USF Libraries
USF Digital Collections

The inflammatory consequences of stress and adiposity

MISSING IMAGE

Material Information

Title:
The inflammatory consequences of stress and adiposity
Physical Description:
Book
Language:
English
Creator:
Bykowski, Cathy A
Publisher:
University of South Florida
Place of Publication:
Tampa, Fla
Publication Date:

Subjects

Subjects / Keywords:
Socioeconomic status
Depression
Body mass index
Waist circumference
C-reactive protein
Dissertations, Academic -- Psychology -- Masters -- USF   ( lcsh )
Genre:
non-fiction   ( marcgt )

Notes

Abstract:
ABSTRACT: The inflammatory process is important in protecting the body against the invasion of pathogens, but recent research has suggested that a long-term inflammatory response may lead to chronic diseases (e.g., Black, 2003; Wu, Dorn, Donahue, Sempos, & Trevisan, 2002). Two factors that have been implicated in the inflammatory and disease processes are stress and obesity (Black, 2003). While their individual lines of research continue to grow, few researchers have attempted to integrate these factors into one model to explain their effects on inflammation. This study aimed to replicate previous findings suggesting relationships between stress, obesity and inflammation and test an integrated model of stress and obesity by examining a possible interaction between the effects of stress and obesity on inflammation. Socioeconomic Status (SES) and depression were employed to examine the association between stress and the inflammatory marker, c-reactive protein (CRP). The study utilized the data resulting from the National Health and Nutrition Examination Survey (NHANES; National Center for Health Statistics, 2006). Included in the dataset are 4998 adults (2416 males and 2582 females) ranging in age from 18 years to over 85 years (M = 47.13, SD = 20.86). A subsample (N = 589) completed the Major Depression module of the Composite International Diagnostic Interview (CDCI). The results indicate that body mass index, waist circumference, income, education, and depression symptoms significantly predict CRP. The data also suggest an interaction between the adiposity variables and the SES variables. This supports the hypothesis that the inflammatory effect of stress on an individual is moderated by adiposity.
Thesis:
Thesis (M.A.)--University of South Florida, 2008.
Bibliography:
Includes bibliographical references.
System Details:
Mode of access: World Wide Web.
System Details:
System requirements: World Wide Web browser and PDF reader.
Statement of Responsibility:
by Cathy A. Bykowski.
General Note:
Title from PDF of title page.
General Note:
Document formatted into pages; contains 107 pages.

Record Information

Source Institution:
University of South Florida Library
Holding Location:
University of South Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 001992346
oclc - 316062586
usfldc doi - E14-SFE0002382
usfldc handle - e14.2382
System ID:
SFS0026700:00001


This item is only available as the following downloads:


Full Text
xml version 1.0 encoding UTF-8 standalone no
record xmlns http:www.loc.govMARC21slim xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.loc.govstandardsmarcxmlschemaMARC21slim.xsd
leader nam Ka
controlfield tag 001 001992346
003 fts
005 20090326095247.0
006 m||||e|||d||||||||
007 cr mnu|||uuuuu
008 090317s2008 flu s 000 0 eng d
datafield ind1 8 ind2 024
subfield code a E14-SFE0002382
035
(OCoLC)316062586
040
FHM
c FHM
049
FHMM
090
BF121 (Online)
1 100
Bykowski, Cathy A.
4 245
The inflammatory consequences of stress and adiposity
h [electronic resource] /
by Cathy A. Bykowski.
260
[Tampa, Fla] :
b University of South Florida,
2008.
500
Title from PDF of title page.
Document formatted into pages; contains 107 pages.
502
Thesis (M.A.)--University of South Florida, 2008.
504
Includes bibliographical references.
516
Text (Electronic thesis) in PDF format.
3 520
ABSTRACT: The inflammatory process is important in protecting the body against the invasion of pathogens, but recent research has suggested that a long-term inflammatory response may lead to chronic diseases (e.g., Black, 2003; Wu, Dorn, Donahue, Sempos, & Trevisan, 2002). Two factors that have been implicated in the inflammatory and disease processes are stress and obesity (Black, 2003). While their individual lines of research continue to grow, few researchers have attempted to integrate these factors into one model to explain their effects on inflammation. This study aimed to replicate previous findings suggesting relationships between stress, obesity and inflammation and test an integrated model of stress and obesity by examining a possible interaction between the effects of stress and obesity on inflammation. Socioeconomic Status (SES) and depression were employed to examine the association between stress and the inflammatory marker, c-reactive protein (CRP). The study utilized the data resulting from the National Health and Nutrition Examination Survey (NHANES; National Center for Health Statistics, 2006). Included in the dataset are 4998 adults (2416 males and 2582 females) ranging in age from 18 years to over 85 years (M = 47.13, SD = 20.86). A subsample (N = 589) completed the Major Depression module of the Composite International Diagnostic Interview (CDCI). The results indicate that body mass index, waist circumference, income, education, and depression symptoms significantly predict CRP. The data also suggest an interaction between the adiposity variables and the SES variables. This supports the hypothesis that the inflammatory effect of stress on an individual is moderated by adiposity.
538
Mode of access: World Wide Web.
System requirements: World Wide Web browser and PDF reader.
590
Co-advisor: William Sacco, Ph.D.
Co-advisor: Kristen Salomon, Ph.D.
653
Socioeconomic status
Depression
Body mass index
Waist circumference
C-reactive protein
0 690
Dissertations, Academic
z USF
x Psychology
Masters.
773
t USF Electronic Theses and Dissertations.
856
u http://digital.lib.usf.edu/?e14.2382



PAGE 1

The Inflammatory Consequences of Stress and Adiposity by Cathy A. Bykowski A thesis submitted in partial fulfillment of the requirements for the degree of Master of Arts Department of Psychology College of Arts and Sciences University of South Florida Co-Major Professor: William P. Sacco, Ph.D. Co-Major Professor: Kristen Salomon, Ph.D. Jonathan Rottenberg, Ph.D. Date of Approval: February 13, 2008 Keywords: socioeconomic status, depression, body mass index, waist circumfere nce, c-reactive protein Copyright 2008, Cathy A. Bykowski

PAGE 2

Acknowledgements I would like to thank my advisors, Dr. William Sacco and Dr. Kristen Salomon, for their instruction, guidance, and patience while I worked to complete this thesi s. I am also very grateful to my labmate, Kristi White, for her valuable input and unending encouragement. In addition, I would like to express my appreciation to Monika Wahi for sharing her statistical expertise and assistance navigating the anal ysis of a large national dataset. Finally, I thank my husband, Jonathan Bykowski, for his enduring love and support throughout this entire process.

PAGE 3

Table of Contents List of Tables ..................................................................................................................... iii List of Figures ......................................................................................................................v Abstract .............................................................................................................................. vi The Inflammatory Consequences of Stress and Adiposity ..................................................1 Stress and the Inflammatory Response ....................................................................2 Stress as a stressor ........................................................................................3 Stress as a psychological experience ...........................................................7 Obesity and Inflammation......................................................................................11 An Integrated Model of Stress and Obesity’s Effect on Inflammation ................. 15 Present Study .........................................................................................................19 Method ...............................................................................................................................22 Participants .............................................................................................................22 Materials ................................................................................................................23 Stressor .......................................................................................................23 Psychological stress ...................................................................................24 Adiposity ....................................................................................................25 Inflammation ..............................................................................................26 Procedure ...............................................................................................................26 Data Analysis .........................................................................................................27 Results ................................................................................................................................31 SES Variables and Adiposity Variables as Predictors of CRP ..............................31 Descriptive statistics ..................................................................................31 Correlations ................................................................................................32 Hierarchical linear regression analyses ......................................................33 Depression Variables and Adiposity Variables as Predictors of CRP ...................40 Descriptive statistics ..................................................................................40 Correlations ................................................................................................41 Hierarchical linear regression analyses ......................................................42 Gender as a Moderator of the Relationship between Stress Variables and CRP ........................................................................................................................45

PAGE 4

ii Gender as a Moderator of the Relationship between Adiposity and CRP .............49 The Relationship of BMI versus WC to CRP ........................................................49 Discussion ..........................................................................................................................51 Stress and Inflammation ........................................................................................51 Adiposity and Inflammation .................................................................................53 Interactions between the Stress and Adiposity Variables ......................................54 Gender ....................................................................................................................56 Strengths of the Study ...........................................................................................58 Limitations of the Study.........................................................................................59 Conclusions ............................................................................................................63 References ..........................................................................................................................66 Appendix A: Hierarchical Linear Regressions that Utilize Sampling Wei ghts: Tables .................................................................................................................................79 Appendix B: 3-way Hierarchical Linear Regressions (Unweighted Data) ........................96 Appendix C: Gender as a Moderator of the Relationship between Adiposity and CRP (Unweighted Data): Tables .....................................................................................105

PAGE 5

iii List of Tables Table 1 Mean Values of Variables in the SES Analyses ........................................31 Table 2 Distribution of Participants Across the Levels of Household Income .......32 Table 3 Correlations Between Variables in the Socioeconomic Status Analyses ..........................................................................................33 Table 4 Summary of Hierarchical Regression Analysis for Education and BMI Predicting LogCRP ....................................................34 Table 5 Summary of Hierarchical Regression Analysis for Education and WC Predicting LogCRP .....................................................34 Table 6 Summary of Hierarchical Regression Analysis for Income and BMI Predicting LogCRP ........................................................35 Table 7 Summary of Hierarchical Regression Analysis for Income and WC Predicting LogCRP .........................................................35 Table 8 Mean Values of Variables in the Depression Analyses .............................40 Table 9 Correlations between Variables in the Depression Analyses .....................41 Table 10 Summary of Hierarchical Regression Analysis for Depression Diagnosis and BMI Predicting LogCRP .................................43 Table 11 Summary of Hierarchical Regression Analysis for Depression Diagnosis and WC Predicting LogCRP ..................................43 Table 12 Summary of Hierarchical Regression Analysis for Depression Symptoms and BM predicting LogCRP .................................44 Table 13 Summary of Hierarchical Regression Analysis for Depression Symptoms and WC Predicting LogCRP .................................44 Table 14 Summary of Hierarchical Regression Analysis for Gender and Education Predicting LogCRP ................................................46

PAGE 6

iv Table 15 Summary of Hierarchical Regression Analysis for Gender and Income Predicting LogCRP ....................................................46 Table 16 Summary of Hierarchical Regression Analysis for Gender and Depression Diagnosis Predicting LogCRP .............................47 Table 17 Summary of Hierarchical Regression Analysis for Gender and Depression Symptoms Predicting LogCRP ...........................47

PAGE 7

v List of Figures Figure 1 BMI & Education Interaction .....................................................................37 Figure 2 BMI & Income Interaction .........................................................................38 Figure 3 WC & Education Interaction ......................................................................39 Figure 4 Gender & Income Interaction .....................................................................48

PAGE 8

vi The Inflammatory Consequences of Stress and Adiposity Cathy A. Bykowski ABSTRACT The inflammatory process is important in protecting the body against the invasi on of pathogens, but recent research has suggested that a long-term inflammatory r esponse may lead to chronic diseases (e.g., Black, 2003; Wu, Dorn, Donahue, Sempos, & Trevisan, 2002). Two factors that have been implicated in the inflammatory and disease processe s are stress and obesity (Black, 2003). While their individual lines of research continue to grow, few researchers have attempted to integrate these factors into one m odel to explain their effects on inflammation. This study aimed to replicate previous finding s suggesting relationships between stress, obesity and inflammation and test an integrated model of stress and obesity by examining a possible interaction between the effects of s tress and obesity on inflammation. Socioeconomic Status (SES) and depression were employed to examine the association between stress and the inflammatory marker, c-rea ctive protein (CRP). The study utilized the data resulting from the National Health and Nut rition Examination Survey (NHANES; National Center for Health Statistics, 2006). Included in the dataset are 4998 adults (2416 males and 2582 females) ranging in age from 18 years to over 85 years ( M = 47.13, SD = 20.86). A subsample (N = 589) completed the Major Depression module of the Composite International Diagnostic Interview ( CDCI). The results indicate that BMI, WC, income, education, and depression symptoms

PAGE 9

vii significantly predict CRP. The data also suggest an interaction between the a diposity variables and the SES variables. This supports the hypothesis that the inflammat ory effect of stress on an individual is moderated by adiposity.

PAGE 10

viii The Inflammatory Consequences of Stress and Adiposity The inflammatory process is important in protecting the body against the invas ion of pathogens, but recent research has suggested that a long-term inflammatory res ponse may lead to chronic diseases such as insulin resistance (Wu, Dorn, Donahue, Sempos, & Trevisan, 2002), atherosclerosis, type 2 diabetes, and metabolic syndrome (Black, 2003) The severity of these illnesses underscores the need to understand the mechanism s that lead to the prolonged inflammation with which they are associated. Two factors that have been implicated in the inflammatory and disease processes are stress and obe sity (Black, 2003). Both factors have been associated with increased inflammatory mar kers (e.g. Brydon, Edwards, Mohamed-Ali, & Steptoe, 2004; Lemieux et al, 2001; McCarty, 1999; Owen, Poulton, Hay, Mohamed-Ali, & Steptoe, 2003) as well as increased risk for these inflammatory diseases (e.g. Burton, Foster, Hirsch & van Itallie 1985; Wellen & Hotamisligil, 2005). While their individual lines of research continue to grow, few researchers have attempted to integrate stress and obesity into one model to explain their effects on inflammation. This paper will discuss the previous research in the disti nct areas of stress and obesity and will then examine a model to explain possible interactions of the two factors. The human body is equipped with a complex security system that is activated when faced with a threat due to injury or infection. Granulocytes are the major group of 1

PAGE 11

2 cells that respond to most threats by migrating to the site of infection or injury, d estroying the threat by secreting toxic chemicals, and then consuming the left-over pa rticles and injured tissue. Some granulocytes also release cytokines which send out message s to the remainder of the body to prepare it for the impending attack (Segerstrom & Mil ler, 2004). The cytokines, particularly interluekin-6 (IL-6) and Tumor Necrosis Factor(TNF), initiate inflammation through the acute phase response (APR). The result is fe ver, a sickness response, and the production of acute phase proteins (APPs), such as c-reacti ve protein (CRP), which enable the body to defend itself against the threat (Black, 2003). The inflammatory response is intended to protect and heal the body. However, recent studies have begun to describe complications that can arise when this process occurs t oo frequently or persists for an extended period of time (Black, 2003; Wu et al., 2002). A better understanding of the factors that cause inflammation may lead to more effective remedies and prevention programs to decrease the incidence of inflammatory di seases. Stress and the Inflammatory Response The definition of stress is at the center of one of psychology’s oldest debates. F or many decades it has been unclear as to whether stress should be defined in terms of a stimulus or response. Hans Selye may be one of the most influential early stres s researchers and his work, which focused on physiological reactions to a stimulus, sha ped the field for many decades. He first used the term “stress” in 1946 to describe a n outside influence that acts on an organism, a stimulus. He continued to use that definition until 1950, when he proposed that “stress” be defined in terms of an internal reaction to an outside influence, or “stressor” (Mason, 1975). Perhaps this surprising change in definitions by one of the great leaders in the field was the beginning of the c onfusion over

PAGE 12

3 how to define stress. Today, researchers are forced to define stress for t heir own project, with some researchers studying the stimulus some studying a response and some studying an interaction of the two (Mason, 1975). Because of the vague nature of the term “stress” it can sometimes be unclear as to whether a particular out come of stress is due to the stimulus or the response to the stimulus. For this reason, this paper examines both aspects of stress. The effects of stress will be measured in terms of the stimuli, or stressor, and in terms of the mental state that is the psychological reaction t o a stressor or stressors. From the early research of Selye and Cannon, scientists have acknowledged the physiological impact of stress (Mason, 1975). As researchers learn more about t he biology of the human body and its responses to stress, it is becoming clearer that st ressors and injury or infection result in the activation of the same pathways and the releas e of the same biochemicals. The body’s response to stress is characterized by t he release of corticosteroids and catecholamines, such as epinephrine and norepinephrine. These hormones, much like granulocytes, initiate the production of cytokines, commencing the APR. The connection between the two systems may be an evolutionary adaptation mechanism. When the body is faced with a threat the sympathetic nervous system i s activated so that the person is ready to fight or flee the threat. In addition, the inflammatory system is activated so that the body is ready to battle any i nfection or injury that results from the fighting or fleeing (Black, 2002). This relationship betwee n the two systems will be examined with a focus on the inflammatory results of stress (both as a stressor and a psychological reaction to a stressor).

PAGE 13

4 Stress as a stressor. As previously mentioned, the first scientific definition of stress was that of a stimulus. This definition is still used in both science and eve ryday life. The field of physics refers to stress when referring to the force tha t is being placed on a material (Mason, 1975). In common language one talks about the stress of a deadline or the stress of school, with the focus on the stressor, not the reaction to the force or the deadline. Stress researchers employ this definition when they e xpose someone to “stress” by requiring them to give a speech, perform mental arit hmetic, or to complete a difficult task, such as tracing a star in a mirror. Often it is unde rstood that the force of the stimulus is the stress, and the focus is not on the individual’s psycholog ical reaction to the stress. It is assumed that a person has been exposed to stress when completing these tasks, although the individual is often not asked about his or her experience or reaction to the task. These studies benefit from the concrete de finition of stress, it is much easier to measure how long a speech is than to have a person quanti fy his or her psychological reaction to a speech. Experimenters that employ the stressor definition of stress in laboratory experiments have more control over the stress t han scientists that define stress in terms of an individual’s reaction (Steptoe & Vogele, 1991). Common laboratory stressors include mental arithmetic, giving a speech, and performing uncommon and difficult tasks (Steptoe & Vogele, 1991). Researchers have also studied common real-world stressors, such as caring for a chronically i ll family member, death of a loved one and socioeconomic status (SES). It is important to point out that many researchers who study these concepts are looking for a specifi c reaction, such as increased heart rate or blood pressure. However, what they define as “st ress” is the stressor They may look for a reaction to stress but the reaction is not the stress. The

PAGE 14

5 definition of stress that is being employed is based on a force outside the body. In o rder to distinguish between “stress” as a stimuli and “stress” as a reaction, the term “stressor” will be used to indicate a force outside the body, or a stimulus. The study of inflammatory responses to stressors, in both animals and humans, has been growing exponentially in recent years. While the results are not always consistent (e.g., Goebel, Mills, Irwin, & Ziegler, 2000; Heinz et al., 2003; Lutge ndorf, Logan, Costanzo, & Lubaroff, 2004; Owen & Steptoe, 2003), many experimenters have found a relationship between inflammatory markers, such as cytokines and APPs, and exposure to a stressor. Some of the first subjects to be studied were rats who demonstrated a relationship between IL-6 and stress. When rats were exposed to a physical stressor (e.g. electric foot-shock), a psychological stressor (e.g., a conditioned aversive stimulus) or a stressor that has both psychological and physical component s (e.g., restraint), the rat’s plasma level of IL-6 increased (Zhou, Kusnecov, S hurin, and DePaoli, 1993). This early animal research quickly led to the study of inflammati on in response to stressors in humans. Increases in concentrations of the cytokines such as IL6 and TNF, have been observed following laboratory speech tasks (Ackerman, Martino, Heyman, Moyna, & Rabin, 1998), physical exercise (Goebel et al., 2000), color-word interference tasks, mirror tracing tasks (Owen and Steptoe, 2003) and acad emic examination (Maes et al., 1998). In addition to an increase in IL-6 in response to an acute laboratory stressor, chronic naturalistic stressors also influence cytokine and APP production. One such stressor is socioeconomic status (SES), an indicator of social position. This constr uct describes types and amounts of resources to which a person has access, both tangible

