ASSOCIATION BETWEEN THYROID FUNCTION AND OBESITY IN EUTHYROID ADULTS

_______________

A Thesis Presented to the Faculty of San Diego State University _______________

In Partial Fulfillment of the Requirements for the Degree Master of Public Health _______________

by Dayana Chanson Spring 2011

iii

Copyright © 2010 by Dayana Chanson All Rights Reserved

iv

DEDICATION To my mother, who not only gave me life, but also instilled in me the love for learning. To my husband, who has supported me and believed in me every step of the way, and to my children who are the love of my life and my reason to live.

v

He who has health, has hope; and he who has hope, has everything. -Owen Arthur

vi

ABSTRACT OF THE THESIS Association between Thyroid Function and Obesity in Euthyroid Adults by Dayana Chanson Master of Public Health San Diego State University, 2010 The obesity epidemic has reached alarming numbers, with about two thirds of U.S. adults being overweight or obese. The problem is multi-faceted and requires the understanding of different areas in order to effectively address the growing obesogenic population and begin to reverse the current trends. The present study is cross-sectional in design and aimed at understanding the association between thyroid function and obesity in individuals with normal thyroid function. A second aim of the study was to look at the association between iodine intake and BMI. The study evaluated data for 1692 euthyroid participants between the ages of 20 years and 49 years old. The study was conducted using data from the National Health and Nutrition Examination Survey for 2007-2008. Polychotomous logistic regressions were employed for multivariate analysis. All analyses were weighted and survey procedures were used to ensure accurate analysis given the complex study design of the data. The outcome of interest was BMI, which was categorized into the traditional cutpoints of normal, overweight and obese. The two main risk factors were serum thyroid-stimulating hormone levels (TSH) and urinary iodine concentration. Findings showed a positive association between TSH levels and BMI. The relationship between TSH and BMI was different among racial groups and among different smoking status categories. Final analysis was stratified by racial and smoking status categories. Findings revealed a significant positive association between TSH levels and BMI among non-Hispanic Whites, Hispanic/Latinos, non-Hispanic Blacks and non-smokers. These associations were significant for those being obese compared to those being normoweight. Noteworthy was the finding among Hispanic/Latinos being overweight relative to being normoweight. The association for these participants was negative, suggesting a decrease in odds of being overweight compared to normoweight for every 1-unit increase in TSH levels. Iodine was also found to be associated with BMI after adjusting for TSH and other factors. A significant negative association was found among Hispanic/Latinos, non-Hispanic Blacks and non-smokers only. Among Hispanic/Latinos and non-smokers, a significant association was found only when comparing obese and normoweight and only for iodine insufficiency relative to iodine sufficiency. Among non-Hispanic Blacks, an association was found when comparing overweight and normoweight for iodine insufficiency and excess compared to iodine sufficiency; and when comparing obese and normoweight for iodine insufficiency and excess relative to iodine sufficiency.

vii

TABLE OF CONTENTS PAGE ABSTRACT ............................................................................................................................. vi LIST OF TABLES ................................................................................................................... ix LIST OF FIGURES ...................................................................................................................x GLOSSARY ............................................................................................................................ xi ACKNOWLEDGEMENTS .................................................................................................... xii CHAPTER 1

INTRODUCTION .........................................................................................................1  Background ..............................................................................................................1  Statement of the Problem .........................................................................................2  Purpose of the Study ................................................................................................2  Theoretical Basis......................................................................................................3  Limitations of the Study...........................................................................................3 

2

LITERATURE REVIEW ..............................................................................................5  The Thyroid Gland and its Hormones .....................................................................5  Thyroid Function .....................................................................................................7  Factors that Affect Thyroid Function ......................................................................9  Iodine .................................................................................................................9  Gender, Age and Race .....................................................................................10  Smoking ...........................................................................................................12  Obesity in US .........................................................................................................13  Thyroid Function and Weight ................................................................................13 

viii 3

METHODS ..................................................................................................................18  Study Design ..........................................................................................................18  Study Population ....................................................................................................18  Data Collection ......................................................................................................19  Instruments .............................................................................................................20  Variables ................................................................................................................21  Outcome ...........................................................................................................21  Risk Factors .....................................................................................................21  Covariates ........................................................................................................23  Statistical Analysis .................................................................................................26 

4

RESULTS ....................................................................................................................29  Descriptive Statistics..............................................................................................29  Bivariate Analysis ..................................................................................................30  Multivariate Analysis .............................................................................................33  Multivariate Analysis by Racial Category .......................................................34  Multivariate Analysis by Smoking Status Category ........................................41 

5

DISCUSSION ..............................................................................................................44  Key Findings ..........................................................................................................44  Strengths ................................................................................................................46  Limitations .............................................................................................................47  Implications and Future Direction .........................................................................48  Conclusion .............................................................................................................49 

REFERENCES ........................................................................................................................50 APPENDIX FLOWCHART OF STUDY SAMPLE .......................................................................55

ix

LIST OF TABLES PAGE Table 1. Characteristics of the Cohort by BMI Level (Weighted Analysis) ...........................31  Table 2. Bivariate Polychotomous Logistic Regression Analysis of Association between BMI levels and Selected Characteristics in the Study Cohort (Weighted Analysis) ....................................................................................................32  Table 3. Multivariate Polychotomous Logistic Regression Model to Examine the Association of BMI Levels with Selected Characteristics Among NonHispanic Whites (n=682) (Weighted Analysis) ...........................................................35  Table 4. Multivariate Polychotomous Logistic Regression Model to Examine the Association of BMI Levels with Selected Characteristics among Hispanic/Latinos (n=510) (Weighted Analysis) ..........................................................36  Table 5. Multivariate Polychotomous Logistic Regression Model to Examine the Association of BMI Levels with Selected Characteristics Among NonHispanic Blacks (n=308) (Weighted Analysis) ...........................................................37  Table 6. Multivariate Polychotomous Logistic Regression Model to Examine the Association of BMI Levels with Selected Characteristics Among Smokers (n=572) (Weighted Analysis) ......................................................................................38  Table 7. Multivariate Polychotomous Logistic Regression Model to Examine the Association of BMI Levels with Selected Characteristics Among NonSmokers (n=928) (Weighted Analysis) .......................................................................39 

x

LIST OF FIGURES PAGE Figure 1. Hypothalamic-Pituitary-Thyroid axis. ........................................................................6  Figure 2. Thyroid hormone synthesis. .......................................................................................6 

xi

GLOSSARY 

Euthyroid- refers to an individual with normal thyroid function. Thyroid function is evaluated using serum TSH levels.



