Asians ( National Heart, Lung, and Blood Institute 2012 ) This d iversity in race and ethnicity allows generalizability of the developed risk score model. This in turn may enhance clinicians decision making regardless of r acial and ethnic difference s for CVD prevention in routine clinical setting s Furthermore, the MESA data consists of several body composition factors such as waist circumference, hip circumference, total body fat (kg) and body surface area. These component s can examine the impact of the use of body composition and consequently,


37 it allows for better selecting of an optimal body composition assessment for the development of a CVD risk model. The MESA was divided into two groups of equal sample size randomly group 1 and group 2. Group 1 was used to develop a risk score and identify an optimal body composition assessment. Group 2 was used in a validation test of the developed model Figure A 1 illustrate s the flow chart of the final sample size. The Natio nal Health and Nutrition Examination Survey (NHANES) was used to report prevalence of BMI misclassification in the US for adults 44 and older who correspond s to the study population when developing a model. The NHANES is a large, national representative c ross sectional dataset using a complex stratified multistage probability cluster sample design. The survey design with weighting variable s and allows for population estimates calculations. NHANES include s both interviews and standardized physical examinati on s such as urine, blood analysis and body composition examination. As such, it can assess the health and nutritional status of the non institutionalized US population ( National Health And Nutrition Examination Survey ) The data has been collected annually since the early 1960s. Participants were recruited at the county level. Health interview s were completed and health examinations were conducted in mobile health centers. Data was released after it was clean ed de identifi ed edit ed review ed via a disclosure review board and finaliz ed. I t is now a publicly available data set ( National Health And Nutrition Examination Survey 2014) The NHANES consists of a variety of body composition assessments such as BMI, WC and sagittal abdominal diameter. It includes Dual energy X ray Absorptiometry (DXA) data which is widely adopted to measure accurate


38 body fat mass and bone density, and can identify true body fat mass. The current study used the data for the years of 1999 to 2006. Although the most recent data exists, the data used in this study is the most recent data with a whole body DXA which measures percent body fat (%BF). %BF captures the most accurate body fat mass and excessive fat mass. Participants Participants were US a dults ages 44 and older who were free of any type of CVD events such as a heart attack, angina, heart failure, resuscitated cardiac arrest and stroke or transient i schemic attack ( TIA ) at baseline and who have had procedures related to CVD (e.g. CABG, angioplasty, valve replacement, pacemaker or defibrillator implantation or heart surgeries) This specific population was ideal for establishing primary CVD preventive care strategy. Participants who did not have all risk factors (i.e., age, sex, systolic blood pressure, treatment of hypertension, diabetes, total cholesterol, HDL C and smoking) and who missed at least one of body composition assessments (i.e., waist circ umference, waist to hip ratio, and waist to height ratio and BMI) were excluded. Outcomes The primary outcome of the study was the time to the first event of CVD including nonfatal myocardial infarction, resuscitated cardiac arrest, definite angina, prob able angina, stroke, and CVD death CVD death includ ed stroke death, CHD death, other a therosclerotic death and other CVD related death ( National Heart, Lung, and Blood Institute 2012) The outcomes were collected from multiple sources including death ce rtificates, medical records from hospitalizations, autopsy reports, and interviews with participants or physicians, relatives or friends (Bild et al., 2002)


39 Independent Variables Eight of the classic risk factors and seven non invasive body composition assessments were included in the risk score model. All risk factors were modified into categorical variables based on the FRS (Wilson et al., 1998) Table 3 1 illustrated with details. Risk Factors For demographics, age was categorized as follows: 44 49, 50 54, 55 59, 60 64, 65 69, 70 74, 75 and older. Sex was used as coded (i.e. men and women). R ace or ethnicity factor was excluded Hypertension definition followed the initial FRS definition established based on the seventh report of the joint national committee on prevention, detection, evaluation, and treatment of high blood pressure (Chobanian et al., 2003) Hypertension treatment was a binary variable asking for use of antihypertensive medication. The s tudy sample was limited to individuals who did n ot use antihypertensive medications Independently, systolic blood pressure (SBP ) was classified into four groups at cut offs of lower than 130, 130 139,140 159 and 160 mm Hg or greater. Diastolic blood pressure (DBP) was categorized into four levels: lowe r than 85, 85 89, 90 99 and 100 or greater. Systolic blood pressure was defined using four levels: 1) normal (SBP<130mmHg), 2) high blood pressure (SBP: 130 139 mmHg), 3) hypertension stage I (SBP:140 159 mmHg), and 4) hypertension stage II Total cholesterol and HDL C were also considered. They were modified into categorical variables based on the third report of the National Cholesterol Education Program (NCEP) ( National Heart, Lung, and Blood Institute 2002) For total cholesterol, cut offs were lower than 160 (4.14mmol/L), 160 199 (2.60 3.36mmol/L), 200 239 (3.37


40 4.14mmol/L), 240 279 (6.22 7.24mmol/L) and 280 mg/dL or greater (7.25mmol/L or greater). Thresholds for HDL C were less than 35 (less than 0.90mmol/L), 35 44 (0.91 1.16mmol/L), 45 49 (1.17 1.29mmol/L), 50 59 (1.30 1.55 mmol/L), and 60 or greater (1.56mmol/L or greater). Diabetes was a dichotomous variable. People who have ever been diagnosed with diabetes by a doctor or who were under treatment with insulin was categorized as d iabetes. P eople whose blood glucose level measured by fasting plasma glucose (i.e., normal<126mg/dL, 7.0mmol/L ) or Hemoglobin A1c of less than 6.5% ( 48mmol/mol ) was defined as normal, and diabetes was defined to people whose blood glucose level of greater or equal to 126mg/dL ( 7.0mmol/L) or HbA1c of greater than 6.5% ( 48mmol/mol) ( A merican D iabetes A ssociation 2016) Smoking status was categorized into two groups: non smokers who have never smoked in their lifetime or who have quit smoking for more than 12 months ; and current smokers those who smoked regularly in past 12 months (Wilson et al., 1998) Table 3 1 presents the description of how to operate risk factors for the risk score. Body Compositions Seven different body composition assessments were select ed on the list of potential variable s for the risk score. As can be seen in T able 3 2, four typical body composition assessments were modified into categorical variable s with var ying cutoffs. Clinical cutoffs used are indicat ive of most adopted cutof fs in clinical setting s and research. For instance, WC and WHR were recoded based on widely adopted clinical cut points. Sex specific threshold s for WC are 88cm (35 inch) for women and 102cm (40 inch) for men ( World Health Organization 2008) In WHR, cut offs are 0.90 for men and 0.85 for women ( World Health Organization 2008) WtoHR does not have


41 established universal cut points. Thus a widely adopted threshold, 0.5, was chosen based on the literature (Ashwell & Gibson, 2016) To account for the potentia l impact of data distribution of the data, tertile and quartiles were used in developing the model. Integrated body composition assessments with BMI were also considered. Zhu and colleagues initially used a combination of WC and BMI to examine CVD risk factors and resulted in better performance when using the combined assessment s rather than using a single BMI (S. Zhu et al., 2004) Integrated body composition may take into account abdominal obesity which is significantly associated with chronic diseases and effect of regional fat on CVD risk. Therefore, the study used integrated body composition assessment s (i.e., WC, WHR and WtoHR) with BMI. Table 3 3 illustrate s the classification of integrated assessments.


42 Table 3 1. Description of Operation for CVD Risk Factors Variable Operation Clinical Range Demographics Age (year old) Categorical 4 4 49 50 54 55 59 60 64 65 69 70 75 Sex Dichotomous Men Women Risk Factors Hypertension Medication Dichotomous Yes No Systolic Blood Pressure (SBP, mm Hg) Categorical lower than 120 120 129 130 139 140 159 160 mm Hg or greater Total Cholesterol (mg/dL, mmol/L) Categorical <200 (<3.36mmol/L) 200 239 (3.37 4.14mmol/L) 240 279 (6.22 7.24mmol/L) 280 mg/dL or greater (7.25mmol/L or greater) High Density Lipoprotein Cholesterol (HDL C) (mg/dL, mmol/L) Categorical less than 35 (less than 0.90mmol/L) 35 44 (0.91 1.16mmol/L) 45 49 (1.17 1.29mmol/L) 50 59 (1.30 1.55 mmol/L) 60 or greater (1.56mmol/L or greater). Diabetes Dichotomous Diabetes: Diagnosed diabetes or abnormal glucose level (FPG>=126mg/dL,7.0mmol/L or HbA1c>=6.5%, 48mmol/mol) Non diabetes: FPG<126mg/dL,7.0mmol/L HbA1c<6.5%, 48mmol/mol Smoking Status Dichotomous Yes No


43 Table 3 2. Description of Sex Specific Cutoffs of Body Composition Assessments Body Composition Assessment Criteria Cutoffs Men Women Body Mass Index (BMI, kg/m 2 ) Clinical Underweight <18.5 Normal 18.5 24.9 Overweight 25.0 29.9 Obese >30.0 Tertile (%) 33 25.77 29.40 25.33 30.52 67 Quartile (%) 25 24.80 27.53 30.34 24.25 27.64 32.24 50 75 Waist Circumference (WC, cm) Clinical 102 (40 inch) 88 (35 inch) Tertile (%) 33 94 103.5 88.5 102 67 Quartile (%) 25 91.3 98.5 106.5 85 95.5 106.4 50 75 50 75 Waist to Hip Ratio (WHR) Clinical 0.9 0.85 Tertile (%) 33 0.93 0.98 0.86 0.94 67 Quartile (%) 25 0.92 0.96 1.00 0.84 0.90 0.96 50 75 Waist to Height Ratio (WtoHR) Clinical 0.9 Tertile (%) 33 0.54 0.59 0.55 0.64 67 Quartile (%) 25 0.53 0.57 0.61 0.53 0.60 0.67 50 75


44 Table 3 3. Description of Classification of Integrated Body Composition Assessments Combining with BMI Integrated Body Composition Assessments Classification BMI Underweight Normal Overweight Obese BMI & WC WC Normal + + + Obese + + + BMI & WHR WHR Normal + + + Obese + + + BMI & WtoHR WtoHR Normal + + + Obese + + +


45 Statistical Analysis The statistical analyses were performed using SAS 9.4 ( Copyright 2013 by SAS Institute Inc., Cary, North Carolina ) and SUDAAN 11.0 (Cary, North Carolina). The dataset was randomly divided into two groups of equal size. The first group was used to create a risk score and the second group was used to test a validation of the model. First, the descriptive analysis of baseline characteristics for the study population was given using Chi square tests to examine significant differences in risk factors and the CVD incidents. P revalence of CVD was presented. Next, several statistical analyses such as c statistics Receiver Operating Characteristics (ROC) curves and Hosmer Lemeshow chi square tests were performed to estimate discrimination, calibration and goodness of fit. Preval ence of BMI misclassification was analyzed. Specific statistics strategies are described as follows. AIM 1 Select an O ptimal Body Composition A ssessment H 1 : The model with an optimal body composition assessment will predi ct CVD risk with better predic tive accuracy compared to the reference model without a body composition assessment. Seven body composition assessments were selected based on preliminary studies. Out of seven, four individual measurements were used as an individual measurement; 1) body mass index (BMI, kg/m 2 ), 2) waist circumference (WC, inch), 3) waist to hip ratio (WHR) and 4) waist to height ratio (WtoHR). They were modified as categorical variables according to clinical cutoffs, tertile and quartile, as shown in T able 3 2 ( World Health Organization 1995, 2008) In addition, three measurements were integrated forms of body composition assessments with BMI in order to improve classification among normal weight and overweight population who are typically misclassified by the use of BMI only; 5) BMI & WC, 6) BMI & WHR and 7) BMI &


46 WtoHR. Among WC, WHR and WtoHR, they are highly correlated theoretically and statistically so that these were not transformed as an integrated form. These integrated assessments were also operated as categor ical variables based on clinical cutoffs. Table 3 3 described integrated assessments. In order to choose an optimal body composition assessment, several tests were performed using Group 1. First, bivariate analyses were performed to test significant diffe rence between body composition assessments and the CVD incidents. A statistically significant variable was eligible to test proportionality assumption for Cox proportional hazards analyses. Second, proportionality assumption test of each assessment was per formed. Proportional hazards assumption test used Schoenfeld residuals. This test focused on correlation of residuals of variables and time. Specifically, when the correlation of body composition assessment and the time to CVD event is not significant, the assumption was met. It indicated that the body composition assessment did not change over the time to the first CVD event. Lastly, multicollinearity test was conducted. A body composition assessment should not be significantly correlated with conventional risk factors. If a correlation between new factor and existing factors exist it may cause a discrepancy in estimated coefficients. Multicollinearity can be detected by using variance inflation factors (VIF). A VIF showed how much the existence of correla tion between additional factor and existing factors inflates the estimated coefficient. When the VIF is statistically significant, it indicated that multicollinearity does not significantly affect the estimation. These process determined candidates of an o ptimal body composition assessment for development of CVD risk score.


