DEVELOPMENT OF A CHAID DECISION TREE FOR ASSESSING RISK OF DETECTING METABOLIC SYNDROME IN ADULTS, AGE YEARS. A Thesis.

Similar documents
Rehabilitation and Research Training Center on Secondary Conditions in Individuals with SCI. James S. Krause, PhD

Know Your Number Aggregate Report Single Analysis Compared to National Averages

Why Do We Treat Obesity? Epidemiology

The Metabolic Syndrome: Is It A Valid Concept? YES

METABOLIC SYNDROME IN OBESE CHILDREN AND ADOLESCENTS

A n aly tical m e t h o d s

RELATIONS AMONG OBESITY, ADULT WEIGHT STATUS AND CANCER IN US ADULTS. A Thesis. with Distinction from the School of Allied Medical

Andrew Cohen, MD and Neil S. Skolnik, MD INTRODUCTION

Metabolic Syndrome among Type-2 Diabetic Patients in Benghazi- Libya: A pilot study. Arab Medical University. Benghazi, Libya

ABSTRACT. Dr. Jiuzhou Song, Department of Avian and Animal Sciences. Blacks in the country suffer from higher prevalences of obesity, diabetes,

Know Your Number Aggregate Report Comparison Analysis Between Baseline & Follow-up

SCIENTIFIC STUDY REPORT

Metabolic Syndrome Update The Metabolic Syndrome: Overview. Global Cardiometabolic Risk

Metabolic Syndrome and Workplace Outcome

Total risk management of Cardiovascular diseases Nobuhiro Yamada

Objectives. Objectives. Alejandro J. de la Torre, MD Cook Children s Hospital May 30, 2015

Joslin Diabetes Center Primary Care Congress for Cardiometabolic Health 2013 The Metabolic Syndrome: Is It a Valid Concept?

Plasma fibrinogen level, BMI and lipid profile in type 2 diabetes mellitus with hypertension

Welcome and Introduction

Biomarkers and undiagnosed disease

Journal of the American College of Cardiology Vol. 48, No. 2, by the American College of Cardiology Foundation ISSN /06/$32.

Text-based Document. Predicting Factors of Body Fat of Metabolic Syndrome Persons. Downloaded 13-May :51:47.

Modelling Reduction of Coronary Heart Disease Risk among people with Diabetes

ISCHEMIC VASCULAR DISEASE (IVD) MEASURES GROUP OVERVIEW

Statistical Fact Sheet Populations

The Physician, the Community and Health Care Reform

Society for Behavioral Medicine 33 rd Annual Meeting New Orleans, LA

Individual Study Table Referring to Item of the Submission: Volume: Page:

Metabolic Syndrome in Asians

Risk Factors for Heart Disease

Association between arterial stiffness and cardiovascular risk factors in a pediatric population

Metabolic Syndrome.

Andrejs Kalvelis 1, MD, PhD, Inga Stukena 2, MD, Guntis Bahs 3 MD, PhD & Aivars Lejnieks 4, MD, PhD ABSTRACT INTRODUCTION. Riga Stradins University

Development of the Automated Diagnosis CT Screening System for Visceral Obesity

Obesity, Metabolic Syndrome, and Diabetes: Making the Connections

Metabolic Syndrome: What s in a name?

3/20/2011. Body Mass Index (kg/[m 2 ]) Age at Issue (*BMI > 30, or ~ 30 lbs overweight for 5 4 woman) Mokdad A.H.

Diabetes Day for Primary Care Clinicians Advances in Diabetes Care

Relationship of Waist Circumference and Lipid Profile in Children

A study of waist hip ratio in identifying cardiovascular risk factors at Government Dharmapuri College Hospital

A Study to Show Postprandial Hypertriglyceridemia as a Risk Factor for Macrovascular Complications in Type 2 Diabetis Mellitus

Is there an association between waist circumference and type 2 diabetes or impaired fasting glucose in US adolescents?

The Metabolic Syndrome Update The Metabolic Syndrome Update. Global Cardiometabolic Risk

Established Risk Factors for Coronary Heart Disease (CHD)

Guidelines on cardiovascular risk assessment and management

Chapter 2: Identification and Care of Patients With Chronic Kidney Disease

DETERMINANTS OF DAY-NIGHT DIFFERENCE IN BLOOD PRESSURE IN SUBJECTS OF AFRICAN ANCESTRY

Global Coronary Heart Disease Risk Assessment of U.S. Persons With the Metabolic. Syndrome. and Nathan D. Wong, PhD, MPH

Table S1. Characteristics associated with frequency of nut consumption (full entire sample; Nn=4,416).

SUPPLEMENTARY DATA. Supplementary Methods

Figure S1. Comparison of fasting plasma lipoprotein levels between males (n=108) and females (n=130). Box plots represent the quartiles distribution

The investigation of serum lipids and prevalence of dyslipidemia in urban adult population of Warangal district, Andhra Pradesh, India

Report Operation Heart to Heart

Roadmap. Diabetes and the Metabolic Syndrome in the Asian Population. Asian. subgroups 8.9. in U.S. (% of total

Clinical Study Synopsis

290 Biomed Environ Sci, 2016; 29(4):

PREVALENCE OF METABOLİC SYNDROME İN CHİLDREN AND ADOLESCENTS

Page 1. Disclosures. Background. No disclosures

The Metabolic Syndrome Prof. Jean-Pierre Després

Test5, Here is Your My5 to Health Profile with Metabolic Syndrome Insight

Metabolic syndrome and insulin resistance in an urban and rural adult population in Sri Lanka

CARDIOVASCULAR RISK FACTORS & TARGET ORGAN DAMAGE IN GREEK HYPERTENSIVES

Metabolic Syndrome: A Preventable & Treatable Cluster of Conditions

Cardiovascular Complications of Diabetes

Supplementary Appendix

4/7/ The stats on heart disease. + Deaths & Age-Adjusted Death Rates for

Clinical Practice Guideline Key Points

Term-End Examination December, 2009 MCC-006 : CARDIOVASCULAR EPIDEMIOLOGY

Retrospective Cohort Study for the Evaluation of Life- Style Risk Factors in Developing Metabolic Syndrome under the Estimated Abdominal Circumference

Zhengtao Liu 1,2,3*, Shuping Que 4*, Lin Zhou 1,2,3 Author affiliation:

Implications of The LookAHEAD Trial: Is Weight Loss Beneficial for Patients with Diabetes?