PAGE 15

6 (e.g. wealth) and intangible (e.g. knowledge and social support). The most common ways to measure SES is through highest level of education attained, amount of income, occupational status, or a combination of the three (Adler & Snibbe, 2003). Researcher s have demonstrated that people of lower SES experience more life stressors, es pecially those stressors associated with a loss of income or ill health. In addition, those stressors have a greater impact on their emotional well-being, compared to individuals of hi gher SES (Kessler, 1979; McLeod & Kessler, 1990). A substantial amount of research has demonstrated a strong negative relations hip between SES and health, in prevalence of chronic diseases (including osteoarthri tis, hypertension, cervical cancer and cardiovascular disease) as well as mor tality rates (Adler et al., 1994). Lower SES has also been associated with higher levels of CRP. Owen et al. (2003) demonstrated an association between occupational status and CRP that was independent of age, sex, body mass, waist-to-hip ratio, smoking, alcohol use, and season of the year. A similar finding was recently reported when using education as a measure of SES (McDade, Hawkley, & Cacioppo, 2006). In addition, Brydon et al. (2004) found that individuals exposed to the chronic stressor of living at a low SES show greate r increases in IL-6 production when they are exposed to a laboratory stressor, compar ed to those of a higher SES (Brydon et al., 2004). The evidence that stress activates the inflammatory response is convincing when stress is defined as a stimulus. In this sense, when people are exposed to stress, their bodies respond by increasing chemicals such as cytokines and APPs, which are responsible for inflammation. However, as mentioned, defining stress as a stre ssor is

PAGE 16

7 only one of the ways in which people have examined the construct. The psychological reaction or mental state that is the response to the stimulus must also be consider ed. Stress as a psychological experience. The other definition of stress is that of the reaction to a stimulus. This implies that stress is an internal condition that may vary from person to person (Hobfoll, 1989). Again, the effects that this stress has on a person may be what are of interest in a particular study. However, this time it is the eff ect of the mental state that is the result of the stimuli This form of stress is not as objective, making it more difficult to measure and understand. Researchers who study this form of stress focus on the psychological condition of a person during or after exposure to a stressor. In this sense stress is an experience that is due to a stimulus, or stressor. However, the stressor is not the stress, the experience is the stress. This ty pe of stress is usually studied with the use of self-report questionnaires which question the parti cipant about their state of mind and the psychological effects of a stimulus or stimuli. T o make a clear distinction between this form of stress and stressors stress that is the internal experience of a stimulus will be referred to as psychological stress. Just as stressors have been associated with increases in the inflammatory response, studies examining psychological stress produce similar results. Re search has shown that those who report severe psychological stress also have significantl y higher levels of CRP (Hapuarachchi et al., 2003), TNF, and IL-6 (Maes et al., 1998) compared to those who report normal levels of psychological stress. Hapuarachchi, Chalmers Winefiled, and Blake-Mortimer (2003) asked healthy volunteers to complete the Gene ral Health Questionnaire (GHQ) to indicate how much psychological stress they ex perienced over the previous two weeks. Within two days of the completion of the GHQ they

PAGE 17

8 reported to the laboratory to allow for collection of a blood sample from which CRP concentrations were measured. Statistical analyses indicated that those wh o received GHQ scores of 0-1 (normal, little to no stress) had significantly lower CRP concentrations compared to those who received scores of 4 or greater (severely st ressed). In a similar study, Maes et al. (1998) collected samples of blood and asked students t o complete the Perceived Stress Scale (PSS) more than one month before and after an exam as well as one day before the exam. Their data indicate that when more psycholog ical stress was perceived, higher concentrations of IL-6 and TNFwere present (Maes et al., 1998). These results suggest a psychological component that is an important part of the inflammatory response. Some researchers have explored specific conditions that result due to exposure to stressors. Burnout is a condition associated with emotional exhaustion, depersonaliza tion and diminished personal accomplishment caused by long-term exposure to stress. T his condition is associated with increased levels of TNF(Grossi, Perski, Evengard, Blomkvist, & Orth-Gomer, 2003) and CRP in women (Toker, Shirom, Shapira, Berliner, & Melamed, 2005). In addition, Tel Aviv women who report a state of fear induced by periodic terrorist attacks also show a positive relationship between the fear of terror (a type of psychological stress) and CRP level (Melamed, Shirom, Toker, Berliner, & Shapira, 2004). Also, individuals with post-traumatic stress disorder, a psychologic al state that is the result of a traumatic stressor, exhibit higher levels of IL-6 (Maes et al., 1999). Depression can also be a mental state that results from exposure to a stres sor, thus it can be considered a form of psychological stress. Depression is characteri zed by

PAGE 18

9 episodes during which an individual experiences a depressed mood or loss of interest or pleasure in nearly all activities. Other depressive symptoms include a signi ficant change in weight, sleep, psychomotor activity, or loss of energy, feelings of worthles sness, inability to concentrate, or recurrent thoughts of death. These symptoms often re sult in clinically significant distress or impairment in important areas of funct ioning (American Psychiatric Association Diagnostic and Statistical Manual of Mental Disorders, IV – text revision, 2000). Many studies have found that females who experience a stressful life eve nt are more likely to have a depressive episode (Kendler, Karkowski, & Prescott, 1999) Kendler, Karkowski, & Prescott (1998) report that many stressful life event s (11 of the 15 they studied) are associated with the onset of major depression (MD) in the mont h in which they occurred. This association indicates a link between the stimulus (st ressor) and reaction to the stimulus (MD). They also found that the severity of the event was positively associated with onset of MD, indicating that the more stressful the person finds the event; the more likely they are to experience psychological stress. A similar line of evidence suggests that a lack of resources creates stres s in a person’s life and also increases depressive symptoms. Hobfoll’s (1989) Conservati on of Resources model defines a resource as anything that is valued by an individual. A resource may be an object (e.g., car or house), condition (e.g., marriage or tenure), personal characteristic (e.g., self-esteem), or energy (e.g., time or know ledge). He goes on to explain stress as a reaction to the loss of resources, a threat to the loss of r esources, or the inability to gain resources after they have been depleted. Researchers have studied this type of stress in relation to depressive symptoms. Resource loss has been as sociated

PAGE 19

10 with an increase in depressive symptoms while resources gain has been associ ated with a decrease in depressive symptoms. It has also been observed that a change in resour ces mediates the relationship between negative life events and depressive sym ptoms (Holahan, Moos, Holahan, & Cronkite, 1999). Therefore, the lack of resources can be viewed as an experience of stress which is also manifested in depressive sym ptoms. Depression has also been associated with increases in cytokines and APPs (e. g., Miller, Stetler, Carney, Freedland, & Banks, 2002; Owen & Steptoe, 2003; Suarez, Krishnan, & Lewis, 2003), although the results are not always consistent. One study di d not find significantly higher CRP levels in depressed patients compared to non-depresse d controls, although they did identify a difference in TNFlevels (Tuglu, Kara, Caliyurt, Vardar, & Abay, 2003). Danner, Kasl, Abramson, and Vaccarino (2003) found that men with a history of a depressive episode were twice as likely to have high levels of CRP compared to men with no history of a depressive episode and that more recent episodes were associated with a greater likelihood of increased CRP levels. Intere stingly this same relationship was not found in females, they hypothesize that this may be because of possible protective effects of estrogen or because CRP levels tend to be higher in even non-depressed women, making a difference difficult to observe. Other studies have found that pharmacological treatment for depression resulte d in significantly decreased inflammatory markers (Lanquillon, Krieg, Bening -Abu-Shach, & Vedder, 2000; Tuglu et al., 2003). Decreases in CRP levels after antidepressant treatment are consistently identified (Lanquillon et al., 2000; Tuglu et al., 2003). I n addition, Tuglu et al. (2003) found decreases in TNFlevels. However, Lanquillon et al. (2000) only found decreases in TNFlevels in those patients who also exhibited less

PAGE 20

11 depressive symptomatology. While IL-6 levels have not shown a significant post-treatment decrease, there is evidence that patients who will respond to treat ment have significantly lower IL-6 levels pre-treatment compared to those who will no t respond to treatment (Lanquillon et al., 2000). Perhaps this indicates that those who have more prolonged depression (or stress) show higher levels of this cytokine. These data support the notion that the psychological experience of a stressor, manifested in conditions such as burnout and depression, is related to the inflammatory response. There is also evidence that when stress is defined in terms of the str essor, inflammatory cytokines and APPs are also increased. These inflammatory responses to stress may be the link between stress and diseases such as atherosclerosi s, which is now thought to be an inflammatory disease (Heinz et al., 2003). Understanding the physiological repercussions of stress may allow for the prevention of such dis eases. Obesity and Inflammation Another factor that influences the inflammatory system is adiposity, the accumulation of adipose, or fat, tissue. Obesity, an excess of adipose tissue, is ass ociated with a number of diseases including hypertension, hypercholesterolemia, diabetes, coronary heart disease, and some types of cancer (Burton et al., 1985). Further, obes ity is becoming an epidemic in the United States. Between the years 2001 and 2002, it was estimated that 65.7% of American adults were overweight or obese and that 30.6% we re obese. In addition, overweight and obesity in children during that time period was estimated to be 31.5% (Hedley et al., 2004). The pervasiveness of obesity is dramati cally increasing. Among American adults there has been a 50% increase in prevalence i n each of the past two decades. A similar pattern is seen in the nation’s children with the

PAGE 21

12 prevalence of overweight tripling in just twenty years (Wyatt, Winters & Dubbert, 2006). Increased body fat has been associated with hyperinsulinemia, diabetes, incr eased lipid levels, hypertension, gallbladder disease, and some forms of cancer (Burton et a l., 1985; Hartz, Rupley, & Rimm, 1984; Ohlson et al., 1985). The prevalence of obesity and its associated complications highlight the need for further research into the condit ion. Historically, adipose tissue was considered to be only an energy store. In re cent years, however, it is becoming apparent that adipose tissue is an active organ of the bod y (Kershaw & Flier, 2004). The endocrine properties of adipose tissue are becoming l ess ambiguous. Research indicates that it is capable of secreting many chemi cals which have effects throughout the body, including proinflammatory cytokines and APPs (Cal abro, Chang, Willerson, & Yeh, 2005; Lemieux et al., 2001; Mohamed-Ali et al., 1997; Mohamed-Ali, Pinkney, & Coppack, 1998; Owen & Steptoe, 2003; Visser, Bouter, McQuillan, Wener, & Harris, 1999; Yudkin, Stehouwer, Emeis, & Coppack, 1999). Understanding the actions of adipose tissue may provide insight into the connections between obesity and the diseases with which it is associated. Numerous researchers have observed increased levels of cytokines and APPs wit h increased adiposity, measured by body mass index, waist-to-hip ratio, and similar procedures (Kern et al., 1995; Lemieux et al., 2001; Mohamed-Ali et al., 1997; Owen & Steptoe, 2003; Visser et al., 1999; Yudkin et al., 1999). One study found that obese women had significantly higher concentrations of IL-6 and TNF, compared to normal weight women. A weight-loss of at least 10% was also associated with a reduct ion in the cytokine levels (Ziccardi et al. 2002).

PAGE 22

13 In addition to these correlational observations, in vivo and in vitro studies have been able to demonstrate the secretions of the tissue directly. For example, TNFproduction by adipose tissue and adipocytes has been demonstrated in vitro. Adipose tissue biopsies from the abdomens of 37 lean and obese premenopausal females and were tested for the presence of TNFmRNA and TNFprotein. The results of this study, as well another that used a similar method, confirmed the presence of TNFmRNA in the adipose tissue of all volunteers (Hotamisligil, Arner, Caro, Atkinson, & Spiegle man, 1995; Kern et al., 1995). In addition, it was demonstrated that the adipose tissue of obese individuals expressed more than double the amount of TNFmRNA as lean controls. All tissue secreted the TNFprotein and the tissue of obese individuals secreted more than that of lean individuals. Weight-loss resulted in a decrease of TNF(Hotamisligil et al., 1995) and TNFmRNA (Hotamisligil et al., 1995; Kern et al., 1995) in most of the obese individuals. This indication that adipose tissue of obese individuals produces excess TNFhas increased our understanding of the role of obesity in health problems such as insulin resistance (Wellen & Hotamisligil, 2005). Similar to the stress literature, studies of the production of cytokines by adipos e tissue do not always result in consistent findings. One in vivo study measured the differences between artery and vein concentrations of IL-6 and TNFacross subcutaneous adipose tissue. It was discovered that the concentration of IL-6 in venous samples, leaving the adipose tissue, were more than twice as high as the ar terial samples, entering the adipose tissue. These results support the hypothesis that IL-6 is produced by adipose tissue. It was also estimated that approximately 30% of the IL-6 ci rculating in the body is secreted by adipose tissue. This same study found no arterio-venous

PAGE 23

14 differences in TNFconcentration (Mohamed-Ali et al., 1997). Although it is still believed that both IL-6 and TNFare produced by adipose tissue, these data indicate that they may be produced by different depots of adipose tissue at different locati ons in the body. IL-6 is produced by the abdominal subcutaneous fat depot that was examined in this particular experiment, however TNFmay be produced by another depot (Mohamed-Ali et al., 1997). The production of CRP by adipose tissue has also been demonstrated in vitro (Calabro et al., 2005; Ouchi et al., 2003). Human adipocytes incubated for 24 hours with IL-1and IL-6 produced about twice the amount of CRP as cells that were not stimulated. Cells incubated with adiponectin and leptin did not produce CRP. These data support the role of inflammatory cytokines in the initiation of production of CRP in human adipose tissue. Furthermore, treatment with anti-inflammatory drugs, suc h as aspirin, decreased the amount of CRP produced by the adipocytes, indicating a possible pharmacological modulator of CRP production (Calabro et al., 2005). This evidence supports the concept that adipose tissue is not a passive energy storage center, but an active organ. Adipose tissue is responsible for the secretion of chemicals that influence the body’s inflammatory process. There is substa ntial evidence that adipose tissue is a source of the proinflammatory cytokines, IL-6 and TN F, as well as the acute phase protein, CRP. However, studies have shown that not all fat depots secrete the same cytokines. The secretions of adipose tissue vary depending on the location of the tissue (Fried, Bunkin, & Greenburg, 1998; Mohamed-Ali et al., 1997). The differing secretions of the tissue help to explain why central adiposity has been shown to be a better predictor of cardiovascular disease, premature death, stroke, a nd

PAGE 24

15 ovarian cancer than overall obesity (Bjorntorp, 1988). One study even suggested that visceral adipose tissue is detrimental to health whereas adipose tissue stor ed on the hips may be beneficial to health (Yusef et al., 2005). The strong associations between c entral adiposity and illness make it likely that more inflammatory markers are p roduced in central adipose tissue than in adiposity tissue located in other regions. Continuing to study the function of adiposity and its location with respect to inflammation is im portant in understanding the role of obesity in inflammatory disease. An Integrated Model of Stress and Obesity’s Effect on Inflammation There is a significant amount of research to suggest that psychological stre ss contributes to the inflammatory process. The result of this relationship is the increased levels of cytokines and acute phase proteins during times of stress. This is evi dent in that IL-6, TNF, and CRP are all associated with both the acute phase response and reactions to stress. These chemicals serve to protect and heal the body when faced with a n infection. However, in excess they result in chronic diseases such as type 2 di abetes (Pradhan, Manson, Rifai, Buring, & Ridker, 2001), metabolic syndrome X (Black, 2003), atherosclerosis (Libby, Ridker, & Maseri, 2002), and cardiovascular disease (Bla ck & Garbutt, 2002). Obesity is a growing epidemic in the United States and it is important to understand both the causes and effects of obesity (Wyatt et al., 2006). The implicati ons of obesity are becoming greater as we learn more about the role of adipose tissue as an active endocrine organ, secreting chemicals that influence all parts of the body. There is sufficient evidence that adipose tissue is responsible for some of the body’s producti on of

PAGE 25

16 cytokines and acute phase proteins (e.g., Calabro et al., 2005; Mohamed-Ali et al., 1997; Ouchi et al., 2003). The relationships between stress and cytokine/APP increases and adipose tiss ue and cytokine/APP production have led to the proposal of a new model. It is known that stress is associated with the release of cytokines and APPs and that adipose t issue is one source of these chemicals. The proposed model combines this information to suggest that stress acts at the level of the adipose tissue to increase the secretion of t hese chemicals. The new model suggests that stress promotes the release of inflammatory markers from adipose tissue. Stress may act on or interact with the adipose tissue to result i n the release of more cytokines and APPs than are produced under normal circumstances The proposed model is based on the research indicating that adipose tissue produces cytokines and APPs. The body needs a certain level of these chemicals t o protect itself from injury and infection, so the adipose tissue may be constantly secreting the inflammatory markers. However, if there is too much adipose tissue, there are more secretion sites, which lead to an abundance of the markers. Another line of research suggests that stress is associated with an increase in inflammatory marke rs. The proposed model suggests that this increase in inflammatory markers may be due to s tress causing the adipose tissue to secrete more cytokines and APPs than it would under normal circumstances. Therefore, the new model posits that the relationship bet ween stress and inflammation is moderated by adiposity. Stress increases the rel ease of cytokines and APPs from adipose tissue, when there is a lot of adipose tissue there w ill be many secretion sites and more cytokines will be released. Thus, the effect t hat stress has on inflammation is dependent on the amount of adipose tissue that is present. This

PAGE 26

17 interaction between stress and adipose tissue may help to explain their relati onships to disease. This model proposes a potential mechanism through which obesity and stress lead to diabetes, atherosclerosis, and other cardiovascular diseases. If the mechanis m can be better understood, more treatment options may become available. Reducing the body’s supply of cytokines and APPs is complicated because there is a level that is nece ssary to fight infection. Researchers have begun to focus on the possibility that drugs may be used to control levels of inflammatory cytokines. Studies indicate that non-steroidal a ntiinflammatory drugs, such as ibuprofen, can prevent monocytes from producing inflammatory cytokines (Jiang, Ting, & Seed, 1998). However, it is possible that t hese drugs may decrease the cytokines to a dangerously low level which will inhibit t he body’s ability to fight infection (Yudkin, Kumari, Humphries, & Mohamed-Ali, 2000). The proposed model suggests that the body is at homeostasis when the correct amount of adipose tissue is present and there is no stress. Therefore, when inflammati on is at a high level, decreasing the amount of adipose tissue or stress may enable an individual to achieve homeostasis safely. Previous studies have focused on the differences between obese and non-obese participants or the impact of stress compared to lack of stress in participants. A study that combines these two areas of research will aid in the determination of the validity of this model. The model predicts that because lean people have less adipose tissue to secrete cytokines and APPs, they will not produce as many of these chemicals even when faced with stress. Conversely, those that are obese have more adipose tiss ue to secrete the inflammatory cytokines and APPs. When obese people are faced wit h stress

PAGE 27

18 they will show a more dramatic increase in cytokines and APPs due to their greater number of production sites (adipose cells). These explorations would help to determine the legitimacy of this model as well as further our understanding of the effe cts of stress and obesity on health. While limited experimental information is available concerning the direct relationships between stress, adiposity and inflammation, there is some evidence of the interactions. For example, some researchers have found that measures of adipos ity reduce or negate the association between stress (and depression) and cytokines a nd/or APPs. The indication that significant associations between stress and infla mmation become insignificant after adjusting for adiposity may indicate that adiposi ty moderates the stress/inflammation relationship (Douglas, Taylor, & O’Malley, 2004; Mil ler et al., 2002). These data have led Miller and colleagues to demonstrate a mediator model in which depression promotes accumulation of adipose tissue and the adipose tissue cause s inflammation directly by releasing cytokines as well as indirectly through the release of leptin which stimulates the production of cytokines (Miller, Freedland, Carney, Ste tler, & Banks, 2003). This type of model does not account for the inflammatory effects of depression that are independent of adiposity. A moderator model may be able to better explain this relationship. It is also limited to the effects of depression a nd not stress in general. A similar model was tested by Ladwig, Marten-Mittag, Lowel, Doring, a nd Koenig (2003) in a German, population-based sample of 3205 men ages 45-74. They found increased CRP concentrations in obese, compared to non-obese men. Also, CRP and BMI were significantly correlated. They also found that depressive mood was

PAGE 28

19 associated with high CRP concentrations in obese, but not non-obese, participants. An ANOVA revealed significant main effects for BMI and a depressive mood as well as a depressed mood x BMI interaction. This study indicates the importance of examining the combined effects of adiposity and depression on inflammation. Interestingly, the re have been very few studies to examine the interactions between stress and adiposit y on inflammation, producing no strong conclusions (McDade et al., 2006). Present Study The goal of this study is to replicate previous findings of the relationships between stress, obesity and inflammation. In addition, this study will test the integrated model by examining a possible interaction between the effects of stress a nd obesity on inflammation. The APP, CRP will be used as a measure of inflammation. This prote in has been identified as one of the most useful in predicting future cardiovascular dis ease (Albert, Ma, Rifai, Stampfer, & Ridker, 2002; Pearson et al., 2003; Ridker, Cushman, Stampfer, Tracy, & Hennekens, 1997; Ridker, Hennekens, Buring, & Rifai, 2000). High levels of CRP are associated with risk for coronary heart disease that is 1.7 ti mes the risk for those with lower levels of CRP (Danesh, Collins, Appleby, & Peto, 1998). Therefore it is important to gain better understanding of the factors that increase the le vels of the APP. CRP has also been associated with both stress (e.g. Hapuarachchi et al., 2003; Melamed et al., 2004; Owen et al., 2003) and obesity (e.g. Lemieux et al., 2001; Visser et al., 1999; Yudkin et al., 1999). To gain a comprehensive understanding of the effects of stress, measures of stressors and psychological stress will be examined. Socioeconomic Status (S ES) will be used to examine the effect that this type of stressor has on the inflammatory marker,

PAGE 29

20 CRP. In addition, depression will be employed to examine the role that psychological reactions to stressors have on CRP. This will add to the evidence that Ladwig et al. (2003) and Miller et al. (2003) presented with regard to only depression. A review of the literature has not resulted in a study that has examined the interactions of SE S and adiposity on inflammation. This study will also examine the association between adiposity and inflammati on. A previous study on the interaction of obesity and depression focused only on BMI (Ladwig et al., 2003). However, recent literature has indicated that the locati on of adiposity may be more important than how much there is (Yusef et al., 2005). In addition, in Miller et al.’s (2002) study of depressed and non-depressed participants, they did not identify an association of waist-to-hip ratio and CRP but did find an association between BMI and CRP. More research is needed to provide a better understanding the effects of adiposity and distribution of adiposity. Therefore, in addition to looking a t the relationship between BMI and inflammation, this study will also examine the e ffects of central adiposity, measured via waist circumference (WC), which will pro vide a more complete picture of the types of adiposity that are associated with the detr imental effects of stress and inflammation. In addition, this study will include both men and women from a population-based sample in the United States. Ladwig et al.’s (2003) study of German men may not be generalizable to the entire US population. In addition, Danner et al. (2003) found that depression in men, but not women, was associated with increased levels of CRP. This indicates that the interactions of adiposity and stress may be different in men and w omen.