47 Several adjusted C ox proportional hazards models including different qualified body composition assessments were specified and analyzed. The reference model containing eight typical CVD risk factors without any body composition assessments was initially analyzed. Following by, models including each selected body composition assessment were analyzed. C statistics and the area under a curve (AUC) were used to compare d iscrimination of each model and Hosmer Lemeshow Chi square tests were performed to test calibration and goodness of fit of the model (Demler, Paynter, & Cook, 2015) Discrimination represents the ability of the prediction model to distinguish population into two groups who have the outcomes of interests and who do not have the outcomes (Harrell, Lee, & Mark, 1996) To test discrimination, a c index which is widely used measurement of predictive discrimination was computed in each model The c index assesse s predictive accura cy derived from predictors in a model and it generates probability of how concordant predictions and actual observed outcomes are (Harrell et al., 1996) I f the predicted survival time to CVD event is longer for person who actually have never had CVD event in lifetime, the predicted value is concordant with the outcome. A range of the c index value is from .5 which means no predictive discrimination to 1.0 that means perfect discrimination with the outcome (Harrell et al., 1996) The ROC curve is a nother technique to present a plot of the tradeoff between sensitivity and specificity at cut off points (Zhu, Zeng, & Wang, 2010) Two curves are drawn the first curve is 45 degree of diagonal that indicates 50% of sensitivity and 50% of specificity, and the ot her curve is the predicted curve derived from the proposed


48 model. T he fact that the predicted curve comes to closer to the diagonal curve indicates less accurate of the model, whereas cut off point reaches to the upper left corner of the plot called as per fect classification (Zhu et al., 2010) Given the ROC curve, the A rea U nder C urve (AUC) indicate d predictive accuracy A range of the AUC is between 0 to 1.0. If the AUC is below .5, the model has worse predictive accuracy wh ereas 1.0 of the AUC indicates perfect accuracy (Zou et al., 2007) The model which ha d the highest value in c statistics was chosen as the final risk score model. ROC curves were presented as figures in Appendix AIM 2. Develop the F inal R isk S core T o develop a risk score scores was developed based on adjusted Hazard Ratios of the variables (Mainous et al., 2007) Specifically, numeric points were assigned to each level of the variable. If HR of a c ertain level of the variable was not significant, it s cored as zero. AIM 3. Test a V alidation of D eveloped M odel and I mprovement of P redictive A ccuracy in N ormal W e ight and O verweight P opulation Prior to the test for validation of the developed model the prevalence of BMI misclassification was reported by us ing the NHANES for the years of 1999 2006. The NHANES is the large nationally representative dataset and it allows population estimates of the US population. Particularly, the dataset contains body fat p ercentage (%BF) measured by the DXA and it allowed to capture true excessive body fat mass. Thus BMI classified population who are normal weight and overweight and %BF identified excessive %BF in each BMI group. Using weighting variable, chi square test was performed to examine prevalence of BMI misclassific ation in the US adults aged 44


49 and older. It gave a robust rationale of following a validation test focusing on normal weight and overweight population. H 2 : The developed model will be validated in different population. A validation test of the developed model was performed in group 2. C statistics and the ROC curve w ere computed and the Hosmer Lemeshow statistics was also conducted to test calibration and goodness of fit H 3 : The developed model will show greater predict ive accuracy in normal weight and overweight population compared to overall population The final model was designed to improve predictive accuracy in population at low risk who would appropriately be classified as intermediate high risk. This misclassifie d population was normal weight and overweight population. Thus, to assess improvement of predictive accuracy in th ese population s c statistics and the AUC for discrimination and the Hosmer Lemeshow statistics for goodness of fit w ere performed only within normal weight and overweight population in group 2.


50 CHAPTER 4 RESULTS The aims of this dissertation were to identify an optimal body composition assessment and to develop a CVD risk score by adding a selected body composition assessment. This study als o performed a validation test of the developed risk score in different population and in subpopulation of normal weight and overweight population s only. Simultaneously, the study provided prevalence of BMI misclassification at the population level and gave a rationale for focusing on normal weight and overweight population s A s the main purpose of the use of the body composition assessment was to improve patient classification for those who may be underestimated or overestimated bec ause of BMI misclassification, and this study conducted additional validation test within normal BMI and overweight population s The findings are presented in four sections including the results of the descriptive analysis for the study sample and the res ults of the three aims. Descriptive Analyses Total sample sizes were 5,483 from the MESA and 5,595 (representing 55,642,860 US adults) from the NHANES. The MESA sample was randomly divided into two groups of equal sizes Group 1 (n=2,741) was used to fo r creat e the risk score and group 2 (n=2,742) was used for a validation test. Table 4 1 summarize s the comparison of risk factors profile between group 1 and group 2. Both groups showed similar prevalence of CVD (group 1 = 6.8% vs. group 2 = 6.6%, p= 0 .78). Group 1 was more likely to have men, current smokers, individuals who took hypertension medication and had diabetes ; these proportions were not significantly different from group 2. The means of age, systolic blood pressure and HDL C were slightly higher in group 1,


51 whereas total cholesterol was lower in group 2 All risk factors except total cholesterol were not significantly different between group 1 and group 2 Table 4 2 present s the baseline characteristics of risk factors of group 1. Of group 1, 6.8 % (n=187) had CVD a mean age of 61.5 years (MESA), and ranged between 44 and 84 years of age Individuals with CVD were significantly older than individuals without CVD (p<.01). Gender was equally distributed w ith men more likely to have CVD than women. The majority of the study population w ere never smoker s (87.5%) and out of the current smokers, 9.7% had CVD. Less than 10% of group 1 had systolic blood pressure over 160mmHg and they showed the highest likelihood of having CVD. Total cholesterol was not statistically significant in different levels of CVD (p=.76). Individuals in group 1 who had diabetes were twice more likely to have CVD (p<.01). Table 4 3 illustrate s the baseline prevalence of CVD in each body composition assessment. The mean BMI was 28.26kg/m 2 (SD=5.37) and t he majority was overweight (39.4%). Prevalence of CVD was highest in those who were obese (7.3%) ; however, it was not statistically significant in BMI (p=.53). According to tertile and quartile cutoffs, highest ranks showed the highest prevalence of CVD but they were not statistically significant (tertile p=.44; quartile p=.16). In WC, the mean WC was 97.74cm (SD=14.19). More than half had abdominal obesity according to clin ical cutoffs and regardless of any cutoffs, individuals with higher WC showed higher likelihood of having CVD. In WHR and WtoHR, the means were 0.92 (SD=0.08) and 0.59 (SD=0.09) respectively. T he majority was obes e according to both assessments. Results s howed


52 that the higher WHR or WtoHR were the y were more likely to have CVD. P revalence of CVD was significantly different in all levels of both assessments. Integrated assessments with BMI allowed estimating more accurate ly the prevalence of obesity and C VD in normal weight and overweight population s Combined BMI and WC showed 20.1% of the sample was classified as overweight while not having abdominal obesity. Particularly, of the overweight population, the proportion of abdominal obesity was slightly low er than normal WC (19.3%). In the normal weight population, 3.8% had abdominal obesity and showed higher CVD prevalence than overweight population s with normal WC (11.5% vs. 6.3%). However this combined measurement was not statistically significant in CVD (p=.11). In contrast, combined BMI and WHR and combined BMI and WtoHR showed higher prevalence of normal weight obesity. Of normal the weight population,16.6% w ere obese based on combined BMI and WHR and 17.2% were obese according to combined BMI and Wt oHR. These proportions were significantly higher than those for the normal weight population with normal range of WHR (p<.01) or WtoHR (p<.01). Normal weight obesity population s showed four times greater likelihood of having CVD than overweight population s having normal range of WHR or WtoHR. Development of Risk Score Aim 1. Select an O ptimal Body Composition A ssessment Seven body composition assessments were considered as a candidate for CVD risk factor. Body composition assessments involved in the study are described in detail in Table 3 2 and Table 3 3. To determine an optimal body composition assessment, the proportionality assumption test between the body composition assessments and the time to the first CVD event and multicollinearity test between typ ical risk factors and the


53 body composition assessments were performed. As can be seen in Table 4 4, none of the body composition assessments were statistically significant with the time to the first CVD event and this indicates that these assessments were assumed to be independent of time. Therefore, all body composition assessments met the proportionality assumption according to the Schoenfeld residuals test. In the multicollinearity tests, variance inflation factors (VIFs) of all body composition assessme nts were greater than 1, and the quartiles of WHR was the only statistically significant variable (p=.01). This means that the quartiles of WHR was not correlated with the other typical risk factors. Therefore, the quartiles of WHR were selected as the opt imal body composition assessment to predict CVD risk in adults. The h igher WHR was the higher the r isk of having CVD With this, individuals who fell in the 4th quartile of WHR were 1.86 times as likely to have CVD. With the selected body composition asse ssment, the final model was specified The final model was compared with the reference model which included only eight typical CVD risk factors. Table 4 5 describes the hazard ratios of each risk factor and c statistics for discrimination and Hosmer Lemesh ow chi square test for calibration and goodness of fit Overall, hazard ratios of typical risk factors were similar in both models. The final model showed slightly greater c statistics and the result of Hosmer Lemeshow test was not statistically significan t (p=0.45). Thus we failed to reject the null hypothesis that predicted CVD events were not significantly different from the observed events. Also there was no significant lack of fit for the final model. That is, the final model could predict the CVD risk successfully. The ROC curves were plotted in Figure B 1 T he AUC of the reference model was 0. 757 and the AUC of the final model with the quartiles of


54 WHR was 0.763 However as a result of the chi square test, the final model was not significantly different from the reference model (p=. 13 ). Aim 2. Develop the F inal R isk S core The final risk score was specified with the quartiles of WHR. N umeric scores were determined based on hazard ratios. As can be seen in T able 4 6, the total score is 24 points and minimum score is zero. If HR of a level of variables was not statistically significant, that level was scored as zero. For instance, all levels of HDL C were not significant and all levels were scored zero. For the quartile of WHR, a score of 4 th quartile was 2. That is, men who have greater than 1.00 or women who have greater than 0.96 had 2 points. Aim 3. Test a V alidation of the D eveloped M odel and I mprovement of P redictive A ccuracy in N ormal W e ight and O verweight P opulation s Prevalence of BMI misclassification in US adults was shown in T able 4 7. The u nweighted sample size was 5,595 representing 55,642,860 US adults aged 44 and older. Of no rmal weight population s 65.6% of US adults who had excessive %BF were misclassified as healthy. Moreover, a lthough 8.4% of the overweight population had normal %BF, they were considered to be unhealthy. W omen who were normal weight (70.4%) were likely to have excessive %BF and the majority of men who were overweight showed normal range of %BF (96.2%). Thus, a substantial proportion of US adults particularly normal weight population s, was misclassified according to the BMI. To assess the performance of the developed risk score, two validation tests were conducted. First, t he risk score was validated in remaining random split of the data, group 2 (n=2,742). Since group 2 presented similar characteristics as group 1 (See Table 4 1) it represented an appropriate study sample for a validation test. T he final


55 model was validated as moderately good performance (AUC, 0.746) and its calibration was not significant (p=.24) (See T able 4 8 and Figure C 1 ). We failed to reject the null h ypothesis that states there is no significant difference between observed events and predicted events of the outcome. T he developed risk score was tested in normal weight and overweight population s only. Results showed better goodness of fit in normal we ight and overweight populations (AUC, 0.771) as shown in Table 4 8 and Figure D 1 The Hos m e r Lemeshow test also failed to reject the null hypothesis that indicated there is no significant differen ce between observed and predicted events of the outcome in normal weight and overweight population s In conclusion, the developed risk score performed better in normal weight and overweight population s