5/28/2010. Pre Test Question

MOLINA HEALTHCARE OF CALIFORNIA

The State of Play of Diabetes Indicators

Metabolic Syndrome. Shon Meek MD, PhD Mayo Clinic Florida Endocrinology

Table of Contents. Page 2 of 20

CHAPTER 3 DIABETES MELLITUS, OBESITY, HYPERTENSION AND DYSLIPIDEMIA IN ADULT CENTRAL KERALA POPULATION

PIEDMONT ACCESS TO HEALTH SERVICES, INC. Guidelines for Screening and Management of Dyslipidemia

3/25/2010. Age-adjusted incidence rates for coronary heart disease according to body mass index and waist circumference tertiles

Relationship of Body Mass Index, Waist Circumference and Cardiovascular Risk Factors in Chinese Adult 1

Screening Results. Juniata College. Juniata College. Screening Results. October 11, October 12, 2016

Cardiovascular Disease Risk Behaviors of Nursing Students in Nursing School

How would you manage Ms. Gold

Effectiveness of a Multidisciplinary Patient Assistance Program in Diabetes Care

Metabolic Syndrome: Why Should We Look For It?

KEEP 2009 Summary Figures

Measure Owner Designation. AMA-PCPI is the measure owner. NCQA is the measure owner. QIP/CMS is the measure owner. AMA-NCQA is the measure owner

METABOLIC SYNDROME IN A JORDANIAN COHORT: DEMOGRAPHY, COMPLICATIONS AND PREDICTORS OF CARDIOVASCULAR DISEASES

Appendix This appendix was part of the submitted manuscript and has been peer reviewed. It is posted as supplied by the authors.

AUTONOMIC FUNCTION IS A HIGH PRIORITY

USRDS UNITED STATES RENAL DATA SYSTEM

Director, Employee Health & Productivity. Coordinator, Employee Health & Productivity

Socioeconomic status risk factors for cardiovascular diseases by sex in Korean adults

Discussion points. The cardiometabolic connection. Cardiometabolic Risk Management in the Primary Care Setting

HSN301 REVISION NOTES TOPIC 1 METABOLIC SYNDROME

LEPTIN AS A NOVEL PREDICTOR OF DEPRESSION IN PATIENTS WITH THE METABOLIC SYNDROME

Since 1980, obesity has more than doubled worldwide, and in 2008 over 1.5 billion adults aged 20 years were overweight.

A COMPREHENSIVE REPORT ISSUED BY THE AMERICAN ASSOCIATION OF CLINICAL ENDOCRINOLOGISTS IN PARTNERSHIP WITH:

Association between Raised Blood Pressure and Dysglycemia in Hong Kong Chinese

Clinical Trial Synopsis TL-OPI-518, NCT#

Transcription:

DEVELOPMENT OF A CHAID DECISION TREE FOR ASSESSING RISK OF DETECTING METABOLIC SYNDROME IN ADULTS, AGE 20-39 YEARS A Thesis Presented to The Graduate Faculty of The University of Akron In Partial Fulfillment of the Requirements for the Degree Master of Science Brian Miller August, 2012

DEVELOPMENT OF A CHAID DECISION TREE FOR ASSESSING RISK OF DETECTING METABOLIC SYNDROME IN ADULTS, AGE 20-39 YEARS Brian Miller Thesis Approved: Advisor Dr. Deborah Marino Committee Member Dr. Pei-Yang Liu Committee Member Dr. Mark Fridline Accepted: Interim School Director Dr. Sandra L. Hudak Interim Dean of the College Dr. Roberta DePompei Dean of the Graduate School Dr. George R. Newkome Date ii

ACKNOWLEDGEMENTS I would like to acknowledge Mrs. Marlene Toot and the faculty of the Department of Nutrition and Dietetics at the University of Akron for their financial and academic support throughout the duration of my graduate coursework and this thesis. Additionally I would like to give thanks to the faculty of the Statistics Department at the University of Akron for their statistical support and education. I would also like to acknowledge my loving fiancée Elise and my family for the emotional support and encouragement needed to pursue my endeavors. iii

TABLE OF CONTENTS Page LIST OF TABLES... vi LIST OF FIGURES... vii CHAPTER I. INTRODUCTION...1 II. BACKGROUND OF THE STUDY...5 Definition of Metabolic Syndrome...5 Demographics of Metabolic Syndrome...7 Prevention and Target Population...9 Dislipidemia and Metabolic Syndrome...12 Hypertension and Metabolic Syndrome...13 Hyperglycemia and Metabolic Syndrome...14 Body Composition and Metabolic Syndrome...15 Chi-Square Automatic Interaction Detection Analysis...17 National Health and Nutrition Examination and Survey...18 III METHODOLOGY...20 iv

Hypothesis and Statement of Objectives...20 NHANES 2009-2010 Decision Tree Development...20 Statistical Analysis...22 IV. RESULTS...24 NHANES 2009-2010 Sample...24 NHANES CHAID Analysis...27 V. DISCUSSION...36 Limitations...40 VI. SUMMARY...41 REFERENCES...43 APPENDIX...45 v

LIST OF TABLES Table Page 1 ATP III Clinical Criteria for Metabolic Syndrome Diagnosis...6 2 IDF Ethnic Specific Waist Circumference Cutoff Points...7 3 Lipid Values and Clinical Interpretations...13 4 Subject Demographics...24 5 Cross-Tabulation of Metabolic Syndrome by Sex, Ethnicity, and Race...26 6 Chi-Squared Metabolic Syndrome Analysis...27 vi

LIST OF FIGURES Table Page 1 Frequency of NCEP Metabolic Syndrome Criteria...25 2 The Full CHAID Decision Tree...28 3 Metabolic Syndrome Decision Tree: Level I...29 4 Metabolic Syndrome Decision Tree: Node I Full...30 5 Metabolic Syndrome Decision Tree: Node 2 Full...31 6 Metabolic Syndrome Decision Tree: Node 3 Full...32 7 Metabolic Syndrome Decision Tree: Node 4 Part I...33 8 Metabolic Syndrome Decision Tree: Node 4 Part II...34 9 Metabolic Syndrome Decision Tree: Node 4 Part III...35 vii

CHAPTER I INTRODUCTION Metabolic Syndrome(MetS) is a collection of cardio-metabolic risk factors increasing related cardiovascular morbidity and/or mortality risk which affects one in three adults in the United States (US). 1 In 1988, Gerald Reaven classified the interrelationship between inflammation, impaired fibrinolysis, hypertension, atherogenic dyslipidemia, visceral obesity, and dysglycemia and its association with developing chronic diseases including cardiovascular disease (CVD), insulin resistance, and hypertension. This constellation was termed Syndrome X. However, these criteria remain cumbersome and costly for clinicians to assess in practice. 2 A decade later the World Health Organization (WHO) and the National Cholesterol Education Program Adult Treatment Panel III (NCEP) identified relative factors associated with the risk of developing Syndrome X, called Metabolic Syndrome. 2 These factors include dyslipidemia characterized by increased triglycerides (TG) and decreased HDL-cholesterol (HDL-C), hypertension, hyperglycemia, and central obesity that, when presented in tandem, exponentially increases the risk of heart attack, stroke, other debilitating occurrences, and death. 3 Relatively, there exists a drastic increase in healthcare cost for adults presenting with MetS when compared to those that do not. 4,5 1