PAGE 30

21 Both genders will be studied in order to develop a more complete understanding of this model. Recent research has dramatically changed our understanding of the physiol ogy of the body. We are now aware of the relationship between the stress response and the inflammatory process. It is becoming clear that stress causes an incre ase in the concentrations of cytokines and acute phase proteins associated with inflammation. W e are also learning more about the endocrine properties of adipose tissue and there is evidence to support the production of cytokines and APPs by adipose tissue. The proposed model joins these two quickly progressing areas of research. It suggest s that if stress causes the increase in certain biochemicals, and those biochemicals are produced by adipose tissue, stress may cause an increase in the adipose tissue’s pr oduction of those chemicals. This model may provide a better understanding of how stress and obesity interact and cause serious, often deadly, diseases. This study will test three main hypotheses in order to demonstrate the relationships between stress, adiposity, and inflammation. The first hypothesis is that stress will be positively related to inflammation. Stress will be defined as both a stressor (SES) and as psychological stress (depression). An increase in stress is expected to be associated with an increase in inflammatory markers (CRP). The second hypot hesis is that adiposity will be positively related to inflammation. Total body mass (BMI ) as well as central adiposity (WC) will be used to measure adiposity. An increase in adiposi ty is expected to be associated with an increase in inflammation (CRP concentration) The third hypothesis states that there will be an interaction between stress and adiposity. Adiposity is expected to moderate the association between stress and inflamm ation.

PAGE 31

22 Method This study utilized the data resulting from the National Health and Nutrition Examination Survey (NHANES). The survey was conducted by the National Center for Health Statistics division of the Centers for Disease Control (CDC) to obta in health, diet and nutrition information from a nationally representative sample. The current ve rsion of the NHANES survey began in 1999. The data were compiled and released in two-year increments; this study involved the analysis of the 2003-2004 data release. The data ar e available to the public and are accessed via the CDC’s website (National Cente r for Health Statistics, 2006). Participants NHANES utilized a complex, stratified random sampling procedure to obtain a sample of participants that is representative of the civilian, non-institutional ized population of the United States. The sampling procedure first sampled counties (or groups of small counties) in the United States. Within the counties, segments (bloc ks or clusters of households) were sampled and from each segment, households were selec ted. One or more member of each household sampled was asked to participate in the study. Participants were compensated for their time and reimbursed for the cost o f transportation and childcare (National Center for Health Statistics, 2006). Included in the dataset were 4998 adults who participated in the household interview and medical examination. The sample was comprised of 48% (N = 2416)

PAGE 32

23 males and 52% (N = 2582) females ranging in age from 18 years to over 85 years ( M = 47.13, SD = 20.86). Mexican Americans, African Americans, low-income individuals, and the elderly were over-sampled. However, the CDC provided sampling weights which were used to correct for the over-sampling and ensure that the data can be generalized to the U.S. population (National Center for Health Statistics, 2006) A subsample was chosen to participate in the Major Depression module of the Composite International Diagnostic Interview (CDCI) 1 Participants eligible for this subsample were those who spoke English or Spanish and were between the ages of 20 and 39 ( M = 28.97, SD = 5.74). The depression subsample consisted of 589 participants (294 male and 295 female). Analysis of the subsample utilized sampling weights to ensure that the subsample is also representative of the U.S. population (National Cente r for Health Statistics, 2006). Materials Stressor. The stressor that was measured was socioeconomic status (SES). As previously mentioned, this is an indicator of social position, which is often quantified through education, income, and occupation. Because occupational status is subjective and hard to quantify in the U.S., studies examining the relationship between SES and health often rely on measures of income and/or education to assess SES (Gallo & 1 This small subsample allowed for the analysis of psychological stress, oper ationalized as depression. However, there was a concern that the restricted age rang e may limit the external validity of the study and may not allow for enough variability in CRP to demonstrate a clear relationship. Therefore, the larger sample was the pri mary sample used for the analyses not requiring the measurement of depression or depressive symptoms.

PAGE 33

24 Matthews, 2003). Therefore, the effects of income and education on inflammation were each explored. The income variable was based on the participant’s annual household income. There were 13 categories from which the participants was asked to select whi ch best describes his/her household income. The categories were divided into five $5,000 ranges between $0 and $24,999 and five $10,000 ranges between $25,000 and $74,999. They were also given the options of “$75,000 and over,” “Over $20,000,” and “Under $20,000” (National Center for Health Statistics, 2006). There is an inverse relati onship between income level and the amount of stressors in a person’s life (Kessler, 1979; McLeod & Kessler, 1990). Participants were also asked to indicate the highest grade or level of school that they have completed or the highest degree that they have received. The answers were then coded as “Less than High School,” “High School Diploma (including GED),” or “More than High School” (National Center for Health Statistics, 2006). There is an inverse relationship between the amount of education received and the amount of stressors a person experiences (Kessler, 1979; McLeod & Kessler, 1990). Psychological stress. Depressive symptoms were measured to indicate the level of psychological stress. The Depression Module of an automated version of the Composite International Diagnostic Interview (CIDI) was used to asses s depressive symptoms and to provide a diagnosis of Major Depressive Disorder (MDD). The CIDI was designed by the World Health Organization to diagnose mental illness accordi ng to the criteria of the 10 th edition of the International Classification of Diseases (ICD-10) and the 4 th edition of the Diagnostic and Statistics Manual of Mental Disorders (DSM-IV)

PAGE 34

25 The CIDI is a structured interview designed to be used by lay people collecting data in epidemiological studies. The items assess the presence of each of the crite ria for MDD over the past 12 months. The participant was asked if there was a period of two weeks or longer during which they felt “sad or depressed or empty” or during which they “ lost interest in most things.” The items also addressed symptoms of MDD, such as appet ite change, insomnia/hypersomnia, reduced psychomotor activity, feelings of worthle ssness, inability to concentrate, or recurrent thoughts of death. The interview result ed in a diagnosis as well as information to assess the severity of the disorder. The m odule was administered during a face-to-face interview at the Mobile Examination Cent er (MEC) during which the interviewer asked each question (in either Spanish or English) as it appeared on a computer screen and entered the answers into a personal computer. The scores were provided as a quantity of depressive symptoms (0-9) as well as a dichotomous variable indicating whether or not criteria for Major Depressiv e Disorder were met (National Center for Health Statistics, 2006). Adiposity. Adiposity was measured using both body mass index (BMI) and waist circumference (WC). BMI is a measure of total body mass that is calculat ed by dividing the total body weight (in kilograms) by height (in meters) squared. These m easurements were conducted by a health technician at the MEC. Weight was measured and rec orded when the participant stepped onto a digital scale that was connected to the Integra ted Survey Information System (ISIS system), which automatically stored t he data. Participants were weighed wearing only underwear, a disposable paper gown, and f oam slippers. Height was measured with an electronic stadiometer that was als o connected to the ISIS system. The participant was instructed to stand straight against a vertical board,

PAGE 35

26 with relaxed arms and shoulders, both feet together and toes pointed out at an angle of approximately 60. The headboard was then lowered to the top of the participant’s head and the height was sent to the ISIS (National Center for Health Statistics 2006). WC is a measure of central adiposity and was measured with a metal tape whic h is placed around the participant’s torso at the level of the right ilium. This mea surement was also conducted at the MEC by a trained health technician. All measurements w ere made to the nearest 0.1 cm (National Center for Health Statistics, 2006). Inflammation. Blood CRP concentration was used to measure inflammation. Venipuncture was performed by a certified phlebotomist at the MEC. The phlebotomi st collected 89 to 92 ml of blood from the participant’s arm. The blood was processed, stored and shipped to the University of Washington to be processed using latex-enhanced nephelometry. A Behring Nephelometer was used to obtain the concentration of CRP in the sample of blood (National Center for Health Statistics, 2006). Procedure Households that were selected to participate in the study were sent a brochur e describing the study’s purpose and procedures. A trained interviewer then visit ed each household and conducted a screening interview to determine if any occupants of the household were eligible for the study. Once the eligible occupants were identi fied, they were asked to participate and provided informed consent. Each participant then completed the household interview to obtain demographic information, including information regarding SES. At the conclusion of the home interview the interviewer scheduled an appointment for the participant at the MEC (National Center for Heal th Statistics, 2006).

PAGE 36

27 The MEC is comprised of four large trailers that were designed for the NHAN ES survey and are equipped to perform extensive medical testing. When the participant reported to the MEC, the measures of adiposity were performed and blood was drawn to determine CRP level. The subset of participants who were selected to undergo ment al health testing also completed the CIDI during this visit to the MEC (National Center for Health Statistics, 2006). The data were then compiled and posted on the CDC’s website to be made available to the public (National Center for Health Statistics, 2006). Data Analysis The NHANES survey used a complex, stratified sampling procedure to produce a sample that is representative of the non-institutionalized civilian population of the United States. To achieve this representation, each participant’s data was weighte d. Analyses that incorporate these weights allow for an estimate of the data that would re sult if the entire non-institutionalized civilian population of the U.S. were sampled (National Ce nter for Health Statistics, 2006). The data were analyzed with and without the sample w eights and the results were similar. The unweighted analyses will be presented in t he results section (see Appendix A for summaries of the weighted analyses). The primary goal of the study was to test the proposed model which suggests that stress interacts with adipose tissue to increase the production of inflamma tory cytokines and APPs. There were three primary hypotheses: (1) stress variables (income/education/depression diagnosis/depression symptoms) will be positively related to the inflammatory marker, CRP, (2) adiposity variables (BMI/WC) will be pos itively

PAGE 37

28 related to CRP, and (3) there will be an interaction between stress variables a nd adiposity variables. To test the hypotheses, eight (4 stress variables x 2 adiposity variables) hierarchical multiple regression analyses were conducted. Other factor s which affect CRP levels, including gender, age, race, use of blood pressure medication, use of cholesterol medication, and smoking status, were statistically controlled to eliminate these constructs as possible confounding variables, increasing confidence in the conclusion that stress and adiposity are related to increased CRP levels 2 The covariates were entered into the model in step one. To test Hypotheses 1 and 2, the stress variable (income or education or depression measure) and adiposity var iable (WC or BMI) were entered at step two. Hypothesis 3 was tested at step three whe n the interaction between the adiposity and stress variable was added to the model. When a statistically significant interaction between a stress variable and a n adiposity variable was detected, regression lines were plotted as described by Preacher (2003) to dem onstrate the nature of the interaction. Previous research has indicated that the relationship between depression and inflammation may differ by gender (Danner et al., 2003). Therefore, further a nalyses were conducted to test the moderating effect of gender on the relationships betw een stress and adiposity and inflammation. To explore this possibility, hierarchical regres sions were conducted in which the covariates (age, race, use of blood pressure medication, use 2 Because race and SES may be related, there was a concern that controlling for race would restrict the range of the SES variable. Therefore, the analyses wer e also conducted excluding race as a covariate. However, the exclusion of race variables pro duced a similar pattern of results. Race is included in all analyses reported.

PAGE 38

29 of cholesterol medication, and smoking status) were entered in step one, the stress variable, adiposity variable, and gender were entered at step two, the two way interactions were added to the model in step three and in step four a three way intera ction between gender, stress, and adiposity was entered. To further examine the two-way interactions between gender and stress var iables and gender and adiposity variables, specific analyses were conducted to examine the role of gender. To explore the moderating effect of gender on the relationship between the stress variables and CRP, a hierarchical regression was conducted in which al l of the covariates mentioned above and the adiposity variables were entered into step one. In step two, gender and the stress variable were entered and in step three the intera ction between gender and the stress variable was entered into the model. Similarly, t o explore the moderating effect of gender on the relationship between adiposity and CR P, a hierarchical regression was conducted in which the covariates and stress varia bles were entered into step 1, the gender and adiposity variable was entered into step two and the gender x adiposity interaction was entered into step 3. When a statistically s ignificant interaction was detected, regression lines were plotted as described by P reacher (2003). A concentration of CRP that is greater than 1 mg/dL (i.e., 10 mg/L) is indicative of an inflammatory response to an infection (Pearson et al., 2003). Therefore, participants with a concentration of CRP that was greater than or equal to 1 mg/d L were excluded from the analysis ( n = 538; 10.78%). In addition, the distribution of CRP concentrations in the sample was positively skewed, necessitating a logarit hmic transformation of the data, which resulted in a normal distribution. Therefore, logCR P was used in all analyses.

PAGE 39

30 The distribution of number of depressive symptoms was also positively skewed, with 87.44% of the sample reporting no depression symptoms. Logarithmic and square root transformations were unsuccessful at normalizing the data; therefore raw data were used in all analyses. Due to the design of NHANES (see above), the sample of participants used for the “stressor” (education and income) analyses was not the same as the sam ple of participants used for the “psychological stress” (depression diagnosis a nd number of depression symptoms) analyses. The sample that was used in the stressor ana lyses is referred to as the SES Sample and the sample that was used in the psychological stress analyses is referred to as the Depression Sample

PAGE 40

31 Results SES Variables and Adiposity Variables as Predictors of CRP Descriptive statistics. Means and standard deviations of logCRP, BMI, and WC for the SES sample are presented in Table 1. The participants were distributed a cross the three levels of education: 1348 (30.26%) reported less than a high school education, 1141 (25.61%) reported having a high school education and 1966 (44.13%) reported having received more than a high school education. They were also distributed across the income brackets (see Table 2). Table 1 Mean Values of Variables in the SES Analyses Variable M SD logCRP -1.86 1.09 BMI 27.57 5.72 WC 95.86 14.80 Note. logCRP = logarithmic transformation of C-reactive protein (mg/dL), BMI = body mass index (kg/m 2 ), WC = waist circumference (cm)

PAGE 41

32 Table 2 Distribution of Participants Across the Levels of Household Income Income Bracket Frequency Percent $0 to $4, 999 96 2.32 $5,000 to $9,999 240 5.81 $10,000 to $14,999 410 9.92 $15,000 to $19,999 354 8.57 $20,000 to $24,999 372 9.01 $25,000 to $34,999 580 10.94 $35,000 to $44,999 452 10.94 $45,000 to $54,999 364 8.81 $55,000 to $64,999 235 5.69 $65,000 to $74,999 203 4.91 $75, 000 and over 825 19.97 Note. 329 participants did not have data regarding househ old income Correlations. Pearson correlation coefficients indicate significant positive relationships between logCRP and the adiposity variables and negative relationships between logCRP and the stress variables (i.e., higher CRP is associated w ith lower SES). In addition, the two adiposity variables and the two stress variables were signif icantly correlated. However, the stress and adiposity variables were not related to eac h other (See Table 3). Table 3 Correlations Between Variables in the Socioeconomic Status Analyses Log CRP BMI WC Education Income logCRP 1.00 0.420*** 0.438*** -0.059*** -0.076*** BMI 1.00 0.879*** -0.027 -0.011 WC 1.00 -0.028 -0.020 Education 1.00 0.334*** Income 1.00 Note. logCRP = logarithmic transformation of C-reactive protein (mg/dL), BMI = body mass index (kg/m 2 ), WC = waist circumference (cm). *** p < 0.001

PAGE 42

33 Hierarchical linear regression analyses. Hierarchical linear regressions were conducted to determine the ability of the stress variables (education and income) a nd the adiposity variables (BMI and WC) to predict logCRP. Tables 4 through 7 summarize the results of the four analyses that were conducted. The models that included education and BMI, education and WC, income and BMI, and income and WC were all statistically significant predictors of l ogCRP (see Tables 4 through 7). All models supported the first and second hypotheses, indicating significant main effects for education, income, BMI, and WC. In addition, the anal yses suggested an interaction between education and BMI, education and WC, and income and BMI. All models that included the main effects (i.e., step 2) accounted for a signi ficant amount of variance. Adding significant interaction terms resulted in statis tically significant increases in R 2 though the changes were small (see Tables 4 through 7).