56 Table 4 1. Comparison of Characteristics between Grou p 1 and Group 2 Factor Group 1 Group 2 p value Sample size 2741 2742 CVD prevalence 6.8 6.6 .78 Age (years) 61.5 61.4 .87 Male (%) 50.3 49.7 .70 Current smoking (%) 50.6 49.4 .74 Systolic Blood Pressure (mmHg) 133.6 133.3 .61 Hypertension Medication (%) 51.5 48.5 .09 Total Cholesterol (ml/dL) 193.5 194.9 .17 HDL Cholesterol (ml/dL) 51.2 50.7 .23 Diabetes Mellitus (%) 51.9 48.1 .31


57 Table 4 2 Baseline Profile of CVD Risk Factors of Study Population from Group 1 of the Multi Ethnic Study of Atherosclerosis (MESA) (n= 2,741 ) Predictors Total Population Cardiovascular Disease Event (%) p value Yes No Sample size 2741 187 2554 <.01 Age (Mean, year old)* 61.5 67.1 61.1 <.01 Gender* Male 47.8 9.4 90.6 <.01 Female 52.2 4.5 95.5 Smoking status* Never smoker 87.5 6.4 93.6 .03 Current smoker 12.5 9.7 90.4 Systolic blood pressure (mmHg)* <130 45.5 4.3 95.8 <.01 130 139 21.7 5.7 94.3 140 159 24.1 10.3 89.7 8.8 13.3 86.7 Hypertension Medication* No 62.9 5.0 95.0 <.01 Yes 37.1 9.9 90.1 Total Cholesterol (mg/dL) <200 60.6 6.8 93.3 .76 200 239 30.2 6.6 93.4 9.2 7.9 92.1 High Density Lipoprotein Cholesterol (HDL C)* (mg/dL) <35 21.2 9.5 90.5 <.01 35 59 55.0 6.8 93.2 23.8 4.4 95.6 Diabetes* No 88.0 6.0 94.0 <.01 Yes 12.0 12.8 87.2 Statistically significant at .05


58 Table 4 3 Baseline Characteristics of Body Composition of Group 1 from the Multi Ethnic Study of Atherosclerosis (MESA) (n= 2.741 ) Variable Total Population Cardiovascular Disease Event (%) p value Yes No BMI Clinical Cutoffs Underweight 0.5 0.0 100.0 .53 Normal 29.1 6.0 94.0 Overweight 39.4 7.1 92.9 Obese 31.0 7.3 92.7 Tertile 1 st 33.4 6.1 93.9 .44 2 nd 34.6 6.8 93.3 3 rd 32.0 7.6 92.4 Quartile 1 st 25.8 5.1 94.9 .16 2 nd 24.8 6.8 93.3 3 rd 25.5 7.6 92.4 4 th 23.9 7.9 92.1 Waist Circumference Clinical Cutoffs* Normal 47.4 5.8 94.2 .04 Ab dominal Obesity 52.6 7.8 92.2 Tertile* 1 st 33.4 4.1 96.0 <.01 2 nd 34.3 8.1 91.9 3 rd 32.4 8.3 91.7 Quartile* 1 st 25.8 3.8 96.2 <.01 2 nd 24.8 6.8 93.2 3 rd 25.2 8.5 91.5 4 th 24.2 8.3 91.7 Waist to Hip Ratio (WHR) Clinical Cutoffs* Normal 23.8 3.1 96.9 <.01 Obesity 76.2 8.0 92.0 Tertile* 1 st 33.2 4.5 95.5 <.01 2 nd 34.3 6.2 93.8 3 rd 32.5 9.9 90.1 Quartile* 1 st 26.2 3.8 96.2 <.01 2 nd 25.9 5.9 94.1 3 rd 24.0 7.2 92.9 4 th 24.0 10.8 89.2 Waist to Height Ratio (WtoHR) Clinical Cutoffs* Normal 14.1 2.6 97.4 <.01 Obes e 85.9 7.5 92.5 Tertile* 1 st 33.4 4.3 95.7 <.01 2 nd 33.4 7.3 92.7 3 rd 33.3 8.9 91.1 Quartile* 1 st 26.9 3.5 96.5 <.01 2 nd 25.8 6.9 93.1


59 Table 4 3. Continued Variable Total Population Cardiovascular Disease Event (%) p value Yes No Waist to Height Ratio (WtoHR) Quartile 3 rd 23.1 8.2 91.8 4 th 24.2 9.1 90.9 BMI & WC Underweight 0.5 0.0 100.0 11 Normal & normal 25.3 5.2 94.8 Normal & obese 3.8 11.5 88.5 Overweight & normal 20.1 6.3 93.7 Overweight & obese 19.3 8.0 92.1 Obese 31.0 7.3 92.7 BMI & WHR Underweight 0.5 0.0 100.0 <.01 Normal & normal 12.5 2.9 97.1 Normal & obese 16.6 8.3 91.7 Overweight & normal 7.8 2.8 97.2 Overweight & obese 31.6 8.2 91.8 Obese 31.0 7.3 92.7 BMI & WtoHR Underweight 0.5 0.0 100.0 .02 Normal & normal 11.9 2.8 97.3 Normal & obese 17.2 8.3 91.7 Overweight & normal 1.7 2.2 97.8 Overweight & obese 37.7 7.4 92.7 Obese 31.0 7.3 92.7 Statistically significant at .05


60 Table 4 4 Results of Proportionality Assumption Test with Body Composition Assessments and the Time to the First CVD Event in Group 1 Using the Multi Ethnic Study of Atherosclerosis (n=2,741) Proportionality VIF Body Composition Assessments p value VIF p value BMI Clinical Cutoffs 0. 52 1.15 0. 62 Tertile 0. 52 1.13 0. 77 Quartile 0. 53 1.15 0. 51 WC Clinical Cutoffs 0. 51 1.21 0.07 Tertile 0. 53 1. 12 0. 16 Quartile 0. 54 1.1 3 0. 20 WHipR Clinical Cutoffs 0. 53 1. 12 0. 28 Tertile 0. 52 1. 13 0. 12 Quartile 0. 53 1. 14 0. 01 WHeightR Clinical Cutoffs 0. 53 1. 09 0. 33 Tertile 0. 53 1. 13 0. 21 Quartile 0. 54 1. 15 0. 11 BMI&WC 0. 51 1.16 0. 92 BMI&WHipR 0. 52 1.1 7 0. 84 BMI&WHeightR 0. 5 3 1.1 6 0.78 Statistically significant at .05


61 Table 4 5 Results of Comparison of CVD Risk Models Using Cox Proportional Hazard Regressions Predictor A (Reference model) B (Quartile of WHR) HR 95% CI HR 95% CI Age (years) <50 1.00 Referent 1.00 Referent 50 54 3.31 0.94 11.63 3.20 0.91 11.26 55 59 7.09 2.15 23.40 6.72 2.03 22.19 60 64 5.97 1.76 20.20 5.45 1.61 18.49 65 69 11.59 3.57 37.58 10.70 3.29 34.77 >=70 12.41 3.85 39.97 11.25 3.48 36.31 Sex Male 1.97 1.42 2.73 2.07 1.49 2.87 Total Cholesterol (mg/dL) <200 1.00 Referent 1.00 Referent 200 239 1.34 0.97 1.87 1.31 0.94 1.83 >=240 1.75 1.08 2.86 1.74 1.07 2.83 HDL C (mg/dL) <40 1.27 0.90 1.78 1.79 0.84 1.67 40 59 1.00 Referent 1.00 Referent >=60 0.73 0.48 1.12 0.77 0.50 1.19 Systolic Blood Pressure (mmHg) <130 1.00 Referent 1.00 Referent 130 139 0.91 0.58 1.41 0.88 0.57 1.36 140 159 1.58 1.09 2.30 1.54 1.06 2.24 >=160 2.04 1.29 3.24 1.96 1.23 3.10 Anti Hypertension Medication Yes 1.39 1.03 1.89 1.37 1.01 1.87 Diabetes Yes 1.70 1.19 2.43 1.56 1.09 2.24


62 Table 4 5 Continued Predictor A (Reference model) B (Quartile of WHR) HR 95% CI HR 95% CI Smoking Current Smoker 2.12 1.44 3.12 2.09 1.42 3.08 WHR quartile 1 st 1.00 Referent 2 nd 1.21 0.74 1.98 3 rd 1.33 0.82 2.15 4 th 1.82 1.15 2.89 c statistics 0.757 0.763 Hosmer Lemeshow Chi square 0.07 0.45 Chi square Referent 0.13 Statistically significant at .05 ** Statistically significant at .01


63 Table 4 6 Scoring on the Final Risk Model Factor HR 95% CI Points Age (years) 45 49 1.00 0 50 54 3.20 0.91 11.26 0 55 59 6.72 2.03 22.19 7 60 64 5.45 1.61 18.49 5 65 69 10.70 3.29 34.77 11 70 74 11.25 3.48 36.31 11 Gender Female 1.00 0 Male 2.07 1.49 2.87 2 Systolic Blood Pressure <130 1.00 0 130 139 0.88 0.57 1.36 0 140 159 1.54 1.06 2.24 2 >=160 1.96 1.23 3.10 2 Hypertension Medication No 1.00 0 Yes 1.37 1.01 1.87 1 Total Cholesterol <200 1.00 0 200 239 1.31 0.94 1.83 0 >=240 1.74 1.07 2.83 2 HDL C <40 1.79 0.84 1.67 0 40 59 1.00 0 >=60 0.77 0.50 1.19 0 Diabetes No 1.00 0 Yes 1.56 1.09 2.24 2 Smoking Status No 1.00 0 Yes 2.09 1.42 3.08 2 Waist to Hip Ratio (WHR) 1 st 1.00 0 2 nd 1.21 0.74 1.98 0 3 rd 1.33 0.82 2.15 0 4 th 1.82 1.15 2.89 2


64 Table 4 7 Summary of Body Mass Index Misclassification in US Adults aged over 44 and Older Using National Health and Nutrition Examination Survey (NHANES), 1999 2006 (unweighted n= 5,595 weighted n= 55,642,860 ) Prevalence of misclassification BMI Normal weight Overweight % Body fat Normal 34.4 8.4 Obese 65.6 91.6


65 Table 4 8 Summary of a Validation Test in Group 2 and Subpopulation of Normal weight and Overweight population only in Group 2 Using the Multi Ethnic Study of Atherosclerosis (MESA) Group 2 (n=2,742) Normal weight and overweight population (n=1,815) Discrimination c statistic 0.746 0.771 Calibration Hosmer Lemeshow c hi square statistics 0.24 0.76


66 CHAPTER 5 DISCUSSION Discussion This study developed the CVD risk score by adding the quartiles of WHR. We found that the quartile of WHR was a significant and the optimal predictor of CVD risk estimation. Adding the quartiles of WHR resulted in better performance while the improvement was not statistically significant from the reference model which excluded the quartile of WHR. The developed risk score was validated in group 2 with moderately good performance. It especially showed greater improvement of performance particularly among no rmal weight and overweight population s. This study is valuable in the sense of improvement of performance in normal weight and overweight population s who are usually misclassified by the use of BMI only This study provide d additional clinical value to th e typical CVD risk score by integrating the o ptimal body composition assessment. Furthermore, it may suggest an alternative assessment to define obesity when considering CVD in clinical setting s Body Composition Assessment The developed risk score selec ted WHR as the optimal body composition assessment in CVD risk assessment. Its selection was validated with the preliminary study that WHR played a significant role in predicting CVD and mortality (Cameron et al., 2012) Moreover, to date, of 363 existing CVD risk scores, 29% selected BMI as a predictor. To our knowledge, this is the first study that equipped WHR to the risk score and resulted in better performance to predict CVD risk Adding WHR is innovative approach to account for current obesity trend. In 2012, more than half of US adults had abdominal obesity and its prevalence has significantly increased since 1999 (Ford,