Prevalence and complications associated with the previously mentioned chronic diseases have become a major health concern in the US. Currently, the majority of research and scientific literature on MetS focuses on adult populations younger than 20 years and older than 40 years of age. Thus, there is limited scientific research that explores MetS risk in younger to middle age adult men and women, 20-39 years of age. Current risk predictive models are useful and cost-effective in identifying risk of developing cardio-metabolic chronic disease. However, since MetS morbidities including dyslipidemia and hypertension are not readily detectable or present until an older age, these morbidities may not be as prevalent in younger adults. 6 The opportunity for preventive intervention is of the upmost importance well before most people qualify as moderate to high risk on current predictive models. 7 Thus, it is imperative to develop preventative tools for clinicians to identify those at risk for MetS before the onset of disease. Furthermore, there exists no validated quantitative risk score for identifying and stratifying relative risk of developing MetS and its subsequent cardiovascular and metabolic implications. 8 The NCEP and International Diabetes Federation (IDF) clinical risk models are limited in their usefulness because they identify either the presence or absence of MetS. 4, 5 However much like obesity, there are varied clinical implications with increased severity of the predictors used in these models. Also, increased severity of certain criteria may prove more imperative than other criteria in predicting the presence of MetS. Creating a quantitative relative risk score that stratifies severity of morbidity and/or mortality risk based on cardio-metabolic and anthropometric data that is easily obtained 2

in a clinical setting may be invaluable for clinicians to provide improved patient centered care. 6 Worachartcheewan et al. created a decision tree model for evaluating risk of developing MetS according to the NCEP criteria using a large Thai population of men and women without regard to age, weight, or health status. The decision tree was not designed with clinical application in mind thus may not be appropriate or effective for clinical use. Although in its infancy, this method proves to be a useful starting point for other models to follow. Furthermore, the clinical application of decision trees has been extensively demonstrated. 9 The focus of this study is to create a decision tree method for detecting MetS in a population of young adults age 20-39 that emphasizes a simple measure for the initial screening. The model created in this study focuses on clinical application and cost of testing to create an effective screening tool. 10 The primary method used for the development of the decision tree in the current study is the Chi Square Automatic Interaction Detection (CHAID) analysis which is a form of analysis that determines how continuous variables best combine to explain the outcome in a given dependent binary response variable. More simply put, a decision tree analysis breaks down to a yes or no statement based on if and then logic. For this study the response variable is the presence or absence of MetS. The benefit of this analysis is that it can visualize the relationship between the binary target variable and the relative factors in a tree image. 11 The central hypothesis states that CHAID analysis will detect risk associated with MetS in adults 20-39 years of age from the NHANES 2009-2010 data to serve as a pilot 3

for the development of clinically relevant screening tool. The first aim of this study is to employ a CHAID analysis on NHANES 2009-2010 data to develop a stratified quantitative risk model for detecting MetS. The second aim of this study is to analyze the distribution of MetS measures across sex, ethnicity, and race in a representative sample of US adult s age 20-39 years. 4

CHAPTER II BACKGROUND OF THE STUDY Definition of Metabolic Syndrome Insulin resistance is the key abnormality associated with atherogenic, prothrombotic, and inflammatory states relative to CVD risk. The link between insulin resistance and the cardio-metabolic risk factors, termed Syndrome X, was established by Gerald Reaven in 1988. This model was based on inflammation, impaired fibrinolysis, hypertension, atherogenic dyslipidemia, visceral obesity, and dysglycemia. 1 However, assessing these criteria remains cumbersome and costly for clinicians to assess in practice. As a consequence many models for clinical risk assessment for developing Syndrome X use cardio-metabolic abnormalities linked to insulin resistance including intra-abdominal and visceral obesity, dyslipidemia, fasting plasma glucose, and hypertension. Organizations such as the WHO, the NCEP, and the IDF have utilized these criteria to assess relative risk of developing Syndrome X and subsequent cardiovascular morbidity and mortality. This collection of risk criteria is termed Metabolic Syndrome. 12, 13 Although these models have proven useful, neither has a direct marker of insulin resistance. Also, these models only look at the presence or absence of the various criteria rather than its quantitative value. Here a person only has to present with three of the five criteria, regardless of which of the five, to have a clinical diagnosis 5

of MetS, which in some cases downplays the severity of insulin resistance and CVD risks. The National Institute of Health: National Heart, Lung, and Blood Institute defined MetS as presenting with three of the following criteria based the NCEP guidelines which include: abdominal obesity, dyslipidemia including high TG and low HDL-C, high blood pressure, and high fasting blood glucose whose values are defined in Table 1. 13 Table 1: ATP III Clinical Criteria for Metabolic Syndrome Diagnosis Risk Factor Defining Level Waist Circumference Men >102 cm (>40in) Women >88 cm (>35in) TG* 150mg/dl HDL Cholesterol* Men <40mg/dl Women <50mg/dl Blood Pressure** 135 Systolic 85 Diastolic Fasting Plasma Glucose* 100mg/dl *Blood panels are based on a 9-12 hour fast **Following a 5 minute steady state Similarly, the IDF shares the same triglyceride, HDL-C, and blood pressure guidelines as the NCEP. The IDF differs in the fasting plasma glucose (FPG 100), and the abdominal obesity is categorized by ethnic-specific waist circumference guidelines which are defined in table 2. In this model, if the subjects Body Mass Index (BMI) is greater than 30 kg/m², abdominal obesity is assumed and does need to be measured (14). However this model is considered by experts to be a work in progress requiring further improvement and validation. 7 6