Free thyroxine (fT4) - refers to the small amount of unbound thyroid hormone. This hormone is deiodinized (an iodine atom is released) and turned into the active T3 form.



Free triiodothyronine (fT3) - refers to the small amount of unbound, bioactive thyroid hormone.



Hypothalamus-Pituitary-Thyroid axis- refers to the set of interactions (feedback system) between the hypothalamus, pituitary gland and thyroid gland that controls production and release of hormones such as TSH, T4 and T3.



Thyroid-releasing hormone (TRH) - refers to the hormone released by the hypothalamus in response to low thyroid hormone levels. This hormone stimulates the production and release of thyroid-stimulating hormone in the pituitary gland.



Thyroid-stimulating hormone (TSH) - refers to the hormone released by the pituitary gland in response to low levels of thyroid hormone. TSH stimulates the production and release of thyroid hormones.

xii

ACKNOWLEDGEMENTS I would like to thank my committee, Dr. Macera, Dr. Slymen, and Dr. Woodruff for their help throughout this process and during my time at San Diego State. Without the encouragement and guidance from Dr. Macera and Dr. Woodruff, and the expertise in statistics of Dr. Slymen I would not have been able to complete this project. I am forever grateful. I would also like to thank all my family and friends for their constant support and love. They believed in me when I lost hope. It was thanks to all these wonderful individuals that I kept going even when I felt overwhelmed and lost. May God bless each of you always. Thank you.

1

CHAPTER 1 INTRODUCTION BACKGROUND Obesity is a growing problem, not only in the United States, but globally. Obesity has increased over the last few decades and currently, the Centers for Disease Control and Prevention (CDC) estimates that about 68% of the US adult population is considered overweight or obese (Flegal, Carroll, Ogden, & Curtin, 2010). Currently, efforts to understand the multi-faceted problem are of top priority in public health. Many factors have been found to affect a person’s weight, including lifestyle choices like nutritional behavior and physical activity, as well as genetics and environmental factors—for example, where someone lives and food availability in the community. Likewise, risk factors associated with being overweight or obese have also been studied. Obesity increases the risk for cardiovascular disease, diabetes, cancers and metabolic syndrome, to name a few. It reduces the quality of life as well as overall life expectancy, and is responsible for a large portion of healthcare costs in America. It is important to understand all the risk factors that contribute to obesity, as well as understand all the co-morbidities that arise with being overweight and obese since the majority of the population suffers from the disease. The study of hormones, especially thyroid hormones and their association with obesity has been well documented in individuals with thyroid disorders. People who have thyroid dysfunction have been found to also suffer from weight problems (Bunevicius, Peceliuniene, Mickuviene, Girdler, & Bunevicius, 2008). Overt hypothyroidism is known to be related to weight gain, while overt hyperthyroidism leads to weight loss. Studies on

2 euthyroid, or people with normal thyroid function, are scarce. However, most studies that have been done on euthyroid individuals have shown that there is a significant association between BMI and thyroid function. More studies are needed to fully understand the extent of the association and translate the findings into practical use in the clinical setting.

STATEMENT OF THE PROBLEM More studies are needed that focus on euthyroid subjects. The present study aims to look at the association between thyroid function, as indicated by thyroid-stimulating hormone (TSH) and urinary iodine concentration, and body mass index (BMI). The study will add to the knowledge base on euthyroid individuals, obesity, and thyroid function. The primary goal is to investigate if TSH levels significantly impacts a person’s weight. Because thyroid hormone synthesis is dependent on iodine levels in the body, a secondary goal is to investigate if urinary iodine concentration significantly impacts a person’s weight. This study explores these relationships for both males and females combined, as the analysis did not point to gender as an effect modifier. The hypotheses for the current study are: 

Hypothesis #1: An increase in thyroid-stimulating hormone (TSH), within the normal range, affects weight.



Hypothesis #2: An association exists between iodine levels in the body and weight.

PURPOSE OF THE STUDY Obesity is a growing problem that needs to be well understood in order to control it and stop the epidemic. Many aspects of lifestyle, genetics, and the environment play a role in determining a person’s weight. In order to effectively reduce the prevalence of obesity and reverse the current trends, it is important to have enough information and knowledge about all the aspects that affect weight. Hormones control bodily functions such as metabolism and

3 cell growth and development, which directly affect how the body uses energy. This link makes hormones a very important aspect of weight management. In this manner, the present study adds to the knowledge base of the problem and provides some insight into one of the many facets that affect obesity. The present study not only focuses on a major current public health problem, but it looks at an aspect that individuals do not have much control over. Understanding the role of thyroid hormones and obesity is not a new focus; however, looking at the impact within normal hormone levels is a relatively new approach.

THEORETICAL BASIS The present study uses thyroid-stimulating hormone (henceforth TSH) as an indicator of thyroid function and status. Because of the mechanism for thyroid hormone synthesis and the role of TSH in it, the measurement of TSH levels is a good indicator of thyroid function (American Thyroid Association, 2008c). The Centers for Disease Control and Prevention uses three cutoffs for BMI. These cutoff levels are a valid and accurate way of measuring the weight status of the average individual (Centers for Disease Control and Prevention, 2009). In large population studies such as the one being presented here, traditional BMI cutoffs of normal, overweight and obese are used.

LIMITATIONS OF THE STUDY While the present study is large in sample size, the design is cross-sectional. This in itself limits the information that can be extracted from the analysis. Due to the study design, causation cannot be established. Additionally, some of the variables that were originally of interest due to their relationship to the risk factors were excluded because of a large number

4 of missing data and lack of variability. These variables included diabetes diagnosis, coronary heart disease diagnosis, and alcohol consumption.