67 Maynard, & Li, 2014) Abdominal obesity was strongly associated with CVD risk rather than total body fat (Myint, Kwok, Luben, Wareha m, & Khaw, 2014) The risk of abdominal obesity resulted in twice the risk of mortality in normal weight populations (Karine R Sahakyan et al., 2015) Therefore, a proxy considering abdominal obesity became a critical role in CVD risk estimation. In additi on, hip circumference also played a role as an independent factor in CVD risk and mortality (Heitmann, Frederiksen, & Lissner, 2004; Lissner, Bjrkelund, Heitmann, Seidell, & Bengtsson, 2001) Epidemiological studies have shown that hip circumference was a ssociated with CVD incidents and mortality and it was correlated with abdominal obesity (Cameron et al., 2012; Heitmann et al., 2004) To account for the association of hip circumference with CVD, the WHR may account for body fat accumulation over the body compared to waist circumference only. In addition, s ince only 33.7% of normal weight population had abdominal obesity, abdominal obesity may not be commonly observed in normal weight population (Mainous et al. 2016). With this reason, the use of waist circumference may miss normal weight individual with excessive body fat again. However Ramsaran and Maharaj showed that WHR was associated with normal weight obesity population (Ramsaran and Maharaj, 201 7 ). There fore, the WHR may capture body fat accumulation over the body with improved accuracy particularly in normal weight population and it may contribute to enhancing CVD risk estimation. Norm al weight obesity defined by %BF (e.g., 25% of men and 32% of women) i ntegrated with BMI (18.5 24.9kg/m 2 ) in middle aged adults may be associated with sarcopenia in older adults Sarcopenia represents age associated loss of muscle mass (Stenholm, Harris, Rantanen, Visser, Kritchevsky & Ferrucci, 2009). It may cause


68 physical function impairment and poor health states in older adults (Stenholm et al., 2009). If adults in middle age who are currently normal weight obese maintain excessive body fat during their lifetime without exercise, they will be likely to have higher rates o f abnormal CVD risk and mortality (Batsis et al., 2013). Such individuals, particularly because they have normal range of BMI, are likely to miss an opportunity to receive preventive care on time. Thus, their CVD risk may go untreated and place them at gre ater risk of adverse CVD outcomes. Furthermore, CVD risk factors including hypertension and diabetes were associated with sarcopenia ( Chin et al., 2013; Han, 2017). Schrager and colleagues articulated that inflammatory markers such as CRP, IL 18 were assoc iated with excessive body fat as well as muscle strength (Schrager et al., 2007). These inflammatory markers may elevate CVD risk factors and individuals with sarcopenic obesity may be more likely to present CVD. Moreover, more than 20% of US older adults had sarcopenia, and its prevalence is projected to increase (Batsis et al., 2013). Therefore, normal weight obesity as a potential factor of sarcopenia should be paid attention to prevent CVD in older adults. Creating risk score is a complex process of con sideration with clinical significant and statistical significance. The developed risk score incorporated categorical forms of predictors. Categorical form is easy and simple way to classify populations at risk on the basis of clinical guidelines. Clinician s use clinical cutoffs of a categorical factor to inform patients to achieve for preventing chronic disease risk. However, statistics wise, categorical form of a predictor is not frequently used because of the principle of statistical significance. For ins tance, the initial FRS attempted to include BMI as a categorical form and simultaneously, BMI was excluded due to no statistical significance


69 (Wilson et al., 1998) The updated risk score, however, added BMI as a significant predictor to the risk model and BMI was incorporated as a continuous form (Wilson et al., 2008) QRISK, which is widely used in the UK also adopted BMI by including interaction term of BMI and age whereas the ASCVD risk score excluded BMI due to non significance (Goff et al., 2014; Hipp isley Cox et al., 2007) This indicates that the form of a predictor played a critical role in determining risk factor in risk assessment. Although statistical significance should be considered when meeting the assumption of regression analysis, it calls i nto question how to deal with a gap between a statistical significance and a clinical significance in selecting a risk score in risk score. Limitation In Terms of a Body Composition Assessment This study has some limitations in regards to body composition assessments. First, the current study selected quartiles of WHR as a threshold instead of most adopted cutoffs. Some epidemiological studies found that quartiles of WHR showed more accurate pow er to predict CVD and mortality compared to BMI (Myint et al., 2014; Sokol et al., 2016) This approach, however, may raise concern about inconsistent cutoffs in different populations. Lear and colleagues argue that cutoffs of WHR associated with CVD risk factors were different in racial and ethnic groups (Lear, James, Ko, & Kumanyika, 2010) More specifically, Asians who were likely to have higher visceral fat and higher risk of CVD showed lower cutoffs of WHR than European and Caucasians (Lear et al., 201 0) Also recent epidemiological studies reported significant association of a high risk for chronic diseases with lower BMI in Asians and suggest race specific BMI cutoffs based on an association of chronic disease risk (Raji et al., 2001) To take into ac count race specific physiological characteristics, the current study used a longitudinal dataset


70 composed of diverse ethnic participants. This reduced bias when determining cutoffs of quartiles. Further efforts would be needed to establish reliable clinica l cutoffs of WHR with diverse population based datasets. Second, this study selected non invasive and conventional body composition assessments as a proxy for obesity; they are also unable to directly measure the amount of body fat. Using high technic equipment such as BIA or DXA allows for the most accurate bod y fat mass measurements. These techniques have not been adopted for use in clinical settings because of the high costs for the sole purpose of screening and absence of health insurance coverage by most health insurers. This study showed a substantial propo rtion of normal weight population s were classified as healthy on the basis of %BF. WHR may miss some population s with true excessive body fat. In future study, %BF may play a critical role to capture true excessive body fat over the body and contribute to identifying obese population who has true excessive body fat By adding %BF to the risk score, the score will estimate more accurate risk score for population who are normal weight obesity population or overweight population who are usually misclassified b y non invasive measurement. Thus, a cost effective body composition assessment which is able to measure accurate body fat mass is needed. Appropriate health care reimbursement policies are also needed. Future research should apply more accurate body compos ition assessment s to predict CVD risk. Risk Factors This study used eight typical CVD risk factors that have been used in most risk scores. To date, in addition to the WHR, there have been several attempts to investigate the effect of variety body composit ion types on CVD. A recent exploratory study revealed that neck circumference was positively associated with CVD risk factors


71 (Preis et al., 2010) Neck circumference represented upper body subcutaneous fat which is associated with insulin resistance and h igh LDL C (Koutsari, Snozek, & Jensen, 2008) While it may be a novel marker of CVD, weak correlations of less than .50 with total cholesterol, LDL C and glucose should be considered to be a significant CVD risk factor (Ben Noun & Laor, 2003) Other studie s focused on sagittal abdominal diameter (SAD) when considering CVD risk (Kahn et al., 2014; Pouliot et al., 1994; Sampaio et al., 2007) SAD represented a simple non invasive measurement of visceral fat located around the organs (Zamboni et al., 1998) Vi sceral fat has been paid attention as a critical cause of systemic inflammation and was significantly associated with CVD (Fontana et al., 2007; Mahabadi et al., 2009) Individuals with high visceral fat showed 1.83 times higher odds of stroke (Mahabadi et al., 2009) Thus, in order to reflect the risk of visceral fat, future risk score can be developed by adding SAD. As mentioned above, a direct body fat measurement would be the best way to assess CVD with improved accuracy. While %BF is a representative s urrogate of body fat mass, it requires a technician and high technique machines such as MRI and CT. Future accessible and cost effectiveness measurement of body fat mass may be a key in measuring CVD risk in the future. Some physiological biomarkers have been added to the CVD risk score. Cardiorespiratory Fitness (CRF) is defined as the ability of the circulatory and respiratory system to supply oxygen as energy source during physical activity (Caspersen, Powell, & Christenson, 1985) It has been discussed as a key factor of (Blair et al., 1995; Hainer, Toplak, & Stich, 2009) Regardless of obesity, improved CRF was associated with reduced


72 mortality rate about 44% in men (Blair et al., 1995) With this advantage, Gander and colleagues recently proposed updating the FRS by adding CRF (J. C. Gander et al., 2017) In addition, Greenland and colleagues added coronary artery calcium score (CACS) to the FRS (Greenland et al., 2004) CACS is the measurement of cumulative calcium on the artery that causes ASCVD and higher CACS was associated with twice higher risk of ASCVD than (Keelan et al., 2001) It is noted that these risk scores contributed to emphasizing the importance of physiological markers in predictin g CVD risk. Lifestyle factors were also considerable risk factors of CVD. The fact that lifestyle factors were modifiable prior to the CVD occurrence is a n advantage in prevention for asymptomatic population s P hysical activity has been an increased focus because of the increase in sedentary lifestyles. According to Mainous and colleague, individuals with abnormal blood glucose were less likely to perform exercise despite normal weight (Mainous et al., 2017) To take into account the significance of lifest yle factors, Chiuve and colleague created a lifestyle based prediction model and incorporated CVD risk score to examine the association between lifestyle factors and CVD risk factors (Chiuve et al., 2014) T hey considered dietary factors and seven behavior al factors ( i.e., exercise, sleep duration, sedentary lifestyle, smoking, alcohol, BMI and WC) and found that unhealthy lifestyle s were significantly associated with elevated risk of CVD (Chiuve et al., 2014; Gooding et al., 2017) Levesque and colleagues also created a simpler lifestyle risk score and showed that a high risk lifestyle score significantly increased elevated CVD risk factors (Levesque, Poirier, Despres, & Almeras, 2017) H ealthy lifestyle affects body compositions and it may be a key in obesity prevention Therefore,


73 on the basis of current physical activity and nutrition guideline s it would be worth considering that healthy lifestyle factors can be integrated to the CVD risk score. Limitation In Terms of Risk Factors This study has a limitation in terms of selecting risk factors. T he developed risk score is the use of typical CVD risk factors without considering new biomarkers and lifestyle factors. A s mentioned above, modifiable behavioral risk factors have been shown to prevent CVD and were effective in predict ing elevated CVD risk (Chiuve et al., 2014; Gooding et al., 2017) O ther research has taken into account socioeconomic status such as poverty as a risk factor and has provided insights for prevention strateg ies using a societ al approach (Hippisley Cox et al., 2007) R ecent stud y has attempted to add genetic factors to the risk score (Knowles et al., 2017) It is important to consider the diverse dimensions of CVD prevention as a key role in improving predictive accuracy. Risk Score This study developed a new CVD risk score by adding a body composition assessment to improve decision making for clinicians. Previous studies have investigated socioeconomic characteristics and physiological features of high CVD risk populations by using CVD risk score (Gulliford et al., 2017; Patel, Taksler, Hu, & Rothberg, 2017; Robinson, Jackson, Wells, Kerr, & Marshall, 2017) Robinson and colleagues found that many primary care physicians used estimated CVD risk score when making a decision (Robinson et al., 2017) Use of CVD risk score in primary care setting played a significant role in identifying high risk populations who were unaware of their CVD risk because of missed opportunities to undergo health check (Gulliford et al., 2017) This showcased the value of the developed risk score in identifying populations with elevated CVD risk who did not get CVD risk estimates at primary care settings


74 because of their healthy body weight ranges. Moreover, by using the non invasive body composition assessment, t he use of the improved risk score may provide an opportunity to check the potential for CVD risk in a primary care clinical setting. There might be some challenges to use the risk score in primary care setting. Unless the risk score is not req uired to use in primary care setting, patients may not estimate their risk without According to Gulliford and colleagues, given an opportunity to estimate their CVD risk by a primary care physician, most patients at high risk ended up recognizing their CVD risk ( Gulliford et al., 2017). Moreover, it is unknown that primary care physicians would widely utilize body composition assessments in addition to BMI in primary care setting Especially, while the WHR was adopted to the current risk score, it may not be universally adopted in primary care setting. T o date, limited studies have investigated the use of CVD risk score and the use of body composition assessments when providing treatment at primary care settings. Therefore, additional studies should investigate the use of risk score in health care settings and evaluate the impact of risk score on CVD prevention care. The developed risk score showed moderately good ability to predict CVD compared to the other risk scores For instance, AU C of the original FRS was .75 and AUC of the updated risk score with BMI developed by Wilson and colleagues was .81 (Wilson et al., 1998; 2008). The lowest AUC was the risk score with CACS developed by Greendland and colleagues and it was .68 (Greenland et al., 2004). In addition when the CRF was added to the FRS, the AUC was .80 (Gander et al., 2017). AUC of the most recent risk score developed by the ACA/AHA was .71 (Muntner et al., 2014).