It is important to note that these criteria are simple surrogate variables for use in clinical practice to identify high-risk individuals likely to have abdominal obesity, insulin resistance, atherogenic dyslipidemia, hypertension, hyperglycemia, and a pro-thrombic profile. Després et al. states that there is insufficient evidence to support the cut-off points for clinical identification of MetS by the NCEP Guidelines. 7 Table 2: IDF Ethnic Specific Waist Circumference Cutoff Points Country/Ethnic Group Sex Waist Circumference Europids, Sub Saharan Africans, Eastern Male Female 94cm 80cm Mediterranean South Asians, Central, and South Americans Male Female 90cm 80cm Chinese Male Female 90cm 80cm Japanese Male 90cm US Clinical Assessment Female Male Female 80cm 102cm 88cm Demographics of Metabolic Syndrome MetS is a constellation of cardio-metabolic risk factors that affects one in three US adults. 1 These factors include dyslipidemia, hypertension, hyperglycemia, and central adiposity that, when present in tandem, exponentially increase the risk of heart attack, stroke, other debilitating occurrences, death, and associated healthcare cost. 8 The US Department of Health and Human Services: National Health and Labor Statistics reported that 34% of US adults met the NCEP criteria for metabolic syndrome. This report utilized odds ratios (OR) or the increase in risk or increased likelihood per unit increase of the relative measure(s). Men and women aged 40-59 years were three 7

times more likely (men OR 2.70; women OR 3.20) than those aged 20-39 years to meet the criteria for MetS. Men over the age of 60 years were four times more likely (OR 4.18, p 0.05) to meet the criteria for MetS than men aged 20-39 years. Women over the age of 60 years were six times (OR 5.50, p 0.05) more likely to meet the criteria for MetS than women aged 20-39 years. Along with age, the prevalence of MetS is proportional to BMI. Based on clinical measurement of BMI, overweight men were six and women were over five times more likely (men OR 6.17; women OR 5.5, p 0.05) to meet the criteria for MetS than normal weight, age matched men and women, respectively. 1 In 2008, CVD affected over 82 million US adults over the age of twenty, or 36.2% of the adult population. Concurrently, the CVD related mortality rate was 813,804 persons in 2007 and cost the US Healthcare System $286.6 billion. 16 The incidence of diabetes mellitus between the years 1998 and 2002 rose from 3.91% (10.48 million) to 4.84% (13.49 million) with the current prevalence of 11.3% (25.6 million) of all American adults greater than 20 years of age. 15 The Center for Disease Control and Prevention found that in 2009, the prevalence of obesity, defined as a BMI 30 kg/m², in the US was 26.7% with no state meeting the Healthy People 2010 goal to reduce overall obesity prevalence to less than 15%. Compared to 2007, there has been a 1.1% increase in obesity. 17 Schultz et al. investigated the healthcare cost of MetS in a sample of working adults. The age and gender adjusted healthcare cost for those at-risk for MetS was $4,016 per year compared to $2,117 per year for those not at-risk for MetS, a difference of $1,899 per year. Those at-risk of MetS had twice the healthcare cost as those that did 8

not. 5 Boudreau et al. showed a 24% increase in overall healthcare costs for every additional risk MetS factor related. 4 Arnlov et al. investigated the impact of obesity on mortality and cardiovascular morbidity in adults with and without MetS. Overweight and obese subjects with insulin resistance had 1.30 and 2.21 times the mortality rate compared to normal weight controls, respectively. Concurrently, overweight and obese subjects with insulin resistance had 1.88 and 2.87 times higher mortality rate from cardiovascular events compared to normal weight controls, respectively. 18 Grundy et al. states that metabolic risk needs to include risk assessment for CVD using the Framingham risk scores to identify those at a higher risk of complications from MetS. 19 Also, the hypertension risk score should be utilized to identify those at a higher risk of complications from MetS. Using multiple risk calculators in tandem with the diagnostic criteria may allow for better clinical care and decreased healthcare cost for adults meeting the criteria for MetS. 19, 20 Furthermore, formulating a quantitative tool for establishing overall risk for both developing MetS and related cardiovascular implications may prove invaluable for clinicians and patients alike. Prevention and Target Population With the state of healthcare shifting its focus from treatment of disease to prevention, it is becoming increasingly important to create and utilize preventive screening tools to identify risk of chronic disease. The NCEP model has been identified as a crude but useful screening tool that identifies those at high risk of associated cardio- 9

metabolic chronic diseases including CVD, coronary heart disease, hypertension, and/or diabetes. 7 Després et al. asserts that MetS is an all or none diagnosis that does not capture the true severity of the condition and further calls for new modeling approaches that address the variables of diagnosis as continuous variables. 7 Clinicians and healthcare professionals utilize risk calculators and simple screening models to provide affordable and accurate information on the risk of developing many debilitating diseases including, but not limited to, CVD risk. Cameron et al. explores the utility of current predictive risk models and how they under-predict overall long-term risk in younger adults. The importance of obesity in these models is often discounted due to the lack of other MetS morbidities such as dyslipidemia and hypertension that often do not manifest until older age or not in time for preventative measures to be taken to prevent morbidity. The opportunity for preventive intervention is of the upmost importance well before people qualify as moderate to high risk on current prediction models. 6 Therefore, it is imperative to develop preventative tools for clinicians to identify those at risk before the onset of disease. Additionally, MetS research has focused on children/adolescence and older adults leaving little research on younger to middle aged adult men and women, 20-39 years of age. Many adults in this age range either do not utilize healthcare or do not have health insurance. In 2008, the US Department of Health and Human Services identified that 13 million or 30% of adults aged 20-29 years did not have health insurance coverage. Furthermore, those same adults were four-times more likely than those with private insurance and twice as likely as those with Medicare to have an unmet medical need or 10

absence of a medical screening. Young adults are susceptible to not having health insurance coverage for various reasons including: being dropped from their parent s policy upon graduation and/or lack of coverage offerings from temporary or part-time employment. 21 For many young adults, socioeconomic disparities contribute to a decreased life expectancy and an increased risk of future chronic diseases such as diabetes, coronary heart disease, hypertension, and CVD. 22 Data from the CDC NCHS National Ambulatory Medical Care Survey found that in the time span from 1992 to 2000, adults from the age of 18-44 had the least physician office visits and hospital outpatient department visits of all age groups. 23 The NCEP recommends screening to identify abnormal lipid values at age twenty including two MetS criteria, TG and HDL. However, this recommendation has not been widely executed in this population. 13, 24 The documented prevalence of MetS in young adults is limited and shows a large variance of 0.6% to 13% due to the lack of healthcare utilization. Furthermore, younger adults have been largely ignored by researchers in MetS where this stage in life may be crucial to the halting and/or prevention of chronic diseases. 24 Screenings in healthcare have been used for early detection and initiation of treatment of disease to reduce mortality and morbidity. Screenings have been used in detecting various cancers including, but not limited to: cervical, colorectal, and breast. These screenings have proven to be cost effective in cancer treatment. 25 Considering MetS, early detection of problematic variance the criteria included in its diagnosis may improve the outcome and/or halt the progression of disease. 11