5

CHAPTER 2 LITERATURE REVIEW THE THYROID GLAND AND ITS HORMONES The thyroid gland, located below the larynx, is the largest endocrine organ, and is essential to mammalian life. The thyroid gland produces hormones, which help in the regulation of metabolism, growth and development, and even reproduction (Watson & Miller, 2004). To maintain normal thyroid function, the hypothalamus and the pituitary gland both impact thyroid status. Known as the hypothalamic-pituitary-thyroid axis, this link between the three is what regulates proper thyroid hormone synthesis and release. The system is a negative-feed process in which the level of circulating thyroid hormone (both T4 and T3) signals the hypothalamus to synthesize and release thyroid-releasing hormone (TRH) or suppress it. The level of thyroid-releasing hormone in turn balances the release of TSH. When more TSH is needed, thyroid-releasing hormone binds to receptors in the pituitary gland that cause the release of TSH. The release of TSH causes the thyroid to stimulate the expression of: (1) the sodium/iodide symporter (NIS), which is responsible for iodide transport and therefore iodine uptake, (2) the enzyme thyroid peroxidase, and (3) thyroglobulin. In addition, TSH also helps in the generation of H2O2. All of these uses of TSH cause more thyroid hormones to be produced and released (Miot, Dupuy, Dumont, Rousset, 2010). Less TSH causes less thyroid hormone to be produced and released from the thyroid gland, causing TRH suppression. Figure 1 illustrates the hypothalamic-pituitarythyroid axis negative feedback process.

6

Figure 1. Hypothalamic-Pituitary-Thyroid axis. Within the thyroid gland, thyroid hormone synthesis requires iodine and the enzyme thyroid peroxidase to turn thyroglobulin (Tg) into thyroxine and triiodothyronine, T4 and T3, respectively. Thyroxine and triiodothyronine are then released into the bloodstream, where they are part of protein synthesis and metabolic processes in a multitude of cells and tissues. Figure 2 illustrates the process.

Figure 2. Thyroid hormone synthesis. The regulation of thyroid hormones is a complex process. In a person with healthy thyroid function, the presence of excess or lack of iodine, for example, leads the thyroid gland to make and release more or less of its hormones. The level of hormone in circulation signals the suppression or the synthesis of other hormones like TRH and TSH that in turn help maintain thyroid hormone balance.

7 Since thyroid hormones are not water-soluble they bind to proteins when they are released from the thyroid gland. The majority travels in the bloodstream bound to the proteins thyroxine-binding globulin (TBG), thyroxine-binding prealbumin (TBPA) and albumin. Free thyroxine, fT4, is the small amount (about 0.03%) of thyroid hormone that is not bound to a protein (Greer, 1990). Similarly, fT3 is the triiodothyronine that is not bound to a protein. While there is more thyroxine in the thyroid gland and circulating in the body, triiodothyronine (T3) is the bioactive form of the two. In addition to being released from the thyroid gland, T3 and 3, 3’, 5’-triiodothyronine (rT3) are also formed from thyroxine locally in the cells by deiodination. Greer (1990) states that in individuals with healthy thyroid function 20% of T3 is produced in the thyroid gland, while 80% is formed from thyroxine peripherally (p. 236).

THYROID FUNCTION The term euthyroid is used to refer to individuals with normal thyroid function. Different studies use some combination of the following criteria to decide who is considered euthyroid. Most studies will primarily check for thyroid autoantibodies, TSH levels, fT4 levels and previous thyroid disease as elements to determine thyroid status (Bastemir, Akin, Alkis, & Kaptanoglu, 2007; Fox, et al., 2008; Iacobellis, Ribaudo, Zappaterreno, Iannucci, & Leonetti, 2005). If all information is readily available to the investigators, a person can be determined euthyroid if: (1) they do not have a previous or current diagnosis of thyroid disorders or dysfunction, (2) are not taking thyroid hormone therapy or medications for thyroid diseases, (3) do not have positive thyroid autoantibodies (anti-Tg and TPOAb) in the serum, and (4) have TSH levels, fT4 levels and fT3 levels within the reference ranges.

8 Abnormal thyroid function, which leads to the body making too little or too much hormone, can be due to different issues. Autoimmune thyroid disorders (AITD) are largely attributed to genetics (Hadj-Kacem et al., 2009), while other forms of hypothyroidism and hyperthyroidism can be iodine-induced (Camargo et al., 2008) or caused by medications that interfere with thyroid function. Hypothyroidism occurs when the thyroid gland does not make enough thyroid hormones. It can be detected through serum TSH tests that show levels of TSH above the reference range. The most severe form of hypothyroidism is myxedema, where the body starts to shut down. This form of hypothyroidism may occur after many years of a person suffering from the dysfunction (American Thyroid Association, 2008c). On the other hand, hyperthyroidism occurs when the body makes too much thyroid hormone and the thyroid gland is said to be “overactive”. The dysfunction can be diagnosed through serum TSH, T4, and T3 tests, which reveal low levels of TSH and higher levels of T4 and T3 than the reference ranges. Besides autoimmune disease, hyperthyroidism can be caused, temporarily, by lower overall immune response (e.g. when a person is fighting a viral infection), the individual is suffering from thyroid inflammation (thyroiditis), or due to a thyroid nodule or goiter forming (American Thyroid Association, 2008b). In addition to these diseases, an individual can develop thyroid cancer. Thyroid nodules can be either benign or cancerous. A thyroid ultrasound and biopsy is used to diagnose cancer, and a sensitive test is used to check for thyroid carcinoma reoccurrence, where serum thyroglobulin (Tg) levels are examined. While thyroid cancer is a common endocrine-related cancer, it is rare that a person is diagnosed as having this type of carcinoma. According to the American Thyroid Association (2008a), the incidence of thyroid cancer in the United States is 20,000 cases per

9 year. Thyroid carcinoma can be managed through surgical removal of the thyroid and continued use of thyroid hormone therapy.