PAGE 43

34 Table 4 Summary of Hierarchical Regression Analysis for Education and BMI Predicting LogCRP Variable B SE B R 2 F (10, 3681) p Step 2 0.137*** 93.50 <0.001 Education 0.028 0.020 0.023 BMI 0.069 0.003 0.381*** Step 3 0.001* 85.52 <0.001 Education 0.225 0.092 0.183* BMI 0.054 0.008 0.296*** Education x BMI 0.007 0.003 0.179* Note. All steps contain the covariates (gender, age, eth nicity, use of blood pressure medication, use of cholesterol medication and smoking) variables prese nted are those that are relevant to the hypotheses. BMI = body mass index (kg/m 2 ). R 2 = 0.066, F (8,3683) = 32.25, p < 0.001 for Step 1; R 2 = 0.203 for Step 2; R 2 = 0.204 for Step 3. p < .05, *** p < .001 Table 5 Summary of Hierarchical Regression Analysis for Education and WC Predicting LogCRP Variable B SE B R 2 F (10, 3621) p Step 2 0.161*** 105.36 <0.001 Education 0.025 0.020 0.020 WC 0.031 0.001 0.426*** Step 3 0.001* 96.36 <0.001 Education 0.308 0.127 0.250* WC 0.024 0.003 0.337*** Education x WC 0.003 0.001 0.243* Note. All steps contain the covariates (gender, age, eth nicity, use of blood pressure medication, use of cholesterol medication and smoking) variables prese nted are those that are relevant to the hypotheses. WC = waist circumference (cm). R 2 = 0.065, F (8, 3623) = 31.38, p < 0.001 for Step 1; R 2 = 0.225 for Step 2; R 2 = 0.227 for Step 3. p < .05, *** p < .001

PAGE 44

35 Table 6 Summary of Hierarchical Regression Analysis for Income and BMI Predicting LogCRP Variable B SE B R 2 F (10, 3434) p Step 2 0.136*** 86.53 <0.001 Income -0.021 0.006 -0.060*** BMI 0.069 0.003 0.377*** Step 3 0.002** 79.75 <0.001 Income -0.102 0.027 -0.293*** BMI 0.050 0.007 0.272*** Income x BMI 0.003 0.001 0.256** Note. All steps contain the covariates (gender, age, eth nicity, use of blood pressure medication, use of cholesterol medication and smoking) variables prese nted are those that are relevant to the hypotheses. BMI = body mass index (kg/m 2 ). R 2 = 0.065, F (8, 3436) = 29.85, p < 0.001 for Step 1; R 2 = .201 for Step 2; R 2 = .204 for Step 3. ** p < .01, *** p < .001 Table 7 Summary of Hierarchical Regression Analysis for Income and WC Predicting LogCRP Variable B SE B R 2 F (10, 3378) p Step 2 0.158*** 96.96 <0.001 Income -0.018 0.006 -0.050** WC 0.031 0.002 0.421*** Step 3 0.001 88.51 <0.001 Income -0.083 0.037 -0.240* WC 0.026 0.003 0.357*** Income x WC 0.001 0.000 0.199 Note. All steps contain the covariates (gender, age, eth nicity, use of blood pressure medication, use of cholesterol medication and smoking) variables prese nted are those that are relevant to the hypotheses. WC = waist circumference (cm). R 2 = 0.065, F (8, 3380) = 29.18, p < 0.001 for Step 1; R 2 = 0.223 for Step 2; R 2 = 0.224 for Step 3. p < .05, *** p < .001

PAGE 45

36 Regression lines were plotted to gain a better understanding of the nature of the interactions. Figures 1 and 2 demonstrate the effect of education and income on logCRP moderated by BMI. The regression lines illustrate the simple slopes and sim ple intercepts for normal (BMI = 21.70), overweight (BMI = 27.45), and obese (BMI = 35) individuals. The BMI chosen to represent each group is the median BMI value for each category. Figure 3 illustrates the effect of education on logCRP moderated b y WC. The regression lines illustrate the simple slopes and simple intercepts for indivi duals with high (106.7 cm) and low (86.0 cm) WC. The values of WC chosen represent the 75 th and 25 th percentile of WC in the SES sample, respectively. They are also above/below t he recommended maximum WC for males (101.6 cm) and females (88.9 cm).

PAGE 46

37 BMI & Education Interaction-4.1 -3.6 -3.1 -2.6 -2.1 -1.6 -1.1 -0.6 -0.1 01234 logCRP Normal* Overweight Obese Figure 1. BMI moderates the effect of education on logCRP concentrations. Values for Education: 1 = less than high school; 2 = high school or GED; 3 = more than high school. Follow-up analyses indicated that the slope for the normal BMI group was signifi cantly different from zero ( B = -0.072, t (3680) = -2.540, p < 0.05). However the slopes for the overweight and obese groups were not significantly different from zero ( B = -0.032, t (3680) = -1.586, p > 0.05; B = 0.021, t (3680) = 0.704, p > 0.05, respectively). *slope is significantly different than zero.

PAGE 47

38 B M I & I n c o m e I n t e r a c t i o n 4 6 3 6 2 6 1 6 0 6 0 2 4 6 8 1 0 1 2 I n c o m e logCRPP N o r m a l O v e r w e i g h t O b e s e Figure 2. BMI moderates the effect of income and BMI on logCRP concentrations. Values for Income represent the income brackets described in the methods secti on: 1 = $0 to $4, 999; 2 = $5,000 to $9,999; 3 = $10,000 to $14,999; 4 = $15,000 to $19,999; 5 = $20,000 to $24,999; 6 = $25,000 to $34,999; 7 = $35,000 to $44,999; 8 = $45,000 to $54,999; 9 = $55,000 to $64,999; 10 = $65,000 to $74,999; 11 = $75, 000 and over. Follow-up analyses indicated that the slopes for the normal BMI and overweight gr oups were significantly different from zero ( B = -0.039, t (3433) = -4.857, p < 0.01; B = -0.023, t (3433) = -4.047, p < 0.01, respectively). However the slope for the obese group was not significantly different from zero ( B = -0.001, t (3433) = -0.095, p > 0.05). *slope is significantly different than zero.

PAGE 48

39 WC & Education Interaction -3.6 -2.6 -1.6 -0.6 01234 logCRPP High WC Low WC* Figure 3. WC moderates the effect of education on logCRP concentrations. Education Values: 1 = less than high school; 2 = high school or GED; 3 = more than high school. Follow-up analyses indicated that the slope for the low WC group was significantl y different from zero ( B = -0.061, t (3620) = -2.386, p < 0.05). However, the slope for high WC group was not significantly different from zero ( B = -0.002, t (3620) = -0.103, p > 0.05). *slope is significantly different than zero.

PAGE 49

40 It was hypothesized that, when faced with more stress (i.e, lower SES), those wi th more adiposity would demonstrate a greater increase in concentrations of CRP than those with less adiposity. Interestingly, this is not the pattern present in the data ( see Figures 1 – 3). Across all analyses of the SES sample, the simple slope of the regression li ne for those with the highest adiposity is not significantly different from zero, sugge sting no effect of education or income on logCRP for this group. Conversely, those with lower adiposity demonstrate the expected negative relationship between education, incom e and logCRP. Thus, the inflammatory marker, CRP, in those with less adiposity appears to be more strongly related to the SES variables. Depression Variables and Adiposity Variables as Predictors of CRP Descriptive statistics. Means and standard deviations of logCRP, BMI, WC, and number of depression symptoms for the Depression sample are presented in Table 8. As previously mentioned, 87.44% of the participants in this sample reported no symptoms of depression. In addition, 550 (93.38%) of the participants did not meet criteria for major depressive disorder. Table 8 Mean Values of Variables in the Depression Analyses Variable M SD logCRP -2.00 1.18 BMI 27.25 6.04 WC 93.52 15.36 Depression Symptoms 0.78 2.15 Note. logCRP = logarithmic transformation of C-reactive protein, BMI = body mass index (kg/m 2 ), WC = waist circumference (cm)

PAGE 50

41 Correlations. Pearson correlation coefficients indicate significant positive relationships between logCRP and the adiposity variables. Further, the two adiposi ty variables (BMI and WC) were related to each other as were the two psychologi cal stress variables (depression diagnosis and number of depression symptoms). In addition, there was a small positive correlation between number of depression symptoms and BMI ( See Table 9). Table 9 Correlations between Variables in the Depression Analyses Log CRP BMI WC Income Education Depression Diagnosis Depression Symptoms logCRP 1.00 0.438*** 0.428*** -0.079 -0.031 0.061 0.080 BMI 1.00 0.909*** -0.120** -0.113** 0.048 0.088* WC 1.00 -0.064 -0.073 0.024 0.063 Income 1.00 0.338*** -0.077 -0.079 Education 1.00 -0.022 -0.003 Depression Diagnosis 1.00 0.767*** Depression Symptoms 1.00 Note. CRP = logarithmic transformation of C-reactive pro tein, BMI = body mass index (kg/m 2 ), WC = waist circumference (cm). p < 0.05, ** p < 0.01, *** p < 0.001

PAGE 51

42 Hierarchical linear regression analyses. Hierarchical linear regressions were conducted to determine the ability of the depression indices (depression diagnosis and depression symptoms) and the adiposity variables (BMI and WC) to predict logCR P. Tables 10 through 13 summarize the results of the four analyses that were conducte d (depression diagnosis and BMI, depression diagnosis and WC, depression symptoms and BMI, and depression symptoms and WC). These analyses indicate that both BMI and WC are significantly related to logCRP. However, neither depression symptom s nor depression diagnosis not significantly predicted logCRP in any analysis. Ther e were also no significant interactions between the depression and adiposity variables.

PAGE 52

43 Table 10 Summary of Hierarchical Regression Analysis for Depression Diagnosis and BMI Predicting LogCRP Variable B SE B R 2 F (10, 572) p Step 2 0.189*** 21.79 <0.001 Depression Diagnosis 0.175 0.171 0.037 BMI 0.087 0.007 0.444*** Step 3 0.000 19.80 <0.001 Depression Diagnosis 0.443 0.626 0.094 BMI 0.088 0.008 0.450*** Depression Diagnosis x BMI -0.010 0.021 -0.060 Note. All steps contain the covariates (gender, age, eth nicity, use of blood pressure medication, use of cholesterol medication and smoking) variables prese nted are those that are relevant to the hypotheses. BMI= body mass index (kg/m 2 ). R 2 = 0.087, F (8, 574) = 6.80, p < 0.001 for Step 1; R 2 = 0.276 for Step 2; R 2 = 0.276 for Step 3. *** p < .001 Table 11 Summary of Hierarchical Regression Analysis for Depression Diagnosis and WC Predicting LogCRP Variable B SE B R 2 F (10, 568) p Step 2 0.201*** 22.72 <0.001 Depression Diagnosis 0.206 0.172 0.043 WC 0.036 0.003 0.463*** Step 3 0.000 20.64 <0.001 Depression Diagnosis 0.569 0.966 0.119 WC 0.036 0.003 0.467*** Depression Diagnosis x WC -0.004 0.010 -0.07 7 Note. All steps contain the covariates (gender, age, eth nicity, use of blood pressure medication, use of cholesterol medication and smoking) variables prese nted are those that are relevant to the hypotheses. WC=Waist Circumference (cm). R 2 = 0.085, F (8, 570) = 6.65, p < 0.001 for Step 1; R 2 = 0.286 for Step 2; R 2 = 0.286 for Step 3. *** p < .001

PAGE 53

44 Table 12 Summary of Hierarchical Regression Analysis for Depression Symptoms and BMI Predicting LogCRP Variable B SE B R 2 F (10, 572) p Step 2 0.189*** 21.79 <0.001 Depression Symptoms 0.020 0.020 0.037 BMI 0.087 0.007 0.443*** Step 3 0.000 19.78 <0.001 Depression Symptoms 0.010 0.079 0.019 BMI 0.086 0.008 0.441*** Depression Symptoms x BMI 0.0003 0.003 0.019 Note. All steps contain the covariates (gender, age, eth nicity, use of blood pressure medication, use of cholesterol medication and smoking) variables prese nted are those that are relevant to the hypotheses. BMI = body mass index (kg/m 2 ). R 2 = 0.087, F (8, 574) = 6.80, p < 0.001 for Step 1; R 2 = 0.276 for Step 2; R 2 = 0.276 for Step 3. *** p < .001 Table 13 Summary of Hierarchical Regression Analysis for Depression Symptoms and WC Predicting LogCRP Variable B SE B R 2 F (10, 548) p Step 2 0.207*** 22.04 <0.001 Depression Symptoms 0.022 0.020 0.041 WC 0.036 0.003 0.466*** Step 3 0.000 20.00 <0.001 Depression Symptoms 0.001 0.118 0.002 WC 0.035 0.003 0.463*** Depression Symptoms x WC 0.0002 0.001 0.039 Note. All steps contain the covariates (age, ethnicity, use of blood pressure medication, use of cholestero l medication and smoking) variables presented are tho se that are relevant to the hypotheses. WC = Waist Circumference (cm). R 2 = 0.080, F (7, 551) = 6.84, p < .0001 for Step 1; R 2 = .287 for Step 2; R 2 = .287 for Step 3. *** p < .001.

PAGE 54

45 Gender as a Moderator of the relationship between Stress Variables and CRP Analyses were conducted to test the moderating effect of gender on the relationships between stress and adiposity and inflammation. Hierarchical r egressions were conducted to determine whether there were interactions between the str ess variables, adiposity variables, and CRP. The results did not suggest a three-way interaction between gender, stress, and adiposity. However, they did indicate tha t gender moderates the effects of the stress variables on levels of CRP and adiposity on levels of CRP (see Appendix B). Hierarchical linear regressions were conducted to further examine the moder ating effect of gender on the relationship between the stress variables and CRP, control ling for adiposity. Tables 14 through 17 summarize the four analyses (gender and education, gender and income, gender and depression diagnosis, and gender and depression symptoms). All analyses indicate significant main effects for gender with females having higher logCRP values than males. In addition, there was a significant main e ffect for income ( t (1) = -3.37, p < 0.01) in the hypothesized directions. There was also a significant interaction between gender and income ( t (1) = 2.42, p < 0.05). The addition of this interaction resulted in a small, yet statistically significant increase in variance accounted for ( R 2 = 0.001; t (3380) = 6.087, p < 0.05).

PAGE 55

46 Table 14 Summary of Hierarchical Regression Analysis for Gender and Education Predicting LogCRP Variable B SE B R 2 F (11,3611) p Step 2 0.037*** 96.31 <0.001 Gender 0.448 0.034 0.215*** Education -0.027 0.020 -0.022 Step 3 0.000 41.73 <0.001 Gender 0.351 0.086 0.169*** Education -0.092 0.057 -0.075 Education x Gender 0.044 0.036 0.074 Note. All steps contain the covariates (age, ethnicity, use of blood pressure medication, use of cholestero l medication, smoking, body mass index, and waist cir cumference) variables presented are those that are relevant to the hypotheses. R 2 = 0.190 for Step 1, F (9,3613) = 93.94, p < .0001; R 2 = .227 for Step 2; R 2 = .227 for Step 3. *** p < .001. Table 15 Summary of Hierarchical Regression Analysis for Gender and Income Predicting LogCRP Variable B SE B R 2 F (11,3368) p Step 2 0.038*** 88.51 <0.001 Gender 0.429 0.036 0.206*** Income -0.018 0.006 0.053*** Step 3 0.002* 81.74 <0.001 Gender 0.252 0.081 0.121** Income -0.057 0.017 0.164*** Gender x Income 0.026 0.011 0.141* Note. All steps contain the covariates (age, ethnicity, use of blood pressure medication, use of cholestero l medication, smoking, body mass index, and waist cir cumference) variables presented are those that are relevant to the hypotheses. R 2 = 0.186 for Step 1, F (9,3370) = 85.56, p < .0001; R 2 = .224 for Step 2; R 2 = .226 for Step 3. p < .05, ** p < .01, *** p < .001.

PAGE 56

47 Table 16 Summary of Hierarchical Regression Analysis for Gender and Depression Diagnosis Predicting LogCRP Variable B SE B R 2 F (11,567) p Step 2 0.069*** 21.01 <0.001 Gender 0.647 0.089 0.273*** Depression Diagnosis 0.193 0.172 0.040 Step 3 0.001 19.35 <0.001 Gender 0.670 0.092 0.282*** Depression Diagnosis 0.757 0.568 0.158 Gender x Depression Diagnosis -0.360 0.345 -0. 125 Note. All steps contain the covariates (age, ethnicity, use of blood pressure medication, use of cholestero l medication, smoking, body mass index, and waist cir cumference) variables presented are those that are relevant to the hypotheses. R 2 = 0.221, F (9, 569) = 17.93, p < .001 for Step 1; R 2 = 0.290 for Step 2; R 2 = 0.291 for Step 3. *** p < .001. Table 17 Summary of Hierarchical Regression Analysis for Gender and Depression Symptoms Predicting LogCRP Variable B SE B R 2 F (11,567) p Step 2 0.068*** 20.97 <0.001 Gender 0.645 0.089 0.272*** Depression Symptoms 0.019 0.020 0.035 Step 3 0.002 19.33 <0.001 Gender 0.678 0.094 0.286*** Depression Symptoms 0.090 0.067 0.163 Gender x Depression Symptoms -0.045 0.040 -0.1 35 Note. All steps contain the covariates (age, ethnicity, use of blood pressure medication, use of cholestero l medication, smoking, body mass index, and waist cir cumference) variables presented are those that are relevant to the hypotheses. R 2 = 0.221, F (9, 569) = 17.93, p < .001 for Step 1; R 2 = 0.289 for Step 2; R 2 = 0.291 for Step 3. *** p < .001.

PAGE 57

48 Regression lines were plotted to gain a better understanding of the nature of t he interaction between gender and income (see Figure 4). The slope of the regres sion line for females did not significantly differ from zero ( B = -0.006; t (3367) = -0.722, p > 0.05), suggesting no effect of income on logCRP for this group. Conversely, males demonstrat e the expected positive relationship between income and logCRP ( B = -0.031, t (3367) = 4.067, p < 0.01). Gender & Income Interaction-3.6 -2.6 -1.6 -0.6 024681012 logCRPP Male* Female Figure 4. Gender moderates the relationship between income and concentration of logCRP. Values for Income represent the income brackets described in the m ethods section: 1 = $0 to $4, 999; 2 = $5,000 to $9,999; 3 = $10,000 to $14,999; 4 = $15,000 to $19,999; 5 = $20,000 to $24,999; 6 = $25,000 to $34,999; 7 = $35,000 to $44,999; 8 = $45,000 to $54,999; 9 = $55,000 to $64,999; 10 = $65,000 to $74,999; 11 = $75, 000 and over. *slope is significantly different than zero.

PAGE 58

49 Gender as a Moderator of the Relationship between Adiposity and CRP Hierarchical linear regressions were conducted to determine whether g ender moderates the relationship between adiposity and CRP, controlling for stress var iables (i.e., education, income, depression diagnosis, and depression symptoms). Analyses in both the SES and depression samples indicate significant main effects for gender and BMI in step 2. However, the interaction term is not significant when entered into t he model in step 3 (see Appendix C, Tables C1 and C2). The analyses that examine the interaction between gender and WC also show significant main effects of gender a nd WC in step 2. In addition, the results indicate a significant, but small, interaction bet ween gender and WC ( t (1) = 1.98, p < 0.05) in the depression sample and an effect that approached significance ( t (1) = 1.91, p = 0.056) in the SES sample. However, the addition of the gender by WC interaction does not significantly increase the R 2 in any model (see Appendix C, Tables C3 and C4). The Relationship of BMI versus WC to CRP It has been suggested that BMI and WC may have different effects on health and inflammation (Bjorntorp, 1988). Therefore, the effects of BMI and WC on CRP were compared. In the SES sample, the correlation between WC and logCRP was slightly higher than the correlation between BMI and logCRP (see Table 3). However, i n the Depression sample the correlation between BMI and logCRP was slightly hi gher than between WC and logCRP (see Table 9). In both cases the differences are very small. To further explore this question, regression analyses were conducted. The analys es controlled for all other variables in step one. In step two BMI or WC was added to the model and in step three the other adiposity measure was included (i.e., WC or BMI). In

PAGE 59

50 the SES sample, each measure of adiposity accounted for significantly more var iance when entered in to the equation. The addition of WC to a model including BMI resulted in R 2 = 0.0194 ( F (1,3611) = 90.30, p < 0.01). The addition of BMI to a model containing WC resulted in R 2 = 0.0012 ( F (1,3611) = 5.59, p = 0.018). In the Depression sample the addition of WC resulted in a significantly higher amount of variance accounted for ( R 2 = 0.017; F (1,527) = 13.01, p < 0.01) but the addition of BMI did not ( R 2 = 0.003; F (1,527) = 1.96, p = 0.163). This suggests that WC, a measure of central adiposity as opposed to overall adiposity, may have a slightly stronger relationship with CRP.