75 Therefore, b ased on these results of goodness of fit, the devel oped risk score with the quartile of WHR was proved to perform moderately good. The developed CVD risk score has common points to metabolic syndrome. Metabolic syndrome is defined as presence of at least three CVD risk factors including insulin resistance/hyperglycemia, abdominal obesity, hypertension and dyslipidemia (American Heart Association, 2016 b ). To diagnose metabolic syndrome, the WHO, the National Cholesterol Education Program (NCEP) Adult Treatment Panel (ATP) III and International Dia betes Federation (IDF) released comprehensive criteria (Huang, 2009). Mostly, they provided clinical thresholds of each CVD risk factors (e.g., waist circumference, triglcerides, HDL C, blood pressure and fasting glucose) and counted a number of occurrence s of these metabolic disorders. When people who had more than three metabolic disorders, they are diagnosed to metabolic syndrome (Huang, 2009). There are some similarities between metabolic syndrome criteria and the CVD risk score used in this study Firs t, the purpose of both methods is to classify patients who are at low/high risk of chronic disease Second, both utilize an adiposity measurement ; the criteria of metabolic syndrome adopted waist circumference and the current developed risk score used the WHR. They considered the impact of excessive body fat accumulation on CVD risk factors On the other hand, there are some differences. While they both used a body composition assessment, metabolic syndrome focuses on abdominal obesity instead of body fat a ccumulation over the body. The current developed risk score used the WHR and it may consider total body fat accumulation. Thus, it will contribute classifying misclassified population particularly, normal weight population. Next difference is risk estimati on. The CVD risk score estimates risk by


76 weighing a level of each risk factor while metabo lic syndrome simply counts number s of risk factors. The risk score provides more accurate potential risk estimates and healthcare providers may suggest effective trea tment or preventive care services. Lastly, the outcomes are different. Metabolic syndrome diagnoses metabolic disorders that significantly link to CVD. It does not provide how much potential risk of CVD people have. The risk score, however, direc tly provid es CVD risk estimates. I t helps not only to classify population at low, intermediate and high, but also to provide objective evidence to initiate appropriate preventive care service. The developed risk score may incorporate clinical guidelines for preven tive care. Current clinical guidelines use risk score as a criteria when recommending preventive care services. The most recent ACC and AHA guideline suggest a clinical threshold of 7.5% of 10 year CVD risk for initiation of statin treatment (Goff et al., 2014) The USPSTF recommends a routine lipid screening and CVD risk assessment for individuals with an estimated 10 year CVD risk of 10% or greater (USPSTF, 2016) These efforts may encourage the use of CVD risk score in clinical settings and may result in clinicians using this tool when delivering care. However, both guidelines used the ASCVD risk score that did not include a body composition assessment and may mi ss some populations with elevated CVD risk. Since these guidelines focus on individuals at high risk of typical CVD risk factors, some people who are at intermediate risk may miss an opportunity to assess their CVD risk. Consequently, they may miss an oppo rtunity to receive preventive care services in advance. A consequence of missing those opportunity can elevate CVD incidents. Additionally, as presented above, the clinical thresholds were not consistent between the two guidelines. More studies are


77 needed to provide sufficient evidence to establish universal cutoffs by using the updated risk score. Limitation In Terms of Risk Score Performance of the developed risk score was not statistically significant with the reference model which excluded the quartile s of WHR whereas the predictive accuracy was slightly improved. It was corresponding to the initial FRS that attempted to add categorical form of BMI to the risk score (Wilson et al., 1998) Despite multiple evidences of significant association between BMI and CVD, BMI did not play a significant role when added to the risk score. Likewise, although some epidemiological studies have shown that WHR was significantly associated with CVD and CVD risk factors (De Koning et al., 2007; Myint et al., 2014) adding the WHR did not improve the predictive accuracy significantly. To account for statistical significance, further study may examine invisible relationship between the WHR and CVD risk factors and reflect this relationship when updating the risk score. The d eveloped risk score focused on middle age of adults over the age of 44 who were the target population for primary CVD preventive care services (USPSTF, 2016) Middle age adults showed elevated risk of chronic disease and CVD (Egan et al., 2010; Menke et al ., 2015) Particularly, they had the highest prevalence of obesity in the US (Flegal et al., 2016) As a result, the majority of risk scores including the current risk score were created by focusing on this population However, s ome studies have argued th e optimal timing of estimating CVD risk is for the purpose of prevention (Gooding et al., 2017; Liu et al., 2012; Shah et al., 2016) As previously stated, even though this middle age group has been focused on as a high risk group, CVD risk estimation in m iddle age may be too late in preventing CVD risk. The p revalence of obesity in US adolescen ts


78 has jumped to 20.6% and its prevalence in young adults between the ages of 20 and 39 has reached 34.3% in 2014 (Flegal et al., 2016; Ogden et al., 2016) These yo ung er populations are a potential high risk group for premature CVD risk. Thus, it is recommended that obesity prevention strategy for young adults to combat increase CVD prevalence. Several studies suggest healthy lifestyle intervention s through young adulthood (Gooding et al., 2017; Liu et al., 2012; Shah et al., 2016) Higher fitness levels and maintaining healthy lifestyle s in young adulthood have been found to be strongly associated with lower risk of CVD (Liu et al., 2012; Shah et al. 2016) Furthermore, the most recent USPSTF guideline include that primary care physicians can offer behavioral counseling to individuals who are not overweight or obese and who do not have any CVD risk factors (e.g., diabetes, hypertension and dyslipidem ia) for CVD prevention treatment (Grossman et al., 2017; U.S. Preventive Services Task Force 2012) While this population has a small net benefit from behavioral counseling, it is a valuable opportunity to modify current lifestyle habits in order to maintain a healthy lifestyle and p revent CVD risk (Grossman et al., 2017) Further risk score should also reflect these changes and need to be updated for young adults. Additional stu dies should investigate an optimal timing to estimate CVD risk score for the purpose of prevention. Next an external validation test of the model was not performed The current study performed an internal validation test using a random split of the dataset. While it prove s an accurate predictive accuracy of the model in different study populations, it still lacks generalizability in applying this risk estimation to other population groups. T he developed risk score was created from middle age adults. It may underestimate the risk


79 of CVD for young adults. Therefore, the developed risk score needs to be tested i n diverse population s including young adults. Implication Cardiovascular Disease (CVD) risk score is a pragmatic tool for both asymptomatic population and healthcare providers. For healthcare providers, i t may improve clinical decision making For asympt omatic population, it allows examine their potential CVD risk and receive preventive care services before the onset of CVD. Particularly, it can be used in primary care setting in which is a gateway of health care. Primary care physicians may provide appro priate preventive care in a timely manner and people can save costs for seeking a cardiologist. T he proposed model may contribute to classifying population s at low risk who would be intermediate high CVD risk because of BMI misclassification. This populat ion has been neglected in receiving appropriate health care or disease prevention strateg ies such as screening or timely counselling on time because they are not a target population in clinical guideline s Thus, the developed risk score may contribute in classifying population s previously neglected from the CVD risk assessment s. Also, c urrent clinical guideline may need to expand the target population for CVD screening. Lastly, focusing on the body composition assessment can contribute to updating curren t healthcare policy. In 2016, The US Equal Employment Opportunity Commission (EEOC) issued a new rule about Employer Wellness program offered by the Affordable Care Act. The wellness program offers health promotion and chronic disease preventive care servi ces to employers with subsidies. However when participants fail to meet a certain outcome such as normal weight, they will be penalized about 30% of costs of health insurance coverage The strategy of adding the WHR as a substitute for BMI may


80 contribute t o classifying population with enhanced performance and prevent unfair penalty.


81 CHAPTER 6 CONCLUSION This study developed a non invasive and affordable CVD risk score for populations aged 44 and older by adding the quartiles of WHR as the optimal body co mposition assessment. The risk score included eight classical risk factors (age, sex, diabetes, hypertension medication, systolic blood pressure, total cholesterol, HDL C and smoking) and the quartiles of WHR It wa s designed to improve predictive accuracy particularly for normal weight and overweight populations who were usually misclassified by the use of BMI only and who were neglected in receiving appropriate preventive care A s obesity becomes epidemic in public health, obesity related measurement may play a key role in CVD risk assessment. Thus, adding WHR may be advantageous for primary care physicians when assess ing CVD risk and may provide more accurate CVD risk estimations. WHR is a stronger predictor of CVD than BMI and it may account for abdomin al obesity which has also become epidemic in the US (De Koning et al., 2007; Ford et al., 2014) T h erefore the developed risk score may contribute to accounting for most recent trend of obesity in US adults and providing scientific evidence to establish efficient health promotion strategy to prevent CVD risk T h e developed risk score may also aid clinicians, specifically primary care physicians, in predicting future CVD risk with objective evidence and impro ve overall decision making.


82 APPENDIX A FLOW CHART OF STUDY POPULATION Figure A 1. Flow Chart of the Multi Ethnic Study of Atherosclerosis (MESA) with Exclusion Criteria Presenting the Final Sample Size


83 APPENDIX B AREA UNDER THE CURVE OF THE FINAL MODEL Figure B 1 Areas Under the Curve (AUC) comparing the predictive ability of the to Hip Ratio) in Group 1 (n=2,74 1)


84 APPENDIX C AREA UNDER THE CURVE OF THE VALIDATION TEST Figure C 1 Area Under the Curve (AUC) of the Validation Test in Group 2 (n=2,742 p= .24 ).


85 APPENDIX D AREA UNDER THE CURVE IN NORMAL WEIGHT AND OVERWEIGHT POPULATION Figure D 1 Area Under the C urve (AUC) of the Final Model in Normal W eight and Overweight Population in Group 2 (n=2,742 p= .76 ).


86 LIST OF REFERENCES Ackland, T. R., Lohman, T. G., Sundgot Borgen, J., Maughan, R. J., Meyer, N. L., Stewart, A. D., & Mller, W. (2012). Current status of body composition assessment in sport. Sports Medicine, 42 (3), 227 249. A merican D iabetes A ssociation (2010). Diagnosis and classification of diabetes mellitus. Diabetes care, 33 (Supplement 1), S62 S69. American Diabetes Association (2016). Standards of Diabetes Care in Diabetes 2016. Diabetes Care, 39 (supplement 1), S1 S106. Adams, K. F., Schatzkin, A., Harris, T. B., Kipnis, V., Mouw, T., Ballard Barbash, R., Leitzmann, M. F. (2006). Overweight, obesity, and mortality in a large prospective cohort of persons 50 to 71 years old. New England Journal of Medicine, 355 (8), 763 778. A merican H eart A ssociation (2014). What is Cardiovascular Disease? Retrieved from ularDi sease/What is Cardiovascular Disease_UCM_301852_Article.jsp#.WIDU8HqVTy0 American Heart Association (2016 a ). Heart disease, stroke and research statistics at a glance. Retrieved on April, 1 2016. American Heart Association. (2016 b ). About Metabolic Syndrome. Retrieved from Metabolic Syndrome_UCM_301920_Article.jsp#.WgsBKmhSw2w American Heart Association / American Stroke Association (2015). African Americans & Cardiovasuclar Diseases: Statistical Fact Sheet 2015 Update Statistical Fact Sheet American Heart Association Albert, M. A., Glynn, R. J., & Ridker, P. M. (2003). Plasma concentration of C reactive protein and the calculated Framingham Coronary Heart Dis ease Risk Score. Circulation, 108 (2), 161 165. Artac, M., Dalton, A. R., Majeed, A., Car, J., & Millett, C. (2013). Effectiveness of a national cardiovascular disease risk assessment program (NHS Health Check): results after one year. Preventive medicine, 57 (2), 129 134. Artham, S. M., Lavie, C. J., Milani, R. V., & Ventura, H. O. (2008). The obesity paradox: impact of obesity on the prevalence and prognosis of cardiovascular diseases. Postgrad Med, 120 (2), 34 41. doi: 10.3810/pgm.2008.07.178810.3810/pgm. 2009.01.1970


87 Asgari, S., Barzin, M., Hosseinpanah, F., Hadaegh, F., Azizi, F., & Khalili, D. (2016). Obesity Paradox and Recurrent Coronary Heart Disease in a Population Based Study: Tehran Lipid and Glucose Study. Int J Endocrinol Metab, 14 (2), e37018. doi: 10.5812/ijem.37018 Ashwell, M., & Gibson, S. (2016). Waist to circumference. BMJ Open, 6 (3), e010159. Aune, D., Sen, A., Nora t, T., Janszky, I., Romundstad, P., Tonstad, S., & Vatten, L. J. (2016). Body mass index, abdominal fatness and heart failure incidence and mortality: a systematic review and dose response meta analysis of prospective studies. Circulation 115.016801. Bar ba, C., Cavalli Sforza, T., Cutter, J., & Darnton Hill, I. (2004). Appropriate body mass index for Asian populations and its implications for policy and intervention strategies. The Lancet, 363 (9403), 157. Barreira, T. V., Staiano, A. E., Harrington, D. M ., Heymsfield, S. B., Smith, S. R., Bouchard, C., & Katzmarzyk, P. T. (2012). Anthropometric correlates of total body fat, abdominal adiposity, and cardiovascular disease risk factors in a biracial sample of men and women. Mayo Clin Proc, 87 (5), 452 460. d oi: 10.1016/j.mayocp.2011.12.017 Batsis, J. A., Sahakyan, K. R., Rodriguez Escudero, J. P., Bartels, S. J., Somers, V. K., & Lopez Jimenez, F. (2013). Normal weight obesity and mortality in United ealth and Nutrition Examination Survey). Am J Cardiol, 112 (10), 1592 1598. Ben Noun, L. L., & Laor, A. (2003). Relationship of neck circumference to cardiovascular risk factors. Obesity, 11 (2), 226 231. Bild, D. E., Bluemke, D. A., Burke, G. L., Detrano, R., Roux, A. V. D., Folsom, A. R., Liu, K. (2002). Multi ethnic study of atherosclerosis: objectives and design. American journal of epidemiology, 156 (9), 871 881. Black, H. R. (1992). Cardiovascular risk factors. The Yale University school of medi cine heart book 23 35. Blair, S. N., Kohl, H. W., Barlow, C. E., Paffenbarger, R. S., Gibbons, L. W., & Macera, C. A. (1995). Changes in physical fitness and all cause mortality: a prospective study of healthy and unhealthy men. JAMA, 273 (14), 1093 1098. Brindle, P., Jonathan, E., Lampe, F., Walker, M., Whincup, P., Fahey, T., & Ebrahim, S. (2003). Predictive accuracy of the Framingham coronary risk score in British men: prospective cohort study. BMJ, 327 (7426), 1267.