FACTORS THAT AFFECT THYROID FUNCTION Different factors affect thyroid function. The present study was able to include urinary iodine concentration, gender, age, smoking and race as variables in the analysis due to their known potential to be confounders when studying thyroid function. Additionally, education, physical activity level, daily caloric intake and poverty status were included because they are known risk factors for BMI. While other known risk factors of thyroid function are known, these are not included in the study due to the unavailability of data.

Iodine Studies have shown that iodine is a vital component of thyroid hormone synthesis, and therefore is of critical importance in thyroid function (Zimmermann, 2010). Without the proper balance of iodine intake, thyroid function suffers greatly. Iodine is needed in thyroid hormone synthesis to turn thyroglobulin into thyroxine or triiodothyronine, T4 or T3, respectively. Too little iodine will cause the thyroid gland to produce less hormone, while too much will lead to an excess of hormone. Iodine deficiency is still a big problem across the world (Zimmermann, Jooste, & Pandav, 2008). The effects range from different size goiters to the very serious complication called cretinism. Since the 1990’s, the U.S. has been considered iodine sufficient. However, excess iodine intake is of concern because it may lead to thyroid dysfunction. The effects of excess iodine intake range from hyperthyroidism to hypothyroidism and chronic autoimmune thyroiditis (CAT) (Camargo et al., 2008). Research shows that in populations where there is iodine deficiency or excess, overt hyperthyroidism and hypothyroidism exist (Laurberg et al., 2006). For example, in areas of

10 iodine deficiency such as in Aalborg, Denmark where there is a moderate iodine deficiency, and in Copenhagen, Denmark where there is a mild iodine deficiency, Pedersen et al. (2002), found that incidence rates of overt hyperthyroidism were much higher in Aalborg than those in Copenhagen. The reverse was also true, where more cases of overt hypothyroidism existed in Copenhagen than in Aalborg (Pedersen et al., 2002). Some studies have found iodine intake to be directly related to serum TSH levels (Alsayed et al., 2008; Camargo et al., 2008; Shan et al., 2005). Laurberg et al. (1998), found that serum TSH levels were low in Jutland, Denmark, an area known to have low iodine intake levels, whereas Iceland, an area known to have high iodine intake levels had high TSH values. Similarly, Vejbjerg et al. (2009) also found higher median serum TSH levels after iodine intake increased in the population due to mandatory salt iodization. While iodine deficiency causes more damage than iodine excess, both extremes are detrimental to a population’s health. In studies of thyroid hormones and obesity, iodine may be a key component in understanding this relationship.

Gender, Age and Race It has been well documented that a difference in thyroid function/dysfunction exists between males and females. Blount, Pirkle, Osterloh, Valentin-Blasini, & Caldwell (2006) noted that women were different from men in terms of thyroid function due to the effect of pregnancy and estrogen on thyroid function, as well as their susceptibility to autoimmune thyroid disease. Higher prevalence rates of thyroid disease have been found in females, especially in premenopausal women (Bunevicius et al., 2008). In their study using data from the National Health and Nutrition Examination Survey III (NHANES III: 1988-1994), Hollowell et al. (2002) reported a significantly higher percent of females with TSH levels >

11 4.5 mIU/liter than males in both the total population studied and in the subgroup of diseasefree individuals. Additionally, all populations showed that there was a higher percentage of females with TSH levels < 0.4 mIU/liter, as well (Hollowell et al., 2002). While the present study included gender as a covariate in the analysis, the variable was not found to be an effect modifier of TSH levels and BMI. This may be because the population being studied had normal TSH levels, where differences between the genders have been found when looking at thyroid dysfunction. This may also be due to the fact that the present study evaluated adults between 20 and 49 years of age, which, for the most part may not include a high percent of post-menopausal women. Shon, Jung, Kim, & Lee (2008) also found a difference between males and females when looking at TSH levels and obesity. Their findings were in agreement with other studies that showed higher levels of TSH in euthyroid women compared to euthyroid men. Age has also been shown to affect thyroid function. As people age, thyroid hormone synthesis is impacted. While studies like the one done by Hollowell et al. (2002) have shown that TSH levels increase with age when iodine intake is sufficient in a population, findings are contradicting. For example, Laurberg et al. (2006) found that in the first cohort, TSH levels decreased with age, but the association was not seen when participants with thyroid abnormalities were excluded. A third characteristic that has been documented as a modifier of TSH levels is race. Studies have shown that different racial groups show different levels of serum TSH. Hollowell et al. (2002) found that more whites than blacks had TSH levels above the normal range, while more blacks than whites showed TSH levels below the normal range. A significant association between race and TSH levels was found among 809 patients (with

12 TSH levels between 0-5mU/L) in a study conducted in 1991, after adjusting for age, sex and chronic medical problems and/or medication use (Schectman, Kallenberg, Hirsch, & Shumacher, 1991). Findings showed that whites were more likely to have higher levels of TSH as compared to blacks. The study also pointed out that it was in agreement with other studies that had been previously published. Moreover, a current study done among a population in Brazil demonstrated that prevalence of hypothyroidism was higher among white participants as compared to mulatto and black participants (Sichieri et al., 2007). Results showed that on average, TSH values were 22% lower for blacks than for whites (finding was statistically significant with a p=001). The current study found that the relationship between serum TSH levels and BMI was different for different racial categories.

Smoking The literature points to a pattern showing TSH levels being lower in smokers compared to non-smokers (Belin, Astor, Powe, & Ladenson, 2004). Additionally, smoking cessation has been associated with an increase in weight. Furthermore, smoking could interfere with iodine intake by way of thiocyanate inhibition of the NIS or sodium/iodide symporter, which is the protein responsible for iodide transport by the thyroid (this topic is beyond the scope of the present study) (Laurberg et al., 2006; Miot et al., 2010). Smoking status is therefore seen as an effect modifier or confounder of the relationship between BMI and TSH levels (Blount et al., 2006; De Pergola, Ciampolillo, Paolotti, Trerotoli, & Giorgino, 2007; Nyrnes, Jorde, & Sundsfjord, 2006). Thus, some studies have stratified the data by smoking status, have only included those who are non-smokers or have adjusted for smoking status in their analysis. The study by Nyrnes et al. (2006) found that in non-smokers, BMI and TSH levels were positively associated. Furthermore, the follow up cohort also showed

13 that among non-smokers only, an association existed between a change in TSH and a change in BMI. Investigators of this study suggest that smoking may mask the relationship between BMI and TSH levels. While Makepeace et al. (2008) did not find a relationship between TSH and BMI, the study did find that TSH levels were significantly lower in current smokers than non-smokers or former smokers.