Dyslipidemia and Metabolic Syndrome Dyslipidemia in MetS, defined as high serum TG and low HDL-C, are crude but clinically effective markers for the presence of cardio-metabolic abnormalities. 15 An analysis of NHANES 2003-2006 data shows that 31% of adults exhibit hypertriglyceridemia which is defined as a fasting serum level greater than or equal to 150mg/dl. Additionally the same analysis shows that 25% of adults exhibit low HDL cholesterol which is defined as serum HDL cholesterol below 40mg/dl and 50mg/dl for 1, 13 men and women, respectively. The results from the Quebec Health Survey showed that when controlling for BMI, men and women exhibited a significant increase in waist circumference between subjects presenting with normo-lipidemia and high TG-low/HDL-C. 2 The current models used for assessing clinical risk of MetS do not regard the extent of dyslipidemia but rather only use cut-off points for a binary diagnosis. 7 The American Heart Association identifies varying degrees of dyslipidemia based on stratified values of TG, HDL-C, LDL-C, and TC. Table 3 illustrates the stratified lipid values and their corresponding risk. Increased waist circumference and fasting TG can be used as a screening phenotype to identify those at high risk of being carriers of MetS. 15 Després et al. found that a tandem increase of TG above 2mmol/L with a waist circumference greater than 90cm showed a greater than 80% probability of developing MetS. Those below the previously mentioned triglyceride and waist circumference levels showed only a 10% probability of developing MetS. 2 12

Table 3: Lipid Values and Clinical Interpretations TC (Total Cholesterol) Normal Borderline High High HDL (High Density Lipoprotein) CVD Risk (Men) CVD Risk (Women) Protective/ CVD Risk LDL (Low Density Lipoprotein) Optimal Above Optimal Borderline High High Very High <200mg/dl 200-239mg/dl 240mg/dl <40mg/dl <50mg/dl >60mg/dl <100mg/dl 100-129mg/dl 130-159mg/dl 160-189mg/dl 190mg/dl TG (TG) Normal <150mg/dl Borderline High 150-199mg/dl High 200-499mg/dl Very High 500mg/dl Reference Ranges and Interpretations adapted from the American Heart Association: Lipid Panel Guidelines 2011 available at: www.heart.org. 33 Hypertension and Metabolic Syndrome Hypertension is a chronic medical condition where the blood pressure in the arteries is elevated, increasing the work load on the heart to circulate blood through the blood vessels. There are two measurements in blood pressure, systolic during ventricular contraction and diastolic during relaxation. Systole represents the maximum pressure in the arteries while diastole is the minimum pressure in the arteries. The American Heart 13

Association defines hypertension as a persistent resting blood pressure greater than or equal to 140/90 mmhg. 18 Hypertension is a major risk factor for ischemia, cerebro-vascular accidents, myocardial infarction, progressive congestive heart failure, aneurysms, peripheral arterial disease and chronic kidney disease. It is notable that longevity is significantly impacted by a moderate elevation of arterial blood pressure. 18 An analysis of NHANES 2003-2006 data shows that 40% of adults exhibited hypertension. 13 A previous study identified a codependent relationship between the other MetS criteria, specifically waist circumference, and blood pressure. 1 Hyperglycemia and Metabolic Syndrome Insulin resistance is one of the primary outcomes of MetS along with CVD and hypertension. An analysis of NHANES 2003-2006 data shows that 39% of adults exhibit hyperglycemia which is defined as FPG greater than or equal to 100mg/dl. 1, 13 Diabetes mellitus is a disease characterized by hyperglycemia from the destruction of insulin secreting islet cells or the cells of the body not being able to utilize or respond to insulin. Hyperglycemia produces the classical symptoms of polyuria, polydispsia, and polyphagia. Acute complications from diabetes include hypoglycemia, diabetic ketoacidosis, or nonketotic hyperosmolar coma while chronic complications include cardiovascular complications, renal failure, diabetic neuropathy, and diabetic retinopathy. 26 14

The risk associated with MetS and its clinical screening tools focus on insulin resistance with most guidelines using FPG levels as an indirect insulin resistance marker to assess risk. The presence of diabetes has been shown to decrease longevity and lead to poorer prognoses in persons with CVD. Additionally, diabetes has been shown to exacerbate or attenuate increased inflammation and malignancy. 26 Obesity and adiposity are correlated strongly with increased fasting plasma glucose and overall insulin resistance. Insulin resistance is recognized as a core metabolic abnormality associated with the atherogenic, and diabetogenic risk factors of MetS. 7 Economically, Boudreau et al. showed that the healthcare costs of patients with MetS and diabetes differed by a magnitude of 1.6 when compared to those with MetS but no diabetes. 4 Body Composition and Metabolic Syndrome Due to limited research, it is unclear how body fat distribution based on waist to hip body ratios impacts the risk of developing metabolic syndrome. An analysis of NHANES 2003-2006 data shows that 53% of adults exhibit high abdominal obesity which is defined as a waist circumference greater than 102cm (40in) for men and 88cm (35in) for women in the NCEP. 1, 13 A key concept of the NCEP and IDF guidelines is the utilization of central adiposity distribution (waist circumference) in addition to obesity as determined by BMI. The cutoff points established for men, women, and ethnicity, used in the IDF model represent either the absence or the presence for that criterion rather than a quantitative measure. 6 15

Camhi et al. showed that waist circumference and BMI measures were highly correlated with visceral adipose tissue (range: r = 0.73-0.77), subcutaneous adipose tissue (range: r = 0.85-0.93), and fat mass (range: r = 0.91-0.94) across sex and race groups. 26 BMI and waist circumference are therefore strong predictors of obesity. Excess intra-abdominal adiposity or visceral adiposity is a stronger predictor of insulin resistance as compared to subcutaneous adiposity. 8 However, subjects matched for visceral adiposity showed no difference in insulin resistance based on variance of subcutaneous adiposity alone. 8 This study shows that BMI, fat mass, and visceral adiposity are all correlated with indices of dysglycemia. The IDF and NCEP models utilize waist circumference due to a high correlation of BMI to waist circumference (r=0.91, P<0.0001), but negates possible variance in girth which may substantially increase overall risk. Waist circumference is included based on its co-linearity with dyslipidemia and abdominal obesity. It is also important to note that waist circumference does not distinguish between visceral and subcutaneous adiposity which was previously mentioned to have a large impact on risk of insulin resistance. 8 However, waist circumference has been recognized as a powerful indicator of current and future health. In Japan, new laws require employers and the local government to measure waist circumference of employees over the age of 40 years. 6 It is important to note that measuring waist circumference is a crucial first step to screen for MetS, but the ability of anthropometrics alone to identify the presence of the MetS is limited. 2 16