88 Brindle, P. M., McConnachie, A., Up ton, M. N., Hart, C. L., Smith, G. D., & Watt, G. C. (2005). The accuracy of the Framingham risk score in different socioeconomic groups: a prospective study. Br J Gen Pract, 55 (520), 838 845. Cameron, A. J., Magliano, D. J., Shaw, J. E., Zimmet, P. Z., C arstensen, B., Alberti, K. G. M., Kowlessur, S. (2012). The influence of hip circumference on the relationship between abdominal obesity and mortality. Int J Epidemiol, 41 (2), 484 494. Caspersen, C. J., Powell, K. E., & Christenson, G. M. (1985). Ph ysical activity, exercise, and physical fitness: definitions and distinctions for health related research. Public health reports, 100 (2), 126. C enters for Disease Control and Prevention. (2014). Smoking and Cardiovascular Disease: Fact Sheet. Centers for Disease Control and Prevention (2008). Smoking attributable mortality, years of potential life lost, and productivity losses -United States, 2000 2004. MMWR Morb Mortal Wkly Rep, 57 (45), 1226 1228. Chiuve, S. E., Cook, N. R., Shay, C. M., Rexrode, K. M., Albert, C. M., Manson, J. E., Rimm, E. B. (2014). Lifestyle Based Prediction Model for the Prevention of CVD: The Healthy Heart Score. J Am Heart Assoc, 3 (6), e000954. Chobanian, A. V., Bakris, G. L., Black, H. R., Cushman, W. C., Green, L. A., Izz o Jr, J. L., Wright Jr, J. T. (2003). The seventh report of the joint national committee on prevention, detection, evaluation, and treatment of high blood pressure: the JNC 7 report. JAMA, 289 (19), 2560 2571. Chin, S. O., Rhee, S. Y., Chon, S., Hwang, Y. C., Jeong, I. K., Oh, S., ... & Kim, J. W. (2013). Sarcopenia is independently associated with cardiovascular disease in older Korean adults: the Korea National Health and Nutrition Examinat ion Survey (KNHANES) from 2009. PLoS One 8 (3), e60119. Collaboration, A. B. I. (2008). Ankle brachial index combined with Framingham Risk Score to predict cardiovascular events and mortality: a meta analysis. JAMA: the journal of the American Medical Association, 300 (2), 197. Conroy, R., Pyrl, K., Fitzgera ld, A. e., Sans, S., Menotti, A., De Backer, G., Keil, U. (2003). Estimation of ten year risk of fatal cardiovascular disease in Europe: the SCORE project. European heart journal, 24 (11), 987 1003. Curtis, J. P., Selter, J. G., Wang, Y., Rathore, S. S., Jovin, I. S., Jadbabaie, F., Bader, F. (2005). The obesity paradox: body mass index and outcomes in patients with heart failure. Archives of internal medicine, 165 (1), 55 61.


89 D'Agostino Sr, R. B., Grundy, S., Sullivan, L. M., & Wilson, P. (2001 ). Validation of the Framingham coronary heart disease prediction scores: results of a multiple ethnic groups investigation. JAMA, 286 (2), 180 187. & Kannel, W. B. ( 2008). General cardiovascular risk profile for use in primary care. Circulation, 117 (6), 743 753. De Koning, L., Merchant, A. T., Pogue, J., & Anand, S. S. (2007). Waist circumference and waist to hip ratio as predictors of cardiovascular events: meta reg ression analysis of prospective studies. European heart journal, 28 (7), 850 856. Demler, O. V., Paynter, N. P., & Cook, N. R. (2015). Tests of calibration and goodness of fit in the survival setting. Statistics in medicine, 34 (10), 1659 1680. Dinarello, C. A. (2000). Proinflammatory cytokines. Chest Journal, 118 (2), 503 508. Egan, B. M., Zhao, Y., & Axon, R. N. (2010). US trends in prevalence, awareness, treatment, and control of hypertension, 1988 2008. JAMA, 303 (20), 2043 2050. illy, S. (2006). Genetics of obesity in humans. Endocrine reviews, 27 (7), 710 718. Flegal, K. M., Kruszon Moran, D., Carroll, M. D., Fryar, C. D., & Ogden, C. L. (2016). Trends in Obesity Among Adults in the United States, 2005 to 2014. JAMA, 315 (21), 228 4 2291. Fonarow, G. C., Srikanthan, P., Costanzo, M. R., Cintron, G. B., Lopatin, M., Committee, A. S. A., & Investigators. (2007). An obesity paradox in acute heart failure: Analysis of body mass index and inhospital mortality for 108927 patients in the Acute Decompensated Heart Failure National Registry. Am Heart J, 153 (1), 74 81. Fontana, L., Eagon, J. C., Trujillo, M. E., Scherer, P. E., & Klein, S. (2007). Visceral fat adipokine secretion is associated with systemic inflammation in obese humans. Diab etes, 56 (4), 1010 1013. Ford, E. S., Maynard, L. M., & Li, C. (2014). Trends in mean waist circumference and abdominal obesity among US adults, 1999 2012. JAMA, 312 (11), 1151 1153. Frank, L. D., Andresen, M. A., & Schmid, T. L. (2004). Obesity relationsh ips with community design, physical activity, and time spent in cars. American journal of preventive medicine, 27 (2), 87 96. Gander, J. (2014). Factors related to coronary heart disease risk among men: validation of the Framingham risk score. Preventing c hronic disease, 11


90 Gander, J. C., Sui, X., Hbert, J. R., Lavie, C. J., Hazlett, L. J., Cai, B., & Blair, S. N. (2017). Addition of estimated cardiorespiratory fitness to the clinical assessment of 10 year coronary heart disease risk in asymptomatic men. Preventive Medicine Reports, 7 30 37. Go, A. S., Mozaffarian, D., Roger, V. L., Benjamin, E. J., Berry, J. D., Blaha, M. J., Franco, S. (2014). Executive summary: heart disease and stroke statistics -2014 update: a report from the American Heart Association. Circulation, 129 (3), 399. Go A. S., Mozaffarian, D., Roger, V. L., Benjamin, E. J., Berry, J. D., Blaha, M. J., Turner, M. B. (2014). Heart Disease and Stroke Statistics 2014 Update. A Report From the American Heart Association, 129 (3), e28 e292. doi: 10.1161/01.cir.0000441139 .02102.80 Goff, D. C., Lloyd cardiovascular risk: a report of the American College of Cardiology/American Heart Association Task Force on Practice Guidelines. Journal of the American College of Cardiology, 63 (25_PA). Gmez Ambrosi, J., Silva, C., Galofr, J., Escalada, J., Santos, S., Milln, D., Valent, V. (2012). Body mass index classification misses subj ects with increased cardiometabolic risk factors related to elevated adiposity. International journal of obesity, 36 (2), 286 294. Gooding, H. C., Ning, H., Gillman, M. W., Shay, C., Allen, N., Goff, D. C., Chiuve, S. (2017). Application of a Lifesty le Based Tool to Estimate Premature Cardiovascular Disease Events in Young Adults: The Coronary Artery Risk Development in Young Adults (CARDIA) Study. JAMA Internal Medicine, 177 (9), 1354 1360. Gordon Larsen, P., Nelson, M. C., Page, P., & Popkin, B. M. (2006). Inequality in the built environment underlies key health disparities in physical activity and obesity. Pediatrics, 117 (2), 417 424. Greenland, P., LaBree, L., Azen, S. P., Doherty, T. M., & Detrano, R. C. (2004). Coronary artery calcium score comb ined with Framingham score for risk prediction in asymptomatic individuals. JAMA, 291 (2), 210 215. Grossman, D. C., Bibbins Domingo, K., Curry, S. J., Barry, M. J., Davidson, K. W., Doubeni, C. A., Kurth, A. E. (2017). Behavioral Counseling to Promo te a Healthful Diet and Physical Activity for Cardiovascular Disease Prevention in Adults Without Cardiovascular Risk Factors: US Preventive Services Task Force Recommendation Statement. JAMA, 318 (2), 167 174.


91 Gulliford, M. C., Khoshaba, B., McDermott, L., Cornelius, V., Ashworth, M., Fuller, F., Wright, A. J. (2017). Cardiovascular risk at health checks performed opportunistically or following an invitation letter. Cohort study. J Public Health (Oxf) 1 6 doi: 10.1093/pubmed/fdx068 Hainer, V., Toplak, H., & Stich, V. (2009). Fat or fit: what is more important? Diabetes care, 32 (suppl 2), S392 S397. Hall, J. E., da Silva, A. A., do Carmo, J. M., Dubinion, J., Hamza, S., Munusamy, S., Stec, D. E. (20 10). Obesity induced hypertension: role of sympathetic nervous system, leptin, and melanocortins. Journal of Biological Chemistry, 285 (23), 17271 17276. Han, P., Yu, H., Ma, Y., Kang, L., Fu, L., Jia, L., ... & Zhang, W. (2017). The increased risk of sarc openia in patients with cardiovascular risk factors in Suburb Dwelling older Chinese using the AWGS defi nition. Scientific Reports 7 (1), 9592. Harrell, F. E., Lee, K. L., Califf, R. M., Pryor, D. B., & Rosati, R. A. (1984). Regression modelling strategies f or improved prognostic prediction. Statistics in medicine, 3 (2), 143 152. Harrell, F. E., Lee, K. L., & Mark, D. B. (1996). Tutorial in biostatistics multivariable prognostic models: issues in developing models, evaluating assumptions and adequacy, and me asuring and reducing errors. Statistics in medicine, 15 361 387. Hastie, C. E., Padmanabhan, S., Slack, R., Pell, A. C., Oldroyd, K. G., Flapan, A. D., Dominiczak, A. F. (2010). Obesity paradox in a cohort of 4880 consecutive patients undergoing pe rcutaneous coronary intervention. European heart journal, 31 (2), 222 226. Heidenreich, P. A., Trogdon, J. G., Khavjou, O. A., Butler, J., Dracup, K., Ezekowitz, M. D., Khera, A. (2011). Forecasting the future of cardiovascular disease in the United States a policy statement from the American heart association. Circulation, 123 (8), 933 944. Heitmann, B. L., Frederiksen, P., & Lissner, L. (2004). Hip circumference and cardiovascular morbidity and mortality in men and women. Obesity, 12 (3), 482 487. Heymsfield, S. B., & Wadden, T. A. (2017). Mechanisms, Pathophysiology, and Management of Obesity. New England Journal of Medicine, 376 (3), 254 266. Hippisley Cox, J., Coupland, C., Vinogradova, Y., Robson, J., May, M., & Brindle, P. (2007). Derivation an d validation of QRISK, a new cardiovascular disease risk score for the United Kingdom: prospective open cohort study. BMJ, 335 (7611), 136.