PAGE 60

51 Discussion Stress and Inflammation The first goal of the study was to determine the relationship between stre ss and inflammation. Low SES has been identified as a chronic stressor (e.g., Kessle r, 1979; Turner, Wheaton, & Lloyd, 1995) and has been associated with increased rates of morbidity and mortality (Adler et al., 1994). Two major components of SES, educational attainment and income level, were examined as proxies for stress in this stud y. Results indicated that those with lower SES had higher concentrations of CRP. These finding s are consistent with previous studies that found higher levels of CRP in individuals in lower social classes compared to their more affluent counterparts (e.g., McDa de et al., 2006; Owen et al., 2003; Panagiotakos et al., 2004). Stress has also been conceptualized as a psychological reaction to a stress or (Mason, 1975). Further, depressive symptoms and major depressive disorder have been associated with the occurrence of stressors (Holahan et al., 1999; Kendler et al ., 1998; Kendler et al., 1999). Thus, these constructs were employed to gain a better understanding of the inflammatory responses to such psychological responses. These analyses did not suggest a relationship between depression and CRP; this is contrary to previous research that has identified relationships between these variables (e .g., Danner et al., 2004; Miller et al., 2002, Owen & Steptoe, 2003; Suarez, et al., 2003). There are multiple explanations for the difference in findings between previous studies and the current one. For example, in a sample of well-functioning elderly

PAGE 61

52 individuals (70 – 79 years of age, M = 73.6 years, SD = 2.8), Penninx et al. (2003) found a significant difference in CRP, IL-6, and TNFbetween those who scored above the clinical cut-off (i.e., 16) on the Center for Epidemiological Studies Depressi on (CES-D) scale and those with scores below the cut-off. Similarly, Vetta et al. (2001) found differences in a number of biological markers, including TNFand CRP concentrations, in depressed compared to control participants in a sample of individuals ranging in a ge from 65 to 94 years ( M = 80.1, SD = 12.4). Conversely, the depression sample in the current study consisted of those 20 to 39 years of age, with a mean age of 28.97 years ( SD = 5.74). Age is positively associated with inflammation (Suarez, 2004), therefore the disparity in ages between the samples may account for the difference in finding s. These data, along with the findings of the present study, suggest that the relationship bet ween CRP and depression may only be present in the elderly. The lack of a relationship between the measures of depression and CRP may also be due to the measure of depression that was employed in this study. The CIDI aske d participants about their depressive episodes over the previous year. The data did not indicate how recently the depressive episode occurred, therefore it is possible t hat participants were recalling depressive episodes that they experienced m onths before and had since recovered. Danner et al. (2003) reported that participants who experience d a depressive episode more than six months before measurement of CRP exhibited leve ls that are similar to those who have never had a depressive episode. Thus, it is possible that the participants in this study had previously experienced a depressive epis ode accompanied by increased levels of CRP, which returned to normal following recovery from the episode.

PAGE 62

53 Although this study did not find a relationship between CRP and depression, one cannot rule out the possibility that inflammation is related to depression. The inflammatory response is very complex and involves many cytokines, proteins, and catecholamines. Therefore, the lack of a relationship between depression and CRP in this sample may not be generalizable to the entire inflammatory process. That i s, inflammation may still be increased in those with depression, through mechanisms other than CRP. In fact, other studies that have examined multiple biomarkers have fail ed to find a relationship between depression and CRP, though they have found relationships with other inflammatory markers (Joyce et al., 1992; Tuglu et al., 2003). Adiposity and Inflammation A second goal of this study was to explore the relationship between adiposity and inflammation. Previous studies have suggested that adipose tissue is involved in secretion of inflammatory markers such as IL-6, TNF, and CRP (Calabro et al., 2005; Lemieux et al., 2001; Mohamed-Ali et al., 1997; Mohamed-Ali et al., 1998; Owen & Steptoe, 2003; Visser et al., 1999; Yudkin et al., 1999). Therefore it was expected that those with more adipose tissue would have higher levels of such inflammatory marke rs. Indeed, across all analyses there were significant associations betwe en measures of adiposity (both BMI and WC) and CRP concentration, which is consistent with the findings of other researchers (e.g., Owen & Steptoe, 2003; Kern et al., 1995; Lemieux et al., 2001; McDade et al., 2006; Mohamed-Ali et al., 1997; Visser et al., 1999; Yudkin et al., 1999). Further, it has been suggested that central adiposity is a better predictor of obesity-related diseases than overall adiposity (Bjorntorp, 1988). Likewise, some

PAGE 63

54 researchers have found that WC is better at predicting CRP than BMI and total bod y fat percentage (McDade et al., 2006). However, others have found a relationship between CRP and BMI, but not waist-to-hip ratio, particularly in samples of depressed parti cipants (Miller et al., 2002). While the data from the current study suggest that WC may be a slightly better predictor of CRP, the differences were very small. Thus, they did not provide convincing evidence that WC is more influential than BMI. Both measures wer e related to inflammation and appear to be useful predictors of the inflammatory re sponse. Interactions between the Stress and Adiposity Variables The third goal of this study was to investigate the proposed model, which integrates the effects of stress and the effects of obesity on the inflamma tory system. Given that adipose tissue secretes pro-inflammatory cytokines and CRP and tha t stress is related to increases in inflammation, it was hypothesized that adiposity woul d moderate the effect of stress on inflammation. The proposed model suggested that individuals with more adiposity would be more affected by stress, demonstrating greate r changes in levels of CRP under stress than those with less adiposity. Analysis of the data revealed interactions between the SES variables and adiposity but no interactions between the depression variables and adiposity. Interestingly, the nature of the interactions differed from the expected pat tern (see Figures 1 – 3). The plots of the regression lines indicated that adiposity was posit ively related to higher CRP. However, while CRP was not related to income or education levels among those with higher levels of adiposity, individuals with lower levels of adiposity demonstrated the expected relationship. That is, they had lower levels of inflammation than those with more adiposity and their level of CRP was positively

PAGE 64

55 related to the amount of stress they were experiencing, as measured by two indic es of SES, income and education. These findings may suggest that under less stress (i.e., higher SES), inflam mation in those with more adiposity remains at levels that are similar to those obser ved in high stress situations. This is consistent with research that shows that having too m uch adiposity is detrimental to health. That is, even in non-stressful situations, those w ith high levels of adiposity exhibit levels of inflammation that are similar to thos e seen in people living in stressful conditions. This explanation lends credence to theories tha t suggest that physiological factors may be more important than psychosocial fa ctors in determining health. Accordingly, if a person has an abundance of adipose tissue, st ress may not dramatically impact the level of inflammation. In terms of the propose d model it may suggest that adipose tissue is constantly secreting inflammatory mark ers to the point where a ceiling is reached, preventing stress from further influencing t he levels of these markers. This interpretation highlights the importance of successful weight ma nagement before stress management can be productive at reducing inflammation because a hi gh level of adiposity may overshadow any effect of stress management on inflamm ation. However, having low levels of adiposity is not enough to ensure healthy levels of CRP. Those with less adiposity are affected by chronic stress, making stress ma nagement more important once weight is at healthy levels. An alternative explanation emphasizes the psychological factors that are associated with obesity. Some have suggested that the state of being obese is i tself a stressor. In a national sample of adults, Carr and Friedman (2005) found that compare d to participants who were in the normal weight range, very obese subjects (BMI 35)

PAGE 65

56 reported less self-acceptance, more discrimination, and poorer overall health. The obese and very obese participants were 40 to 50% more likely to report experiences of major discrimination than normal weight participants. It is possible that the stres s of being obese may be more detrimental than the stress of living at low SES. It may be t hat the level of inflammation is higher across all obese participants regardless of their SES because they are all faced with equal stressors that because of their weig ht. The stress of living at low SES may not add a significant amount of stress on top of being obese. This interpretation also suggests the utility of weight management as a way to m anage stress, improve psychological well-being and lower levels of inflammation. Unfortunately the validity of this explanation is not clear. In this sample the re were no significant correlations between the measures of adiposity and depressi on diagnosis and there was only a small correlation between depression symptoms and BMI. Thus, those with more adiposity did not appear to be more depressed. Further, there wer e no correlations between the SES variables and the adiposity variables. The lack of relationships between the adiposity and stress variables suggests that those wit h more adiposity may not be experiencing more stress. Future research would benefit fr om measuring and controlling for these types of stress and stigmatization to de termine the pure effect of SES. Gender A secondary goal of this study was to examine the relationship of gender to inflammation. The only previous study that examined the interaction between obesity and depression on inflammation studied only men (Ladwig et al., 2003). Therefore, it was not clear that the results could be generalized to women. In addition, other studies

PAGE 66

57 have suggested that depression is related to increased levels of CRP in men but not women (Danner et al., 2003; Ford & Erlinger, 2004; Liukkonen et al., 2006). However, others have not found gender differences in the effect of SES on CRP (Owen et al., 2003). Given the conflicting data regarding the effects of gender and stress on CRP, analyses were conducted to determine the role of gender in these data. All anal yses that included gender demonstrate a main effect of gender, indicating that female s have higher levels of CRP than males, which is consistent with previous research (McDade, 2006). There were no significant gender x stress x adiposity interactions, suggest ing that the relationships between stress and adiposity do not differ across genders. However, there was a small but significant two-way interaction between gende r and income, indicating that the relationship between income and CRP is negative for males, but not females. This is contrary to the findings of Owen et al. (2003) but consistent with previously reported data on the effect of depression on CRP concentrations (Danner et al., 2003; Ford & Erlinger, 2004; Liukkonen et al., 2006). Interestingly, the interactions between depression variables and gender w ere not significant. This analysis suggests that CRP levels in males and females may be differentially affected by stress, underscoring the need to examine inf lammatory responses in males and females separately rather than simply controlling f or gender. It also suggests that care should be taken when generalizing results, as the relati onships seen in males may not be present in females.

PAGE 67

58 Strengths of the Study This study had several important strengths. The size of the sample indicates a high degree of external validity. The analyses were conducted using a large, na tionallyrepresentative sample, making the results generalizable to the population of the Uni ted States. The analyses that are discussed are those that did not utilize the sampl ing weights. However, the results were similar when the sample weights were used, demonstrating the generalizability of the data even without the use of the weight s. The size of the sample also allows for adequate representation of minorities and indivi duals at all levels of SES. Another strength of this study is that it included measures of two types of str ess: SES (including education and income) and depression (including symptoms and diagnosis). The ability to look at multiple stressors and psychological factors allowed for more specific knowledge of how stress affects CRP levels and interacts wi th adiposity. In addition, the variables were treated as continuous. Researchers often spli t SES variables into high and low groups, losing important information about the gradation of the effect. By using continuous variables in the current analysis, it was possi ble to determine the effects of stress and adiposity on CRP at all values of the variable s. These analyses also statistically controlled for factors that are kno wn to be associated with CRP, SES, and adiposity. By controlling for race, smoking, age, gende r and use of medications, the effects of stress and adiposity on CRP independent of these confounds could be identified. In addition, those who had levels of CRP that are indicative of active infections were eliminated from the analyses. This ensur ed that the

PAGE 68

59 effects that were seen in these data are common in normal conditions and are not accounted for by another infection or acute inflammatory reaction. Limitations of the Study Unfortunately, there were also a number of limitations to this study. Due to the correlational nature of this study, one cannot rule out the presence of an unknown third variable influencing the stress or adiposity variables and CRP. These variabl es, and CRP in particular, are known to be affected by many factors. While this study statis tically controlled for the effects of race, age, gender (where appropriate), use of blood pr essure and cholesterol medication, and smoking there are other factors that may have influenc ed the relationship between CRP, adiposity and stress. For example, other researche rs have controlled for alcohol intake, history of coronary heart disease, cancer, psychologi cal illnesses, endocrinological illnesses, use of other anti-inflammatory drug s, and season (e.g., Owen et al., 2003). While, the elimination of individuals with high concentrations of CRP (>1 mg/dL) should have reduced the effects that other illnesses had on the dat a, we did not have access to information on these factors, so their influence could not be directly controlled. The nature of the study also introduced some limitations to the analyses. This study was conducted on data obtained from a national epidemiological study. The primary goal of the larger study was to gain a better understanding of Americ ans’ health and nutritional status and there were no direct measures of stress available in t he dataset. Therefore, two proxies for stress were used (SES and depression). It is possible that a direct measure of stress would have provided different information than those measure s that were available. Regardless, this study provided valuable insight into the eff ects of

PAGE 69

60 SES and depression on CRP levels. Both of these factors have been associated with increased risk for cardiovascular disease (eg., Adler & Snibbe, 2003; Frasure-Sm ith & Lesprance, 2005). Given the relationship between CRP and cardiovascular disease ( e.g., Albert, Ma, Rifai, Stampfer, & Ridker, 2002; Danesh et al., 1998; Mendall et al., 2000; Ridker et al., 2000), it is important to understand how SES and depression affect CRP. Similarly, only one measure of inflammation was available in the dataset. CR P is only one of a number of elements involved in the inflammatory process. It is possible that a more comprehensive examination of multiple inflammatory markers would ha ve provided more information about the effect of adiposity and stress on inflammation. Also, some researchers have found relationships between stress and other marker s of inflammation, such as haptogloblin, -l-antichymotrypsin, IL-6 and TNF, but not with CRP (e.g., Grossi et al., 2003; Joyce et al., 1992; Sutherland et al., 2003; Tuglu et al., 2003). However, CRP is believed to be one of the best markers of risk for cardiovascular disease (Albert et al., 2002; Cushman, Stampfer, Tracy, & Hennekens, 1997; Danesh, Collins, Appleby, & Peto, 1998; Pearson et al., 2003; Ridker et al., 2000), therefore, it is important that we understand the influences on this inflammatory marker in particul ar. Another constraint on the current analysis was that the raw SES data was not made available to the public. Only information regarding general level of educ ation was provided, i.e. if the individual had less than a high school education, a high school education, or more than a high school education. In addition, the income data was also only available in income brackets. Use of categories may have masked the ef fects of SES on CRP. If this data could have been analyzed in its more detailed form (i.e., exact income or number of years of education), it may have provided better information about

PAGE 70

61 the relationships between education, income and CRP. However, these multiple categories provide a more continuous variable than analyses in which the SES data wa s divided into only high and low groups. Therefore, though not ideal, this type of data was useful in illustrating the gradation of CRP concentrations across levels of SE S. The depression analyses also suffered due to the design of the study. Some studies that have found CRP and depression relationships have sampled equal groups of depressed and non-depressed, whereas the current study was population-based. While this increased external validity, it also decreased the proportion of depressed pa rticipants. Only 6.62% ( n = 39) of the sample met the criteria for major depression. In addition, only 12.56% ( n = 74) of the participants reported one or more symptoms of depression. If a sample of depressed participants were specifically selected for this study a more normal distribution of symptom frequency may have been obtained, which may have affected the results. The scope of the CIDI may have also affected the data. The questionnaire asked about symptoms of depression over the previous year. The participants’ responses may have been subject to retrospective memory bias which may have affected their abi lity to accurately recall the events of a major depressive episode that occurred mont hs ago. As previously discussed, only a subsample of participants ages 20 to 39 years were aske d to complete the CIDI. This reduced sample of relatively young participants m ay have had a substantial impact on the findings. There was also no way to know if the participants were currently taking antidepressant medications. Lanquillon et al.’s (2000 ) study suggested that these medications impact levels of CRP. It is possible that t hose who had

PAGE 71

62 experienced depressive symptoms had been taking medication that affected the relationship between the symptoms and the CRP level. The correlational and cross-sectional nature of the analyses does not allow f or conclusions regarding causality. One cannot be sure if low SES causes changes i n CRP or if high CRP levels affect SES. There are multiple hypotheses regarding the di rection of the relationship between SES and health. There is a possibility that an increase in inflammation could inhibit an individual’s ability to achieve higher education or income However, the consensus in the literature seems to be that low SES leads to poorer healt h (Adler et al., 1994). The directionality of the relationship between depression and inflammation has raised even more questions and hypotheses. It is possible that the presence of depr ession symptoms results in an increase in inflammation. Indeed, researchers have de monstrated that alleviating depression lowers levels of inflammatory markers (La nquillon et al., 2000; Tuglu et al., 2003). However, others have found that increased inflammation can lead to depressed mood. Wright, Strike, Brydon and Steptoe (2005) injected participants with a low dose of S. typhi capsular polysaccharide vaccine or saline solution in the morning. The vaccination contained a level of antigen that was sufficient to cause a n inflammatory response (increase in IL-6) but not cause physical symptoms in the participants. They found that the participants who had been injected with the antigen reported a more negative mood throughout the day, compared to those who received saline. There was also a correlation between the amount of increase of IL6 and the change in mood, with those who had more IL-6 production showing an increase in measures of negative mood. It is likely that the relationship between depressi on and

PAGE 72

63 inflammation is even more complex. Miller and Blackwell (2006) present a model in which chronic stressors elicit an increase in inflammation. In response to the i ncreased inflammation, people may develop depressive symptoms and cardiovascular disease. The depressive symptoms and cardiovascular disease then participate in perpetuati ng the heightened inflammation, creating a positive feedback loop (Miller & Blac kwell, 2006). These contradictory studies and models that suggest bidirectional effects make it difficult to interpret the results of this study. There is obviously a need for more experi mental and longitudinal studies to tease apart the cause and effect or bidirectional relat ionships between stress, adiposity, and CRP. There were also some statistical limitations to the analyses. While a large sample increases reliability and external validity, one must keep in mind that it als o has its drawbacks. The large sample size may have overpowered the analyses such that e ven small effects could be statistically significant. In addition, multiple statistical tests were conducted, which may have inflated the alpha, resulting in an increase in type 1 errors. Another statistical difficulty was the non-normality of the depression dat a. The data was skewed due to the fact that many participants reported no symptoms of depressi on. Transformations of the data were unsuccessful at normalizing, causing the r esearcher to conduct the analyses on non-normal data. This may have reduced the confidence that one has in the results and conclusions. Conclusions The purpose of this study was to integrate two lines of research. There is substantial evidence suggesting that psychological stress is related to infl ammation (e.g. Hapuarachchi et al., 2003; Melamed et al., 2004; Owen et al., 2003) and that adiposity is

PAGE 73

64 also related to inflammation (e.g. Lemieux et al., 2001; Visser et al., 1999; Yudkin et al ., 1999). The proposed model led to the hypothesis that stress would cause adipose tissue to secrete more cytokines and APPs than it would under normal circumstances. Therefore, the relationship between stress and inflammation was expected to be moderated by adiposity. Indeed, adiposity did moderate the relationship between str ess and inflammation; however, not in the hypothesized direction. The obese participants i n this sample exhibited higher levels of inflammation than the normal weight partic ipants, regardless of income and education. In contrast, among normal weight individuals, income and education were related to CRP levels. While the relationship between SE S and CRP is different for obese and normal weight individuals, the reason for these differences is still unclear. Epidemiological data suggest that obesity is reaching epidemic proportions in t he United States in both children and adults (Hedley et al., 2004; Wyatt, Winters & Dubbert 2006). Further, high levels of adiposity have been linked to a number of chronic diseases (e.g., Burton et al., 1985; Hartz, Rupley, & Rimm, 1984; Ohlson et al., 1985). In addition, stress (specifically SES and depression) is known to be a risk factor for cardiovascular disease. It has been suggested that inflammation may be one pat hway through which adiposity and stress lead to these health complications. This study demonstrates the relationship between these variables and especially highlig hts the dramatic impact of adiposity on inflammation. While lowering stress may be beneficial at reducing CRP in individuals at healthy weights, it does not appear to be enough in those with high levels of adiposity. This is probably due to both the biological and psychological factors that are associated with obesity. Adipose cells rele ase

PAGE 74

65 inflammatory materials. Therefore, those with larger or a greater num ber of adipose cells may be releasing more of these dangerous substances. Further, obesity is a ssociated with psychological distress such as low self-acceptance and experiences of dis crimination (Carr & Friedman, 2005). These psychological factors possibly increase infl ammatory markers on top of any other stress with which these people are faced. This highlight s the importance of successful weight management to improve both physiological and psychological health.

PAGE 75

66 References Ackerman, K. D., Martino, M., Heyman, R., Moyna, N. M., & Rabin, B. S. (1998). Stressor-induced alteration of cytokine production in multiple sclerosis patient s and controls. Psychosomatic Medicine, 60, 484-491. Adler, N. E., Boyce, T., Chesney, M. A., Cohen, S., Folkman, S., Kahn, R. L., & Syme, S. L. (1994). Socioeconomic status and health: The challenge of the gradient. American Psychologist, 49, 15-24 Adler, N. E. & Snibbe, A. C. (2003). The role of psychosocial processes in explaining the gradient between socioeconomic status and health. Current Directions in Psychological Science, 12, 119-123. Albert, C. M., Ma, J., Rifai, N., Stampfer, M. J., & Ridker, P. M. (2002). Prospective study of C-reactive protein, homocysteine, and plasma lipid levels as predictors of sudden cardiac death. Circulation, 105, 2595-2599. American Psychiatric Association: Diagnostic and Statistical Manual of Mental Disorders Fourth Edition, Text Revision. Washington, DC, American Psychiatric Association, 2000. Bjorntorp, P. (1988). The associations between obesity, adipose tissue distribution and disease. Acta Medica Scandinavica, Suppl. 723, 121-134. Black, P. H. & Garbutt, L. D. (2002). Stress, inflammation and cardiovascular disease. Journal of Psychosomatic Research, 52, 1-23.