Chi-Square Automatic Interaction Detection Decision tree algorithms have been used in multiple facets of healthcare and medicine for the past twenty years. Decision trees are a reliable and effective decision making technique that has high accuracy in the classification of multiple measures or variables. This technique uses a series of if and then logic statements to make a yes or no decision on the response variable, in this case the presence or absence of MetS using the NCEP guidelines as the criteria for the decision. Podgorelec et al. highlights the use of decision algorithms in medicine. For example, models have been used for: early detection of myocardial infarction to better improve treatment and reduce post-myocardial infarction mortality, another example highlighted was a model designed to detect adverse medication reactions and interactions to reduce pharmacological errors, and for improving alarm systems for patients in intensive care units based on telemetry of continuous vital sign measures to reduce the incidence of false alarms. 9 Decision trees allow for the application of multiple variables based on an ordinal stratification of continuous measures in a step-wise manner in binary decision making. For instance, credit companies use decision tree algorithms when assessing an applicant s eligibility for a credit card or approval of a loan. The primary method used for the development of the decision tree in this study is the Chi Square Automatic Interaction Detection (CHAID) analysis which determines how continuous variables best combine to explain the outcome in a given dependent binary response variable. The benefit of this analysis is that it can visualize the relationship between the binary target variable and the relative factors in a tree image. 11 17

The CHAID analysis categorizes continuous variables as ordinal based on a division into multiple groups. Next, multiple pair-wise Chi-Square independence tests are conducted with a level of significance determined using a Bonferoni Adjustment to control for type I error. This merges each level of analysis or variables into homogenous groups stratified based on the significant variance between ordinal groups. A decision tree analysis uses nodes, or division points where data is homogenous and can either split or terminate to make a decision. The next process is termed splitting, which determines which predictor best divides the parent node or previous variable. Note that Statistical Package for the Social Sciences (SPSS) allows for the definition of the first split or variable which divides the data first. These divisions continue until the maximum depth of the tree is met or if the node size falls below the user-specified 9, 11 minimum node size which both results in a terminal or decision node. National Health and Nutrition Examination Survey The National Health Survey Act was passed in 1956 which addressed the need for a continuing survey of statistical data based on the prevalence, distribution, and effects of illness and disability in the US population. Included in this survey system are direct interviews, clinical tests, laboratory measures, and physical examinations of subjects. The National Center for Health Statistics (NCHS), a branch of the US Public Health Service in the US Department of Health and Human Services conducted the first set of surveys called National Health Examination Surveys. 28 18

In 1970, to address nutrition and its relationship to health status, the Nationals Nutrition Surveillance System was established to further research unveiling the links between dietary habits and disease. With this added component, the survey was now called the National Health and Nutrition Examination Survey or NHANES. 28 Approximately 7,000 randomly-selected subjects across the US participate of the latest NHANES. Participation in this survey in confidential, voluntary, and selected participants received personal interview with a standardized physical examination. The latest available publication of the NHANES is for the time period of 2009 and 2010. The results of the survey have been utilized in multiple facets of research and have been invaluable for understanding the ever-changing status of the nation s health and identifying needs, trends, and disparities. 28 19

CHAPTER III METHODOLOGY Hypothesis and Statement of Objectives A CHAID analysis will detect risk associated with metabolic syndrome in young adults, 20-39 years of age using the NHANES 2009-2010 data to serve as a pilot for the development of clinically relevant screening tool. Objective 1: To employ a CHAID analysis on NHANES 2009-2010 data to develop a stratified quantitative risk score for detecting Metabolic Syndrome. Objective 2: To analyze the distribution of metabolic syndrome measures across sex, ethnicity, and race in a representative sample of US adults age 20-39 years. NHANES Decision Tree Development NHANES 2009-2010 data acquired from the Centers for Disease Control and Prevention website was used for this study. The dataset used for analysis is a portion of the NHANES 2009-2010. The contents of this dataset are made public by the Center for Disease Control and Prevention National Center for Health Statistics. The data collection for the NHANES 2009-2010 is under the continuation of protocol #2005-06 of NCHS 20

Research Ethics Review Board. Collection of this data started January 2009 and ended December of 2010 and this data was made public in September of 2011. 28 The full dataset includes 10,537 subjects designed to represent the population of the US across age, sex, and ethnicity. Subjects with missing criteria were excluded from the present analysis because of the inability to make a complete diagnosis of MetS (n = 7589). Subjects not meeting the inclusion criteria of an age between 20-39 years, BMI less than 20kg/m² were also excluded (total n = 2203: n = 522 for age <20 years, n = 1622 for age greater than 39 years, and n = 59 for BMI less than 20kg/m²). The final sample size meeting the inclusion criteria was 745 subjects. Data obtained on the 745 subjects include demographic information included age, sex, and ethnicity represented in binary form of ethnic or non-ethnic). In addition, anthropometric information included weight (kg), height (cm), BMI, and waist circumference (cm). Laboratories included HDL-C (mg/dl), TG (mg/dl), FPG (mg/dl), and blood pressure was expressed as systolic and diastolic pressures (mmhg). The MetS criteria follow the risk assessment model of the NCEP guidelines. Presence of MetS is defined as presence of three or more of the following factors as follows: central obesity, defined as a waist circumference greater than or equal to 102cm and 88cm for men and women, respectively, a fasting triglyceride of greater than 150mg/dl, a fasting HDL cholesterol of less than 40mg/dl and 50mg/dl for men and women, respectively, a resting blood pressure of greater than 135mmHg systolic and 85mmHg diastolic, and a fasting plasma glucose greater than or equal to 100mg/dl. All blood panels are based on a 9-12 hour fast. 28 Approval for this analysis is granted by the University of Akron s Institutional Review Board. 21

Statistical Analysis The dataset was setup in a column-wise format based on the different measures that were organized row-wise by a given sequence number. Data management performed using dataset merging and data subset functions and statistical analysis performed using IBM SPSS version 19. A chi-squared automatic interaction detection, or CHAID, analysis was used to develop the decision tree model. The first level or first division was user-specified as waist circumference, due to the measurement of this parameter having the lowest cost in MetS screening. 12, 27, 28 Parent nodes defined at 20 subjects and the child node defined at 5 subjects. The parent node is level where the data set divides into child nodes that can themselves become parent nodes or end in a terminal or decision node. Within individual categories, data stratified and merged based on Chi-squared test with significance values less than or equal to 0.05. Accuracy and risk defined by the CHAID analysis expressed as percentage. Other statistical tests include: binary logistic regression preformed on individual MetS criteria to develop predictive odds ratios relative to the likelihood of developing MetS per unit increase of the individual criteria measure from 1.0, a Chi-Squared independence test used to variance in the frequency of MetS based on sex, race, and ethnic status, a sample demographics values expressed as mean ± standard deviation, and a Pearson s correlation between BMI and waist circumference individually grouped by sex, ethnicity, and race. 11 22