92 Huang, P. L. (2009). A comprehensive definition for metabolic syndrome. Disease models & mechanisms 2 (5 6), 231 237 Hu, F. B., Stampfer, M. J., Solomon, C. G., Liu, S., Willett, W. C., Speizer, F. E., Manson, J. E. (2001). The impact of diabetes mellitus on mortality from all causes and coronary heart disease in women: 20 years of follow up. Archives of internal medicine, 161 (14), 1717 1723. Jensen, M. D., Ryan, D. H., Apovian, C. M., Ard, J. D., Comuzzie, A. G., Donato, K. A., Kushner, R. F. (2014). 2013 AHA/ACC/TOS guideline for the management of overweight and obesity in adults: a report of the American College of Cardiology/American Heart Association Task Force on Practice Guidelines and The Obesity Society. Journal of the American College of Cardiology, 63 (25_PA). Jousilahti, P., Vartiainen, E., Tuomilehto, J., & Puska, P. (1999). Sex, age, cardiovasc ular risk factors, and coronary heart disease. Circulation, 99 (9), 1165 1172. Kahn, H. S., Gu, Q., Bullard, K. M., Freedman, D. S., Ahluwalia, N., & Ogden, C. L. (2014). Population distribution of the sagittal abdominal diameter (SAD) from a representativ e sample of US adults: comparison of SAD, waist circumference and body mass index for identifying dysglycemia. PloS one, 9 (10), e108707. Kalantar Zadeh, K., Streja, E., Kovesdy, C. P., Oreopoulos, A., Noori, N., Jing, J., Mehrotra, R. (2010). The ob esity paradox and mortality associated with surrogates of body size and muscle mass in patients receiving hemodialysis. Paper presented at the Mayo Clin Proc. Kartheuser, A. H., Leonard, D. F., Penninckx, F., Paterson, H. M., Brandt, D., Remue, C., R is, F. (2013). Waist circumference and waist/hip ratio are better predictive risk factors for mortality and morbidity after colorectal surgery than body mass index and body surface area. Annals of surgery, 258 (5), 722 730. Katsuki, A., Sumida, Y., Urakawa H., Gabazza, E. C., Murashima, S., Maruyama, N., Adachi, Y. (2003). Increased visceral fat and serum levels of triglyceride are associated with insulin resistance in Japanese metabolically obese, normal weight subjects with normal glucose tolerance Diabetes care, 26 (8), 2341 2344. Keelan, P. C., Bielak, L. F., Ashai, K., Jamjoum, L. S., Denktas, A. E., Rumberger, J. A., Schwartz, R. S. (2001). Long term prognostic value of coronary calcification detected by electron beam computed tomography in patients undergoing coronary angiography. Circulation, 104 (4), 412 417. Kennel, W. B., O'Agustino, R. B., & Betanger, A. J. (1968). Framingham Study. An epidemiological investigation of cardiovascular disease, sect, 30


93 Kim, M. K., Han, K., Kwon, H. S ., Song, K. H., Yim, H. W., Lee, W. C., & Park, Y. M. (2014). Normal weight obesity in Korean adults. Clinical endocrinology, 80 (2), 214 220. Knowles, J. W., Zarafshar, S., Pavlovic, A., Goldstein, B. A., Tsai, S., Li, J., Kiernan, M. (2017). Impact of a Genetic Risk Score for Coronary Artery Disease on Reducing Cardiovascular Risk: A Pilot Randomized Controlled Study. Frontiers in Cardiovascular Medicine, 4 53. Koopman, R. J., & Mainous, A. (2008). Evaluating multivariate risk scores for clinical decision making. FAMILY MEDICINE KANSAS CITY 40 (6), 412. Koutsari, C., Snozek, C. L., & Jensen, M. D. (2008). Plasma NEFA storage in adipose tissue in the postprandial state: sex related and regional differences. Diabetologia, 51 (11), 2041 2048. Larsen B. A., Allison, M. A., Kang, E., Saad, S., Laughlin, G. A., Araneta, M. R. G., Wassel, C. L. (2014). Associations of physical activity and sedentary behavior with regional fat deposition. Medicine and science in sports and exercise, 46 (3), 520. La vie, C. J., De Schutter, A., Parto, P., Jahangir, E., Kokkinos, P., Ortega, F. B., Milani, R. V. (2016). Obesity and Prevalence of Cardiovascular Diseases and Prognosis The Obesity Paradox Updated. Progress in cardiovascular diseases, 58 (5), 537 547. Lavie, C. J., Milani, R. V., & Ventura, H. O. (2009). Obesity and cardiovascular disease. Journal of the American College of Cardiology, 53 (21), 1925 1932. Lear, S., James, P., Ko, G., & Kumanyika, S. (2010). Appropriateness of waist circumference and waist to hip ratio cutoffs for different ethnic groups. European journal of clinical nutrition, 64 (1), 42. Levesque, V., Poirier, P., Despres, J. P., & Almeras, N. (2017). Relation Between a Simple Lifestyle Risk Score and Established Biological Risk Factors for Cardiovascular Disease. Am J Cardiol doi: 10.1016/j.amjcard.2017.08.008 Li, C., Ford, E. S., McGuir e, L. C., & Mokdad, A. H. (2007). Increasing trends in waist circumference and abdominal obesity among US adults. Obesity, 15 (1), 216 216. Lissner, L., Bjrkelund, C., Heitmann, B. L., Seidell, J. C., & Bengtsson, C. (2001). Larger hip circumference indep endently predicts health and longevity in a Swedish female cohort. Obesity, 9 (10), 644 646. Liu, K., Daviglus, M. L., Loria, C. M., Colangelo, L. A., Spring, B., Moller, A. C., & Lloyd Jones, D. M. (2012). Healthy lifestyle through young adulthood and the presence of low cardiovascular disease risk profile in middle age. Circulation, 125 (8), 996 1004.


94 Lloyd Jones, D. M., Hong, Y., Labarthe, D., Mozaffarian, D., Appel, L. J., Van Horn, L., Tomaselli, G. F. (2010). Defining and setting national goals for cardiovascular Strategic Impact Goal through 2020 and beyond. Circulation, 121 (4), 586 613. Lloyd Jones, D. M., Wilson, P. W., Larson, M. G., Beiser, A., Leip, E. P., D'Agostino, R. B., & Levy, D. (2004). Framingham risk score and prediction of lifetime risk for coronary heart disease. Am J Cardiol, 94 (1), 20 24. Ma, J., Ward, E. M., Siegel, R. L., & Jemal, A. (2015). Temporal Trends in Mortality in the United States, 1969 2013. J AMA, 314 (16), 1731 1739. doi: 10.1001/jama.2015.12319 MacKay, M. F., Haffner, S. M., Wagenknecht, L. E., D'agostino, R. B., & Hanley, A. J. (2009). Prediction of Type 2 diabetes using alternate anthropometric measures in a multi ethnic cohort. Diabetes car e, 32 (5), 956 958. Mahabadi, A. A., Massaro, J. M., Rosito, G. A., Levy, D., Murabito, J. M., Wolf, P. A., Hoffmann, U. (2009). Association of pericardial fat, intrathoracic fat, and visceral abdominal fat with cardiovascular disease burden: the Fra mingham Heart Study. European heart journal, 30 (7), 850 856. Mainous, A. G., Koopman, R. J., Diaz, V. A., Everett, C. J., Wilson, P. W., & Tilley, B. C. (2007). A coronary heart disease risk score based on patient reported information. Am J Cardiol, 99 (9) 1236 1241. Mainous, A. G., Tanner, R. J., Anton, S. D., & Jo, A. (2015). Grip strength as a marker of hypertension and diabetes in healthy weight adults. American journal of preventive medicine, 49 (6), 850 858. Mainous, A. G., Tanner, R. J., Anton, S. D., & Jo, A. (2016). Low Grip Strength and Prediabetes in Normal Weight Adults. The Journal of the American Board of Family Medicine, 29 (2), 280 282. Mainous, A. G., Tanner, R. J., Anton, S. D., Jo, A., & Luetke, M. C. (2017). Physical Activity and Abnorm al Blood Glucose Among Healthy Weight Adults. American journal of preventive medicine Mainous, A. G., Tanner, R. J., Jo, A., & Anton, S. D. (2016). Prevalence of prediabetes and abdominal obesity among healthy weight adults: 18 Year trend. The Annals of Family Medicine, 14 (4), 304 310. Menke, A., Casagrande, S., Geiss, L., & Cowie, C. C. (2015). Prevalence of and Trends in Diabetes Among Adults in the United States, 1988 2012. JAMA, 314 (10), 1021 1029. doi: 10.1001/jama.2015.10029


95 Mokdad, A. H., Ford, E. S., Bowman, B. A., Dietz, W. H., Vinicor, F., Bales, V. S., & Marks, J. S. (2003). Prevalence of obesity, diabetes, and obesity related health risk factors, 2001. JAMA, 289 (1), 76 79. Muntner, P., Colantonio, L. D., Cushman, M., Goff, D. C., Howard, G., Howard, V. J., Safford, M. M. (2014). Validation of the atherosclerotic cardiovascular disease Pooled Cohort risk equations. JAMA, 311 (14), 1406 1415. Myint, P. K., Kwok, C. S., Luben, R. N., Wareham, N. J., & Khaw, K. T. (2014). Body fat percentage body mass index and waist to hip ratio as predictors of mortality and cardiovascular disease. Heart heartjnl 2014 305816. Nana, A., Slater, G. J., Hopkins, W. G., & Burke, L. M. (2013). Effects of exercise sessions on DXA measurements of body compositi on in active people. Med Sci Sports Exerc, 45 (1), 178 185. N ational H ealth A nd N utrition E xamination S urvey Survey Design Factors, from https:// National Health And Nutrition Examination Survey. (2014). [Overview; National Health and Nutrition Examination Survey]. National Heart, Lung, and Blood Institute The Multi Ethnic Study of Atherosclerosis (MESA): 2000 2012 Retrieved from: https:// www.mesa National Heart, Lung, and Blood Institute (2002). Detection, evaluation and treatment of high blood cholesterol in adults (Adult Treatment Panel III): Final report. NIH Publication (02 5215). Ogden, C. L., Carroll M. D., Lawman, H. G., Fryar, C. D., Kruszon Moran, D., Kit, B. K., & Flegal, K. M. (2016). Trends in obesity prevalence among children and adolescents in the United States, 1988 1994 through 2013 2014. JAMA, 315 (21), 2292 2299. Oreopoulos, A., Padwal, R ., Kalantar Zadeh, K., Fonarow, G. C., Norris, C. M., & McAlister, F. A. (2008). Body mass index and mortality in heart failure: a meta analysis. Am Heart J, 156 (1), 13 22. doi: 10.1016/j.ahj.2008.02.014 Ortega, F. B., Lavie, C. J., & Blair, S. N. (2016). Obesity and Cardiovascular Disease. Circ Res, 118 (11), 1752 1770. doi: 10.1161/circresaha.115.306883 Pasco, J. A., Holloway, K. L., Dobbins, A. G., Kotowicz, M. A., Williams, L. J., & Brennan, S. L. (2014 ). Body mass index and measures of body fat for defining obesity and underweight: a cross sectional, population based study. BMC obesity, 1 (1), 1.


96 Patel, K., Taksler, G., Hu, B., & Rothberg, M. (2017). Prevalence of Elevated Cardiovascular Risks in Young Adults. Peterson, M. D., Al Snih, S., Stoddard, J., Shekar, A., & Hurvitz, E. A. (2014). Obesity misclassification and the metabolic syndrome in adults with functional mobility impairments: Nutrition Examination Survey 2003 2006. Preventive medicine, 60 71 76. Pouliot, M. C., Desprs, J. P., Lemieux, S., Moorjani, S., Bouchard, C., Tremblay, A., Lupien, P. J. (1994). Waist circumference and abdominal sagittal diameter: best simple anthropometric indexes of abdominal visceral adipose tissue accumula tion and related cardiovascular risk in men and women. Am J Cardiol, 73 (7), 460 468. Preis, S. R., Massaro, J. M., Hoffmann, U., D'Agostino Sr, R. B., Levy, D., Robins, S. J., Fox, C. S. (2010). Neck circumference as a novel measure of cardiometabol ic risk: the Framingham Heart study. The Journal of Clinical Endocrinology & Metabolism, 95 (8), 3701 3710. Pyrl, K., De Backer, G., Graham, I., Poole Wilson, P., & Wood, D. (1994). Prevention of coronary heart disease in clinical practice. European hea rt journal, 15 (10), 1300 1331. Raji, A., Seely, E. W., Arky, R. A., & Simonson, D. C. (2001). Body fat distribution and insulin resistance in healthy Asian Indians and Caucasians. The Journal of Clinical Endocrinology & Metabolism, 86 (11), 5366 5371. Ram saran, C., & Maharaj, R. G. (2017). Normal weight obesity among young adults in Trinidad and Tobago: pre valence and associated factors. International journal of adolescent medicine and health 29 (2). Ridker, P. M., Buring, J. E., Rifai, N., & Cook, N. R. (20 07). Development and validation of improved algorithms for the assessment of global cardiovascular risk in women: the Reynolds Risk Score. JAMA, 297 (6), 611 619. Robinson, T., Jackson, R., Wells, S., Kerr, A., & Marshall, R. (2017). An observational study of how clinicians use cardiovascular risk assessment to inform statin prescribing decisions. N Z Med J, 130 (1463), 28 38. Rodondi, N., Locatelli, I., Aujesky, D., Butler, J., Vittinghoff, E., Simonsick, E., Pletcher, M. J. (2012). Framingham risk score and alternatives for prediction of coronary heart disease in older adults. PloS one, 7 (3), e34287.