OBESITY IN US The obesity epidemic we are facing as a nation is alarming. Currently, estimates show that about 68% of adults in the United States are overweight and obese (Flegal et al., 2010). The problems related with obesity have been well documented in numerous studies. Many studies have looked at the impact of environmental factors on weight, while others have looked at the genetic aspect and psychological and/or behavioral aspects related to weight. A third focus of research in obesity is the hormonal aspect that contributes to body weight. The area of hormonal influence, specifically thyroid hormones, on weight is the focus of the current study.

THYROID FUNCTION AND WEIGHT There are different ways that thyroid hormones have been said to affect weight. First, they are thought of as hormones that contribute to energy expenditure and thermogenesis. Additionally, some researchers have suggested that increased levels of TSH are an adaptive mechanism due to increase adiposity (Chikunguwo et al., 2007). Others have suggested that being obese may lead to thyroid hormone resistance in peripheral metabolism. These mechanisms are beyond the scope of this study; however a brief explanation is provided regarding the major role of thyroid hormones in resting energy expenditure (REE).

14 Thyroid hormones are critical to energy expenditure and thermogenesis. Triiodothyronine (T3) and thyroxine (T4) deiodinized to T3 are both used by cells to increase the metabolic rate known as adaptive thermogenesis. This complex mechanism—which has been documented by others—allows for humans and other mammals to increase the metabolic rate, thus creating more heat to maintain proper energy homeostasis (Bianco, Maia, Da Silva, & Christoffolete, 2005). In human adults, skeletal muscle provides a place for thermogenesis to occur. Thyroid hormones have been found to play an important role in energy expenditure in these cells. They are part of complex pathways and processes that lead to increasing ATP and energy expenditure. A decrease in thyroid hormone would therefore lead to slower metabolic rate. This is seen in the condition known as hypothyroidism. The opposite is true for people with the condition known as hyperthyroidism, where metabolic rate increases and loss of body fat is seen. Al-Adsani, Hoffer, & Silva (1997) found significant changes in resting energy expenditure (REE) when there were small changes in T4 dosage of patients who were on chronic thyroid hormone replacement. They used TSH as the indicator for thyroid status since TSH “reflects an end effect of thyroid hormone” and showed how REE decreased as TSH levels increased. The investigators noted that this kind of change found in REE could account for weight gain over a 5-10 year time span. Significant findings between REE and TSH levels further confirm that TSH levels are an important factor to consider when looking at weight and obesity in our population. This is of high importance considering that these changes in REE are seen when changes happen within the normal range of TSH levels. Thyroid hormones have been found to affect weight, even when hormone levels are within the normal range. Knudsen et al. (2005) found a significant positive association

15 between TSH levels and BMI. Other investigators have also found positive associations between TSH levels and obesity in their studies. For example, a recent study among euthyroid individuals showed that TSH levels were significantly higher among obese participants compared to lean participants (Nannipieri et al., 2009). Bastemir et al. (2007) also found that the degree of obesity as measured by BMI was associated with higher TSH levels. While Rotondi et al. (2009) did not find a significant association between BMI and TSH in euthyroid morbidly obese patients compared to euthyroid normal weight (henceforth “normoweight”) patients, findings did show that TSH levels were higher in the morbidly obese group relative to the normoweight group. Furthermore, Iacobellis et al. (2005) also found that in euthyroid individuals, TSH levels and BMI were positively correlated. Additionally, the study also found that after adjusting for BMI, TSH levels and Leptin levels were also correlated. They suggested that TSH could therefore be an early marker for energy balance in individuals with severe obesity (Iacobellis et al., 2005). Studies have also found significant associations between thyroid hormones and other obesity-related measures such as waist circumference and weight in kilograms. De Pergola et al. (2007) found that TSH levels were positively correlated with waist circumference in a study of 201 overweight and obese women who were euthyroid. Interestingly, this study also found TSH to be negatively correlated with age, something that other studies have found to be a positive association (Blount et al., 2006). Similarly, Fox et al. (2008) also found a positive association between TSH and weight in kilograms in a cohort of euthyroid individuals. The study was able to assess the association over time given that it was a prospective design. Findings showed that over time, when TSH levels increased, weight also increased (Fox et al., 2008). In the same manner, the study by Sari, Balci, Altunbas, &

16 Karayalcin (2003) also showed an association between TSH and weight. As with the study by Fox et al. (2008), the prospective design of the study by Sari et al. (2003) allowed for evaluation of change over time. The study found that those who lost more than 10 percent body weight showed lowered levels of TSH at follow up. A few studies did result in no association between TSH levels and BMI (Makepeace et al., 2008; Manji et al., 2006; Shon et al., 2008). However, the study by Makepeace et al. (2008) did find that levels of TSH were lower among smokers. Their findings on lower TSH levels among smokers were consistent with the literature. The study by Manji et al. (2006) had a sample size of 401 euthyroid men and women; however, a potential limitation of the study was that the participants had been referred to the clinic for thyroid nodules or goiter. The investigators classified them as euthyroid because their serum TSH levels were within the normal range. Nevertheless, most other studies exclude subjects with confirmed goiters or thyroid nodules when looking at euthyroid individuals, since these usually occur when thyroid dysfunction is present or iodine levels are not sufficient. Lastly, the study by Shon et al. (2008) did acknowledge that their findings with regards to TSH and BMI did not agree with findings of other studies. However, they found that TSH levels were higher in euthyroid women than in men, which is consistent with the literature. Despite the scarcity of studies about thyroid function and obesity in euthyroid individuals, the recent literature does point to a pattern that suggests thyroid hormones, such as TSH levels within the normal range, affect, to an extent, the weight status of individuals. Investigators need to study this area more closely to gain better understanding of the role these hormones play even when individuals are deemed healthy. While many more studies are needed to confirm this—especially prospective designs—the literature thus far points in

17 this direction. The present study adds to the body of knowledge in the topic and uses data from a large study population the represents individuals across the nation. As more studies are conducted in the area of thyroid hormones and obesity in euthyroid individuals, there is hope to fully understand another potential facet in the complex chronic disease that is obesity. The knowledge gained will better prepare public health officials and other health providers in dealing with a growing epidemic in the United States as well as other parts of the world.