The data in this analysis contained no potentially harmful identifiers outside of basic cross-tabulation statistics on the frequency of MetS in relation to sex, ethnicity, and race. The identities of subjects in this data set are confidential and not available to the researchers. Identifiable and administrative data have not been made public and were not considered for this analysis. 23

CHAPTER IV RESULTS NHANES 2009-2010 Sample NHANES 2009-2010 subject demographics are illustrated in table 1. The final sample size includes 745 subjects. Of the 745 subjects between the ages of 20-39 years, 149 or 20% present with the NCEP criteria for MetS. Table 4: Subject Demographics Parameter Mean Std. Deviation Age (yr) 29.31 5.770 Weight(Kg) 82.70 21.261 Height(cm) 168.18 9.943 Body Mass Index 29.16 6.768 SBP (mmhg) 113.80 11.672 DBP (mmhg) 66.74 11.768 Waist Circumference (cm) 96.82 15.806 HDL (mg/dl) 51.30 14.861 TG (mg/dl) 126.70 114.283 FPG (mg/dl) 97.99 24.560 Fast Time (Hr) 11.51 3.030 Subject demographics displayed Mean ± Standard Deviation. Sample size for all parameters is n=745 The frequency of the MetS criteria across the MetS, no-mets, and total sample is displayed in figure 1. The three criteria with the highest frequency are waist circumference at or above the NCEP cut-offs presenting in 86.6% of those with MetS, 24

40.0% of those without MetS, and 49.1% of the total sample, low HDL presenting in 89.2% of those with MetS, 22.3% of those without MetS, and 35.7% of the total sample, and elevated FPG presenting in 75.8% of those with MetS, 19.5% of those without MetS, and 30.7% of the total sample. Blood pressure elevation has the lowest frequency presenting in only 10.1% of those with MetS, 0.8% of those without MetS, and 2.7% of the total sample. Figure 1: Frequency of NCEP Metabolic Syndrome Criteria Above represents the frequency of NCEP Criteria for Metabolic Syndrome represented as WC (waist circumference), BP (blood pressure), TG, HDL, and FPG meeting the NCEP Metabolic Syndrome criteria. Total (n=745), no-mets (n=596), and MetS (n=149) A Chi-Squared independence test showed no significant difference of MetS frequency by sex (p=0.46), ethnicity (p=0.45), or race (p=0.15). The relative frequency of MetS remained similar across the previously mentioned categories. The cross-tabulation of MetS frequency in relation to sex, ethnicity, and race is displayed in figure 9. 25

Table 5: Cross-Tabulation of Metabolic Syndrome by Sex, Ethnicity, and Race Metabolic Syndrome No MetS MetS Total Sex Male 264 78.8% 71 21.2% 335 Female 332 81.0% 78 19.0% 410 Ethnicity Non-Ethnic 240 81.4% 55 18.6% 295 Ethnic 356 79.1% 94 20.9% 450 Race Mexican- 141 73.8% 50 26.2% 191 American Hispanic 78 81.3% 18 18.8% 96 White 240 81.4% 55 18.6% 295 Black 102 84.3% 19 15.7% 121 Other 35 83.3% 7 16.7% 42 Total 596 80.0% 149 20.0% 745 The above is a cross-tabulation of metabolic syndrome frequency based on stratification by sex, ethnicity, and race. Values displayed as proportion of the total sample and percentage of total sample. Individual logistic regression identified an odds ratio, OR, for each of the MetS criteria defined as the increase in risk of developing MetS based on the per unit increase of each criteria independent of one another. The odds ratios are as follows: waist circumference (OR 1.07), systolic and diastolic blood pressure (OR 1.04 and 1.06, respectively), HDL-C (OR 0.88), TG (OR 1.02), and FPG (OR 1.06). All criteria reached significance (p<0.001). The odds ratios above represent the increased likelihood or risk in detecting MetS based on the per unit variance of the NCEP criteria. A Pearson s correlation identified a significant relationship between BMI and waist circumference (r = 0.93, p<0.001). Also, the correlation between BMI and waist circumference when grouped individually by sex, ethnicity, and race ranged between a correlation of 0.93-0.94, (all p<0.001). 26

NHANES CHAID Analysis Each branching, or progression path through the model in this study, is a mutually exclusive path of testing for the detection of MetS. Each diagram displays a set of if and then logic statements to detect MetS based on the NCEP measures for diagnosis MetS. Each node displays a percentage based on the binary dependent variable, in this case the presence or absence of MetS, and in clinical application for the purpose of this study is used as the risk or the likelihood of detecting MetS given the variable of that branch. The outcome of the CHAID analysis is displayed in table 6. The risk or total inaccurate prediction of MetS, of an inaccurate prediction in this model is 7.7%. This model correctly identified 71.8% of those with MetS and 97.5% of those without MetS for an overall accuracy of 92.3%. Table 6: Chi-Squared Metabolic Syndrome Analysis Predicted Percent Observed No MetS MetS Correct No MetS 581 15 97.5% MetS 42 107 71.8% Overall % 83.6% 16.4% 92.3% The full CHAID analysis is displayed in figure 2. The CHAID decision tree has a depth of 5 levels with the first level user-specified as waist circumference. Each terminal node in this model is mutually exclusive of one another with the dependent variable defined as MetS. 27

Figure 2: The Full CHAID Decision Tree Displayed above is a left to right, fully opened CHAID decision tree with MetS as the dependent variable. Progression through the tree s branches is mutually exclusive. There are a total of 35 total nodes with 20 being terminal nodes. The first level is displayed in figure 3 and has a total of 4 splits all based on waist circumference. The waist circumference stratifications with associated risks are as follows: less than 86cm has a risk of 0.5%, 86cm to less than 94cm has a risk of 8.8%, 94cm to less than 103cm has a risk of 21.5%, and greater than 103 cm has a risk of 45.8%. 28

Figure 3: Metabolic Syndrome Decision Tree: Level I The above is the first split of the CHAID analysis forced on the waist circumference variable. Waist circumference measured in cm. The next figures are based off of the first level split and include nodes 1 through 4. Based on this model those in the first split presenting with a waist circumference measure less than 86cm have 0.5% chance of having MetS. Notable is that this branch only had 1 detected case of MetS. The first split in waist circumference is displayed in figure 4. 29