97 Roger, V. L., Go, A. S., Lloyd Jones, D. M., Adams, R. J., Berry, J. D., Brown, T. M., Ford, E. S. (2011). Heart Disease and Stroke Statistics 2011 Update1. About 1. About These Statistics2. American Heart Association's 2020 Impact Goals3. Cardiovascular Diseases4. S ubclinical Atherosclerosis5. Coronary Heart Disease, Acute Coronary Syndrome, and Angina Pectoris6. Stroke (Cerebrovascular Disease) 7. High Blood Pressure8. Congenital Cardiovascular Defects9. Cardiomyopathy and Heart Failure10. Other Cardiovascular Disea ses11. Family History and Genetics12. Risk Factor: Smoking/Tobacco Use13. Risk Factor: High Blood Cholesterol and Other Lipids14. Risk Factor: Physical Inactivity15. Risk Factor: Overweight and Obesity16. Risk Factor: Diabetes Mellitus17. End Stage Renal D isease and Chronic Kidney Disease18. Metabolic Syndrome19. Nutrition20. Quality of Care21. Medical Procedures22. Economic Cost of Cardiovascular Disease23. At a Glance Summary Tables24. Glossary. Circulation, 123 (4), e18 e209. Romero Corral, A., Montori, V. M., Somers, V. K., Korinek, J., Thomas, R. J., Allison, T. G., Lopez Jimenez, F. (2006). Association of bodyweight with total mortality and with cardiovascular events in coronary artery disease: a systematic review of cohort studies. The Lancet, 3 68 (9536), 666 678. Romero Corral, A., Somers, V. K., Sierra Johnson, J., Korenfeld, Y., Boarin, S., Korinek, J., Lopez Jimenez, F. (2009). Normal weight obesity: a risk factor for cardiometabolic dysregulation and cardiovascular mortality. European heart journal ehp487. Sahakyan, K. R., Somers, V. K., Rodriguez Escudero, J. P., Hodge, D. O., Carter, R. E., Sochor, O., Singh, P. (2015). Normal weight central obesity: implications for total and cardiovascular mortality. Annals of internal medic ine, 163 (11), 827 835. Sahakyan, K. R., Somers, V. K., Rodriguez Escudero, J. P., Hodge, D. O., Carter, R. E., Sochor, O., Lopez Jimenez, F. (2015). Normal Weight Central Obesity: Implications for Total and Cardiovascular Mortality. Ann Intern Med, 163 (11), 827 835. doi: 10.7326/m14 2525 Sampaio, L. R., Simes, E. J., Assis, A. M. O., & Ramos, L. R. (2007). Validity and reliability of the sagittal abdominal diameter as a predictor of visceral abdominal fat. Arquivos Brasileiros de Endocrinologia & Me tabologia, 51 (6), 980 986. Schrager, M. A., Metter, E. J., Simonsick, E., Ble, A., Bandinelli, S., Lauretani, F., & Ferrucci, L. (2007). Sarcopenic obesity and infla mmation in the InCHIANTI study. Journal of Applied Physiology 102 (3), 919 925. Shah, R. V., Murthy, V. L., Colangelo, L. A., Reis, J., Venkatesh, B. A., Sharma, R., Rana, J. S. (2016). Association of fitness in young adulthood with survival and cardiovascular risk: the Coronary Artery Risk Development in Young Adults (CARDIA) study. JAMA I nternal Medicine, 176 (1), 87 95.


98 Shakiba, M., Soori, H., Mansournia, M. A., Nazari, S. S., & Salimi, Y. (2016). Adjusting for reverse causation to estimate the effect of obesity on mortality after incident heart failure in the Atherosclerosis Risk in Comm unities (ARIC) study. Epidemiol Health, 38 e2016025. doi: 10.4178/epih.e2016025 Shea, J., King, M., Yi, Y., Gulliver, W., & Sun, G. (2012). Body fat percentage is associated with cardiometabolic dysregulation in BMI defined normal weight subjects. Nutriti on, Metabolism and Cardiovascular Diseases, 22 (9), 741 747. Sidney, S., Quesenberry, C. P., Jaffe, M. G., Sorel, M., Nguyen Huynh, M. N., Kushi, L. H., Rana, J. S. (2016). Recent Trends in Cardiovascular Mortality in the United States and Public Hea lth Goals. JAMA Cardiology Silventoinen, K., Magnusson, P. K., Tynelius, P., Batty, G. D., & Rasmussen, F. (2009). Association of body size and muscle strength with incidence of coronary heart disease and cerebrovascular diseases: a population based coho rt study of one million Swedish men. Int J Epidemiol, 38 (1), 110 118. doi: 10.1093/ije/dyn231 Smith, D. (2016). Waist To Hip Ratio vs. Body Mass Index as a Predictor of Total Mortality for People with Normal Weight and Central Obesity. Sokol, A., Wirth, M D., Manczuk, M., Shivappa, N., Zatonska, K., Hurley, T. G., & Hbert, J. R. (2016). Association between the dietary inflammatory index, waist to hip ratio and metabolic syndrome. Nutrition research, 36 (11), 1298 1303. Stampfer, M. J., & Colditz, G. A. (1991). Estrogen replacement therapy and coronary heart disease: a quantitative assessment of the epidemiologic evidence. Preventive medicine, 20 (1), 47 63. Stenholm, S., Harris, T. B., Rantanen, T., Visser, M., Kritchevsky, S. B., & Ferrucci, L. (2008). Sarcopenic obesity definit ion, etiology and consequences. Current opinion in clinical nutrition and metabolic care 11 (6), 693. Sullivan, L. M., Massaro, J. M., & D'Agostino, R. B. (2004). Presentation of multivaria te data for clinical use: The Framingham Study risk score functions. Statistics in medicine, 23 (10), 1631 1660. Tchkonia, T., Thomou, T., Zhu, Y., Karagiannides, I., Pothoulakis, C., Jensen, M. D., & Kirkland, J. L. (2013). Mechanisms and metabolic implic ations of regional differences among fat depots. Cell metabolism, 17 (5), 644 656. Tucker, A. M., Vogel, R. A., Lincoln, A. E., Dunn, R. E., Ahrensfield, D. C., Allen, T. W., Yates, A. P. (2009). Prevalence of cardiovascular disease risk factors amon g National Football League players. JAMA, 301 (20), 2111 2119. doi: 10.1001/jama.2009.716 Tunstall Pedoe, H., Woodward, M., Tavendale, R., A'brook, R., & McCluskey, M. K. (1997). Comparison of the prediction by 27 different factors of coronary heart


99 disease and death in men and women of the Scottish Heart Health Study: cohort study. BMJ, 315 (7110), 722 729. U.S. Preventive Services Task Force (USPSTF) (2014). Final Recommendation Statement : Healthful Diet and Physical Activity: Counseling Adults with High Risk for CVD: US Preventive Services Task Force from nStatementFinal/healthy diet and physical activity counseling adults with high risk of cvd U.S. Preventive Services Task Force ( USPSTF) (2012). Obesity In Adults: Screening and Management, from https:// Final/obesity in adults screening and management U.S. Preventive Services Task Force (USPSTF) (2015). Final Recomme ndation Statement : Abnormal Blood Glucose and Type 2 Diabetes Mellitus: Screening: US Preventive Services Task Force from nStatementFinal/screening for abnormal blood glucose and typ e 2 diabetes U.S. Preventive Services Task Force (USPSTF) (2012). Healthful Diet and Physical Activity for Cardiovascular Disease Prevention in Adults: Behavioral Counseling: US Preventive Services Task Force from https://www.uspreventiveservicestaskforc Final/obesity in adults screening and management U.S. Preventive Services Task Force (USPSTF) (2016). Final Recommendation Statement: Statin Use for the Primary Prevention of Cardiovascular Disease in Adults: Preventive Me dication, from https:// nStatementFinal/statin use in adults preventive medication1 Voulgari, C., Tentolouris, N., Dilaveris, P., Tousoulis, D., Katsilambros, N., & Stefanadis, C. (2011). Incr eased heart failure risk in normal weight people with metabolic syndrome compared with metabolically healthy obese individuals. Journal of the American College of Cardiology, 58 (13), 1343 1350. Wang, Z. J., Zhou, Y. J., Galper, B. Z., Gao, F., Yeh, R. W., & Mauri, L. (2015). Association of body mass index with mortality and cardiovascular events for patients with coronary artery disease: a systematic review and meta analysis. Heart heartjnl 2014 307119. Wei, M., Gaskill, S. P., Haffner, S. M., & Stern, M. P. (1998). Effects of diabetes and level of glycemia on all cause and cardiovascular mortality: the San Antonio Heart Study. Diabetes care, 21 (7), 1167 1172.

PAGE 100

100 W orld H ealth O rganization (1995). Physical status: The use of and interpretation of anthropometry, Report of a WHO Expert Committee. World Health Organization (2008). Waist circumference and waist hip ratio. Report of a WHO Expert Consultation Geneva: World Health Organization 1 47. W orld Health Organization (2016). Obesity and Overweight Fact sheet, from Wilson, P. W., Abbott, R. D., & Castelli, W. P. (1988). High density lipoprotein cholesterol and mortality. The Framingham Heart S tudy. Arteriosclerosis, thrombosis, and vascular biology, 8 (6), 737 741. Wilson, P. W., Bozeman, S. R., Burton, T. M., Hoaglin, D. C., Ben Joseph, R., & Pashos, C. L. (2008). Prediction of first events of coronary heart disease and stroke with considerati on of adiposity. Circulation, 118 (2), 124 130. Wilson, P. W., D'agostino, R. B., Sullivan, L., Parise, H., & Kannel, W. B. (2002). Overweight and obesity as determinants of cardiovascular risk: the Framingham experience. Archives of internal medicine, 162 (16), 1867 1872. W. B. (1998). Prediction of coronary heart disease using risk factor categories. Circulation, 97 (18), 1837 1847. Yoon, S. S., Burt, V., Louis, T., & Carroll, M. D. (2012). Hypertension among adults in the United States, 2009 2010. NCHS data brief (107), 1 8. Zamboni, M., Turcato, E., Armellini, F., Kahn, H., Zivelonghi, A., Santana, H., Bosello, O. (1998). Sagittal abdominal diameter as a practic al predictor of visceral fat. International journal of obesity, 22 (7), 655 660. Zhu, S., Heshka, S., Wang, Z., Shen, W., Allison, D. B., Ross, R., & Heymsfield, S. B. (2004). Combination of BMI and waist circumference for identifying cardiovascular risk factors in whites. Obesity, 12 (4), 633 645. Zhu, W., Zeng, N., & Wang, N. (2010). Sensitivity, specificity, accuracy, associated confidence interval and ROC analysis with practical SAS implementations. NESUG proceedings: health care and life sciences, Baltimore, Maryland 1 9. operating characteristic analysis for evaluating diagnostic tests and predictive models. Circulation, 115 (5), 654 657.

PAGE 101

101 BIOGRAPHICAL SKETCH Ara Jo was born in Seoul, Republic of Korea. She had diverse education backgrounds. She had Bachelor of Economics a nd Exercise Science at Ewha Womans University. After graduation, she mo degree in sport m anagement at Florida State University. After finishing her master degree she decided to study health services research at the Department of Health Services Research, Management and Policy. Her research interests stemmed from these broad backgrounds. During her PhD program, her primary research focused on preventive care services for chronic diseases such as screening and lif estyle intervention and body composition assessments.