18

CHAPTER 3 METHODS STUDY DESIGN The current study is of a cross-sectional design, done using national data from the National Health and Nutrition Examination Survey (NHANES) 2007-2008. The National Health and Nutrition Examination Survey (henceforth NHANES) is continuously conducted by the National Center for Health Statistics (NCHS), which is part of the Centers for Disease Control and Prevention (CDC). It is a population-based survey that collects health and nutrition information about the United States civilian, non-institutionalized population. The survey sample is representative of the larger population and is used to understand the current health trends of Americans and also to influence national health policies.

STUDY POPULATION NHANES is a complex, stratified, multi-stage probability sample, representative of the U.S. civilian, non-institutionalized population. Subjects for the current study included adult males and females of age 20 years to 49 years old. Participants completed both the interview and the examination portions of the survey. To be included in the study, participants had to: (1) be within the age range of 20-49 years old at the time of the examination part of the survey, (2) females included in the study were not pregnant (this was assessed by self-report and a urine pregnancy test), (3) have serum thyroid autoantibodies (both TgAb and TPO) within the normal value range, (4) have thyroglobulin (Tg) levels within the normal value range, (5) have a self-report of no thyroid cancer or other thyroid problems diagnosis, (6) have serum TSH levels within the normal range of 0.3-3.0 mcIU/mL

19 (this study used the narrower range recommended by the American Association of Clinical Endocrinologists in 2003), and (7) have reported being from one of the following four race categories: Hispanic/Latino, Mexican American, non-Hispanic White or non-Hispanic Black. Original data included 10149 participants. After removal of all subjects who did not meet the inclusion criteria described, this study had data from 1692 subjects (see Appendix for flowchart of study sample).

DATA COLLECTION This study is considered a secondary analysis, where data was readily available and the investigator did not collect it. Data collection involved complex probability sampling. The most current analytical guidelines published by the CDC state that the sampling frame was all counties in the U.S. Primary Sampling Units were selected and from there, a selection of clusters of households was made. From these households, a person was screened for demographic characteristics and one or more people from the household were chosen to participate. For the NHANES 2007-2008 cycle, methodology was changed from previous years. New methods for sampling include oversampling all Hispanics, not just Mexicans (National Center for Health Statistics, 2005). The survey consisted of questionnaires done at home and in mobile units (MEC). Additionally, subjects were given an examination, which included physiological measurements and serum and urine samples. Interviews done in person used the computerassisted personal interview (CAPI) software that has automatic edits and checks. These edits and checks alert the interviewer of potential errors or inconsistencies in the responses. Furthermore, the NHANES field office staff checked the interviews for completeness and accuracy. A highly trained medical team conducted all medical exams in the mobile units and

20 all laboratory tests were done using the most current and accurate methods and equipment available. Interviewers went through a two-week, intensive training prior to administering the survey to participants. A large number of interviewers were bilingual in English and Spanish, and most had previous interviewer experience. To ensure that high quality data was obtained, NCHS and contractor staff closely monitored interviewers. The laboratory staff also received extensive training in standardized procedures. All phlebotomists and medical technologists were certified by organizations such as the American Society for Clinical Pathologists. Staff from NCHS performed unscheduled visits to the mobile units and contract laboratories repeated tests on five percent of specimens to ensure accurate testing (National Center for Health Statistics, 2007).

INSTRUMENTS Instruments used included health and nutrition questionnaires administered at home and in mobile units, laboratory serum tests and urine sample tests. Data was electronically recorded and sent to a central survey database system. Electronic questionnaire forms administered at home were programmed in Blaise©. Questionnaires administered in mobile units had some non- Blaise© formatting included. To ensure minimal errors, the programs had built-in edits and checks. Laboratory data was also automated. Collection materials were all checked to ensure that they were free of contaminants. All procedures involved data checks for quality control. Instruments were translated into Spanish and staff was trained to handle questions and interactions with respondents in both English and Spanish (National Center for Health Statistics, 2007).

21

VARIABLES The current study included a total of ten independent variables and one categorical outcome. All independent variables, except for TSH levels, were categorical in nature.

Outcome The outcome of interest in the present study was body mass index (BMI). BMI was used as a categorical variable with 3 categories: underweight or normal range (BMI < 25.0 kg/m2), overweight (BMI 25 to 29.99 kg/m2) and obese (BMI ≥ 30.0 kg/m2). Although at first the BMI variable was to be divided into more than three categories to better understand TSH levels among different levels of obesity (Rotondi et al., 2009), after looking at the distribution of subjects along the multiple categories it became clear that there were not enough participants in the higher obese classes to gain statistical significance and power to interpret results. Therefore, the decision was made to use traditional BMI cutoff points to categorize participants within this variable.