Figure 4: Metabolic Syndrome Decision Tree: Node I Full The above is the display of the first node from the initial split of the CHAID analysis. Waist circumference measured in cm and HDL measured in mg/dl The second split represents a waist circumference from 86cm up to 94cm representing an 8.8% chance of having MetS. Node 2 is displayed in figure 5. The next split off of node 2 is represented by the triglyceride criteria divided into 2 nodes, each themselves splitting into 2 nodes based on fasting plasma glucose. 30

Figure 5: Metabolic Syndrome Decision Tree: Node 2 Full The above is the display of the second node from the initial split of the CHAID analysis. Waist circumference measured in cm and TG and FPG measured in mg/dl The third split represents a waist circumference from 94cm up to 103cm representing a 21.5% chance of having MetS. Node 2 is displayed in figure 6. The next split off of node 2 is represented by the triglyceride criteria splitting into 2 nodes, each themselves splitting into 2 nodes with the lower based on FPG and the higher on HDL. 31

The terminal nodes from the HDL split is defined by two FPG with a risk of 40% for less than 94mg/dl and 94.4% for greater than or equal 94mg/dl. Figure 6: Metabolic Syndrome Decision Tree: Node 3 The above is the display of the third node from the initial split of CHAID analysis. Waist circumference measured in cm and TG and FPG measured in mg/dl The fourth split represents those with a waist circumference greater than 103cm representing a 45.8% chance of having MetS. The branches of Node 4 are displayed in 32

figures 7 through 9. The next split of node 4 is 3 levels of HDL which each split into 2 nodes based on fasting plasma glucose. The first split of node 4 is displayed in figure 7. Figure 7: Metabolic Syndrome Decision Tree: Node 4 Part I The above is the display of the fourth node from the initial split and the first node of the second level of the CHAID analysis. Waist circumference measured in cm and TG and FPG measured in mg/dl The next split of the second level of node four is represented by a split into 2 nodes based on fasting plasma glucose with the lower splitting into 2 nodes based on TG and the higher based on systolic blood pressure. The second split of node 4 is displayed in figure 8. 33

Figure 8: Metabolic Syndrome Decision Tree: Node 4 Part II The above is the display of the fourth node from the initial split and the second node of the second level of the CHAID analysis. Waist circumference measured in cm, HDL, TG, and FPG measured in mg/dl, and SBP measured in mmhg The third split of the second level of node four is represented by a split into 2 nodes based on FPG. The third split of node 4 is displayed in figure 9. The third terminal nodes from the HDL split is defined by two FPG with a risk of 2.1% for equal to or less than 103mg/dl and 26.7% for greater than 103mg/dl. 34

Figure 9: Metabolic Syndrome Decision Tree: Node 4 Part III Above is the display of the fourth node from the initial split and the third node of the second level of the CHAID analysis. Waist circumference measured in cm, HDL, TG, and FPG measured in mg/dl 35

CHAPTER V DISCUSSION The first objective of this pilot study was to develop a decision tree algorithm to serve as a first step in creating a reliable screening tool for accurate detection of MetS. The clinical applicability of this model is its ability to identify the need, or lack thereof for further testing in identifying MetS risk based on a initial screening procedure. 1 The uniqueness of this type of analysis is that it allows healthcare providers to identify the risk of MetS with the least amount of testing in turn that decreases the overall invasiveness and costs. Also, screening for MetS is often overlooked by healthcare providers even in young adults with healthcare coverage. For young adults that do not have the means to utilize healthcare, this tool can identify risk of MetS based on a simple measurement of waist circumference or BMI and can alert healthcare providers to have more in-depth screening in certain at-risk young adults. Surrogate variables to identify risk of developing the actual diagnosis and its relative morbidities. The American Heart Association uses systolic and diastolic blood pressure measurements of 120-139mmHg and 80-89mmHg, respectively to identify a subject as pre-hypertensive. 29, 32 Similarly, the American Diabetes Association uses impaired FPG of 100-125mg/dl to identify a subject as pre-diabetic. 16 Preliminary syndromes are related to the risk of developing the full syndrome and its consequences. 36

There are no clinically established criteria for pre-metabolic syndrome. If established, pre-metabolic syndrome diagnostic criteria could improve outcomes associated with the development of MetS or could halt the progression of MetS and its relative consequences. It is important to note that the NCEP criteria themselves are surrogate variables designed to clinically identify the associated risk of presenting with MetS which is characterized by the dyslipidemia, dysinsulemia, pro-thrombotic, and pro-inflammatory states of Gerald Reaven s original classification of MetS. According to this analysis in addition to previous studies, waist circumference may be a candidate to diagnose premetabolic syndrome. 1 The mean waist circumference of this sample was 96.82cm with 49.1% of total population and 86.6% of the population with MetS presenting with the NCEP waist circumference criteria. It is important to note that the detection model derived in this study does delineate between men and women and is designed as a pilot to a universal screening tool. The disadvantage of this all-inclusive analysis is that it negates the anthropometric differences and the different implications from the HDL measure between men and women. The first split in the decision tree derived in this analysis was forced based on waist circumference which was stratified into four groupings. In clinical application, subjects presenting with a waist circumference less than 86cm can be discounted in further testing for metabolic syndrome. In the other three groups further testing is warranted. Waist circumference is a simple measure that has little or no cost to obtain. 37

The last group identifies anyone of a waist circumference of 103cm as having a 45.8% chance of developing MetS. The results of this study show a strong correlation between BMI and waist circumference across sex, ethnicity, and race. Després et al. demonstrated a strong correlation between BMI and waist circumference (r = 0.91) which is comparable to this study (r = 0.93). However Després et al. does highlight that BMI does not take into consideration actual body composition. This may account for large variances of girth measurements in large epidemiological samples weakening the clinical interchangeability of BMI and waist circumference. 8 Camhi et al. found that waist circumference and BMI were more tightly associated with actual fat mass and subcutaneous adipose tissue rather than visceral adipose tissue. 27 Adiposity has been identified as a strong predictor of MetS and a strong contributor to BMI and waist circumference. However, there is limited scientific evidence investigating the relationship of segmental body fat compartmentalization, or where the fat is, and MetS risk. It is important to note that although measuring waist circumference is a crucial first step, the ability of simple anthropometrics alone to both identify visceral adiposity and the presence of the MetS is limited. 8 Further anthropometric measures beyond waist circumference and BMI have promise in being cost-effective and clinically useful for detecting cardio-metabolic disease risk and 30, 31 establishing need for further assessment. For future screenings to be cost effective and less cumbersome to collect, simple point-of-care techniques need to be adopted. In addition to waist circumference, FPG and 38