PAGE 76

67 Black, P. H. (2002). Stress and the inflammatory response: A review of neurogenic inflammation. Brain, Behavior, and Immunity, 16, 622-653. Black, P. H. (2003). The inflammatory response is an integral part of the stress re sponse: Implications for atherosclerosis, insulin resistance, type II diabetes a nd metabolic syndrome X. Brain, Behavior, and Immunity, 17, 350-364. Brydon, L., Edwards, S., Mohamed-Ali, V., & Steptoe, A. (2004). Socioeconomic status and stress-induced increases in interleukin-6. Brain, Behavior, and Immunity, 18, 281-290. Burton, B. T., Foster, W. R., Hirsch, J., & van Itallie, T. B. (1985). Health implications of obesity: an NIH consensus development conference. International Journal of Obesity, 9, 155-169. Calabro, P., Chang, D. W., Willerson, J. T., & Yeh, E. T. H. (2005). Release of Creactive protein in response to inflammatory cytokines by human adipocytes: Linking obesity to vascular inflammation. Journal of the American College of Cardiology, 46, 1112-1113. Carr, D. & Friedman, M. A. (2005). Is obesity stigmatizing? Body weight, perce ived discrimination, and psychological well-being in the United States. Journal of Health and Social Behavior, 46, 244-259. Danesh, J., Collins, R., Appleby, P., & Peto, R. (1998). Association of fibrinogen, creactive protein, albumin, or leukocyte count with coronary heart disease: Meta analyses of prospective studies. The Journal of the American Medical Association, 279, 1477-1482.

PAGE 77

68 Danner, M., Kasl, S. V., Abramson, J. L., & Vaccarino, V. (2003). Association between depression and elevated c-reactive protein. Psychosomatic Medicine, 65, 347356. Douglas, K. M., Taylor, A. J., & O’Malley. (2004). Relationship between depression and c-reactive protein in a screening population. Psychosomatic Medicine, 66, 679-683. Ford, D. E. & Erlinger, T. P. (2004). Depression and c-reactive protein in US adults: Data from the Third National Health and Nutrition Examination Survey. Archives of Internal Medicine, 164, 1010-1014. Frasure-Smith, N.& Lesperance, F. (2005). Depression and coronary heart diseas e: Complex synergism of mind, body and environment. Current Directions in Psychological Science, 14, 39 – 43. Fried, S. K., Bunkin, D. A., & Greenberg, A. S. (1998). Omental and subcutaneous adipose tissues of obese subjects release interleukin-6: Depot difference and regulation by glucocorticoid. Journal of Clinical Endocrinology and Metabolism, 83, 847-850. Gallo, L. C. & Matthews, K. A. (2003). Understanding the association between socioeconomic status and physical health: Do negative emotions play a role? Psychological Bulletin, 129, 10-51. Goebel, M. U., Mills, P. J., Irwin, M. R., & Ziegler, M. G. (2000). Interleukin-6 and tumor necrosis factorproduction after acute psychological stress, exercise, and infused Isoproterenol: Differential effects and pathways. Psychosomatic Medicine, 62, 591-598.

PAGE 78

69 Grossi, G., Perski, A., Evengard, B., Blomkvist, V., & Orth-Gomer, K. (2003). Physiological correlates of burnout among women. Journal of Psychosomatic Research, 55, 309-316. Hapuarachchi, J. R., Chalmers, A. H., Winefiled, A. H., & Blake-Mortimer, J. S. (2003). Changes in clinically relevant metabolites with psychological stress parameters. Behavioral Medicine, 29, 52-59. Hartz, A. J., Rupley, D. C., & Rimm, A. A. (1984). The association of girth measurement with disease in 32, 856 women. American Journal of Endocrinology, 119, 71-80. Hedley, A. A., Ogden, C. L., Johnson, C. L., Carroll, M. D., Curtin, L. R., Flegal, K. M. (2004). Prevalence of overweight and obesity among US children, adolescents, and adults, 1999-2002. The Journal of the American Medical Association, 291, 2847-2850. Heinz, A., Hermann, D., Smolka, M.N., Rieks, M., Graf, K., Pohlau, D., Kuhn, W., & Bauer, M. (2003). Effects of acute psychological stress on adhesion molecules, interleukins and sex hormones: Implications for coronary heart disease. Psychopharmacology, 165, 111-117. Hobfoll, S. E. (1989). Conservation of resources: A new attempt at conceptualization of stress. American Psychologist, 44, 513-524. Holahan, C. J., Moos, R. H., Holahan, C. K., Cronkite, R. C. (1999). Resource loss, resource gain, and depressive symptoms: A 10-year model. Journal of Personality and Social Psychology, 77, 620-629.

PAGE 79

70 Hotamisligil, G. S., Arner, P., Caro, J. F., Atkinson, R. L., & Spiegelman, B. M. (1995). Increased adipose tissue expression of tumor necrosis factorin human obesity and insulin resistance. The Journal of Clinical Investigation, 95, 2409-2415. Jiang, C., Ting, A. T., & Seed, B. (1998). PPARagonists inhibit production of monocyte inflammatory cytokines. Nature, 391, 82-86. Joyce, P. R., Hawes, C. R., Mulder, R. T., Sellman, J. D., Wilson, D. A., & Boswell, D. R. (1992). Elevated levels of acute phase plasma proteins in major depression. Biological Psychiatry, 32, 1035-1041. Kendler, K. S., Karkowski, L. M., & Prescott, C. A. (1998). Stressful life events and major depression: Risk period, long-term contextual threat, and diagnostic specificity. The Journal of Nervous and Mental Disease, 186, 661-669. Kendler, K. S., Karkowski, L. M., & Prescott, C. A. (1999). Causal relationship between stressful life events and the onset of major depression. American Journal of Psychiatry, 156, 837-841. Kern, P. A., Saghizadeh, M., Ong, J. M., Bosch, R. J., Deem, R., & Simsolo, R. B. (1995). The expression of tumor necrosis factor in human adipose tissue: Regulation by obesity, weight loss, and relationship to lipoprotein lipase. The Journal of Clinical Investigation, 95, 2111-2119. Kershaw, E. E. & Flier, J. S. (2004). Adipose tissue as an endocrine organ. The Journal of Endocrinology and Metabolism, 89, 2548-2556. Kessler, R. C. (1979). Stress, social status, and psychological distress. Journal of Health and Social Behavior, 20, 259-272.

PAGE 80

71 Ladwig, K. H., Marten-Mittag, B., Lowel, H., Doring, A., & Koenig, W. (2003). Influence of depressive mood on the association of CRP and obesity in 3205 middle aged healthy men. Brain, Behavior, and Immunity, 17, 268-375. Lanquillon, S., Krieg, J.-C., Bening-Abu-Schach, U., & Vedder, H. (2000). Cytokine production and treatment response in major depressive disorder. Neuropsychopharmacology, 22, 370-378. Libby, P., Ridker, P. M., & Maseri, A. (2002). Inflammation and atherosclerosis. Circulation, 105, 1135-1143. Liukkonen, T., Silvennoinen-Kassinen, S., Jokelainen, J., Rsnen, P., Leinonen, M., Meyer-Rochow, V. B., & Timonen, M. (2006). The association between c-reactive protein levels and depression: Results from the Northern Finland 1966 Birth Cohort Study. Biological Psychiatry, 60, 825-830. Lemieux, I., Pascot, A., Prud’homme, D., Almeras, N., Bogaty, P., Nadeau, A., Bergeron, J., & Despres, J. (2001). Elevated C-reactive protein: Another component of the atherothrombotic profile of abdominal obesity. Arteriosclerosis, Thrombosis, and Vascular Biology, 21, 961-967. Lutgendorf, S. K., Logan, H., Costanzo, E., & Lubaroff, D. (2004). Effects of acute stress, relaxation, and neurogenic inflammatory stimulus on interleukin-6 in humans. Brain, Behavior, and Immunity, 18, 55-56. Maes, M., Lin, A., Delmeire, L., Van Gastel, A., Kenis, G., De Jongh, R., & Bosmans, E. (1999). Elevated serum interleukin-6 (IL-6) and IL-6 receptor concentrations in posttraumatic stress disorder following accidental man-made traumatic events. Biological Psychiatry, 45, 833-839.

PAGE 81

72 Maes, M., Song, C., Lin, A., De Jongh, R., Van Gastel, A., Kenis, G., Bosmans, E., De Meester, I., Benoy, I., Neels, H., Demedts, P., Janca, A., Scharpe, S., & Smith, R .S. (1998). The effects of psychological stress on humans: Increased production of pro-inflammatory cytokines and a Th1-like response in stress-induced anxi ety. Cytokine, 10, 313-318. Mason, J. W. (1975). A historical view of the stress field (Parts I and II). Journal of Human Stress, 1, 6-12 & 22-36. McCarty, M. F. (1999). Interleukin-6 as a central mediator of cardiovascular r isk associated with chronic inflammation, smoking, diabetes, and visceral obesity: Down-regulation with essential fatty acids, ethanol and pentoxifyline. Medical Hypotheses, 52, 465-477. McDade, T. W., Hawkley, L. C., & Cacioppo, J. T. (2006). Psychosocial and behavioral predictors of inflammation in middle-aged and older adults: The Chicago Health, Aging, and Social Relations Study. Psychosomatic Medicine, 68, 376-381. McLeod, J. D. & Kessler, R. D. (1990). Socioeconomic status differences in vulnerability to undesirable life events. Journal of Health and Social Behavior, 31, 162-172. Melamed, S., Shirom, A., Toker, S., Berliner, S., & Shapira, I. (2004). Association of fear of terror with low-grade inflammation among apparently healthy empl oyed adults. Psychosomatic Medicine, 66, 484-491.

PAGE 82

73 Mendall, M. A., Strachan, D. P., Butland, B. K., Ballam, L., Morris, J., Sweetnam, P. M., & Elwood, P. C. (2000). C-reactive protein: Relation to total mortality, cardiovascular mortality and cardiovascular risk factors in men. European Heart Journal, 21, 1584-1590. Miller, G. E., Freedland, K. E., Carney, R. M., Stetler, C. A., & Banks, W. A. (2003). Pathways linking depression, adiposity, and inflammatory markers in healthy young adults. Brain, Behavior, and Immunity, 17, 276-285. Miller, G. E., Stetler, C. A., Carney, R. M., Freedland, K. E., & Banks, W. A. (2002). Clinical Depression and Inflammatory Risk Markers for Coronary Heart Dise ase. American Journal of Cardiology, 90, 1279–1283. Mohamed-Ali, V., Goodrick, S., Rawesh, A., Katz, D. R., Miles, J. M., Yudkin, J. S., Klein, S. & Coppack, S. W. (1997). Subcutaneous adipose tissue releases interleukin-6, but not tumor necrosis factor, in vivo. Journal of Clinical Endocrinology and Metabolism, 82, 4196-4200. Mohamed-Ali, V., Pinkney, J. H., & Coppack, S. W. (1998). Adipose tissue as an endocrine and paracrine organ. International Journal of Obesity, 22, 1145-1158. National Center for Health Statistics. (2006). NHANES 2001-2002: National Health and Nutrition Examination Survey Questionnaire, Examination Protocol, and Laboratory Protocol. Retrieved July 15, 2006 from U.S. Department of Health and Human Services, Centers for Disease Control and Prevention Website: http://www.cdc.gov/nchs/about/major/nhanes/nhanes01-02.htm

PAGE 83

74 Ohlson, L-O., Larsson, B., Svcardsudd, K., Welin, L, Eriksson, H., Wilhemsen, L., Bjorntorp, P., & Tibblin, G. (1985). The influence of body fat distribution on the incidence of diabetes mellitus: 13.5 years of follow-up of the participants in the study of men born in 1913. Diabetes, 34, 1055-1058. Ouchi, N., Kihara, S., Funahashi, T., Nakamura, T., Nishada, M., Kumada, M., Okamoto, Y., Ohashi, K., Nagaretani, H., Kishida, K., Nishizawa, H., Maeda, N., Kobayashi, H., Hiraoka, H., Matsuzawa, Y. (2003). Reciprocal association of c-reactive protein with adiponectin in blood stream and adipose tissue. Circulation, 107, 671-674. Owen, N., Poulton, T., Hay, F. C., Mohamed-Ali, V., & Steptoe, A. (2003). Socioeconomic status, C-reactive protein, immune factors, and responses to acute mental stress. Brain, Behavior, and Immunity, 17, 286-295. Owen, N. & Steptoe, A. (2003). Natural killer cell and proinflammatory cytokine responses to mental stress: Associations with heart rate and heart rate vari ability. Biological Psychology, 63, 101-115. Panagiotakos, D. B., Pitsavos, C. E., Chrysohoou, C. A., Skoumas, J., Toutouza, M., Belegrinos, D., Toutouzas, P. K., & Stefanadis, C. (2004). The association between educational status and risk factors related to cardiovascular diseas e in health individuals: The ATTICA Study. Annals of Epidemiology, 14, 188-194.

PAGE 84

75 Pearson, T. A., Mensah, G. A., Alexander, R. W., Anderson, J. L., Cannon, R. O., Criqui, M., Fadl, Y. Y., Fortmann, S. P., Hong, Y., Myers, G. L., Rifai, N., Smith, S. C., Taubert, K., Tracy, R. P., & Vinicor, F. (2003). Markers of inflammation and cardiovascular disease: Application to clinical and public health practice: A statement for healthcare professionals from the Centers for Disease Cont rol and Prevention and the American Heart Association. Circulation, 107, 499-511. Penninx, B. W. J. H., Kirtchevsky, S. B., Yaffe, K., Newman, A. B., Simonsick, E. M., Rubin, S., Ferrucci, L., Harris, T., & Pahor, M. (2003). Inflammatory markers and depressed mood in older persons: Results from the health, aging and body composition study. Society of Biological Psychiatry, 54, 566-572. Pradhan, A. D., Manson, J. E., Rifai, N., Buring, J. E., & Ridker, P. M. (2001). Creactive protein, interleukin-6, and risk of developing type 2 diabetes mellitus. The Journal of the American Medical Association, 286, 327-334. Preacher, K. J. (2003). A Primer on Interaction Effects in Multiple Linear R egression. Retrieved January 16, 2008, from www.psych.ku.edu/preacher/interact/interactions.htm. Ridker, P. M., Cushman, M., Stampfer, M. J., Tracy, R. P., & Hennekens, C. H. (1997). Inflammation, aspirin, and the risk of cardiovascular disease in apparently hea lthy men. The New England Journal of Medicine, 336, 973-979. Ridker, P. M., Hennekens, C. H., Burin J. E., & Rifai, N. (2000). C-reactive protein and other markers of inflammation in the prediction of cardiovascular disease in women. The New England Journal of Medicine, 342, 836-843.

PAGE 85

76 Segerstrom, S. C. & Miller, G. E. (2004). Psychological stress and the human immune system: A meta-analytic study of 30 years of inquiry. Psychological B ulletin, 130, 601-630. Suarez, E. C., Krishnan, R. R., & Lewis, J. G. (2003). The relation of severity of depressive symptoms to monocyte-associated proinflammatory cytokines and chemokines in apparently healthy men. Psychosomatic Medicine, 65, 362-368. Sutherland, A. G., Alexander, D. A., & Hutchison, J. D. (2003). Disturbance of proinflammatory cytokines in post-traumatic psychopathology. Cytokine, 24, 219225. Steptoe, A. & Vogele, C. (1991). Methodology of mental stress testing in cardiovascul ar research. Circulation, 83(suppl. 2), II-14 – II-24. Toker, S., Shirom, A., Shapira, I., Berliner, S., & Melamed, S. (2005). The association between burnout, depression, anxiety, and inflammation biomarkers: C-reactive protein and fibrinogen in men and women. Journal of Occupational Health Psychology, 10, 344-362. Tuglu, C., Kara, S. H., Caliyurt, O., Vardar, E., & Abay, E. (2003). Increased serum tumor necrosis factor-alpha levels and treatment response in major depressive disorder. Psychopharmacology, 170, 429-433. Turner, R. J., Wheaton, B., & Lloyd, D. A. (1995). The epidemiology of social stress. American Sociological Review, 60, 104-125.

PAGE 86

77 Vetta, F., Ronzoni, S., Lupattelli, M. R., Novi, B., Fabbriconi, P., Ficoneri, C., Cicconetti, P., Bruno, A., Russo, F., & Bollea, M. R. (2001). Tumor necrosis factorand mood disorders in the elderly. Archives of Gerontology and Geriatrics, suppl. 7, 435-442. Visser, M., Bouter, L. M., McQuillan, G. M., Wener, M. H., & Harris, T. B. (1999). Elevated C-reactive protein levels in overweight and obese adults. The Journal of the American Medical Association, 282, 2131-2135. Wellen, K. E. & Hotamisligil, G. S. (2005). Inflammation, stress, and diabetes. The Journal of Clinical Investigation, 115, 1111-1119. Wright, C. E., Strike, P. C., Brydon, L., & Steptoe, A. (2005). Acute inflammation and negative mood: Mediation by cytokine activation. Brain, Behavior, and Immunity, 19, 345-350. Wu, T. Dorn, J. P. Donahue, R. P. Sempos, C. T., Trevisan, M. (2002). Associations of serum c-reactive protein with fasting insulin, glucose and glycosylated Hemoglobin: The third national health and nutrition examination survey, 1988-1994. American Journal of Epidemiology, 155, 65-71. Wyatt, S. B., Winters, K. P., Dubbert, P. M. (2006). Overweight and obesity: Prevalence, consequences, and causes of a growing public health problem. The American Journal of the Medical Sciences, 331, 166-174. Yudkin, J. S., Kumari, M., Humphries, S. E., & Mohamed-Ali, V. (2000). Inflammation, obesity, stress and coronary heart disease: Is interleuki n-6 the link? Atherosclerosis, 148, 209-214.

PAGE 87

78 Yudkin, J. S., Stehouwer, C. D. A., Emeis, J. J., & Coppack, S. W. (1999). C-reactive protein in healthy subjects: Associations with obesity, insulin resistance, an d endothelial dysfunction: A potential role for cytokines originating from adipose tissue? Arteriosclerosis, Thrombosis, and Vascular Biology, 19, 972-978. Yusuf, S., Hawken, S., Ounpuu, S., Bautista, L., Franzosi, M. G., Commerford, P., Lang, C. C., Rumboldt, Z., Onen, C. L., Lisheng, L., Tanomsup, S., Wangaijr, P., Razak, F., Sharma, A. M., Anand, A. S. (2005). Obesity and the risk of myocardial infarction in 27,000 participants from 52 countries: A case-control study. Lancet, 366, 1640-1649. Zhou, D., Kusnecov, A. W., Shurin, M. R., DePaoli, M., & Rabin, B. (1993). Exposure to physical and psychological stressors elevates plasma interleukin-6: Relationship to the activation of hypothalamic-pituitary-adrenal axis. Endocrinology, 133, 2523-2530. Ziccardi, P., Nappo, F., Giugliano, G., Esposito, K., Marfella, R., Cioffi, M., D’Andrea, F., Molinari, A. M., & Giugliano, G. (2002). Reduction of inflammatory cytokine concentrations and improvement of endothelial functions in obese women after weight loss over one year. Circulation, 105, 804-809.