Risk Factors The main risk factor studied was thyroid-stimulating hormone (TSH). TSH was chosen as an indicator for thyroid status based on the literature indicating that levels of TSH directly correlate with the function of the thyroid gland and its production and secretion of thyroxine and triiodothyronine into the bloodstream (American Association of Clinical Endocrinologists, 2004). TSH levels were treated as a continuous variable in the model, and the sample mean of TSH was the measurement that was reported since no outliers existed in the data. While the normal range for the test was 0.34-5.6 mcIU/mL, for the purpose of this study, the normal range was set at 0.3-3.0 mcIU/mL. This narrower range was due to the new

22 2003 American Association of Clinical Endocrinologists (AACE) (2003) recommendations. Serum TSH was assessed by using the Access HYPERsensitive human thyroid-stimulating hormone (hTSH) assay, which is a 3rd generation assay. The Access HYPERsensitive hTSH assay was used because it has virtually no cross-reactivity with other peptide hormones, and this allows for better sensitivity and specificity of the test, which translates into better detection of hyperthyroid and euthyroid subjects (National Center for Health Statistics, 2009a). To understand how iodine intake may be related to BMI (with serum TSH in the model due to their relationship), urinary iodine concentration was the second risk factor of interest. The original variable was continuous and was recoded into a categorical variable for the present study. Iodine intake was categorized into four groups: (1) insufficient (less than 100 ug/L), (2) sufficient (100-1199 ug/L), (3) above recommended (200-299ug/L), and (4) excess (300 ug/L or above). The “sufficient” category was the reference category for all analysis. Urinary iodine concentration has been considered the best way to understand iodine intake in a population and it is the preferred method to use. The method is both accurate and rapid (National Center for Health Statistics, 2009b). Urinary iodine concentration was measured using ICP-DRC-MS (Inductively Coupled Plasma Dynamic Reaction Cell Mass Spectroscopy). Documentation for the NHANES 2007-2008 urinary iodine laboratory test states that: Iodine (isotope mass 127)… [was] measured in urine by ICP-DRC-MS using 100% argon as the Dynamic Reaction Cell TM (DRC) gas utilizing collisional focusing. Urine samples are diluted 1+1+8 (sample + water+ diluents) with water and diluents containing tellurium and bismuth for internal standardization. (National Center for Health Statistics, 2009b)

23

Covariates For the current study, gender, age, education, race, daily caloric intake, physical activity level, poverty status, and smoking status were the covariate variables. They were included in the analysis because they have been found to be related to BMI and/or modify or confound the relationship between the main risk factor (TSH) and the outcome (BMI). All covariates for the present study were of a categorical nature. Gender was a dichotomous variable with males as the category of reference for all analysis. Originally, age was not categorical, but was categorized in order to better understand the differences between age ranges. Age was categorized into three groups: (1) 20-29 year olds (reference category), (2) 30-39 year olds, and (3) 40-49 year olds. Education was recoded from its original groups into three groups: (1) up to high school diploma (those individuals who completed any level of school up to getting a high school diploma or GED), (2) some college/AA (those who completed some college courses or up to an Associate in Arts degree), and (3) college graduate or above (those who completed a 4 year degree or more). Race was also recoded from its original groups. The Mexican American and Other Hispanic groups were recoded into one category as Hispanic/Latino, while Non-Hispanic White and Non-Hispanic Black were kept the same as the original categories. Due to the small number of subjects in the “other race/multi-racial” category (n=77), these individuals were excluded from further analysis. The present study therefore only looked at individuals who self-reported being Hispanic/Latino, non-Hispanic White or non-Hispanic Black. Daily caloric intake was assessed using only the Day 1 total calories variable. Persons with values on this variable all had reliable information, and therefore it was felt that this variable would accurately represent the individual’s daily caloric pattern. The variable was categorized into four groups: (1) below recommended (which were individuals who consume

24 less than 1800 calories/day), (2) recommended (those who consumed anywhere from 1800 to less than 3000 calories/day), (3) over the recommended (those who consumed 3000 to less than 3700 calories/day, and (4) well above the recommended (those who consumed 3700 calories or more in a day). These categories were created based on the recommendations given in the Dietary Guidelines for Americans, 2005. The guidelines provide a table for estimated calorie requirements for each gender and age group across three levels of physical activity (U.S. Department of Health and Human Services, 2005, p. 12). To create the groups in this study, the investigator looked at the lowest calorie recommendations for adults 19-50 years of age. The lowest recommended calorie intake was found to be that of the sedentary females of ages 31-50 years old (1800 calories). Therefore this was considered the low end of the recommended category. The highest recommended value was for active males of ages 1930 years old (3000 calories). This was then considered the upper limit for the recommended category. Anything below 1800 calories was considered to be below the recommended and anything above 3000 was considered above. However, due to the large range of values above the 3000-calorie limit, the above recommended was further divided into two categories. The cut off to include a fourth category was evaluated by checking the value where 10% of the population would be included in the highest category. This would ensure a large enough sample within the category to have an accurate analysis. Results indicated that an appropriate cut off would be 3700 calories. Physical activity level was also recoded from the original variables. Three original variables were used to determine the physical activity groups used in the current study. For the present study, physical activity was of a recreational nature only. Additionally, both vigorous as well as moderate recreational activities were considered. The first questions

25 asked were if a person did vigorous/moderate activities for more than 10 minutes (original variables were PAQ 650 and PAQ 665). Those who answered “no” were categorized as not doing any physical activity (this was the first category and the reference category for analysis in the present study). For individuals who answered that they did do vigorous/moderate activities for more than 10 minutes, two other sets of questions were used to determine the level of physical activity they did in a week (original variables were PAQ 655 and PAD 660, and PAQ 670 and PAD 675). If individuals answered that they did physical activity (either vigorous or moderate) one to 3 times a week for more than 30 minutes, they were placed in the second category “1-3 days of PA a week”. If participants answered that they did physical activity for 4 or more days a week and it was for 30 minutes or more, they were placed in the third category “4 or more days of PA a week”. Using all three questions and ensuring that participants were doing at least 30 minutes each time was seen as a way to accurately report recommended levels of physical activity. Poverty Status was recoded into a categorical variable from the original ratio of family income to poverty threshold variable. The variable was categorized into six categories: (1) below poverty level, (2) 100-199% poverty line, (3) 200-299% poverty line, (4) 300-399% poverty line, (5) 400-499% poverty line, and (6) 500% and above poverty line. Lastly, smoking status was a variable created using the original variable that measured serum cotinine levels. Cotinine is a metabolite of nicotine and is considered a biomarker for tobacco exposure. As Benowitz, Bernert, Caraballo, Holiday, & Wang (2009) stated, the cutoff between smokers and nonsmokers that has been mostly used is 14 ng/mL. However, their study, using NHANES data, revealed that a cutoff of 3.08 ng/mL for adults has both high sensitivity and specificity (96.3% and 97.4%, respectively). The original variable was of a

26 continuous nature and was recoded into a dichotomous variable. A cutoff of 3 ng/mL was used in the present study to classify smokers/heavily exposed to tobacco smoke and nonsmokers. Those with serum cotinine levels below 3 ng/mL were considered non-smokers and this group was the reference category for all analysis. While Benowitz et al. (2009) did evaluate the different smoking patterns between racial groups and found other cotinine cutoffs for different races, the present study used the overall 3 ng/mL cutoff across all races.

STATISTICAL ANALYSIS Analysis was performed using SAS® 9.2 Software. Because of the complex, stratified, multi-stage sampling of the survey used, SAS procedures that accommodate survey data with a such design were used. These included: PROC SURVEYMEANS, PROC SURVEYFREQ and PROC SURVEYLOGISTIC. In order to produce unbiased results, the sample weight provided by CDC was used. Therefore all results are weighted measures. Different data files from the NHANES 2007-2008 website were merged per CDC protocol, using a unique identifier common among all files. Missing data was dealt with as follows: (1) if participants had missing data for the outcome variable or TSH, then they were excluded from the study and (2) missing data for any of the other variables was kept as missing data and in the analyses that included these variables, these cases were excluded. Descriptive statistics (mean and standard deviation) were used for the serum TSH level variable, which was continuous. All other variables in the model were of a categorical nature and frequencies for each were reported. All reported percentages are weighted values and a significant alpha level of 0.05 was used for all analyses. For all analysis with categorical data the Rao-Scott Chi-square test was used due to the complex design of the sample. It allows for more accurate results when dealing with multistage stratification and

27 weighted study samples. While the continuous TSH variable was not normally distributed, a transformation was not performed because the sample size was large enough for the Central Limit Theorem to be applied to the situation. Contingency tables and Rao-Scott Chi-square tests were performed to evaluate the bivariate analysis between each categorical independent variable and the outcome variable (BMI). Analysis showed that not adjusting for other variables, gender, race, physical activity level, urinary iodine concentration and daily caloric intake were each significantly associated with BMI at the 0.05 alpha level. Three variables, which were kept in the final model, were not significantly associated with BMI in the initial bivariate analysis. These variables were education, smoking status and poverty status. Despite the fact that the variables mentioned were not statistically significant, they were kept in the model due to their known association with the outcome variable. In addition to the chisquare tests, simple logistic regressions were performed for each independent variable with the outcome variable. These regression analyses provided unadjusted odds ratios, which are reported on Table 2 (p. 32) in the Results chapter. Originally, for multivariate analysis, ordinal logistic regression was planned; however the proportional odds assumption test revealed a significant finding. A significant proportional odds assumption test indicated that the assumption of proportional odds had been violated and the odds for the different levels of the variable were not found to be equal. Because the odds were not equal, a different method had to be used. Polychotomous logistic regression was therefore used in the present study. Polychotomous logistic regression is a method that involves logits. Each logit produces an odds ratio, which compares the reference category, in this case normal BMI, to each of the other categories in the variable. Because the outcome variable for the present study consisted of three categories (normal, overweight and

28 obese), the two logits created by the regression were: (1) overweight compared to normoweight and (2) obese compared to normoweight. All possible interaction terms between TSH and each of the other independent variables were considered and models were run excluding each interaction term that was not significant in the previous model. The final model revealed two significant interaction terms: (1) TSH and race and (2) TSH and smoking status. Because interaction terms were found to be significant, and therefore effect modifiers of the BMI and TSH relationship, final analysis involved stratification by race and by smoking status. Results are presented for each racial group as well as for the two smoking categories.

29

CHAPTER 4 RESULTS DESCRIPTIVE STATISTICS Of 10149 subjects, a total of 1692 participants met the inclusion criteria and were part of the current study. These 1692 participants represented 74, 865,215 individuals in the U.S. non-institutionalized civilian population. All analyses and therefore reported percentages are weighted. Of the 1692 subjects, 908 were male and 784 were female (55% and 45%, respectively). To be included in the study, participants had to be 20 to 49 years of age, and roughly a third of participants were in each of the age categories—34% were 20-29 years old, 33% were 30-39 years old, and 33% were 40-49 years old. Approximately 69% of the sample identified themselves as Non-Hispanic White (n=757), while 18% of participants were Latino/Hispanic (n=583) and 13% were Non-Hispanic Black (n=352). Almost half of the sample (45%) had completed an education level up to a High School diploma, while 31% had attended some college or received an Associate of Arts degree (AA), and 24% were college graduates or higher. Data for the poverty status was only available for 1568 participants. Of these subjects, 17% (n=371) were identified as living below the poverty line. Using serum cotinine levels to assess smoking status, the study found that of 1690 participants with available data, 63% were considered non-smokers, while 37% were current smokers or people who were heavily exposed to tobacco smoke. Of the 1624 participants who had data available for urinary iodine concentration, 33% were considered to have sufficient iodine intake (100-199 ug/L). A third of these participants had insufficient iodine intake, which was below100 ug/L, and 37% had an excess of iodine intake (200 ug/L and

30 above). The majority of the study sample was considered sedentary (77%), while only 43% had a daily caloric intake that was considered to be within the recommended limits for the age group of the study. Table 1 shows demographics and clinical characteristics of the study sample. Sixty-four percent of the sample was overweight or obese, consistent with the current national estimates for obesity in the adult U.S. population. There were 599 participants who were overweight and 529 were obese (BMI of 25-29.9 and 30+, respectively). All participants in this study were considered to be euthyroid according to serum TSH levels. The sample’s average TSH level was 1.47mIU/L (±0.02). Contingency table results for TSH levels alone were higher in the overweight and obese BMI categories, suggesting an association between BMI and TSH levels within the normal range (p-value