PAGE 88

79 Appendix A Hierarchical Linear Regressions that Utilize Sampling Weights: T ables

PAGE 89

80 Table A1 Summary of Hierarchical Regression Analysis for Education and BMI Predicting LogCRP : Weighted Data Variable B SE B R 2 F (10, 15) p Step 2 0.156*** 123.33 <0.001 Education 0.015 0.027 BMI 0.076*** 0.003 Step 3 0.000 168.40 <0.001 Education 0.157 0.102 BMI 0.064*** 0.010 Education x BMI 0.005 0.003 Note. All steps contain the covariates (gender, age, eth nicity, use of blood pressure medication, use of cholesterol medication and smoking) variables prese nted are those that are relevant to the hypotheses. BMI = body mass index (kg/m 2 ). R 2 = 0.067, F (8,15) = 48.58, p < 0.001 for Step 1; R 2 = 0.223 for Step 2; R 2 = 0.223 for Step 3. *** p < .001 Table A2 Summary of Hierarchical Regression Analysis for Education and WC Predicting LogCRP: Weighted Data Variable B SE B R 2 F (10, 15) p Step 2 0.164*** 125.65 <0.001 Education 0.014 0.027 WC 0.032*** 0.001 Step 3 0.000 147.22 <0.001 Education 0.138 0.168 WC 0.029*** 0.005 Education x WC 0.001 0.002 Note. All steps contain the covariates (gender, age, eth nicity, use of blood pressure medication, use of cholesterol medication and smoking) variables prese nted are those that are relevant to the hypotheses. WC = waist circumference (cm). R 2 = 0.067, F (8, 15) = 48.58, p < 0.001 for Step 1; R 2 = 0.232 for Step 2; R 2 = 0.232 for Step 3. p < .05, *** p < .001

PAGE 90

Appendix A (continued) 81 Table A3 Summary of Hierarchical Regression Analysis for Income and BMI Predicting LogCRP: Weighted Data Variable B SE B R 2 F (10, 15) p Step 2 0.159*** 230.58 <0.001 Income 0.023*** 0.006 BMI 0.077*** 0.003 Step 3 0.002* 285.22 <0.001 Income 0.106** 0.030 BMI 0.055*** 0.009 Income x BMI 0.003** 0.001 Note. All steps contain the covariates (gender, age, eth nicity, use of blood pressure medication, use of cholesterol medication and smoking) variables prese nted are those that are relevant to the hypotheses. BMI = body mass index (kg/m 2 ). R 2 = 0.067, F (8, 15) = 48.58, p < 0.001 for Step 1; R 2 = .226for Step 2; R 2 = .228 for Step 3. ** p < .01, *** p < .001 Table A4 Summary of Hierarchical Regression Analysis for Income and WC Predicting LogCRP: Weighted Data Variable B SE B R 2 F (10, 15) p Step 2 0.167*** 132.26 <0.001 Income 0.021** 0.006 WC 0.032*** 0.001 Step 3 0.001 144.39 <0.001 Income 0.078 0.049 WC 0.028*** 0.004 Income x WC 0.001 0.000 Note. All steps contain the covariates (gender, age, eth nicity, use of blood pressure medication, use of cholesterol medication and smoking) variables prese nted are those that are relevant to the hypotheses. WC = waist circumference (cm). R 2 = 0.067, F (8, 15) = 48.58, p < 0.001 for Step 1; R 2 = 0.234 for Step 2; R 2 = 0.235 for Step 3. ** p < .01, *** p < .001

PAGE 91

Appendix A (continued) 82 Table A5 Summary of Hierarchical Regression Analysis for Depression Diagnosis and BMI Predicting LogCRP: Weighted Data Variable B SE B R 2 F (10, 15) p Step 2 0.181*** 498.77 <0.001 Depression Diagnosis 0.090 0.156 BMI 0.085*** 0.006 Step 3 0.000 666.85 <0.001 Depression Diagnosis 0.472 0.529 BMI 0.087*** 0.006 Depression Diagnosis x BMI -0.014 0.014 Note. All steps contain the covariates (gender, age, eth nicity, use of blood pressure medication, use of cholesterol medication and smoking) variables prese nted are those that are relevant to the hypotheses. BMI= body mass index (kg/m 2 ). R 2 = 0.077, F (8, 15) = 83.84, p < 0.001 for Step 1; R 2 = 0.258 for Step 2; R 2 = 0.2581 for Step 3. *** p < .001 Table A6 Summary of Hierarchical Regression Analysis for Depression Diagnosis and WC Predicting LogCRP: Weighted Data Variable B SE B R 2 F (10, 15) p Step 2 0.167*** 214.67 <0.001 Depression Diagnosis 0.093 0.189 WC 0.032 0.003 Step 3 0.000 311.30 <0.001 Depression Diagnosis 0.564 0.596 WC 0.033*** 0.003 Depression Diagnosis x WC 0.005 0.005 Note. All steps contain the covariates (gender, age, eth nicity, use of blood pressure medication, use of cholesterol medication and smoking) variables prese nted are those that are relevant to the hypotheses. WC=Waist Circumference (cm). R 2 = 0.077, F (8, 15) = 83.84, p < 0.001 for Step 1; R 2 = 0.244 for Step 2; R 2 = 0.244 for Step 3. *** p < .001

PAGE 92

Appendix A (continued) 83 Table A7 Summary of Hierarchical Regression Analysis for Depression Symptoms and BMI Predicting LogCRP: Weighted Data Variable B SE B R 2 F (10, 15) p Step 2 0.180*** 834.67 <0.001 Depression Symptoms 0.007 0.011 BMI 0.085*** 0.005 Step 3 0.000 1017.33 <0.001 Depression Symptoms -0.002 0.081 BMI 0.084*** 0.007 Depression Symptoms x BMI 0.000 0.002 Note. All steps contain the covariates (gender, age, eth nicity, use of blood pressure medication, use of cholesterol medication and smoking) variables prese nted are those that are relevant to the hypotheses. BMI = body mass index (kg/m 2 ). R 2 = 0.077, F (8, 15) = 83.84, p < 0.001 for Step 1; R 2 = 0.257 for Step 2; R 2 = 0.257 for Step 3. *** p < .001 Table A8 Summary of Hierarchical Regression Analysis for Depression Symptoms and WC Predicting LogCRP: Weighted Data Variable B SE B R 2 F (10, 15) p Step 2 0.161*** 272.21 <0.001 Depression Symptoms 0.007 0.013 WC 0.033*** 0.003 Step 3 0.000 356.19 <0.001 Depression Symptoms -0.001 0.107 WC 0.032*** 0.003 Depression Symptoms x WC 0.000 0.001 Note. All steps contain the covariates (gender, age, eth nicity, use of blood pressure medication, use of cholesterol medication and smoking) variables prese nted are those that are relevant to the hypotheses. WC = Waist Circumference (cm). R 2 = 0.077, F (8,15) = 83.84, p < .0001 for Step 1; R 2 = .238 for Step 2; R 2 = .238 for Step 3. *** p < .001.

PAGE 93

Appendix A (continued) 84 Table A9 Summary of Hierarchical Regression Analysis for Education, BMI, and Gender Predicting LogCRP : Weighted Data Variable B SE B R 2 F (10, 15) p Step 2 0.167*** 123.33 <0.001 Education -0.015 0.027 BMI 0.076*** 0.003 Gender 0.276*** 0.037 Step 3 0.001 195.75 <0.001 Education -0.245 0.126 BMI 0.057** 0.015 Gender 0.014 0.239 Education x BMI 0.005 0.003 Education x Gender 0.064 0.049 BMI x Gender 0.005 0.007 Step 4 0.000 204.13 <0.001 Education -0.546 0.502 BMI 0.032 0.042 Gender -0.467 0.771 Education x BMI 0.016 0.016 Education x Gender 0.255 0.307 BMI x Gender 0.021 0.025 Education x BMI x Gender -0.007 0.010 Note. All steps contain the covariates (age, ethnicity, use of blood pressure medication, use of cholestero l medication and smoking) variables presented are tho se that are relevant to the hypotheses. BMI= body mass index (kg/m 2 ). R 2 = 0.056, F (7, 15) = 46.44, p < 0.001 for Step 1; R 2 = 0.223 for Step 2; R 2 = 0.224for Step 3; R 2 = 0.224 for Step 4. ** p < .01, *** p < .001.

PAGE 94

Appendix A (continued) 85 Table A10 Summary of Hierarchical Regression Analysis for Education, WC, and Gender Predicting LogCRP: Weighted Data Variable B SE B R 2 F (10, 15) p Step 2 0.176*** 125.65 <0.001 Education -0.014 0.027 WC 0.032*** 0.001 Gender 0.476*** 0.042 Step 3 0.002** 186.40 <0.001 Education -0.311 0.206 WC 0.019* 0.007 Gender 0.002 0.002 Education x WC -0.016 0.015 Education x Gender 0.096 0.051 WC x Gender 0.006* 0.003 Step 4 0.000 181.71 <0.001 Education -0.228 0.726 WC 0.021 0.017 Gender -0.207 1.152 Education x WC 0.001 0.007 Education x Gender 0.040 0.471 WC x Gender 0.005 0.011 Education x WC x Gender 0.001 0.005 Note. All steps contain the covariates (age, ethnicity, use of blood pressure medication, use of cholestero l medication and smoking) variables presented are tho se that are relevant to the hypotheses. WC = waist circumference (cm). R 2 = 0.056, F (7, 15) = 46.44, p < 0.001 for Step 1; R 2 = 0.232 for Step 2; R 2 = 0.234 for Step 3; R 2 = 0.234 for Step 4. *** p < .001, ** p < .01, p < .05

PAGE 95

Appendix A (continued) 86 Table A11 Summary of Hierarchical Regression Analysis for Income, BMI, and Gender Predicting LogCRP: Weighted Data Variable B SE B R 2 F (10, 15) p Step 2 0.170*** 230.58 <0.001 Income -0.023*** 0.006 BMI 0.077*** 0.003 Gender 0.251*** 0.035 Step 3 0.004** 293.17 <0.001 Income -0.140** 0.036 BMI 0.043** 0.012 Gender -0.096 0.227 Income x BMI 0.003** 0.001 Income x Gender 0.021 0.012 BMI x Gender 0.007 0.006 Step 4 0.000 921.96 <0.001 Income -0.133 0.111 BMI 0.045 0.027 Gender -0.064 0.480 Income x BMI 0.002 0.004 Income x Gender 0.016 0.065 BMI x Gender 0.006 0.015 Income x BMI x Gender 0.0001 0.002 Note. All steps contain the covariates (age, ethnicity, use of blood pressure medication, use of cholestero l medication and smoking) variables presented are tho se that are relevant to the hypotheses. BMI= body mass index (kg/m 2 ). R 2 = 0.056, F (7, 15) = 46.44, p < 0.001 for Step 1; R 2 = 0.226 for Step 2; R 2 = 0.230 for Step 3; R 2 = 0.230 for Step 4. ** p < .01, *** p < .001.

PAGE 96

Appendix A (continued) 87 Table A12 Summary of Hierarchical Regression Analysis for Income, WC, and Gender Predicting LogCRP: Weighted Data Variable B SE B R 2 F (10, 15) p Step 2 0.178*** 132.26 <0.001 Income -0.021** 0.006 WC 0.032*** 0.001 Gender 0.454*** 0.041 Step 3 0.005** 233.75 <0.001 Income -0.157* 0.062 WC 0.016* 0.006 Gender -0.440 0.254 Income x WC 0.001 0.0005 Income x Gender 0.033* 0.012 WC x Gender 0.007** 0.002 Step 4 0.000 276.35 <0.001 Income -0.105 0.128 WC 0.020* 0.009 Gender -0.189 0.545 Income x WC 0.0003 0.001 Income x Gender -0.001 0.074 WC x Gender 0.004 0.005 Income x WC x Gender 0.0003 0.00 Note All steps contain the covariates (age, ethnicity, use of blood pressure medication, use of cholester ol medication and smoking) variables presented are tho se that are relevant to the hypotheses. WC = Waist Circumference (cm). R 2 = 0.056, F (7, 15) = 46.44, p < 0.001 for Step 1; R 2 = 0.234 for Step 2; R 2 = 0.239 for Step 3; R 2 = 0.239 for Step 4. p < .05, ** p < .01, *** p < .001.

PAGE 97

Appendix A (continued) 88 Table A13 Summary of Hierarchical Regression Analysis for Depression Diagnosis, BMI, and Gender Predicting LogCRP: Weighted Data Variable B SE B R 2 F (10, 15) p Step 2 0.219*** 498.77 <0.001 Depression Diagnosis 0.090 0.156 BMI 0.085*** 0.006 Gender 0.548*** 0.094 Step 3 0.002 591.49 <0.001 Depression Diagnosis 0.991* 0.461 BMI 0.074** 0.020 Gender 0.347 0.339 Depression Diagnosis x BMI -0.016 0.015 Depression Diagnosis x Gender -0.297 0.322 BMI x Gender -0.008 0.012 Step 4 0.000 869.54 <0.001 Depression Diagnosis 0.623 2.024 BMI 0.073** 0.022 Gender 0.329 0.376 Depression Diagnosis x BMI -0.002 0.070 Depression Diagnosis x Gender -0.094 1.153 BMI x Gender 0.009 0.014 Depression Diagnosis x BMI x Gender -0.008 0.037 Note. All steps contain the covariates (age, ethnicity, use of blood pressure medication, use of cholestero l medication and smoking) variables presented are tho se that are relevant to the hypotheses. BMI= body mass index (kg/m 2 ). R 2 = 0.039, F (7, 15) = 105.87, p < 0.001 for Step 1; R 2 = 0.258 for Step 2; R 2 = 0.260 for Step 3; R 2 = 0.260 for Step 4. *** p < .001, ** p < .01, p < .05

PAGE 98

Appendix A (continued) 89 Table A14 Summary of Hierarchical Regression Analysis for Depression Diagnosis, WC, and Gender Predicting LogCRP: Weighted Data Variable B SE B R 2 F (10, 15) p Step 2 0.205*** 214.67 <0.001 Depression Diagnosis 0.093 0.189 WC 0.032 0.003 Gender 0.685*** 0.095 Step 3 0.004 604.34 <0.001 Depression Diagnosis 1.377* 0.588 WC 0.021 0.010 Gender -0.078 0.629 Depression Diagnosis x WC -0.009* 0.004 Depression Diagnosis x Gender -0.286 0.317 WC x Gender 0.009 0.007 Step 4 0.000 419.16 <0.001 Depression Diagnosis 2.538 2.704 WC 0.022 0.011 Gender -0.029 0.664 Depression Diagnosis x WC -0.021 0.028 Depression Diagnosis x Gender -0.945 1.514 WC x Gender 0.008 0.007 Depression Diagnosis x WC x Gender 0.007 0.016 Note. All steps contain the covariates (age, ethnicity, use of blood pressure medication, use of cholestero l medication and smoking) variables presented are tho se that are relevant to the hypotheses. WC=Waist Circumference (cm). R 2 = 0.039, F (7, 15) = 105.87, p < 0.001 for Step 1; R 2 = 0.244 for Step 2; R 2 = 0.248 for Step 3, R 2 = 0.248 for Step 4. p < .05, *** p < .001

PAGE 99

Appendix A (continued) 90 Table A15 Summary of Hierarchical Regression Analysis for Depression Symptoms, BMI, and Gender Predicting LogCRP: Weighted Data Variable B SE B R 2 F (10, 15) p Step 2 0.218*** 834.67 <0.001 Depression Symptoms 0.007 0.011 BMI 0.085*** 0.005 Gender 0.549*** 0.094 Step 3 0.001 1680.61 <0.001 Depression Symptoms 0.042 0.083 BMI 0.076** 0.020 Gender 0.417 0.324 Depression Symptoms x BMI 0.0002 0.003 Depression Symptoms x Gender -0.027 0.043 BMI x Gender 0.006 0.012 Step 4 0.001 1681.86 <0.001 Depression Symptoms 0.192 0.266 BMI 0.080** 0.023 Gender 0.488 0.385 Depression Symptoms x BMI -0.005 0.009 Depression Symptoms x Gender -0.111 0.154 BMI x Gender 0.003 0.014 Depression Symptoms x BMI x Gender 0.003 0.00 5 Note. All steps contain the covariates (age, ethnicity, use of blood pressure medication, use of cholestero l medication and smoking) variables presented are tho se that are relevant to the hypotheses. BMI = bod y mass index (kg/m 2 ). R 2 = 0.039, F (7, 15) = 105.87, p < 0.001 for Step 1; R 2 = 0.257 for Step 2; R 2 = 0.258 for Step 3; R 2 = 0.259 for Step 4. ** p < .01, *** p < .001

PAGE 100

Appendix A (continued) 91 Table A16 Summary of Hierarchical Regression Analysis for Depression Symptoms, WC, and Gender Predicting LogCRP: Weighted Data Variable B SE B R 2 F (10, 15) p Step 2 0.199*** 272.21 <0.001 Depression Symptoms 0.007 0.013 WC 0.033*** 0.003 Gender 0.673*** 0.098 Step 3 0.008 133.82 <0.001 Depression Symptoms 0.020 0.080 WC 0.022* 0.010 Gender 0.014 0.609 Depression Symptoms x WC 0.00003 0.001 Depression Symptoms x Gender -0.011 0.040 WC x Gender 0.007 0.007 Step 4 0.001 375.69 <0.001 Depression Symptoms 0.444 0.350 WC 0.025* 0.012 Gender 0.238 0.698 Depression Symptoms x WC -0.004 0.004 Depression Symptoms x Gender -0.266 0.193 WC x Gender 0.005 0.008 Depression Symptoms x WC x Gender 0.003 0.002 Note. All steps contain the covariates (age, ethnicity, use of blood pressure medication, use of cholestero l medication and smoking) variables presented are tho se that are relevant to the hypotheses. WC = Waist Circumference (cm). R 2 = 0.039, F (7, 15) = 105.87, p < 0.001 for Step 1; R 2 = .238 for Step 2; R 2 = .246 for Step 3; R 2 = .247 for Step 4. p < .05, *** p < .001.

PAGE 101

Appendix A (continued) 92 Table A17 Summary of Hierarchical Regression Analysis for Gender and Education Predicting LogCRP: Weighted Data Variable B SE B R 2 F (11,15) p Step 2 0.026*** 122.18 <0.001 Gender 0.398*** 0.049 Education -0.015 0.028 Step 3 0.001 132.23 <0.001 Gender 0.217 0.120 Education -0.128 0.089 Education x Gender 0.076 0.051 Note. All steps contain the covariates (age, ethnicity, use of blood pressure medication, use of cholestero l medication, smoking, body mass index, and waist cir cumference) variables presented are those that are relevant to the hypotheses. R 2 = 0.211, F (9,15) = 134.60, p < .001 for Step 1; R 2 = .237 for Step 2; R 2 = .238 for Step 3. *** p < .001. Table A18 Summary of Hierarchical Regression Analysis for Gender and Income Predicting LogCRP: Weighted Data Variable B SE B R 2 F (11,15) p Step 2 0.028*** 157.15 <0.001 Gender 0.377*** 0.051 Income -0.022** 0.006 Step 3 0.002* 155.17 <0.001 Gender 0.186 0.096 Income -0.060** 0.019 Gender x Income 0.025* 0.011 Note. All steps contain the covariates (age, ethnicity, use of blood pressure medication, use of cholestero l medication, smoking, body mass index, and waist cir cumference) variables presented are those that are relevant to the hypotheses. R 2 = 0.211, F (9,15) = 134.60, p < .0001 for Step 1; R 2 = .239 for Step 2; R 2 = .241 for Step 3. p < .05, ** p < .01, *** p < .001.

PAGE 102

Appendix A (continued) 93 Table A19 Summary of Hierarchical Regression Analysis for Gender and Depression Diagnosis Predicting LogCRP: Weighted Data Variable B SE B R 2 F (11,15) p Step 2 0.055*** 836.16 <0.001 Gender 0.592*** 0.088 Depression Diagnosis 0.089 0.166 Step 3 0.001 910.08 <0.001 Gender 0.614*** 0.087 Depression Diagnosis 0.550 0.518 Gender x Depression Diagnosis -0.294 0.348 Note. All steps contain the covariates (age, ethnicity, use of blood pressure medication, use of cholestero l medication, smoking, body mass index, and waist cir cumference) variables presented are those that are relevant to the hypotheses. R 2 = 0.203 F (9, 15) = 219.63, p < .001 for Step 1; R 2 = 0.258 for Step 2; R 2 = 0.259 for Step 3. *** p < .001. Table A20 Summary of Hierarchical Regression Analysis for Gender and Depression Symptoms Predicting LogCRP: Weighted Data Variable B SE B R 2 F (11,15) p Step 2 0.055*** 1265.33 <0.001 Gender 0.593*** 0.089 Depression Symptoms 0.007 0.012 Step 3 0.000 1062.00 <0.001 Gender 0.611*** 0.090 Depression Symptoms 0.040 0.058 Gender x Depression Symptoms -0.021 0.042 Note. All steps contain the covariates (age, ethnicity, use of blood pressure medication, use of cholestero l medication, smoking, body mass index, and waist cir cumference) variables presented are those that are relevant to the hypotheses. R 2 = 0.203 F (9, 15) = 219.63, p < .001 for Step 1; R 2 = 0.258 for Step 2; R 2 = 0.258 for Step 3. *** p < .001.

PAGE 103

Appendix A (continued) 94 Table A21 Summary of Hierarchical Regression Analysis for Gender and BMI Predicting LogCRP (The SES Sample): Weighted Data Variable B SE B R 2 F (11,15) p Step 2 0.164*** 207.45 <0.001 Gender 0.251*** 0.035 BMI 0.077*** 0.003 Step 3 0.000 217.57 <0.001 Gender 0.098 0.183 BMI 0.068*** 0.010 Gender x BMI 0.005 0.006 Note. All steps contain the covariates (age, ethnicity, use of blood pressure medication, use of cholestero l medication, smoking, education, and income) variabl es presented are those that are relevant to the hypotheses. These analyses utilized the sampling w eights. BMI = body mass index. R 2 = 0.063, F (9, 15) = 33.41, p < .0001 for Step 1; R 2 = 0.227 for Step 2; R 2 = 0.227 for Step 3. *** p < .001 Table A22 Summary of Hierarchical Regression Analysis for Gender and BMI Predicting LogCRP (The Depression Sample): Weighted Data Variable B SE B R 2 F (11,583) p Step 2 0.215*** 811.48 <0.001 Gender 0.548*** 0.094 BMI 0.085*** 0.006 Step 3 0.000 719.22 <0.001 Gender 0.401 0.312 BMI 0.076** 0.021 Gender x BMI 0.005 0.013 Note. All steps contain the covariates (age, ethnicity, use of blood pressure medication, use of cholestero l medication, smoking, depression diagnosis, and depr ession symptoms) variables presented are those that are relevant to the hypotheses. These analyses uti lized sampling weights. BMI = body mass index. R 2 = 0.043, F (9, 15) = 94.82, p < .001 for Step 1; R 2 = 0.258 for Step 2; R 2 = 0.258 for Step 3. ** p < .01, *** p < .001.

PAGE 104

Appendix A (continued) 95 Table A23 Summary of Hierarchical Regression Analysis for Gender and WC Predicting LogCRP (the SES sample): Weighted Data Variable B SE B R 2 F (11,15) p Step 2 0.172*** 119.08 <0.001 Gender 0.453*** 0.041 WC 0.032*** 0.001 Step 3 0.001** 190.49 <0.001 Gender -0.139 0.224 WC 0.023*** 0.004 Gender x WC 0.006* 0.002 Note. All steps contain the covariates (age, ethnicity, use of blood pressure medication, use of cholestero l medication, smoking, education, and income) variabl es presented are those that are relevant to the hypotheses. These analyses utilized sampling weigh ts. WC = waist circumference (cm). R 2 = 0.063, F (9, 15) = 33.41, p < .001 for Step 1; R 2 = .235 for Step 2; R 2 = 0.236 for Step 3. ** p < .01, *** p < .001. Table A24 Summary of Hierarchical Regression Analysis for Gender and WC Predicting LogCRP (the Depression Sample): Weighted Data Variable B SE B R 2 F (11,15) p Step 2 0.201*** 554.23 <0.001 Gender 0.685*** 0.095 WC 0.032*** 0.003 Step 3 0.002 1600.32 <0.001 Gender 0.015 0.022 WC 0.022 0.010 Gender x WC 0.007 0.007 Note. All steps contain the covariates (age, ethnicity, use of blood pressure medication, use of cholestero l medication, smoking, depression diagnosis, and depr ession symptoms) variables presented are those that are relevant to the hypotheses. These analyses uti lized sampling weights. WC = waist circumference ( cm). R 2 = 0.043, F (9, 15) = 94.82, p < .001 for Step 1; R 2 = 0.244 for Step 2; R 2 = 0.246 for Step 3. *** p < .001.

PAGE 105

96 Appendix B 3-way Hierarchical Linear Regressions (Unweighted Data)

PAGE 106

97 Table B1 Summary of Hierarchical Regression Analysis for Education, BMI, and Gender Predicting LogCRP : Unweighted Data Variable B SE B R 2 F (10, 3681) p Step 2 0.159*** 93.50 <0.001 Education -0.028 0.020 -0.023 BMI 0.069 0.003 0.381*** Gender 0.304 0.031 0.146*** Step 3 0.001 72.40 <0.001 Education -0.266 0.104 -0.217* BMI 0.055 0.011 0.303*** Gender 0.256 0.180 0.123 Education x BMI 0.007 0.003 0.176* Education x Gender 0.031 0.037 0.051 BMI x Gender -0.001 0.006 -0.009 Step 4 0.000 67.46 <0.001 Education -0.750 0.304 -0.611* BMI 0.017 0.025 0.095 Gender -0.422 0.438 -0.203 Education x BMI -0.024 0.011 0.621* Education x Gender 0.339 0.185 0.564 BMI x Gender 0.024 0.015 0.386 Education x BMI x Gender -0.011 0.007 -0.550 Note. All steps contain the covariates (age, ethnicity, use of blood pressure medication, use of cholestero l medication and smoking) variables presented are tho se that are relevant to the hypotheses. BMI= body mass index (kg/m 2 ). R 2 = 0.044, F (7, 3684) = 24.33, p < 0.001 for Step 1; R 2 = 0.203 for Step 2; R 2 = 0.204 for Step 3; R 2 = 0.204 for Step 4. p < .05, ** p < .01, *** p < .001.

PAGE 107

Appendix B (continued) 98 Table B2 Summary of Hierarchical Regression Analysis for Education, WC, and Gender Predicting LogCRP: Unweighted Data Variable B SE B R 2 F (10, 3621) p Step 2 0.181*** 105.36 <0.001 Education -0.025 0.020 -0.020 WC 0.031 0.001 0.426*** Gender 0.482 0.032 0.231*** Step 3 0.003** 82.01 <0.001 Education -0.417 0.145 -0.339** WC 0.019 0.004 0.257*** Gender -0.0002 0.234 -0.0001 Education x WC 0.003 0.001 0.261* Education x Gender 0.062 0.037 0.103 WC x Gender 0.004 0.002 0.174 Step 4 0.000 76.15 <0.001 Education -0.587 0.405 -0.477 WC 0.015 0.010 0.205 Gender -0.252 0.607 -0.121 Education x WC 0.005 0.004 0.406 Education x Gender 0.175 0.254 0.291 WC x Gender 0.006 0.006 0.297 Education x WC x Gender -0.001 0.003 -0.184 Note. All steps contain the covariates (age, ethnicity, use of blood pressure medication, use of cholestero l medication and smoking) variables presented are tho se that are relevant to the hypotheses. WC = waist circumference (cm). R 2 = 0.044, F (7, 3684) = 24.33, p < 0.001 for Step 1; R 2 = 0.225 for Step 2; R 2 = 0.228 for Step 3; R 2 = 0.228 for Step 4. *** p < .001, ** p < .01, p < .05

PAGE 108

Appendix B (continued) 99 Table B3 Summary of Hierarchical Regression Analysis for Income, BMI, and Gender Predicting LogCRP: Unweighted Data Variable B SE B R 2 F (10,3434) p Step 2 0.157*** 86.53 <0.001 Income -0.021 0.006 -0.060*** BMI 0.069 0.003 0.377*** Gender 0.280 0.032 0.135*** Step 3 0.004** 67.84 <0.001 Income -0.134 0.031 -0.384*** BMI 0.046 0.012 0.253*** Gender 0.072 0.179 0.035 Income x BMI 0.003 0.001 0.256** Income x Gender 0.021 0.011 0.253* BMI x Gender 0.003 0.006 0.117 Step 4 0.000 62.98 <0.001 Income -0.155 0.092 -0.446 BMI 0.041 0.024 0.225 Gender -0.017 0.399 -0.008 Income x BMI 0.004 0.003 0.325 Income x Gender 0.035 0.055 0.190 BMI x Gender 0.006 0.014 0.093 Income x BMI x Gender 0.0004 0.002 -0.077 Note. All steps contain the covariates (age, ethnicity, use of blood pressure medication, use of cholestero l medication and smoking) variables presented are tho se that are relevant to the hypotheses. BMI= body mass index (kg/m 2 ). R 2 = 0.044, F (7, 3684) = 24.33, p < 0.001 for Step 1; R 2 = 0.201 for Step 2; R 2 = 0.205 for Step 3; R 2 = 0.205 for Step 4. *** p < .001, ** p < .01, p < .05

PAGE 109

Appendix B (continued) 100 Table B4 Summary of Hierarchical Regression Analysis for Income, WC, and Gender Predicting LogCRP: Unweighted Data Variable B SE B R 2 F (10, 3378) p Step 2 0.179*** 96.96 <0.001 Income -0.018 0.006 -0.050** WC 0.031 0.001 0.421*** Gender 0.459 0.033 0.221*** Step 3 0.004** 76.09 <0.001 Income -0.151 0.043 -0.433*** WC 0.017 0.004 0.238*** Gender -0.233 0.240 -0.112 Income x WC 0.001 0.0004 0.263* Income x Gender 0.032 0.011 0.173** WC x Gender 0.005 0.002 0.239* Step 4 0.000 70.64 <0.001 Income -0.159 0.121 -0.456 WC 0.017 0.009 0.231 Gender -0.268 0.553 -0.129 Income x WC 0.001 0.001 0.287 Income x Gender 0.037 0.074 0.200 WC x Gender 0.005 0.006 0.256 Income x WC x Gender -0.0001 0.001 -0.027 Note. All steps contain the covariates (age, ethnicity, use of blood pressure medication, use of cholestero l medication and smoking) variables presented are tho se that are relevant to the hypotheses. WC=Waist Circumference (cm). R 2 = 0.044, F (7, 3684) = 24.33, p < 0.001 for Step 1; R 2 = 0.223 for Step 2; R 2 = 0.227 for Step 3; R 2 = 0.227 for Step 4. *** p < .001, ** p < .01, p < .05

PAGE 110

Appendix B (continued) 101 Table B5 Summary of Hierarchical Regression Analysis for Depression Diagnosis, BMI, and Gender Predicting LogCRP: Unweighted Data Variable B SE B R 2 F (10, 572) p Step 2 0.239*** 21.79 <0.001 Depression Diagnosis 0.175 0.171 0.037 BMI 0.087 0.007 0.444*** Gender 0.569 0.087 0.241*** Step 3 0.002 16.87 <0.001 Depression Diagnosis 1.133 0.821 0.239 BMI 0.080 0.023 0.407*** Gender 0.447 0.406 0.189 Depression Diagnosis x BMI -0.010 0.022 -0.062 Depression Diagnosis x Gender -0.432 0.345 -0. 152 BMI x Gender -0.006 0.015 0.079 Step 4 0.000 15.65 <0.001 Depression Diagnosis 0.297 2.833 0.063 BMI 0.078 0.024 0.397** Gender 0.409 0.424 0.173 Depression Diagnosis x BMI 0.020 0.100 0.126 Depression Diagnosis x Gender 0.034 1.551 0. 012 BMI x Gender 0.007 0.015 0.098 Depression Diagnosis x BMI x Gender -0.017 0.0 54 -0.178 Note. All steps contain the covariates (age, ethnicity, use of blood pressure medication, use of cholestero l medication and smoking) variables presented are tho se that are relevant to the hypotheses. BMI= body mass index (kg/m 2 ). R 2 = 0.037, F (7, 575) = 3.17, p < 0.01 for Step 1; R 2 = 0.276 for Step 2; R 2 = 0.278 for Step 3; R 2 = 0.278 for Step 4. *** p < .001, ** p < .01, p < .05

PAGE 111

Appendix B (continued) 102 Table B6 Summary of Hierarchical Regression Analysis for Depression Diagnosis, WC, and Gender Predicting LogCRP: Unweighted Data Variable B SE B R 2 F (10, 568) p Step 2 0.249*** 22.72 <0.001 Depression Diagnosis 0.206 0.172 0.043 WC 0.036 0.003 0.463*** Gender 0.679 0.087 0.287*** Step 3 0.005 17.81 <0.001 Depression Diagnosis 1.551 1.191 0.324 WC 0.022 0.009 0.284* Gender -0.197 0.537 -0.083 Depression Diagnosis x WC -0.008 0.010 -0.162 Depression Diagnosis x Gender -0.371 0.348 -0. 129 WC x Gender 0.010 0.006 0.406 Step 4 0.000 16.51 <0.001 Depression Diagnosis 1.113 4.263 0.233 WC 0.022 0.009 0.281* Gender -0.212 0.556 -0.090 Depression Diagnosis x WC -0.004 0.043 -0.071 Depression Diagnosis x Gender -0.124 2.339 -0. 043 WC x Gender 0.010 0.006 0.413 Depression Diagnosis x WC x Gender -0.003 0.02 4 -0.085 Note. All steps contain the covariates (age, ethnicity, use of blood pressure medication, use of cholestero l medication and smoking) variables presented are tho se that are relevant to the hypotheses. WC=Waist Circumference (cm). R 2 = 0.037, F (7, 571) = 3.11, p < 0.01 for Step 1; R 2 = 0.286 for Step 2; R 2 = 0.291 for Step 3, R 2 = 0.291 for Step 4. p < .05, *** p < .001

PAGE 112

Appendix B (continued) 103 Table B7 Summary of Hierarchical Regression Analysis for Depression Symptoms, BMI, and Gender Predicting LogCRP: Unweighted Data Variable B SE B R 2 F (10, 572) p Step 2 0.239*** 21.79 <0.001 Depression Symptoms 0.020 0.020 0.037 BMI 0.087 0.007 0.443*** Gender 0.568 0.087 0.244*** Step 3 0.003 16.94 <0.001 Depression Symptoms 0.103 0.099 0.189 BMI 0.080 0.023 0.407*** Gender 0.497 0.406 0.210 Depression Symptoms x BMI 0.001 0.003 0.030 Depression Symptoms x Gender -0.063 0.040 -0.1 93 BMI x Gender 0.005 0.015 0.062 Step 4 0.001 15.74 <0.001 Depression Symptoms 0.302 0.346 0.551 BMI 0.085 0.025 0.434*** Gender 0.590 0.435 0.249 Depression Symptoms x BMI -0.007 0.012 -0.358 Depression Symptoms x Gender -0.175 0.191 -0.5 34 BMI x Gender 0.002 0.016 0.015 Depression Symptoms x BMI x Gender 0.004 0.00 7 0.373 Note. All steps contain the covariates (age, ethnicity, use of blood pressure medication, use of cholestero l medication and smoking) variables presented are tho se that are relevant to the hypotheses. BMI = bod y mass index (kg/m 2 ). R 2 = 0.037, F (7, 575) = 3.17, p < 0.01 for Step 1; R 2 = 0.276 for Step 2; R 2 = 0.279 for Step 3; R 2 = 0.280 for Step 4. *** p < .001

PAGE 113

Appendix B (continued) 104 Table B8 Summary of Hierarchical Regression Analysis for Depression Symptoms, WC, and Gender Predicting LogCRP: Unweighted Data Variable B SE B R 2 F (10, 568) p Step 2 0.248*** 22.68 <0.001 Depression Symptoms 0.021 0.020 0.039 WC 0.036 0.003 0.461*** Gender 0.677 0.088 0.286*** Step 3 0.005 17.73 <0.001 Depression Symptoms 0.098 0.143 0.178 WC 0.022 0.009 0.282* Gender -0.153 0.538 -0.065 Depression Symptoms x WC -0.0001 0.001 -0.020 Depression Symptoms x Gender -0.042 0.041 -0.1 27 WC x Gender 0.009 0.006 0.389 Step 4 0.001 16.49 <0.001 Depression Symptoms 0.453 0.480 0.817 WC 0.024 0.009 0.314* Gender -0.003 0.573 -0.001 Depression Symptoms x WC -0.004 0.005 -0.662 Depression Symptoms x Gender -0.248 0.270 -0.7 50 WC x Gender 0.008 0.006 0.321 Depression Symptoms x WC x Gender 0.002 0.003 0.624 Note. All steps contain the covariates (age, ethnicity, use of blood pressure medication, use of cholestero l medication and smoking) variables presented are tho se that are relevant to the hypotheses. WC = Waist Circumference (cm). R 2 = 0.037, F (7, 571) = 3.11, p < 0.01 for Step 1; R 2 = .285 for Step 2; R 2 = .290 for Step 3; R 2 = .291 for Step 4. p < .05, *** p < .001.

PAGE 114

105 Appendix C Gender as a Moderator of the Relationship between Adiposity and CRP (Unweighted Data): Tables

PAGE 115

106 Table C1 Summary of Hierarchical Regression Analysis for Gender and BMI Predicting LogCRP (The SES Sample) Variable B SE B R 2 F (11,3430) p Step 2 0.151*** 78.69 <0.001 Gender 0.282 0.032 0.136*** BMI 0.069 0.003 0.377*** Step 3 0.000 72.12 <0.001 Gender 0.249 0.163 0.120 BMI 0.067 0.010 0.366*** Gender x BMI 0.001 0.006 0.019 Note. All steps contain the covariates (age, ethnicity, use of blood pressure medication, use of cholestero l medication, smoking, education, and income) variabl es presented are those that are relevant to the hypotheses. These analyses did not utilize the sam pling weights. BMI = body mass index. R 2 = 0.051, F (9, 3432) = 20.48, p < .0001 for Step 1; R 2 = 0.202 for Step 2; R 2 = 0.202 for Step 3. *** p < .001 Table C2 Summary of Hierarchical Regression Analysis for Gender and BMI Predicting LogCRP (The Depression Sample) Variable B SE B R 2 F (11,571) p Step 2 0.231*** 19.79 <0.001 Gender 0.568 0.087 0.240*** BMI 0.087 0.007 0.443*** Step 3 0.000 18.12 <0.001 Gender 0.479 0.398 0.203 BMI 0.082 0.023 0.418*** Gender x BMI 0.003 0.014 0.045 Note. All steps contain the covariates (age, ethnicity, use of blood pressure medication, use of cholestero l medication, smoking, depression diagnosis, and depr ession symptoms) variables presented are those that are relevant to the hypotheses. These analyses did not utilize sampling weights. BMI = body mass ind ex. R 2 = 0.045, F (9, 573) = 3.01, p < 0.01 for Step 1; R 2 = 0.276 for Step 2; R 2 = 0.276 for Step 3.

PAGE 116

Appendix C (continued) 107 Table C3 Summary of Hierarchical Regression Analysis for Gender and WC Predicting LogCRP (the SES sample) Variable B SE B R 2 F (11,3374) p Step 2 0.174*** 87.97 <0.001 Gender 0.461 0.033 0.222*** WC 0.030 0.001 0.420*** Step 3 0.001 81.01 <0.001 Gender 0.037 0.224 0.018 WC 0.024 0.004 0.330*** Gender x WC 0.004 0.002 0.209 Note. All steps contain the covariates (age, ethnicity, use of blood pressure medication, use of cholestero l medication, smoking, education, and income) variabl es presented are those that are relevant to the hypotheses. These analyses did not utilize samplin g weights. WC = waist circumference (cm). R 2 = 0.049, F (9, 3376) = 19.50, p < .001 for Step 1; R 2 = .223 for Step 2; R 2 = 0.224 for Step 3. *** p < .001. Table C4 Summary of Hierarchical Regression Analysis for Gender and WC Predicting LogCRP (the Depression Sample) Variable B SE B R 2 F (11,567) p Step 2 0.242*** 20.63 <0.001 Gender 0.677 0.086 0.286*** WC 0.036 0.003 0.462*** Step 3 0.003 19.16 <0.001 Gender -0.134 0.531 -0.057 WC 0.023 0.009 0.293* Gender x WC 0.009 0.006 0.366 Note. All steps contain the covariates (age, ethnicity, use of blood pressure medication, use of cholestero l medication, smoking, depression diagnosis, and depr ession symptoms) variables presented are those that are relevant to the hypotheses. These analyses did not utilize sampling weights. WC = waist circumference. R 2 = 0.044, F (9,569) = 2.90, p < 0.01 for Step 1; R 2 = 0.286 for Step 2; R 2 = 0.289 for Step 3.